4 * Copyright(c) 2010-2014 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/queue.h>
44 #include <netinet/in.h>
46 #include <linux/if_tun.h>
48 #include <sys/ioctl.h>
52 #include <rte_common.h>
54 #include <rte_memory.h>
55 #include <rte_memcpy.h>
56 #include <rte_memzone.h>
57 #include <rte_tailq.h>
59 #include <rte_per_lcore.h>
60 #include <rte_launch.h>
61 #include <rte_atomic.h>
62 #include <rte_lcore.h>
63 #include <rte_branch_prediction.h>
64 #include <rte_interrupts.h>
66 #include <rte_debug.h>
67 #include <rte_ether.h>
68 #include <rte_ethdev.h>
71 #include <rte_mempool.h>
73 #include <rte_string_fns.h>
74 #include <rte_cycles.h>
76 /* Macros for printing using RTE_LOG */
77 #define RTE_LOGTYPE_APP RTE_LOGTYPE_USER1
78 #define FATAL_ERROR(fmt, args...) rte_exit(EXIT_FAILURE, fmt "\n", ##args)
79 #define PRINT_INFO(fmt, args...) RTE_LOG(INFO, APP, fmt "\n", ##args)
81 /* Max ports than can be used (each port is associated with two lcores) */
82 #define MAX_PORTS (RTE_MAX_LCORE / 2)
84 /* Max size of a single packet */
85 #define MAX_PACKET_SZ 2048
87 /* Number of bytes needed for each mbuf */
89 (MAX_PACKET_SZ + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
91 /* Number of mbufs in mempool that is created */
94 /* How many packets to attempt to read from NIC in one go */
95 #define PKT_BURST_SZ 32
97 /* How many objects (mbufs) to keep in per-lcore mempool cache */
98 #define MEMPOOL_CACHE_SZ PKT_BURST_SZ
100 /* Number of RX ring descriptors */
103 /* Number of TX ring descriptors */
107 * RX and TX Prefetch, Host, and Write-back threshold values should be
108 * carefully set for optimal performance. Consult the network
109 * controller's datasheet and supporting DPDK documentation for guidance
110 * on how these parameters should be set.
113 /* Options for configuring ethernet port */
114 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 = 0, /* CRC stripped by hardware */
123 .mq_mode = ETH_MQ_TX_NONE,
127 /* Mempool for mbufs */
128 static struct rte_mempool * pktmbuf_pool = NULL;
130 /* Mask of enabled ports */
131 static uint32_t ports_mask = 0;
133 /* Mask of cores that read from NIC and write to tap */
134 static uint64_t input_cores_mask = 0;
136 /* Mask of cores that read from tap and write to NIC */
137 static uint64_t output_cores_mask = 0;
139 /* Array storing port_id that is associated with each lcore */
140 static uint8_t port_ids[RTE_MAX_LCORE];
142 /* Structure type for recording lcore-specific stats */
149 /* Array of lcore-specific stats */
150 static struct stats lcore_stats[RTE_MAX_LCORE];
152 /* Print out statistics on packets handled */
158 printf("\n**Exception-Path example application statistics**\n"
159 "======= ====== ============ ============ ===============\n"
160 " Lcore Port RX TX Dropped on TX\n"
161 "------- ------ ------------ ------------ ---------------\n");
162 RTE_LCORE_FOREACH(i) {
163 printf("%6u %7u %13"PRIu64" %13"PRIu64" %16"PRIu64"\n",
164 i, (unsigned)port_ids[i],
165 lcore_stats[i].rx, lcore_stats[i].tx,
166 lcore_stats[i].dropped);
168 printf("======= ====== ============ ============ ===============\n");
171 /* Custom handling of signals to handle stats */
173 signal_handler(int signum)
175 /* When we receive a USR1 signal, print stats */
176 if (signum == SIGUSR1) {
180 /* When we receive a USR2 signal, reset stats */
181 if (signum == SIGUSR2) {
182 memset(&lcore_stats, 0, sizeof(lcore_stats));
183 printf("\n**Statistics have been reset**\n");
189 * Create a tap network interface, or use existing one with same name.
190 * If name[0]='\0' then a name is automatically assigned and returned in name.
192 static int tap_create(char *name)
197 fd = open("/dev/net/tun", O_RDWR);
201 memset(&ifr, 0, sizeof(ifr));
203 /* TAP device without packet information */
204 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
207 snprintf(ifr.ifr_name, IFNAMSIZ, "%s", name);
209 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
216 snprintf(name, IFNAMSIZ, "%s", ifr.ifr_name);
221 /* Main processing loop */
223 main_loop(__attribute__((unused)) void *arg)
225 const unsigned lcore_id = rte_lcore_id();
226 char tap_name[IFNAMSIZ];
229 if ((1ULL << lcore_id) & input_cores_mask) {
230 /* Create new tap interface */
231 snprintf(tap_name, IFNAMSIZ, "tap_dpdk_%.2u", lcore_id);
232 tap_fd = tap_create(tap_name);
234 FATAL_ERROR("Could not create tap interface \"%s\" (%d)",
237 PRINT_INFO("Lcore %u is reading from port %u and writing to %s",
238 lcore_id, (unsigned)port_ids[lcore_id], tap_name);
240 /* Loop forever reading from NIC and writing to tap */
242 struct rte_mbuf *pkts_burst[PKT_BURST_SZ];
244 const unsigned nb_rx =
245 rte_eth_rx_burst(port_ids[lcore_id], 0,
246 pkts_burst, PKT_BURST_SZ);
247 lcore_stats[lcore_id].rx += nb_rx;
248 for (i = 0; likely(i < nb_rx); i++) {
249 struct rte_mbuf *m = pkts_burst[i];
250 /* Ignore return val from write() */
251 int ret = write(tap_fd,
252 rte_pktmbuf_mtod(m, void*),
253 rte_pktmbuf_data_len(m));
255 if (unlikely(ret < 0))
256 lcore_stats[lcore_id].dropped++;
258 lcore_stats[lcore_id].tx++;
262 else if ((1ULL << lcore_id) & output_cores_mask) {
263 /* Create new tap interface */
264 snprintf(tap_name, IFNAMSIZ, "tap_dpdk_%.2u", lcore_id);
265 tap_fd = tap_create(tap_name);
267 FATAL_ERROR("Could not create tap interface \"%s\" (%d)",
270 PRINT_INFO("Lcore %u is reading from %s and writing to port %u",
271 lcore_id, tap_name, (unsigned)port_ids[lcore_id]);
273 /* Loop forever reading from tap and writing to NIC */
276 struct rte_mbuf *m = rte_pktmbuf_alloc(pktmbuf_pool);
280 ret = read(tap_fd, rte_pktmbuf_mtod(m, void *),
282 lcore_stats[lcore_id].rx++;
283 if (unlikely(ret < 0)) {
284 FATAL_ERROR("Reading from %s interface failed",
289 m->pkt_len = (uint16_t)ret;
290 m->data_len = (uint16_t)ret;
291 ret = rte_eth_tx_burst(port_ids[lcore_id], 0, &m, 1);
292 if (unlikely(ret < 1)) {
294 lcore_stats[lcore_id].dropped++;
297 lcore_stats[lcore_id].tx++;
302 PRINT_INFO("Lcore %u has nothing to do", lcore_id);
306 * Tap file is closed automatically when program exits. Putting close()
307 * here will cause the compiler to give an error about unreachable code.
311 /* Display usage instructions */
313 print_usage(const char *prgname)
315 PRINT_INFO("\nUsage: %s [EAL options] -- -p PORTMASK -i IN_CORES -o OUT_CORES\n"
316 " -p PORTMASK: hex bitmask of ports to use\n"
317 " -i IN_CORES: hex bitmask of cores which read from NIC\n"
318 " -o OUT_CORES: hex bitmask of cores which write to NIC",
322 /* Convert string to unsigned number. 0 is returned if error occurs */
324 parse_unsigned(const char *portmask)
329 num = strtoull(portmask, &end, 16);
330 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
333 return (uint64_t)num;
336 /* Record affinities between ports and lcores in global port_ids[] array */
338 setup_port_lcore_affinities(void)
344 /* Setup port_ids[] array, and check masks were ok */
345 RTE_LCORE_FOREACH(i) {
346 if (input_cores_mask & (1ULL << i)) {
347 /* Skip ports that are not enabled */
348 while ((ports_mask & (1 << rx_port)) == 0) {
350 if (rx_port > (sizeof(ports_mask) * 8))
351 goto fail; /* not enough ports */
354 port_ids[i] = rx_port++;
356 else if (output_cores_mask & (1ULL << i)) {
357 /* Skip ports that are not enabled */
358 while ((ports_mask & (1 << tx_port)) == 0) {
360 if (tx_port > (sizeof(ports_mask) * 8))
361 goto fail; /* not enough ports */
364 port_ids[i] = tx_port++;
368 if (rx_port != tx_port)
369 goto fail; /* uneven number of cores in masks */
371 if (ports_mask & (~((1 << rx_port) - 1)))
372 goto fail; /* unused ports */
376 FATAL_ERROR("Invalid core/port masks specified on command line");
379 /* Parse the arguments given in the command line of the application */
381 parse_args(int argc, char **argv)
384 const char *prgname = argv[0];
386 /* Disable printing messages within getopt() */
389 /* Parse command line */
390 while ((opt = getopt(argc, argv, "i:o:p:")) != EOF) {
393 input_cores_mask = parse_unsigned(optarg);
396 output_cores_mask = parse_unsigned(optarg);
399 ports_mask = parse_unsigned(optarg);
402 print_usage(prgname);
403 FATAL_ERROR("Invalid option specified");
407 /* Check that options were parsed ok */
408 if (input_cores_mask == 0) {
409 print_usage(prgname);
410 FATAL_ERROR("IN_CORES not specified correctly");
412 if (output_cores_mask == 0) {
413 print_usage(prgname);
414 FATAL_ERROR("OUT_CORES not specified correctly");
416 if (ports_mask == 0) {
417 print_usage(prgname);
418 FATAL_ERROR("PORTMASK not specified correctly");
421 setup_port_lcore_affinities();
424 /* Initialise a single port on an Ethernet device */
426 init_port(uint8_t port)
430 /* Initialise device and RX/TX queues */
431 PRINT_INFO("Initialising port %u ...", (unsigned)port);
433 ret = rte_eth_dev_configure(port, 1, 1, &port_conf);
435 FATAL_ERROR("Could not configure port%u (%d)",
436 (unsigned)port, ret);
438 ret = rte_eth_rx_queue_setup(port, 0, NB_RXD, rte_eth_dev_socket_id(port),
442 FATAL_ERROR("Could not setup up RX queue for port%u (%d)",
443 (unsigned)port, ret);
445 ret = rte_eth_tx_queue_setup(port, 0, NB_TXD, rte_eth_dev_socket_id(port),
448 FATAL_ERROR("Could not setup up TX queue for port%u (%d)",
449 (unsigned)port, ret);
451 ret = rte_eth_dev_start(port);
453 FATAL_ERROR("Could not start port%u (%d)", (unsigned)port, ret);
455 rte_eth_promiscuous_enable(port);
458 /* Check the link status of all ports in up to 9s, and print them finally */
460 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
462 #define CHECK_INTERVAL 100 /* 100ms */
463 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
464 uint8_t portid, count, all_ports_up, print_flag = 0;
465 struct rte_eth_link link;
467 printf("\nChecking link status");
469 for (count = 0; count <= MAX_CHECK_TIME; count++) {
471 for (portid = 0; portid < port_num; portid++) {
472 if ((port_mask & (1 << portid)) == 0)
474 memset(&link, 0, sizeof(link));
475 rte_eth_link_get_nowait(portid, &link);
476 /* print link status if flag set */
477 if (print_flag == 1) {
478 if (link.link_status)
479 printf("Port %d Link Up - speed %u "
480 "Mbps - %s\n", (uint8_t)portid,
481 (unsigned)link.link_speed,
482 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
483 ("full-duplex") : ("half-duplex\n"));
485 printf("Port %d Link Down\n",
489 /* clear all_ports_up flag if any link down */
490 if (link.link_status == 0) {
495 /* after finally printing all link status, get out */
499 if (all_ports_up == 0) {
502 rte_delay_ms(CHECK_INTERVAL);
505 /* set the print_flag if all ports up or timeout */
506 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
513 /* Initialise ports/queues etc. and start main loop on each core */
515 main(int argc, char** argv)
518 unsigned i,high_port;
519 uint8_t nb_sys_ports, port;
521 /* Associate signal_hanlder function with USR signals */
522 signal(SIGUSR1, signal_handler);
523 signal(SIGUSR2, signal_handler);
526 ret = rte_eal_init(argc, argv);
528 FATAL_ERROR("Could not initialise EAL (%d)", ret);
532 /* Parse application arguments (after the EAL ones) */
533 parse_args(argc, argv);
535 /* Create the mbuf pool */
536 pktmbuf_pool = rte_mempool_create("mbuf_pool", NB_MBUF, MBUF_SZ,
538 sizeof(struct rte_pktmbuf_pool_private),
539 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
541 if (pktmbuf_pool == NULL) {
542 FATAL_ERROR("Could not initialise mbuf pool");
546 /* Get number of ports found in scan */
547 nb_sys_ports = rte_eth_dev_count();
548 if (nb_sys_ports == 0)
549 FATAL_ERROR("No supported Ethernet device found");
550 /* Find highest port set in portmask */
551 for (high_port = (sizeof(ports_mask) * 8) - 1;
552 (high_port != 0) && !(ports_mask & (1 << high_port));
555 if (high_port > nb_sys_ports)
556 FATAL_ERROR("Port mask requires more ports than available");
558 /* Initialise each port */
559 for (port = 0; port < nb_sys_ports; port++) {
560 /* Skip ports that are not enabled */
561 if ((ports_mask & (1 << port)) == 0) {
566 check_all_ports_link_status(nb_sys_ports, ports_mask);
568 /* Launch per-lcore function on every lcore */
569 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
570 RTE_LCORE_FOREACH_SLAVE(i) {
571 if (rte_eal_wait_lcore(i) < 0)