xref: /dpdk/examples/ipv4_multicast/main.c (revision 68fa37e021a1c44c6b2a947cefc20eb61c729947)
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 
45 #include <rte_common.h>
46 #include <rte_byteorder.h>
47 #include <rte_log.h>
48 #include <rte_tailq.h>
49 #include <rte_memory.h>
50 #include <rte_memcpy.h>
51 #include <rte_memzone.h>
52 #include <rte_eal.h>
53 #include <rte_per_lcore.h>
54 #include <rte_launch.h>
55 #include <rte_atomic.h>
56 #include <rte_cycles.h>
57 #include <rte_prefetch.h>
58 #include <rte_lcore.h>
59 #include <rte_per_lcore.h>
60 #include <rte_branch_prediction.h>
61 #include <rte_interrupts.h>
62 #include <rte_pci.h>
63 #include <rte_random.h>
64 #include <rte_debug.h>
65 #include <rte_ether.h>
66 #include <rte_ethdev.h>
67 #include <rte_ring.h>
68 #include <rte_mempool.h>
69 #include <rte_mbuf.h>
70 #include <rte_malloc.h>
71 #include <rte_fbk_hash.h>
72 #include <rte_ip.h>
73 
74 #include "main.h"
75 
76 #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1
77 
78 #define MAX_PORTS 16
79 
80 #define	MCAST_CLONE_PORTS	2
81 #define	MCAST_CLONE_SEGS	2
82 
83 #define	PKT_MBUF_SIZE	(2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
84 #define	NB_PKT_MBUF	8192
85 
86 #define	HDR_MBUF_SIZE	(sizeof(struct rte_mbuf) + 2 * RTE_PKTMBUF_HEADROOM)
87 #define	NB_HDR_MBUF	(NB_PKT_MBUF * MAX_PORTS)
88 
89 #define	CLONE_MBUF_SIZE	(sizeof(struct rte_mbuf))
90 #define	NB_CLONE_MBUF	(NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2)
91 
92 /* allow max jumbo frame 9.5 KB */
93 #define	JUMBO_FRAME_MAX_SIZE	0x2600
94 
95 /*
96  * RX and TX Prefetch, Host, and Write-back threshold values should be
97  * carefully set for optimal performance. Consult the network
98  * controller's datasheet and supporting DPDK documentation for guidance
99  * on how these parameters should be set.
100  */
101 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
102 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
103 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
104 
105 /*
106  * These default values are optimized for use with the Intel(R) 82599 10 GbE
107  * Controller and the DPDK ixgbe PMD. Consider using other values for other
108  * network controllers and/or network drivers.
109  */
110 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
111 #define TX_HTHRESH 0  /**< Default values of TX host threshold reg. */
112 #define TX_WTHRESH 0  /**< Default values of TX write-back threshold reg. */
113 
114 #define MAX_PKT_BURST 32
115 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
116 
117 /* Configure how many packets ahead to prefetch, when reading packets */
118 #define PREFETCH_OFFSET	3
119 
120 /*
121  * Construct Ethernet multicast address from IPv4 multicast address.
122  * Citing RFC 1112, section 6.4:
123  * "An IP host group address is mapped to an Ethernet multicast address
124  * by placing the low-order 23-bits of the IP address into the low-order
125  * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)."
126  */
127 #define	ETHER_ADDR_FOR_IPV4_MCAST(x)	\
128 	(rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16)
129 
130 /*
131  * Configurable number of RX/TX ring descriptors
132  */
133 #define RTE_TEST_RX_DESC_DEFAULT 128
134 #define RTE_TEST_TX_DESC_DEFAULT 512
135 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
136 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
137 
138 /* ethernet addresses of ports */
139 static struct ether_addr ports_eth_addr[MAX_PORTS];
140 
141 /* mask of enabled ports */
142 static uint32_t enabled_port_mask = 0;
143 
144 static uint8_t nb_ports = 0;
145 
146 static int rx_queue_per_lcore = 1;
147 
148 struct mbuf_table {
149 	uint16_t len;
150 	struct rte_mbuf *m_table[MAX_PKT_BURST];
151 };
152 
153 #define MAX_RX_QUEUE_PER_LCORE 16
154 #define MAX_TX_QUEUE_PER_PORT 16
155 struct lcore_queue_conf {
156 	uint64_t tx_tsc;
157 	uint16_t n_rx_queue;
158 	uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
159 	uint16_t tx_queue_id[MAX_PORTS];
160 	struct mbuf_table tx_mbufs[MAX_PORTS];
161 } __rte_cache_aligned;
162 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
163 
164 static const struct rte_eth_conf port_conf = {
165 	.rxmode = {
166 		.max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
167 		.split_hdr_size = 0,
168 		.header_split   = 0, /**< Header Split disabled */
169 		.hw_ip_checksum = 0, /**< IP checksum offload disabled */
170 		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
171 		.jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
172 		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
173 	},
174 	.txmode = {
175 		.mq_mode = ETH_MQ_TX_NONE,
176 	},
177 };
178 
179 static const struct rte_eth_rxconf rx_conf = {
180 	.rx_thresh = {
181 		.pthresh = RX_PTHRESH,
182 		.hthresh = RX_HTHRESH,
183 		.wthresh = RX_WTHRESH,
184 	},
185 };
186 
187 static const struct rte_eth_txconf tx_conf = {
188 	.tx_thresh = {
189 		.pthresh = TX_PTHRESH,
190 		.hthresh = TX_HTHRESH,
191 		.wthresh = TX_WTHRESH,
192 	},
193 	.tx_free_thresh = 0, /* Use PMD default values */
194 	.tx_rs_thresh = 0, /* Use PMD default values */
195 };
196 
197 static struct rte_mempool *packet_pool, *header_pool, *clone_pool;
198 
199 
200 /* Multicast */
201 static struct rte_fbk_hash_params mcast_hash_params = {
202 	.name = "MCAST_HASH",
203 	.entries = 1024,
204 	.entries_per_bucket = 4,
205 	.socket_id = 0,
206 	.hash_func = NULL,
207 	.init_val = 0,
208 };
209 
210 struct rte_fbk_hash_table *mcast_hash = NULL;
211 
212 struct mcast_group_params {
213 	uint32_t ip;
214 	uint16_t port_mask;
215 };
216 
217 static struct mcast_group_params mcast_group_table[] = {
218 		{IPv4(224,0,0,101), 0x1},
219 		{IPv4(224,0,0,102), 0x2},
220 		{IPv4(224,0,0,103), 0x3},
221 		{IPv4(224,0,0,104), 0x4},
222 		{IPv4(224,0,0,105), 0x5},
223 		{IPv4(224,0,0,106), 0x6},
224 		{IPv4(224,0,0,107), 0x7},
225 		{IPv4(224,0,0,108), 0x8},
226 		{IPv4(224,0,0,109), 0x9},
227 		{IPv4(224,0,0,110), 0xA},
228 		{IPv4(224,0,0,111), 0xB},
229 		{IPv4(224,0,0,112), 0xC},
230 		{IPv4(224,0,0,113), 0xD},
231 		{IPv4(224,0,0,114), 0xE},
232 		{IPv4(224,0,0,115), 0xF},
233 };
234 
235 #define N_MCAST_GROUPS \
236 	(sizeof (mcast_group_table) / sizeof (mcast_group_table[0]))
237 
238 
239 /* Send burst of packets on an output interface */
240 static void
241 send_burst(struct lcore_queue_conf *qconf, uint8_t port)
242 {
243 	struct rte_mbuf **m_table;
244 	uint16_t n, queueid;
245 	int ret;
246 
247 	queueid = qconf->tx_queue_id[port];
248 	m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
249 	n = qconf->tx_mbufs[port].len;
250 
251 	ret = rte_eth_tx_burst(port, queueid, m_table, n);
252 	while (unlikely (ret < n)) {
253 		rte_pktmbuf_free(m_table[ret]);
254 		ret++;
255 	}
256 
257 	qconf->tx_mbufs[port].len = 0;
258 }
259 
260 /* Get number of bits set. */
261 static inline uint32_t
262 bitcnt(uint32_t v)
263 {
264 	uint32_t n;
265 
266 	for (n = 0; v != 0; v &= v - 1, n++)
267 		;
268 
269 	return (n);
270 }
271 
272 /**
273  * Create the output multicast packet based on the given input packet.
274  * There are two approaches for creating outgoing packet, though both
275  * are based on data zero-copy idea, they differ in few details:
276  * First one creates a clone of the input packet, e.g - walk though all
277  * segments of the input packet, and for each of them create a new packet
278  * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone()
279  * for more details). Then new mbuf is allocated for the packet header
280  * and is prepended to the 'clone' mbuf.
281  * Second approach doesn't make a clone, it just increment refcnt for all
282  * input packet segments. Then it allocates new mbuf for the packet header
283  * and prepends it to the input packet.
284  * Basically first approach reuses only input packet's data, but creates
285  * it's own copy of packet's metadata. Second approach reuses both input's
286  * packet data and metadata.
287  * The advantage of first approach - is that each outgoing packet has it's
288  * own copy of metadata, so we can safely modify data pointer of the
289  * input packet. That allows us to skip creation if the output packet for
290  * the last destination port, but instead modify input packet's header inplace,
291  * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times.
292  * The advantage of second approach - less work for each outgoing packet,
293  * e.g: we skip "clone" operation completely. Though it comes with a price -
294  * input packet's metadata has to be intact. So for N destination ports we
295  * need to invoke mcast_out_pkt N times.
296  * So for small number of outgoing ports (and segments in the input packet)
297  * first approach will be faster.
298  * As number of outgoing ports (and/or input segments) will grow,
299  * second way will become more preferable.
300  *
301  *  @param pkt
302  *  Input packet mbuf.
303  *  @param use_clone
304  *  Control which of the two approaches described above should be used:
305  *  - 0 - use second approach:
306  *    Don't "clone" input packet.
307  *    Prepend new header directly to the input packet
308  *  - 1 - use first approach:
309  *    Make a "clone" of input packet first.
310  *    Prepend new header to the clone of the input packet
311  *  @return
312  *  - The pointer to the new outgoing packet.
313  *  - NULL if operation failed.
314  */
315 static inline struct rte_mbuf *
316 mcast_out_pkt(struct rte_mbuf *pkt, int use_clone)
317 {
318 	struct rte_mbuf *hdr;
319 
320 	/* Create new mbuf for the header. */
321 	if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL))
322 		return (NULL);
323 
324 	/* If requested, then make a new clone packet. */
325 	if (use_clone != 0 &&
326 	    unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) {
327 		rte_pktmbuf_free(hdr);
328 		return (NULL);
329 	}
330 
331 	/* prepend new header */
332 	hdr->next = pkt;
333 
334 
335 	/* update header's fields */
336 	hdr->pkt_len = (uint16_t)(hdr->data_len + pkt->pkt_len);
337 	hdr->nb_segs = (uint8_t)(pkt->nb_segs + 1);
338 
339 	/* copy metadata from source packet*/
340 	hdr->port = pkt->port;
341 	hdr->vlan_tci = pkt->vlan_tci;
342 	hdr->l2_l3_len = pkt->l2_l3_len;
343 	hdr->hash = pkt->hash;
344 
345 	hdr->ol_flags = pkt->ol_flags;
346 
347 	__rte_mbuf_sanity_check(hdr, 1);
348 	return (hdr);
349 }
350 
351 /*
352  * Write new Ethernet header to the outgoing packet,
353  * and put it into the outgoing queue for the given port.
354  */
355 static inline void
356 mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr,
357 		struct lcore_queue_conf *qconf, uint8_t port)
358 {
359 	struct ether_hdr *ethdr;
360 	uint16_t len;
361 
362 	/* Construct Ethernet header. */
363 	ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr));
364 	RTE_MBUF_ASSERT(ethdr != NULL);
365 
366 	ether_addr_copy(dest_addr, &ethdr->d_addr);
367 	ether_addr_copy(&ports_eth_addr[port], &ethdr->s_addr);
368 	ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
369 
370 	/* Put new packet into the output queue */
371 	len = qconf->tx_mbufs[port].len;
372 	qconf->tx_mbufs[port].m_table[len] = pkt;
373 	qconf->tx_mbufs[port].len = ++len;
374 
375 	/* Transmit packets */
376 	if (unlikely(MAX_PKT_BURST == len))
377 		send_burst(qconf, port);
378 }
379 
380 /* Multicast forward of the input packet */
381 static inline void
382 mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf)
383 {
384 	struct rte_mbuf *mc;
385 	struct ipv4_hdr *iphdr;
386 	uint32_t dest_addr, port_mask, port_num, use_clone;
387 	int32_t hash;
388 	uint8_t port;
389 	union {
390 		uint64_t as_int;
391 		struct ether_addr as_addr;
392 	} dst_eth_addr;
393 
394 	/* Remove the Ethernet header from the input packet */
395 	iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
396 	RTE_MBUF_ASSERT(iphdr != NULL);
397 
398 	dest_addr = rte_be_to_cpu_32(iphdr->dst_addr);
399 
400 	/*
401 	 * Check that it is a valid multicast address and
402 	 * we have some active ports assigned to it.
403 	 */
404 	if(!IS_IPV4_MCAST(dest_addr) ||
405 	    (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 ||
406 	    (port_mask = hash & enabled_port_mask) == 0) {
407 		rte_pktmbuf_free(m);
408 		return;
409 	}
410 
411 	/* Calculate number of destination ports. */
412 	port_num = bitcnt(port_mask);
413 
414 	/* Should we use rte_pktmbuf_clone() or not. */
415 	use_clone = (port_num <= MCAST_CLONE_PORTS &&
416 	    m->nb_segs <= MCAST_CLONE_SEGS);
417 
418 	/* Mark all packet's segments as referenced port_num times */
419 	if (use_clone == 0)
420 		rte_pktmbuf_refcnt_update(m, (uint16_t)port_num);
421 
422 	/* construct destination ethernet address */
423 	dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr);
424 
425 	for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) {
426 
427 		/* Prepare output packet and send it out. */
428 		if ((port_mask & 1) != 0) {
429 			if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL))
430 				mcast_send_pkt(mc, &dst_eth_addr.as_addr,
431 						qconf, port);
432 			else if (use_clone == 0)
433 				rte_pktmbuf_free(m);
434 		}
435 	}
436 
437 	/*
438 	 * If we making clone packets, then, for the last destination port,
439 	 * we can overwrite input packet's metadata.
440 	 */
441 	if (use_clone != 0)
442 		mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port);
443 	else
444 		rte_pktmbuf_free(m);
445 }
446 
447 /* Send burst of outgoing packet, if timeout expires. */
448 static inline void
449 send_timeout_burst(struct lcore_queue_conf *qconf)
450 {
451 	uint64_t cur_tsc;
452 	uint8_t portid;
453 	const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
454 
455 	cur_tsc = rte_rdtsc();
456 	if (likely (cur_tsc < qconf->tx_tsc + drain_tsc))
457 		return;
458 
459 	for (portid = 0; portid < MAX_PORTS; portid++) {
460 		if (qconf->tx_mbufs[portid].len != 0)
461 			send_burst(qconf, portid);
462 	}
463 	qconf->tx_tsc = cur_tsc;
464 }
465 
466 /* main processing loop */
467 static int
468 main_loop(__rte_unused void *dummy)
469 {
470 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
471 	unsigned lcore_id;
472 	int i, j, nb_rx;
473 	uint8_t portid;
474 	struct lcore_queue_conf *qconf;
475 
476 	lcore_id = rte_lcore_id();
477 	qconf = &lcore_queue_conf[lcore_id];
478 
479 
480 	if (qconf->n_rx_queue == 0) {
481 		RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n",
482 		    lcore_id);
483 		return 0;
484 	}
485 
486 	RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n",
487 	    lcore_id);
488 
489 	for (i = 0; i < qconf->n_rx_queue; i++) {
490 
491 		portid = qconf->rx_queue_list[i];
492 		RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n",
493 		    lcore_id, (int) portid);
494 	}
495 
496 	while (1) {
497 
498 		/*
499 		 * Read packet from RX queues
500 		 */
501 		for (i = 0; i < qconf->n_rx_queue; i++) {
502 
503 			portid = qconf->rx_queue_list[i];
504 			nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
505 						 MAX_PKT_BURST);
506 
507 			/* Prefetch first packets */
508 			for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
509 				rte_prefetch0(rte_pktmbuf_mtod(
510 						pkts_burst[j], void *));
511 			}
512 
513 			/* Prefetch and forward already prefetched packets */
514 			for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
515 				rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
516 						j + PREFETCH_OFFSET], void *));
517 				mcast_forward(pkts_burst[j], qconf);
518 			}
519 
520 			/* Forward remaining prefetched packets */
521 			for (; j < nb_rx; j++) {
522 				mcast_forward(pkts_burst[j], qconf);
523 			}
524 		}
525 
526 		/* Send out packets from TX queues */
527 		send_timeout_burst(qconf);
528 	}
529 }
530 
531 /* display usage */
532 static void
533 print_usage(const char *prgname)
534 {
535 	printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
536 	    "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
537 	    "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
538 	    prgname);
539 }
540 
541 static uint32_t
542 parse_portmask(const char *portmask)
543 {
544 	char *end = NULL;
545 	unsigned long pm;
546 
547 	/* parse hexadecimal string */
548 	pm = strtoul(portmask, &end, 16);
549 	if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
550 		return 0;
551 
552 	return ((uint32_t)pm);
553 }
554 
555 static int
556 parse_nqueue(const char *q_arg)
557 {
558 	char *end = NULL;
559 	unsigned long n;
560 
561 	/* parse numerical string */
562 	errno = 0;
563 	n = strtoul(q_arg, &end, 0);
564 	if (errno != 0 || end == NULL || *end != '\0' ||
565 			n == 0 || n >= MAX_RX_QUEUE_PER_LCORE)
566 		return (-1);
567 
568 	return (n);
569 }
570 
571 /* Parse the argument given in the command line of the application */
572 static int
573 parse_args(int argc, char **argv)
574 {
575 	int opt, ret;
576 	char **argvopt;
577 	int option_index;
578 	char *prgname = argv[0];
579 	static struct option lgopts[] = {
580 		{NULL, 0, 0, 0}
581 	};
582 
583 	argvopt = argv;
584 
585 	while ((opt = getopt_long(argc, argvopt, "p:q:",
586 				  lgopts, &option_index)) != EOF) {
587 
588 		switch (opt) {
589 		/* portmask */
590 		case 'p':
591 			enabled_port_mask = parse_portmask(optarg);
592 			if (enabled_port_mask == 0) {
593 				printf("invalid portmask\n");
594 				print_usage(prgname);
595 				return -1;
596 			}
597 			break;
598 
599 		/* nqueue */
600 		case 'q':
601 			rx_queue_per_lcore = parse_nqueue(optarg);
602 			if (rx_queue_per_lcore < 0) {
603 				printf("invalid queue number\n");
604 				print_usage(prgname);
605 				return -1;
606 			}
607 			break;
608 
609 		default:
610 			print_usage(prgname);
611 			return -1;
612 		}
613 	}
614 
615 	if (optind >= 0)
616 		argv[optind-1] = prgname;
617 
618 	ret = optind-1;
619 	optind = 0; /* reset getopt lib */
620 	return ret;
621 }
622 
623 static void
624 print_ethaddr(const char *name, struct ether_addr *eth_addr)
625 {
626 	printf("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
627 	       eth_addr->addr_bytes[0],
628 	       eth_addr->addr_bytes[1],
629 	       eth_addr->addr_bytes[2],
630 	       eth_addr->addr_bytes[3],
631 	       eth_addr->addr_bytes[4],
632 	       eth_addr->addr_bytes[5]);
633 }
634 
635 static int
636 init_mcast_hash(void)
637 {
638 	uint32_t i;
639 
640 	mcast_hash_params.socket_id = rte_socket_id();
641 	mcast_hash = rte_fbk_hash_create(&mcast_hash_params);
642 	if (mcast_hash == NULL){
643 		return -1;
644 	}
645 
646 	for (i = 0; i < N_MCAST_GROUPS; i ++){
647 		if (rte_fbk_hash_add_key(mcast_hash,
648 			mcast_group_table[i].ip,
649 			mcast_group_table[i].port_mask) < 0) {
650 			return -1;
651 		}
652 	}
653 
654 	return 0;
655 }
656 
657 /* Check the link status of all ports in up to 9s, and print them finally */
658 static void
659 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
660 {
661 #define CHECK_INTERVAL 100 /* 100ms */
662 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
663 	uint8_t portid, count, all_ports_up, print_flag = 0;
664 	struct rte_eth_link link;
665 
666 	printf("\nChecking link status");
667 	fflush(stdout);
668 	for (count = 0; count <= MAX_CHECK_TIME; count++) {
669 		all_ports_up = 1;
670 		for (portid = 0; portid < port_num; portid++) {
671 			if ((port_mask & (1 << portid)) == 0)
672 				continue;
673 			memset(&link, 0, sizeof(link));
674 			rte_eth_link_get_nowait(portid, &link);
675 			/* print link status if flag set */
676 			if (print_flag == 1) {
677 				if (link.link_status)
678 					printf("Port %d Link Up - speed %u "
679 						"Mbps - %s\n", (uint8_t)portid,
680 						(unsigned)link.link_speed,
681 				(link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
682 					("full-duplex") : ("half-duplex\n"));
683 				else
684 					printf("Port %d Link Down\n",
685 							(uint8_t)portid);
686 				continue;
687 			}
688 			/* clear all_ports_up flag if any link down */
689 			if (link.link_status == 0) {
690 				all_ports_up = 0;
691 				break;
692 			}
693 		}
694 		/* after finally printing all link status, get out */
695 		if (print_flag == 1)
696 			break;
697 
698 		if (all_ports_up == 0) {
699 			printf(".");
700 			fflush(stdout);
701 			rte_delay_ms(CHECK_INTERVAL);
702 		}
703 
704 		/* set the print_flag if all ports up or timeout */
705 		if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
706 			print_flag = 1;
707 			printf("done\n");
708 		}
709 	}
710 }
711 
712 int
713 MAIN(int argc, char **argv)
714 {
715 	struct lcore_queue_conf *qconf;
716 	int ret;
717 	uint16_t queueid;
718 	unsigned lcore_id = 0, rx_lcore_id = 0;
719 	uint32_t n_tx_queue, nb_lcores;
720 	uint8_t portid;
721 
722 	/* init EAL */
723 	ret = rte_eal_init(argc, argv);
724 	if (ret < 0)
725 		rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
726 	argc -= ret;
727 	argv += ret;
728 
729 	/* parse application arguments (after the EAL ones) */
730 	ret = parse_args(argc, argv);
731 	if (ret < 0)
732 		rte_exit(EXIT_FAILURE, "Invalid IPV4_MULTICAST parameters\n");
733 
734 	/* create the mbuf pools */
735 	packet_pool = rte_mempool_create("packet_pool", NB_PKT_MBUF,
736 	    PKT_MBUF_SIZE, 32, sizeof(struct rte_pktmbuf_pool_private),
737 	    rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
738 	    rte_socket_id(), 0);
739 
740 	if (packet_pool == NULL)
741 		rte_exit(EXIT_FAILURE, "Cannot init packet mbuf pool\n");
742 
743 	header_pool = rte_mempool_create("header_pool", NB_HDR_MBUF,
744 	    HDR_MBUF_SIZE, 32, 0, NULL, NULL, rte_pktmbuf_init, NULL,
745 	    rte_socket_id(), 0);
746 
747 	if (header_pool == NULL)
748 		rte_exit(EXIT_FAILURE, "Cannot init header mbuf pool\n");
749 
750 	clone_pool = rte_mempool_create("clone_pool", NB_CLONE_MBUF,
751 	    CLONE_MBUF_SIZE, 32, 0, NULL, NULL, rte_pktmbuf_init, NULL,
752 	    rte_socket_id(), 0);
753 
754 	if (clone_pool == NULL)
755 		rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n");
756 
757 	nb_ports = rte_eth_dev_count();
758 	if (nb_ports == 0)
759 		rte_exit(EXIT_FAILURE, "No physical ports!\n");
760 	if (nb_ports > MAX_PORTS)
761 		nb_ports = MAX_PORTS;
762 
763 	nb_lcores = rte_lcore_count();
764 
765 	/* initialize all ports */
766 	for (portid = 0; portid < nb_ports; portid++) {
767 		/* skip ports that are not enabled */
768 		if ((enabled_port_mask & (1 << portid)) == 0) {
769 			printf("Skipping disabled port %d\n", portid);
770 			continue;
771 		}
772 
773 		qconf = &lcore_queue_conf[rx_lcore_id];
774 
775 		/* get the lcore_id for this port */
776 		while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
777 		       qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
778 
779 			rx_lcore_id ++;
780 			qconf = &lcore_queue_conf[rx_lcore_id];
781 
782 			if (rx_lcore_id >= RTE_MAX_LCORE)
783 				rte_exit(EXIT_FAILURE, "Not enough cores\n");
784 		}
785 		qconf->rx_queue_list[qconf->n_rx_queue] = portid;
786 		qconf->n_rx_queue++;
787 
788 		/* init port */
789 		printf("Initializing port %d on lcore %u... ", portid,
790 		       rx_lcore_id);
791 		fflush(stdout);
792 
793 		n_tx_queue = nb_lcores;
794 		if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
795 			n_tx_queue = MAX_TX_QUEUE_PER_PORT;
796 		ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
797 					    &port_conf);
798 		if (ret < 0)
799 			rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
800 				  ret, portid);
801 
802 		rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
803 		print_ethaddr(" Address:", &ports_eth_addr[portid]);
804 		printf(", ");
805 
806 		/* init one RX queue */
807 		queueid = 0;
808 		printf("rxq=%hu ", queueid);
809 		fflush(stdout);
810 		ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
811 					     rte_eth_dev_socket_id(portid), &rx_conf,
812 					     packet_pool);
813 		if (ret < 0)
814 			rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n",
815 				  ret, portid);
816 
817 		/* init one TX queue per couple (lcore,port) */
818 		queueid = 0;
819 
820 		RTE_LCORE_FOREACH(lcore_id) {
821 			if (rte_lcore_is_enabled(lcore_id) == 0)
822 				continue;
823 			printf("txq=%u,%hu ", lcore_id, queueid);
824 			fflush(stdout);
825 			ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
826 						     rte_lcore_to_socket_id(lcore_id), &tx_conf);
827 			if (ret < 0)
828 				rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
829 					  "port=%d\n", ret, portid);
830 
831 			qconf = &lcore_queue_conf[lcore_id];
832 			qconf->tx_queue_id[portid] = queueid;
833 			queueid++;
834 		}
835 
836 		/* Start device */
837 		ret = rte_eth_dev_start(portid);
838 		if (ret < 0)
839 			rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
840 				  ret, portid);
841 
842 		printf("done:\n");
843 	}
844 
845 	check_all_ports_link_status(nb_ports, enabled_port_mask);
846 
847 	/* initialize the multicast hash */
848 	int retval = init_mcast_hash();
849 	if (retval != 0)
850 		rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n");
851 
852 	/* launch per-lcore init on every lcore */
853 	rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
854 	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
855 		if (rte_eal_wait_lcore(lcore_id) < 0)
856 			return -1;
857 	}
858 
859 	return 0;
860 }
861