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>
48 #include <rte_common.h>
50 #include <rte_malloc.h>
51 #include <rte_memory.h>
52 #include <rte_memcpy.h>
53 #include <rte_memzone.h>
55 #include <rte_per_lcore.h>
56 #include <rte_launch.h>
57 #include <rte_atomic.h>
58 #include <rte_cycles.h>
59 #include <rte_prefetch.h>
60 #include <rte_lcore.h>
61 #include <rte_per_lcore.h>
62 #include <rte_branch_prediction.h>
63 #include <rte_interrupts.h>
65 #include <rte_random.h>
66 #include <rte_debug.h>
67 #include <rte_ether.h>
68 #include <rte_ethdev.h>
70 #include <rte_mempool.h>
72 #include <rte_timer.h>
73 #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 = 0, /**< 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 /* Print out statistics on packets dropped */
146 print_stats(__attribute__((unused)) struct rte_timer *ptr_timer,
147 __attribute__((unused)) void *ptr_data)
149 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
152 total_packets_dropped = 0;
153 total_packets_tx = 0;
154 total_packets_rx = 0;
156 const char clr[] = { 27, '[', '2', 'J', '\0' };
157 const char topLeft[] = { 27, '[', '1', ';', '1', 'H', '\0' };
159 /* Clear screen and move to top left */
160 printf("%s%s", clr, topLeft);
162 printf("\nPort statistics ====================================");
164 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
165 /* skip disabled ports */
166 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
168 printf("\nStatistics for port %u ------------------------------"
169 "\nPackets sent: %24"PRIu64
170 "\nPackets received: %20"PRIu64
171 "\nPackets dropped: %21"PRIu64,
173 port_statistics[portid].tx,
174 port_statistics[portid].rx,
175 port_statistics[portid].dropped);
177 total_packets_dropped += port_statistics[portid].dropped;
178 total_packets_tx += port_statistics[portid].tx;
179 total_packets_rx += port_statistics[portid].rx;
181 printf("\nAggregate statistics ==============================="
182 "\nTotal packets sent: %18"PRIu64
183 "\nTotal packets received: %14"PRIu64
184 "\nTotal packets dropped: %15"PRIu64,
187 total_packets_dropped);
188 printf("\n====================================================\n");
192 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
194 struct ether_hdr *eth;
198 struct rte_eth_dev_tx_buffer *buffer;
200 dst_port = l2fwd_dst_ports[portid];
201 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
203 /* 02:00:00:00:00:xx */
204 tmp = ð->d_addr.addr_bytes[0];
205 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
208 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
210 buffer = tx_buffer[dst_port];
211 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
213 port_statistics[dst_port].tx += sent;
216 /* main processing loop */
218 l2fwd_main_loop(void)
220 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
224 uint64_t prev_tsc, diff_tsc, cur_tsc;
225 unsigned i, j, portid, nb_rx;
226 struct lcore_queue_conf *qconf;
227 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
228 / US_PER_S * BURST_TX_DRAIN_US;
229 struct rte_eth_dev_tx_buffer *buffer;
233 lcore_id = rte_lcore_id();
234 qconf = &lcore_queue_conf[lcore_id];
236 if (qconf->n_rx_port == 0) {
237 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
241 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
243 for (i = 0; i < qconf->n_rx_port; i++) {
245 portid = qconf->rx_port_list[i];
246 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
250 uint64_t tsc_initial = rte_rdtsc();
251 uint64_t tsc_lifetime = (rand()&0x07) * rte_get_tsc_hz();
254 /* Keepalive heartbeat */
255 rte_keepalive_mark_alive(rte_global_keepalive_info);
257 cur_tsc = rte_rdtsc();
260 * Die randomly within 7 secs for demo purposes if
263 if (check_period > 0 && cur_tsc - tsc_initial > tsc_lifetime)
267 * TX burst queue drain
269 diff_tsc = cur_tsc - prev_tsc;
270 if (unlikely(diff_tsc > drain_tsc)) {
272 for (i = 0; i < qconf->n_rx_port; i++) {
274 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
275 buffer = tx_buffer[portid];
277 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
279 port_statistics[portid].tx += sent;
287 * Read packet from RX queues
289 for (i = 0; i < qconf->n_rx_port; i++) {
291 portid = qconf->rx_port_list[i];
292 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
293 pkts_burst, MAX_PKT_BURST);
295 port_statistics[portid].rx += nb_rx;
297 for (j = 0; j < nb_rx; j++) {
299 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
300 l2fwd_simple_forward(m, portid);
307 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
315 l2fwd_usage(const char *prgname)
317 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
318 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
319 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
320 " -K PERIOD: Keepalive check period (5 default; 86400 max)\n"
321 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
326 l2fwd_parse_portmask(const char *portmask)
331 /* parse hexadecimal string */
332 pm = strtoul(portmask, &end, 16);
333 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
343 l2fwd_parse_nqueue(const char *q_arg)
348 /* parse hexadecimal string */
349 n = strtoul(q_arg, &end, 10);
350 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
354 if (n >= MAX_RX_QUEUE_PER_LCORE)
361 l2fwd_parse_timer_period(const char *q_arg)
366 /* parse number string */
367 n = strtol(q_arg, &end, 10);
368 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
370 if (n >= MAX_TIMER_PERIOD)
377 l2fwd_parse_check_period(const char *q_arg)
382 /* parse number string */
383 n = strtol(q_arg, &end, 10);
384 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
386 if (n >= MAX_TIMER_PERIOD)
392 /* Parse the argument given in the command line of the application */
394 l2fwd_parse_args(int argc, char **argv)
399 char *prgname = argv[0];
400 static struct option lgopts[] = {
406 while ((opt = getopt_long(argc, argvopt, "p:q:T:K:",
407 lgopts, &option_index)) != EOF) {
412 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
413 if (l2fwd_enabled_port_mask == 0) {
414 printf("invalid portmask\n");
415 l2fwd_usage(prgname);
422 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
423 if (l2fwd_rx_queue_per_lcore == 0) {
424 printf("invalid queue number\n");
425 l2fwd_usage(prgname);
432 timer_period = l2fwd_parse_timer_period(optarg)
433 * (int64_t)(1000 * TIMER_MILLISECOND);
434 if (timer_period < 0) {
435 printf("invalid timer period\n");
436 l2fwd_usage(prgname);
443 check_period = l2fwd_parse_check_period(optarg);
444 if (check_period < 0) {
445 printf("invalid check period\n");
446 l2fwd_usage(prgname);
453 l2fwd_usage(prgname);
457 l2fwd_usage(prgname);
463 argv[optind-1] = prgname;
466 optind = 0; /* reset getopt lib */
470 /* Check the link status of all ports in up to 9s, and print them finally */
472 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
474 #define CHECK_INTERVAL 100 /* 100ms */
475 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
476 uint8_t portid, count, all_ports_up, print_flag = 0;
477 struct rte_eth_link link;
479 printf("\nChecking link status");
481 for (count = 0; count <= MAX_CHECK_TIME; count++) {
483 for (portid = 0; portid < port_num; portid++) {
484 if ((port_mask & (1 << portid)) == 0)
486 memset(&link, 0, sizeof(link));
487 rte_eth_link_get_nowait(portid, &link);
488 /* print link status if flag set */
489 if (print_flag == 1) {
490 if (link.link_status)
491 printf("Port %d Link Up - speed %u "
492 "Mbps - %s\n", (uint8_t)portid,
493 (unsigned)link.link_speed,
494 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
495 ("full-duplex") : ("half-duplex\n"));
497 printf("Port %d Link Down\n",
501 /* clear all_ports_up flag if any link down */
502 if (link.link_status == ETH_LINK_DOWN) {
507 /* after finally printing all link status, get out */
511 if (all_ports_up == 0) {
514 rte_delay_ms(CHECK_INTERVAL);
517 /* set the print_flag if all ports up or timeout */
518 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
526 dead_core(__attribute__((unused)) void *ptr_data, const int id_core)
528 printf("Dead core %i - restarting..\n", id_core);
529 if (rte_eal_get_lcore_state(id_core) == FINISHED) {
530 rte_eal_wait_lcore(id_core);
531 rte_eal_remote_launch(l2fwd_launch_one_lcore, NULL, id_core);
533 printf("..false positive!\n");
538 main(int argc, char **argv)
540 struct lcore_queue_conf *qconf;
541 struct rte_eth_dev_info dev_info;
544 uint8_t nb_ports_available;
545 uint8_t portid, last_port;
546 unsigned lcore_id, rx_lcore_id;
547 unsigned nb_ports_in_mask = 0;
550 ret = rte_eal_init(argc, argv);
552 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
556 l2fwd_enabled_port_mask = 0;
558 /* parse application arguments (after the EAL ones) */
559 ret = l2fwd_parse_args(argc, argv);
561 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
563 /* create the mbuf pool */
564 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF, 32,
565 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
566 if (l2fwd_pktmbuf_pool == NULL)
567 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
569 nb_ports = rte_eth_dev_count();
571 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
573 /* reset l2fwd_dst_ports */
574 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
575 l2fwd_dst_ports[portid] = 0;
579 * Each logical core is assigned a dedicated TX queue on each port.
581 for (portid = 0; portid < nb_ports; portid++) {
582 /* skip ports that are not enabled */
583 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
586 if (nb_ports_in_mask % 2) {
587 l2fwd_dst_ports[portid] = last_port;
588 l2fwd_dst_ports[last_port] = portid;
594 rte_eth_dev_info_get(portid, &dev_info);
596 if (nb_ports_in_mask % 2) {
597 printf("Notice: odd number of ports in portmask.\n");
598 l2fwd_dst_ports[last_port] = last_port;
604 /* Initialize the port/queue configuration of each logical core */
605 for (portid = 0; portid < nb_ports; portid++) {
606 /* skip ports that are not enabled */
607 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
610 /* get the lcore_id for this port */
611 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
612 lcore_queue_conf[rx_lcore_id].n_rx_port ==
613 l2fwd_rx_queue_per_lcore) {
615 if (rx_lcore_id >= RTE_MAX_LCORE)
616 rte_exit(EXIT_FAILURE, "Not enough cores\n");
619 if (qconf != &lcore_queue_conf[rx_lcore_id])
620 /* Assigned a new logical core in the loop above. */
621 qconf = &lcore_queue_conf[rx_lcore_id];
623 qconf->rx_port_list[qconf->n_rx_port] = portid;
625 printf("Lcore %u: RX port %u\n",
626 rx_lcore_id, (unsigned) portid);
629 nb_ports_available = nb_ports;
631 /* Initialise each port */
632 for (portid = 0; portid < nb_ports; portid++) {
633 /* skip ports that are not enabled */
634 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
635 printf("Skipping disabled port %u\n",
637 nb_ports_available--;
641 printf("Initializing port %u... ", (unsigned) portid);
643 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
645 rte_exit(EXIT_FAILURE,
646 "Cannot configure device: err=%d, port=%u\n",
647 ret, (unsigned) portid);
649 rte_eth_macaddr_get(portid, &l2fwd_ports_eth_addr[portid]);
651 /* init one RX queue */
653 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
654 rte_eth_dev_socket_id(portid),
658 rte_exit(EXIT_FAILURE,
659 "rte_eth_rx_queue_setup:err=%d, port=%u\n",
660 ret, (unsigned) portid);
662 /* init one TX queue on each port */
664 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
665 rte_eth_dev_socket_id(portid),
668 rte_exit(EXIT_FAILURE,
669 "rte_eth_tx_queue_setup:err=%d, port=%u\n",
670 ret, (unsigned) portid);
672 /* Initialize TX buffers */
673 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
674 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
675 rte_eth_dev_socket_id(portid));
676 if (tx_buffer[portid] == NULL)
677 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
680 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
682 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
683 rte_eth_tx_buffer_count_callback,
684 &port_statistics[portid].dropped);
686 rte_exit(EXIT_FAILURE, "Cannot set error callback for "
687 "tx buffer on port %u\n", (unsigned) portid);
690 ret = rte_eth_dev_start(portid);
692 rte_exit(EXIT_FAILURE,
693 "rte_eth_dev_start:err=%d, port=%u\n",
694 ret, (unsigned) portid);
696 rte_eth_promiscuous_enable(portid);
698 printf("Port %u, MAC address: "
699 "%02X:%02X:%02X:%02X:%02X:%02X\n\n",
701 l2fwd_ports_eth_addr[portid].addr_bytes[0],
702 l2fwd_ports_eth_addr[portid].addr_bytes[1],
703 l2fwd_ports_eth_addr[portid].addr_bytes[2],
704 l2fwd_ports_eth_addr[portid].addr_bytes[3],
705 l2fwd_ports_eth_addr[portid].addr_bytes[4],
706 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
708 /* initialize port stats */
709 memset(&port_statistics, 0, sizeof(port_statistics));
712 if (!nb_ports_available) {
713 rte_exit(EXIT_FAILURE,
714 "All available ports are disabled. Please set portmask.\n");
717 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
719 struct rte_timer hb_timer, stats_timer;
721 rte_timer_subsystem_init();
722 rte_timer_init(&stats_timer);
724 if (check_period > 0) {
725 rte_global_keepalive_info =
726 rte_keepalive_create(&dead_core, NULL);
727 if (rte_global_keepalive_info == NULL)
728 rte_exit(EXIT_FAILURE, "init_keep_alive() failed");
729 rte_timer_init(&hb_timer);
730 if (rte_timer_reset(&hb_timer,
731 (check_period * rte_get_timer_hz()) / 1000,
734 (void(*)(struct rte_timer*, void*))
735 &rte_keepalive_dispatch_pings,
736 rte_global_keepalive_info
738 rte_exit(EXIT_FAILURE, "Keepalive setup failure.\n");
740 if (timer_period > 0) {
741 if (rte_timer_reset(&stats_timer,
742 (timer_period * rte_get_timer_hz()) / 1000,
747 rte_exit(EXIT_FAILURE, "Stats setup failure.\n");
749 /* launch per-lcore init on every slave lcore */
750 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
751 struct lcore_queue_conf *qconf = &lcore_queue_conf[lcore_id];
753 if (qconf->n_rx_port == 0)
755 "lcore %u has nothing to do\n",
759 rte_eal_remote_launch(
760 l2fwd_launch_one_lcore,
764 rte_keepalive_register_core(rte_global_keepalive_info,
773 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
774 if (rte_eal_wait_lcore(lcore_id) < 0)