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>
50 #include <rte_common.h>
52 #include <rte_malloc.h>
53 #include <rte_memory.h>
54 #include <rte_memcpy.h>
55 #include <rte_memzone.h>
57 #include <rte_per_lcore.h>
58 #include <rte_launch.h>
59 #include <rte_atomic.h>
60 #include <rte_cycles.h>
61 #include <rte_prefetch.h>
62 #include <rte_lcore.h>
63 #include <rte_per_lcore.h>
64 #include <rte_branch_prediction.h>
65 #include <rte_interrupts.h>
67 #include <rte_random.h>
68 #include <rte_debug.h>
69 #include <rte_ether.h>
70 #include <rte_ethdev.h>
72 #include <rte_mempool.h>
75 static volatile bool force_quit;
77 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
81 #define MAX_PKT_BURST 32
82 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
83 #define MEMPOOL_CACHE_SIZE 256
86 * Configurable number of RX/TX ring descriptors
88 #define RTE_TEST_RX_DESC_DEFAULT 128
89 #define RTE_TEST_TX_DESC_DEFAULT 512
90 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
91 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
93 /* ethernet addresses of ports */
94 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
96 /* mask of enabled ports */
97 static uint32_t l2fwd_enabled_port_mask = 0;
99 /* list of enabled ports */
100 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
102 static unsigned int l2fwd_rx_queue_per_lcore = 1;
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 } __rte_cache_aligned;
110 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
112 static struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
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 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
139 /* A tsc-based timer responsible for triggering statistics printout */
140 static uint64_t timer_period = 10; /* default period is 10 seconds */
142 /* Print out statistics on packets dropped */
146 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
149 total_packets_dropped = 0;
150 total_packets_tx = 0;
151 total_packets_rx = 0;
153 const char clr[] = { 27, '[', '2', 'J', '\0' };
154 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
156 /* Clear screen and move to top left */
157 printf("%s%s", clr, topLeft);
159 printf("\nPort statistics ====================================");
161 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
162 /* skip disabled ports */
163 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
165 printf("\nStatistics for port %u ------------------------------"
166 "\nPackets sent: %24"PRIu64
167 "\nPackets received: %20"PRIu64
168 "\nPackets dropped: %21"PRIu64,
170 port_statistics[portid].tx,
171 port_statistics[portid].rx,
172 port_statistics[portid].dropped);
174 total_packets_dropped += port_statistics[portid].dropped;
175 total_packets_tx += port_statistics[portid].tx;
176 total_packets_rx += port_statistics[portid].rx;
178 printf("\nAggregate statistics ==============================="
179 "\nTotal packets sent: %18"PRIu64
180 "\nTotal packets received: %14"PRIu64
181 "\nTotal packets dropped: %15"PRIu64,
184 total_packets_dropped);
185 printf("\n====================================================\n");
189 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
191 struct ether_hdr *eth;
195 struct rte_eth_dev_tx_buffer *buffer;
197 dst_port = l2fwd_dst_ports[portid];
198 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
200 /* 02:00:00:00:00:xx */
201 tmp = ð->d_addr.addr_bytes[0];
202 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
205 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
207 buffer = tx_buffer[dst_port];
208 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
210 port_statistics[dst_port].tx += sent;
213 /* main processing loop */
215 l2fwd_main_loop(void)
217 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
221 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
222 unsigned i, j, portid, nb_rx;
223 struct lcore_queue_conf *qconf;
224 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S *
226 struct rte_eth_dev_tx_buffer *buffer;
231 lcore_id = rte_lcore_id();
232 qconf = &lcore_queue_conf[lcore_id];
234 if (qconf->n_rx_port == 0) {
235 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
239 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
241 for (i = 0; i < qconf->n_rx_port; i++) {
243 portid = qconf->rx_port_list[i];
244 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
249 while (!force_quit) {
251 cur_tsc = rte_rdtsc();
254 * TX burst queue drain
256 diff_tsc = cur_tsc - prev_tsc;
257 if (unlikely(diff_tsc > drain_tsc)) {
259 for (i = 0; i < qconf->n_rx_port; i++) {
261 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
262 buffer = tx_buffer[portid];
264 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
266 port_statistics[portid].tx += sent;
270 /* if timer is enabled */
271 if (timer_period > 0) {
273 /* advance the timer */
274 timer_tsc += diff_tsc;
276 /* if timer has reached its timeout */
277 if (unlikely(timer_tsc >= timer_period)) {
279 /* do this only on master core */
280 if (lcore_id == rte_get_master_lcore()) {
282 /* reset the timer */
292 * Read packet from RX queues
294 for (i = 0; i < qconf->n_rx_port; i++) {
296 portid = qconf->rx_port_list[i];
297 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
298 pkts_burst, MAX_PKT_BURST);
300 port_statistics[portid].rx += nb_rx;
302 for (j = 0; j < nb_rx; j++) {
304 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
305 l2fwd_simple_forward(m, portid);
312 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
320 l2fwd_usage(const char *prgname)
322 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
323 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
324 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
325 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
330 l2fwd_parse_portmask(const char *portmask)
335 /* parse hexadecimal string */
336 pm = strtoul(portmask, &end, 16);
337 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
347 l2fwd_parse_nqueue(const char *q_arg)
352 /* parse hexadecimal string */
353 n = strtoul(q_arg, &end, 10);
354 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
358 if (n >= MAX_RX_QUEUE_PER_LCORE)
365 l2fwd_parse_timer_period(const char *q_arg)
370 /* parse number string */
371 n = strtol(q_arg, &end, 10);
372 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
374 if (n >= MAX_TIMER_PERIOD)
380 /* Parse the argument given in the command line of the application */
382 l2fwd_parse_args(int argc, char **argv)
384 int opt, ret, timer_secs;
387 char *prgname = argv[0];
388 static struct option lgopts[] = {
394 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
395 lgopts, &option_index)) != EOF) {
400 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
401 if (l2fwd_enabled_port_mask == 0) {
402 printf("invalid portmask\n");
403 l2fwd_usage(prgname);
410 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
411 if (l2fwd_rx_queue_per_lcore == 0) {
412 printf("invalid queue number\n");
413 l2fwd_usage(prgname);
420 timer_secs = l2fwd_parse_timer_period(optarg);
421 if (timer_secs < 0) {
422 printf("invalid timer period\n");
423 l2fwd_usage(prgname);
426 timer_period = timer_secs;
431 l2fwd_usage(prgname);
435 l2fwd_usage(prgname);
441 argv[optind-1] = prgname;
444 optind = 0; /* reset getopt lib */
448 /* Check the link status of all ports in up to 9s, and print them finally */
450 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
452 #define CHECK_INTERVAL 100 /* 100ms */
453 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
454 uint8_t portid, count, all_ports_up, print_flag = 0;
455 struct rte_eth_link link;
457 printf("\nChecking link status");
459 for (count = 0; count <= MAX_CHECK_TIME; count++) {
463 for (portid = 0; portid < port_num; portid++) {
466 if ((port_mask & (1 << portid)) == 0)
468 memset(&link, 0, sizeof(link));
469 rte_eth_link_get_nowait(portid, &link);
470 /* print link status if flag set */
471 if (print_flag == 1) {
472 if (link.link_status)
473 printf("Port %d Link Up - speed %u "
474 "Mbps - %s\n", (uint8_t)portid,
475 (unsigned)link.link_speed,
476 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
477 ("full-duplex") : ("half-duplex\n"));
479 printf("Port %d Link Down\n",
483 /* clear all_ports_up flag if any link down */
484 if (link.link_status == ETH_LINK_DOWN) {
489 /* after finally printing all link status, get out */
493 if (all_ports_up == 0) {
496 rte_delay_ms(CHECK_INTERVAL);
499 /* set the print_flag if all ports up or timeout */
500 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
508 signal_handler(int signum)
510 if (signum == SIGINT || signum == SIGTERM) {
511 printf("\n\nSignal %d received, preparing to exit...\n",
518 main(int argc, char **argv)
520 struct lcore_queue_conf *qconf;
521 struct rte_eth_dev_info dev_info;
524 uint8_t nb_ports_available;
525 uint8_t portid, last_port;
526 unsigned lcore_id, rx_lcore_id;
527 unsigned nb_ports_in_mask = 0;
530 ret = rte_eal_init(argc, argv);
532 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
537 signal(SIGINT, signal_handler);
538 signal(SIGTERM, signal_handler);
540 /* parse application arguments (after the EAL ones) */
541 ret = l2fwd_parse_args(argc, argv);
543 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
545 /* convert to number of cycles */
546 timer_period *= rte_get_timer_hz();
548 /* create the mbuf pool */
549 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
550 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
552 if (l2fwd_pktmbuf_pool == NULL)
553 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
555 nb_ports = rte_eth_dev_count();
557 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
559 /* reset l2fwd_dst_ports */
560 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
561 l2fwd_dst_ports[portid] = 0;
565 * Each logical core is assigned a dedicated TX queue on each port.
567 for (portid = 0; portid < nb_ports; portid++) {
568 /* skip ports that are not enabled */
569 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
572 if (nb_ports_in_mask % 2) {
573 l2fwd_dst_ports[portid] = last_port;
574 l2fwd_dst_ports[last_port] = portid;
581 rte_eth_dev_info_get(portid, &dev_info);
583 if (nb_ports_in_mask % 2) {
584 printf("Notice: odd number of ports in portmask.\n");
585 l2fwd_dst_ports[last_port] = last_port;
591 /* Initialize the port/queue configuration of each logical core */
592 for (portid = 0; portid < nb_ports; portid++) {
593 /* skip ports that are not enabled */
594 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
597 /* get the lcore_id for this port */
598 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
599 lcore_queue_conf[rx_lcore_id].n_rx_port ==
600 l2fwd_rx_queue_per_lcore) {
602 if (rx_lcore_id >= RTE_MAX_LCORE)
603 rte_exit(EXIT_FAILURE, "Not enough cores\n");
606 if (qconf != &lcore_queue_conf[rx_lcore_id])
607 /* Assigned a new logical core in the loop above. */
608 qconf = &lcore_queue_conf[rx_lcore_id];
610 qconf->rx_port_list[qconf->n_rx_port] = portid;
612 printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
615 nb_ports_available = nb_ports;
617 /* Initialise each port */
618 for (portid = 0; portid < nb_ports; portid++) {
619 /* skip ports that are not enabled */
620 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
621 printf("Skipping disabled port %u\n", (unsigned) portid);
622 nb_ports_available--;
626 printf("Initializing port %u... ", (unsigned) portid);
628 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
630 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
631 ret, (unsigned) portid);
633 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
635 /* init one RX queue */
637 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
638 rte_eth_dev_socket_id(portid),
642 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
643 ret, (unsigned) portid);
645 /* init one TX queue on each port */
647 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
648 rte_eth_dev_socket_id(portid),
651 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
652 ret, (unsigned) portid);
654 /* Initialize TX buffers */
655 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
656 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
657 rte_eth_dev_socket_id(portid));
658 if (tx_buffer[portid] == NULL)
659 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
662 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
664 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
665 rte_eth_tx_buffer_count_callback,
666 &port_statistics[portid].dropped);
668 rte_exit(EXIT_FAILURE, "Cannot set error callback for "
669 "tx buffer on port %u\n", (unsigned) portid);
672 ret = rte_eth_dev_start(portid);
674 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
675 ret, (unsigned) portid);
679 rte_eth_promiscuous_enable(portid);
681 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
683 l2fwd_ports_eth_addr[portid].addr_bytes[0],
684 l2fwd_ports_eth_addr[portid].addr_bytes[1],
685 l2fwd_ports_eth_addr[portid].addr_bytes[2],
686 l2fwd_ports_eth_addr[portid].addr_bytes[3],
687 l2fwd_ports_eth_addr[portid].addr_bytes[4],
688 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
690 /* initialize port stats */
691 memset(&port_statistics, 0, sizeof(port_statistics));
694 if (!nb_ports_available) {
695 rte_exit(EXIT_FAILURE,
696 "All available ports are disabled. Please set portmask.\n");
699 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
702 /* launch per-lcore init on every lcore */
703 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
704 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
705 if (rte_eal_wait_lcore(lcore_id) < 0) {
711 for (portid = 0; portid < nb_ports; portid++) {
712 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
714 printf("Closing port %d...", portid);
715 rte_eth_dev_stop(portid);
716 rte_eth_dev_close(portid);