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>
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>
64 #include <rte_random.h>
65 #include <rte_debug.h>
66 #include <rte_ether.h>
67 #include <rte_ethdev.h>
68 #include <rte_mempool.h>
71 static volatile bool force_quit;
73 /* MAC updating enabled by default */
74 static int mac_updating = 1;
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 = 1, /**< 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_mac_updating(struct rte_mbuf *m, unsigned dest_portid)
190 struct ether_hdr *eth;
193 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
195 /* 02:00:00:00:00:xx */
196 tmp = ð->d_addr.addr_bytes[0];
197 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dest_portid << 40);
200 ether_addr_copy(&l2fwd_ports_eth_addr[dest_portid], ð->s_addr);
204 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
208 struct rte_eth_dev_tx_buffer *buffer;
210 dst_port = l2fwd_dst_ports[portid];
213 l2fwd_mac_updating(m, dst_port);
215 buffer = tx_buffer[dst_port];
216 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
218 port_statistics[dst_port].tx += sent;
221 /* main processing loop */
223 l2fwd_main_loop(void)
225 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
229 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
230 unsigned i, j, portid, nb_rx;
231 struct lcore_queue_conf *qconf;
232 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S *
234 struct rte_eth_dev_tx_buffer *buffer;
239 lcore_id = rte_lcore_id();
240 qconf = &lcore_queue_conf[lcore_id];
242 if (qconf->n_rx_port == 0) {
243 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
247 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
249 for (i = 0; i < qconf->n_rx_port; i++) {
251 portid = qconf->rx_port_list[i];
252 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
257 while (!force_quit) {
259 cur_tsc = rte_rdtsc();
262 * TX burst queue drain
264 diff_tsc = cur_tsc - prev_tsc;
265 if (unlikely(diff_tsc > drain_tsc)) {
267 for (i = 0; i < qconf->n_rx_port; i++) {
269 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
270 buffer = tx_buffer[portid];
272 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
274 port_statistics[portid].tx += sent;
278 /* if timer is enabled */
279 if (timer_period > 0) {
281 /* advance the timer */
282 timer_tsc += diff_tsc;
284 /* if timer has reached its timeout */
285 if (unlikely(timer_tsc >= timer_period)) {
287 /* do this only on master core */
288 if (lcore_id == rte_get_master_lcore()) {
290 /* reset the timer */
300 * Read packet from RX queues
302 for (i = 0; i < qconf->n_rx_port; i++) {
304 portid = qconf->rx_port_list[i];
305 nb_rx = rte_eth_rx_burst(portid, 0,
306 pkts_burst, MAX_PKT_BURST);
308 port_statistics[portid].rx += nb_rx;
310 for (j = 0; j < nb_rx; j++) {
312 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
313 l2fwd_simple_forward(m, portid);
320 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
328 l2fwd_usage(const char *prgname)
330 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
331 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
332 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
333 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n"
334 " --[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default)\n"
336 " - The source MAC address is replaced by the TX port MAC address\n"
337 " - The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID\n",
342 l2fwd_parse_portmask(const char *portmask)
347 /* parse hexadecimal string */
348 pm = strtoul(portmask, &end, 16);
349 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
359 l2fwd_parse_nqueue(const char *q_arg)
364 /* parse hexadecimal string */
365 n = strtoul(q_arg, &end, 10);
366 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
370 if (n >= MAX_RX_QUEUE_PER_LCORE)
377 l2fwd_parse_timer_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 static const char short_options[] =
394 "q:" /* number of queues */
395 "T:" /* timer period */
398 #define CMD_LINE_OPT_MAC_UPDATING "mac-updating"
399 #define CMD_LINE_OPT_NO_MAC_UPDATING "no-mac-updating"
402 /* long options mapped to a short option */
404 /* first long only option value must be >= 256, so that we won't
405 * conflict with short options */
406 CMD_LINE_OPT_MIN_NUM = 256,
409 static const struct option lgopts[] = {
410 { CMD_LINE_OPT_MAC_UPDATING, no_argument, &mac_updating, 1},
411 { CMD_LINE_OPT_NO_MAC_UPDATING, no_argument, &mac_updating, 0},
415 /* Parse the argument given in the command line of the application */
417 l2fwd_parse_args(int argc, char **argv)
419 int opt, ret, timer_secs;
422 char *prgname = argv[0];
426 while ((opt = getopt_long(argc, argvopt, short_options,
427 lgopts, &option_index)) != EOF) {
432 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
433 if (l2fwd_enabled_port_mask == 0) {
434 printf("invalid portmask\n");
435 l2fwd_usage(prgname);
442 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
443 if (l2fwd_rx_queue_per_lcore == 0) {
444 printf("invalid queue number\n");
445 l2fwd_usage(prgname);
452 timer_secs = l2fwd_parse_timer_period(optarg);
453 if (timer_secs < 0) {
454 printf("invalid timer period\n");
455 l2fwd_usage(prgname);
458 timer_period = timer_secs;
466 l2fwd_usage(prgname);
472 argv[optind-1] = prgname;
475 optind = 1; /* reset getopt lib */
479 /* Check the link status of all ports in up to 9s, and print them finally */
481 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
483 #define CHECK_INTERVAL 100 /* 100ms */
484 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
486 uint8_t count, all_ports_up, print_flag = 0;
487 struct rte_eth_link link;
489 printf("\nChecking link status");
491 for (count = 0; count <= MAX_CHECK_TIME; count++) {
495 for (portid = 0; portid < port_num; portid++) {
498 if ((port_mask & (1 << portid)) == 0)
500 memset(&link, 0, sizeof(link));
501 rte_eth_link_get_nowait(portid, &link);
502 /* print link status if flag set */
503 if (print_flag == 1) {
504 if (link.link_status)
506 "Port%d Link Up. Speed %u Mbps - %s\n",
507 portid, link.link_speed,
508 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
509 ("full-duplex") : ("half-duplex\n"));
511 printf("Port %d Link Down\n", portid);
514 /* clear all_ports_up flag if any link down */
515 if (link.link_status == ETH_LINK_DOWN) {
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 signal_handler(int signum)
541 if (signum == SIGINT || signum == SIGTERM) {
542 printf("\n\nSignal %d received, preparing to exit...\n",
549 main(int argc, char **argv)
551 struct lcore_queue_conf *qconf;
552 struct rte_eth_dev_info dev_info;
555 uint16_t nb_ports_available;
556 uint16_t portid, last_port;
557 unsigned lcore_id, rx_lcore_id;
558 unsigned nb_ports_in_mask = 0;
561 ret = rte_eal_init(argc, argv);
563 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
568 signal(SIGINT, signal_handler);
569 signal(SIGTERM, signal_handler);
571 /* parse application arguments (after the EAL ones) */
572 ret = l2fwd_parse_args(argc, argv);
574 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
576 printf("MAC updating %s\n", mac_updating ? "enabled" : "disabled");
578 /* convert to number of cycles */
579 timer_period *= rte_get_timer_hz();
581 /* create the mbuf pool */
582 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
583 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
585 if (l2fwd_pktmbuf_pool == NULL)
586 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
588 nb_ports = rte_eth_dev_count();
590 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
592 /* reset l2fwd_dst_ports */
593 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
594 l2fwd_dst_ports[portid] = 0;
598 * Each logical core is assigned a dedicated TX queue on each port.
600 for (portid = 0; portid < nb_ports; portid++) {
601 /* skip ports that are not enabled */
602 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
605 if (nb_ports_in_mask % 2) {
606 l2fwd_dst_ports[portid] = last_port;
607 l2fwd_dst_ports[last_port] = portid;
614 rte_eth_dev_info_get(portid, &dev_info);
616 if (nb_ports_in_mask % 2) {
617 printf("Notice: odd number of ports in portmask.\n");
618 l2fwd_dst_ports[last_port] = last_port;
624 /* Initialize the port/queue configuration of each logical core */
625 for (portid = 0; portid < nb_ports; portid++) {
626 /* skip ports that are not enabled */
627 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
630 /* get the lcore_id for this port */
631 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
632 lcore_queue_conf[rx_lcore_id].n_rx_port ==
633 l2fwd_rx_queue_per_lcore) {
635 if (rx_lcore_id >= RTE_MAX_LCORE)
636 rte_exit(EXIT_FAILURE, "Not enough cores\n");
639 if (qconf != &lcore_queue_conf[rx_lcore_id])
640 /* Assigned a new logical core in the loop above. */
641 qconf = &lcore_queue_conf[rx_lcore_id];
643 qconf->rx_port_list[qconf->n_rx_port] = portid;
645 printf("Lcore %u: RX port %u\n", rx_lcore_id, portid);
648 nb_ports_available = nb_ports;
650 /* Initialise each port */
651 for (portid = 0; portid < nb_ports; portid++) {
652 /* skip ports that are not enabled */
653 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
654 printf("Skipping disabled port %u\n", portid);
655 nb_ports_available--;
659 printf("Initializing port %u... ", portid);
661 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
663 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
666 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
669 rte_exit(EXIT_FAILURE,
670 "Cannot adjust number of descriptors: err=%d, port=%u\n",
673 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
675 /* init one RX queue */
677 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
678 rte_eth_dev_socket_id(portid),
682 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
685 /* init one TX queue on each port */
687 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
688 rte_eth_dev_socket_id(portid),
691 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
694 /* Initialize TX buffers */
695 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
696 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
697 rte_eth_dev_socket_id(portid));
698 if (tx_buffer[portid] == NULL)
699 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
702 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
704 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
705 rte_eth_tx_buffer_count_callback,
706 &port_statistics[portid].dropped);
708 rte_exit(EXIT_FAILURE,
709 "Cannot set error callback for tx buffer on port %u\n",
713 ret = rte_eth_dev_start(portid);
715 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
720 rte_eth_promiscuous_enable(portid);
722 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
724 l2fwd_ports_eth_addr[portid].addr_bytes[0],
725 l2fwd_ports_eth_addr[portid].addr_bytes[1],
726 l2fwd_ports_eth_addr[portid].addr_bytes[2],
727 l2fwd_ports_eth_addr[portid].addr_bytes[3],
728 l2fwd_ports_eth_addr[portid].addr_bytes[4],
729 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
731 /* initialize port stats */
732 memset(&port_statistics, 0, sizeof(port_statistics));
735 if (!nb_ports_available) {
736 rte_exit(EXIT_FAILURE,
737 "All available ports are disabled. Please set portmask.\n");
740 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
743 /* launch per-lcore init on every lcore */
744 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
745 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
746 if (rte_eal_wait_lcore(lcore_id) < 0) {
752 for (portid = 0; portid < nb_ports; portid++) {
753 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
755 printf("Closing port %d...", portid);
756 rte_eth_dev_stop(portid);
757 rte_eth_dev_close(portid);