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>
52 #include <rte_tailq.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>
64 #include <rte_random.h>
65 #include <rte_debug.h>
66 #include <rte_ether.h>
67 #include <rte_ethdev.h>
69 #include <rte_mempool.h>
74 #include <rte_string_fns.h>
78 #define APP_LOOKUP_EXACT_MATCH 0
79 #define APP_LOOKUP_LPM 1
80 #define DO_RFC_1812_CHECKS
82 //#define APP_LOOKUP_METHOD APP_LOOKUP_EXACT_MATCH
83 #ifndef APP_LOOKUP_METHOD
84 #define APP_LOOKUP_METHOD APP_LOOKUP_LPM
87 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
89 #elif (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
92 #error "APP_LOOKUP_METHOD set to incorrect value"
95 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
97 #define MEMPOOL_CACHE_SIZE 256
99 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
102 * This expression is used to calculate the number of mbufs needed depending on user input, taking
103 * into account memory for rx and tx hardware rings, cache per lcore and mtable per port per lcore.
104 * RTE_MAX is used to ensure that NB_MBUF never goes below a minimum value of 8192
107 #define NB_MBUF RTE_MAX ( \
108 (nb_ports*nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
109 nb_ports*nb_lcores*MAX_PKT_BURST + \
110 nb_ports*n_tx_queue*RTE_TEST_TX_DESC_DEFAULT + \
111 nb_lcores*MEMPOOL_CACHE_SIZE), \
115 * RX and TX Prefetch, Host, and Write-back threshold values should be
116 * carefully set for optimal performance. Consult the network
117 * controller's datasheet and supporting DPDK documentation for guidance
118 * on how these parameters should be set.
120 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
121 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
122 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
125 * These default values are optimized for use with the Intel(R) 82599 10 GbE
126 * Controller and the DPDK ixgbe PMD. Consider using other values for other
127 * network controllers and/or network drivers.
129 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
130 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */
131 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */
133 #define MAX_PKT_BURST 32
134 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
140 /* Configure how many packets ahead to prefetch, when reading packets */
141 #define PREFETCH_OFFSET 3
144 * Configurable number of RX/TX ring descriptors
146 #define RTE_TEST_RX_DESC_DEFAULT 128
147 #define RTE_TEST_TX_DESC_DEFAULT 512
148 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
149 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
151 /* ethernet addresses of ports */
152 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
154 /* mask of enabled ports */
155 static uint32_t enabled_port_mask = 0;
156 static int numa_on = 1; /**< NUMA is enabled by default. */
160 struct rte_mbuf *m_table[MAX_PKT_BURST];
163 struct lcore_rx_queue {
166 } __rte_cache_aligned;
168 #define MAX_RX_QUEUE_PER_LCORE 16
169 #define MAX_TX_QUEUE_PER_PORT 1
170 #define MAX_RX_QUEUE_PER_PORT 1
172 #define MAX_LCORE_PARAMS 1024
173 struct lcore_params {
177 } __rte_cache_aligned;
179 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
180 static struct lcore_params lcore_params_array_default[] = {
192 static struct lcore_params * lcore_params = lcore_params_array_default;
193 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
194 sizeof(lcore_params_array_default[0]);
196 static struct rte_eth_conf port_conf = {
198 .mq_mode = ETH_MQ_RX_RSS,
199 .max_rx_pkt_len = ETHER_MAX_LEN,
201 .header_split = 0, /**< Header Split disabled */
202 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
203 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
204 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
205 .hw_strip_crc = 0, /**< CRC stripped by hardware */
210 .rss_hf = ETH_RSS_IPV4 | ETH_RSS_IPV6,
214 .mq_mode = ETH_MQ_TX_NONE,
218 static const struct rte_eth_rxconf rx_conf = {
220 .pthresh = RX_PTHRESH,
221 .hthresh = RX_HTHRESH,
222 .wthresh = RX_WTHRESH,
224 .rx_free_thresh = 32,
227 static const struct rte_eth_txconf tx_conf = {
229 .pthresh = TX_PTHRESH,
230 .hthresh = TX_HTHRESH,
231 .wthresh = TX_WTHRESH,
233 .tx_free_thresh = 0, /* Use PMD default values */
234 .tx_rs_thresh = 0, /* Use PMD default values */
235 .txq_flags = (ETH_TXQ_FLAGS_NOMULTSEGS |
236 ETH_TXQ_FLAGS_NOVLANOFFL |
237 ETH_TXQ_FLAGS_NOXSUMSCTP |
238 ETH_TXQ_FLAGS_NOXSUMUDP |
239 ETH_TXQ_FLAGS_NOXSUMTCP)
242 static struct rte_mempool * pktmbuf_pool[NB_SOCKETS];
245 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
247 #ifdef RTE_MACHINE_CPUFLAG_SSE4_2
248 #include <rte_hash_crc.h>
249 #define DEFAULT_HASH_FUNC rte_hash_crc
251 #include <rte_jhash.h>
252 #define DEFAULT_HASH_FUNC rte_jhash
261 } __attribute__((__packed__));
264 struct ipv4_5tuple key;
268 static struct l3fwd_route l3fwd_route_array[] = {
269 {{IPv4(100,10,0,1), IPv4(200,10,0,1), 101, 11, IPPROTO_TCP}, 0},
270 {{IPv4(100,20,0,2), IPv4(200,20,0,2), 102, 12, IPPROTO_TCP}, 1},
271 {{IPv4(100,30,0,3), IPv4(200,30,0,3), 103, 13, IPPROTO_TCP}, 2},
272 {{IPv4(100,40,0,4), IPv4(200,40,0,4), 104, 14, IPPROTO_TCP}, 3},
275 typedef struct rte_hash lookup_struct_t;
276 static lookup_struct_t *l3fwd_lookup_struct[NB_SOCKETS];
278 #define L3FWD_HASH_ENTRIES 1024
279 struct rte_hash_parameters l3fwd_hash_params = {
280 .name = "l3fwd_hash_0",
281 .entries = L3FWD_HASH_ENTRIES,
283 .key_len = sizeof(struct ipv4_5tuple),
284 .hash_func = DEFAULT_HASH_FUNC,
285 .hash_func_init_val = 0,
286 .socket_id = SOCKET0,
289 #define L3FWD_NUM_ROUTES \
290 (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
292 static uint8_t l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
295 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
302 static struct l3fwd_route l3fwd_route_array[] = {
303 {IPv4(1,1,1,0), 24, 0},
304 {IPv4(2,1,1,0), 24, 1},
305 {IPv4(3,1,1,0), 24, 2},
306 {IPv4(4,1,1,0), 24, 3},
307 {IPv4(5,1,1,0), 24, 4},
308 {IPv4(6,1,1,0), 24, 5},
309 {IPv4(7,1,1,0), 24, 6},
310 {IPv4(8,1,1,0), 24, 7},
313 #define L3FWD_NUM_ROUTES \
314 (sizeof(l3fwd_route_array) / sizeof(l3fwd_route_array[0]))
316 #define L3FWD_LPM_MAX_RULES 1024
318 typedef struct rte_lpm lookup_struct_t;
319 static lookup_struct_t *l3fwd_lookup_struct[NB_SOCKETS];
324 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
325 uint16_t tx_queue_id;
326 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
327 lookup_struct_t * lookup_struct;
328 } __rte_cache_aligned;
330 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
332 /* Send burst of packets on an output interface */
334 send_burst(struct lcore_conf *qconf, uint16_t n, uint8_t port)
336 struct rte_mbuf **m_table;
340 queueid = qconf->tx_queue_id;
341 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
343 ret = rte_eth_tx_burst(port, queueid, m_table, n);
344 if (unlikely(ret < n)) {
346 rte_pktmbuf_free(m_table[ret]);
353 /* Enqueue a single packet, and send burst if queue is filled */
355 send_single_packet(struct rte_mbuf *m, uint8_t port)
359 struct lcore_conf *qconf;
361 lcore_id = rte_lcore_id();
363 qconf = &lcore_conf[lcore_id];
364 len = qconf->tx_mbufs[port].len;
365 qconf->tx_mbufs[port].m_table[len] = m;
368 /* enough pkts to be sent */
369 if (unlikely(len == MAX_PKT_BURST)) {
370 send_burst(qconf, MAX_PKT_BURST, port);
374 qconf->tx_mbufs[port].len = len;
378 #ifdef DO_RFC_1812_CHECKS
380 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
382 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
384 * 1. The packet length reported by the Link Layer must be large
385 * enough to hold the minimum length legal IP datagram (20 bytes).
387 if (link_len < sizeof(struct ipv4_hdr))
390 /* 2. The IP checksum must be correct. */
391 /* this is checked in H/W */
394 * 3. The IP version number must be 4. If the version number is not 4
395 * then the packet may be another version of IP, such as IPng or
398 if (((pkt->version_ihl) >> 4) != 4)
401 * 4. The IP header length field must be large enough to hold the
402 * minimum length legal IP datagram (20 bytes = 5 words).
404 if ((pkt->version_ihl & 0xf) < 5)
408 * 5. The IP total length field must be large enough to hold the IP
409 * datagram header, whose length is specified in the IP header length
412 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
419 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
421 print_key(struct ipv4_5tuple key)
423 printf("IP dst = %08x, IP src = %08x, port dst = %d, port src = %d, proto = %d\n",
424 (unsigned)key.ip_dst, (unsigned)key.ip_src, key.port_dst, key.port_src, key.proto);
427 static inline uint8_t
428 get_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
430 struct ipv4_5tuple key;
435 key.ip_dst = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
436 key.ip_src = rte_be_to_cpu_32(ipv4_hdr->src_addr);
437 key.proto = ipv4_hdr->next_proto_id;
439 switch (ipv4_hdr->next_proto_id) {
441 tcp = (struct tcp_hdr *)((unsigned char *) ipv4_hdr +
442 sizeof(struct ipv4_hdr));
443 key.port_dst = rte_be_to_cpu_16(tcp->dst_port);
444 key.port_src = rte_be_to_cpu_16(tcp->src_port);
448 udp = (struct udp_hdr *)((unsigned char *) ipv4_hdr +
449 sizeof(struct ipv4_hdr));
450 key.port_dst = rte_be_to_cpu_16(udp->dst_port);
451 key.port_src = rte_be_to_cpu_16(udp->src_port);
459 /* Find destination port */
460 ret = rte_hash_lookup(l3fwd_lookup_struct, (const void *)&key);
461 return (uint8_t)((ret < 0)? portid : l3fwd_out_if[ret]);
465 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
466 static inline uint8_t
467 get_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
471 return (uint8_t) ((rte_lpm_lookup(l3fwd_lookup_struct,
472 rte_be_to_cpu_32(ipv4_hdr->dst_addr), &next_hop) == 0)?
478 l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid, lookup_struct_t * l3fwd_lookup_struct)
480 struct ether_hdr *eth_hdr;
481 struct ipv4_hdr *ipv4_hdr;
485 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
487 ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(m, unsigned char *) +
488 sizeof(struct ether_hdr));
490 #ifdef DO_RFC_1812_CHECKS
491 /* Check to make sure the packet is valid (RFC1812) */
492 if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt.pkt_len) < 0) {
498 dst_port = get_dst_port(ipv4_hdr, portid, l3fwd_lookup_struct);
499 if (dst_port >= RTE_MAX_ETHPORTS || (enabled_port_mask & 1 << dst_port) == 0)
502 /* 02:00:00:00:00:xx */
503 tmp = ð_hdr->d_addr.addr_bytes[0];
504 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
506 #ifdef DO_RFC_1812_CHECKS
507 /* Update time to live and header checksum */
508 --(ipv4_hdr->time_to_live);
509 ++(ipv4_hdr->hdr_checksum);
513 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr);
515 send_single_packet(m, dst_port);
519 /* main processing loop */
521 main_loop(__attribute__((unused)) void *dummy)
523 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
525 uint64_t prev_tsc, diff_tsc, cur_tsc;
527 uint8_t portid, queueid;
528 struct lcore_conf *qconf;
529 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
533 lcore_id = rte_lcore_id();
534 qconf = &lcore_conf[lcore_id];
536 if (qconf->n_rx_queue == 0) {
537 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
541 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
543 for (i = 0; i < qconf->n_rx_queue; i++) {
545 portid = qconf->rx_queue_list[i].port_id;
546 queueid = qconf->rx_queue_list[i].queue_id;
547 RTE_LOG(INFO, L3FWD, " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n", lcore_id,
553 cur_tsc = rte_rdtsc();
556 * TX burst queue drain
558 diff_tsc = cur_tsc - prev_tsc;
559 if (unlikely(diff_tsc > drain_tsc)) {
562 * This could be optimized (use queueid instead of
563 * portid), but it is not called so often
565 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
566 if (qconf->tx_mbufs[portid].len == 0)
568 send_burst(&lcore_conf[lcore_id],
569 qconf->tx_mbufs[portid].len,
571 qconf->tx_mbufs[portid].len = 0;
578 * Read packet from RX queues
580 for (i = 0; i < qconf->n_rx_queue; ++i) {
582 portid = qconf->rx_queue_list[i].port_id;
583 queueid = qconf->rx_queue_list[i].queue_id;
584 nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst, MAX_PKT_BURST);
586 /* Prefetch first packets */
587 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
588 rte_prefetch0(rte_pktmbuf_mtod(
589 pkts_burst[j], void *));
592 /* Prefetch and forward already prefetched packets */
593 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
594 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
595 j + PREFETCH_OFFSET], void *));
596 l3fwd_simple_forward(pkts_burst[j], portid, qconf->lookup_struct);
599 /* Forward remaining prefetched packets */
600 for (; j < nb_rx; j++) {
601 l3fwd_simple_forward(pkts_burst[j], portid, qconf->lookup_struct);
608 check_lcore_params(void)
610 uint8_t queue, lcore;
614 for (i = 0; i < nb_lcore_params; ++i) {
615 queue = lcore_params[i].queue_id;
616 if (queue >= MAX_RX_QUEUE_PER_PORT) {
617 printf("invalid queue number: %hhu\n", queue);
620 lcore = lcore_params[i].lcore_id;
621 if (!rte_lcore_is_enabled(lcore)) {
622 printf("error: lcore %hhu is not enabled in lcore mask\n", lcore);
625 if ((socketid = rte_lcore_to_socket_id(lcore) != 0) &&
627 printf("warning: lcore %hhu is on socket %d with numa off \n",
635 check_port_config(const unsigned nb_ports)
640 for (i = 0; i < nb_lcore_params; ++i) {
641 portid = lcore_params[i].port_id;
642 if ((enabled_port_mask & (1 << portid)) == 0) {
643 printf("port %u is not enabled in port mask\n", portid);
646 if (portid >= nb_ports) {
647 printf("port %u is not present on the board\n", portid);
655 get_port_n_rx_queues(const uint8_t port)
660 for (i = 0; i < nb_lcore_params; ++i) {
661 if (lcore_params[i].port_id == port && lcore_params[i].queue_id > queue)
662 queue = lcore_params[i].queue_id;
664 return (uint8_t)(++queue);
668 init_lcore_rx_queues(void)
670 uint16_t i, nb_rx_queue;
673 for (i = 0; i < nb_lcore_params; ++i) {
674 lcore = lcore_params[i].lcore_id;
675 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
676 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
677 printf("error: too many queues (%u) for lcore: %u\n",
678 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
681 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
682 lcore_params[i].port_id;
683 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
684 lcore_params[i].queue_id;
685 lcore_conf[lcore].n_rx_queue++;
693 print_usage(const char *prgname)
695 printf ("%s [EAL options] -- -p PORTMASK"
696 " [--config (port,queue,lcore)[,(port,queue,lcore]]\n"
697 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
698 " --config (port,queue,lcore): rx queues configuration\n"
699 " --no-numa: optional, disable numa awareness\n",
703 /* Custom handling of signals to handle process terminal */
705 signal_handler(int signum)
708 uint8_t nb_ports = rte_eth_dev_count();
710 /* When we receive a SIGINT signal */
711 if (signum == SIGINT) {
712 for (portid = 0; portid < nb_ports; portid++) {
713 /* skip ports that are not enabled */
714 if ((enabled_port_mask & (1 << portid)) == 0)
716 rte_eth_dev_close(portid);
719 rte_exit(EXIT_SUCCESS, "\n User forced exit\n");
722 parse_portmask(const char *portmask)
727 /* parse hexadecimal string */
728 pm = strtoul(portmask, &end, 16);
729 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
739 parse_config(const char *q_arg)
742 const char *p, *p0 = q_arg;
750 unsigned long int_fld[_NUM_FLD];
751 char *str_fld[_NUM_FLD];
757 while ((p = strchr(p0,'(')) != NULL) {
759 if((p0 = strchr(p,')')) == NULL)
763 if(size >= sizeof(s))
766 rte_snprintf(s, sizeof(s), "%.*s", size, p);
767 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') != _NUM_FLD)
769 for (i = 0; i < _NUM_FLD; i++){
771 int_fld[i] = strtoul(str_fld[i], &end, 0);
772 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
775 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
776 printf("exceeded max number of lcore params: %hu\n",
780 lcore_params_array[nb_lcore_params].port_id = (uint8_t)int_fld[FLD_PORT];
781 lcore_params_array[nb_lcore_params].queue_id = (uint8_t)int_fld[FLD_QUEUE];
782 lcore_params_array[nb_lcore_params].lcore_id = (uint8_t)int_fld[FLD_LCORE];
785 lcore_params = lcore_params_array;
789 /* Parse the argument given in the command line of the application */
791 parse_args(int argc, char **argv)
796 char *prgname = argv[0];
797 static struct option lgopts[] = {
799 {"no-numa", 0, 0, 0},
805 while ((opt = getopt_long(argc, argvopt, "p:",
806 lgopts, &option_index)) != EOF) {
811 enabled_port_mask = parse_portmask(optarg);
812 if (enabled_port_mask == 0) {
813 printf("invalid portmask\n");
814 print_usage(prgname);
821 if (!strcmp(lgopts[option_index].name, "config")) {
822 ret = parse_config(optarg);
824 printf("invalid config\n");
825 print_usage(prgname);
830 if (!strcmp(lgopts[option_index].name, "no-numa")) {
831 printf("numa is disabled \n");
837 print_usage(prgname);
843 argv[optind-1] = prgname;
846 optind = 0; /* reset getopt lib */
851 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
853 printf ("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
854 eth_addr->addr_bytes[0],
855 eth_addr->addr_bytes[1],
856 eth_addr->addr_bytes[2],
857 eth_addr->addr_bytes[3],
858 eth_addr->addr_bytes[4],
859 eth_addr->addr_bytes[5]);
862 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
864 setup_hash(int socketid)
871 rte_snprintf(s, sizeof(s), "l3fwd_hash_%d", socketid);
872 l3fwd_hash_params.name = s;
873 l3fwd_hash_params.socket_id = socketid;
874 l3fwd_lookup_struct[socketid] = rte_hash_create(&l3fwd_hash_params);
875 if (l3fwd_lookup_struct[socketid] == NULL)
876 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd hash on "
877 "socket %d\n", socketid);
879 /* populate the hash */
880 for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
881 ret = rte_hash_add_key (l3fwd_lookup_struct[socketid],
882 (void *) &l3fwd_route_array[i].key);
884 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the"
885 "l3fwd hash on socket %d\n", i, socketid);
887 l3fwd_out_if[ret] = l3fwd_route_array[i].if_out;
888 printf("Hash: Adding key\n");
889 print_key(l3fwd_route_array[i].key);
894 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
896 setup_lpm(int socketid)
902 /* create the LPM table */
903 rte_snprintf(s, sizeof(s), "L3FWD_LPM_%d", socketid);
904 l3fwd_lookup_struct[socketid] = rte_lpm_create(s, socketid,
905 L3FWD_LPM_MAX_RULES, 0);
906 if (l3fwd_lookup_struct[socketid] == NULL)
907 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd LPM table"
908 " on socket %d\n", socketid);
910 /* populate the LPM table */
911 for (i = 0; i < L3FWD_NUM_ROUTES; i++) {
912 ret = rte_lpm_add(l3fwd_lookup_struct[socketid],
913 l3fwd_route_array[i].ip,
914 l3fwd_route_array[i].depth,
915 l3fwd_route_array[i].if_out);
918 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the "
919 "l3fwd LPM table on socket %d\n",
923 printf("LPM: Adding route 0x%08x / %d (%d)\n",
924 (unsigned)l3fwd_route_array[i].ip,
925 l3fwd_route_array[i].depth,
926 l3fwd_route_array[i].if_out);
932 init_mem(unsigned nb_mbuf)
934 struct lcore_conf *qconf;
939 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
940 if (rte_lcore_is_enabled(lcore_id) == 0)
944 socketid = rte_lcore_to_socket_id(lcore_id);
948 if (socketid >= NB_SOCKETS) {
949 rte_exit(EXIT_FAILURE, "Socket %d of lcore %u is out of range %d\n",
950 socketid, lcore_id, NB_SOCKETS);
952 if (pktmbuf_pool[socketid] == NULL) {
953 rte_snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
954 pktmbuf_pool[socketid] =
955 rte_mempool_create(s, nb_mbuf, MBUF_SIZE,
957 sizeof(struct rte_pktmbuf_pool_private),
958 rte_pktmbuf_pool_init, NULL,
959 rte_pktmbuf_init, NULL,
961 if (pktmbuf_pool[socketid] == NULL)
962 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool on socket %d\n", socketid);
964 printf("Allocated mbuf pool on socket %d\n", socketid);
966 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
969 setup_hash(socketid);
972 qconf = &lcore_conf[lcore_id];
973 qconf->lookup_struct = l3fwd_lookup_struct[socketid];
979 MAIN(int argc, char **argv)
981 struct lcore_conf *qconf;
988 uint8_t portid, nb_rx_queue, queue, socketid;
990 signal(SIGINT, signal_handler);
992 ret = rte_eal_init(argc, argv);
994 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
998 /* parse application arguments (after the EAL ones) */
999 ret = parse_args(argc, argv);
1001 rte_exit(EXIT_FAILURE, "Invalid L3FWD-VF parameters\n");
1003 if (check_lcore_params() < 0)
1004 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1006 ret = init_lcore_rx_queues();
1008 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1011 if (rte_pmd_init_all() < 0)
1012 rte_exit(EXIT_FAILURE, "Cannot init pmd\n");
1014 if (rte_eal_pci_probe() < 0)
1015 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
1017 nb_ports = rte_eth_dev_count();
1018 if (nb_ports > RTE_MAX_ETHPORTS)
1019 nb_ports = RTE_MAX_ETHPORTS;
1021 if (check_port_config(nb_ports) < 0)
1022 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1024 nb_lcores = rte_lcore_count();
1026 /* initialize all ports */
1027 for (portid = 0; portid < nb_ports; portid++) {
1028 /* skip ports that are not enabled */
1029 if ((enabled_port_mask & (1 << portid)) == 0) {
1030 printf("\nSkipping disabled port %d\n", portid);
1035 printf("Initializing port %d ... ", portid );
1038 /* must always equal(=1) */
1039 nb_rx_queue = get_port_n_rx_queues(portid);
1040 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1042 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1043 nb_rx_queue, (unsigned)1 );
1044 ret = rte_eth_dev_configure(portid, nb_rx_queue, n_tx_queue, &port_conf);
1046 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
1049 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1050 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1053 ret = init_mem(NB_MBUF);
1055 rte_exit(EXIT_FAILURE, "init_mem failed\n");
1057 /* init one TX queue */
1058 socketid = (uint8_t)rte_lcore_to_socket_id(rte_get_master_lcore());
1060 printf("txq=%d,%d,%d ", portid, 0, socketid);
1062 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
1063 socketid, &tx_conf);
1065 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
1066 "port=%d\n", ret, portid);
1071 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1072 if (rte_lcore_is_enabled(lcore_id) == 0)
1074 qconf = &lcore_conf[lcore_id];
1075 qconf->tx_queue_id = 0;
1077 printf("\nInitializing rx queues on lcore %u ... ", lcore_id );
1079 /* init RX queues */
1080 for(queue = 0; queue < qconf->n_rx_queue; ++queue) {
1081 portid = qconf->rx_queue_list[queue].port_id;
1082 queueid = qconf->rx_queue_list[queue].queue_id;
1085 socketid = (uint8_t)rte_lcore_to_socket_id(lcore_id);
1089 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
1092 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
1093 socketid, &rx_conf, pktmbuf_pool[socketid]);
1095 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: err=%d,"
1096 "port=%d\n", ret, portid);
1102 for (portid = 0; portid < nb_ports; portid++) {
1103 if ((enabled_port_mask & (1 << portid)) == 0) {
1107 ret = rte_eth_dev_start(portid);
1109 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1112 printf("done: Port %d\n", portid);
1116 /* launch per-lcore init on every lcore */
1117 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1118 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1119 if (rte_eal_wait_lcore(lcore_id) < 0)