4 * Copyright(c) 2010-2014 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.
38 #include <sys/types.h>
40 #include <sys/queue.h>
46 #include <rte_common.h>
47 #include <rte_byteorder.h>
49 #include <rte_memory.h>
50 #include <rte_memcpy.h>
51 #include <rte_memzone.h>
53 #include <rte_per_lcore.h>
54 #include <rte_launch.h>
55 #include <rte_atomic.h>
56 #include <rte_spinlock.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>
64 #include <rte_random.h>
65 #include <rte_debug.h>
66 #include <rte_ether.h>
67 #include <rte_ethdev.h>
68 #include <rte_mempool.h>
73 #include <rte_string_fns.h>
75 #define APP_LOOKUP_EXACT_MATCH 0
76 #define APP_LOOKUP_LPM 1
77 #define DO_RFC_1812_CHECKS
79 //#define APP_LOOKUP_METHOD APP_LOOKUP_EXACT_MATCH
80 #ifndef APP_LOOKUP_METHOD
81 #define APP_LOOKUP_METHOD APP_LOOKUP_LPM
84 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
86 #elif (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
89 #error "APP_LOOKUP_METHOD set to incorrect value"
92 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
94 #define MEMPOOL_CACHE_SIZE 256
97 * This expression is used to calculate the number of mbufs needed depending on user input, taking
98 * into account memory for rx and tx hardware rings, cache per lcore and mtable per port per lcore.
99 * RTE_MAX is used to ensure that NB_MBUF never goes below a minimum value of 8192
102 #define NB_MBUF RTE_MAX ( \
103 (nb_ports*nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
104 nb_ports*nb_lcores*MAX_PKT_BURST + \
105 nb_ports*n_tx_queue*RTE_TEST_TX_DESC_DEFAULT + \
106 nb_lcores*MEMPOOL_CACHE_SIZE), \
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.
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. */
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.
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. */
128 #define MAX_PKT_BURST 32
129 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
135 /* Configure how many packets ahead to prefetch, when reading packets */
136 #define PREFETCH_OFFSET 3
139 * Configurable number of RX/TX ring descriptors
141 #define RTE_TEST_RX_DESC_DEFAULT 128
142 #define RTE_TEST_TX_DESC_DEFAULT 512
143 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
144 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
146 /* ethernet addresses of ports */
147 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
149 /* mask of enabled ports */
150 static uint32_t enabled_port_mask = 0;
151 static int numa_on = 1; /**< NUMA is enabled by default. */
155 struct rte_mbuf *m_table[MAX_PKT_BURST];
158 struct lcore_rx_queue {
161 } __rte_cache_aligned;
163 #define MAX_RX_QUEUE_PER_LCORE 16
164 #define MAX_TX_QUEUE_PER_PORT 1
165 #define MAX_RX_QUEUE_PER_PORT 1
167 #define MAX_LCORE_PARAMS 1024
168 struct lcore_params {
172 } __rte_cache_aligned;
174 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
175 static struct lcore_params lcore_params_array_default[] = {
187 static struct lcore_params * lcore_params = lcore_params_array_default;
188 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
189 sizeof(lcore_params_array_default[0]);
191 static struct rte_eth_conf port_conf = {
193 .mq_mode = ETH_MQ_RX_RSS,
194 .max_rx_pkt_len = ETHER_MAX_LEN,
196 .header_split = 0, /**< Header Split disabled */
197 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
198 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
199 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
200 .hw_strip_crc = 1, /**< CRC stripped by hardware */
205 .rss_hf = ETH_RSS_IP,
209 .mq_mode = ETH_MQ_TX_NONE,
213 static struct rte_mempool * pktmbuf_pool[NB_SOCKETS];
216 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
218 #ifdef RTE_MACHINE_CPUFLAG_SSE4_2
219 #include <rte_hash_crc.h>
220 #define DEFAULT_HASH_FUNC rte_hash_crc
222 #include <rte_jhash.h>
223 #define DEFAULT_HASH_FUNC rte_jhash
232 } __attribute__((__packed__));
235 struct ipv4_5tuple key;
239 static struct l3fwd_route l3fwd_route_array[] = {
240 {{IPv4(100,10,0,1), IPv4(200,10,0,1), 101, 11, IPPROTO_TCP}, 0},
241 {{IPv4(100,20,0,2), IPv4(200,20,0,2), 102, 12, IPPROTO_TCP}, 1},
242 {{IPv4(100,30,0,3), IPv4(200,30,0,3), 103, 13, IPPROTO_TCP}, 2},
243 {{IPv4(100,40,0,4), IPv4(200,40,0,4), 104, 14, IPPROTO_TCP}, 3},
246 typedef struct rte_hash lookup_struct_t;
247 static lookup_struct_t *l3fwd_lookup_struct[NB_SOCKETS];
249 #define L3FWD_HASH_ENTRIES 1024
250 struct rte_hash_parameters l3fwd_hash_params = {
251 .name = "l3fwd_hash_0",
252 .entries = L3FWD_HASH_ENTRIES,
253 .key_len = sizeof(struct ipv4_5tuple),
254 .hash_func = DEFAULT_HASH_FUNC,
255 .hash_func_init_val = 0,
256 .socket_id = SOCKET0,
259 #define L3FWD_NUM_ROUTES \
260 (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
262 static uint8_t l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
265 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
272 static struct l3fwd_route l3fwd_route_array[] = {
273 {IPv4(1,1,1,0), 24, 0},
274 {IPv4(2,1,1,0), 24, 1},
275 {IPv4(3,1,1,0), 24, 2},
276 {IPv4(4,1,1,0), 24, 3},
277 {IPv4(5,1,1,0), 24, 4},
278 {IPv4(6,1,1,0), 24, 5},
279 {IPv4(7,1,1,0), 24, 6},
280 {IPv4(8,1,1,0), 24, 7},
283 #define L3FWD_NUM_ROUTES \
284 (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
286 #define L3FWD_LPM_MAX_RULES 1024
288 typedef struct rte_lpm lookup_struct_t;
289 static lookup_struct_t *l3fwd_lookup_struct[NB_SOCKETS];
294 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
295 uint16_t tx_queue_id;
296 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
297 lookup_struct_t * lookup_struct;
298 } __rte_cache_aligned;
300 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
301 static rte_spinlock_t spinlock_conf[RTE_MAX_ETHPORTS] = {RTE_SPINLOCK_INITIALIZER};
302 /* Send burst of packets on an output interface */
304 send_burst(struct lcore_conf *qconf, uint16_t n, uint8_t port)
306 struct rte_mbuf **m_table;
310 queueid = qconf->tx_queue_id;
311 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
313 rte_spinlock_lock(&spinlock_conf[port]);
314 ret = rte_eth_tx_burst(port, queueid, m_table, n);
315 rte_spinlock_unlock(&spinlock_conf[port]);
317 if (unlikely(ret < n)) {
319 rte_pktmbuf_free(m_table[ret]);
326 /* Enqueue a single packet, and send burst if queue is filled */
328 send_single_packet(struct rte_mbuf *m, uint8_t port)
332 struct lcore_conf *qconf;
334 lcore_id = rte_lcore_id();
336 qconf = &lcore_conf[lcore_id];
337 len = qconf->tx_mbufs[port].len;
338 qconf->tx_mbufs[port].m_table[len] = m;
341 /* enough pkts to be sent */
342 if (unlikely(len == MAX_PKT_BURST)) {
343 send_burst(qconf, MAX_PKT_BURST, port);
347 qconf->tx_mbufs[port].len = len;
351 #ifdef DO_RFC_1812_CHECKS
353 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
355 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
357 * 1. The packet length reported by the Link Layer must be large
358 * enough to hold the minimum length legal IP datagram (20 bytes).
360 if (link_len < sizeof(struct ipv4_hdr))
363 /* 2. The IP checksum must be correct. */
364 /* this is checked in H/W */
367 * 3. The IP version number must be 4. If the version number is not 4
368 * then the packet may be another version of IP, such as IPng or
371 if (((pkt->version_ihl) >> 4) != 4)
374 * 4. The IP header length field must be large enough to hold the
375 * minimum length legal IP datagram (20 bytes = 5 words).
377 if ((pkt->version_ihl & 0xf) < 5)
381 * 5. The IP total length field must be large enough to hold the IP
382 * datagram header, whose length is specified in the IP header length
385 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
392 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
394 print_key(struct ipv4_5tuple key)
396 printf("IP dst = %08x, IP src = %08x, port dst = %d, port src = %d, proto = %d\n",
397 (unsigned)key.ip_dst, (unsigned)key.ip_src, key.port_dst, key.port_src, key.proto);
400 static inline uint8_t
401 get_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
403 struct ipv4_5tuple key;
408 key.ip_dst = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
409 key.ip_src = rte_be_to_cpu_32(ipv4_hdr->src_addr);
410 key.proto = ipv4_hdr->next_proto_id;
412 switch (ipv4_hdr->next_proto_id) {
414 tcp = (struct tcp_hdr *)((unsigned char *) ipv4_hdr +
415 sizeof(struct ipv4_hdr));
416 key.port_dst = rte_be_to_cpu_16(tcp->dst_port);
417 key.port_src = rte_be_to_cpu_16(tcp->src_port);
421 udp = (struct udp_hdr *)((unsigned char *) ipv4_hdr +
422 sizeof(struct ipv4_hdr));
423 key.port_dst = rte_be_to_cpu_16(udp->dst_port);
424 key.port_src = rte_be_to_cpu_16(udp->src_port);
432 /* Find destination port */
433 ret = rte_hash_lookup(l3fwd_lookup_struct, (const void *)&key);
434 return (uint8_t)((ret < 0)? portid : l3fwd_out_if[ret]);
438 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
439 static inline uint8_t
440 get_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
444 return (uint8_t) ((rte_lpm_lookup(l3fwd_lookup_struct,
445 rte_be_to_cpu_32(ipv4_hdr->dst_addr), &next_hop) == 0)?
451 l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
453 struct ether_hdr *eth_hdr;
454 struct ipv4_hdr *ipv4_hdr;
458 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
460 ipv4_hdr = rte_pktmbuf_mtod_offset(m, struct ipv4_hdr *,
461 sizeof(struct ether_hdr));
463 #ifdef DO_RFC_1812_CHECKS
464 /* Check to make sure the packet is valid (RFC1812) */
465 if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt_len) < 0) {
471 dst_port = get_dst_port(ipv4_hdr, portid, l3fwd_lookup_struct);
472 if (dst_port >= RTE_MAX_ETHPORTS || (enabled_port_mask & 1 << dst_port) == 0)
475 /* 02:00:00:00:00:xx */
476 tmp = ð_hdr->d_addr.addr_bytes[0];
477 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
479 #ifdef DO_RFC_1812_CHECKS
480 /* Update time to live and header checksum */
481 --(ipv4_hdr->time_to_live);
482 ++(ipv4_hdr->hdr_checksum);
486 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr);
488 send_single_packet(m, dst_port);
492 /* main processing loop */
494 main_loop(__attribute__((unused)) void *dummy)
496 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
498 uint64_t prev_tsc, diff_tsc, cur_tsc;
500 uint8_t portid, queueid;
501 struct lcore_conf *qconf;
502 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
506 lcore_id = rte_lcore_id();
507 qconf = &lcore_conf[lcore_id];
509 if (qconf->n_rx_queue == 0) {
510 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
514 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
516 for (i = 0; i < qconf->n_rx_queue; i++) {
518 portid = qconf->rx_queue_list[i].port_id;
519 queueid = qconf->rx_queue_list[i].queue_id;
520 RTE_LOG(INFO, L3FWD, " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n", lcore_id,
526 cur_tsc = rte_rdtsc();
529 * TX burst queue drain
531 diff_tsc = cur_tsc - prev_tsc;
532 if (unlikely(diff_tsc > drain_tsc)) {
535 * This could be optimized (use queueid instead of
536 * portid), but it is not called so often
538 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
539 if (qconf->tx_mbufs[portid].len == 0)
541 send_burst(&lcore_conf[lcore_id],
542 qconf->tx_mbufs[portid].len,
544 qconf->tx_mbufs[portid].len = 0;
551 * Read packet from RX queues
553 for (i = 0; i < qconf->n_rx_queue; ++i) {
555 portid = qconf->rx_queue_list[i].port_id;
556 queueid = qconf->rx_queue_list[i].queue_id;
557 nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst, MAX_PKT_BURST);
559 /* Prefetch first packets */
560 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
561 rte_prefetch0(rte_pktmbuf_mtod(
562 pkts_burst[j], void *));
565 /* Prefetch and forward already prefetched packets */
566 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
567 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
568 j + PREFETCH_OFFSET], void *));
569 l3fwd_simple_forward(pkts_burst[j], portid, qconf->lookup_struct);
572 /* Forward remaining prefetched packets */
573 for (; j < nb_rx; j++) {
574 l3fwd_simple_forward(pkts_burst[j], portid, qconf->lookup_struct);
581 check_lcore_params(void)
583 uint8_t queue, lcore;
587 for (i = 0; i < nb_lcore_params; ++i) {
588 queue = lcore_params[i].queue_id;
589 if (queue >= MAX_RX_QUEUE_PER_PORT) {
590 printf("invalid queue number: %hhu\n", queue);
593 lcore = lcore_params[i].lcore_id;
594 if (!rte_lcore_is_enabled(lcore)) {
595 printf("error: lcore %hhu is not enabled in lcore mask\n", lcore);
598 if ((socketid = rte_lcore_to_socket_id(lcore) != 0) &&
600 printf("warning: lcore %hhu is on socket %d with numa off \n",
608 check_port_config(const unsigned nb_ports)
613 for (i = 0; i < nb_lcore_params; ++i) {
614 portid = lcore_params[i].port_id;
615 if ((enabled_port_mask & (1 << portid)) == 0) {
616 printf("port %u is not enabled in port mask\n", portid);
619 if (portid >= nb_ports) {
620 printf("port %u is not present on the board\n", portid);
628 get_port_n_rx_queues(const uint8_t port)
633 for (i = 0; i < nb_lcore_params; ++i) {
634 if (lcore_params[i].port_id == port && lcore_params[i].queue_id > queue)
635 queue = lcore_params[i].queue_id;
637 return (uint8_t)(++queue);
641 init_lcore_rx_queues(void)
643 uint16_t i, nb_rx_queue;
646 for (i = 0; i < nb_lcore_params; ++i) {
647 lcore = lcore_params[i].lcore_id;
648 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
649 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
650 printf("error: too many queues (%u) for lcore: %u\n",
651 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
654 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
655 lcore_params[i].port_id;
656 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
657 lcore_params[i].queue_id;
658 lcore_conf[lcore].n_rx_queue++;
666 print_usage(const char *prgname)
668 printf ("%s [EAL options] -- -p PORTMASK"
669 " [--config (port,queue,lcore)[,(port,queue,lcore]]\n"
670 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
671 " --config (port,queue,lcore): rx queues configuration\n"
672 " --no-numa: optional, disable numa awareness\n",
676 /* Custom handling of signals to handle process terminal */
678 signal_handler(int signum)
681 uint8_t nb_ports = rte_eth_dev_count();
683 /* When we receive a SIGINT signal */
684 if (signum == SIGINT) {
685 for (portid = 0; portid < nb_ports; portid++) {
686 /* skip ports that are not enabled */
687 if ((enabled_port_mask & (1 << portid)) == 0)
689 rte_eth_dev_close(portid);
692 rte_exit(EXIT_SUCCESS, "\n User forced exit\n");
695 parse_portmask(const char *portmask)
700 /* parse hexadecimal string */
701 pm = strtoul(portmask, &end, 16);
702 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
712 parse_config(const char *q_arg)
715 const char *p, *p0 = q_arg;
723 unsigned long int_fld[_NUM_FLD];
724 char *str_fld[_NUM_FLD];
730 while ((p = strchr(p0,'(')) != NULL) {
732 if((p0 = strchr(p,')')) == NULL)
736 if(size >= sizeof(s))
739 snprintf(s, sizeof(s), "%.*s", size, p);
740 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') != _NUM_FLD)
742 for (i = 0; i < _NUM_FLD; i++){
744 int_fld[i] = strtoul(str_fld[i], &end, 0);
745 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
748 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
749 printf("exceeded max number of lcore params: %hu\n",
753 lcore_params_array[nb_lcore_params].port_id = (uint8_t)int_fld[FLD_PORT];
754 lcore_params_array[nb_lcore_params].queue_id = (uint8_t)int_fld[FLD_QUEUE];
755 lcore_params_array[nb_lcore_params].lcore_id = (uint8_t)int_fld[FLD_LCORE];
758 lcore_params = lcore_params_array;
762 /* Parse the argument given in the command line of the application */
764 parse_args(int argc, char **argv)
769 char *prgname = argv[0];
770 static struct option lgopts[] = {
772 {"no-numa", 0, 0, 0},
778 while ((opt = getopt_long(argc, argvopt, "p:",
779 lgopts, &option_index)) != EOF) {
784 enabled_port_mask = parse_portmask(optarg);
785 if (enabled_port_mask == 0) {
786 printf("invalid portmask\n");
787 print_usage(prgname);
794 if (!strcmp(lgopts[option_index].name, "config")) {
795 ret = parse_config(optarg);
797 printf("invalid config\n");
798 print_usage(prgname);
803 if (!strcmp(lgopts[option_index].name, "no-numa")) {
804 printf("numa is disabled \n");
810 print_usage(prgname);
816 argv[optind-1] = prgname;
819 optind = 1; /* reset getopt lib */
824 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
826 char buf[ETHER_ADDR_FMT_SIZE];
827 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
828 printf("%s%s", name, buf);
831 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
833 setup_hash(int socketid)
840 snprintf(s, sizeof(s), "l3fwd_hash_%d", socketid);
841 l3fwd_hash_params.name = s;
842 l3fwd_hash_params.socket_id = socketid;
843 l3fwd_lookup_struct[socketid] = rte_hash_create(&l3fwd_hash_params);
844 if (l3fwd_lookup_struct[socketid] == NULL)
845 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd hash on "
846 "socket %d\n", socketid);
848 /* populate the hash */
849 for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
850 ret = rte_hash_add_key (l3fwd_lookup_struct[socketid],
851 (void *) &l3fwd_route_array[i].key);
853 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the"
854 "l3fwd hash on socket %d\n", i, socketid);
856 l3fwd_out_if[ret] = l3fwd_route_array[i].if_out;
857 printf("Hash: Adding key\n");
858 print_key(l3fwd_route_array[i].key);
863 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
865 setup_lpm(int socketid)
871 struct rte_lpm_config lpm_ipv4_config;
873 lpm_ipv4_config.max_rules = L3FWD_LPM_MAX_RULES;
874 lpm_ipv4_config.number_tbl8s = 256;
875 lpm_ipv4_config.flags = 0;
877 /* create the LPM table */
878 snprintf(s, sizeof(s), "L3FWD_LPM_%d", socketid);
879 l3fwd_lookup_struct[socketid] =
880 rte_lpm_create(s, socketid, &lpm_ipv4_config);
881 if (l3fwd_lookup_struct[socketid] == NULL)
882 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd LPM table"
883 " on socket %d\n", socketid);
885 /* populate the LPM table */
886 for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
887 ret = rte_lpm_add(l3fwd_lookup_struct[socketid],
888 l3fwd_route_array[i].ip,
889 l3fwd_route_array[i].depth,
890 l3fwd_route_array[i].if_out);
893 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the "
894 "l3fwd LPM table on socket %d\n",
898 printf("LPM: Adding route 0x%08x / %d (%d)\n",
899 (unsigned)l3fwd_route_array[i].ip,
900 l3fwd_route_array[i].depth,
901 l3fwd_route_array[i].if_out);
907 init_mem(unsigned nb_mbuf)
909 struct lcore_conf *qconf;
914 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
915 if (rte_lcore_is_enabled(lcore_id) == 0)
919 socketid = rte_lcore_to_socket_id(lcore_id);
923 if (socketid >= NB_SOCKETS) {
924 rte_exit(EXIT_FAILURE, "Socket %d of lcore %u is out of range %d\n",
925 socketid, lcore_id, NB_SOCKETS);
927 if (pktmbuf_pool[socketid] == NULL) {
928 snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
929 pktmbuf_pool[socketid] = rte_pktmbuf_pool_create(s,
930 nb_mbuf, MEMPOOL_CACHE_SIZE, 0,
931 RTE_MBUF_DEFAULT_BUF_SIZE, socketid);
932 if (pktmbuf_pool[socketid] == NULL)
933 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool on socket %d\n", socketid);
935 printf("Allocated mbuf pool on socket %d\n", socketid);
937 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
940 setup_hash(socketid);
943 qconf = &lcore_conf[lcore_id];
944 qconf->lookup_struct = l3fwd_lookup_struct[socketid];
950 main(int argc, char **argv)
952 struct lcore_conf *qconf;
953 struct rte_eth_dev_info dev_info;
954 struct rte_eth_txconf *txconf;
961 uint8_t portid, nb_rx_queue, queue, socketid;
963 signal(SIGINT, signal_handler);
965 ret = rte_eal_init(argc, argv);
967 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
971 /* parse application arguments (after the EAL ones) */
972 ret = parse_args(argc, argv);
974 rte_exit(EXIT_FAILURE, "Invalid L3FWD-VF parameters\n");
976 if (check_lcore_params() < 0)
977 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
979 ret = init_lcore_rx_queues();
981 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
983 nb_ports = rte_eth_dev_count();
985 if (check_port_config(nb_ports) < 0)
986 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
988 nb_lcores = rte_lcore_count();
990 /* initialize all ports */
991 for (portid = 0; portid < nb_ports; portid++) {
992 /* skip ports that are not enabled */
993 if ((enabled_port_mask & (1 << portid)) == 0) {
994 printf("\nSkipping disabled port %d\n", portid);
999 printf("Initializing port %d ... ", portid );
1002 /* must always equal(=1) */
1003 nb_rx_queue = get_port_n_rx_queues(portid);
1004 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1006 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1007 nb_rx_queue, (unsigned)1 );
1008 ret = rte_eth_dev_configure(portid, nb_rx_queue, n_tx_queue, &port_conf);
1010 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
1013 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1014 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1017 ret = init_mem(NB_MBUF);
1019 rte_exit(EXIT_FAILURE, "init_mem failed\n");
1021 /* init one TX queue */
1022 socketid = (uint8_t)rte_lcore_to_socket_id(rte_get_master_lcore());
1024 printf("txq=%d,%d,%d ", portid, 0, socketid);
1027 rte_eth_dev_info_get(portid, &dev_info);
1028 txconf = &dev_info.default_txconf;
1029 if (port_conf.rxmode.jumbo_frame)
1030 txconf->txq_flags = 0;
1031 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
1034 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
1035 "port=%d\n", ret, portid);
1040 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1041 if (rte_lcore_is_enabled(lcore_id) == 0)
1043 qconf = &lcore_conf[lcore_id];
1044 qconf->tx_queue_id = 0;
1046 printf("\nInitializing rx queues on lcore %u ... ", lcore_id );
1048 /* init RX queues */
1049 for(queue = 0; queue < qconf->n_rx_queue; ++queue) {
1050 portid = qconf->rx_queue_list[queue].port_id;
1051 queueid = qconf->rx_queue_list[queue].queue_id;
1054 socketid = (uint8_t)rte_lcore_to_socket_id(lcore_id);
1058 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
1061 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
1063 pktmbuf_pool[socketid]);
1065 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: err=%d,"
1066 "port=%d\n", ret, portid);
1072 for (portid = 0; portid < nb_ports; portid++) {
1073 if ((enabled_port_mask & (1 << portid)) == 0) {
1077 ret = rte_eth_dev_start(portid);
1079 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1082 printf("done: Port %d\n", portid);
1086 /* launch per-lcore init on every lcore */
1087 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1088 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1089 if (rte_eal_wait_lcore(lcore_id) < 0)