4 * Copyright(c) 2010-2013 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>
48 #include <rte_common.h>
50 #include <rte_memory.h>
51 #include <rte_memcpy.h>
52 #include <rte_memzone.h>
53 #include <rte_tailq.h>
55 #include <rte_per_lcore.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>
70 #include <rte_mempool.h>
75 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
77 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
81 * RX and TX Prefetch, Host, and Write-back threshold values should be
82 * carefully set for optimal performance. Consult the network
83 * controller's datasheet and supporting DPDK documentation for guidance
84 * on how these parameters should be set.
86 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
87 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
88 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
91 * These default values are optimized for use with the Intel(R) 82599 10 GbE
92 * Controller and the DPDK ixgbe PMD. Consider using other values for other
93 * network controllers and/or network drivers.
95 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
96 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */
97 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */
99 #define MAX_PKT_BURST 32
100 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
103 * Configurable number of RX/TX ring descriptors
105 #define RTE_TEST_RX_DESC_DEFAULT 128
106 #define RTE_TEST_TX_DESC_DEFAULT 512
107 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
108 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
110 /* ethernet addresses of ports */
111 static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
113 /* mask of enabled ports */
114 static uint32_t l2fwd_enabled_port_mask = 0;
116 /* list of enabled ports */
117 static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
119 static unsigned int l2fwd_rx_queue_per_lcore = 1;
123 struct rte_mbuf *m_table[MAX_PKT_BURST];
126 #define MAX_RX_QUEUE_PER_LCORE 16
127 #define MAX_TX_QUEUE_PER_PORT 16
128 struct lcore_queue_conf {
130 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
131 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
133 } __rte_cache_aligned;
134 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
136 static const struct rte_eth_conf port_conf = {
139 .header_split = 0, /**< Header Split disabled */
140 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
141 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
142 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
143 .hw_strip_crc = 0, /**< CRC stripped by hardware */
146 .mq_mode = ETH_MQ_TX_NONE,
150 static const struct rte_eth_rxconf rx_conf = {
152 .pthresh = RX_PTHRESH,
153 .hthresh = RX_HTHRESH,
154 .wthresh = RX_WTHRESH,
158 static const struct rte_eth_txconf tx_conf = {
160 .pthresh = TX_PTHRESH,
161 .hthresh = TX_HTHRESH,
162 .wthresh = TX_WTHRESH,
164 .tx_free_thresh = 0, /* Use PMD default values */
165 .tx_rs_thresh = 0, /* Use PMD default values */
168 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
170 /* Per-port statistics struct */
171 struct l2fwd_port_statistics {
175 } __rte_cache_aligned;
176 struct l2fwd_port_statistics port_statistics[RTE_MAX_ETHPORTS];
178 /* A tsc-based timer responsible for triggering statistics printout */
179 #define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
180 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
181 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
183 /* Print out statistics on packets dropped */
187 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
190 total_packets_dropped = 0;
191 total_packets_tx = 0;
192 total_packets_rx = 0;
194 const char clr[] = { 27, '[', '2', 'J', '\0' };
195 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
197 /* Clear screen and move to top left */
198 printf("%s%s", clr, topLeft);
200 printf("\nPort statistics ====================================");
202 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
203 /* skip disabled ports */
204 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
206 printf("\nStatistics for port %u ------------------------------"
207 "\nPackets sent: %24"PRIu64
208 "\nPackets received: %20"PRIu64
209 "\nPackets dropped: %21"PRIu64,
211 port_statistics[portid].tx,
212 port_statistics[portid].rx,
213 port_statistics[portid].dropped);
215 total_packets_dropped += port_statistics[portid].dropped;
216 total_packets_tx += port_statistics[portid].tx;
217 total_packets_rx += port_statistics[portid].rx;
219 printf("\nAggregate statistics ==============================="
220 "\nTotal packets sent: %18"PRIu64
221 "\nTotal packets received: %14"PRIu64
222 "\nTotal packets dropped: %15"PRIu64,
225 total_packets_dropped);
226 printf("\n====================================================\n");
229 /* Send the burst of packets on an output interface */
231 l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
233 struct rte_mbuf **m_table;
237 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
239 ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
240 port_statistics[port].tx += ret;
241 if (unlikely(ret < n)) {
242 port_statistics[port].dropped += (n - ret);
244 rte_pktmbuf_free(m_table[ret]);
251 /* Enqueue packets for TX and prepare them to be sent */
253 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
255 unsigned lcore_id, len;
256 struct lcore_queue_conf *qconf;
258 lcore_id = rte_lcore_id();
260 qconf = &lcore_queue_conf[lcore_id];
261 len = qconf->tx_mbufs[port].len;
262 qconf->tx_mbufs[port].m_table[len] = m;
265 /* enough pkts to be sent */
266 if (unlikely(len == MAX_PKT_BURST)) {
267 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
271 qconf->tx_mbufs[port].len = len;
276 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
278 struct ether_hdr *eth;
282 dst_port = l2fwd_dst_ports[portid];
283 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
285 /* 02:00:00:00:00:xx */
286 tmp = ð->d_addr.addr_bytes[0];
287 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
290 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
292 l2fwd_send_packet(m, (uint8_t) dst_port);
295 /* main processing loop */
297 l2fwd_main_loop(void)
299 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
302 uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
303 unsigned i, j, portid, nb_rx;
304 struct lcore_queue_conf *qconf;
305 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
310 lcore_id = rte_lcore_id();
311 qconf = &lcore_queue_conf[lcore_id];
313 if (qconf->n_rx_port == 0) {
314 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
318 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
320 for (i = 0; i < qconf->n_rx_port; i++) {
322 portid = qconf->rx_port_list[i];
323 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
329 cur_tsc = rte_rdtsc();
332 * TX burst queue drain
334 diff_tsc = cur_tsc - prev_tsc;
335 if (unlikely(diff_tsc > drain_tsc)) {
337 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
338 if (qconf->tx_mbufs[portid].len == 0)
340 l2fwd_send_burst(&lcore_queue_conf[lcore_id],
341 qconf->tx_mbufs[portid].len,
343 qconf->tx_mbufs[portid].len = 0;
346 /* if timer is enabled */
347 if (timer_period > 0) {
349 /* advance the timer */
350 timer_tsc += diff_tsc;
352 /* if timer has reached its timeout */
353 if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
355 /* do this only on master core */
356 if (lcore_id == rte_get_master_lcore()) {
358 /* reset the timer */
368 * Read packet from RX queues
370 for (i = 0; i < qconf->n_rx_port; i++) {
372 portid = qconf->rx_port_list[i];
373 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
374 pkts_burst, MAX_PKT_BURST);
376 port_statistics[portid].rx += nb_rx;
378 for (j = 0; j < nb_rx; j++) {
380 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
381 l2fwd_simple_forward(m, portid);
388 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
396 l2fwd_usage(const char *prgname)
398 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
399 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
400 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
401 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
406 l2fwd_parse_portmask(const char *portmask)
411 /* parse hexadecimal string */
412 pm = strtoul(portmask, &end, 16);
413 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
423 l2fwd_parse_nqueue(const char *q_arg)
428 /* parse hexadecimal string */
429 n = strtoul(q_arg, &end, 10);
430 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
434 if (n >= MAX_RX_QUEUE_PER_LCORE)
441 l2fwd_parse_timer_period(const char *q_arg)
446 /* parse number string */
447 n = strtol(q_arg, &end, 10);
448 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
450 if (n >= MAX_TIMER_PERIOD)
456 /* Parse the argument given in the command line of the application */
458 l2fwd_parse_args(int argc, char **argv)
463 char *prgname = argv[0];
464 static struct option lgopts[] = {
470 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
471 lgopts, &option_index)) != EOF) {
476 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
477 if (l2fwd_enabled_port_mask == 0) {
478 printf("invalid portmask\n");
479 l2fwd_usage(prgname);
486 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
487 if (l2fwd_rx_queue_per_lcore == 0) {
488 printf("invalid queue number\n");
489 l2fwd_usage(prgname);
496 timer_period = l2fwd_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
497 if (timer_period < 0) {
498 printf("invalid timer period\n");
499 l2fwd_usage(prgname);
506 l2fwd_usage(prgname);
510 l2fwd_usage(prgname);
516 argv[optind-1] = prgname;
519 optind = 0; /* reset getopt lib */
523 /* Check the link status of all ports in up to 9s, and print them finally */
525 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
527 #define CHECK_INTERVAL 100 /* 100ms */
528 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
529 uint8_t portid, count, all_ports_up, print_flag = 0;
530 struct rte_eth_link link;
532 printf("\nChecking link status");
534 for (count = 0; count <= MAX_CHECK_TIME; count++) {
536 for (portid = 0; portid < port_num; portid++) {
537 if ((port_mask & (1 << portid)) == 0)
539 memset(&link, 0, sizeof(link));
540 rte_eth_link_get_nowait(portid, &link);
541 /* print link status if flag set */
542 if (print_flag == 1) {
543 if (link.link_status)
544 printf("Port %d Link Up - speed %u "
545 "Mbps - %s\n", (uint8_t)portid,
546 (unsigned)link.link_speed,
547 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
548 ("full-duplex") : ("half-duplex\n"));
550 printf("Port %d Link Down\n",
554 /* clear all_ports_up flag if any link down */
555 if (link.link_status == 0) {
560 /* after finally printing all link status, get out */
564 if (all_ports_up == 0) {
567 rte_delay_ms(CHECK_INTERVAL);
570 /* set the print_flag if all ports up or timeout */
571 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
579 MAIN(int argc, char **argv)
581 struct lcore_queue_conf *qconf;
582 struct rte_eth_dev_info dev_info;
585 uint8_t nb_ports_available;
586 uint8_t portid, last_port;
587 unsigned lcore_id, rx_lcore_id;
588 unsigned nb_ports_in_mask = 0;
591 ret = rte_eal_init(argc, argv);
593 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
597 /* parse application arguments (after the EAL ones) */
598 ret = l2fwd_parse_args(argc, argv);
600 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
602 /* create the mbuf pool */
604 rte_mempool_create("mbuf_pool", NB_MBUF,
606 sizeof(struct rte_pktmbuf_pool_private),
607 rte_pktmbuf_pool_init, NULL,
608 rte_pktmbuf_init, NULL,
610 if (l2fwd_pktmbuf_pool == NULL)
611 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
614 if (rte_pmd_init_all() < 0)
615 rte_exit(EXIT_FAILURE, "Cannot init pmd\n");
617 if (rte_eal_pci_probe() < 0)
618 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
620 nb_ports = rte_eth_dev_count();
622 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
624 if (nb_ports > RTE_MAX_ETHPORTS)
625 nb_ports = RTE_MAX_ETHPORTS;
627 /* reset l2fwd_dst_ports */
628 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
629 l2fwd_dst_ports[portid] = 0;
633 * Each logical core is assigned a dedicated TX queue on each port.
635 for (portid = 0; portid < nb_ports; portid++) {
636 /* skip ports that are not enabled */
637 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
640 if (nb_ports_in_mask % 2) {
641 l2fwd_dst_ports[portid] = last_port;
642 l2fwd_dst_ports[last_port] = portid;
649 rte_eth_dev_info_get(portid, &dev_info);
651 if (nb_ports_in_mask % 2) {
652 printf("Notice: odd number of ports in portmask.\n");
653 l2fwd_dst_ports[last_port] = last_port;
659 /* Initialize the port/queue configuration of each logical core */
660 for (portid = 0; portid < nb_ports; portid++) {
661 /* skip ports that are not enabled */
662 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
665 /* get the lcore_id for this port */
666 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
667 lcore_queue_conf[rx_lcore_id].n_rx_port ==
668 l2fwd_rx_queue_per_lcore) {
670 if (rx_lcore_id >= RTE_MAX_LCORE)
671 rte_exit(EXIT_FAILURE, "Not enough cores\n");
674 if (qconf != &lcore_queue_conf[rx_lcore_id])
675 /* Assigned a new logical core in the loop above. */
676 qconf = &lcore_queue_conf[rx_lcore_id];
678 qconf->rx_port_list[qconf->n_rx_port] = portid;
680 printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
683 nb_ports_available = nb_ports;
685 /* Initialise each port */
686 for (portid = 0; portid < nb_ports; portid++) {
687 /* skip ports that are not enabled */
688 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
689 printf("Skipping disabled port %u\n", (unsigned) portid);
690 nb_ports_available--;
694 printf("Initializing port %u... ", (unsigned) portid);
696 ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
698 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
699 ret, (unsigned) portid);
701 rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
703 /* init one RX queue */
705 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
706 rte_eth_dev_socket_id(portid), &rx_conf,
709 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
710 ret, (unsigned) portid);
712 /* init one TX queue on each port */
714 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
715 rte_eth_dev_socket_id(portid), &tx_conf);
717 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
718 ret, (unsigned) portid);
721 ret = rte_eth_dev_start(portid);
723 rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
724 ret, (unsigned) portid);
728 rte_eth_promiscuous_enable(portid);
730 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
732 l2fwd_ports_eth_addr[portid].addr_bytes[0],
733 l2fwd_ports_eth_addr[portid].addr_bytes[1],
734 l2fwd_ports_eth_addr[portid].addr_bytes[2],
735 l2fwd_ports_eth_addr[portid].addr_bytes[3],
736 l2fwd_ports_eth_addr[portid].addr_bytes[4],
737 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
739 /* initialize port stats */
740 memset(&port_statistics, 0, sizeof(port_statistics));
743 if (!nb_ports_available) {
744 rte_exit(EXIT_FAILURE,
745 "All available ports are disabled. Please set portmask.\n");
748 check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
750 /* launch per-lcore init on every lcore */
751 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
752 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
753 if (rte_eal_wait_lcore(lcore_id) < 0)