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>
63 #include <rte_random.h>
64 #include <rte_debug.h>
65 #include <rte_ether.h>
66 #include <rte_ethdev.h>
67 #include <rte_mempool.h>
69 #include <rte_timer.h>
70 #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;
100 #define MAX_RX_QUEUE_PER_LCORE 16
101 #define MAX_TX_QUEUE_PER_PORT 16
102 struct lcore_queue_conf {
104 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
105 } __rte_cache_aligned;
106 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
108 struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
110 static const struct rte_eth_conf port_conf = {
113 .header_split = 0, /**< Header Split disabled */
114 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
115 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
116 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
117 .hw_strip_crc = 1, /**< CRC stripped by hardware */
120 .mq_mode = ETH_MQ_TX_NONE,
124 struct rte_mempool *l2fwd_pktmbuf_pool = NULL;
126 /* Per-port statistics struct */
127 struct l2fwd_port_statistics {
131 } __rte_cache_aligned;
132 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
134 /* A tsc-based timer responsible for triggering statistics printout */
135 #define TIMER_MILLISECOND 1
136 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
137 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* 10 seconds */
138 static int64_t check_period = 5; /* default check cycle is 5ms */
140 /* Keepalive structure */
141 struct rte_keepalive *rte_global_keepalive_info;
143 /* Termination signalling */
144 static int terminate_signal_received;
146 /* Termination signal handler */
147 static void handle_sigterm(__rte_unused int value)
149 terminate_signal_received = 1;
152 /* Print out statistics on packets dropped */
154 print_stats(__attribute__((unused)) struct rte_timer *ptr_timer,
155 __attribute__((unused)) void *ptr_data)
157 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
160 total_packets_dropped = 0;
161 total_packets_tx = 0;
162 total_packets_rx = 0;
164 const char clr[] = { 27, '[', '2', 'J', '\0' };
165 const char topLeft[] = { 27, '[', '1', ';', '1', 'H', '\0' };
167 /* Clear screen and move to top left */
168 printf("%s%s", clr, topLeft);
170 printf("\nPort statistics ====================================");
172 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
173 /* skip disabled ports */
174 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
176 printf("\nStatistics for port %u ------------------------------"
177 "\nPackets sent: %24"PRIu64
178 "\nPackets received: %20"PRIu64
179 "\nPackets dropped: %21"PRIu64,
181 port_statistics[portid].tx,
182 port_statistics[portid].rx,
183 port_statistics[portid].dropped);
185 total_packets_dropped += port_statistics[portid].dropped;
186 total_packets_tx += port_statistics[portid].tx;
187 total_packets_rx += port_statistics[portid].rx;
189 printf("\nAggregate statistics ==============================="
190 "\nTotal packets sent: %18"PRIu64
191 "\nTotal packets received: %14"PRIu64
192 "\nTotal packets dropped: %15"PRIu64,
195 total_packets_dropped);
196 printf("\n====================================================\n");
200 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
202 struct ether_hdr *eth;
206 struct rte_eth_dev_tx_buffer *buffer;
208 dst_port = l2fwd_dst_ports[portid];
209 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
211 /* 02:00:00:00:00:xx */
212 tmp = ð->d_addr.addr_bytes[0];
213 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
216 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
218 buffer = tx_buffer[dst_port];
219 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
221 port_statistics[dst_port].tx += sent;
224 /* main processing loop */
226 l2fwd_main_loop(void)
228 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
232 uint64_t prev_tsc, diff_tsc, cur_tsc;
233 unsigned i, j, portid, nb_rx;
234 struct lcore_queue_conf *qconf;
235 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
236 / US_PER_S * BURST_TX_DRAIN_US;
237 struct rte_eth_dev_tx_buffer *buffer;
241 lcore_id = rte_lcore_id();
242 qconf = &lcore_queue_conf[lcore_id];
244 if (qconf->n_rx_port == 0) {
245 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
249 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
251 for (i = 0; i < qconf->n_rx_port; i++) {
253 portid = qconf->rx_port_list[i];
254 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
258 uint64_t tsc_initial = rte_rdtsc();
259 uint64_t tsc_lifetime = (rand()&0x07) * rte_get_tsc_hz();
261 while (!terminate_signal_received) {
262 /* Keepalive heartbeat */
263 rte_keepalive_mark_alive(rte_global_keepalive_info);
265 cur_tsc = rte_rdtsc();
268 * Die randomly within 7 secs for demo purposes if
271 if (check_period > 0 && cur_tsc - tsc_initial > tsc_lifetime)
275 * TX burst queue drain
277 diff_tsc = cur_tsc - prev_tsc;
278 if (unlikely(diff_tsc > drain_tsc)) {
280 for (i = 0; i < qconf->n_rx_port; i++) {
282 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
283 buffer = tx_buffer[portid];
285 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
287 port_statistics[portid].tx += sent;
295 * Read packet from RX queues
297 for (i = 0; i < qconf->n_rx_port; i++) {
299 portid = qconf->rx_port_list[i];
300 nb_rx = rte_eth_rx_burst(portid, 0,
301 pkts_burst, MAX_PKT_BURST);
303 port_statistics[portid].rx += nb_rx;
305 for (j = 0; j < nb_rx; j++) {
307 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
308 l2fwd_simple_forward(m, portid);
315 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
323 l2fwd_usage(const char *prgname)
325 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
326 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
327 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
328 " -K PERIOD: Keepalive check period (5 default; 86400 max)\n"
329 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
334 l2fwd_parse_portmask(const char *portmask)
339 /* parse hexadecimal string */
340 pm = strtoul(portmask, &end, 16);
341 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
351 l2fwd_parse_nqueue(const char *q_arg)
356 /* parse hexadecimal string */
357 n = strtoul(q_arg, &end, 10);
358 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
362 if (n >= MAX_RX_QUEUE_PER_LCORE)
369 l2fwd_parse_timer_period(const char *q_arg)
374 /* parse number string */
375 n = strtol(q_arg, &end, 10);
376 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
378 if (n >= MAX_TIMER_PERIOD)
385 l2fwd_parse_check_period(const char *q_arg)
390 /* parse number string */
391 n = strtol(q_arg, &end, 10);
392 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
394 if (n >= MAX_TIMER_PERIOD)
400 /* Parse the argument given in the command line of the application */
402 l2fwd_parse_args(int argc, char **argv)
407 char *prgname = argv[0];
408 static struct option lgopts[] = {
414 while ((opt = getopt_long(argc, argvopt, "p:q:T:K:",
415 lgopts, &option_index)) != EOF) {
420 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
421 if (l2fwd_enabled_port_mask == 0) {
422 printf("invalid portmask\n");
423 l2fwd_usage(prgname);
430 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
431 if (l2fwd_rx_queue_per_lcore == 0) {
432 printf("invalid queue number\n");
433 l2fwd_usage(prgname);
440 timer_period = l2fwd_parse_timer_period(optarg)
441 * (int64_t)(1000 * TIMER_MILLISECOND);
442 if (timer_period < 0) {
443 printf("invalid timer period\n");
444 l2fwd_usage(prgname);
451 check_period = l2fwd_parse_check_period(optarg);
452 if (check_period < 0) {
453 printf("invalid check period\n");
454 l2fwd_usage(prgname);
461 l2fwd_usage(prgname);
465 l2fwd_usage(prgname);
471 argv[optind-1] = prgname;
474 optind = 1; /* reset getopt lib */
478 /* Check the link status of all ports in up to 9s, and print them finally */
480 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
482 #define CHECK_INTERVAL 100 /* 100ms */
483 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
485 uint8_t count, all_ports_up, print_flag = 0;
486 struct rte_eth_link link;
488 printf("\nChecking link status");
490 for (count = 0; count <= MAX_CHECK_TIME; count++) {
492 for (portid = 0; portid < port_num; portid++) {
493 if ((port_mask & (1 << portid)) == 0)
495 memset(&link, 0, sizeof(link));
496 rte_eth_link_get_nowait(portid, &link);
497 /* print link status if flag set */
498 if (print_flag == 1) {
499 if (link.link_status)
501 "Port%d Link Up. Speed %u Mbps - %s\n",
502 portid, link.link_speed,
503 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
504 ("full-duplex") : ("half-duplex\n"));
506 printf("Port %d Link Down\n", portid);
509 /* clear all_ports_up flag if any link down */
510 if (link.link_status == ETH_LINK_DOWN) {
515 /* after finally printing all link status, get out */
519 if (all_ports_up == 0) {
522 rte_delay_ms(CHECK_INTERVAL);
525 /* set the print_flag if all ports up or timeout */
526 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
534 dead_core(__rte_unused void *ptr_data, const int id_core)
536 if (terminate_signal_received)
538 printf("Dead core %i - restarting..\n", id_core);
539 if (rte_eal_get_lcore_state(id_core) == FINISHED) {
540 rte_eal_wait_lcore(id_core);
541 rte_eal_remote_launch(l2fwd_launch_one_lcore, NULL, id_core);
543 printf("..false positive!\n");
548 relay_core_state(void *ptr_data, const int id_core,
549 const enum rte_keepalive_state core_state, uint64_t last_alive)
551 rte_keepalive_relayed_state((struct rte_keepalive_shm *)ptr_data,
552 id_core, core_state, last_alive);
556 main(int argc, char **argv)
558 struct lcore_queue_conf *qconf;
559 struct rte_eth_dev_info dev_info;
562 uint16_t nb_ports_available;
563 uint16_t portid, last_port;
564 unsigned lcore_id, rx_lcore_id;
565 unsigned nb_ports_in_mask = 0;
566 struct sigaction signal_handler;
567 struct rte_keepalive_shm *ka_shm;
569 memset(&signal_handler, 0, sizeof(signal_handler));
570 terminate_signal_received = 0;
571 signal_handler.sa_handler = &handle_sigterm;
572 if (sigaction(SIGINT, &signal_handler, NULL) == -1 ||
573 sigaction(SIGTERM, &signal_handler, NULL) == -1)
574 rte_exit(EXIT_FAILURE, "SIGNAL\n");
578 ret = rte_eal_init(argc, argv);
580 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
584 l2fwd_enabled_port_mask = 0;
586 /* parse application arguments (after the EAL ones) */
587 ret = l2fwd_parse_args(argc, argv);
589 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
591 /* create the mbuf pool */
592 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF, 32,
593 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
594 if (l2fwd_pktmbuf_pool == NULL)
595 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
597 nb_ports = rte_eth_dev_count();
599 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
601 /* reset l2fwd_dst_ports */
602 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
603 l2fwd_dst_ports[portid] = 0;
607 * Each logical core is assigned a dedicated TX queue on each port.
609 for (portid = 0; portid < nb_ports; portid++) {
610 /* skip ports that are not enabled */
611 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
614 if (nb_ports_in_mask % 2) {
615 l2fwd_dst_ports[portid] = last_port;
616 l2fwd_dst_ports[last_port] = portid;
622 rte_eth_dev_info_get(portid, &dev_info);
624 if (nb_ports_in_mask % 2) {
625 printf("Notice: odd number of ports in portmask.\n");
626 l2fwd_dst_ports[last_port] = last_port;
632 /* Initialize the port/queue configuration of each logical core */
633 for (portid = 0; portid < nb_ports; portid++) {
634 /* skip ports that are not enabled */
635 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
638 /* get the lcore_id for this port */
639 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
640 lcore_queue_conf[rx_lcore_id].n_rx_port ==
641 l2fwd_rx_queue_per_lcore) {
643 if (rx_lcore_id >= RTE_MAX_LCORE)
644 rte_exit(EXIT_FAILURE, "Not enough cores\n");
647 if (qconf != &lcore_queue_conf[rx_lcore_id])
648 /* Assigned a new logical core in the loop above. */
649 qconf = &lcore_queue_conf[rx_lcore_id];
651 qconf->rx_port_list[qconf->n_rx_port] = portid;
653 printf("Lcore %u: RX port %u\n",
654 rx_lcore_id, portid);
657 nb_ports_available = nb_ports;
659 /* Initialise each port */
660 for (portid = 0; portid < nb_ports; portid++) {
661 /* skip ports that are not enabled */
662 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
663 printf("Skipping disabled port %u\n", portid);
664 nb_ports_available--;
668 printf("Initializing port %u... ", portid);
670 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
672 rte_exit(EXIT_FAILURE,
673 "Cannot configure device: err=%d, port=%u\n",
676 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
679 rte_exit(EXIT_FAILURE,
680 "Cannot adjust number of descriptors: err=%d, port=%u\n",
683 rte_eth_macaddr_get(portid, &l2fwd_ports_eth_addr[portid]);
685 /* init one RX queue */
687 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
688 rte_eth_dev_socket_id(portid),
692 rte_exit(EXIT_FAILURE,
693 "rte_eth_rx_queue_setup:err=%d, port=%u\n",
696 /* init one TX queue on each port */
698 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
699 rte_eth_dev_socket_id(portid),
702 rte_exit(EXIT_FAILURE,
703 "rte_eth_tx_queue_setup:err=%d, port=%u\n",
706 /* Initialize TX buffers */
707 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
708 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
709 rte_eth_dev_socket_id(portid));
710 if (tx_buffer[portid] == NULL)
711 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
714 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
716 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
717 rte_eth_tx_buffer_count_callback,
718 &port_statistics[portid].dropped);
720 rte_exit(EXIT_FAILURE,
721 "Cannot set error callback for tx buffer on port %u\n",
725 ret = rte_eth_dev_start(portid);
727 rte_exit(EXIT_FAILURE,
728 "rte_eth_dev_start:err=%d, port=%u\n",
731 rte_eth_promiscuous_enable(portid);
733 printf("Port %u, MAC address: "
734 "%02X:%02X:%02X:%02X:%02X:%02X\n\n",
736 l2fwd_ports_eth_addr[portid].addr_bytes[0],
737 l2fwd_ports_eth_addr[portid].addr_bytes[1],
738 l2fwd_ports_eth_addr[portid].addr_bytes[2],
739 l2fwd_ports_eth_addr[portid].addr_bytes[3],
740 l2fwd_ports_eth_addr[portid].addr_bytes[4],
741 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
743 /* initialize port stats */
744 memset(&port_statistics, 0, sizeof(port_statistics));
747 if (!nb_ports_available) {
748 rte_exit(EXIT_FAILURE,
749 "All available ports are disabled. Please set portmask.\n");
752 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
754 struct rte_timer hb_timer, stats_timer;
756 rte_timer_subsystem_init();
757 rte_timer_init(&stats_timer);
760 if (check_period > 0) {
761 ka_shm = rte_keepalive_shm_create();
763 rte_exit(EXIT_FAILURE,
764 "rte_keepalive_shm_create() failed");
765 rte_global_keepalive_info =
766 rte_keepalive_create(&dead_core, ka_shm);
767 if (rte_global_keepalive_info == NULL)
768 rte_exit(EXIT_FAILURE, "init_keep_alive() failed");
769 rte_keepalive_register_relay_callback(rte_global_keepalive_info,
770 relay_core_state, ka_shm);
771 rte_timer_init(&hb_timer);
772 if (rte_timer_reset(&hb_timer,
773 (check_period * rte_get_timer_hz()) / 1000,
776 (void(*)(struct rte_timer*, void*))
777 &rte_keepalive_dispatch_pings,
778 rte_global_keepalive_info
780 rte_exit(EXIT_FAILURE, "Keepalive setup failure.\n");
782 if (timer_period > 0) {
783 if (rte_timer_reset(&stats_timer,
784 (timer_period * rte_get_timer_hz()) / 1000,
789 rte_exit(EXIT_FAILURE, "Stats setup failure.\n");
791 /* launch per-lcore init on every slave lcore */
792 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
793 struct lcore_queue_conf *qconf = &lcore_queue_conf[lcore_id];
795 if (qconf->n_rx_port == 0)
797 "lcore %u has nothing to do\n",
801 rte_eal_remote_launch(
802 l2fwd_launch_one_lcore,
806 rte_keepalive_register_core(rte_global_keepalive_info,
810 while (!terminate_signal_received) {
815 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
816 if (rte_eal_wait_lcore(lcore_id) < 0)
821 rte_keepalive_shm_cleanup(ka_shm);