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>
71 #include <rte_mempool.h>
74 static volatile bool force_quit;
76 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
80 #define MAX_PKT_BURST 32
81 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
82 #define MEMPOOL_CACHE_SIZE 256
85 * Configurable number of RX/TX ring descriptors
87 #define RTE_TEST_RX_DESC_DEFAULT 128
88 #define RTE_TEST_TX_DESC_DEFAULT 512
89 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
90 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
92 /* ethernet addresses of ports */
93 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
95 /* mask of enabled ports */
96 static uint32_t l2fwd_enabled_port_mask = 0;
98 /* list of enabled ports */
99 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
101 static unsigned int l2fwd_rx_queue_per_lcore = 1;
103 #define MAX_RX_QUEUE_PER_LCORE 16
104 #define MAX_TX_QUEUE_PER_PORT 16
105 struct lcore_queue_conf {
107 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
108 } __rte_cache_aligned;
109 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
111 static struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
113 static const struct rte_eth_conf port_conf = {
116 .header_split = 0, /**< Header Split disabled */
117 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
118 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
119 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
120 .hw_strip_crc = 0, /**< CRC stripped by hardware */
123 .mq_mode = ETH_MQ_TX_NONE,
127 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
129 /* Per-port statistics struct */
130 struct l2fwd_port_statistics {
134 } __rte_cache_aligned;
135 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
137 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
138 /* A tsc-based timer responsible for triggering statistics printout */
139 static uint64_t timer_period = 10; /* default period is 10 seconds */
141 /* Print out statistics on packets dropped */
145 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
148 total_packets_dropped = 0;
149 total_packets_tx = 0;
150 total_packets_rx = 0;
152 const char clr[] = { 27, '[', '2', 'J', '\0' };
153 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
155 /* Clear screen and move to top left */
156 printf("%s%s", clr, topLeft);
158 printf("\nPort statistics ====================================");
160 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
161 /* skip disabled ports */
162 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
164 printf("\nStatistics for port %u ------------------------------"
165 "\nPackets sent: %24"PRIu64
166 "\nPackets received: %20"PRIu64
167 "\nPackets dropped: %21"PRIu64,
169 port_statistics[portid].tx,
170 port_statistics[portid].rx,
171 port_statistics[portid].dropped);
173 total_packets_dropped += port_statistics[portid].dropped;
174 total_packets_tx += port_statistics[portid].tx;
175 total_packets_rx += port_statistics[portid].rx;
177 printf("\nAggregate statistics ==============================="
178 "\nTotal packets sent: %18"PRIu64
179 "\nTotal packets received: %14"PRIu64
180 "\nTotal packets dropped: %15"PRIu64,
183 total_packets_dropped);
184 printf("\n====================================================\n");
188 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
190 struct ether_hdr *eth;
194 struct rte_eth_dev_tx_buffer *buffer;
196 dst_port = l2fwd_dst_ports[portid];
197 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
199 /* 02:00:00:00:00:xx */
200 tmp = ð->d_addr.addr_bytes[0];
201 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
204 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
206 buffer = tx_buffer[dst_port];
207 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
209 port_statistics[dst_port].tx += sent;
212 /* main processing loop */
214 l2fwd_main_loop(void)
216 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
220 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
221 unsigned i, j, portid, nb_rx;
222 struct lcore_queue_conf *qconf;
223 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S *
225 struct rte_eth_dev_tx_buffer *buffer;
230 lcore_id = rte_lcore_id();
231 qconf = &lcore_queue_conf[lcore_id];
233 if (qconf->n_rx_port == 0) {
234 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
238 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
240 for (i = 0; i < qconf->n_rx_port; i++) {
242 portid = qconf->rx_port_list[i];
243 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
248 while (!force_quit) {
250 cur_tsc = rte_rdtsc();
253 * TX burst queue drain
255 diff_tsc = cur_tsc - prev_tsc;
256 if (unlikely(diff_tsc > drain_tsc)) {
258 for (i = 0; i < qconf->n_rx_port; i++) {
260 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
261 buffer = tx_buffer[portid];
263 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
265 port_statistics[portid].tx += sent;
269 /* if timer is enabled */
270 if (timer_period > 0) {
272 /* advance the timer */
273 timer_tsc += diff_tsc;
275 /* if timer has reached its timeout */
276 if (unlikely(timer_tsc >= timer_period)) {
278 /* do this only on master core */
279 if (lcore_id == rte_get_master_lcore()) {
281 /* reset the timer */
291 * Read packet from RX queues
293 for (i = 0; i < qconf->n_rx_port; i++) {
295 portid = qconf->rx_port_list[i];
296 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
297 pkts_burst, MAX_PKT_BURST);
299 port_statistics[portid].rx += nb_rx;
301 for (j = 0; j < nb_rx; j++) {
303 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
304 l2fwd_simple_forward(m, portid);
311 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
319 l2fwd_usage(const char *prgname)
321 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
322 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
323 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
324 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
329 l2fwd_parse_portmask(const char *portmask)
334 /* parse hexadecimal string */
335 pm = strtoul(portmask, &end, 16);
336 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
346 l2fwd_parse_nqueue(const char *q_arg)
351 /* parse hexadecimal string */
352 n = strtoul(q_arg, &end, 10);
353 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
357 if (n >= MAX_RX_QUEUE_PER_LCORE)
364 l2fwd_parse_timer_period(const char *q_arg)
369 /* parse number string */
370 n = strtol(q_arg, &end, 10);
371 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
373 if (n >= MAX_TIMER_PERIOD)
379 /* Parse the argument given in the command line of the application */
381 l2fwd_parse_args(int argc, char **argv)
383 int opt, ret, timer_secs;
386 char *prgname = argv[0];
387 static struct option lgopts[] = {
393 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
394 lgopts, &option_index)) != EOF) {
399 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
400 if (l2fwd_enabled_port_mask == 0) {
401 printf("invalid portmask\n");
402 l2fwd_usage(prgname);
409 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
410 if (l2fwd_rx_queue_per_lcore == 0) {
411 printf("invalid queue number\n");
412 l2fwd_usage(prgname);
419 timer_secs = l2fwd_parse_timer_period(optarg);
420 if (timer_secs < 0) {
421 printf("invalid timer period\n");
422 l2fwd_usage(prgname);
425 timer_period = timer_secs;
430 l2fwd_usage(prgname);
434 l2fwd_usage(prgname);
440 argv[optind-1] = prgname;
443 optind = 0; /* reset getopt lib */
447 /* Check the link status of all ports in up to 9s, and print them finally */
449 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
451 #define CHECK_INTERVAL 100 /* 100ms */
452 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
453 uint8_t portid, count, all_ports_up, print_flag = 0;
454 struct rte_eth_link link;
456 printf("\nChecking link status");
458 for (count = 0; count <= MAX_CHECK_TIME; count++) {
462 for (portid = 0; portid < port_num; portid++) {
465 if ((port_mask & (1 << portid)) == 0)
467 memset(&link, 0, sizeof(link));
468 rte_eth_link_get_nowait(portid, &link);
469 /* print link status if flag set */
470 if (print_flag == 1) {
471 if (link.link_status)
472 printf("Port %d Link Up - speed %u "
473 "Mbps - %s\n", (uint8_t)portid,
474 (unsigned)link.link_speed,
475 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
476 ("full-duplex") : ("half-duplex\n"));
478 printf("Port %d Link Down\n",
482 /* clear all_ports_up flag if any link down */
483 if (link.link_status == ETH_LINK_DOWN) {
488 /* after finally printing all link status, get out */
492 if (all_ports_up == 0) {
495 rte_delay_ms(CHECK_INTERVAL);
498 /* set the print_flag if all ports up or timeout */
499 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
507 signal_handler(int signum)
509 if (signum == SIGINT || signum == SIGTERM) {
510 printf("\n\nSignal %d received, preparing to exit...\n",
517 main(int argc, char **argv)
519 struct lcore_queue_conf *qconf;
520 struct rte_eth_dev_info dev_info;
523 uint8_t nb_ports_available;
524 uint8_t portid, last_port;
525 unsigned lcore_id, rx_lcore_id;
526 unsigned nb_ports_in_mask = 0;
529 ret = rte_eal_init(argc, argv);
531 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
536 signal(SIGINT, signal_handler);
537 signal(SIGTERM, signal_handler);
539 /* parse application arguments (after the EAL ones) */
540 ret = l2fwd_parse_args(argc, argv);
542 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
544 /* convert to number of cycles */
545 timer_period *= rte_get_timer_hz();
547 /* create the mbuf pool */
548 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
549 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
551 if (l2fwd_pktmbuf_pool == NULL)
552 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
554 nb_ports = rte_eth_dev_count();
556 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
558 /* reset l2fwd_dst_ports */
559 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
560 l2fwd_dst_ports[portid] = 0;
564 * Each logical core is assigned a dedicated TX queue on each port.
566 for (portid = 0; portid < nb_ports; portid++) {
567 /* skip ports that are not enabled */
568 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
571 if (nb_ports_in_mask % 2) {
572 l2fwd_dst_ports[portid] = last_port;
573 l2fwd_dst_ports[last_port] = portid;
580 rte_eth_dev_info_get(portid, &dev_info);
582 if (nb_ports_in_mask % 2) {
583 printf("Notice: odd number of ports in portmask.\n");
584 l2fwd_dst_ports[last_port] = last_port;
590 /* Initialize the port/queue configuration of each logical core */
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 /* get the lcore_id for this port */
597 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
598 lcore_queue_conf[rx_lcore_id].n_rx_port ==
599 l2fwd_rx_queue_per_lcore) {
601 if (rx_lcore_id >= RTE_MAX_LCORE)
602 rte_exit(EXIT_FAILURE, "Not enough cores\n");
605 if (qconf != &lcore_queue_conf[rx_lcore_id])
606 /* Assigned a new logical core in the loop above. */
607 qconf = &lcore_queue_conf[rx_lcore_id];
609 qconf->rx_port_list[qconf->n_rx_port] = portid;
611 printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
614 nb_ports_available = nb_ports;
616 /* Initialise each port */
617 for (portid = 0; portid < nb_ports; portid++) {
618 /* skip ports that are not enabled */
619 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
620 printf("Skipping disabled port %u\n", (unsigned) portid);
621 nb_ports_available--;
625 printf("Initializing port %u... ", (unsigned) portid);
627 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
629 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
630 ret, (unsigned) portid);
632 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
634 /* init one RX queue */
636 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
637 rte_eth_dev_socket_id(portid),
641 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
642 ret, (unsigned) portid);
644 /* init one TX queue on each port */
646 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
647 rte_eth_dev_socket_id(portid),
650 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
651 ret, (unsigned) portid);
653 /* Initialize TX buffers */
654 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
655 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
656 rte_eth_dev_socket_id(portid));
657 if (tx_buffer[portid] == NULL)
658 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
661 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
663 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
664 rte_eth_tx_buffer_count_callback,
665 &port_statistics[portid].dropped);
667 rte_exit(EXIT_FAILURE, "Cannot set error callback for "
668 "tx buffer on port %u\n", (unsigned) portid);
671 ret = rte_eth_dev_start(portid);
673 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
674 ret, (unsigned) portid);
678 rte_eth_promiscuous_enable(portid);
680 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
682 l2fwd_ports_eth_addr[portid].addr_bytes[0],
683 l2fwd_ports_eth_addr[portid].addr_bytes[1],
684 l2fwd_ports_eth_addr[portid].addr_bytes[2],
685 l2fwd_ports_eth_addr[portid].addr_bytes[3],
686 l2fwd_ports_eth_addr[portid].addr_bytes[4],
687 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
689 /* initialize port stats */
690 memset(&port_statistics, 0, sizeof(port_statistics));
693 if (!nb_ports_available) {
694 rte_exit(EXIT_FAILURE,
695 "All available ports are disabled. Please set portmask.\n");
698 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
701 /* launch per-lcore init on every lcore */
702 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
703 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
704 if (rte_eal_wait_lcore(lcore_id) < 0) {
710 for (portid = 0; portid < nb_ports; portid++) {
711 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
713 printf("Closing port %d...", portid);
714 rte_eth_dev_stop(portid);
715 rte_eth_dev_close(portid);