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>
65 #include <rte_random.h>
66 #include <rte_debug.h>
67 #include <rte_ether.h>
68 #include <rte_ethdev.h>
69 #include <rte_mempool.h>
72 static volatile bool force_quit;
74 /* MAC updating enabled by default */
75 static int mac_updating = 1;
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 = 1, /**< 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_mac_updating(struct rte_mbuf *m, unsigned dest_portid)
191 struct ether_hdr *eth;
194 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
196 /* 02:00:00:00:00:xx */
197 tmp = ð->d_addr.addr_bytes[0];
198 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dest_portid << 40);
201 ether_addr_copy(&l2fwd_ports_eth_addr[dest_portid], ð->s_addr);
205 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
209 struct rte_eth_dev_tx_buffer *buffer;
211 dst_port = l2fwd_dst_ports[portid];
214 l2fwd_mac_updating(m, dst_port);
216 buffer = tx_buffer[dst_port];
217 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
219 port_statistics[dst_port].tx += sent;
222 /* main processing loop */
224 l2fwd_main_loop(void)
226 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
230 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
231 unsigned i, j, portid, nb_rx;
232 struct lcore_queue_conf *qconf;
233 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S *
235 struct rte_eth_dev_tx_buffer *buffer;
240 lcore_id = rte_lcore_id();
241 qconf = &lcore_queue_conf[lcore_id];
243 if (qconf->n_rx_port == 0) {
244 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
248 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
250 for (i = 0; i < qconf->n_rx_port; i++) {
252 portid = qconf->rx_port_list[i];
253 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
258 while (!force_quit) {
260 cur_tsc = rte_rdtsc();
263 * TX burst queue drain
265 diff_tsc = cur_tsc - prev_tsc;
266 if (unlikely(diff_tsc > drain_tsc)) {
268 for (i = 0; i < qconf->n_rx_port; i++) {
270 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
271 buffer = tx_buffer[portid];
273 sent = rte_eth_tx_buffer_flush(portid, 0, buffer);
275 port_statistics[portid].tx += sent;
279 /* if timer is enabled */
280 if (timer_period > 0) {
282 /* advance the timer */
283 timer_tsc += diff_tsc;
285 /* if timer has reached its timeout */
286 if (unlikely(timer_tsc >= timer_period)) {
288 /* do this only on master core */
289 if (lcore_id == rte_get_master_lcore()) {
291 /* reset the timer */
301 * Read packet from RX queues
303 for (i = 0; i < qconf->n_rx_port; i++) {
305 portid = qconf->rx_port_list[i];
306 nb_rx = rte_eth_rx_burst(portid, 0,
307 pkts_burst, MAX_PKT_BURST);
309 port_statistics[portid].rx += nb_rx;
311 for (j = 0; j < nb_rx; j++) {
313 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
314 l2fwd_simple_forward(m, portid);
321 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
329 l2fwd_usage(const char *prgname)
331 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
332 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
333 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
334 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n"
335 " --[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default)\n"
337 " - The source MAC address is replaced by the TX port MAC address\n"
338 " - The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID\n",
343 l2fwd_parse_portmask(const char *portmask)
348 /* parse hexadecimal string */
349 pm = strtoul(portmask, &end, 16);
350 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
360 l2fwd_parse_nqueue(const char *q_arg)
365 /* parse hexadecimal string */
366 n = strtoul(q_arg, &end, 10);
367 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
371 if (n >= MAX_RX_QUEUE_PER_LCORE)
378 l2fwd_parse_timer_period(const char *q_arg)
383 /* parse number string */
384 n = strtol(q_arg, &end, 10);
385 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
387 if (n >= MAX_TIMER_PERIOD)
393 static const char short_options[] =
395 "q:" /* number of queues */
396 "T:" /* timer period */
399 #define CMD_LINE_OPT_MAC_UPDATING "mac-updating"
400 #define CMD_LINE_OPT_NO_MAC_UPDATING "no-mac-updating"
403 /* long options mapped to a short option */
405 /* first long only option value must be >= 256, so that we won't
406 * conflict with short options */
407 CMD_LINE_OPT_MIN_NUM = 256,
410 static const struct option lgopts[] = {
411 { CMD_LINE_OPT_MAC_UPDATING, no_argument, &mac_updating, 1},
412 { CMD_LINE_OPT_NO_MAC_UPDATING, no_argument, &mac_updating, 0},
416 /* Parse the argument given in the command line of the application */
418 l2fwd_parse_args(int argc, char **argv)
420 int opt, ret, timer_secs;
423 char *prgname = argv[0];
427 while ((opt = getopt_long(argc, argvopt, short_options,
428 lgopts, &option_index)) != EOF) {
433 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
434 if (l2fwd_enabled_port_mask == 0) {
435 printf("invalid portmask\n");
436 l2fwd_usage(prgname);
443 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
444 if (l2fwd_rx_queue_per_lcore == 0) {
445 printf("invalid queue number\n");
446 l2fwd_usage(prgname);
453 timer_secs = l2fwd_parse_timer_period(optarg);
454 if (timer_secs < 0) {
455 printf("invalid timer period\n");
456 l2fwd_usage(prgname);
459 timer_period = timer_secs;
467 l2fwd_usage(prgname);
473 argv[optind-1] = prgname;
476 optind = 1; /* reset getopt lib */
480 /* Check the link status of all ports in up to 9s, and print them finally */
482 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
484 #define CHECK_INTERVAL 100 /* 100ms */
485 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
487 uint8_t 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)
507 "Port%d Link Up. Speed %u Mbps - %s\n",
508 portid, link.link_speed,
509 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
510 ("full-duplex") : ("half-duplex\n"));
512 printf("Port %d Link Down\n", portid);
515 /* clear all_ports_up flag if any link down */
516 if (link.link_status == ETH_LINK_DOWN) {
521 /* after finally printing all link status, get out */
525 if (all_ports_up == 0) {
528 rte_delay_ms(CHECK_INTERVAL);
531 /* set the print_flag if all ports up or timeout */
532 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
540 signal_handler(int signum)
542 if (signum == SIGINT || signum == SIGTERM) {
543 printf("\n\nSignal %d received, preparing to exit...\n",
550 main(int argc, char **argv)
552 struct lcore_queue_conf *qconf;
553 struct rte_eth_dev_info dev_info;
556 uint16_t nb_ports_available;
557 uint16_t portid, last_port;
558 unsigned lcore_id, rx_lcore_id;
559 unsigned nb_ports_in_mask = 0;
562 ret = rte_eal_init(argc, argv);
564 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
569 signal(SIGINT, signal_handler);
570 signal(SIGTERM, signal_handler);
572 /* parse application arguments (after the EAL ones) */
573 ret = l2fwd_parse_args(argc, argv);
575 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
577 printf("MAC updating %s\n", mac_updating ? "enabled" : "disabled");
579 /* convert to number of cycles */
580 timer_period *= rte_get_timer_hz();
582 /* create the mbuf pool */
583 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
584 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
586 if (l2fwd_pktmbuf_pool == NULL)
587 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
589 nb_ports = rte_eth_dev_count();
591 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
593 /* reset l2fwd_dst_ports */
594 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
595 l2fwd_dst_ports[portid] = 0;
599 * Each logical core is assigned a dedicated TX queue on each port.
601 for (portid = 0; portid < nb_ports; portid++) {
602 /* skip ports that are not enabled */
603 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
606 if (nb_ports_in_mask % 2) {
607 l2fwd_dst_ports[portid] = last_port;
608 l2fwd_dst_ports[last_port] = portid;
615 rte_eth_dev_info_get(portid, &dev_info);
617 if (nb_ports_in_mask % 2) {
618 printf("Notice: odd number of ports in portmask.\n");
619 l2fwd_dst_ports[last_port] = last_port;
625 /* Initialize the port/queue configuration of each logical core */
626 for (portid = 0; portid < nb_ports; portid++) {
627 /* skip ports that are not enabled */
628 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
631 /* get the lcore_id for this port */
632 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
633 lcore_queue_conf[rx_lcore_id].n_rx_port ==
634 l2fwd_rx_queue_per_lcore) {
636 if (rx_lcore_id >= RTE_MAX_LCORE)
637 rte_exit(EXIT_FAILURE, "Not enough cores\n");
640 if (qconf != &lcore_queue_conf[rx_lcore_id])
641 /* Assigned a new logical core in the loop above. */
642 qconf = &lcore_queue_conf[rx_lcore_id];
644 qconf->rx_port_list[qconf->n_rx_port] = portid;
646 printf("Lcore %u: RX port %u\n", rx_lcore_id, portid);
649 nb_ports_available = nb_ports;
651 /* Initialise each port */
652 for (portid = 0; portid < nb_ports; portid++) {
653 /* skip ports that are not enabled */
654 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
655 printf("Skipping disabled port %u\n", portid);
656 nb_ports_available--;
660 printf("Initializing port %u... ", portid);
662 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
664 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
667 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
670 rte_exit(EXIT_FAILURE,
671 "Cannot adjust number of descriptors: err=%d, port=%u\n",
674 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
676 /* init one RX queue */
678 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
679 rte_eth_dev_socket_id(portid),
683 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
686 /* init one TX queue on each port */
688 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
689 rte_eth_dev_socket_id(portid),
692 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
695 /* Initialize TX buffers */
696 tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
697 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
698 rte_eth_dev_socket_id(portid));
699 if (tx_buffer[portid] == NULL)
700 rte_exit(EXIT_FAILURE, "Cannot allocate buffer for tx on port %u\n",
703 rte_eth_tx_buffer_init(tx_buffer[portid], MAX_PKT_BURST);
705 ret = rte_eth_tx_buffer_set_err_callback(tx_buffer[portid],
706 rte_eth_tx_buffer_count_callback,
707 &port_statistics[portid].dropped);
709 rte_exit(EXIT_FAILURE,
710 "Cannot set error callback for tx buffer on port %u\n",
714 ret = rte_eth_dev_start(portid);
716 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
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);