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>
54 #include <rte_per_lcore.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>
69 #include <rte_mempool.h>
72 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
74 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
77 #define MAX_PKT_BURST 32
78 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
81 * Configurable number of RX/TX ring descriptors
83 #define RTE_TEST_RX_DESC_DEFAULT 128
84 #define RTE_TEST_TX_DESC_DEFAULT 512
85 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
86 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
88 /* ethernet addresses of ports */
89 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
91 /* mask of enabled ports */
92 static uint32_t l2fwd_enabled_port_mask = 0;
94 /* list of enabled ports */
95 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
97 static unsigned int l2fwd_rx_queue_per_lcore = 1;
101 struct rte_mbuf *m_table[MAX_PKT_BURST];
104 #define MAX_RX_QUEUE_PER_LCORE 16
105 #define MAX_TX_QUEUE_PER_PORT 16
106 struct lcore_queue_conf {
108 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
109 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
111 } __rte_cache_aligned;
112 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
114 static const struct rte_eth_conf port_conf = {
117 .header_split = 0, /**< Header Split disabled */
118 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
119 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
120 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
121 .hw_strip_crc = 0, /**< CRC stripped by hardware */
124 .mq_mode = ETH_MQ_TX_NONE,
128 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
130 /* Per-port statistics struct */
131 struct l2fwd_port_statistics {
135 } __rte_cache_aligned;
136 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
138 /* A tsc-based timer responsible for triggering statistics printout */
139 #define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
140 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
141 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
143 /* Print out statistics on packets dropped */
147 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
150 total_packets_dropped = 0;
151 total_packets_tx = 0;
152 total_packets_rx = 0;
154 const char clr[] = { 27, '[', '2', 'J', '\0' };
155 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
157 /* Clear screen and move to top left */
158 printf("%s%s", clr, topLeft);
160 printf("\nPort statistics ====================================");
162 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
163 /* skip disabled ports */
164 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
166 printf("\nStatistics for port %u ------------------------------"
167 "\nPackets sent: %24"PRIu64
168 "\nPackets received: %20"PRIu64
169 "\nPackets dropped: %21"PRIu64,
171 port_statistics[portid].tx,
172 port_statistics[portid].rx,
173 port_statistics[portid].dropped);
175 total_packets_dropped += port_statistics[portid].dropped;
176 total_packets_tx += port_statistics[portid].tx;
177 total_packets_rx += port_statistics[portid].rx;
179 printf("\nAggregate statistics ==============================="
180 "\nTotal packets sent: %18"PRIu64
181 "\nTotal packets received: %14"PRIu64
182 "\nTotal packets dropped: %15"PRIu64,
185 total_packets_dropped);
186 printf("\n====================================================\n");
189 /* Send the burst of packets on an output interface */
191 l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
193 struct rte_mbuf **m_table;
197 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
199 ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
200 port_statistics[port].tx += ret;
201 if (unlikely(ret < n)) {
202 port_statistics[port].dropped += (n - ret);
204 rte_pktmbuf_free(m_table[ret]);
211 /* Enqueue packets for TX and prepare them to be sent */
213 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
215 unsigned lcore_id, len;
216 struct lcore_queue_conf *qconf;
218 lcore_id = rte_lcore_id();
220 qconf = &lcore_queue_conf[lcore_id];
221 len = qconf->tx_mbufs[port].len;
222 qconf->tx_mbufs[port].m_table[len] = m;
225 /* enough pkts to be sent */
226 if (unlikely(len == MAX_PKT_BURST)) {
227 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
231 qconf->tx_mbufs[port].len = len;
236 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
238 struct ether_hdr *eth;
242 dst_port = l2fwd_dst_ports[portid];
243 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
245 /* 02:00:00:00:00:xx */
246 tmp = ð->d_addr.addr_bytes[0];
247 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
250 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
252 l2fwd_send_packet(m, (uint8_t) dst_port);
255 /* main processing loop */
257 l2fwd_main_loop(void)
259 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
262 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
263 unsigned i, j, portid, nb_rx;
264 struct lcore_queue_conf *qconf;
265 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
270 lcore_id = rte_lcore_id();
271 qconf = &lcore_queue_conf[lcore_id];
273 if (qconf->n_rx_port == 0) {
274 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
278 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
280 for (i = 0; i < qconf->n_rx_port; i++) {
282 portid = qconf->rx_port_list[i];
283 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
289 cur_tsc = rte_rdtsc();
292 * TX burst queue drain
294 diff_tsc = cur_tsc - prev_tsc;
295 if (unlikely(diff_tsc > drain_tsc)) {
297 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
298 if (qconf->tx_mbufs[portid].len == 0)
300 l2fwd_send_burst(&lcore_queue_conf[lcore_id],
301 qconf->tx_mbufs[portid].len,
303 qconf->tx_mbufs[portid].len = 0;
306 /* if timer is enabled */
307 if (timer_period > 0) {
309 /* advance the timer */
310 timer_tsc += diff_tsc;
312 /* if timer has reached its timeout */
313 if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
315 /* do this only on master core */
316 if (lcore_id == rte_get_master_lcore()) {
318 /* reset the timer */
328 * Read packet from RX queues
330 for (i = 0; i < qconf->n_rx_port; i++) {
332 portid = qconf->rx_port_list[i];
333 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
334 pkts_burst, MAX_PKT_BURST);
336 port_statistics[portid].rx += nb_rx;
338 for (j = 0; j < nb_rx; j++) {
340 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
341 l2fwd_simple_forward(m, portid);
348 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
356 l2fwd_usage(const char *prgname)
358 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
359 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
360 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
361 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
366 l2fwd_parse_portmask(const char *portmask)
371 /* parse hexadecimal string */
372 pm = strtoul(portmask, &end, 16);
373 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
383 l2fwd_parse_nqueue(const char *q_arg)
388 /* parse hexadecimal string */
389 n = strtoul(q_arg, &end, 10);
390 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
394 if (n >= MAX_RX_QUEUE_PER_LCORE)
401 l2fwd_parse_timer_period(const char *q_arg)
406 /* parse number string */
407 n = strtol(q_arg, &end, 10);
408 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
410 if (n >= MAX_TIMER_PERIOD)
416 /* Parse the argument given in the command line of the application */
418 l2fwd_parse_args(int argc, char **argv)
423 char *prgname = argv[0];
424 static struct option lgopts[] = {
430 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
431 lgopts, &option_index)) != EOF) {
436 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
437 if (l2fwd_enabled_port_mask == 0) {
438 printf("invalid portmask\n");
439 l2fwd_usage(prgname);
446 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
447 if (l2fwd_rx_queue_per_lcore == 0) {
448 printf("invalid queue number\n");
449 l2fwd_usage(prgname);
456 timer_period = l2fwd_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
457 if (timer_period < 0) {
458 printf("invalid timer period\n");
459 l2fwd_usage(prgname);
466 l2fwd_usage(prgname);
470 l2fwd_usage(prgname);
476 argv[optind-1] = prgname;
479 optind = 0; /* reset getopt lib */
483 /* Check the link status of all ports in up to 9s, and print them finally */
485 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
487 #define CHECK_INTERVAL 100 /* 100ms */
488 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
489 uint8_t portid, count, all_ports_up, print_flag = 0;
490 struct rte_eth_link link;
492 printf("\nChecking link status");
494 for (count = 0; count <= MAX_CHECK_TIME; count++) {
496 for (portid = 0; portid < port_num; portid++) {
497 if ((port_mask & (1 << portid)) == 0)
499 memset(&link, 0, sizeof(link));
500 rte_eth_link_get_nowait(portid, &link);
501 /* print link status if flag set */
502 if (print_flag == 1) {
503 if (link.link_status)
504 printf("Port %d Link Up - speed %u "
505 "Mbps - %s\n", (uint8_t)portid,
506 (unsigned)link.link_speed,
507 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
508 ("full-duplex") : ("half-duplex\n"));
510 printf("Port %d Link Down\n",
514 /* clear all_ports_up flag if any link down */
515 if (link.link_status == 0) {
520 /* after finally printing all link status, get out */
524 if (all_ports_up == 0) {
527 rte_delay_ms(CHECK_INTERVAL);
530 /* set the print_flag if all ports up or timeout */
531 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
539 main(int argc, char **argv)
541 struct lcore_queue_conf *qconf;
542 struct rte_eth_dev_info dev_info;
545 uint8_t nb_ports_available;
546 uint8_t portid, last_port;
547 unsigned lcore_id, rx_lcore_id;
548 unsigned nb_ports_in_mask = 0;
551 ret = rte_eal_init(argc, argv);
553 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
557 /* parse application arguments (after the EAL ones) */
558 ret = l2fwd_parse_args(argc, argv);
560 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
562 /* create the mbuf pool */
564 rte_mempool_create("mbuf_pool", NB_MBUF,
566 sizeof(struct rte_pktmbuf_pool_private),
567 rte_pktmbuf_pool_init, NULL,
568 rte_pktmbuf_init, NULL,
570 if (l2fwd_pktmbuf_pool == NULL)
571 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
573 nb_ports = rte_eth_dev_count();
575 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
577 if (nb_ports > RTE_MAX_ETHPORTS)
578 nb_ports = RTE_MAX_ETHPORTS;
580 /* reset l2fwd_dst_ports */
581 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
582 l2fwd_dst_ports[portid] = 0;
586 * Each logical core is assigned a dedicated TX queue on each port.
588 for (portid = 0; portid < nb_ports; portid++) {
589 /* skip ports that are not enabled */
590 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
593 if (nb_ports_in_mask % 2) {
594 l2fwd_dst_ports[portid] = last_port;
595 l2fwd_dst_ports[last_port] = portid;
602 rte_eth_dev_info_get(portid, &dev_info);
604 if (nb_ports_in_mask % 2) {
605 printf("Notice: odd number of ports in portmask.\n");
606 l2fwd_dst_ports[last_port] = last_port;
612 /* Initialize the port/queue configuration of each logical core */
613 for (portid = 0; portid < nb_ports; portid++) {
614 /* skip ports that are not enabled */
615 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
618 /* get the lcore_id for this port */
619 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
620 lcore_queue_conf[rx_lcore_id].n_rx_port ==
621 l2fwd_rx_queue_per_lcore) {
623 if (rx_lcore_id >= RTE_MAX_LCORE)
624 rte_exit(EXIT_FAILURE, "Not enough cores\n");
627 if (qconf != &lcore_queue_conf[rx_lcore_id])
628 /* Assigned a new logical core in the loop above. */
629 qconf = &lcore_queue_conf[rx_lcore_id];
631 qconf->rx_port_list[qconf->n_rx_port] = portid;
633 printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
636 nb_ports_available = nb_ports;
638 /* Initialise each port */
639 for (portid = 0; portid < nb_ports; portid++) {
640 /* skip ports that are not enabled */
641 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
642 printf("Skipping disabled port %u\n", (unsigned) portid);
643 nb_ports_available--;
647 printf("Initializing port %u... ", (unsigned) portid);
649 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
651 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
652 ret, (unsigned) portid);
654 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
656 /* init one RX queue */
658 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
659 rte_eth_dev_socket_id(portid),
663 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
664 ret, (unsigned) portid);
666 /* init one TX queue on each port */
668 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
669 rte_eth_dev_socket_id(portid),
672 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
673 ret, (unsigned) portid);
676 ret = rte_eth_dev_start(portid);
678 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
679 ret, (unsigned) portid);
683 rte_eth_promiscuous_enable(portid);
685 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
687 l2fwd_ports_eth_addr[portid].addr_bytes[0],
688 l2fwd_ports_eth_addr[portid].addr_bytes[1],
689 l2fwd_ports_eth_addr[portid].addr_bytes[2],
690 l2fwd_ports_eth_addr[portid].addr_bytes[3],
691 l2fwd_ports_eth_addr[portid].addr_bytes[4],
692 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
694 /* initialize port stats */
695 memset(&port_statistics, 0, sizeof(port_statistics));
698 if (!nb_ports_available) {
699 rte_exit(EXIT_FAILURE,
700 "All available ports are disabled. Please set portmask.\n");
703 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
705 /* launch per-lcore init on every lcore */
706 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
707 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
708 if (rte_eal_wait_lcore(lcore_id) < 0)