4 * Copyright(c) 2010-2014 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>
53 #include <rte_tailq.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>
75 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
77 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
80 #define MAX_PKT_BURST 32
81 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
84 * Configurable number of RX/TX ring descriptors
86 #define RTE_TEST_RX_DESC_DEFAULT 128
87 #define RTE_TEST_TX_DESC_DEFAULT 512
88 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
89 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
91 /* ethernet addresses of ports */
92 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
94 /* mask of enabled ports */
95 static uint32_t l2fwd_enabled_port_mask = 0;
97 /* list of enabled ports */
98 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
100 static unsigned int l2fwd_rx_queue_per_lcore = 1;
104 struct rte_mbuf *m_table[MAX_PKT_BURST];
107 #define MAX_RX_QUEUE_PER_LCORE 16
108 #define MAX_TX_QUEUE_PER_PORT 16
109 struct lcore_queue_conf {
111 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
112 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
114 } __rte_cache_aligned;
115 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
117 static const struct rte_eth_conf port_conf = {
120 .header_split = 0, /**< Header Split disabled */
121 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
122 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
123 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
124 .hw_strip_crc = 0, /**< CRC stripped by hardware */
127 .mq_mode = ETH_MQ_TX_NONE,
131 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
133 /* Per-port statistics struct */
134 struct l2fwd_port_statistics {
138 } __rte_cache_aligned;
139 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
141 /* A tsc-based timer responsible for triggering statistics printout */
142 #define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
143 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
144 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
146 /* Print out statistics on packets dropped */
150 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
153 total_packets_dropped = 0;
154 total_packets_tx = 0;
155 total_packets_rx = 0;
157 const char clr[] = { 27, '[', '2', 'J', '\0' };
158 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
160 /* Clear screen and move to top left */
161 printf("%s%s", clr, topLeft);
163 printf("\nPort statistics ====================================");
165 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
166 /* skip disabled ports */
167 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
169 printf("\nStatistics for port %u ------------------------------"
170 "\nPackets sent: %24"PRIu64
171 "\nPackets received: %20"PRIu64
172 "\nPackets dropped: %21"PRIu64,
174 port_statistics[portid].tx,
175 port_statistics[portid].rx,
176 port_statistics[portid].dropped);
178 total_packets_dropped += port_statistics[portid].dropped;
179 total_packets_tx += port_statistics[portid].tx;
180 total_packets_rx += port_statistics[portid].rx;
182 printf("\nAggregate statistics ==============================="
183 "\nTotal packets sent: %18"PRIu64
184 "\nTotal packets received: %14"PRIu64
185 "\nTotal packets dropped: %15"PRIu64,
188 total_packets_dropped);
189 printf("\n====================================================\n");
192 /* Send the burst of packets on an output interface */
194 l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
196 struct rte_mbuf **m_table;
200 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
202 ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
203 port_statistics[port].tx += ret;
204 if (unlikely(ret < n)) {
205 port_statistics[port].dropped += (n - ret);
207 rte_pktmbuf_free(m_table[ret]);
214 /* Enqueue packets for TX and prepare them to be sent */
216 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
218 unsigned lcore_id, len;
219 struct lcore_queue_conf *qconf;
221 lcore_id = rte_lcore_id();
223 qconf = &lcore_queue_conf[lcore_id];
224 len = qconf->tx_mbufs[port].len;
225 qconf->tx_mbufs[port].m_table[len] = m;
228 /* enough pkts to be sent */
229 if (unlikely(len == MAX_PKT_BURST)) {
230 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
234 qconf->tx_mbufs[port].len = len;
239 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
241 struct ether_hdr *eth;
245 dst_port = l2fwd_dst_ports[portid];
246 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
248 /* 02:00:00:00:00:xx */
249 tmp = ð->d_addr.addr_bytes[0];
250 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
253 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
255 l2fwd_send_packet(m, (uint8_t) dst_port);
258 /* main processing loop */
260 l2fwd_main_loop(void)
262 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
265 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
266 unsigned i, j, portid, nb_rx;
267 struct lcore_queue_conf *qconf;
268 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
273 lcore_id = rte_lcore_id();
274 qconf = &lcore_queue_conf[lcore_id];
276 if (qconf->n_rx_port == 0) {
277 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
281 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
283 for (i = 0; i < qconf->n_rx_port; i++) {
285 portid = qconf->rx_port_list[i];
286 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
292 cur_tsc = rte_rdtsc();
295 * TX burst queue drain
297 diff_tsc = cur_tsc - prev_tsc;
298 if (unlikely(diff_tsc > drain_tsc)) {
300 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
301 if (qconf->tx_mbufs[portid].len == 0)
303 l2fwd_send_burst(&lcore_queue_conf[lcore_id],
304 qconf->tx_mbufs[portid].len,
306 qconf->tx_mbufs[portid].len = 0;
309 /* if timer is enabled */
310 if (timer_period > 0) {
312 /* advance the timer */
313 timer_tsc += diff_tsc;
315 /* if timer has reached its timeout */
316 if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
318 /* do this only on master core */
319 if (lcore_id == rte_get_master_lcore()) {
321 /* reset the timer */
331 * Read packet from RX queues
333 for (i = 0; i < qconf->n_rx_port; i++) {
335 portid = qconf->rx_port_list[i];
336 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
337 pkts_burst, MAX_PKT_BURST);
339 port_statistics[portid].rx += nb_rx;
341 for (j = 0; j < nb_rx; j++) {
343 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
344 l2fwd_simple_forward(m, portid);
351 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
359 l2fwd_usage(const char *prgname)
361 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
362 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
363 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
364 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
369 l2fwd_parse_portmask(const char *portmask)
374 /* parse hexadecimal string */
375 pm = strtoul(portmask, &end, 16);
376 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
386 l2fwd_parse_nqueue(const char *q_arg)
391 /* parse hexadecimal string */
392 n = strtoul(q_arg, &end, 10);
393 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
397 if (n >= MAX_RX_QUEUE_PER_LCORE)
404 l2fwd_parse_timer_period(const char *q_arg)
409 /* parse number string */
410 n = strtol(q_arg, &end, 10);
411 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
413 if (n >= MAX_TIMER_PERIOD)
419 /* Parse the argument given in the command line of the application */
421 l2fwd_parse_args(int argc, char **argv)
426 char *prgname = argv[0];
427 static struct option lgopts[] = {
433 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
434 lgopts, &option_index)) != EOF) {
439 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
440 if (l2fwd_enabled_port_mask == 0) {
441 printf("invalid portmask\n");
442 l2fwd_usage(prgname);
449 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
450 if (l2fwd_rx_queue_per_lcore == 0) {
451 printf("invalid queue number\n");
452 l2fwd_usage(prgname);
459 timer_period = l2fwd_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
460 if (timer_period < 0) {
461 printf("invalid timer period\n");
462 l2fwd_usage(prgname);
469 l2fwd_usage(prgname);
473 l2fwd_usage(prgname);
479 argv[optind-1] = prgname;
482 optind = 0; /* reset getopt lib */
486 /* Check the link status of all ports in up to 9s, and print them finally */
488 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
490 #define CHECK_INTERVAL 100 /* 100ms */
491 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
492 uint8_t portid, count, all_ports_up, print_flag = 0;
493 struct rte_eth_link link;
495 printf("\nChecking link status");
497 for (count = 0; count <= MAX_CHECK_TIME; count++) {
499 for (portid = 0; portid < port_num; portid++) {
500 if ((port_mask & (1 << portid)) == 0)
502 memset(&link, 0, sizeof(link));
503 rte_eth_link_get_nowait(portid, &link);
504 /* print link status if flag set */
505 if (print_flag == 1) {
506 if (link.link_status)
507 printf("Port %d Link Up - speed %u "
508 "Mbps - %s\n", (uint8_t)portid,
509 (unsigned)link.link_speed,
510 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
511 ("full-duplex") : ("half-duplex\n"));
513 printf("Port %d Link Down\n",
517 /* clear all_ports_up flag if any link down */
518 if (link.link_status == 0) {
523 /* after finally printing all link status, get out */
527 if (all_ports_up == 0) {
530 rte_delay_ms(CHECK_INTERVAL);
533 /* set the print_flag if all ports up or timeout */
534 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
542 MAIN(int argc, char **argv)
544 struct lcore_queue_conf *qconf;
545 struct rte_eth_dev_info dev_info;
548 uint8_t nb_ports_available;
549 uint8_t portid, last_port;
550 unsigned lcore_id, rx_lcore_id;
551 unsigned nb_ports_in_mask = 0;
554 ret = rte_eal_init(argc, argv);
556 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
560 /* parse application arguments (after the EAL ones) */
561 ret = l2fwd_parse_args(argc, argv);
563 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
565 /* create the mbuf pool */
567 rte_mempool_create("mbuf_pool", NB_MBUF,
569 sizeof(struct rte_pktmbuf_pool_private),
570 rte_pktmbuf_pool_init, NULL,
571 rte_pktmbuf_init, NULL,
573 if (l2fwd_pktmbuf_pool == NULL)
574 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
576 nb_ports = rte_eth_dev_count();
578 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
580 if (nb_ports > RTE_MAX_ETHPORTS)
581 nb_ports = RTE_MAX_ETHPORTS;
583 /* reset l2fwd_dst_ports */
584 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
585 l2fwd_dst_ports[portid] = 0;
589 * Each logical core is assigned a dedicated TX queue on each port.
591 for (portid = 0; portid < nb_ports; portid++) {
592 /* skip ports that are not enabled */
593 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
596 if (nb_ports_in_mask % 2) {
597 l2fwd_dst_ports[portid] = last_port;
598 l2fwd_dst_ports[last_port] = portid;
605 rte_eth_dev_info_get(portid, &dev_info);
607 if (nb_ports_in_mask % 2) {
608 printf("Notice: odd number of ports in portmask.\n");
609 l2fwd_dst_ports[last_port] = last_port;
615 /* Initialize the port/queue configuration of each logical core */
616 for (portid = 0; portid < nb_ports; portid++) {
617 /* skip ports that are not enabled */
618 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
621 /* get the lcore_id for this port */
622 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
623 lcore_queue_conf[rx_lcore_id].n_rx_port ==
624 l2fwd_rx_queue_per_lcore) {
626 if (rx_lcore_id >= RTE_MAX_LCORE)
627 rte_exit(EXIT_FAILURE, "Not enough cores\n");
630 if (qconf != &lcore_queue_conf[rx_lcore_id])
631 /* Assigned a new logical core in the loop above. */
632 qconf = &lcore_queue_conf[rx_lcore_id];
634 qconf->rx_port_list[qconf->n_rx_port] = portid;
636 printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
639 nb_ports_available = nb_ports;
641 /* Initialise each port */
642 for (portid = 0; portid < nb_ports; portid++) {
643 /* skip ports that are not enabled */
644 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
645 printf("Skipping disabled port %u\n", (unsigned) portid);
646 nb_ports_available--;
650 printf("Initializing port %u... ", (unsigned) portid);
652 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
654 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
655 ret, (unsigned) portid);
657 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
659 /* init one RX queue */
661 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
662 rte_eth_dev_socket_id(portid),
666 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
667 ret, (unsigned) portid);
669 /* init one TX queue on each port */
671 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
672 rte_eth_dev_socket_id(portid),
675 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
676 ret, (unsigned) portid);
679 ret = rte_eth_dev_start(portid);
681 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
682 ret, (unsigned) portid);
686 rte_eth_promiscuous_enable(portid);
688 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
690 l2fwd_ports_eth_addr[portid].addr_bytes[0],
691 l2fwd_ports_eth_addr[portid].addr_bytes[1],
692 l2fwd_ports_eth_addr[portid].addr_bytes[2],
693 l2fwd_ports_eth_addr[portid].addr_bytes[3],
694 l2fwd_ports_eth_addr[portid].addr_bytes[4],
695 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
697 /* initialize port stats */
698 memset(&port_statistics, 0, sizeof(port_statistics));
701 if (!nb_ports_available) {
702 rte_exit(EXIT_FAILURE,
703 "All available ports are disabled. Please set portmask.\n");
706 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
708 /* launch per-lcore init on every lcore */
709 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
710 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
711 if (rte_eal_wait_lcore(lcore_id) < 0)