4 * Copyright(c) 2010-2015 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.
39 #include <sys/types.h>
40 #include <sys/queue.h>
41 #include <netinet/in.h>
48 #include <rte_common.h>
50 #include <rte_memory.h>
51 #include <rte_memcpy.h>
52 #include <rte_memzone.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>
71 #include <rte_timer.h>
72 #include <rte_keepalive.h>
74 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
78 #define MAX_PKT_BURST 32
79 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
82 * Configurable number of RX/TX ring descriptors
84 #define RTE_TEST_RX_DESC_DEFAULT 128
85 #define RTE_TEST_TX_DESC_DEFAULT 512
86 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
87 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
89 /* ethernet addresses of ports */
90 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
92 /* mask of enabled ports */
93 static uint32_t l2fwd_enabled_port_mask;
95 /* list of enabled ports */
96 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
98 static unsigned int l2fwd_rx_queue_per_lcore = 1;
102 struct rte_mbuf *m_table[MAX_PKT_BURST];
105 #define MAX_RX_QUEUE_PER_LCORE 16
106 #define MAX_TX_QUEUE_PER_PORT 16
107 struct lcore_queue_conf {
109 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
110 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
112 } __rte_cache_aligned;
113 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
115 static const struct rte_eth_conf port_conf = {
118 .header_split = 0, /**< Header Split disabled */
119 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
120 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
121 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
122 .hw_strip_crc = 0, /**< CRC stripped by hardware */
125 .mq_mode = ETH_MQ_TX_NONE,
129 struct rte_mempool *l2fwd_pktmbuf_pool = NULL;
131 /* Per-port statistics struct */
132 struct l2fwd_port_statistics {
136 } __rte_cache_aligned;
137 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
139 /* A tsc-based timer responsible for triggering statistics printout */
140 #define TIMER_MILLISECOND 1
141 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
142 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* 10 seconds */
143 static int64_t check_period = 5; /* default check cycle is 5ms */
145 /* Keepalive structure */
146 struct rte_keepalive *rte_global_keepalive_info;
148 /* Print out statistics on packets dropped */
150 print_stats(__attribute__((unused)) struct rte_timer *ptr_timer,
151 __attribute__((unused)) void *ptr_data)
153 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
156 total_packets_dropped = 0;
157 total_packets_tx = 0;
158 total_packets_rx = 0;
160 const char clr[] = { 27, '[', '2', 'J', '\0' };
161 const char topLeft[] = { 27, '[', '1', ';', '1', 'H', '\0' };
163 /* Clear screen and move to top left */
164 printf("%s%s", clr, topLeft);
166 printf("\nPort statistics ====================================");
168 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
169 /* skip disabled ports */
170 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
172 printf("\nStatistics for port %u ------------------------------"
173 "\nPackets sent: %24"PRIu64
174 "\nPackets received: %20"PRIu64
175 "\nPackets dropped: %21"PRIu64,
177 port_statistics[portid].tx,
178 port_statistics[portid].rx,
179 port_statistics[portid].dropped);
181 total_packets_dropped += port_statistics[portid].dropped;
182 total_packets_tx += port_statistics[portid].tx;
183 total_packets_rx += port_statistics[portid].rx;
185 printf("\nAggregate statistics ==============================="
186 "\nTotal packets sent: %18"PRIu64
187 "\nTotal packets received: %14"PRIu64
188 "\nTotal packets dropped: %15"PRIu64,
191 total_packets_dropped);
192 printf("\n====================================================\n");
195 /* Send the burst of packets on an output interface */
197 l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
199 struct rte_mbuf **m_table;
201 unsigned queueid = 0;
203 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
205 ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
206 port_statistics[port].tx += ret;
207 if (unlikely(ret < n)) {
208 port_statistics[port].dropped += (n - ret);
210 rte_pktmbuf_free(m_table[ret]);
217 /* Enqueue packets for TX and prepare them to be sent */
219 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
221 unsigned lcore_id, len;
222 struct lcore_queue_conf *qconf;
224 lcore_id = rte_lcore_id();
226 qconf = &lcore_queue_conf[lcore_id];
227 len = qconf->tx_mbufs[port].len;
228 qconf->tx_mbufs[port].m_table[len] = m;
231 /* enough pkts to be sent */
232 if (unlikely(len == MAX_PKT_BURST)) {
233 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
237 qconf->tx_mbufs[port].len = len;
242 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
244 struct ether_hdr *eth;
248 dst_port = l2fwd_dst_ports[portid];
249 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
251 /* 02:00:00:00:00:xx */
252 tmp = ð->d_addr.addr_bytes[0];
253 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
256 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
258 l2fwd_send_packet(m, (uint8_t) dst_port);
261 /* main processing loop */
263 l2fwd_main_loop(void)
265 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
268 uint64_t prev_tsc, diff_tsc, cur_tsc;
269 unsigned i, j, portid, nb_rx;
270 struct lcore_queue_conf *qconf;
271 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
272 / US_PER_S * BURST_TX_DRAIN_US;
276 lcore_id = rte_lcore_id();
277 qconf = &lcore_queue_conf[lcore_id];
279 if (qconf->n_rx_port == 0) {
280 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
284 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
286 for (i = 0; i < qconf->n_rx_port; i++) {
288 portid = qconf->rx_port_list[i];
289 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
293 uint64_t tsc_initial = rte_rdtsc();
294 uint64_t tsc_lifetime = (rand()&0x07) * rte_get_tsc_hz();
297 /* Keepalive heartbeat */
298 rte_keepalive_mark_alive(rte_global_keepalive_info);
300 cur_tsc = rte_rdtsc();
303 * Die randomly within 7 secs for demo purposes if
306 if (check_period > 0 && cur_tsc - tsc_initial > tsc_lifetime)
310 * TX burst queue drain
312 diff_tsc = cur_tsc - prev_tsc;
313 if (unlikely(diff_tsc > drain_tsc)) {
315 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
316 if (qconf->tx_mbufs[portid].len == 0)
318 l2fwd_send_burst(&lcore_queue_conf[lcore_id],
319 qconf->tx_mbufs[portid].len,
321 qconf->tx_mbufs[portid].len = 0;
328 * Read packet from RX queues
330 for (i = 0; i < qconf->n_rx_port; i++) {
332 portid = qconf->rx_port_list[i];
333 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
334 pkts_burst, MAX_PKT_BURST);
336 port_statistics[portid].rx += nb_rx;
338 for (j = 0; j < nb_rx; j++) {
340 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
341 l2fwd_simple_forward(m, portid);
348 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
356 l2fwd_usage(const char *prgname)
358 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
359 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
360 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
361 " -K PERIOD: Keepalive check period (5 default; 86400 max)\n"
362 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
367 l2fwd_parse_portmask(const char *portmask)
372 /* parse hexadecimal string */
373 pm = strtoul(portmask, &end, 16);
374 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
384 l2fwd_parse_nqueue(const char *q_arg)
389 /* parse hexadecimal string */
390 n = strtoul(q_arg, &end, 10);
391 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
395 if (n >= MAX_RX_QUEUE_PER_LCORE)
402 l2fwd_parse_timer_period(const char *q_arg)
407 /* parse number string */
408 n = strtol(q_arg, &end, 10);
409 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
411 if (n >= MAX_TIMER_PERIOD)
418 l2fwd_parse_check_period(const char *q_arg)
423 /* parse number string */
424 n = strtol(q_arg, &end, 10);
425 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
427 if (n >= MAX_TIMER_PERIOD)
433 /* Parse the argument given in the command line of the application */
435 l2fwd_parse_args(int argc, char **argv)
440 char *prgname = argv[0];
441 static struct option lgopts[] = {
447 while ((opt = getopt_long(argc, argvopt, "p:q:T:K:",
448 lgopts, &option_index)) != EOF) {
453 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
454 if (l2fwd_enabled_port_mask == 0) {
455 printf("invalid portmask\n");
456 l2fwd_usage(prgname);
463 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
464 if (l2fwd_rx_queue_per_lcore == 0) {
465 printf("invalid queue number\n");
466 l2fwd_usage(prgname);
473 timer_period = l2fwd_parse_timer_period(optarg)
474 * (int64_t)(1000 * TIMER_MILLISECOND);
475 if (timer_period < 0) {
476 printf("invalid timer period\n");
477 l2fwd_usage(prgname);
484 check_period = l2fwd_parse_check_period(optarg);
485 if (check_period < 0) {
486 printf("invalid check period\n");
487 l2fwd_usage(prgname);
494 l2fwd_usage(prgname);
498 l2fwd_usage(prgname);
504 argv[optind-1] = prgname;
507 optind = 0; /* reset getopt lib */
511 /* Check the link status of all ports in up to 9s, and print them finally */
513 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
515 #define CHECK_INTERVAL 100 /* 100ms */
516 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
517 uint8_t portid, count, all_ports_up, print_flag = 0;
518 struct rte_eth_link link;
520 printf("\nChecking link status");
522 for (count = 0; count <= MAX_CHECK_TIME; count++) {
524 for (portid = 0; portid < port_num; portid++) {
525 if ((port_mask & (1 << portid)) == 0)
527 memset(&link, 0, sizeof(link));
528 rte_eth_link_get_nowait(portid, &link);
529 /* print link status if flag set */
530 if (print_flag == 1) {
531 if (link.link_status)
532 printf("Port %d Link Up - speed %u "
533 "Mbps - %s\n", (uint8_t)portid,
534 (unsigned)link.link_speed,
535 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
536 ("full-duplex") : ("half-duplex\n"));
538 printf("Port %d Link Down\n",
542 /* clear all_ports_up flag if any link down */
543 if (link.link_status == 0) {
548 /* after finally printing all link status, get out */
552 if (all_ports_up == 0) {
555 rte_delay_ms(CHECK_INTERVAL);
558 /* set the print_flag if all ports up or timeout */
559 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
567 dead_core(__attribute__((unused)) void *ptr_data, const int id_core)
569 printf("Dead core %i - restarting..\n", id_core);
570 if (rte_eal_get_lcore_state(id_core) == FINISHED) {
571 rte_eal_wait_lcore(id_core);
572 rte_eal_remote_launch(l2fwd_launch_one_lcore, NULL, id_core);
574 printf("..false positive!\n");
579 main(int argc, char **argv)
581 struct lcore_queue_conf *qconf;
582 struct rte_eth_dev_info dev_info;
585 uint8_t nb_ports_available;
586 uint8_t portid, last_port;
587 unsigned lcore_id, rx_lcore_id;
588 unsigned nb_ports_in_mask = 0;
591 ret = rte_eal_init(argc, argv);
593 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
597 l2fwd_enabled_port_mask = 0;
599 /* parse application arguments (after the EAL ones) */
600 ret = l2fwd_parse_args(argc, argv);
602 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
604 /* create the mbuf pool */
605 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF, 32,
606 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
607 if (l2fwd_pktmbuf_pool == NULL)
608 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
610 nb_ports = rte_eth_dev_count();
612 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
614 if (nb_ports > RTE_MAX_ETHPORTS)
615 nb_ports = RTE_MAX_ETHPORTS;
617 /* reset l2fwd_dst_ports */
618 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
619 l2fwd_dst_ports[portid] = 0;
623 * Each logical core is assigned a dedicated TX queue on each port.
625 for (portid = 0; portid < nb_ports; portid++) {
626 /* skip ports that are not enabled */
627 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
630 if (nb_ports_in_mask % 2) {
631 l2fwd_dst_ports[portid] = last_port;
632 l2fwd_dst_ports[last_port] = portid;
638 rte_eth_dev_info_get(portid, &dev_info);
640 if (nb_ports_in_mask % 2) {
641 printf("Notice: odd number of ports in portmask.\n");
642 l2fwd_dst_ports[last_port] = last_port;
648 /* Initialize the port/queue configuration of each logical core */
649 for (portid = 0; portid < nb_ports; portid++) {
650 /* skip ports that are not enabled */
651 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
654 /* get the lcore_id for this port */
655 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
656 lcore_queue_conf[rx_lcore_id].n_rx_port ==
657 l2fwd_rx_queue_per_lcore) {
659 if (rx_lcore_id >= RTE_MAX_LCORE)
660 rte_exit(EXIT_FAILURE, "Not enough cores\n");
663 if (qconf != &lcore_queue_conf[rx_lcore_id])
664 /* Assigned a new logical core in the loop above. */
665 qconf = &lcore_queue_conf[rx_lcore_id];
667 qconf->rx_port_list[qconf->n_rx_port] = portid;
669 printf("Lcore %u: RX port %u\n",
670 rx_lcore_id, (unsigned) portid);
673 nb_ports_available = nb_ports;
675 /* Initialise each port */
676 for (portid = 0; portid < nb_ports; portid++) {
677 /* skip ports that are not enabled */
678 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
679 printf("Skipping disabled port %u\n",
681 nb_ports_available--;
685 printf("Initializing port %u... ", (unsigned) portid);
687 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
689 rte_exit(EXIT_FAILURE,
690 "Cannot configure device: err=%d, port=%u\n",
691 ret, (unsigned) portid);
693 rte_eth_macaddr_get(portid, &l2fwd_ports_eth_addr[portid]);
695 /* init one RX queue */
697 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
698 rte_eth_dev_socket_id(portid),
702 rte_exit(EXIT_FAILURE,
703 "rte_eth_rx_queue_setup:err=%d, port=%u\n",
704 ret, (unsigned) portid);
706 /* init one TX queue on each port */
708 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
709 rte_eth_dev_socket_id(portid),
712 rte_exit(EXIT_FAILURE,
713 "rte_eth_tx_queue_setup:err=%d, port=%u\n",
714 ret, (unsigned) portid);
717 ret = rte_eth_dev_start(portid);
719 rte_exit(EXIT_FAILURE,
720 "rte_eth_dev_start:err=%d, port=%u\n",
721 ret, (unsigned) portid);
723 rte_eth_promiscuous_enable(portid);
725 printf("Port %u, MAC address: "
726 "%02X:%02X:%02X:%02X:%02X:%02X\n\n",
728 l2fwd_ports_eth_addr[portid].addr_bytes[0],
729 l2fwd_ports_eth_addr[portid].addr_bytes[1],
730 l2fwd_ports_eth_addr[portid].addr_bytes[2],
731 l2fwd_ports_eth_addr[portid].addr_bytes[3],
732 l2fwd_ports_eth_addr[portid].addr_bytes[4],
733 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
735 /* initialize port stats */
736 memset(&port_statistics, 0, sizeof(port_statistics));
739 if (!nb_ports_available) {
740 rte_exit(EXIT_FAILURE,
741 "All available ports are disabled. Please set portmask.\n");
744 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
746 struct rte_timer hb_timer, stats_timer;
748 rte_timer_subsystem_init();
749 rte_timer_init(&stats_timer);
751 if (check_period > 0) {
752 rte_global_keepalive_info =
753 rte_keepalive_create(&dead_core, NULL);
754 if (rte_global_keepalive_info == NULL)
755 rte_exit(EXIT_FAILURE, "init_keep_alive() failed");
756 rte_timer_init(&hb_timer);
757 if (rte_timer_reset(&hb_timer,
758 (check_period * rte_get_timer_hz()) / 1000,
761 (void(*)(struct rte_timer*, void*))
762 &rte_keepalive_dispatch_pings,
763 rte_global_keepalive_info
765 rte_exit(EXIT_FAILURE, "Keepalive setup failure.\n");
767 if (timer_period > 0) {
768 if (rte_timer_reset(&stats_timer,
769 (timer_period * rte_get_timer_hz()) / 1000,
774 rte_exit(EXIT_FAILURE, "Stats setup failure.\n");
776 /* launch per-lcore init on every slave lcore */
777 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
778 struct lcore_queue_conf *qconf = &lcore_queue_conf[lcore_id];
780 if (qconf->n_rx_port == 0)
782 "lcore %u has nothing to do\n",
786 rte_eal_remote_launch(
787 l2fwd_launch_one_lcore,
791 rte_keepalive_register_core(rte_global_keepalive_info,
800 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
801 if (rte_eal_wait_lcore(lcore_id) < 0)