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_launch.h>
58 #include <rte_atomic.h>
59 #include <rte_cycles.h>
60 #include <rte_prefetch.h>
61 #include <rte_lcore.h>
62 #include <rte_per_lcore.h>
63 #include <rte_branch_prediction.h>
64 #include <rte_interrupts.h>
66 #include <rte_random.h>
67 #include <rte_debug.h>
68 #include <rte_ether.h>
69 #include <rte_ethdev.h>
70 #include <rte_mempool.h>
73 static volatile bool force_quit;
75 /* MAC updating enabled by default */
76 static int mac_updating = 1;
78 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
82 #define MAX_PKT_BURST 32
83 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
84 #define MEMPOOL_CACHE_SIZE 256
87 * Configurable number of RX/TX ring descriptors
89 #define RTE_TEST_RX_DESC_DEFAULT 128
90 #define RTE_TEST_TX_DESC_DEFAULT 512
91 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
92 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
94 /* ethernet addresses of ports */
95 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
97 /* mask of enabled ports */
98 static uint32_t l2fwd_enabled_port_mask = 0;
100 /* list of enabled ports */
101 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
103 static unsigned int l2fwd_rx_queue_per_lcore = 1;
105 #define MAX_RX_QUEUE_PER_LCORE 16
106 #define MAX_TX_QUEUE_PER_PORT 16
107 struct lcore_queue_conf {
109 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
110 } __rte_cache_aligned;
111 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
113 static struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
115 static const struct rte_eth_conf port_conf = {
118 .header_split = 0, /**< Header Split disabled */
119 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
120 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
121 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
122 .hw_strip_crc = 1, /**< CRC stripped by hardware */
125 .mq_mode = ETH_MQ_TX_NONE,
129 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
131 /* Per-port statistics struct */
132 struct l2fwd_port_statistics {
136 } __rte_cache_aligned;
137 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
139 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
140 /* A tsc-based timer responsible for triggering statistics printout */
141 static uint64_t timer_period = 10; /* 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");
190 l2fwd_mac_updating(struct rte_mbuf *m, unsigned dest_portid)
192 struct ether_hdr *eth;
195 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
197 /* 02:00:00:00:00:xx */
198 tmp = ð->d_addr.addr_bytes[0];
199 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dest_portid << 40);
202 ether_addr_copy(&l2fwd_ports_eth_addr[dest_portid], ð->s_addr);
206 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
210 struct rte_eth_dev_tx_buffer *buffer;
212 dst_port = l2fwd_dst_ports[portid];
215 l2fwd_mac_updating(m, dst_port);
217 buffer = tx_buffer[dst_port];
218 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
220 port_statistics[dst_port].tx += sent;
223 /* main processing loop */
225 l2fwd_main_loop(void)
227 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
231 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
232 unsigned i, j, portid, nb_rx;
233 struct lcore_queue_conf *qconf;
234 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S *
236 struct rte_eth_dev_tx_buffer *buffer;
241 lcore_id = rte_lcore_id();
242 qconf = &lcore_queue_conf[lcore_id];
244 if (qconf->n_rx_port == 0) {
245 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
249 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
251 for (i = 0; i < qconf->n_rx_port; i++) {
253 portid = qconf->rx_port_list[i];
254 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
259 while (!force_quit) {
261 cur_tsc = rte_rdtsc();
264 * TX burst queue drain
266 diff_tsc = cur_tsc - prev_tsc;
267 if (unlikely(diff_tsc > drain_tsc)) {
269 for (i = 0; i < qconf->n_rx_port; i++) {
271 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
272 buffer = tx_buffer[portid];
274 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
276 port_statistics[portid].tx += sent;
280 /* if timer is enabled */
281 if (timer_period > 0) {
283 /* advance the timer */
284 timer_tsc += diff_tsc;
286 /* if timer has reached its timeout */
287 if (unlikely(timer_tsc >= timer_period)) {
289 /* do this only on master core */
290 if (lcore_id == rte_get_master_lcore()) {
292 /* reset the timer */
302 * Read packet from RX queues
304 for (i = 0; i < qconf->n_rx_port; i++) {
306 portid = qconf->rx_port_list[i];
307 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
308 pkts_burst, MAX_PKT_BURST);
310 port_statistics[portid].rx += nb_rx;
312 for (j = 0; j < nb_rx; j++) {
314 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
315 l2fwd_simple_forward(m, portid);
322 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
330 l2fwd_usage(const char *prgname)
332 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
333 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
334 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
335 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n"
336 " --[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default)\n"
338 " - The source MAC address is replaced by the TX port MAC address\n"
339 " - The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID\n",
344 l2fwd_parse_portmask(const char *portmask)
349 /* parse hexadecimal string */
350 pm = strtoul(portmask, &end, 16);
351 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
361 l2fwd_parse_nqueue(const char *q_arg)
366 /* parse hexadecimal string */
367 n = strtoul(q_arg, &end, 10);
368 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
372 if (n >= MAX_RX_QUEUE_PER_LCORE)
379 l2fwd_parse_timer_period(const char *q_arg)
384 /* parse number string */
385 n = strtol(q_arg, &end, 10);
386 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
388 if (n >= MAX_TIMER_PERIOD)
394 static const char short_options[] =
396 "q:" /* number of queues */
397 "T:" /* timer period */
400 #define CMD_LINE_OPT_MAC_UPDATING "mac-updating"
401 #define CMD_LINE_OPT_NO_MAC_UPDATING "no-mac-updating"
404 /* long options mapped to a short option */
406 /* first long only option value must be >= 256, so that we won't
407 * conflict with short options */
408 CMD_LINE_OPT_MIN_NUM = 256,
411 static const struct option lgopts[] = {
412 { CMD_LINE_OPT_MAC_UPDATING, no_argument, &mac_updating, 1},
413 { CMD_LINE_OPT_NO_MAC_UPDATING, no_argument, &mac_updating, 0},
417 /* Parse the argument given in the command line of the application */
419 l2fwd_parse_args(int argc, char **argv)
421 int opt, ret, timer_secs;
424 char *prgname = argv[0];
428 while ((opt = getopt_long(argc, argvopt, short_options,
429 lgopts, &option_index)) != EOF) {
434 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
435 if (l2fwd_enabled_port_mask == 0) {
436 printf("invalid portmask\n");
437 l2fwd_usage(prgname);
444 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
445 if (l2fwd_rx_queue_per_lcore == 0) {
446 printf("invalid queue number\n");
447 l2fwd_usage(prgname);
454 timer_secs = l2fwd_parse_timer_period(optarg);
455 if (timer_secs < 0) {
456 printf("invalid timer period\n");
457 l2fwd_usage(prgname);
460 timer_period = timer_secs;
468 l2fwd_usage(prgname);
474 argv[optind-1] = prgname;
477 optind = 1; /* reset getopt lib */
481 /* Check the link status of all ports in up to 9s, and print them finally */
483 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
485 #define CHECK_INTERVAL 100 /* 100ms */
486 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
487 uint8_t portid, count, all_ports_up, print_flag = 0;
488 struct rte_eth_link link;
490 printf("\nChecking link status");
492 for (count = 0; count <= MAX_CHECK_TIME; count++) {
496 for (portid = 0; portid < port_num; portid++) {
499 if ((port_mask & (1 << portid)) == 0)
501 memset(&link, 0, sizeof(link));
502 rte_eth_link_get_nowait(portid, &link);
503 /* print link status if flag set */
504 if (print_flag == 1) {
505 if (link.link_status)
506 printf("Port %d Link Up - speed %u "
507 "Mbps - %s\n", (uint8_t)portid,
508 (unsigned)link.link_speed,
509 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
510 ("full-duplex") : ("half-duplex\n"));
512 printf("Port %d Link Down\n",
516 /* clear all_ports_up flag if any link down */
517 if (link.link_status == ETH_LINK_DOWN) {
522 /* after finally printing all link status, get out */
526 if (all_ports_up == 0) {
529 rte_delay_ms(CHECK_INTERVAL);
532 /* set the print_flag if all ports up or timeout */
533 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
541 signal_handler(int signum)
543 if (signum == SIGINT || signum == SIGTERM) {
544 printf("\n\nSignal %d received, preparing to exit...\n",
551 main(int argc, char **argv)
553 struct lcore_queue_conf *qconf;
554 struct rte_eth_dev_info dev_info;
557 uint8_t nb_ports_available;
558 uint8_t portid, last_port;
559 unsigned lcore_id, rx_lcore_id;
560 unsigned nb_ports_in_mask = 0;
563 ret = rte_eal_init(argc, argv);
565 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
570 signal(SIGINT, signal_handler);
571 signal(SIGTERM, signal_handler);
573 /* parse application arguments (after the EAL ones) */
574 ret = l2fwd_parse_args(argc, argv);
576 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
578 printf("MAC updating %s\n", mac_updating ? "enabled" : "disabled");
580 /* convert to number of cycles */
581 timer_period *= rte_get_timer_hz();
583 /* create the mbuf pool */
584 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
585 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
587 if (l2fwd_pktmbuf_pool == NULL)
588 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
590 nb_ports = rte_eth_dev_count();
592 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
594 /* reset l2fwd_dst_ports */
595 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
596 l2fwd_dst_ports[portid] = 0;
600 * Each logical core is assigned a dedicated TX queue on each port.
602 for (portid = 0; portid < nb_ports; portid++) {
603 /* skip ports that are not enabled */
604 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
607 if (nb_ports_in_mask % 2) {
608 l2fwd_dst_ports[portid] = last_port;
609 l2fwd_dst_ports[last_port] = portid;
616 rte_eth_dev_info_get(portid, &dev_info);
618 if (nb_ports_in_mask % 2) {
619 printf("Notice: odd number of ports in portmask.\n");
620 l2fwd_dst_ports[last_port] = last_port;
626 /* Initialize the port/queue configuration of each logical core */
627 for (portid = 0; portid < nb_ports; portid++) {
628 /* skip ports that are not enabled */
629 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
632 /* get the lcore_id for this port */
633 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
634 lcore_queue_conf[rx_lcore_id].n_rx_port ==
635 l2fwd_rx_queue_per_lcore) {
637 if (rx_lcore_id >= RTE_MAX_LCORE)
638 rte_exit(EXIT_FAILURE, "Not enough cores\n");
641 if (qconf != &lcore_queue_conf[rx_lcore_id])
642 /* Assigned a new logical core in the loop above. */
643 qconf = &lcore_queue_conf[rx_lcore_id];
645 qconf->rx_port_list[qconf->n_rx_port] = portid;
647 printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
650 nb_ports_available = nb_ports;
652 /* Initialise each port */
653 for (portid = 0; portid < nb_ports; portid++) {
654 /* skip ports that are not enabled */
655 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
656 printf("Skipping disabled port %u\n", (unsigned) portid);
657 nb_ports_available--;
661 printf("Initializing port %u... ", (unsigned) portid);
663 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
665 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
666 ret, (unsigned) portid);
668 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
671 rte_exit(EXIT_FAILURE,
672 "Cannot adjust number of descriptors: err=%d, port=%u\n",
673 ret, (unsigned) portid);
675 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
677 /* init one RX queue */
679 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
680 rte_eth_dev_socket_id(portid),
684 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
685 ret, (unsigned) portid);
687 /* init one TX queue on each port */
689 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
690 rte_eth_dev_socket_id(portid),
693 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
694 ret, (unsigned) portid);
696 /* Initialize TX buffers */
697 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
698 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
699 rte_eth_dev_socket_id(portid));
700 if (tx_buffer[portid] == NULL)
701 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
704 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
706 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
707 rte_eth_tx_buffer_count_callback,
708 &port_statistics[portid].dropped);
710 rte_exit(EXIT_FAILURE, "Cannot set error callback for "
711 "tx buffer on port %u\n", (unsigned) portid);
714 ret = rte_eth_dev_start(portid);
716 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
717 ret, (unsigned) portid);
721 rte_eth_promiscuous_enable(portid);
723 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
725 l2fwd_ports_eth_addr[portid].addr_bytes[0],
726 l2fwd_ports_eth_addr[portid].addr_bytes[1],
727 l2fwd_ports_eth_addr[portid].addr_bytes[2],
728 l2fwd_ports_eth_addr[portid].addr_bytes[3],
729 l2fwd_ports_eth_addr[portid].addr_bytes[4],
730 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
732 /* initialize port stats */
733 memset(&port_statistics, 0, sizeof(port_statistics));
736 if (!nb_ports_available) {
737 rte_exit(EXIT_FAILURE,
738 "All available ports are disabled. Please set portmask.\n");
741 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
744 /* launch per-lcore init on every lcore */
745 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
746 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
747 if (rte_eal_wait_lcore(lcore_id) < 0) {
753 for (portid = 0; portid < nb_ports; portid++) {
754 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
756 printf("Closing port %d...", portid);
757 rte_eth_dev_stop(portid);
758 rte_eth_dev_close(portid);