xref: /dpdk/examples/ipv4_multicast/main.c (revision d737e95480f4893f47b28acd7ac935d919456682)
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 #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1
75 
76 #define MAX_PORTS 16
77 
78 #define	MCAST_CLONE_PORTS	2
79 #define	MCAST_CLONE_SEGS	2
80 
81 #define	PKT_MBUF_SIZE	(2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
82 #define	NB_PKT_MBUF	8192
83 
84 #define	HDR_MBUF_SIZE	(sizeof(struct rte_mbuf) + 2 * RTE_PKTMBUF_HEADROOM)
85 #define	NB_HDR_MBUF	(NB_PKT_MBUF * MAX_PORTS)
86 
87 #define	CLONE_MBUF_SIZE	(sizeof(struct rte_mbuf))
88 #define	NB_CLONE_MBUF	(NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2)
89 
90 /* allow max jumbo frame 9.5 KB */
91 #define	JUMBO_FRAME_MAX_SIZE	0x2600
92 
93 #define MAX_PKT_BURST 32
94 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
95 
96 /* Configure how many packets ahead to prefetch, when reading packets */
97 #define PREFETCH_OFFSET	3
98 
99 /*
100  * Construct Ethernet multicast address from IPv4 multicast address.
101  * Citing RFC 1112, section 6.4:
102  * "An IP host group address is mapped to an Ethernet multicast address
103  * by placing the low-order 23-bits of the IP address into the low-order
104  * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)."
105  */
106 #define	ETHER_ADDR_FOR_IPV4_MCAST(x)	\
107 	(rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16)
108 
109 /*
110  * Configurable number of RX/TX ring descriptors
111  */
112 #define RTE_TEST_RX_DESC_DEFAULT 128
113 #define RTE_TEST_TX_DESC_DEFAULT 512
114 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
115 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
116 
117 /* ethernet addresses of ports */
118 static struct ether_addr ports_eth_addr[MAX_PORTS];
119 
120 /* mask of enabled ports */
121 static uint32_t enabled_port_mask = 0;
122 
123 static uint8_t nb_ports = 0;
124 
125 static int rx_queue_per_lcore = 1;
126 
127 struct mbuf_table {
128 	uint16_t len;
129 	struct rte_mbuf *m_table[MAX_PKT_BURST];
130 };
131 
132 #define MAX_RX_QUEUE_PER_LCORE 16
133 #define MAX_TX_QUEUE_PER_PORT 16
134 struct lcore_queue_conf {
135 	uint64_t tx_tsc;
136 	uint16_t n_rx_queue;
137 	uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
138 	uint16_t tx_queue_id[MAX_PORTS];
139 	struct mbuf_table tx_mbufs[MAX_PORTS];
140 } __rte_cache_aligned;
141 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
142 
143 static const struct rte_eth_conf port_conf = {
144 	.rxmode = {
145 		.max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
146 		.split_hdr_size = 0,
147 		.header_split   = 0, /**< Header Split disabled */
148 		.hw_ip_checksum = 0, /**< IP checksum offload disabled */
149 		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
150 		.jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
151 		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
152 	},
153 	.txmode = {
154 		.mq_mode = ETH_MQ_TX_NONE,
155 	},
156 };
157 
158 static struct rte_mempool *packet_pool, *header_pool, *clone_pool;
159 
160 
161 /* Multicast */
162 static struct rte_fbk_hash_params mcast_hash_params = {
163 	.name = "MCAST_HASH",
164 	.entries = 1024,
165 	.entries_per_bucket = 4,
166 	.socket_id = 0,
167 	.hash_func = NULL,
168 	.init_val = 0,
169 };
170 
171 struct rte_fbk_hash_table *mcast_hash = NULL;
172 
173 struct mcast_group_params {
174 	uint32_t ip;
175 	uint16_t port_mask;
176 };
177 
178 static struct mcast_group_params mcast_group_table[] = {
179 		{IPv4(224,0,0,101), 0x1},
180 		{IPv4(224,0,0,102), 0x2},
181 		{IPv4(224,0,0,103), 0x3},
182 		{IPv4(224,0,0,104), 0x4},
183 		{IPv4(224,0,0,105), 0x5},
184 		{IPv4(224,0,0,106), 0x6},
185 		{IPv4(224,0,0,107), 0x7},
186 		{IPv4(224,0,0,108), 0x8},
187 		{IPv4(224,0,0,109), 0x9},
188 		{IPv4(224,0,0,110), 0xA},
189 		{IPv4(224,0,0,111), 0xB},
190 		{IPv4(224,0,0,112), 0xC},
191 		{IPv4(224,0,0,113), 0xD},
192 		{IPv4(224,0,0,114), 0xE},
193 		{IPv4(224,0,0,115), 0xF},
194 };
195 
196 #define N_MCAST_GROUPS \
197 	(sizeof (mcast_group_table) / sizeof (mcast_group_table[0]))
198 
199 
200 /* Send burst of packets on an output interface */
201 static void
202 send_burst(struct lcore_queue_conf *qconf, uint8_t port)
203 {
204 	struct rte_mbuf **m_table;
205 	uint16_t n, queueid;
206 	int ret;
207 
208 	queueid = qconf->tx_queue_id[port];
209 	m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
210 	n = qconf->tx_mbufs[port].len;
211 
212 	ret = rte_eth_tx_burst(port, queueid, m_table, n);
213 	while (unlikely (ret < n)) {
214 		rte_pktmbuf_free(m_table[ret]);
215 		ret++;
216 	}
217 
218 	qconf->tx_mbufs[port].len = 0;
219 }
220 
221 /* Get number of bits set. */
222 static inline uint32_t
223 bitcnt(uint32_t v)
224 {
225 	uint32_t n;
226 
227 	for (n = 0; v != 0; v &= v - 1, n++)
228 		;
229 
230 	return (n);
231 }
232 
233 /**
234  * Create the output multicast packet based on the given input packet.
235  * There are two approaches for creating outgoing packet, though both
236  * are based on data zero-copy idea, they differ in few details:
237  * First one creates a clone of the input packet, e.g - walk though all
238  * segments of the input packet, and for each of them create a new packet
239  * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone()
240  * for more details). Then new mbuf is allocated for the packet header
241  * and is prepended to the 'clone' mbuf.
242  * Second approach doesn't make a clone, it just increment refcnt for all
243  * input packet segments. Then it allocates new mbuf for the packet header
244  * and prepends it to the input packet.
245  * Basically first approach reuses only input packet's data, but creates
246  * it's own copy of packet's metadata. Second approach reuses both input's
247  * packet data and metadata.
248  * The advantage of first approach - is that each outgoing packet has it's
249  * own copy of metadata, so we can safely modify data pointer of the
250  * input packet. That allows us to skip creation if the output packet for
251  * the last destination port, but instead modify input packet's header inplace,
252  * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times.
253  * The advantage of second approach - less work for each outgoing packet,
254  * e.g: we skip "clone" operation completely. Though it comes with a price -
255  * input packet's metadata has to be intact. So for N destination ports we
256  * need to invoke mcast_out_pkt N times.
257  * So for small number of outgoing ports (and segments in the input packet)
258  * first approach will be faster.
259  * As number of outgoing ports (and/or input segments) will grow,
260  * second way will become more preferable.
261  *
262  *  @param pkt
263  *  Input packet mbuf.
264  *  @param use_clone
265  *  Control which of the two approaches described above should be used:
266  *  - 0 - use second approach:
267  *    Don't "clone" input packet.
268  *    Prepend new header directly to the input packet
269  *  - 1 - use first approach:
270  *    Make a "clone" of input packet first.
271  *    Prepend new header to the clone of the input packet
272  *  @return
273  *  - The pointer to the new outgoing packet.
274  *  - NULL if operation failed.
275  */
276 static inline struct rte_mbuf *
277 mcast_out_pkt(struct rte_mbuf *pkt, int use_clone)
278 {
279 	struct rte_mbuf *hdr;
280 
281 	/* Create new mbuf for the header. */
282 	if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL))
283 		return (NULL);
284 
285 	/* If requested, then make a new clone packet. */
286 	if (use_clone != 0 &&
287 	    unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) {
288 		rte_pktmbuf_free(hdr);
289 		return (NULL);
290 	}
291 
292 	/* prepend new header */
293 	hdr->next = pkt;
294 
295 
296 	/* update header's fields */
297 	hdr->pkt_len = (uint16_t)(hdr->data_len + pkt->pkt_len);
298 	hdr->nb_segs = (uint8_t)(pkt->nb_segs + 1);
299 
300 	/* copy metadata from source packet*/
301 	hdr->port = pkt->port;
302 	hdr->vlan_tci = pkt->vlan_tci;
303 	hdr->tx_offload = pkt->tx_offload;
304 	hdr->hash = pkt->hash;
305 
306 	hdr->ol_flags = pkt->ol_flags;
307 
308 	__rte_mbuf_sanity_check(hdr, 1);
309 	return (hdr);
310 }
311 
312 /*
313  * Write new Ethernet header to the outgoing packet,
314  * and put it into the outgoing queue for the given port.
315  */
316 static inline void
317 mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr,
318 		struct lcore_queue_conf *qconf, uint8_t port)
319 {
320 	struct ether_hdr *ethdr;
321 	uint16_t len;
322 
323 	/* Construct Ethernet header. */
324 	ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr));
325 	RTE_MBUF_ASSERT(ethdr != NULL);
326 
327 	ether_addr_copy(dest_addr, &ethdr->d_addr);
328 	ether_addr_copy(&ports_eth_addr[port], &ethdr->s_addr);
329 	ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
330 
331 	/* Put new packet into the output queue */
332 	len = qconf->tx_mbufs[port].len;
333 	qconf->tx_mbufs[port].m_table[len] = pkt;
334 	qconf->tx_mbufs[port].len = ++len;
335 
336 	/* Transmit packets */
337 	if (unlikely(MAX_PKT_BURST == len))
338 		send_burst(qconf, port);
339 }
340 
341 /* Multicast forward of the input packet */
342 static inline void
343 mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf)
344 {
345 	struct rte_mbuf *mc;
346 	struct ipv4_hdr *iphdr;
347 	uint32_t dest_addr, port_mask, port_num, use_clone;
348 	int32_t hash;
349 	uint8_t port;
350 	union {
351 		uint64_t as_int;
352 		struct ether_addr as_addr;
353 	} dst_eth_addr;
354 
355 	/* Remove the Ethernet header from the input packet */
356 	iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
357 	RTE_MBUF_ASSERT(iphdr != NULL);
358 
359 	dest_addr = rte_be_to_cpu_32(iphdr->dst_addr);
360 
361 	/*
362 	 * Check that it is a valid multicast address and
363 	 * we have some active ports assigned to it.
364 	 */
365 	if(!IS_IPV4_MCAST(dest_addr) ||
366 	    (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 ||
367 	    (port_mask = hash & enabled_port_mask) == 0) {
368 		rte_pktmbuf_free(m);
369 		return;
370 	}
371 
372 	/* Calculate number of destination ports. */
373 	port_num = bitcnt(port_mask);
374 
375 	/* Should we use rte_pktmbuf_clone() or not. */
376 	use_clone = (port_num <= MCAST_CLONE_PORTS &&
377 	    m->nb_segs <= MCAST_CLONE_SEGS);
378 
379 	/* Mark all packet's segments as referenced port_num times */
380 	if (use_clone == 0)
381 		rte_pktmbuf_refcnt_update(m, (uint16_t)port_num);
382 
383 	/* construct destination ethernet address */
384 	dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr);
385 
386 	for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) {
387 
388 		/* Prepare output packet and send it out. */
389 		if ((port_mask & 1) != 0) {
390 			if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL))
391 				mcast_send_pkt(mc, &dst_eth_addr.as_addr,
392 						qconf, port);
393 			else if (use_clone == 0)
394 				rte_pktmbuf_free(m);
395 		}
396 	}
397 
398 	/*
399 	 * If we making clone packets, then, for the last destination port,
400 	 * we can overwrite input packet's metadata.
401 	 */
402 	if (use_clone != 0)
403 		mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port);
404 	else
405 		rte_pktmbuf_free(m);
406 }
407 
408 /* Send burst of outgoing packet, if timeout expires. */
409 static inline void
410 send_timeout_burst(struct lcore_queue_conf *qconf)
411 {
412 	uint64_t cur_tsc;
413 	uint8_t portid;
414 	const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
415 
416 	cur_tsc = rte_rdtsc();
417 	if (likely (cur_tsc < qconf->tx_tsc + drain_tsc))
418 		return;
419 
420 	for (portid = 0; portid < MAX_PORTS; portid++) {
421 		if (qconf->tx_mbufs[portid].len != 0)
422 			send_burst(qconf, portid);
423 	}
424 	qconf->tx_tsc = cur_tsc;
425 }
426 
427 /* main processing loop */
428 static int
429 main_loop(__rte_unused void *dummy)
430 {
431 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
432 	unsigned lcore_id;
433 	int i, j, nb_rx;
434 	uint8_t portid;
435 	struct lcore_queue_conf *qconf;
436 
437 	lcore_id = rte_lcore_id();
438 	qconf = &lcore_queue_conf[lcore_id];
439 
440 
441 	if (qconf->n_rx_queue == 0) {
442 		RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n",
443 		    lcore_id);
444 		return 0;
445 	}
446 
447 	RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n",
448 	    lcore_id);
449 
450 	for (i = 0; i < qconf->n_rx_queue; i++) {
451 
452 		portid = qconf->rx_queue_list[i];
453 		RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n",
454 		    lcore_id, (int) portid);
455 	}
456 
457 	while (1) {
458 
459 		/*
460 		 * Read packet from RX queues
461 		 */
462 		for (i = 0; i < qconf->n_rx_queue; i++) {
463 
464 			portid = qconf->rx_queue_list[i];
465 			nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
466 						 MAX_PKT_BURST);
467 
468 			/* Prefetch first packets */
469 			for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
470 				rte_prefetch0(rte_pktmbuf_mtod(
471 						pkts_burst[j], void *));
472 			}
473 
474 			/* Prefetch and forward already prefetched packets */
475 			for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
476 				rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
477 						j + PREFETCH_OFFSET], void *));
478 				mcast_forward(pkts_burst[j], qconf);
479 			}
480 
481 			/* Forward remaining prefetched packets */
482 			for (; j < nb_rx; j++) {
483 				mcast_forward(pkts_burst[j], qconf);
484 			}
485 		}
486 
487 		/* Send out packets from TX queues */
488 		send_timeout_burst(qconf);
489 	}
490 }
491 
492 /* display usage */
493 static void
494 print_usage(const char *prgname)
495 {
496 	printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
497 	    "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
498 	    "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
499 	    prgname);
500 }
501 
502 static uint32_t
503 parse_portmask(const char *portmask)
504 {
505 	char *end = NULL;
506 	unsigned long pm;
507 
508 	/* parse hexadecimal string */
509 	pm = strtoul(portmask, &end, 16);
510 	if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
511 		return 0;
512 
513 	return ((uint32_t)pm);
514 }
515 
516 static int
517 parse_nqueue(const char *q_arg)
518 {
519 	char *end = NULL;
520 	unsigned long n;
521 
522 	/* parse numerical string */
523 	errno = 0;
524 	n = strtoul(q_arg, &end, 0);
525 	if (errno != 0 || end == NULL || *end != '\0' ||
526 			n == 0 || n >= MAX_RX_QUEUE_PER_LCORE)
527 		return (-1);
528 
529 	return (n);
530 }
531 
532 /* Parse the argument given in the command line of the application */
533 static int
534 parse_args(int argc, char **argv)
535 {
536 	int opt, ret;
537 	char **argvopt;
538 	int option_index;
539 	char *prgname = argv[0];
540 	static struct option lgopts[] = {
541 		{NULL, 0, 0, 0}
542 	};
543 
544 	argvopt = argv;
545 
546 	while ((opt = getopt_long(argc, argvopt, "p:q:",
547 				  lgopts, &option_index)) != EOF) {
548 
549 		switch (opt) {
550 		/* portmask */
551 		case 'p':
552 			enabled_port_mask = parse_portmask(optarg);
553 			if (enabled_port_mask == 0) {
554 				printf("invalid portmask\n");
555 				print_usage(prgname);
556 				return -1;
557 			}
558 			break;
559 
560 		/* nqueue */
561 		case 'q':
562 			rx_queue_per_lcore = parse_nqueue(optarg);
563 			if (rx_queue_per_lcore < 0) {
564 				printf("invalid queue number\n");
565 				print_usage(prgname);
566 				return -1;
567 			}
568 			break;
569 
570 		default:
571 			print_usage(prgname);
572 			return -1;
573 		}
574 	}
575 
576 	if (optind >= 0)
577 		argv[optind-1] = prgname;
578 
579 	ret = optind-1;
580 	optind = 0; /* reset getopt lib */
581 	return ret;
582 }
583 
584 static void
585 print_ethaddr(const char *name, struct ether_addr *eth_addr)
586 {
587 	char buf[ETHER_ADDR_FMT_SIZE];
588 	ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
589 	printf("%s%s", name, buf);
590 }
591 
592 static int
593 init_mcast_hash(void)
594 {
595 	uint32_t i;
596 
597 	mcast_hash_params.socket_id = rte_socket_id();
598 	mcast_hash = rte_fbk_hash_create(&mcast_hash_params);
599 	if (mcast_hash == NULL){
600 		return -1;
601 	}
602 
603 	for (i = 0; i < N_MCAST_GROUPS; i ++){
604 		if (rte_fbk_hash_add_key(mcast_hash,
605 			mcast_group_table[i].ip,
606 			mcast_group_table[i].port_mask) < 0) {
607 			return -1;
608 		}
609 	}
610 
611 	return 0;
612 }
613 
614 /* Check the link status of all ports in up to 9s, and print them finally */
615 static void
616 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
617 {
618 #define CHECK_INTERVAL 100 /* 100ms */
619 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
620 	uint8_t portid, count, all_ports_up, print_flag = 0;
621 	struct rte_eth_link link;
622 
623 	printf("\nChecking link status");
624 	fflush(stdout);
625 	for (count = 0; count <= MAX_CHECK_TIME; count++) {
626 		all_ports_up = 1;
627 		for (portid = 0; portid < port_num; portid++) {
628 			if ((port_mask & (1 << portid)) == 0)
629 				continue;
630 			memset(&link, 0, sizeof(link));
631 			rte_eth_link_get_nowait(portid, &link);
632 			/* print link status if flag set */
633 			if (print_flag == 1) {
634 				if (link.link_status)
635 					printf("Port %d Link Up - speed %u "
636 						"Mbps - %s\n", (uint8_t)portid,
637 						(unsigned)link.link_speed,
638 				(link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
639 					("full-duplex") : ("half-duplex\n"));
640 				else
641 					printf("Port %d Link Down\n",
642 							(uint8_t)portid);
643 				continue;
644 			}
645 			/* clear all_ports_up flag if any link down */
646 			if (link.link_status == 0) {
647 				all_ports_up = 0;
648 				break;
649 			}
650 		}
651 		/* after finally printing all link status, get out */
652 		if (print_flag == 1)
653 			break;
654 
655 		if (all_ports_up == 0) {
656 			printf(".");
657 			fflush(stdout);
658 			rte_delay_ms(CHECK_INTERVAL);
659 		}
660 
661 		/* set the print_flag if all ports up or timeout */
662 		if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
663 			print_flag = 1;
664 			printf("done\n");
665 		}
666 	}
667 }
668 
669 int
670 main(int argc, char **argv)
671 {
672 	struct lcore_queue_conf *qconf;
673 	struct rte_eth_dev_info dev_info;
674 	struct rte_eth_txconf *txconf;
675 	int ret;
676 	uint16_t queueid;
677 	unsigned lcore_id = 0, rx_lcore_id = 0;
678 	uint32_t n_tx_queue, nb_lcores;
679 	uint8_t portid;
680 
681 	/* init EAL */
682 	ret = rte_eal_init(argc, argv);
683 	if (ret < 0)
684 		rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
685 	argc -= ret;
686 	argv += ret;
687 
688 	/* parse application arguments (after the EAL ones) */
689 	ret = parse_args(argc, argv);
690 	if (ret < 0)
691 		rte_exit(EXIT_FAILURE, "Invalid IPV4_MULTICAST parameters\n");
692 
693 	/* create the mbuf pools */
694 	packet_pool = rte_mempool_create("packet_pool", NB_PKT_MBUF,
695 	    PKT_MBUF_SIZE, 32, sizeof(struct rte_pktmbuf_pool_private),
696 	    rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
697 	    rte_socket_id(), 0);
698 
699 	if (packet_pool == NULL)
700 		rte_exit(EXIT_FAILURE, "Cannot init packet mbuf pool\n");
701 
702 	header_pool = rte_mempool_create("header_pool", NB_HDR_MBUF,
703 	    HDR_MBUF_SIZE, 32, 0, NULL, NULL, rte_pktmbuf_init, NULL,
704 	    rte_socket_id(), 0);
705 
706 	if (header_pool == NULL)
707 		rte_exit(EXIT_FAILURE, "Cannot init header mbuf pool\n");
708 
709 	clone_pool = rte_mempool_create("clone_pool", NB_CLONE_MBUF,
710 	    CLONE_MBUF_SIZE, 32, 0, NULL, NULL, rte_pktmbuf_init, NULL,
711 	    rte_socket_id(), 0);
712 
713 	if (clone_pool == NULL)
714 		rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n");
715 
716 	nb_ports = rte_eth_dev_count();
717 	if (nb_ports == 0)
718 		rte_exit(EXIT_FAILURE, "No physical ports!\n");
719 	if (nb_ports > MAX_PORTS)
720 		nb_ports = MAX_PORTS;
721 
722 	nb_lcores = rte_lcore_count();
723 
724 	/* initialize all ports */
725 	for (portid = 0; portid < nb_ports; portid++) {
726 		/* skip ports that are not enabled */
727 		if ((enabled_port_mask & (1 << portid)) == 0) {
728 			printf("Skipping disabled port %d\n", portid);
729 			continue;
730 		}
731 
732 		qconf = &lcore_queue_conf[rx_lcore_id];
733 
734 		/* get the lcore_id for this port */
735 		while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
736 		       qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
737 
738 			rx_lcore_id ++;
739 			qconf = &lcore_queue_conf[rx_lcore_id];
740 
741 			if (rx_lcore_id >= RTE_MAX_LCORE)
742 				rte_exit(EXIT_FAILURE, "Not enough cores\n");
743 		}
744 		qconf->rx_queue_list[qconf->n_rx_queue] = portid;
745 		qconf->n_rx_queue++;
746 
747 		/* init port */
748 		printf("Initializing port %d on lcore %u... ", portid,
749 		       rx_lcore_id);
750 		fflush(stdout);
751 
752 		n_tx_queue = nb_lcores;
753 		if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
754 			n_tx_queue = MAX_TX_QUEUE_PER_PORT;
755 		ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
756 					    &port_conf);
757 		if (ret < 0)
758 			rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
759 				  ret, portid);
760 
761 		rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
762 		print_ethaddr(" Address:", &ports_eth_addr[portid]);
763 		printf(", ");
764 
765 		/* init one RX queue */
766 		queueid = 0;
767 		printf("rxq=%hu ", queueid);
768 		fflush(stdout);
769 		ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
770 					     rte_eth_dev_socket_id(portid),
771 					     NULL,
772 					     packet_pool);
773 		if (ret < 0)
774 			rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n",
775 				  ret, portid);
776 
777 		/* init one TX queue per couple (lcore,port) */
778 		queueid = 0;
779 
780 		RTE_LCORE_FOREACH(lcore_id) {
781 			if (rte_lcore_is_enabled(lcore_id) == 0)
782 				continue;
783 			printf("txq=%u,%hu ", lcore_id, queueid);
784 			fflush(stdout);
785 
786 			rte_eth_dev_info_get(portid, &dev_info);
787 			txconf = &dev_info.default_txconf;
788 			txconf->txq_flags = 0;
789 			ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
790 						     rte_lcore_to_socket_id(lcore_id), txconf);
791 			if (ret < 0)
792 				rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
793 					  "port=%d\n", ret, portid);
794 
795 			qconf = &lcore_queue_conf[lcore_id];
796 			qconf->tx_queue_id[portid] = queueid;
797 			queueid++;
798 		}
799 
800 		/* Start device */
801 		ret = rte_eth_dev_start(portid);
802 		if (ret < 0)
803 			rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
804 				  ret, portid);
805 
806 		printf("done:\n");
807 	}
808 
809 	check_all_ports_link_status(nb_ports, enabled_port_mask);
810 
811 	/* initialize the multicast hash */
812 	int retval = init_mcast_hash();
813 	if (retval != 0)
814 		rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n");
815 
816 	/* launch per-lcore init on every lcore */
817 	rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
818 	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
819 		if (rte_eal_wait_lcore(lcore_id) < 0)
820 			return -1;
821 	}
822 
823 	return 0;
824 }
825