xref: /dpdk/examples/ip_fragmentation/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 <sys/param.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_memory.h>
50 #include <rte_memcpy.h>
51 #include <rte_memzone.h>
52 #include <rte_tailq.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_lpm.h>
72 #include <rte_lpm6.h>
73 #include <rte_ip.h>
74 #include <rte_string_fns.h>
75 
76 #include <rte_ip_frag.h>
77 
78 #include "main.h"
79 
80 #define RTE_LOGTYPE_IP_FRAG RTE_LOGTYPE_USER1
81 
82 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
83 
84 /* allow max jumbo frame 9.5 KB */
85 #define JUMBO_FRAME_MAX_SIZE	0x2600
86 
87 #define	ROUNDUP_DIV(a, b)	(((a) + (b) - 1) / (b))
88 
89 /*
90  * Default byte size for the IPv6 Maximum Transfer Unit (MTU).
91  * This value includes the size of IPv6 header.
92  */
93 #define	IPV4_MTU_DEFAULT	ETHER_MTU
94 #define	IPV6_MTU_DEFAULT	ETHER_MTU
95 
96 /*
97  * Default payload in bytes for the IPv6 packet.
98  */
99 #define	IPV4_DEFAULT_PAYLOAD	(IPV4_MTU_DEFAULT - sizeof(struct ipv4_hdr))
100 #define	IPV6_DEFAULT_PAYLOAD	(IPV6_MTU_DEFAULT - sizeof(struct ipv6_hdr))
101 
102 /*
103  * Max number of fragments per packet expected - defined by config file.
104  */
105 #define	MAX_PACKET_FRAG RTE_LIBRTE_IP_FRAG_MAX_FRAG
106 
107 #define NB_MBUF   8192
108 
109 /*
110  * RX and TX Prefetch, Host, and Write-back threshold values should be
111  * carefully set for optimal performance. Consult the network
112  * controller's datasheet and supporting DPDK documentation for guidance
113  * on how these parameters should be set.
114  */
115 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
116 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
117 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
118 
119 /*
120  * These default values are optimized for use with the Intel(R) 82599 10 GbE
121  * Controller and the DPDK ixgbe PMD. Consider using other values for other
122  * network controllers and/or network drivers.
123  */
124 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
125 #define TX_HTHRESH 0  /**< Default values of TX host threshold reg. */
126 #define TX_WTHRESH 0  /**< Default values of TX write-back threshold reg. */
127 
128 #define MAX_PKT_BURST	32
129 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
130 
131 /* Configure how many packets ahead to prefetch, when reading packets */
132 #define PREFETCH_OFFSET	3
133 
134 /*
135  * Configurable number of RX/TX ring descriptors
136  */
137 #define RTE_TEST_RX_DESC_DEFAULT 128
138 #define RTE_TEST_TX_DESC_DEFAULT 512
139 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
140 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
141 
142 /* ethernet addresses of ports */
143 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
144 
145 #ifndef IPv4_BYTES
146 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8
147 #define IPv4_BYTES(addr) \
148 		(uint8_t) (((addr) >> 24) & 0xFF),\
149 		(uint8_t) (((addr) >> 16) & 0xFF),\
150 		(uint8_t) (((addr) >> 8) & 0xFF),\
151 		(uint8_t) ((addr) & 0xFF)
152 #endif
153 
154 #ifndef IPv6_BYTES
155 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
156                        "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
157 #define IPv6_BYTES(addr) \
158 	addr[0],  addr[1], addr[2],  addr[3], \
159 	addr[4],  addr[5], addr[6],  addr[7], \
160 	addr[8],  addr[9], addr[10], addr[11],\
161 	addr[12], addr[13],addr[14], addr[15]
162 #endif
163 
164 #define IPV6_ADDR_LEN 16
165 
166 /* mask of enabled ports */
167 static int enabled_port_mask = 0;
168 
169 static int rx_queue_per_lcore = 1;
170 
171 #define MBUF_TABLE_SIZE  (2 * MAX(MAX_PKT_BURST, MAX_PACKET_FRAG))
172 
173 struct mbuf_table {
174 	uint16_t len;
175 	struct rte_mbuf *m_table[MBUF_TABLE_SIZE];
176 };
177 
178 struct rx_queue {
179 	struct rte_mempool *direct_pool;
180 	struct rte_mempool *indirect_pool;
181 	struct rte_lpm *lpm;
182 	struct rte_lpm6 *lpm6;
183 	uint8_t portid;
184 };
185 
186 #define MAX_RX_QUEUE_PER_LCORE 16
187 #define MAX_TX_QUEUE_PER_PORT 16
188 struct lcore_queue_conf {
189 	uint16_t n_rx_queue;
190 	uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
191 	struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
192 	struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
193 } __rte_cache_aligned;
194 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
195 
196 static const struct rte_eth_conf port_conf = {
197 	.rxmode = {
198 		.max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
199 		.split_hdr_size = 0,
200 		.header_split   = 0, /**< Header Split disabled */
201 		.hw_ip_checksum = 1, /**< IP checksum offload enabled */
202 		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
203 		.jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
204 		.hw_strip_crc   = 0, /**< CRC stripped by hardware */
205 	},
206 	.txmode = {
207 		.mq_mode = ETH_MQ_TX_NONE,
208 	},
209 };
210 
211 static const struct rte_eth_rxconf rx_conf = {
212 	.rx_thresh = {
213 		.pthresh = RX_PTHRESH,
214 		.hthresh = RX_HTHRESH,
215 		.wthresh = RX_WTHRESH,
216 	},
217 };
218 
219 static const struct rte_eth_txconf tx_conf = {
220 	.tx_thresh = {
221 		.pthresh = TX_PTHRESH,
222 		.hthresh = TX_HTHRESH,
223 		.wthresh = TX_WTHRESH,
224 	},
225 	.tx_free_thresh = 0, /* Use PMD default values */
226 	.tx_rs_thresh = 0, /* Use PMD default values */
227 };
228 
229 /*
230  * IPv4 forwarding table
231  */
232 struct l3fwd_ipv4_route {
233 	uint32_t ip;
234 	uint8_t  depth;
235 	uint8_t  if_out;
236 };
237 
238 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = {
239 		{IPv4(100,10,0,0), 16, 0},
240 		{IPv4(100,20,0,0), 16, 1},
241 		{IPv4(100,30,0,0), 16, 2},
242 		{IPv4(100,40,0,0), 16, 3},
243 		{IPv4(100,50,0,0), 16, 4},
244 		{IPv4(100,60,0,0), 16, 5},
245 		{IPv4(100,70,0,0), 16, 6},
246 		{IPv4(100,80,0,0), 16, 7},
247 };
248 
249 /*
250  * IPv6 forwarding table
251  */
252 
253 struct l3fwd_ipv6_route {
254 	uint8_t ip[IPV6_ADDR_LEN];
255 	uint8_t depth;
256 	uint8_t if_out;
257 };
258 
259 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = {
260 	{{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
261 	{{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
262 	{{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
263 	{{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
264 	{{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
265 	{{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
266 	{{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
267 	{{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
268 };
269 
270 #define LPM_MAX_RULES         1024
271 #define LPM6_MAX_RULES         1024
272 #define LPM6_NUMBER_TBL8S (1 << 16)
273 
274 struct rte_lpm6_config lpm6_config = {
275 		.max_rules = LPM6_MAX_RULES,
276 		.number_tbl8s = LPM6_NUMBER_TBL8S,
277 		.flags = 0
278 };
279 
280 static struct rte_mempool *socket_direct_pool[RTE_MAX_NUMA_NODES];
281 static struct rte_mempool *socket_indirect_pool[RTE_MAX_NUMA_NODES];
282 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES];
283 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES];
284 
285 /* Send burst of packets on an output interface */
286 static inline int
287 send_burst(struct lcore_queue_conf *qconf, uint16_t n, uint8_t port)
288 {
289 	struct rte_mbuf **m_table;
290 	int ret;
291 	uint16_t queueid;
292 
293 	queueid = qconf->tx_queue_id[port];
294 	m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
295 
296 	ret = rte_eth_tx_burst(port, queueid, m_table, n);
297 	if (unlikely(ret < n)) {
298 		do {
299 			rte_pktmbuf_free(m_table[ret]);
300 		} while (++ret < n);
301 	}
302 
303 	return 0;
304 }
305 
306 static inline void
307 l3fwd_simple_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf,
308 		uint8_t queueid, uint8_t port_in)
309 {
310 	struct rx_queue *rxq;
311 	uint32_t i, len;
312 	uint8_t next_hop, port_out, ipv6;
313 	int32_t len2;
314 
315 	ipv6 = 0;
316 	rxq = &qconf->rx_queue_list[queueid];
317 
318 	/* by default, send everything back to the source port */
319 	port_out = port_in;
320 
321 	/* Remove the Ethernet header and trailer from the input packet */
322 	rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
323 
324 	/* Build transmission burst */
325 	len = qconf->tx_mbufs[port_out].len;
326 
327 	/* if this is an IPv4 packet */
328 	if (m->ol_flags & PKT_RX_IPV4_HDR) {
329 		struct ipv4_hdr *ip_hdr;
330 		uint32_t ip_dst;
331 		/* Read the lookup key (i.e. ip_dst) from the input packet */
332 		ip_hdr = rte_pktmbuf_mtod(m, struct ipv4_hdr *);
333 		ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
334 
335 		/* Find destination port */
336 		if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop) == 0 &&
337 				(enabled_port_mask & 1 << next_hop) != 0) {
338 			port_out = next_hop;
339 
340 			/* Build transmission burst for new port */
341 			len = qconf->tx_mbufs[port_out].len;
342 		}
343 
344 		/* if we don't need to do any fragmentation */
345 		if (likely (IPV4_MTU_DEFAULT >= m->pkt_len)) {
346 			qconf->tx_mbufs[port_out].m_table[len] = m;
347 			len2 = 1;
348 		} else {
349 			len2 = rte_ipv4_fragment_packet(m,
350 				&qconf->tx_mbufs[port_out].m_table[len],
351 				(uint16_t)(MBUF_TABLE_SIZE - len),
352 				IPV4_MTU_DEFAULT,
353 				rxq->direct_pool, rxq->indirect_pool);
354 
355 			/* Free input packet */
356 			rte_pktmbuf_free(m);
357 
358 			/* If we fail to fragment the packet */
359 			if (unlikely (len2 < 0))
360 				return;
361 		}
362 	}
363 	/* if this is an IPv6 packet */
364 	else if (m->ol_flags & PKT_RX_IPV6_HDR) {
365 		struct ipv6_hdr *ip_hdr;
366 
367 		ipv6 = 1;
368 
369 		/* Read the lookup key (i.e. ip_dst) from the input packet */
370 		ip_hdr = rte_pktmbuf_mtod(m, struct ipv6_hdr *);
371 
372 		/* Find destination port */
373 		if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, &next_hop) == 0 &&
374 				(enabled_port_mask & 1 << next_hop) != 0) {
375 			port_out = next_hop;
376 
377 			/* Build transmission burst for new port */
378 			len = qconf->tx_mbufs[port_out].len;
379 		}
380 
381 		/* if we don't need to do any fragmentation */
382 		if (likely (IPV6_MTU_DEFAULT >= m->pkt_len)) {
383 			qconf->tx_mbufs[port_out].m_table[len] = m;
384 			len2 = 1;
385 		} else {
386 			len2 = rte_ipv6_fragment_packet(m,
387 				&qconf->tx_mbufs[port_out].m_table[len],
388 				(uint16_t)(MBUF_TABLE_SIZE - len),
389 				IPV6_MTU_DEFAULT,
390 				rxq->direct_pool, rxq->indirect_pool);
391 
392 			/* Free input packet */
393 			rte_pktmbuf_free(m);
394 
395 			/* If we fail to fragment the packet */
396 			if (unlikely (len2 < 0))
397 				return;
398 		}
399 	}
400 	/* else, just forward the packet */
401 	else {
402 		qconf->tx_mbufs[port_out].m_table[len] = m;
403 		len2 = 1;
404 	}
405 
406 	for (i = len; i < len + len2; i ++) {
407 		void *d_addr_bytes;
408 
409 		m = qconf->tx_mbufs[port_out].m_table[i];
410 		struct ether_hdr *eth_hdr = (struct ether_hdr *)
411 			rte_pktmbuf_prepend(m, (uint16_t)sizeof(struct ether_hdr));
412 		if (eth_hdr == NULL) {
413 			rte_panic("No headroom in mbuf.\n");
414 		}
415 
416 		m->l2_len = sizeof(struct ether_hdr);
417 
418 		/* 02:00:00:00:00:xx */
419 		d_addr_bytes = &eth_hdr->d_addr.addr_bytes[0];
420 		*((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)port_out << 40);
421 
422 		/* src addr */
423 		ether_addr_copy(&ports_eth_addr[port_out], &eth_hdr->s_addr);
424 		if (ipv6)
425 			eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6);
426 		else
427 			eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
428 	}
429 
430 	len += len2;
431 
432 	if (likely(len < MAX_PKT_BURST)) {
433 		qconf->tx_mbufs[port_out].len = (uint16_t)len;
434 		return;
435 	}
436 
437 	/* Transmit packets */
438 	send_burst(qconf, (uint16_t)len, port_out);
439 	qconf->tx_mbufs[port_out].len = 0;
440 }
441 
442 /* main processing loop */
443 static int
444 main_loop(__attribute__((unused)) void *dummy)
445 {
446 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
447 	unsigned lcore_id;
448 	uint64_t prev_tsc, diff_tsc, cur_tsc;
449 	int i, j, nb_rx;
450 	uint8_t portid;
451 	struct lcore_queue_conf *qconf;
452 	const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
453 
454 	prev_tsc = 0;
455 
456 	lcore_id = rte_lcore_id();
457 	qconf = &lcore_queue_conf[lcore_id];
458 
459 	if (qconf->n_rx_queue == 0) {
460 		RTE_LOG(INFO, IP_FRAG, "lcore %u has nothing to do\n", lcore_id);
461 		return 0;
462 	}
463 
464 	RTE_LOG(INFO, IP_FRAG, "entering main loop on lcore %u\n", lcore_id);
465 
466 	for (i = 0; i < qconf->n_rx_queue; i++) {
467 
468 		portid = qconf->rx_queue_list[i].portid;
469 		RTE_LOG(INFO, IP_FRAG, " -- lcoreid=%u portid=%d\n", lcore_id,
470 				(int) portid);
471 	}
472 
473 	while (1) {
474 
475 		cur_tsc = rte_rdtsc();
476 
477 		/*
478 		 * TX burst queue drain
479 		 */
480 		diff_tsc = cur_tsc - prev_tsc;
481 		if (unlikely(diff_tsc > drain_tsc)) {
482 
483 			/*
484 			 * This could be optimized (use queueid instead of
485 			 * portid), but it is not called so often
486 			 */
487 			for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
488 				if (qconf->tx_mbufs[portid].len == 0)
489 					continue;
490 				send_burst(&lcore_queue_conf[lcore_id],
491 					   qconf->tx_mbufs[portid].len,
492 					   portid);
493 				qconf->tx_mbufs[portid].len = 0;
494 			}
495 
496 			prev_tsc = cur_tsc;
497 		}
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].portid;
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 				l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
519 			}
520 
521 			/* Forward remaining prefetched packets */
522 			for (; j < nb_rx; j++) {
523 				l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
524 			}
525 		}
526 	}
527 }
528 
529 /* display usage */
530 static void
531 print_usage(const char *prgname)
532 {
533 	printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
534 	       "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
535 	       "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
536 	       prgname);
537 }
538 
539 static int
540 parse_portmask(const char *portmask)
541 {
542 	char *end = NULL;
543 	unsigned long pm;
544 
545 	/* parse hexadecimal string */
546 	pm = strtoul(portmask, &end, 16);
547 	if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
548 		return -1;
549 
550 	if (pm == 0)
551 		return -1;
552 
553 	return 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 hexadecimal string */
563 	n = strtoul(q_arg, &end, 10);
564 	if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
565 		return -1;
566 	if (n == 0)
567 		return -1;
568 	if (n >= MAX_RX_QUEUE_PER_LCORE)
569 		return -1;
570 
571 	return n;
572 }
573 
574 /* Parse the argument given in the command line of the application */
575 static int
576 parse_args(int argc, char **argv)
577 {
578 	int opt, ret;
579 	char **argvopt;
580 	int option_index;
581 	char *prgname = argv[0];
582 	static struct option lgopts[] = {
583 		{NULL, 0, 0, 0}
584 	};
585 
586 	argvopt = argv;
587 
588 	while ((opt = getopt_long(argc, argvopt, "p:q:",
589 				  lgopts, &option_index)) != EOF) {
590 
591 		switch (opt) {
592 		/* portmask */
593 		case 'p':
594 			enabled_port_mask = parse_portmask(optarg);
595 			if (enabled_port_mask < 0) {
596 				printf("invalid portmask\n");
597 				print_usage(prgname);
598 				return -1;
599 			}
600 			break;
601 
602 		/* nqueue */
603 		case 'q':
604 			rx_queue_per_lcore = parse_nqueue(optarg);
605 			if (rx_queue_per_lcore < 0) {
606 				printf("invalid queue number\n");
607 				print_usage(prgname);
608 				return -1;
609 			}
610 			break;
611 
612 		/* long options */
613 		case 0:
614 			print_usage(prgname);
615 			return -1;
616 
617 		default:
618 			print_usage(prgname);
619 			return -1;
620 		}
621 	}
622 
623 	if (enabled_port_mask == 0) {
624 		printf("portmask not specified\n");
625 		print_usage(prgname);
626 		return -1;
627 	}
628 
629 	if (optind >= 0)
630 		argv[optind-1] = prgname;
631 
632 	ret = optind-1;
633 	optind = 0; /* reset getopt lib */
634 	return ret;
635 }
636 
637 static void
638 print_ethaddr(const char *name, struct ether_addr *eth_addr)
639 {
640 	printf("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
641 	       eth_addr->addr_bytes[0],
642 	       eth_addr->addr_bytes[1],
643 	       eth_addr->addr_bytes[2],
644 	       eth_addr->addr_bytes[3],
645 	       eth_addr->addr_bytes[4],
646 	       eth_addr->addr_bytes[5]);
647 }
648 
649 /* Check the link status of all ports in up to 9s, and print them finally */
650 static void
651 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
652 {
653 #define CHECK_INTERVAL 100 /* 100ms */
654 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
655 	uint8_t portid, count, all_ports_up, print_flag = 0;
656 	struct rte_eth_link link;
657 
658 	printf("\nChecking link status");
659 	fflush(stdout);
660 	for (count = 0; count <= MAX_CHECK_TIME; count++) {
661 		all_ports_up = 1;
662 		for (portid = 0; portid < port_num; portid++) {
663 			if ((port_mask & (1 << portid)) == 0)
664 				continue;
665 			memset(&link, 0, sizeof(link));
666 			rte_eth_link_get_nowait(portid, &link);
667 			/* print link status if flag set */
668 			if (print_flag == 1) {
669 				if (link.link_status)
670 					printf("Port %d Link Up - speed %u "
671 						"Mbps - %s\n", (uint8_t)portid,
672 						(unsigned)link.link_speed,
673 				(link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
674 					("full-duplex") : ("half-duplex\n"));
675 				else
676 					printf("Port %d Link Down\n",
677 							(uint8_t)portid);
678 				continue;
679 			}
680 			/* clear all_ports_up flag if any link down */
681 			if (link.link_status == 0) {
682 				all_ports_up = 0;
683 				break;
684 			}
685 		}
686 		/* after finally printing all link status, get out */
687 		if (print_flag == 1)
688 			break;
689 
690 		if (all_ports_up == 0) {
691 			printf(".");
692 			fflush(stdout);
693 			rte_delay_ms(CHECK_INTERVAL);
694 		}
695 
696 		/* set the print_flag if all ports up or timeout */
697 		if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
698 			print_flag = 1;
699 			printf("\ndone\n");
700 		}
701 	}
702 }
703 
704 static int
705 init_routing_table(void)
706 {
707 	struct rte_lpm *lpm;
708 	struct rte_lpm6 *lpm6;
709 	int socket, ret;
710 	unsigned i;
711 
712 	for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
713 		if (socket_lpm[socket]) {
714 			lpm = socket_lpm[socket];
715 			/* populate the LPM table */
716 			for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) {
717 				ret = rte_lpm_add(lpm,
718 					l3fwd_ipv4_route_array[i].ip,
719 					l3fwd_ipv4_route_array[i].depth,
720 					l3fwd_ipv4_route_array[i].if_out);
721 
722 				if (ret < 0) {
723 					RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
724 						"LPM table\n", i);
725 					return -1;
726 				}
727 
728 				RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv4_BYTES_FMT
729 						"/%d (port %d)\n",
730 					socket,
731 					IPv4_BYTES(l3fwd_ipv4_route_array[i].ip),
732 					l3fwd_ipv4_route_array[i].depth,
733 					l3fwd_ipv4_route_array[i].if_out);
734 			}
735 		}
736 
737 		if (socket_lpm6[socket]) {
738 			lpm6 = socket_lpm6[socket];
739 			/* populate the LPM6 table */
740 			for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) {
741 				ret = rte_lpm6_add(lpm6,
742 					l3fwd_ipv6_route_array[i].ip,
743 					l3fwd_ipv6_route_array[i].depth,
744 					l3fwd_ipv6_route_array[i].if_out);
745 
746 				if (ret < 0) {
747 					RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
748 						"LPM6 table\n", i);
749 					return -1;
750 				}
751 
752 				RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv6_BYTES_FMT
753 						"/%d (port %d)\n",
754 					socket,
755 					IPv6_BYTES(l3fwd_ipv6_route_array[i].ip),
756 					l3fwd_ipv6_route_array[i].depth,
757 					l3fwd_ipv6_route_array[i].if_out);
758 			}
759 		}
760 	}
761 	return 0;
762 }
763 
764 static int
765 init_mem(void)
766 {
767 	char buf[PATH_MAX];
768 	struct rte_mempool *mp;
769 	struct rte_lpm *lpm;
770 	struct rte_lpm6 *lpm6;
771 	int socket;
772 	unsigned lcore_id;
773 
774 	/* traverse through lcores and initialize structures on each socket */
775 
776 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
777 
778 		if (rte_lcore_is_enabled(lcore_id) == 0)
779 			continue;
780 
781 		socket = rte_lcore_to_socket_id(lcore_id);
782 
783 		if (socket == SOCKET_ID_ANY)
784 			socket = 0;
785 
786 		if (socket_direct_pool[socket] == NULL) {
787 			RTE_LOG(INFO, IP_FRAG, "Creating direct mempool on socket %i\n",
788 					socket);
789 			snprintf(buf, sizeof(buf), "pool_direct_%i", socket);
790 
791 			mp = rte_mempool_create(buf, NB_MBUF,
792 						   MBUF_SIZE, 32,
793 						   sizeof(struct rte_pktmbuf_pool_private),
794 						   rte_pktmbuf_pool_init, NULL,
795 						   rte_pktmbuf_init, NULL,
796 						   socket, 0);
797 			if (mp == NULL) {
798 				RTE_LOG(ERR, IP_FRAG, "Cannot create direct mempool\n");
799 				return -1;
800 			}
801 			socket_direct_pool[socket] = mp;
802 		}
803 
804 		if (socket_indirect_pool[socket] == NULL) {
805 			RTE_LOG(INFO, IP_FRAG, "Creating indirect mempool on socket %i\n",
806 					socket);
807 			snprintf(buf, sizeof(buf), "pool_indirect_%i", socket);
808 
809 			mp = rte_mempool_create(buf, NB_MBUF,
810 							   sizeof(struct rte_mbuf), 32,
811 							   0,
812 							   NULL, NULL,
813 							   rte_pktmbuf_init, NULL,
814 							   socket, 0);
815 			if (mp == NULL) {
816 				RTE_LOG(ERR, IP_FRAG, "Cannot create indirect mempool\n");
817 				return -1;
818 			}
819 			socket_indirect_pool[socket] = mp;
820 		}
821 
822 		if (socket_lpm[socket] == NULL) {
823 			RTE_LOG(INFO, IP_FRAG, "Creating LPM table on socket %i\n", socket);
824 			snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
825 
826 			lpm = rte_lpm_create(buf, socket, LPM_MAX_RULES, 0);
827 			if (lpm == NULL) {
828 				RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
829 				return -1;
830 			}
831 			socket_lpm[socket] = lpm;
832 		}
833 
834 		if (socket_lpm6[socket] == NULL) {
835 			RTE_LOG(INFO, IP_FRAG, "Creating LPM6 table on socket %i\n", socket);
836 			snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
837 
838 			lpm6 = rte_lpm6_create("IP_FRAG_LPM6", socket, &lpm6_config);
839 			if (lpm6 == NULL) {
840 				RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
841 				return -1;
842 			}
843 			socket_lpm6[socket] = lpm6;
844 		}
845 	}
846 
847 	return 0;
848 }
849 
850 int
851 MAIN(int argc, char **argv)
852 {
853 	struct lcore_queue_conf *qconf;
854 	struct rx_queue *rxq;
855 	int socket, ret;
856 	unsigned nb_ports;
857 	uint16_t queueid = 0;
858 	unsigned lcore_id = 0, rx_lcore_id = 0;
859 	uint32_t n_tx_queue, nb_lcores;
860 	uint8_t portid;
861 
862 	/* init EAL */
863 	ret = rte_eal_init(argc, argv);
864 	if (ret < 0)
865 		rte_exit(EXIT_FAILURE, "rte_eal_init failed");
866 	argc -= ret;
867 	argv += ret;
868 
869 	/* parse application arguments (after the EAL ones) */
870 	ret = parse_args(argc, argv);
871 	if (ret < 0)
872 		rte_exit(EXIT_FAILURE, "Invalid arguments");
873 
874 	nb_ports = rte_eth_dev_count();
875 	if (nb_ports > RTE_MAX_ETHPORTS)
876 		nb_ports = RTE_MAX_ETHPORTS;
877 	else if (nb_ports == 0)
878 		rte_exit(EXIT_FAILURE, "No ports found!\n");
879 
880 	nb_lcores = rte_lcore_count();
881 
882 	/* initialize structures (mempools, lpm etc.) */
883 	if (init_mem() < 0)
884 		rte_panic("Cannot initialize memory structures!\n");
885 
886 	/* check if portmask has non-existent ports */
887 	if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned)))
888 		rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n");
889 
890 	/* initialize all ports */
891 	for (portid = 0; portid < nb_ports; portid++) {
892 		/* skip ports that are not enabled */
893 		if ((enabled_port_mask & (1 << portid)) == 0) {
894 			printf("Skipping disabled port %d\n", portid);
895 			continue;
896 		}
897 
898 		qconf = &lcore_queue_conf[rx_lcore_id];
899 
900 		/* get the lcore_id for this port */
901 		while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
902 		       qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
903 
904 			rx_lcore_id ++;
905 			if (rx_lcore_id >= RTE_MAX_LCORE)
906 				rte_exit(EXIT_FAILURE, "Not enough cores\n");
907 
908 			qconf = &lcore_queue_conf[rx_lcore_id];
909 		}
910 
911 		socket = (int) rte_lcore_to_socket_id(rx_lcore_id);
912 		if (socket == SOCKET_ID_ANY)
913 			socket = 0;
914 
915 		rxq = &qconf->rx_queue_list[qconf->n_rx_queue];
916 		rxq->portid = portid;
917 		rxq->direct_pool = socket_direct_pool[socket];
918 		rxq->indirect_pool = socket_indirect_pool[socket];
919 		rxq->lpm = socket_lpm[socket];
920 		rxq->lpm6 = socket_lpm6[socket];
921 		qconf->n_rx_queue++;
922 
923 		/* init port */
924 		printf("Initializing port %d on lcore %u...", portid,
925 		       rx_lcore_id);
926 		fflush(stdout);
927 
928 		n_tx_queue = nb_lcores;
929 		if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
930 			n_tx_queue = MAX_TX_QUEUE_PER_PORT;
931 		ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
932 					    &port_conf);
933 		if (ret < 0) {
934 			printf("\n");
935 			rte_exit(EXIT_FAILURE, "Cannot configure device: "
936 				"err=%d, port=%d\n",
937 				ret, portid);
938 		}
939 
940 		/* init one RX queue */
941 		ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
942 					     socket, &rx_conf,
943 					     socket_direct_pool[socket]);
944 		if (ret < 0) {
945 			printf("\n");
946 			rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
947 				"err=%d, port=%d\n",
948 				ret, portid);
949 		}
950 
951 		rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
952 		print_ethaddr(" Address:", &ports_eth_addr[portid]);
953 		printf("\n");
954 
955 		/* init one TX queue per couple (lcore,port) */
956 		queueid = 0;
957 		for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
958 			if (rte_lcore_is_enabled(lcore_id) == 0)
959 				continue;
960 
961 			socket = (int) rte_lcore_to_socket_id(lcore_id);
962 			printf("txq=%u,%d ", lcore_id, queueid);
963 			fflush(stdout);
964 			ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
965 						     socket, &tx_conf);
966 			if (ret < 0) {
967 				printf("\n");
968 				rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: "
969 					"err=%d, port=%d\n", ret, portid);
970 			}
971 
972 			qconf = &lcore_queue_conf[lcore_id];
973 			qconf->tx_queue_id[portid] = queueid;
974 			queueid++;
975 		}
976 
977 		printf("\n");
978 	}
979 
980 	printf("\n");
981 
982 	/* start ports */
983 	for (portid = 0; portid < nb_ports; portid++) {
984 		if ((enabled_port_mask & (1 << portid)) == 0) {
985 			continue;
986 		}
987 		/* Start device */
988 		ret = rte_eth_dev_start(portid);
989 		if (ret < 0)
990 			rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
991 				ret, portid);
992 
993 		rte_eth_promiscuous_enable(portid);
994 	}
995 
996 	if (init_routing_table() < 0)
997 		rte_exit(EXIT_FAILURE, "Cannot init routing table\n");
998 
999 	check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
1000 
1001 	/* launch per-lcore init on every lcore */
1002 	rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1003 	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1004 		if (rte_eal_wait_lcore(lcore_id) < 0)
1005 			return -1;
1006 	}
1007 
1008 	return 0;
1009 }
1010