1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 2010-2016 Intel Corporation
10 #include <sys/types.h>
11 #include <sys/queue.h>
12 #include <netinet/in.h>
20 #include <rte_common.h>
22 #include <rte_malloc.h>
23 #include <rte_memory.h>
24 #include <rte_memcpy.h>
26 #include <rte_launch.h>
27 #include <rte_atomic.h>
28 #include <rte_cycles.h>
29 #include <rte_prefetch.h>
30 #include <rte_lcore.h>
31 #include <rte_per_lcore.h>
32 #include <rte_branch_prediction.h>
33 #include <rte_interrupts.h>
34 #include <rte_random.h>
35 #include <rte_debug.h>
36 #include <rte_ether.h>
37 #include <rte_ethdev.h>
38 #include <rte_mempool.h>
40 #include <rte_timer.h>
41 #include <rte_keepalive.h>
45 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
49 #define MAX_PKT_BURST 32
50 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
53 * Configurable number of RX/TX ring descriptors
55 #define RTE_TEST_RX_DESC_DEFAULT 128
56 #define RTE_TEST_TX_DESC_DEFAULT 512
57 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
58 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
60 /* ethernet addresses of ports */
61 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
63 /* mask of enabled ports */
64 static uint32_t l2fwd_enabled_port_mask;
66 /* list of enabled ports */
67 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
69 static unsigned int l2fwd_rx_queue_per_lcore = 1;
71 #define MAX_RX_QUEUE_PER_LCORE 16
72 #define MAX_TX_QUEUE_PER_PORT 16
73 struct lcore_queue_conf {
75 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
76 } __rte_cache_aligned;
77 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
79 struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
81 static const struct rte_eth_conf port_conf = {
84 .header_split = 0, /**< Header Split disabled */
85 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
86 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
87 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
88 .hw_strip_crc = 1, /**< CRC stripped by hardware */
91 .mq_mode = ETH_MQ_TX_NONE,
95 struct rte_mempool *l2fwd_pktmbuf_pool = NULL;
97 /* Per-port statistics struct */
98 struct l2fwd_port_statistics {
102 } __rte_cache_aligned;
103 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
105 /* A tsc-based timer responsible for triggering statistics printout */
106 #define TIMER_MILLISECOND 1
107 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
108 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* 10 seconds */
109 static int64_t check_period = 5; /* default check cycle is 5ms */
111 /* Keepalive structure */
112 struct rte_keepalive *rte_global_keepalive_info;
114 /* Termination signalling */
115 static int terminate_signal_received;
117 /* Termination signal handler */
118 static void handle_sigterm(__rte_unused int value)
120 terminate_signal_received = 1;
123 /* Print out statistics on packets dropped */
125 print_stats(__attribute__((unused)) struct rte_timer *ptr_timer,
126 __attribute__((unused)) void *ptr_data)
128 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
131 total_packets_dropped = 0;
132 total_packets_tx = 0;
133 total_packets_rx = 0;
135 const char clr[] = { 27, '[', '2', 'J', '\0' };
136 const char topLeft[] = { 27, '[', '1', ';', '1', 'H', '\0' };
138 /* Clear screen and move to top left */
139 printf("%s%s", clr, topLeft);
141 printf("\nPort statistics ====================================");
143 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
144 /* skip disabled ports */
145 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
147 printf("\nStatistics for port %u ------------------------------"
148 "\nPackets sent: %24"PRIu64
149 "\nPackets received: %20"PRIu64
150 "\nPackets dropped: %21"PRIu64,
152 port_statistics[portid].tx,
153 port_statistics[portid].rx,
154 port_statistics[portid].dropped);
156 total_packets_dropped += port_statistics[portid].dropped;
157 total_packets_tx += port_statistics[portid].tx;
158 total_packets_rx += port_statistics[portid].rx;
160 printf("\nAggregate statistics ==============================="
161 "\nTotal packets sent: %18"PRIu64
162 "\nTotal packets received: %14"PRIu64
163 "\nTotal packets dropped: %15"PRIu64,
166 total_packets_dropped);
167 printf("\n====================================================\n");
171 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
173 struct ether_hdr *eth;
177 struct rte_eth_dev_tx_buffer *buffer;
179 dst_port = l2fwd_dst_ports[portid];
180 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
182 /* 02:00:00:00:00:xx */
183 tmp = ð->d_addr.addr_bytes[0];
184 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
187 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
189 buffer = tx_buffer[dst_port];
190 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
192 port_statistics[dst_port].tx += sent;
195 /* main processing loop */
197 l2fwd_main_loop(void)
199 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
203 uint64_t prev_tsc, diff_tsc, cur_tsc;
204 unsigned i, j, portid, nb_rx;
205 struct lcore_queue_conf *qconf;
206 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
207 / US_PER_S * BURST_TX_DRAIN_US;
208 struct rte_eth_dev_tx_buffer *buffer;
212 lcore_id = rte_lcore_id();
213 qconf = &lcore_queue_conf[lcore_id];
215 if (qconf->n_rx_port == 0) {
216 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
220 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
222 for (i = 0; i < qconf->n_rx_port; i++) {
224 portid = qconf->rx_port_list[i];
225 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
229 uint64_t tsc_initial = rte_rdtsc();
230 uint64_t tsc_lifetime = (rand()&0x07) * rte_get_tsc_hz();
232 while (!terminate_signal_received) {
233 /* Keepalive heartbeat */
234 rte_keepalive_mark_alive(rte_global_keepalive_info);
236 cur_tsc = rte_rdtsc();
239 * Die randomly within 7 secs for demo purposes if
242 if (check_period > 0 && cur_tsc - tsc_initial > tsc_lifetime)
246 * TX burst queue drain
248 diff_tsc = cur_tsc - prev_tsc;
249 if (unlikely(diff_tsc > drain_tsc)) {
251 for (i = 0; i < qconf->n_rx_port; i++) {
253 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
254 buffer = tx_buffer[portid];
256 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
258 port_statistics[portid].tx += sent;
266 * Read packet from RX queues
268 for (i = 0; i < qconf->n_rx_port; i++) {
270 portid = qconf->rx_port_list[i];
271 nb_rx = rte_eth_rx_burst(portid, 0,
272 pkts_burst, MAX_PKT_BURST);
274 port_statistics[portid].rx += nb_rx;
276 for (j = 0; j < nb_rx; j++) {
278 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
279 l2fwd_simple_forward(m, portid);
286 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
294 l2fwd_usage(const char *prgname)
296 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
297 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
298 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
299 " -K PERIOD: Keepalive check period (5 default; 86400 max)\n"
300 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
305 l2fwd_parse_portmask(const char *portmask)
310 /* parse hexadecimal string */
311 pm = strtoul(portmask, &end, 16);
312 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
322 l2fwd_parse_nqueue(const char *q_arg)
327 /* parse hexadecimal string */
328 n = strtoul(q_arg, &end, 10);
329 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
333 if (n >= MAX_RX_QUEUE_PER_LCORE)
340 l2fwd_parse_timer_period(const char *q_arg)
345 /* parse number string */
346 n = strtol(q_arg, &end, 10);
347 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
349 if (n >= MAX_TIMER_PERIOD)
356 l2fwd_parse_check_period(const char *q_arg)
361 /* parse number string */
362 n = strtol(q_arg, &end, 10);
363 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
365 if (n >= MAX_TIMER_PERIOD)
371 /* Parse the argument given in the command line of the application */
373 l2fwd_parse_args(int argc, char **argv)
378 char *prgname = argv[0];
379 static struct option lgopts[] = {
385 while ((opt = getopt_long(argc, argvopt, "p:q:T:K:",
386 lgopts, &option_index)) != EOF) {
391 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
392 if (l2fwd_enabled_port_mask == 0) {
393 printf("invalid portmask\n");
394 l2fwd_usage(prgname);
401 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
402 if (l2fwd_rx_queue_per_lcore == 0) {
403 printf("invalid queue number\n");
404 l2fwd_usage(prgname);
411 timer_period = l2fwd_parse_timer_period(optarg)
412 * (int64_t)(1000 * TIMER_MILLISECOND);
413 if (timer_period < 0) {
414 printf("invalid timer period\n");
415 l2fwd_usage(prgname);
422 check_period = l2fwd_parse_check_period(optarg);
423 if (check_period < 0) {
424 printf("invalid check period\n");
425 l2fwd_usage(prgname);
432 l2fwd_usage(prgname);
436 l2fwd_usage(prgname);
442 argv[optind-1] = prgname;
445 optind = 1; /* reset getopt lib */
449 /* Check the link status of all ports in up to 9s, and print them finally */
451 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
453 #define CHECK_INTERVAL 100 /* 100ms */
454 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
456 uint8_t count, all_ports_up, print_flag = 0;
457 struct rte_eth_link link;
459 printf("\nChecking link status");
461 for (count = 0; count <= MAX_CHECK_TIME; count++) {
463 for (portid = 0; portid < port_num; portid++) {
464 if ((port_mask & (1 << portid)) == 0)
466 memset(&link, 0, sizeof(link));
467 rte_eth_link_get_nowait(portid, &link);
468 /* print link status if flag set */
469 if (print_flag == 1) {
470 if (link.link_status)
472 "Port%d Link Up. Speed %u Mbps - %s\n",
473 portid, link.link_speed,
474 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
475 ("full-duplex") : ("half-duplex\n"));
477 printf("Port %d Link Down\n", portid);
480 /* clear all_ports_up flag if any link down */
481 if (link.link_status == ETH_LINK_DOWN) {
486 /* after finally printing all link status, get out */
490 if (all_ports_up == 0) {
493 rte_delay_ms(CHECK_INTERVAL);
496 /* set the print_flag if all ports up or timeout */
497 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
505 dead_core(__rte_unused void *ptr_data, const int id_core)
507 if (terminate_signal_received)
509 printf("Dead core %i - restarting..\n", id_core);
510 if (rte_eal_get_lcore_state(id_core) == FINISHED) {
511 rte_eal_wait_lcore(id_core);
512 rte_eal_remote_launch(l2fwd_launch_one_lcore, NULL, id_core);
514 printf("..false positive!\n");
519 relay_core_state(void *ptr_data, const int id_core,
520 const enum rte_keepalive_state core_state, uint64_t last_alive)
522 rte_keepalive_relayed_state((struct rte_keepalive_shm *)ptr_data,
523 id_core, core_state, last_alive);
527 main(int argc, char **argv)
529 struct lcore_queue_conf *qconf;
532 uint16_t nb_ports_available;
533 uint16_t portid, last_port;
534 unsigned lcore_id, rx_lcore_id;
535 unsigned nb_ports_in_mask = 0;
536 struct sigaction signal_handler;
537 struct rte_keepalive_shm *ka_shm;
539 memset(&signal_handler, 0, sizeof(signal_handler));
540 terminate_signal_received = 0;
541 signal_handler.sa_handler = &handle_sigterm;
542 if (sigaction(SIGINT, &signal_handler, NULL) == -1 ||
543 sigaction(SIGTERM, &signal_handler, NULL) == -1)
544 rte_exit(EXIT_FAILURE, "SIGNAL\n");
548 ret = rte_eal_init(argc, argv);
550 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
554 l2fwd_enabled_port_mask = 0;
556 /* parse application arguments (after the EAL ones) */
557 ret = l2fwd_parse_args(argc, argv);
559 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
561 /* create the mbuf pool */
562 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF, 32,
563 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
564 if (l2fwd_pktmbuf_pool == NULL)
565 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
567 nb_ports = rte_eth_dev_count();
569 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
571 /* reset l2fwd_dst_ports */
572 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
573 l2fwd_dst_ports[portid] = 0;
577 * Each logical core is assigned a dedicated TX queue on each port.
579 for (portid = 0; portid < nb_ports; portid++) {
580 /* skip ports that are not enabled */
581 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
584 if (nb_ports_in_mask % 2) {
585 l2fwd_dst_ports[portid] = last_port;
586 l2fwd_dst_ports[last_port] = portid;
592 if (nb_ports_in_mask % 2) {
593 printf("Notice: odd number of ports in portmask.\n");
594 l2fwd_dst_ports[last_port] = last_port;
600 /* Initialize the port/queue configuration of each logical core */
601 for (portid = 0; portid < nb_ports; portid++) {
602 /* skip ports that are not enabled */
603 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
606 /* get the lcore_id for this port */
607 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
608 lcore_queue_conf[rx_lcore_id].n_rx_port ==
609 l2fwd_rx_queue_per_lcore) {
611 if (rx_lcore_id >= RTE_MAX_LCORE)
612 rte_exit(EXIT_FAILURE, "Not enough cores\n");
615 if (qconf != &lcore_queue_conf[rx_lcore_id])
616 /* Assigned a new logical core in the loop above. */
617 qconf = &lcore_queue_conf[rx_lcore_id];
619 qconf->rx_port_list[qconf->n_rx_port] = portid;
621 printf("Lcore %u: RX port %u\n",
622 rx_lcore_id, portid);
625 nb_ports_available = nb_ports;
627 /* Initialise each port */
628 for (portid = 0; portid < nb_ports; portid++) {
629 /* skip ports that are not enabled */
630 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
631 printf("Skipping disabled port %u\n", portid);
632 nb_ports_available--;
636 printf("Initializing port %u... ", portid);
638 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
640 rte_exit(EXIT_FAILURE,
641 "Cannot configure device: err=%d, port=%u\n",
644 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
647 rte_exit(EXIT_FAILURE,
648 "Cannot adjust number of descriptors: err=%d, port=%u\n",
651 rte_eth_macaddr_get(portid, &l2fwd_ports_eth_addr[portid]);
653 /* init one RX queue */
655 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
656 rte_eth_dev_socket_id(portid),
660 rte_exit(EXIT_FAILURE,
661 "rte_eth_rx_queue_setup:err=%d, port=%u\n",
664 /* init one TX queue on each port */
666 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
667 rte_eth_dev_socket_id(portid),
670 rte_exit(EXIT_FAILURE,
671 "rte_eth_tx_queue_setup:err=%d, port=%u\n",
674 /* Initialize TX buffers */
675 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
676 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
677 rte_eth_dev_socket_id(portid));
678 if (tx_buffer[portid] == NULL)
679 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
682 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
684 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
685 rte_eth_tx_buffer_count_callback,
686 &port_statistics[portid].dropped);
688 rte_exit(EXIT_FAILURE,
689 "Cannot set error callback for tx buffer on port %u\n",
693 ret = rte_eth_dev_start(portid);
695 rte_exit(EXIT_FAILURE,
696 "rte_eth_dev_start:err=%d, port=%u\n",
699 rte_eth_promiscuous_enable(portid);
701 printf("Port %u, MAC address: "
702 "%02X:%02X:%02X:%02X:%02X:%02X\n\n",
704 l2fwd_ports_eth_addr[portid].addr_bytes[0],
705 l2fwd_ports_eth_addr[portid].addr_bytes[1],
706 l2fwd_ports_eth_addr[portid].addr_bytes[2],
707 l2fwd_ports_eth_addr[portid].addr_bytes[3],
708 l2fwd_ports_eth_addr[portid].addr_bytes[4],
709 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
711 /* initialize port stats */
712 memset(&port_statistics, 0, sizeof(port_statistics));
715 if (!nb_ports_available) {
716 rte_exit(EXIT_FAILURE,
717 "All available ports are disabled. Please set portmask.\n");
720 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
722 struct rte_timer hb_timer, stats_timer;
724 rte_timer_subsystem_init();
725 rte_timer_init(&stats_timer);
728 if (check_period > 0) {
729 ka_shm = rte_keepalive_shm_create();
731 rte_exit(EXIT_FAILURE,
732 "rte_keepalive_shm_create() failed");
733 rte_global_keepalive_info =
734 rte_keepalive_create(&dead_core, ka_shm);
735 if (rte_global_keepalive_info == NULL)
736 rte_exit(EXIT_FAILURE, "init_keep_alive() failed");
737 rte_keepalive_register_relay_callback(rte_global_keepalive_info,
738 relay_core_state, ka_shm);
739 rte_timer_init(&hb_timer);
740 if (rte_timer_reset(&hb_timer,
741 (check_period * rte_get_timer_hz()) / 1000,
744 (void(*)(struct rte_timer*, void*))
745 &rte_keepalive_dispatch_pings,
746 rte_global_keepalive_info
748 rte_exit(EXIT_FAILURE, "Keepalive setup failure.\n");
750 if (timer_period > 0) {
751 if (rte_timer_reset(&stats_timer,
752 (timer_period * rte_get_timer_hz()) / 1000,
757 rte_exit(EXIT_FAILURE, "Stats setup failure.\n");
759 /* launch per-lcore init on every slave lcore */
760 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
761 struct lcore_queue_conf *qconf = &lcore_queue_conf[lcore_id];
763 if (qconf->n_rx_port == 0)
765 "lcore %u has nothing to do\n",
769 rte_eal_remote_launch(
770 l2fwd_launch_one_lcore,
774 rte_keepalive_register_core(rte_global_keepalive_info,
778 while (!terminate_signal_received) {
783 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
784 if (rte_eal_wait_lcore(lcore_id) < 0)
789 rte_keepalive_shm_cleanup(ka_shm);