xref: /dpdk/examples/ipv4_multicast/main.c (revision b6df9fc8715f9a925136006b18fdd65f9c621757)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2013 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 
35 #include <stdio.h>
36 #include <stdlib.h>
37 #include <stdint.h>
38 #include <inttypes.h>
39 #include <sys/types.h>
40 #include <string.h>
41 #include <sys/queue.h>
42 #include <stdarg.h>
43 #include <errno.h>
44 #include <getopt.h>
45 
46 #include <rte_common.h>
47 #include <rte_byteorder.h>
48 #include <rte_log.h>
49 #include <rte_tailq.h>
50 #include <rte_memory.h>
51 #include <rte_memcpy.h>
52 #include <rte_memzone.h>
53 #include <rte_eal.h>
54 #include <rte_per_lcore.h>
55 #include <rte_launch.h>
56 #include <rte_atomic.h>
57 #include <rte_cycles.h>
58 #include <rte_prefetch.h>
59 #include <rte_lcore.h>
60 #include <rte_per_lcore.h>
61 #include <rte_branch_prediction.h>
62 #include <rte_interrupts.h>
63 #include <rte_pci.h>
64 #include <rte_random.h>
65 #include <rte_debug.h>
66 #include <rte_ether.h>
67 #include <rte_ethdev.h>
68 #include <rte_ring.h>
69 #include <rte_mempool.h>
70 #include <rte_mbuf.h>
71 #include <rte_malloc.h>
72 #include <rte_fbk_hash.h>
73 #include <rte_ip.h>
74 
75 #include "main.h"
76 
77 #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1
78 
79 #define MAX_PORTS 16
80 
81 #define	MCAST_CLONE_PORTS	2
82 #define	MCAST_CLONE_SEGS	2
83 
84 #define	PKT_MBUF_SIZE	(2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
85 #define	NB_PKT_MBUF	8192
86 
87 #define	HDR_MBUF_SIZE	(sizeof(struct rte_mbuf) + 2 * RTE_PKTMBUF_HEADROOM)
88 #define	NB_HDR_MBUF	(NB_PKT_MBUF * MAX_PORTS)
89 
90 #define	CLONE_MBUF_SIZE	(sizeof(struct rte_mbuf))
91 #define	NB_CLONE_MBUF	(NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2)
92 
93 /* allow max jumbo frame 9.5 KB */
94 #define	JUMBO_FRAME_MAX_SIZE	0x2600
95 
96 /*
97  * RX and TX Prefetch, Host, and Write-back threshold values should be
98  * carefully set for optimal performance. Consult the network
99  * controller's datasheet and supporting DPDK documentation for guidance
100  * on how these parameters should be set.
101  */
102 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
103 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
104 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
105 
106 /*
107  * These default values are optimized for use with the Intel(R) 82599 10 GbE
108  * Controller and the DPDK ixgbe PMD. Consider using other values for other
109  * network controllers and/or network drivers.
110  */
111 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
112 #define TX_HTHRESH 0  /**< Default values of TX host threshold reg. */
113 #define TX_WTHRESH 0  /**< Default values of TX write-back threshold reg. */
114 
115 #define MAX_PKT_BURST 32
116 #define BURST_TX_DRAIN 200000ULL /* around 100us at 2 Ghz */
117 
118 #define SOCKET0 0
119 
120 /* Configure how many packets ahead to prefetch, when reading packets */
121 #define PREFETCH_OFFSET	3
122 
123 /*
124  * Construct Ethernet multicast address from IPv4 multicast address.
125  * Citing RFC 1112, section 6.4:
126  * "An IP host group address is mapped to an Ethernet multicast address
127  * by placing the low-order 23-bits of the IP address into the low-order
128  * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)."
129  */
130 #define	ETHER_ADDR_FOR_IPV4_MCAST(x)	\
131 	(rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16)
132 
133 /*
134  * Configurable number of RX/TX ring descriptors
135  */
136 #define RTE_TEST_RX_DESC_DEFAULT 128
137 #define RTE_TEST_TX_DESC_DEFAULT 512
138 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
139 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
140 
141 /* ethernet addresses of ports */
142 static struct ether_addr ports_eth_addr[MAX_PORTS];
143 
144 /* mask of enabled ports */
145 static uint32_t enabled_port_mask = 0;
146 
147 static uint8_t nb_ports = 0;
148 
149 static int rx_queue_per_lcore = 1;
150 
151 struct mbuf_table {
152 	uint16_t len;
153 	struct rte_mbuf *m_table[MAX_PKT_BURST];
154 };
155 
156 #define MAX_RX_QUEUE_PER_LCORE 16
157 #define MAX_TX_QUEUE_PER_PORT 16
158 struct lcore_queue_conf {
159 	uint64_t tx_tsc;
160 	uint16_t n_rx_queue;
161 	uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
162 	uint16_t tx_queue_id[MAX_PORTS];
163 	struct mbuf_table tx_mbufs[MAX_PORTS];
164 } __rte_cache_aligned;
165 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
166 
167 static const struct rte_eth_conf port_conf = {
168 	.rxmode = {
169 		.max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
170 		.split_hdr_size = 0,
171 		.header_split   = 0, /**< Header Split disabled */
172 		.hw_ip_checksum = 0, /**< IP checksum offload disabled */
173 		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
174 		.jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
175 		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
176 	},
177 	.txmode = {
178 		.mq_mode = ETH_DCB_NONE,
179 	},
180 };
181 
182 static const struct rte_eth_rxconf rx_conf = {
183 	.rx_thresh = {
184 		.pthresh = RX_PTHRESH,
185 		.hthresh = RX_HTHRESH,
186 		.wthresh = RX_WTHRESH,
187 	},
188 };
189 
190 static const struct rte_eth_txconf tx_conf = {
191 	.tx_thresh = {
192 		.pthresh = TX_PTHRESH,
193 		.hthresh = TX_HTHRESH,
194 		.wthresh = TX_WTHRESH,
195 	},
196 	.tx_free_thresh = 0, /* Use PMD default values */
197 	.tx_rs_thresh = 0, /* Use PMD default values */
198 };
199 
200 static struct rte_mempool *packet_pool, *header_pool, *clone_pool;
201 
202 
203 /* Multicast */
204 static struct rte_fbk_hash_params mcast_hash_params = {
205 	.name = "MCAST_HASH",
206 	.entries = 1024,
207 	.entries_per_bucket = 4,
208 	.socket_id = SOCKET0,
209 	.hash_func = NULL,
210 	.init_val = 0,
211 };
212 
213 struct rte_fbk_hash_table *mcast_hash = NULL;
214 
215 struct mcast_group_params {
216 	uint32_t ip;
217 	uint16_t port_mask;
218 };
219 
220 static struct mcast_group_params mcast_group_table[] = {
221 		{IPv4(224,0,0,101), 0x1},
222 		{IPv4(224,0,0,102), 0x2},
223 		{IPv4(224,0,0,103), 0x3},
224 		{IPv4(224,0,0,104), 0x4},
225 		{IPv4(224,0,0,105), 0x5},
226 		{IPv4(224,0,0,106), 0x6},
227 		{IPv4(224,0,0,107), 0x7},
228 		{IPv4(224,0,0,108), 0x8},
229 		{IPv4(224,0,0,109), 0x9},
230 		{IPv4(224,0,0,110), 0xA},
231 		{IPv4(224,0,0,111), 0xB},
232 		{IPv4(224,0,0,112), 0xC},
233 		{IPv4(224,0,0,113), 0xD},
234 		{IPv4(224,0,0,114), 0xE},
235 		{IPv4(224,0,0,115), 0xF},
236 };
237 
238 #define N_MCAST_GROUPS \
239 	(sizeof (mcast_group_table) / sizeof (mcast_group_table[0]))
240 
241 
242 /* Send burst of packets on an output interface */
243 static void
244 send_burst(struct lcore_queue_conf *qconf, uint8_t port)
245 {
246 	struct rte_mbuf **m_table;
247 	uint16_t n, queueid;
248 	int ret;
249 
250 	queueid = qconf->tx_queue_id[port];
251 	m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
252 	n = qconf->tx_mbufs[port].len;
253 
254 	ret = rte_eth_tx_burst(port, queueid, m_table, n);
255 	while (unlikely (ret < n)) {
256 		rte_pktmbuf_free(m_table[ret]);
257 		ret++;
258 	}
259 
260 	qconf->tx_mbufs[port].len = 0;
261 }
262 
263 /* Get number of bits set. */
264 static inline uint32_t
265 bitcnt(uint32_t v)
266 {
267 	uint32_t n;
268 
269 	for (n = 0; v != 0; v &= v - 1, n++)
270 		;
271 
272 	return (n);
273 }
274 
275 /**
276  * Create the output multicast packet based on the given input packet.
277  * There are two approaches for creating outgoing packet, though both
278  * are based on data zero-copy idea, they differ in few details:
279  * First one creates a clone of the input packet, e.g - walk though all
280  * segments of the input packet, and for each of them create a new packet
281  * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone()
282  * for more details). Then new mbuf is allocated for the packet header
283  * and is prepended to the 'clone' mbuf.
284  * Second approach doesn't make a clone, it just increment refcnt for all
285  * input packet segments. Then it allocates new mbuf for the packet header
286  * and prepends it to the input packet.
287  * Basically first approach reuses only input packet's data, but creates
288  * it's own copy of packet's metadata. Second approach reuses both input's
289  * packet data and metadata.
290  * The advantage of first approach - is that each outgoing packet has it's
291  * own copy of metadata, so we can safely modify data pointer of the
292  * input packet. That allows us to skip creation if the output packet for
293  * the last destination port, but instead modify input packet's header inplace,
294  * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times.
295  * The advantage of second approach - less work for each outgoing packet,
296  * e.g: we skip "clone" operation completely. Though it comes with a price -
297  * input packet's metadata has to be intact. So for N destination ports we
298  * need to invoke mcast_out_pkt N times.
299  * So for small number of outgoing ports (and segments in the input packet)
300  * first approach will be faster.
301  * As number of outgoing ports (and/or input segments) will grow,
302  * second way will become more preferable.
303  *
304  *  @param pkt
305  *  Input packet mbuf.
306  *  @param use_clone
307  *  Control which of the two approaches described above should be used:
308  *  - 0 - use second approach:
309  *    Don't "clone" input packet.
310  *    Prepend new header directly to the input packet
311  *  - 1 - use first approach:
312  *    Make a "clone" of input packet first.
313  *    Prepend new header to the clone of the input packet
314  *  @return
315  *  - The pointer to the new outgoing packet.
316  *  - NULL if operation failed.
317  */
318 static inline struct rte_mbuf *
319 mcast_out_pkt(struct rte_mbuf *pkt, int use_clone)
320 {
321 	struct rte_mbuf *hdr;
322 
323 	/* Create new mbuf for the header. */
324 	if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL))
325 		return (NULL);
326 
327 	/* If requested, then make a new clone packet. */
328 	if (use_clone != 0 &&
329 	    unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) {
330 		rte_pktmbuf_free(hdr);
331 		return (NULL);
332 	}
333 
334 	/* prepend new header */
335 	hdr->pkt.next = pkt;
336 
337 
338 	/* update header's fields */
339 	hdr->pkt.pkt_len = (uint16_t)(hdr->pkt.data_len + pkt->pkt.pkt_len);
340 	hdr->pkt.nb_segs = (uint8_t)(pkt->pkt.nb_segs + 1);
341 
342 	/* copy metadata from source packet*/
343 	hdr->pkt.in_port = pkt->pkt.in_port;
344 	hdr->pkt.vlan_macip = pkt->pkt.vlan_macip;
345 	hdr->pkt.hash = pkt->pkt.hash;
346 
347 	hdr->ol_flags = pkt->ol_flags;
348 
349 	__rte_mbuf_sanity_check(hdr, RTE_MBUF_PKT, 1);
350 	return (hdr);
351 }
352 
353 /*
354  * Write new Ethernet header to the outgoing packet,
355  * and put it into the outgoing queue for the given port.
356  */
357 static inline void
358 mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr,
359 		struct lcore_queue_conf *qconf, uint8_t port)
360 {
361 	struct ether_hdr *ethdr;
362 	uint16_t len;
363 
364 	/* Construct Ethernet header. */
365 	ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr));
366 	RTE_MBUF_ASSERT(ethdr != NULL);
367 
368 	ether_addr_copy(dest_addr, &ethdr->d_addr);
369 	ether_addr_copy(&ports_eth_addr[port], &ethdr->s_addr);
370 	ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
371 
372 	/* Put new packet into the output queue */
373 	len = qconf->tx_mbufs[port].len;
374 	qconf->tx_mbufs[port].m_table[len] = pkt;
375 	qconf->tx_mbufs[port].len = ++len;
376 
377 	/* Transmit packets */
378 	if (unlikely(MAX_PKT_BURST == len))
379 		send_burst(qconf, port);
380 }
381 
382 /* Multicast forward of the input packet */
383 static inline void
384 mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf)
385 {
386 	struct rte_mbuf *mc;
387 	struct ipv4_hdr *iphdr;
388 	uint32_t dest_addr, port_mask, port_num, use_clone;
389 	int32_t hash;
390 	uint8_t port;
391 	union {
392 		uint64_t as_int;
393 		struct ether_addr as_addr;
394 	} dst_eth_addr;
395 
396 	/* Remove the Ethernet header from the input packet */
397 	iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
398 	RTE_MBUF_ASSERT(iphdr != NULL);
399 
400 	dest_addr = rte_be_to_cpu_32(iphdr->dst_addr);
401 
402 	/*
403 	 * Check that it is a valid multicast address and
404 	 * we have some active ports assigned to it.
405 	 */
406 	if(!IS_IPV4_MCAST(dest_addr) ||
407 	    (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 ||
408 	    (port_mask = hash & enabled_port_mask) == 0) {
409 		rte_pktmbuf_free(m);
410 		return;
411 	}
412 
413 	/* Calculate number of destination ports. */
414 	port_num = bitcnt(port_mask);
415 
416 	/* Should we use rte_pktmbuf_clone() or not. */
417 	use_clone = (port_num <= MCAST_CLONE_PORTS &&
418 	    m->pkt.nb_segs <= MCAST_CLONE_SEGS);
419 
420 	/* Mark all packet's segments as referenced port_num times */
421 	if (use_clone == 0)
422 		rte_pktmbuf_refcnt_update(m, (uint16_t)port_num);
423 
424 	/* construct destination ethernet address */
425 	dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr);
426 
427 	for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) {
428 
429 		/* Prepare output packet and send it out. */
430 		if ((port_mask & 1) != 0) {
431 			if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL))
432 				mcast_send_pkt(mc, &dst_eth_addr.as_addr,
433 						qconf, port);
434 			else if (use_clone == 0)
435 				rte_pktmbuf_free(m);
436 		}
437 	}
438 
439 	/*
440 	 * If we making clone packets, then, for the last destination port,
441 	 * we can overwrite input packet's metadata.
442 	 */
443 	if (use_clone != 0)
444 		mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port);
445 	else
446 		rte_pktmbuf_free(m);
447 }
448 
449 /* Send burst of outgoing packet, if timeout expires. */
450 static inline void
451 send_timeout_burst(struct lcore_queue_conf *qconf)
452 {
453 	uint64_t cur_tsc;
454 	uint8_t portid;
455 
456 	cur_tsc = rte_rdtsc();
457 	if (likely (cur_tsc < qconf->tx_tsc + BURST_TX_DRAIN))
458 		return;
459 
460 	for (portid = 0; portid < MAX_PORTS; portid++) {
461 		if (qconf->tx_mbufs[portid].len != 0)
462 			send_burst(qconf, portid);
463 	}
464 	qconf->tx_tsc = cur_tsc;
465 }
466 
467 /* main processing loop */
468 static __attribute__((noreturn)) int
469 main_loop(__rte_unused void *dummy)
470 {
471 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
472 	unsigned lcore_id;
473 	int i, j, nb_rx;
474 	uint8_t portid;
475 	struct lcore_queue_conf *qconf;
476 
477 	lcore_id = rte_lcore_id();
478 	qconf = &lcore_queue_conf[lcore_id];
479 
480 
481 	if (qconf->n_rx_queue == 0) {
482 		RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n",
483 		    lcore_id);
484 		while(1);
485 	}
486 
487 	RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n",
488 	    lcore_id);
489 
490 	for (i = 0; i < qconf->n_rx_queue; i++) {
491 
492 		portid = qconf->rx_queue_list[i];
493 		RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n",
494 		    lcore_id, (int) portid);
495 	}
496 
497 	while (1) {
498 
499 		/*
500 		 * Read packet from RX queues
501 		 */
502 		for (i = 0; i < qconf->n_rx_queue; i++) {
503 
504 			portid = qconf->rx_queue_list[i];
505 			nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
506 						 MAX_PKT_BURST);
507 
508 			/* Prefetch first packets */
509 			for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
510 				rte_prefetch0(rte_pktmbuf_mtod(
511 						pkts_burst[j], void *));
512 			}
513 
514 			/* Prefetch and forward already prefetched packets */
515 			for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
516 				rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
517 						j + PREFETCH_OFFSET], void *));
518 				mcast_forward(pkts_burst[j], qconf);
519 			}
520 
521 			/* Forward remaining prefetched packets */
522 			for (; j < nb_rx; j++) {
523 				mcast_forward(pkts_burst[j], qconf);
524 			}
525 		}
526 
527 		/* Send out packets from TX queues */
528 		send_timeout_burst(qconf);
529 	}
530 }
531 
532 /* display usage */
533 static void
534 print_usage(const char *prgname)
535 {
536 	printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
537 	    "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
538 	    "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
539 	    prgname);
540 }
541 
542 static uint32_t
543 parse_portmask(const char *portmask)
544 {
545 	char *end = NULL;
546 	unsigned long pm;
547 
548 	/* parse hexadecimal string */
549 	pm = strtoul(portmask, &end, 16);
550 	if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
551 		return 0;
552 
553 	return ((uint32_t)pm);
554 }
555 
556 static int
557 parse_nqueue(const char *q_arg)
558 {
559 	char *end = NULL;
560 	unsigned long n;
561 
562 	/* parse numerical string */
563 	errno = 0;
564 	n = strtoul(q_arg, &end, 0);
565 	if (errno != 0 || end == NULL || *end != '\0' ||
566 			n == 0 || n >= MAX_RX_QUEUE_PER_LCORE)
567 		return (-1);
568 
569 	return (n);
570 }
571 
572 /* Parse the argument given in the command line of the application */
573 static int
574 parse_args(int argc, char **argv)
575 {
576 	int opt, ret;
577 	char **argvopt;
578 	int option_index;
579 	char *prgname = argv[0];
580 	static struct option lgopts[] = {
581 		{NULL, 0, 0, 0}
582 	};
583 
584 	argvopt = argv;
585 
586 	while ((opt = getopt_long(argc, argvopt, "p:q:",
587 				  lgopts, &option_index)) != EOF) {
588 
589 		switch (opt) {
590 		/* portmask */
591 		case 'p':
592 			enabled_port_mask = parse_portmask(optarg);
593 			if (enabled_port_mask == 0) {
594 				printf("invalid portmask\n");
595 				print_usage(prgname);
596 				return -1;
597 			}
598 			break;
599 
600 		/* nqueue */
601 		case 'q':
602 			rx_queue_per_lcore = parse_nqueue(optarg);
603 			if (rx_queue_per_lcore < 0) {
604 				printf("invalid queue number\n");
605 				print_usage(prgname);
606 				return -1;
607 			}
608 			break;
609 
610 		default:
611 			print_usage(prgname);
612 			return -1;
613 		}
614 	}
615 
616 	if (optind >= 0)
617 		argv[optind-1] = prgname;
618 
619 	ret = optind-1;
620 	optind = 0; /* reset getopt lib */
621 	return ret;
622 }
623 
624 static void
625 print_ethaddr(const char *name, struct ether_addr *eth_addr)
626 {
627 	printf("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
628 	       eth_addr->addr_bytes[0],
629 	       eth_addr->addr_bytes[1],
630 	       eth_addr->addr_bytes[2],
631 	       eth_addr->addr_bytes[3],
632 	       eth_addr->addr_bytes[4],
633 	       eth_addr->addr_bytes[5]);
634 }
635 
636 static int
637 init_mcast_hash(void)
638 {
639 	uint32_t i;
640 
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 	    SOCKET0, 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 	    SOCKET0, 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 	    SOCKET0, 0);
753 
754 	if (clone_pool == NULL)
755 		rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n");
756 
757 	/* init driver */
758 	if (rte_pmd_init_all() < 0)
759 		rte_exit(EXIT_FAILURE, "Cannot init pmd\n");
760 
761 	if (rte_eal_pci_probe() < 0)
762 		rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
763 
764 	nb_ports = rte_eth_dev_count();
765 	if (nb_ports == 0)
766 		rte_exit(EXIT_FAILURE, "No physical ports!\n");
767 	if (nb_ports > MAX_PORTS)
768 		nb_ports = MAX_PORTS;
769 
770 	nb_lcores = rte_lcore_count();
771 
772 	/* initialize all ports */
773 	for (portid = 0; portid < nb_ports; portid++) {
774 		/* skip ports that are not enabled */
775 		if ((enabled_port_mask & (1 << portid)) == 0) {
776 			printf("Skipping disabled port %d\n", portid);
777 			continue;
778 		}
779 
780 		qconf = &lcore_queue_conf[rx_lcore_id];
781 
782 		/* get the lcore_id for this port */
783 		while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
784 		       qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
785 
786 			rx_lcore_id ++;
787 			qconf = &lcore_queue_conf[rx_lcore_id];
788 
789 			if (rx_lcore_id >= RTE_MAX_LCORE)
790 				rte_exit(EXIT_FAILURE, "Not enough cores\n");
791 		}
792 		qconf->rx_queue_list[qconf->n_rx_queue] = portid;
793 		qconf->n_rx_queue++;
794 
795 		/* init port */
796 		printf("Initializing port %d on lcore %u... ", portid,
797 		       rx_lcore_id);
798 		fflush(stdout);
799 
800 		n_tx_queue = nb_lcores;
801 		if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
802 			n_tx_queue = MAX_TX_QUEUE_PER_PORT;
803 		ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
804 					    &port_conf);
805 		if (ret < 0)
806 			rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
807 				  ret, portid);
808 
809 		rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
810 		print_ethaddr(" Address:", &ports_eth_addr[portid]);
811 		printf(", ");
812 
813 		/* init one RX queue */
814 		queueid = 0;
815 		printf("rxq=%hu ", queueid);
816 		fflush(stdout);
817 		ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
818 					     SOCKET0, &rx_conf,
819 					     packet_pool);
820 		if (ret < 0)
821 			rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n",
822 				  ret, portid);
823 
824 		/* init one TX queue per couple (lcore,port) */
825 		queueid = 0;
826 
827 		RTE_LCORE_FOREACH(lcore_id) {
828 			if (rte_lcore_is_enabled(lcore_id) == 0)
829 				continue;
830 			printf("txq=%u,%hu ", lcore_id, queueid);
831 			fflush(stdout);
832 			ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
833 						     SOCKET0, &tx_conf);
834 			if (ret < 0)
835 				rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
836 					  "port=%d\n", ret, portid);
837 
838 			qconf = &lcore_queue_conf[lcore_id];
839 			qconf->tx_queue_id[portid] = queueid;
840 			queueid++;
841 		}
842 
843 		/* Start device */
844 		ret = rte_eth_dev_start(portid);
845 		if (ret < 0)
846 			rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
847 				  ret, portid);
848 
849 		printf("done:\n");
850 	}
851 
852 	check_all_ports_link_status(nb_ports, enabled_port_mask);
853 
854 	/* initialize the multicast hash */
855 	int retval = init_mcast_hash();
856 	if (retval != 0)
857 		rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n");
858 
859 	/* launch per-lcore init on every lcore */
860 	rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
861 	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
862 		if (rte_eal_wait_lcore(lcore_id) < 0)
863 			return -1;
864 	}
865 
866 	return 0;
867 }
868