4 * Copyright(c) 2010-2012 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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.
33 * version: DPDK.L.1.2.3-3
40 #include <sys/types.h>
42 #include <sys/queue.h>
47 #include <rte_common.h>
48 #include <rte_byteorder.h>
50 #include <rte_memory.h>
51 #include <rte_memcpy.h>
52 #include <rte_memzone.h>
53 #include <rte_tailq.h>
55 #include <rte_per_lcore.h>
56 #include <rte_launch.h>
57 #include <rte_atomic.h>
58 #include <rte_cycles.h>
59 #include <rte_prefetch.h>
60 #include <rte_lcore.h>
61 #include <rte_per_lcore.h>
62 #include <rte_branch_prediction.h>
63 #include <rte_interrupts.h>
65 #include <rte_random.h>
66 #include <rte_debug.h>
67 #include <rte_ether.h>
68 #include <rte_ethdev.h>
70 #include <rte_mempool.h>
75 #include <rte_string_fns.h>
79 #define APP_LOOKUP_EXACT_MATCH 0
80 #define APP_LOOKUP_LPM 1
81 #define DO_RFC_1812_CHECKS
83 //#define APP_LOOKUP_METHOD APP_LOOKUP_EXACT_MATCH
84 #ifndef APP_LOOKUP_METHOD
85 #define APP_LOOKUP_METHOD APP_LOOKUP_LPM
88 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
90 #include <rte_hash_crc.h>
91 #include <rte_jhash.h>
92 #elif (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
95 #error "APP_LOOKUP_METHOD set to incorrect value"
98 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
102 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
106 * RX and TX Prefetch, Host, and Write-back threshold values should be
107 * carefully set for optimal performance. Consult the network
108 * controller's datasheet and supporting DPDK documentation for guidance
109 * on how these parameters should be set.
111 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
112 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
113 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
116 * These default values are optimized for use with the Intel(R) 82599 10 GbE
117 * Controller and the DPDK ixgbe PMD. Consider using other values for other
118 * network controllers and/or network drivers.
120 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
121 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */
122 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */
124 #define MAX_PKT_BURST 32
125 #define BURST_TX_DRAIN 200000ULL /* around 100us at 2 Ghz */
131 /* Configure how many packets ahead to prefetch, when reading packets */
132 #define PREFETCH_OFFSET 3
135 * Configurable number of RX/TX ring descriptors
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;
142 /* ethernet addresses of ports */
143 static struct ether_addr ports_eth_addr[MAX_PORTS];
145 /* mask of enabled ports */
146 static uint32_t enabled_port_mask = 0;
147 static int promiscuous_on = 0; /**< Ports set in promiscuous mode off by default. */
148 static int numa_on = 1; /**< NUMA is enabled by default. */
152 struct rte_mbuf *m_table[MAX_PKT_BURST];
155 struct lcore_rx_queue {
158 } __rte_cache_aligned;
160 #define MAX_RX_QUEUE_PER_LCORE 16
161 #define MAX_TX_QUEUE_PER_PORT MAX_PORTS
162 #define MAX_RX_QUEUE_PER_PORT 128
164 #define MAX_LCORE_PARAMS 1024
165 struct lcore_params {
169 } __rte_cache_aligned;
171 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
172 static struct lcore_params lcore_params_array_default[] = {
184 static struct lcore_params * lcore_params = lcore_params_array_default;
185 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
186 sizeof(lcore_params_array_default[0]);
188 static struct rte_eth_conf port_conf = {
191 .header_split = 0, /**< Header Split disabled */
192 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
193 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
194 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
195 .hw_strip_crc = 0, /**< CRC stripped by hardware */
200 .rss_hf = ETH_RSS_IPV4,
207 static const struct rte_eth_rxconf rx_conf = {
209 .pthresh = RX_PTHRESH,
210 .hthresh = RX_HTHRESH,
211 .wthresh = RX_WTHRESH,
215 static const struct rte_eth_txconf tx_conf = {
217 .pthresh = TX_PTHRESH,
218 .hthresh = TX_HTHRESH,
219 .wthresh = TX_WTHRESH,
221 .tx_free_thresh = 0, /* Use PMD default values */
222 .tx_rs_thresh = 0, /* Use PMD default values */
225 static struct rte_mempool * pktmbuf_pool[NB_SOCKETS];
228 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
235 } __attribute__((__packed__));
238 struct ipv4_5tuple key;
242 static struct l3fwd_route l3fwd_route_array[] = {
243 {{IPv4(100,10,0,1), IPv4(200,10,0,1), 101, 11, IPPROTO_TCP}, 0},
244 {{IPv4(100,20,0,2), IPv4(200,20,0,2), 102, 12, IPPROTO_TCP}, 1},
245 {{IPv4(100,30,0,3), IPv4(200,30,0,3), 103, 13, IPPROTO_TCP}, 2},
246 {{IPv4(100,40,0,4), IPv4(200,40,0,4), 104, 14, IPPROTO_TCP}, 3},
249 typedef struct rte_hash lookup_struct_t;
250 static lookup_struct_t *l3fwd_lookup_struct[NB_SOCKETS];
252 #define L3FWD_HASH_ENTRIES 1024
253 struct rte_hash_parameters l3fwd_hash_params = {
254 .name = "l3fwd_hash_0",
255 .entries = L3FWD_HASH_ENTRIES,
257 .key_len = sizeof(struct ipv4_5tuple),
258 .hash_func = rte_hash_crc,
259 .hash_func_init_val = 0,
260 .socket_id = SOCKET0,
263 #define L3FWD_NUM_ROUTES \
264 (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
266 static uint8_t l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
269 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
276 static struct l3fwd_route l3fwd_route_array[] = {
277 {IPv4(1,1,1,0), 24, 0},
278 {IPv4(2,1,1,0), 24, 1},
279 {IPv4(3,1,1,0), 24, 2},
280 {IPv4(4,1,1,0), 24, 3},
281 {IPv4(5,1,1,0), 24, 4},
282 {IPv4(6,1,1,0), 24, 5},
283 {IPv4(7,1,1,0), 24, 6},
284 {IPv4(8,1,1,0), 24, 7},
287 #define L3FWD_NUM_ROUTES \
288 (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
290 #define L3FWD_LPM_MAX_RULES 1024
292 typedef struct rte_lpm lookup_struct_t;
293 static lookup_struct_t *l3fwd_lookup_struct[NB_SOCKETS];
298 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
299 uint16_t tx_queue_id[MAX_PORTS];
300 struct mbuf_table tx_mbufs[MAX_PORTS];
301 lookup_struct_t * lookup_struct;
302 } __rte_cache_aligned;
304 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
306 /* Send burst of packets on an output interface */
308 send_burst(struct lcore_conf *qconf, uint16_t n, uint8_t port)
310 struct rte_mbuf **m_table;
314 queueid = qconf->tx_queue_id[port];
315 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
317 ret = rte_eth_tx_burst(port, queueid, m_table, n);
318 if (unlikely(ret < n)) {
320 rte_pktmbuf_free(m_table[ret]);
327 /* Enqueue a single packet, and send burst if queue is filled */
329 send_single_packet(struct rte_mbuf *m, uint8_t port)
333 struct lcore_conf *qconf;
335 lcore_id = rte_lcore_id();
337 qconf = &lcore_conf[lcore_id];
338 len = qconf->tx_mbufs[port].len;
339 qconf->tx_mbufs[port].m_table[len] = m;
342 /* enough pkts to be sent */
343 if (unlikely(len == MAX_PKT_BURST)) {
344 send_burst(qconf, MAX_PKT_BURST, port);
348 qconf->tx_mbufs[port].len = len;
352 #ifdef DO_RFC_1812_CHECKS
354 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
356 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
358 * 1. The packet length reported by the Link Layer must be large
359 * enough to hold the minimum length legal IP datagram (20 bytes).
361 if (link_len < sizeof(struct ipv4_hdr))
364 /* 2. The IP checksum must be correct. */
365 /* this is checked in H/W */
368 * 3. The IP version number must be 4. If the version number is not 4
369 * then the packet may be another version of IP, such as IPng or
372 if (((pkt->version_ihl) >> 4) != 4)
375 * 4. The IP header length field must be large enough to hold the
376 * minimum length legal IP datagram (20 bytes = 5 words).
378 if ((pkt->version_ihl & 0xf) < 5)
382 * 5. The IP total length field must be large enough to hold the IP
383 * datagram header, whose length is specified in the IP header length
386 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
393 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
395 print_key(struct ipv4_5tuple key)
397 printf("IP dst = %08x, IP src = %08x, port dst = %d, port src = %d, proto = %d\n",
398 (unsigned)key.ip_dst, (unsigned)key.ip_src, key.port_dst, key.port_src, key.proto);
401 static inline uint8_t
402 get_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
404 struct ipv4_5tuple key;
409 key.ip_dst = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
410 key.ip_src = rte_be_to_cpu_32(ipv4_hdr->src_addr);
411 key.proto = ipv4_hdr->next_proto_id;
413 switch (ipv4_hdr->next_proto_id) {
415 tcp = (struct tcp_hdr *)((unsigned char *) ipv4_hdr +
416 sizeof(struct ipv4_hdr));
417 key.port_dst = rte_be_to_cpu_16(tcp->dst_port);
418 key.port_src = rte_be_to_cpu_16(tcp->src_port);
422 udp = (struct udp_hdr *)((unsigned char *) ipv4_hdr +
423 sizeof(struct ipv4_hdr));
424 key.port_dst = rte_be_to_cpu_16(udp->dst_port);
425 key.port_src = rte_be_to_cpu_16(udp->src_port);
433 /* Find destination port */
434 ret = rte_hash_lookup(l3fwd_lookup_struct, (const void *)&key);
435 return (uint8_t)((ret < 0)? portid : l3fwd_out_if[ret]);
439 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
440 static inline uint8_t
441 get_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
445 return (uint8_t) ((rte_lpm_lookup(l3fwd_lookup_struct,
446 rte_be_to_cpu_32(ipv4_hdr->dst_addr), &next_hop) == 0)?
452 l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
454 struct ether_hdr *eth_hdr;
455 struct ipv4_hdr *ipv4_hdr;
459 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
461 ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(m, unsigned char *) +
462 sizeof(struct ether_hdr));
464 #ifdef DO_RFC_1812_CHECKS
465 /* Check to make sure the packet is valid (RFC1812) */
466 if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt.pkt_len) < 0) {
472 dst_port = get_dst_port(ipv4_hdr, portid, l3fwd_lookup_struct);
473 if (dst_port >= MAX_PORTS || (enabled_port_mask & 1 << dst_port) == 0)
476 /* 00:09:c0:00:00:xx */
477 tmp = ð_hdr->d_addr.addr_bytes[0];
478 *((uint64_t *)tmp) = 0x000000c00900 + (dst_port << 24);
480 #ifdef DO_RFC_1812_CHECKS
481 /* Update time to live and header checksum */
482 --(ipv4_hdr->time_to_live);
483 ++(ipv4_hdr->hdr_checksum);
487 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr);
489 send_single_packet(m, dst_port);
493 /* main processing loop */
494 static __attribute__((noreturn)) int
495 main_loop(__attribute__((unused)) void *dummy)
497 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
499 uint64_t prev_tsc = 0;
500 uint64_t diff_tsc, cur_tsc;
502 uint8_t portid, queueid;
503 struct lcore_conf *qconf;
505 lcore_id = rte_lcore_id();
506 qconf = &lcore_conf[lcore_id];
508 if (qconf->n_rx_queue == 0) {
509 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
513 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
515 for (i = 0; i < qconf->n_rx_queue; i++) {
517 portid = qconf->rx_queue_list[i].port_id;
518 queueid = qconf->rx_queue_list[i].queue_id;
519 RTE_LOG(INFO, L3FWD, " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n", lcore_id,
525 cur_tsc = rte_rdtsc();
528 * TX burst queue drain
530 diff_tsc = cur_tsc - prev_tsc;
531 if (unlikely(diff_tsc > BURST_TX_DRAIN)) {
534 * This could be optimized (use queueid instead of
535 * portid), but it is not called so often
537 for (portid = 0; portid < MAX_PORTS; portid++) {
538 if (qconf->tx_mbufs[portid].len == 0)
540 send_burst(&lcore_conf[lcore_id],
541 qconf->tx_mbufs[portid].len,
543 qconf->tx_mbufs[portid].len = 0;
550 * Read packet from RX queues
552 for (i = 0; i < qconf->n_rx_queue; ++i) {
554 portid = qconf->rx_queue_list[i].port_id;
555 queueid = qconf->rx_queue_list[i].queue_id;
556 nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst, MAX_PKT_BURST);
558 /* Prefetch first packets */
559 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
560 rte_prefetch0(rte_pktmbuf_mtod(
561 pkts_burst[j], void *));
564 /* Prefetch and forward already prefetched packets */
565 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
566 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
567 j + PREFETCH_OFFSET], void *));
568 l3fwd_simple_forward(pkts_burst[j], portid, qconf->lookup_struct);
571 /* Forward remaining prefetched packets */
572 for (; j < nb_rx; j++) {
573 l3fwd_simple_forward(pkts_burst[j], portid, qconf->lookup_struct);
580 check_lcore_params(void)
582 uint8_t queue, lcore;
586 for (i = 0; i < nb_lcore_params; ++i) {
587 queue = lcore_params[i].queue_id;
588 if (queue >= MAX_RX_QUEUE_PER_PORT) {
589 printf("invalid queue number: %hhu\n", queue);
592 lcore = lcore_params[i].lcore_id;
593 if (!rte_lcore_is_enabled(lcore)) {
594 printf("error: lcore %hhu is not enabled in lcore mask\n", lcore);
597 if ((socketid = rte_lcore_to_socket_id(lcore) != 0) &&
599 printf("warning: lcore %hhu is on socket %d with numa off \n",
607 check_port_config(const unsigned nb_ports)
612 for (i = 0; i < nb_lcore_params; ++i) {
613 portid = lcore_params[i].port_id;
614 if ((enabled_port_mask & (1 << portid)) == 0) {
615 printf("port %u is not enabled in port mask\n", portid);
618 if (portid >= nb_ports) {
619 printf("port %u is not present on the board\n", portid);
627 get_port_n_rx_queues(const uint8_t port)
632 for (i = 0; i < nb_lcore_params; ++i) {
633 if (lcore_params[i].port_id == port && lcore_params[i].queue_id > queue)
634 queue = lcore_params[i].queue_id;
636 return (uint8_t)(++queue);
640 init_lcore_rx_queues(void)
642 uint16_t i, nb_rx_queue;
645 for (i = 0; i < nb_lcore_params; ++i) {
646 lcore = lcore_params[i].lcore_id;
647 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
648 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
649 printf("error: too many queues (%u) for lcore: %u\n",
650 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
653 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
654 lcore_params[i].port_id;
655 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
656 lcore_params[i].queue_id;
657 lcore_conf[lcore].n_rx_queue++;
665 print_usage(const char *prgname)
667 printf ("%s [EAL options] -- -p PORTMASK -P"
668 " [--config (port,queue,lcore)[,(port,queue,lcore]]\n"
669 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
670 " --config (port,queue,lcore): rx queues configuration\n"
671 " --no-numa: optional, disable numa awareness\n",
676 parse_portmask(const char *portmask)
681 /* parse hexadecimal string */
682 pm = strtoul(portmask, &end, 16);
683 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
693 parse_config(const char *q_arg)
696 const char *p, *p0 = q_arg;
704 unsigned long int_fld[_NUM_FLD];
705 char *str_fld[_NUM_FLD];
711 while ((p = strchr(p0,'(')) != NULL) {
713 if((p0 = strchr(p,')')) == NULL)
717 if(size >= sizeof(s))
720 rte_snprintf(s, sizeof(s), "%.*s", size, p);
721 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') != _NUM_FLD)
723 for (i = 0; i < _NUM_FLD; i++){
725 int_fld[i] = strtoul(str_fld[i], &end, 0);
726 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
729 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
730 printf("exceeded max number of lcore params: %hu\n",
734 lcore_params_array[nb_lcore_params].port_id = (uint8_t)int_fld[FLD_PORT];
735 lcore_params_array[nb_lcore_params].queue_id = (uint8_t)int_fld[FLD_QUEUE];
736 lcore_params_array[nb_lcore_params].lcore_id = (uint8_t)int_fld[FLD_LCORE];
739 lcore_params = lcore_params_array;
743 /* Parse the argument given in the command line of the application */
745 parse_args(int argc, char **argv)
750 char *prgname = argv[0];
751 static struct option lgopts[] = {
753 {"no-numa", 0, 0, 0},
759 while ((opt = getopt_long(argc, argvopt, "p:P",
760 lgopts, &option_index)) != EOF) {
765 enabled_port_mask = parse_portmask(optarg);
766 if (enabled_port_mask == 0) {
767 printf("invalid portmask\n");
768 print_usage(prgname);
773 printf("Promiscuous mode selected\n");
779 if (!strcmp(lgopts[option_index].name, "config")) {
780 ret = parse_config(optarg);
782 printf("invalid config\n");
783 print_usage(prgname);
788 if (!strcmp(lgopts[option_index].name, "no-numa")) {
789 printf("numa is disabled \n");
795 print_usage(prgname);
801 argv[optind-1] = prgname;
804 optind = 0; /* reset getopt lib */
809 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
811 printf ("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
812 eth_addr->addr_bytes[0],
813 eth_addr->addr_bytes[1],
814 eth_addr->addr_bytes[2],
815 eth_addr->addr_bytes[3],
816 eth_addr->addr_bytes[4],
817 eth_addr->addr_bytes[5]);
820 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
822 setup_hash(int socketid)
829 rte_snprintf(s, sizeof(s), "l3fwd_hash_%d", socketid);
830 l3fwd_hash_params.name = s;
831 l3fwd_hash_params.socket_id = socketid;
832 l3fwd_lookup_struct[socketid] = rte_hash_create(&l3fwd_hash_params);
833 if (l3fwd_lookup_struct[socketid] == NULL)
834 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd hash on "
835 "socket %d\n", socketid);
837 /* populate the hash */
838 for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
839 ret = rte_hash_add_key (l3fwd_lookup_struct[socketid],
840 (void *) &l3fwd_route_array[i].key);
842 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the"
843 "l3fwd hash on socket %d\n", i, socketid);
845 l3fwd_out_if[ret] = l3fwd_route_array[i].if_out;
846 printf("Hash: Adding key\n");
847 print_key(l3fwd_route_array[i].key);
852 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
854 setup_lpm(int socketid)
860 /* create the LPM table */
861 rte_snprintf(s, sizeof(s), "L3FWD_LPM_%d", socketid);
862 l3fwd_lookup_struct[socketid] = rte_lpm_create(s, socketid,
863 L3FWD_LPM_MAX_RULES, RTE_LPM_MEMZONE);
864 if (l3fwd_lookup_struct[socketid] == NULL)
865 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd LPM table"
866 " on socket %d\n", socketid);
868 /* populate the LPM table */
869 for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
870 ret = rte_lpm_add(l3fwd_lookup_struct[socketid],
871 l3fwd_route_array[i].ip,
872 l3fwd_route_array[i].depth,
873 l3fwd_route_array[i].if_out);
876 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the "
877 "l3fwd LPM table on socket %d\n",
881 printf("LPM: Adding route 0x%08x / %d (%d)\n",
882 (unsigned)l3fwd_route_array[i].ip,
883 l3fwd_route_array[i].depth,
884 l3fwd_route_array[i].if_out);
892 struct lcore_conf *qconf;
897 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
898 if (rte_lcore_is_enabled(lcore_id) == 0)
902 socketid = rte_lcore_to_socket_id(lcore_id);
906 if (socketid >= NB_SOCKETS) {
907 rte_exit(EXIT_FAILURE, "Socket %d of lcore %u is out of range %d\n",
908 socketid, lcore_id, NB_SOCKETS);
910 if (pktmbuf_pool[socketid] == NULL) {
911 rte_snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
912 pktmbuf_pool[socketid] =
913 rte_mempool_create(s, NB_MBUF, MBUF_SIZE, 32,
914 sizeof(struct rte_pktmbuf_pool_private),
915 rte_pktmbuf_pool_init, NULL,
916 rte_pktmbuf_init, NULL,
918 if (pktmbuf_pool[socketid] == NULL)
919 rte_exit(EXIT_FAILURE,
920 "Cannot init mbuf pool on socket %d\n", socketid);
922 printf("Allocated mbuf pool on socket %d\n", socketid);
924 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
927 setup_hash(socketid);
930 qconf = &lcore_conf[lcore_id];
931 qconf->lookup_struct = l3fwd_lookup_struct[socketid];
937 MAIN(int argc, char **argv)
939 struct lcore_conf *qconf;
940 struct rte_eth_link link;
945 uint32_t n_tx_queue, nb_lcores;
946 uint8_t portid, nb_rx_queue, queue, socketid;
949 ret = rte_eal_init(argc, argv);
951 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
955 /* parse application arguments (after the EAL ones) */
956 ret = parse_args(argc, argv);
958 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
960 if (check_lcore_params() < 0)
961 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
963 ret = init_lcore_rx_queues();
965 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
969 rte_exit(EXIT_FAILURE, "init_mem failed\n");
972 #ifdef RTE_LIBRTE_IGB_PMD
973 if (rte_igb_pmd_init() < 0)
974 rte_exit(EXIT_FAILURE, "Cannot init igb pmd\n");
976 #ifdef RTE_LIBRTE_IXGBE_PMD
977 if (rte_ixgbe_pmd_init() < 0)
978 rte_exit(EXIT_FAILURE, "Cannot init ixgbe pmd\n");
981 if (rte_eal_pci_probe() < 0)
982 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
984 nb_ports = rte_eth_dev_count();
985 if (nb_ports > MAX_PORTS)
986 nb_ports = MAX_PORTS;
988 if (check_port_config(nb_ports) < 0)
989 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
991 nb_lcores = rte_lcore_count();
993 /* initialize all ports */
994 for (portid = 0; portid < nb_ports; portid++) {
995 /* skip ports that are not enabled */
996 if ((enabled_port_mask & (1 << portid)) == 0) {
997 printf("\nSkipping disabled port %d\n", portid);
1002 printf("Initializing port %d ... ", portid );
1005 nb_rx_queue = get_port_n_rx_queues(portid);
1006 n_tx_queue = nb_lcores;
1007 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1008 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1009 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1010 nb_rx_queue, (unsigned)n_tx_queue );
1011 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1012 (uint16_t)n_tx_queue, &port_conf);
1014 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
1017 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1018 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1022 /* init one TX queue per couple (lcore,port) */
1024 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1025 if (rte_lcore_is_enabled(lcore_id) == 0)
1029 socketid = (uint8_t)rte_lcore_to_socket_id(lcore_id);
1033 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
1035 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
1036 socketid, &tx_conf);
1038 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
1039 "port=%d\n", ret, portid);
1041 qconf = &lcore_conf[lcore_id];
1042 qconf->tx_queue_id[portid] = queueid;
1048 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1049 if (rte_lcore_is_enabled(lcore_id) == 0)
1051 qconf = &lcore_conf[lcore_id];
1052 printf("\nInitializing rx queues on lcore %u ... ", lcore_id );
1054 /* init RX queues */
1055 for(queue = 0; queue < qconf->n_rx_queue; ++queue) {
1056 portid = qconf->rx_queue_list[queue].port_id;
1057 queueid = qconf->rx_queue_list[queue].queue_id;
1060 socketid = (uint8_t)rte_lcore_to_socket_id(lcore_id);
1064 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
1067 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
1068 socketid, &rx_conf, pktmbuf_pool[socketid]);
1070 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: err=%d,"
1071 "port=%d\n", ret, portid);
1078 for (portid = 0; portid < nb_ports; portid++) {
1079 if ((enabled_port_mask & (1 << portid)) == 0) {
1083 ret = rte_eth_dev_start(portid);
1085 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1088 printf("done: Port %d ", portid);
1090 /* get link status */
1091 rte_eth_link_get(portid, &link);
1092 if (link.link_status) {
1093 printf(" Link Up - speed %u Mbps - %s\n",
1094 (unsigned) link.link_speed,
1095 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1096 ("full-duplex") : ("half-duplex\n"));
1098 printf(" Link Down\n");
1101 * If enabled, put device in promiscuous mode.
1102 * This allows IO forwarding mode to forward packets
1103 * to itself through 2 cross-connected ports of the
1107 rte_eth_promiscuous_enable(portid);
1110 /* launch per-lcore init on every lcore */
1111 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1112 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1113 if (rte_eal_wait_lcore(lcore_id) < 0)