4 * Copyright(c) 2010-2016 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>
49 #include <rte_common.h>
51 #include <rte_malloc.h>
52 #include <rte_memory.h>
53 #include <rte_memcpy.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>
68 #include <rte_mempool.h>
70 #include <rte_timer.h>
71 #include <rte_keepalive.h>
75 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
79 #define MAX_PKT_BURST 32
80 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
83 * Configurable number of RX/TX ring descriptors
85 #define RTE_TEST_RX_DESC_DEFAULT 128
86 #define RTE_TEST_TX_DESC_DEFAULT 512
87 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
88 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
90 /* ethernet addresses of ports */
91 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
93 /* mask of enabled ports */
94 static uint32_t l2fwd_enabled_port_mask;
96 /* list of enabled ports */
97 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
99 static unsigned int l2fwd_rx_queue_per_lcore = 1;
101 #define MAX_RX_QUEUE_PER_LCORE 16
102 #define MAX_TX_QUEUE_PER_PORT 16
103 struct lcore_queue_conf {
105 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
106 } __rte_cache_aligned;
107 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
109 struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
111 static const struct rte_eth_conf port_conf = {
114 .header_split = 0, /**< Header Split disabled */
115 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
116 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
117 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
118 .hw_strip_crc = 1, /**< CRC stripped by hardware */
121 .mq_mode = ETH_MQ_TX_NONE,
125 struct rte_mempool *l2fwd_pktmbuf_pool = NULL;
127 /* Per-port statistics struct */
128 struct l2fwd_port_statistics {
132 } __rte_cache_aligned;
133 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
135 /* A tsc-based timer responsible for triggering statistics printout */
136 #define TIMER_MILLISECOND 1
137 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
138 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* 10 seconds */
139 static int64_t check_period = 5; /* default check cycle is 5ms */
141 /* Keepalive structure */
142 struct rte_keepalive *rte_global_keepalive_info;
144 /* Termination signalling */
145 static int terminate_signal_received;
147 /* Termination signal handler */
148 static void handle_sigterm(__rte_unused int value)
150 terminate_signal_received = 1;
153 /* Print out statistics on packets dropped */
155 print_stats(__attribute__((unused)) struct rte_timer *ptr_timer,
156 __attribute__((unused)) void *ptr_data)
158 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
161 total_packets_dropped = 0;
162 total_packets_tx = 0;
163 total_packets_rx = 0;
165 const char clr[] = { 27, '[', '2', 'J', '\0' };
166 const char topLeft[] = { 27, '[', '1', ';', '1', 'H', '\0' };
168 /* Clear screen and move to top left */
169 printf("%s%s", clr, topLeft);
171 printf("\nPort statistics ====================================");
173 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
174 /* skip disabled ports */
175 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
177 printf("\nStatistics for port %u ------------------------------"
178 "\nPackets sent: %24"PRIu64
179 "\nPackets received: %20"PRIu64
180 "\nPackets dropped: %21"PRIu64,
182 port_statistics[portid].tx,
183 port_statistics[portid].rx,
184 port_statistics[portid].dropped);
186 total_packets_dropped += port_statistics[portid].dropped;
187 total_packets_tx += port_statistics[portid].tx;
188 total_packets_rx += port_statistics[portid].rx;
190 printf("\nAggregate statistics ==============================="
191 "\nTotal packets sent: %18"PRIu64
192 "\nTotal packets received: %14"PRIu64
193 "\nTotal packets dropped: %15"PRIu64,
196 total_packets_dropped);
197 printf("\n====================================================\n");
201 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
203 struct ether_hdr *eth;
207 struct rte_eth_dev_tx_buffer *buffer;
209 dst_port = l2fwd_dst_ports[portid];
210 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
212 /* 02:00:00:00:00:xx */
213 tmp = ð->d_addr.addr_bytes[0];
214 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
217 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
219 buffer = tx_buffer[dst_port];
220 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
222 port_statistics[dst_port].tx += sent;
225 /* main processing loop */
227 l2fwd_main_loop(void)
229 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
233 uint64_t prev_tsc, diff_tsc, cur_tsc;
234 unsigned i, j, portid, nb_rx;
235 struct lcore_queue_conf *qconf;
236 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
237 / US_PER_S * BURST_TX_DRAIN_US;
238 struct rte_eth_dev_tx_buffer *buffer;
242 lcore_id = rte_lcore_id();
243 qconf = &lcore_queue_conf[lcore_id];
245 if (qconf->n_rx_port == 0) {
246 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
250 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
252 for (i = 0; i < qconf->n_rx_port; i++) {
254 portid = qconf->rx_port_list[i];
255 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
259 uint64_t tsc_initial = rte_rdtsc();
260 uint64_t tsc_lifetime = (rand()&0x07) * rte_get_tsc_hz();
262 while (!terminate_signal_received) {
263 /* Keepalive heartbeat */
264 rte_keepalive_mark_alive(rte_global_keepalive_info);
266 cur_tsc = rte_rdtsc();
269 * Die randomly within 7 secs for demo purposes if
272 if (check_period > 0 && cur_tsc - tsc_initial > tsc_lifetime)
276 * TX burst queue drain
278 diff_tsc = cur_tsc - prev_tsc;
279 if (unlikely(diff_tsc > drain_tsc)) {
281 for (i = 0; i < qconf->n_rx_port; i++) {
283 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
284 buffer = tx_buffer[portid];
286 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
288 port_statistics[portid].tx += sent;
296 * Read packet from RX queues
298 for (i = 0; i < qconf->n_rx_port; i++) {
300 portid = qconf->rx_port_list[i];
301 nb_rx = rte_eth_rx_burst(portid, 0,
302 pkts_burst, MAX_PKT_BURST);
304 port_statistics[portid].rx += nb_rx;
306 for (j = 0; j < nb_rx; j++) {
308 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
309 l2fwd_simple_forward(m, portid);
316 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
324 l2fwd_usage(const char *prgname)
326 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
327 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
328 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
329 " -K PERIOD: Keepalive check period (5 default; 86400 max)\n"
330 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
335 l2fwd_parse_portmask(const char *portmask)
340 /* parse hexadecimal string */
341 pm = strtoul(portmask, &end, 16);
342 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
352 l2fwd_parse_nqueue(const char *q_arg)
357 /* parse hexadecimal string */
358 n = strtoul(q_arg, &end, 10);
359 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
363 if (n >= MAX_RX_QUEUE_PER_LCORE)
370 l2fwd_parse_timer_period(const char *q_arg)
375 /* parse number string */
376 n = strtol(q_arg, &end, 10);
377 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
379 if (n >= MAX_TIMER_PERIOD)
386 l2fwd_parse_check_period(const char *q_arg)
391 /* parse number string */
392 n = strtol(q_arg, &end, 10);
393 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
395 if (n >= MAX_TIMER_PERIOD)
401 /* Parse the argument given in the command line of the application */
403 l2fwd_parse_args(int argc, char **argv)
408 char *prgname = argv[0];
409 static struct option lgopts[] = {
415 while ((opt = getopt_long(argc, argvopt, "p:q:T:K:",
416 lgopts, &option_index)) != EOF) {
421 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
422 if (l2fwd_enabled_port_mask == 0) {
423 printf("invalid portmask\n");
424 l2fwd_usage(prgname);
431 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
432 if (l2fwd_rx_queue_per_lcore == 0) {
433 printf("invalid queue number\n");
434 l2fwd_usage(prgname);
441 timer_period = l2fwd_parse_timer_period(optarg)
442 * (int64_t)(1000 * TIMER_MILLISECOND);
443 if (timer_period < 0) {
444 printf("invalid timer period\n");
445 l2fwd_usage(prgname);
452 check_period = l2fwd_parse_check_period(optarg);
453 if (check_period < 0) {
454 printf("invalid check period\n");
455 l2fwd_usage(prgname);
462 l2fwd_usage(prgname);
466 l2fwd_usage(prgname);
472 argv[optind-1] = prgname;
475 optind = 1; /* reset getopt lib */
479 /* Check the link status of all ports in up to 9s, and print them finally */
481 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
483 #define CHECK_INTERVAL 100 /* 100ms */
484 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
486 uint8_t count, all_ports_up, print_flag = 0;
487 struct rte_eth_link link;
489 printf("\nChecking link status");
491 for (count = 0; count <= MAX_CHECK_TIME; count++) {
493 for (portid = 0; portid < port_num; portid++) {
494 if ((port_mask & (1 << portid)) == 0)
496 memset(&link, 0, sizeof(link));
497 rte_eth_link_get_nowait(portid, &link);
498 /* print link status if flag set */
499 if (print_flag == 1) {
500 if (link.link_status)
502 "Port%d Link Up. Speed %u Mbps - %s\n",
503 portid, link.link_speed,
504 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
505 ("full-duplex") : ("half-duplex\n"));
507 printf("Port %d Link Down\n", portid);
510 /* clear all_ports_up flag if any link down */
511 if (link.link_status == ETH_LINK_DOWN) {
516 /* after finally printing all link status, get out */
520 if (all_ports_up == 0) {
523 rte_delay_ms(CHECK_INTERVAL);
526 /* set the print_flag if all ports up or timeout */
527 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
535 dead_core(__rte_unused void *ptr_data, const int id_core)
537 if (terminate_signal_received)
539 printf("Dead core %i - restarting..\n", id_core);
540 if (rte_eal_get_lcore_state(id_core) == FINISHED) {
541 rte_eal_wait_lcore(id_core);
542 rte_eal_remote_launch(l2fwd_launch_one_lcore, NULL, id_core);
544 printf("..false positive!\n");
549 relay_core_state(void *ptr_data, const int id_core,
550 const enum rte_keepalive_state core_state, uint64_t last_alive)
552 rte_keepalive_relayed_state((struct rte_keepalive_shm *)ptr_data,
553 id_core, core_state, last_alive);
557 main(int argc, char **argv)
559 struct lcore_queue_conf *qconf;
560 struct rte_eth_dev_info dev_info;
563 uint16_t nb_ports_available;
564 uint16_t portid, last_port;
565 unsigned lcore_id, rx_lcore_id;
566 unsigned nb_ports_in_mask = 0;
567 struct sigaction signal_handler;
568 struct rte_keepalive_shm *ka_shm;
570 memset(&signal_handler, 0, sizeof(signal_handler));
571 terminate_signal_received = 0;
572 signal_handler.sa_handler = &handle_sigterm;
573 if (sigaction(SIGINT, &signal_handler, NULL) == -1 ||
574 sigaction(SIGTERM, &signal_handler, NULL) == -1)
575 rte_exit(EXIT_FAILURE, "SIGNAL\n");
579 ret = rte_eal_init(argc, argv);
581 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
585 l2fwd_enabled_port_mask = 0;
587 /* parse application arguments (after the EAL ones) */
588 ret = l2fwd_parse_args(argc, argv);
590 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
592 /* create the mbuf pool */
593 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF, 32,
594 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
595 if (l2fwd_pktmbuf_pool == NULL)
596 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
598 nb_ports = rte_eth_dev_count();
600 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
602 /* reset l2fwd_dst_ports */
603 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
604 l2fwd_dst_ports[portid] = 0;
608 * Each logical core is assigned a dedicated TX queue on each port.
610 for (portid = 0; portid < nb_ports; portid++) {
611 /* skip ports that are not enabled */
612 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
615 if (nb_ports_in_mask % 2) {
616 l2fwd_dst_ports[portid] = last_port;
617 l2fwd_dst_ports[last_port] = portid;
623 rte_eth_dev_info_get(portid, &dev_info);
625 if (nb_ports_in_mask % 2) {
626 printf("Notice: odd number of ports in portmask.\n");
627 l2fwd_dst_ports[last_port] = last_port;
633 /* Initialize the port/queue configuration of each logical core */
634 for (portid = 0; portid < nb_ports; portid++) {
635 /* skip ports that are not enabled */
636 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
639 /* get the lcore_id for this port */
640 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
641 lcore_queue_conf[rx_lcore_id].n_rx_port ==
642 l2fwd_rx_queue_per_lcore) {
644 if (rx_lcore_id >= RTE_MAX_LCORE)
645 rte_exit(EXIT_FAILURE, "Not enough cores\n");
648 if (qconf != &lcore_queue_conf[rx_lcore_id])
649 /* Assigned a new logical core in the loop above. */
650 qconf = &lcore_queue_conf[rx_lcore_id];
652 qconf->rx_port_list[qconf->n_rx_port] = portid;
654 printf("Lcore %u: RX port %u\n",
655 rx_lcore_id, portid);
658 nb_ports_available = nb_ports;
660 /* Initialise each port */
661 for (portid = 0; portid < nb_ports; portid++) {
662 /* skip ports that are not enabled */
663 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
664 printf("Skipping disabled port %u\n", portid);
665 nb_ports_available--;
669 printf("Initializing port %u... ", portid);
671 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
673 rte_exit(EXIT_FAILURE,
674 "Cannot configure device: err=%d, port=%u\n",
677 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
680 rte_exit(EXIT_FAILURE,
681 "Cannot adjust number of descriptors: err=%d, port=%u\n",
684 rte_eth_macaddr_get(portid, &l2fwd_ports_eth_addr[portid]);
686 /* init one RX queue */
688 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
689 rte_eth_dev_socket_id(portid),
693 rte_exit(EXIT_FAILURE,
694 "rte_eth_rx_queue_setup:err=%d, port=%u\n",
697 /* init one TX queue on each port */
699 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
700 rte_eth_dev_socket_id(portid),
703 rte_exit(EXIT_FAILURE,
704 "rte_eth_tx_queue_setup:err=%d, port=%u\n",
707 /* Initialize TX buffers */
708 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
709 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
710 rte_eth_dev_socket_id(portid));
711 if (tx_buffer[portid] == NULL)
712 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
715 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
717 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
718 rte_eth_tx_buffer_count_callback,
719 &port_statistics[portid].dropped);
721 rte_exit(EXIT_FAILURE,
722 "Cannot set error callback for tx buffer on port %u\n",
726 ret = rte_eth_dev_start(portid);
728 rte_exit(EXIT_FAILURE,
729 "rte_eth_dev_start:err=%d, port=%u\n",
732 rte_eth_promiscuous_enable(portid);
734 printf("Port %u, MAC address: "
735 "%02X:%02X:%02X:%02X:%02X:%02X\n\n",
737 l2fwd_ports_eth_addr[portid].addr_bytes[0],
738 l2fwd_ports_eth_addr[portid].addr_bytes[1],
739 l2fwd_ports_eth_addr[portid].addr_bytes[2],
740 l2fwd_ports_eth_addr[portid].addr_bytes[3],
741 l2fwd_ports_eth_addr[portid].addr_bytes[4],
742 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
744 /* initialize port stats */
745 memset(&port_statistics, 0, sizeof(port_statistics));
748 if (!nb_ports_available) {
749 rte_exit(EXIT_FAILURE,
750 "All available ports are disabled. Please set portmask.\n");
753 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
755 struct rte_timer hb_timer, stats_timer;
757 rte_timer_subsystem_init();
758 rte_timer_init(&stats_timer);
761 if (check_period > 0) {
762 ka_shm = rte_keepalive_shm_create();
764 rte_exit(EXIT_FAILURE,
765 "rte_keepalive_shm_create() failed");
766 rte_global_keepalive_info =
767 rte_keepalive_create(&dead_core, ka_shm);
768 if (rte_global_keepalive_info == NULL)
769 rte_exit(EXIT_FAILURE, "init_keep_alive() failed");
770 rte_keepalive_register_relay_callback(rte_global_keepalive_info,
771 relay_core_state, ka_shm);
772 rte_timer_init(&hb_timer);
773 if (rte_timer_reset(&hb_timer,
774 (check_period * rte_get_timer_hz()) / 1000,
777 (void(*)(struct rte_timer*, void*))
778 &rte_keepalive_dispatch_pings,
779 rte_global_keepalive_info
781 rte_exit(EXIT_FAILURE, "Keepalive setup failure.\n");
783 if (timer_period > 0) {
784 if (rte_timer_reset(&stats_timer,
785 (timer_period * rte_get_timer_hz()) / 1000,
790 rte_exit(EXIT_FAILURE, "Stats setup failure.\n");
792 /* launch per-lcore init on every slave lcore */
793 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
794 struct lcore_queue_conf *qconf = &lcore_queue_conf[lcore_id];
796 if (qconf->n_rx_port == 0)
798 "lcore %u has nothing to do\n",
802 rte_eal_remote_launch(
803 l2fwd_launch_one_lcore,
807 rte_keepalive_register_core(rte_global_keepalive_info,
811 while (!terminate_signal_received) {
816 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
817 if (rte_eal_wait_lcore(lcore_id) < 0)
822 rte_keepalive_shm_cleanup(ka_shm);