xref: /dpdk/examples/ipv4_multicast/main.c (revision 4c00cfdc0ea225f2518a35db928ad1ab02b2a724)
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_memory.h>
49 #include <rte_memcpy.h>
50 #include <rte_eal.h>
51 #include <rte_launch.h>
52 #include <rte_atomic.h>
53 #include <rte_cycles.h>
54 #include <rte_prefetch.h>
55 #include <rte_lcore.h>
56 #include <rte_per_lcore.h>
57 #include <rte_branch_prediction.h>
58 #include <rte_interrupts.h>
59 #include <rte_pci.h>
60 #include <rte_random.h>
61 #include <rte_debug.h>
62 #include <rte_ether.h>
63 #include <rte_ethdev.h>
64 #include <rte_mempool.h>
65 #include <rte_mbuf.h>
66 #include <rte_malloc.h>
67 #include <rte_fbk_hash.h>
68 #include <rte_ip.h>
69 
70 #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1
71 
72 #define MAX_PORTS 16
73 
74 #define	MCAST_CLONE_PORTS	2
75 #define	MCAST_CLONE_SEGS	2
76 
77 #define	PKT_MBUF_DATA_SIZE	RTE_MBUF_DEFAULT_BUF_SIZE
78 #define	NB_PKT_MBUF	8192
79 
80 #define	HDR_MBUF_DATA_SIZE	(2 * RTE_PKTMBUF_HEADROOM)
81 #define	NB_HDR_MBUF	(NB_PKT_MBUF * MAX_PORTS)
82 
83 #define	NB_CLONE_MBUF	(NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2)
84 
85 /* allow max jumbo frame 9.5 KB */
86 #define	JUMBO_FRAME_MAX_SIZE	0x2600
87 
88 #define MAX_PKT_BURST 32
89 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
90 
91 /* Configure how many packets ahead to prefetch, when reading packets */
92 #define PREFETCH_OFFSET	3
93 
94 /*
95  * Construct Ethernet multicast address from IPv4 multicast address.
96  * Citing RFC 1112, section 6.4:
97  * "An IP host group address is mapped to an Ethernet multicast address
98  * by placing the low-order 23-bits of the IP address into the low-order
99  * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)."
100  */
101 #define	ETHER_ADDR_FOR_IPV4_MCAST(x)	\
102 	(rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16)
103 
104 /*
105  * Configurable number of RX/TX ring descriptors
106  */
107 #define RTE_TEST_RX_DESC_DEFAULT 128
108 #define RTE_TEST_TX_DESC_DEFAULT 512
109 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
110 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
111 
112 /* ethernet addresses of ports */
113 static struct ether_addr ports_eth_addr[MAX_PORTS];
114 
115 /* mask of enabled ports */
116 static uint32_t enabled_port_mask = 0;
117 
118 static uint16_t nb_ports;
119 
120 static int rx_queue_per_lcore = 1;
121 
122 struct mbuf_table {
123 	uint16_t len;
124 	struct rte_mbuf *m_table[MAX_PKT_BURST];
125 };
126 
127 #define MAX_RX_QUEUE_PER_LCORE 16
128 #define MAX_TX_QUEUE_PER_PORT 16
129 struct lcore_queue_conf {
130 	uint64_t tx_tsc;
131 	uint16_t n_rx_queue;
132 	uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
133 	uint16_t tx_queue_id[MAX_PORTS];
134 	struct mbuf_table tx_mbufs[MAX_PORTS];
135 } __rte_cache_aligned;
136 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
137 
138 static struct rte_eth_conf port_conf = {
139 	.rxmode = {
140 		.max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
141 		.split_hdr_size = 0,
142 		.header_split   = 0, /**< Header Split disabled */
143 		.hw_ip_checksum = 0, /**< IP checksum offload disabled */
144 		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
145 		.jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
146 		.hw_strip_crc   = 1, /**< CRC stripped by hardware */
147 	},
148 	.txmode = {
149 		.mq_mode = ETH_MQ_TX_NONE,
150 	},
151 };
152 
153 static struct rte_mempool *packet_pool, *header_pool, *clone_pool;
154 
155 
156 /* Multicast */
157 static struct rte_fbk_hash_params mcast_hash_params = {
158 	.name = "MCAST_HASH",
159 	.entries = 1024,
160 	.entries_per_bucket = 4,
161 	.socket_id = 0,
162 	.hash_func = NULL,
163 	.init_val = 0,
164 };
165 
166 struct rte_fbk_hash_table *mcast_hash = NULL;
167 
168 struct mcast_group_params {
169 	uint32_t ip;
170 	uint16_t port_mask;
171 };
172 
173 static struct mcast_group_params mcast_group_table[] = {
174 		{IPv4(224,0,0,101), 0x1},
175 		{IPv4(224,0,0,102), 0x2},
176 		{IPv4(224,0,0,103), 0x3},
177 		{IPv4(224,0,0,104), 0x4},
178 		{IPv4(224,0,0,105), 0x5},
179 		{IPv4(224,0,0,106), 0x6},
180 		{IPv4(224,0,0,107), 0x7},
181 		{IPv4(224,0,0,108), 0x8},
182 		{IPv4(224,0,0,109), 0x9},
183 		{IPv4(224,0,0,110), 0xA},
184 		{IPv4(224,0,0,111), 0xB},
185 		{IPv4(224,0,0,112), 0xC},
186 		{IPv4(224,0,0,113), 0xD},
187 		{IPv4(224,0,0,114), 0xE},
188 		{IPv4(224,0,0,115), 0xF},
189 };
190 
191 #define N_MCAST_GROUPS \
192 	(sizeof (mcast_group_table) / sizeof (mcast_group_table[0]))
193 
194 
195 /* Send burst of packets on an output interface */
196 static void
197 send_burst(struct lcore_queue_conf *qconf, uint16_t port)
198 {
199 	struct rte_mbuf **m_table;
200 	uint16_t n, queueid;
201 	int ret;
202 
203 	queueid = qconf->tx_queue_id[port];
204 	m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
205 	n = qconf->tx_mbufs[port].len;
206 
207 	ret = rte_eth_tx_burst(port, queueid, m_table, n);
208 	while (unlikely (ret < n)) {
209 		rte_pktmbuf_free(m_table[ret]);
210 		ret++;
211 	}
212 
213 	qconf->tx_mbufs[port].len = 0;
214 }
215 
216 /* Get number of bits set. */
217 static inline uint32_t
218 bitcnt(uint32_t v)
219 {
220 	uint32_t n;
221 
222 	for (n = 0; v != 0; v &= v - 1, n++)
223 		;
224 
225 	return n;
226 }
227 
228 /**
229  * Create the output multicast packet based on the given input packet.
230  * There are two approaches for creating outgoing packet, though both
231  * are based on data zero-copy idea, they differ in few details:
232  * First one creates a clone of the input packet, e.g - walk though all
233  * segments of the input packet, and for each of them create a new packet
234  * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone()
235  * for more details). Then new mbuf is allocated for the packet header
236  * and is prepended to the 'clone' mbuf.
237  * Second approach doesn't make a clone, it just increment refcnt for all
238  * input packet segments. Then it allocates new mbuf for the packet header
239  * and prepends it to the input packet.
240  * Basically first approach reuses only input packet's data, but creates
241  * it's own copy of packet's metadata. Second approach reuses both input's
242  * packet data and metadata.
243  * The advantage of first approach - is that each outgoing packet has it's
244  * own copy of metadata, so we can safely modify data pointer of the
245  * input packet. That allows us to skip creation if the output packet for
246  * the last destination port, but instead modify input packet's header inplace,
247  * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times.
248  * The advantage of second approach - less work for each outgoing packet,
249  * e.g: we skip "clone" operation completely. Though it comes with a price -
250  * input packet's metadata has to be intact. So for N destination ports we
251  * need to invoke mcast_out_pkt N times.
252  * So for small number of outgoing ports (and segments in the input packet)
253  * first approach will be faster.
254  * As number of outgoing ports (and/or input segments) will grow,
255  * second way will become more preferable.
256  *
257  *  @param pkt
258  *  Input packet mbuf.
259  *  @param use_clone
260  *  Control which of the two approaches described above should be used:
261  *  - 0 - use second approach:
262  *    Don't "clone" input packet.
263  *    Prepend new header directly to the input packet
264  *  - 1 - use first approach:
265  *    Make a "clone" of input packet first.
266  *    Prepend new header to the clone of the input packet
267  *  @return
268  *  - The pointer to the new outgoing packet.
269  *  - NULL if operation failed.
270  */
271 static inline struct rte_mbuf *
272 mcast_out_pkt(struct rte_mbuf *pkt, int use_clone)
273 {
274 	struct rte_mbuf *hdr;
275 
276 	/* Create new mbuf for the header. */
277 	if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL))
278 		return NULL;
279 
280 	/* If requested, then make a new clone packet. */
281 	if (use_clone != 0 &&
282 	    unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) {
283 		rte_pktmbuf_free(hdr);
284 		return NULL;
285 	}
286 
287 	/* prepend new header */
288 	hdr->next = pkt;
289 
290 
291 	/* update header's fields */
292 	hdr->pkt_len = (uint16_t)(hdr->data_len + pkt->pkt_len);
293 	hdr->nb_segs = (uint8_t)(pkt->nb_segs + 1);
294 
295 	/* copy metadata from source packet*/
296 	hdr->port = pkt->port;
297 	hdr->vlan_tci = pkt->vlan_tci;
298 	hdr->vlan_tci_outer = pkt->vlan_tci_outer;
299 	hdr->tx_offload = pkt->tx_offload;
300 	hdr->hash = pkt->hash;
301 
302 	hdr->ol_flags = pkt->ol_flags;
303 
304 	__rte_mbuf_sanity_check(hdr, 1);
305 	return hdr;
306 }
307 
308 /*
309  * Write new Ethernet header to the outgoing packet,
310  * and put it into the outgoing queue for the given port.
311  */
312 static inline void
313 mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr,
314 		struct lcore_queue_conf *qconf, uint16_t port)
315 {
316 	struct ether_hdr *ethdr;
317 	uint16_t len;
318 
319 	/* Construct Ethernet header. */
320 	ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr));
321 	RTE_ASSERT(ethdr != NULL);
322 
323 	ether_addr_copy(dest_addr, &ethdr->d_addr);
324 	ether_addr_copy(&ports_eth_addr[port], &ethdr->s_addr);
325 	ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
326 
327 	/* Put new packet into the output queue */
328 	len = qconf->tx_mbufs[port].len;
329 	qconf->tx_mbufs[port].m_table[len] = pkt;
330 	qconf->tx_mbufs[port].len = ++len;
331 
332 	/* Transmit packets */
333 	if (unlikely(MAX_PKT_BURST == len))
334 		send_burst(qconf, port);
335 }
336 
337 /* Multicast forward of the input packet */
338 static inline void
339 mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf)
340 {
341 	struct rte_mbuf *mc;
342 	struct ipv4_hdr *iphdr;
343 	uint32_t dest_addr, port_mask, port_num, use_clone;
344 	int32_t hash;
345 	uint16_t port;
346 	union {
347 		uint64_t as_int;
348 		struct ether_addr as_addr;
349 	} dst_eth_addr;
350 
351 	/* Remove the Ethernet header from the input packet */
352 	iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
353 	RTE_ASSERT(iphdr != NULL);
354 
355 	dest_addr = rte_be_to_cpu_32(iphdr->dst_addr);
356 
357 	/*
358 	 * Check that it is a valid multicast address and
359 	 * we have some active ports assigned to it.
360 	 */
361 	if(!IS_IPV4_MCAST(dest_addr) ||
362 	    (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 ||
363 	    (port_mask = hash & enabled_port_mask) == 0) {
364 		rte_pktmbuf_free(m);
365 		return;
366 	}
367 
368 	/* Calculate number of destination ports. */
369 	port_num = bitcnt(port_mask);
370 
371 	/* Should we use rte_pktmbuf_clone() or not. */
372 	use_clone = (port_num <= MCAST_CLONE_PORTS &&
373 	    m->nb_segs <= MCAST_CLONE_SEGS);
374 
375 	/* Mark all packet's segments as referenced port_num times */
376 	if (use_clone == 0)
377 		rte_pktmbuf_refcnt_update(m, (uint16_t)port_num);
378 
379 	/* construct destination ethernet address */
380 	dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr);
381 
382 	for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) {
383 
384 		/* Prepare output packet and send it out. */
385 		if ((port_mask & 1) != 0) {
386 			if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL))
387 				mcast_send_pkt(mc, &dst_eth_addr.as_addr,
388 						qconf, port);
389 			else if (use_clone == 0)
390 				rte_pktmbuf_free(m);
391 		}
392 	}
393 
394 	/*
395 	 * If we making clone packets, then, for the last destination port,
396 	 * we can overwrite input packet's metadata.
397 	 */
398 	if (use_clone != 0)
399 		mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port);
400 	else
401 		rte_pktmbuf_free(m);
402 }
403 
404 /* Send burst of outgoing packet, if timeout expires. */
405 static inline void
406 send_timeout_burst(struct lcore_queue_conf *qconf)
407 {
408 	uint64_t cur_tsc;
409 	uint16_t portid;
410 	const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
411 
412 	cur_tsc = rte_rdtsc();
413 	if (likely (cur_tsc < qconf->tx_tsc + drain_tsc))
414 		return;
415 
416 	for (portid = 0; portid < MAX_PORTS; portid++) {
417 		if (qconf->tx_mbufs[portid].len != 0)
418 			send_burst(qconf, portid);
419 	}
420 	qconf->tx_tsc = cur_tsc;
421 }
422 
423 /* main processing loop */
424 static int
425 main_loop(__rte_unused void *dummy)
426 {
427 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
428 	unsigned lcore_id;
429 	int i, j, nb_rx;
430 	uint16_t portid;
431 	struct lcore_queue_conf *qconf;
432 
433 	lcore_id = rte_lcore_id();
434 	qconf = &lcore_queue_conf[lcore_id];
435 
436 
437 	if (qconf->n_rx_queue == 0) {
438 		RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n",
439 		    lcore_id);
440 		return 0;
441 	}
442 
443 	RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n",
444 	    lcore_id);
445 
446 	for (i = 0; i < qconf->n_rx_queue; i++) {
447 
448 		portid = qconf->rx_queue_list[i];
449 		RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n",
450 		    lcore_id, portid);
451 	}
452 
453 	while (1) {
454 
455 		/*
456 		 * Read packet from RX queues
457 		 */
458 		for (i = 0; i < qconf->n_rx_queue; i++) {
459 
460 			portid = qconf->rx_queue_list[i];
461 			nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
462 						 MAX_PKT_BURST);
463 
464 			/* Prefetch first packets */
465 			for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
466 				rte_prefetch0(rte_pktmbuf_mtod(
467 						pkts_burst[j], void *));
468 			}
469 
470 			/* Prefetch and forward already prefetched packets */
471 			for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
472 				rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
473 						j + PREFETCH_OFFSET], void *));
474 				mcast_forward(pkts_burst[j], qconf);
475 			}
476 
477 			/* Forward remaining prefetched packets */
478 			for (; j < nb_rx; j++) {
479 				mcast_forward(pkts_burst[j], qconf);
480 			}
481 		}
482 
483 		/* Send out packets from TX queues */
484 		send_timeout_burst(qconf);
485 	}
486 }
487 
488 /* display usage */
489 static void
490 print_usage(const char *prgname)
491 {
492 	printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
493 	    "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
494 	    "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
495 	    prgname);
496 }
497 
498 static uint32_t
499 parse_portmask(const char *portmask)
500 {
501 	char *end = NULL;
502 	unsigned long pm;
503 
504 	/* parse hexadecimal string */
505 	pm = strtoul(portmask, &end, 16);
506 	if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
507 		return 0;
508 
509 	return (uint32_t)pm;
510 }
511 
512 static int
513 parse_nqueue(const char *q_arg)
514 {
515 	char *end = NULL;
516 	unsigned long n;
517 
518 	/* parse numerical string */
519 	errno = 0;
520 	n = strtoul(q_arg, &end, 0);
521 	if (errno != 0 || end == NULL || *end != '\0' ||
522 			n == 0 || n >= MAX_RX_QUEUE_PER_LCORE)
523 		return -1;
524 
525 	return n;
526 }
527 
528 /* Parse the argument given in the command line of the application */
529 static int
530 parse_args(int argc, char **argv)
531 {
532 	int opt, ret;
533 	char **argvopt;
534 	int option_index;
535 	char *prgname = argv[0];
536 	static struct option lgopts[] = {
537 		{NULL, 0, 0, 0}
538 	};
539 
540 	argvopt = argv;
541 
542 	while ((opt = getopt_long(argc, argvopt, "p:q:",
543 				  lgopts, &option_index)) != EOF) {
544 
545 		switch (opt) {
546 		/* portmask */
547 		case 'p':
548 			enabled_port_mask = parse_portmask(optarg);
549 			if (enabled_port_mask == 0) {
550 				printf("invalid portmask\n");
551 				print_usage(prgname);
552 				return -1;
553 			}
554 			break;
555 
556 		/* nqueue */
557 		case 'q':
558 			rx_queue_per_lcore = parse_nqueue(optarg);
559 			if (rx_queue_per_lcore < 0) {
560 				printf("invalid queue number\n");
561 				print_usage(prgname);
562 				return -1;
563 			}
564 			break;
565 
566 		default:
567 			print_usage(prgname);
568 			return -1;
569 		}
570 	}
571 
572 	if (optind >= 0)
573 		argv[optind-1] = prgname;
574 
575 	ret = optind-1;
576 	optind = 1; /* reset getopt lib */
577 	return ret;
578 }
579 
580 static void
581 print_ethaddr(const char *name, struct ether_addr *eth_addr)
582 {
583 	char buf[ETHER_ADDR_FMT_SIZE];
584 	ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
585 	printf("%s%s", name, buf);
586 }
587 
588 static int
589 init_mcast_hash(void)
590 {
591 	uint32_t i;
592 
593 	mcast_hash_params.socket_id = rte_socket_id();
594 	mcast_hash = rte_fbk_hash_create(&mcast_hash_params);
595 	if (mcast_hash == NULL){
596 		return -1;
597 	}
598 
599 	for (i = 0; i < N_MCAST_GROUPS; i ++){
600 		if (rte_fbk_hash_add_key(mcast_hash,
601 			mcast_group_table[i].ip,
602 			mcast_group_table[i].port_mask) < 0) {
603 			return -1;
604 		}
605 	}
606 
607 	return 0;
608 }
609 
610 /* Check the link status of all ports in up to 9s, and print them finally */
611 static void
612 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
613 {
614 #define CHECK_INTERVAL 100 /* 100ms */
615 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
616 	uint16_t portid;
617 	uint8_t count, all_ports_up, print_flag = 0;
618 	struct rte_eth_link link;
619 
620 	printf("\nChecking link status");
621 	fflush(stdout);
622 	for (count = 0; count <= MAX_CHECK_TIME; count++) {
623 		all_ports_up = 1;
624 		for (portid = 0; portid < port_num; portid++) {
625 			if ((port_mask & (1 << portid)) == 0)
626 				continue;
627 			memset(&link, 0, sizeof(link));
628 			rte_eth_link_get_nowait(portid, &link);
629 			/* print link status if flag set */
630 			if (print_flag == 1) {
631 				if (link.link_status)
632 					printf(
633 					"Port%d Link Up. Speed %u Mbps - %s\n",
634 					portid, link.link_speed,
635 				(link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
636 					("full-duplex") : ("half-duplex\n"));
637 				else
638 					printf("Port %d Link Down\n", portid);
639 				continue;
640 			}
641 			/* clear all_ports_up flag if any link down */
642 			if (link.link_status == ETH_LINK_DOWN) {
643 				all_ports_up = 0;
644 				break;
645 			}
646 		}
647 		/* after finally printing all link status, get out */
648 		if (print_flag == 1)
649 			break;
650 
651 		if (all_ports_up == 0) {
652 			printf(".");
653 			fflush(stdout);
654 			rte_delay_ms(CHECK_INTERVAL);
655 		}
656 
657 		/* set the print_flag if all ports up or timeout */
658 		if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
659 			print_flag = 1;
660 			printf("done\n");
661 		}
662 	}
663 }
664 
665 int
666 main(int argc, char **argv)
667 {
668 	struct lcore_queue_conf *qconf;
669 	struct rte_eth_dev_info dev_info;
670 	struct rte_eth_txconf *txconf;
671 	int ret;
672 	uint16_t queueid;
673 	unsigned lcore_id = 0, rx_lcore_id = 0;
674 	uint32_t n_tx_queue, nb_lcores;
675 	uint16_t portid;
676 
677 	/* init EAL */
678 	ret = rte_eal_init(argc, argv);
679 	if (ret < 0)
680 		rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
681 	argc -= ret;
682 	argv += ret;
683 
684 	/* parse application arguments (after the EAL ones) */
685 	ret = parse_args(argc, argv);
686 	if (ret < 0)
687 		rte_exit(EXIT_FAILURE, "Invalid IPV4_MULTICAST parameters\n");
688 
689 	/* create the mbuf pools */
690 	packet_pool = rte_pktmbuf_pool_create("packet_pool", NB_PKT_MBUF, 32,
691 		0, PKT_MBUF_DATA_SIZE, rte_socket_id());
692 
693 	if (packet_pool == NULL)
694 		rte_exit(EXIT_FAILURE, "Cannot init packet mbuf pool\n");
695 
696 	header_pool = rte_pktmbuf_pool_create("header_pool", NB_HDR_MBUF, 32,
697 		0, HDR_MBUF_DATA_SIZE, rte_socket_id());
698 
699 	if (header_pool == NULL)
700 		rte_exit(EXIT_FAILURE, "Cannot init header mbuf pool\n");
701 
702 	clone_pool = rte_pktmbuf_pool_create("clone_pool", NB_CLONE_MBUF, 32,
703 		0, 0, rte_socket_id());
704 
705 	if (clone_pool == NULL)
706 		rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n");
707 
708 	nb_ports = rte_eth_dev_count();
709 	if (nb_ports == 0)
710 		rte_exit(EXIT_FAILURE, "No physical ports!\n");
711 	if (nb_ports > MAX_PORTS)
712 		nb_ports = MAX_PORTS;
713 
714 	nb_lcores = rte_lcore_count();
715 
716 	/* initialize all ports */
717 	for (portid = 0; portid < nb_ports; portid++) {
718 		/* skip ports that are not enabled */
719 		if ((enabled_port_mask & (1 << portid)) == 0) {
720 			printf("Skipping disabled port %d\n", portid);
721 			continue;
722 		}
723 
724 		qconf = &lcore_queue_conf[rx_lcore_id];
725 
726 		/* limit the frame size to the maximum supported by NIC */
727 		rte_eth_dev_info_get(portid, &dev_info);
728 		port_conf.rxmode.max_rx_pkt_len = RTE_MIN(
729 		    dev_info.max_rx_pktlen, port_conf.rxmode.max_rx_pkt_len);
730 
731 		/* get the lcore_id for this port */
732 		while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
733 		       qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
734 
735 			rx_lcore_id ++;
736 			qconf = &lcore_queue_conf[rx_lcore_id];
737 
738 			if (rx_lcore_id >= RTE_MAX_LCORE)
739 				rte_exit(EXIT_FAILURE, "Not enough cores\n");
740 		}
741 		qconf->rx_queue_list[qconf->n_rx_queue] = portid;
742 		qconf->n_rx_queue++;
743 
744 		/* init port */
745 		printf("Initializing port %d on lcore %u... ", portid,
746 		       rx_lcore_id);
747 		fflush(stdout);
748 
749 		n_tx_queue = nb_lcores;
750 		if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
751 			n_tx_queue = MAX_TX_QUEUE_PER_PORT;
752 		ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
753 					    &port_conf);
754 		if (ret < 0)
755 			rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
756 				  ret, portid);
757 
758 		ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
759 						       &nb_txd);
760 		if (ret < 0)
761 			rte_exit(EXIT_FAILURE,
762 				 "Cannot adjust number of descriptors: err=%d, port=%d\n",
763 				 ret, portid);
764 
765 		rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
766 		print_ethaddr(" Address:", &ports_eth_addr[portid]);
767 		printf(", ");
768 
769 		/* init one RX queue */
770 		queueid = 0;
771 		printf("rxq=%hu ", queueid);
772 		fflush(stdout);
773 		ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
774 					     rte_eth_dev_socket_id(portid),
775 					     NULL,
776 					     packet_pool);
777 		if (ret < 0)
778 			rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n",
779 				  ret, portid);
780 
781 		/* init one TX queue per couple (lcore,port) */
782 		queueid = 0;
783 
784 		RTE_LCORE_FOREACH(lcore_id) {
785 			if (rte_lcore_is_enabled(lcore_id) == 0)
786 				continue;
787 			printf("txq=%u,%hu ", lcore_id, queueid);
788 			fflush(stdout);
789 
790 			txconf = &dev_info.default_txconf;
791 			txconf->txq_flags = 0;
792 			ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
793 						     rte_lcore_to_socket_id(lcore_id), txconf);
794 			if (ret < 0)
795 				rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
796 					  "port=%d\n", ret, portid);
797 
798 			qconf = &lcore_queue_conf[lcore_id];
799 			qconf->tx_queue_id[portid] = queueid;
800 			queueid++;
801 		}
802 
803 		/* Start device */
804 		ret = rte_eth_dev_start(portid);
805 		if (ret < 0)
806 			rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
807 				  ret, portid);
808 
809 		printf("done:\n");
810 	}
811 
812 	check_all_ports_link_status(nb_ports, enabled_port_mask);
813 
814 	/* initialize the multicast hash */
815 	int retval = init_mcast_hash();
816 	if (retval != 0)
817 		rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n");
818 
819 	/* launch per-lcore init on every lcore */
820 	rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
821 	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
822 		if (rte_eal_wait_lcore(lcore_id) < 0)
823 			return -1;
824 	}
825 
826 	return 0;
827 }
828