4 * Copyright(c) 2010-2012 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.
33 * version: DPDK.L.1.2.3-3
41 #include <sys/types.h>
43 #include <sys/queue.h>
44 #include <netinet/in.h>
51 #include <rte_common.h>
53 #include <rte_memory.h>
54 #include <rte_memcpy.h>
55 #include <rte_memzone.h>
56 #include <rte_tailq.h>
58 #include <rte_per_lcore.h>
59 #include <rte_launch.h>
60 #include <rte_atomic.h>
61 #include <rte_cycles.h>
62 #include <rte_prefetch.h>
63 #include <rte_lcore.h>
64 #include <rte_per_lcore.h>
65 #include <rte_branch_prediction.h>
66 #include <rte_interrupts.h>
68 #include <rte_random.h>
69 #include <rte_debug.h>
70 #include <rte_ether.h>
71 #include <rte_ethdev.h>
73 #include <rte_mempool.h>
78 #define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
80 #define L2FWD_MAX_PORTS 32
82 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
86 * RX and TX Prefetch, Host, and Write-back threshold values should be
87 * carefully set for optimal performance. Consult the network
88 * controller's datasheet and supporting DPDK documentation for guidance
89 * on how these parameters should be set.
91 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
92 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
93 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
96 * These default values are optimized for use with the Intel(R) 82599 10 GbE
97 * Controller and the DPDK ixgbe PMD. Consider using other values for other
98 * network controllers and/or network drivers.
100 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
101 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */
102 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */
104 #define MAX_PKT_BURST 32
105 #define BURST_TX_DRAIN 200000ULL /* around 100us at 2 Ghz */
110 * Configurable number of RX/TX ring descriptors
112 #define RTE_TEST_RX_DESC_DEFAULT 128
113 #define RTE_TEST_TX_DESC_DEFAULT 512
114 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
115 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
117 /* ethernet addresses of ports */
118 static struct ether_addr l2fwd_ports_eth_addr[L2FWD_MAX_PORTS];
120 /* mask of enabled ports */
121 static uint32_t l2fwd_enabled_port_mask = 0;
123 /* list of enabled ports */
124 static uint32_t l2fwd_dst_ports[L2FWD_MAX_PORTS];
126 static unsigned int l2fwd_rx_queue_per_lcore = 1;
128 #define MAX_PKT_BURST 32
131 struct rte_mbuf *m_table[MAX_PKT_BURST];
134 #define MAX_RX_QUEUE_PER_LCORE 16
136 /* Each VF(port) has one Rx/Tx queue (with queueid: 0) */
137 struct lcore_queue_conf {
140 unsigned rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
141 unsigned tx_queue_id;
142 struct mbuf_table tx_mbufs[L2FWD_MAX_PORTS];
144 } __rte_cache_aligned;
145 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
147 static const struct rte_eth_conf port_conf = {
150 .header_split = 0, /**< Header Split disabled */
151 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
152 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
153 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
154 .hw_strip_crc = 1, /**< CRC stripped by hardware */
160 static const struct rte_eth_rxconf rx_conf = {
162 .pthresh = RX_PTHRESH,
163 .hthresh = RX_HTHRESH,
164 .wthresh = RX_WTHRESH,
168 static const struct rte_eth_txconf tx_conf = {
170 .pthresh = TX_PTHRESH,
171 .hthresh = TX_HTHRESH,
172 .wthresh = TX_WTHRESH,
174 .tx_free_thresh = 0, /* Use PMD default values */
175 .tx_rs_thresh = 0, /* Use PMD default values */
178 struct rte_mempool * l2fwd_pktmbuf_pool = NULL;
180 /* Per-port statistics struct */
181 struct l2fwd_port_statistics {
185 } __rte_cache_aligned;
186 struct l2fwd_port_statistics port_statistics[L2FWD_MAX_PORTS];
188 /* A tsc-based timer responsible for triggering statistics printout */
189 #define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
190 #define MAX_TIMER_PERIOD 86400 /* 1 day max */
191 static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
193 /* Print out statistics on packets dropped */
197 uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
200 total_packets_dropped = 0;
201 total_packets_tx = 0;
202 total_packets_rx = 0;
204 const char clr[] = { 27, '[', '2', 'J', '\0' };
205 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
207 /* Clear screen and move to top left */
208 printf("%s%s", clr, topLeft);
210 printf("\nPort statistics ====================================");
212 for (portid = 0; portid < L2FWD_MAX_PORTS; portid++) {
213 /* skip ports that are not enabled */
214 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
217 printf("\nStatistics for port %u ------------------------------"
218 "\nPackets sent: %24"PRIu64
219 "\nPackets received: %20"PRIu64
220 "\nPackets dropped: %21"PRIu64,
222 port_statistics[portid].tx,
223 port_statistics[portid].rx,
224 port_statistics[portid].dropped);
226 total_packets_dropped += port_statistics[portid].dropped;
227 total_packets_tx += port_statistics[portid].tx;
228 total_packets_rx += port_statistics[portid].rx;
230 printf("\nAggregate statistics ==============================="
231 "\nTotal packets sent: %18"PRIu64
232 "\nTotal packets received: %14"PRIu64
233 "\nTotal packets dropped: %15"PRIu64,
236 total_packets_dropped);
237 printf("\n====================================================\n");
240 /* Send the packet on an output interface */
242 l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
244 struct rte_mbuf **m_table;
248 queueid = (uint16_t) qconf->tx_queue_id;
249 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
251 ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
253 port_statistics[port].tx += ret;
254 if (unlikely(ret < n)) {
255 port_statistics[port].dropped += (n - ret);
257 rte_pktmbuf_free(m_table[ret]);
264 /* Send the packet on an output interface */
266 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
268 unsigned lcore_id, len;
269 struct lcore_queue_conf *qconf;
271 lcore_id = rte_lcore_id();
273 qconf = &lcore_queue_conf[lcore_id];
274 len = qconf->tx_mbufs[port].len;
275 qconf->tx_mbufs[port].m_table[len] = m;
278 /* enough pkts to be sent */
279 if (unlikely(len == MAX_PKT_BURST)) {
280 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
284 qconf->tx_mbufs[port].len = len;
289 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
291 struct ether_hdr *eth;
295 dst_port = l2fwd_dst_ports[portid];
296 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
298 /* 00:09:c0:00:00:xx */
299 tmp = ð->d_addr.addr_bytes[0];
300 *((uint64_t *)tmp) = 0x000000c00900 + (dst_port << 24);
303 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
305 l2fwd_send_packet(m, (uint8_t) dst_port);
308 /* main processing loop */
310 l2fwd_main_loop(void)
312 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
315 unsigned int nb_ports;
316 uint64_t prev_tsc = 0;
317 uint64_t diff_tsc, cur_tsc, timer_tsc;
318 unsigned i, j, portid, nb_rx;
319 struct lcore_queue_conf *qconf;
323 lcore_id = rte_lcore_id();
324 qconf = &lcore_queue_conf[lcore_id];
326 nb_ports = rte_eth_dev_count();
327 if (nb_ports > L2FWD_MAX_PORTS)
328 nb_ports = L2FWD_MAX_PORTS;
330 if (qconf->n_rx_queue == 0) {
331 RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
335 RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
337 for (i = 0; i < qconf->n_rx_queue; i++) {
339 portid = qconf->rx_queue_list[i];
340 RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
346 cur_tsc = rte_rdtsc();
349 * TX burst queue drain
351 diff_tsc = cur_tsc - prev_tsc;
352 if (unlikely(diff_tsc > BURST_TX_DRAIN)) {
354 for (portid = 0; portid < nb_ports; portid++) {
355 if (qconf->tx_mbufs[portid].len == 0)
357 l2fwd_send_burst(&lcore_queue_conf[lcore_id],
358 qconf->tx_mbufs[portid].len,
360 qconf->tx_mbufs[portid].len = 0;
363 /* if timer is enabled */
364 if (timer_period > 0) {
366 /* advance the timer */
367 timer_tsc += diff_tsc;
369 /* if timer has reached its timeout */
370 if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
372 /* do this only on master core */
373 if (lcore_id == rte_get_master_lcore()) {
375 /* reset the timer */
384 * Read packet from RX queues
386 for (i = 0; i < qconf->n_rx_queue; i++) {
388 portid = qconf->rx_queue_list[i];
389 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
390 pkts_burst, MAX_PKT_BURST);
392 port_statistics[portid].rx += nb_rx;
394 for (j = 0; j < nb_rx; j++) {
396 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
397 l2fwd_simple_forward(m, portid);
404 l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
412 l2fwd_usage(const char *prgname)
414 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
415 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
416 " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
417 " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
422 l2fwd_parse_portmask(const char *portmask)
427 /* parse hexadecimal string */
428 pm = strtoul(portmask, &end, 16);
429 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
439 l2fwd_parse_nqueue(const char *q_arg)
444 /* parse hexadecimal string */
445 n = strtoul(q_arg, &end, 10);
446 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
450 if (n >= MAX_RX_QUEUE_PER_LCORE)
457 l2fwd_parse_timer_period(const char *q_arg)
462 /* parse number string */
463 n = strtol(q_arg, &end, 10);
464 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
466 if (n >= MAX_TIMER_PERIOD)
472 /* Parse the argument given in the command line of the application */
474 l2fwd_parse_args(int argc, char **argv)
479 char *prgname = argv[0];
480 static struct option lgopts[] = {
486 while ((opt = getopt_long(argc, argvopt, "p:q:T:",
487 lgopts, &option_index)) != EOF) {
492 l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
493 if (l2fwd_enabled_port_mask == 0) {
494 printf("invalid portmask\n");
495 l2fwd_usage(prgname);
502 l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
503 if (l2fwd_rx_queue_per_lcore == 0) {
504 printf("invalid queue number\n");
505 l2fwd_usage(prgname);
512 timer_period = l2fwd_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
513 if (timer_period < 0) {
514 printf("invalid timer period\n");
515 l2fwd_usage(prgname);
522 l2fwd_usage(prgname);
526 l2fwd_usage(prgname);
532 argv[optind-1] = prgname;
535 optind = 0; /* reset getopt lib */
540 MAIN(int argc, char **argv)
542 struct lcore_queue_conf *qconf;
544 unsigned int nb_ports;
546 unsigned lcore_id, rx_lcore_id;
548 unsigned nb_ports_in_mask = 0;
551 ret = rte_eal_init(argc, argv);
553 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
557 /* parse application arguments (after the EAL ones) */
558 ret = l2fwd_parse_args(argc, argv);
560 rte_exit(EXIT_FAILURE, "Invalid L2FWD-VF parameters\n");
562 /* create the mbuf pool */
564 rte_mempool_create("mbuf_pool", NB_MBUF,
566 sizeof(struct rte_pktmbuf_pool_private),
567 rte_pktmbuf_pool_init, NULL,
568 rte_pktmbuf_init, NULL,
570 if (l2fwd_pktmbuf_pool == NULL)
571 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
574 #ifdef RTE_LIBRTE_IGB_PMD
575 if (rte_igb_pmd_init() < 0)
576 rte_exit(EXIT_FAILURE, "Cannot init igb pmd\n");
578 #ifdef RTE_LIBRTE_IXGBE_PMD
579 if (rte_ixgbe_pmd_init() < 0)
580 rte_exit(EXIT_FAILURE, "Cannot init ixgbe pmd\n");
582 if (rte_ixgbevf_pmd_init() < 0)
583 rte_exit(EXIT_FAILURE, "Cannot init ixgbevf pmd\n");
586 if (rte_eal_pci_probe() < 0)
587 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
589 nb_ports = rte_eth_dev_count();
591 rte_exit(EXIT_FAILURE, "No Ethernet port - bye\n");
593 if (nb_ports > L2FWD_MAX_PORTS)
594 nb_ports = L2FWD_MAX_PORTS;
596 /* reset l2fwd_dst_ports */
597 for (portid = 0; portid < L2FWD_MAX_PORTS; portid++)
598 l2fwd_dst_ports[portid] = 0;
602 qconf = &lcore_queue_conf[rx_lcore_id];
605 * Initialize the lcore/port-rx-queue configuration of each lcore.
606 * NOTE: Each logical core sends packets out to all port-tx-queues
608 for (portid = 0; portid < nb_ports; portid++) {
610 /* skip ports that are not enabled */
611 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
614 if (nb_ports_in_mask % 2) {
615 l2fwd_dst_ports[portid] = last_port;
616 l2fwd_dst_ports[last_port] = portid;
623 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
624 lcore_queue_conf[rx_lcore_id].n_rx_queue ==
625 l2fwd_rx_queue_per_lcore) {
628 if (rx_lcore_id >= RTE_MAX_LCORE)
629 rte_exit(EXIT_FAILURE, "Not enough cores\n");
632 qconf = &lcore_queue_conf[rx_lcore_id];
633 qconf->tx_queue_id = 0;
634 qconf->rx_queue_list[qconf->n_rx_queue] = portid;
637 printf("Lcore %u: RX port %u\n", rx_lcore_id, portid);
641 /* Initialise each port */
642 for (portid = 0; portid < nb_ports; portid++) {
644 /* skip ports that are not enabled */
645 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
646 printf("Skipping disabled port %u\n", portid);
651 printf("Initializing port %u... ", portid);
653 ret = rte_eth_dev_configure((uint8_t) portid, 1, 1, &port_conf);
655 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
658 rte_eth_macaddr_get((uint8_t) portid,
659 &l2fwd_ports_eth_addr[portid]);
661 /* init one RX queue */
663 ret = rte_eth_rx_queue_setup((uint8_t) portid, 0, nb_rxd,
667 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%u\n",
670 /* init one TX queue */
672 ret = rte_eth_tx_queue_setup((uint8_t) portid, 0, nb_txd,
675 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%u\n",
679 ret = rte_eth_dev_start((uint8_t) portid);
681 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%u\n",
687 printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
689 l2fwd_ports_eth_addr[portid].addr_bytes[0],
690 l2fwd_ports_eth_addr[portid].addr_bytes[1],
691 l2fwd_ports_eth_addr[portid].addr_bytes[2],
692 l2fwd_ports_eth_addr[portid].addr_bytes[3],
693 l2fwd_ports_eth_addr[portid].addr_bytes[4],
694 l2fwd_ports_eth_addr[portid].addr_bytes[5]);
696 /* initialize port stats */
697 memset(&port_statistics, 0, sizeof(port_statistics));
700 /* launch per-lcore init on every lcore */
701 rte_eal_mp_remote_launch(l2fwd_launch_one_lcore, NULL, CALL_MASTER);
702 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
703 if (rte_eal_wait_lcore(lcore_id) < 0)