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.
40 #include <sys/queue.h>
45 #include <netinet/in.h>
47 #include <linux/if_tun.h>
49 #include <sys/ioctl.h>
53 #include <rte_common.h>
55 #include <rte_memory.h>
56 #include <rte_memcpy.h>
57 #include <rte_memzone.h>
58 #include <rte_tailq.h>
60 #include <rte_per_lcore.h>
61 #include <rte_launch.h>
62 #include <rte_atomic.h>
63 #include <rte_lcore.h>
64 #include <rte_branch_prediction.h>
65 #include <rte_interrupts.h>
67 #include <rte_debug.h>
68 #include <rte_ether.h>
69 #include <rte_ethdev.h>
72 #include <rte_mempool.h>
74 #include <rte_string_fns.h>
75 #include <rte_cycles.h>
77 /* Macros for printing using RTE_LOG */
78 #define RTE_LOGTYPE_APP RTE_LOGTYPE_USER1
79 #define FATAL_ERROR(fmt, args...) rte_exit(EXIT_FAILURE, fmt "\n", ##args)
80 #define PRINT_INFO(fmt, args...) RTE_LOG(INFO, APP, fmt "\n", ##args)
82 /* NUMA socket to allocate mbuf pool on */
85 /* Max ports than can be used (each port is associated with two lcores) */
86 #define MAX_PORTS (RTE_MAX_LCORE / 2)
88 /* Max size of a single packet */
89 #define MAX_PACKET_SZ 2048
91 /* Number of bytes needed for each mbuf */
93 (MAX_PACKET_SZ + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
95 /* Number of mbufs in mempool that is created */
98 /* How many packets to attempt to read from NIC in one go */
99 #define PKT_BURST_SZ 32
101 /* How many objects (mbufs) to keep in per-lcore mempool cache */
102 #define MEMPOOL_CACHE_SZ PKT_BURST_SZ
104 /* Number of RX ring descriptors */
107 /* Number of TX ring descriptors */
111 * RX and TX Prefetch, Host, and Write-back threshold values should be
112 * carefully set for optimal performance. Consult the network
113 * controller's datasheet and supporting DPDK documentation for guidance
114 * on how these parameters should be set.
116 /* RX ring configuration */
117 static const struct rte_eth_rxconf rx_conf = {
119 .pthresh = 8, /* Ring prefetch threshold */
120 .hthresh = 8, /* Ring host threshold */
121 .wthresh = 4, /* Ring writeback threshold */
123 .rx_free_thresh = 0, /* Immediately free RX descriptors */
127 * These default values are optimized for use with the Intel(R) 82599 10 GbE
128 * Controller and the DPDK ixgbe PMD. Consider using other values for other
129 * network controllers and/or network drivers.
131 /* TX ring configuration */
132 static const struct rte_eth_txconf tx_conf = {
134 .pthresh = 36, /* Ring prefetch threshold */
135 .hthresh = 0, /* Ring host threshold */
136 .wthresh = 0, /* Ring writeback threshold */
138 .tx_free_thresh = 0, /* Use PMD default values */
139 .tx_rs_thresh = 0, /* Use PMD default values */
142 /* Options for configuring ethernet port */
143 static const struct rte_eth_conf port_conf = {
145 .header_split = 0, /* Header Split disabled */
146 .hw_ip_checksum = 0, /* IP checksum offload disabled */
147 .hw_vlan_filter = 0, /* VLAN filtering disabled */
148 .jumbo_frame = 0, /* Jumbo Frame Support disabled */
149 .hw_strip_crc = 0, /* CRC stripped by hardware */
152 .mq_mode = ETH_DCB_NONE,
156 /* Mempool for mbufs */
157 static struct rte_mempool * pktmbuf_pool = NULL;
159 /* Mask of enabled ports */
160 static uint32_t ports_mask = 0;
162 /* Mask of cores that read from NIC and write to tap */
163 static uint32_t input_cores_mask = 0;
165 /* Mask of cores that read from tap and write to NIC */
166 static uint32_t output_cores_mask = 0;
168 /* Array storing port_id that is associated with each lcore */
169 static uint8_t port_ids[RTE_MAX_LCORE];
171 /* Structure type for recording lcore-specific stats */
178 /* Array of lcore-specific stats */
179 static struct stats lcore_stats[RTE_MAX_LCORE];
181 /* Print out statistics on packets handled */
187 printf("\n**Exception-Path example application statistics**\n"
188 "======= ====== ============ ============ ===============\n"
189 " Lcore Port RX TX Dropped on TX\n"
190 "------- ------ ------------ ------------ ---------------\n");
191 RTE_LCORE_FOREACH(i) {
192 printf("%6u %7u %13"PRIu64" %13"PRIu64" %16"PRIu64"\n",
193 i, (unsigned)port_ids[i],
194 lcore_stats[i].rx, lcore_stats[i].tx,
195 lcore_stats[i].dropped);
197 printf("======= ====== ============ ============ ===============\n");
200 /* Custom handling of signals to handle stats */
202 signal_handler(int signum)
204 /* When we receive a USR1 signal, print stats */
205 if (signum == SIGUSR1) {
209 /* When we receive a USR2 signal, reset stats */
210 if (signum == SIGUSR2) {
211 memset(&lcore_stats, 0, sizeof(lcore_stats));
212 printf("\n**Statistics have been reset**\n");
218 * Create a tap network interface, or use existing one with same name.
219 * If name[0]='\0' then a name is automatically assigned and returned in name.
221 static int tap_create(char *name)
226 fd = open("/dev/net/tun", O_RDWR);
230 memset(&ifr, 0, sizeof(ifr));
232 /* TAP device without packet information */
233 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
236 rte_snprintf(ifr.ifr_name, IFNAMSIZ, name);
238 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
245 rte_snprintf(name, IFNAMSIZ, ifr.ifr_name);
250 /* Main processing loop */
251 static __attribute__((noreturn)) int
252 main_loop(__attribute__((unused)) void *arg)
254 const unsigned lcore_id = rte_lcore_id();
255 char tap_name[IFNAMSIZ];
258 /* Create new tap interface */
259 rte_snprintf(tap_name, IFNAMSIZ, "tap_dpdk_%.2u", lcore_id);
260 tap_fd = tap_create(tap_name);
262 FATAL_ERROR("Could not create tap interface \"%s\" (%d)",
265 if ((1 << lcore_id) & input_cores_mask) {
266 PRINT_INFO("Lcore %u is reading from port %u and writing to %s",
267 lcore_id, (unsigned)port_ids[lcore_id], tap_name);
269 /* Loop forever reading from NIC and writing to tap */
271 struct rte_mbuf *pkts_burst[PKT_BURST_SZ];
273 const unsigned nb_rx =
274 rte_eth_rx_burst(port_ids[lcore_id], 0,
275 pkts_burst, PKT_BURST_SZ);
276 lcore_stats[lcore_id].rx += nb_rx;
277 for (i = 0; likely(i < nb_rx); i++) {
278 struct rte_mbuf *m = pkts_burst[i];
279 /* Ignore return val from write() */
280 int ret = write(tap_fd,
281 rte_pktmbuf_mtod(m, void*),
282 rte_pktmbuf_data_len(m));
284 if (unlikely(ret < 0))
285 lcore_stats[lcore_id].dropped++;
287 lcore_stats[lcore_id].tx++;
291 else if ((1 << lcore_id) & output_cores_mask) {
292 PRINT_INFO("Lcore %u is reading from %s and writing to port %u",
293 lcore_id, tap_name, (unsigned)port_ids[lcore_id]);
295 /* Loop forever reading from tap and writing to NIC */
298 struct rte_mbuf *m = rte_pktmbuf_alloc(pktmbuf_pool);
302 ret = read(tap_fd, m->pkt.data, MAX_PACKET_SZ);
303 lcore_stats[lcore_id].rx++;
304 if (unlikely(ret < 0)) {
305 FATAL_ERROR("Reading from %s interface failed",
310 m->pkt.pkt_len = (uint16_t)ret;
311 m->pkt.data_len = (uint16_t)ret;
312 ret = rte_eth_tx_burst(port_ids[lcore_id], 0, &m, 1);
313 if (unlikely(ret < 1)) {
315 lcore_stats[lcore_id].dropped++;
318 lcore_stats[lcore_id].tx++;
323 PRINT_INFO("Lcore %u has nothing to do", lcore_id);
325 ; /* loop doing nothing */
328 * Tap file is closed automatically when program exits. Putting close()
329 * here will cause the compiler to give an error about unreachable code.
333 /* Display usage instructions */
335 print_usage(const char *prgname)
337 PRINT_INFO("\nUsage: %s [EAL options] -- -p PORTMASK -i IN_CORES -o OUT_CORES\n"
338 " -p PORTMASK: hex bitmask of ports to use\n"
339 " -i IN_CORES: hex bitmask of cores which read from NIC\n"
340 " -o OUT_CORES: hex bitmask of cores which write to NIC",
344 /* Convert string to unsigned number. 0 is returned if error occurs */
346 parse_unsigned(const char *portmask)
351 num = strtoul(portmask, &end, 16);
352 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
355 return (uint32_t)num;
358 /* Record affinities between ports and lcores in global port_ids[] array */
360 setup_port_lcore_affinities(void)
366 /* Setup port_ids[] array, and check masks were ok */
367 RTE_LCORE_FOREACH(i) {
368 if (input_cores_mask & (1 << i)) {
369 /* Skip ports that are not enabled */
370 while ((ports_mask & (1 << rx_port)) == 0) {
372 if (rx_port > (sizeof(ports_mask) * 8))
373 goto fail; /* not enough ports */
376 port_ids[i] = rx_port++;
378 else if (output_cores_mask & (1 << i)) {
379 /* Skip ports that are not enabled */
380 while ((ports_mask & (1 << tx_port)) == 0) {
382 if (tx_port > (sizeof(ports_mask) * 8))
383 goto fail; /* not enough ports */
386 port_ids[i] = tx_port++;
390 if (rx_port != tx_port)
391 goto fail; /* uneven number of cores in masks */
393 if (ports_mask & (~((1 << rx_port) - 1)))
394 goto fail; /* unused ports */
398 FATAL_ERROR("Invalid core/port masks specified on command line");
401 /* Parse the arguments given in the command line of the application */
403 parse_args(int argc, char **argv)
406 const char *prgname = argv[0];
408 /* Disable printing messages within getopt() */
411 /* Parse command line */
412 while ((opt = getopt(argc, argv, "i:o:p:")) != EOF) {
415 input_cores_mask = parse_unsigned(optarg);
418 output_cores_mask = parse_unsigned(optarg);
421 ports_mask = parse_unsigned(optarg);
424 print_usage(prgname);
425 FATAL_ERROR("Invalid option specified");
429 /* Check that options were parsed ok */
430 if (input_cores_mask == 0) {
431 print_usage(prgname);
432 FATAL_ERROR("IN_CORES not specified correctly");
434 if (output_cores_mask == 0) {
435 print_usage(prgname);
436 FATAL_ERROR("OUT_CORES not specified correctly");
438 if (ports_mask == 0) {
439 print_usage(prgname);
440 FATAL_ERROR("PORTMASK not specified correctly");
443 setup_port_lcore_affinities();
446 /* Initialise a single port on an Ethernet device */
448 init_port(uint8_t port)
452 /* Initialise device and RX/TX queues */
453 PRINT_INFO("Initialising port %u ...", (unsigned)port);
455 ret = rte_eth_dev_configure(port, 1, 1, &port_conf);
457 FATAL_ERROR("Could not configure port%u (%d)",
458 (unsigned)port, ret);
460 ret = rte_eth_rx_queue_setup(port, 0, NB_RXD, SOCKET, &rx_conf,
463 FATAL_ERROR("Could not setup up RX queue for port%u (%d)",
464 (unsigned)port, ret);
466 ret = rte_eth_tx_queue_setup(port, 0, NB_TXD, SOCKET, &tx_conf);
468 FATAL_ERROR("Could not setup up TX queue for port%u (%d)",
469 (unsigned)port, ret);
471 ret = rte_eth_dev_start(port);
473 FATAL_ERROR("Could not start port%u (%d)", (unsigned)port, ret);
475 rte_eth_promiscuous_enable(port);
478 /* Check the link status of all ports in up to 9s, and print them finally */
480 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
482 #define CHECK_INTERVAL 100 /* 100ms */
483 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
484 uint8_t portid, count, all_ports_up, print_flag = 0;
485 struct rte_eth_link link;
487 printf("\nChecking link status");
489 for (count = 0; count <= MAX_CHECK_TIME; count++) {
491 for (portid = 0; portid < port_num; portid++) {
492 if ((port_mask & (1 << portid)) == 0)
494 memset(&link, 0, sizeof(link));
495 rte_eth_link_get_nowait(portid, &link);
496 /* print link status if flag set */
497 if (print_flag == 1) {
498 if (link.link_status)
499 printf("Port %d Link Up - speed %u "
500 "Mbps - %s\n", (uint8_t)portid,
501 (unsigned)link.link_speed,
502 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
503 ("full-duplex") : ("half-duplex\n"));
505 printf("Port %d Link Down\n",
509 /* clear all_ports_up flag if any link down */
510 if (link.link_status == 0) {
515 /* after finally printing all link status, get out */
519 if (all_ports_up == 0) {
522 rte_delay_ms(CHECK_INTERVAL);
525 /* set the print_flag if all ports up or timeout */
526 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
533 /* Initialise ports/queues etc. and start main loop on each core */
535 main(int argc, char** argv)
538 unsigned i,high_port;
539 uint8_t nb_sys_ports, port;
541 /* Associate signal_hanlder function with USR signals */
542 signal(SIGUSR1, signal_handler);
543 signal(SIGUSR2, signal_handler);
546 ret = rte_eal_init(argc, argv);
548 FATAL_ERROR("Could not initialise EAL (%d)", ret);
552 /* Parse application arguments (after the EAL ones) */
553 parse_args(argc, argv);
555 /* Create the mbuf pool */
556 pktmbuf_pool = rte_mempool_create("mbuf_pool", NB_MBUF, MBUF_SZ,
558 sizeof(struct rte_pktmbuf_pool_private),
559 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
561 if (pktmbuf_pool == NULL) {
562 FATAL_ERROR("Could not initialise mbuf pool");
566 /* Initialise PMD driver(s) */
567 ret = rte_pmd_init_all();
569 FATAL_ERROR("Could not probe PMD (%d)", ret);
571 /* Scan PCI bus for recognised devices */
572 ret = rte_eal_pci_probe();
574 FATAL_ERROR("Could not probe PCI (%d)", ret);
576 /* Get number of ports found in scan */
577 nb_sys_ports = rte_eth_dev_count();
578 if (nb_sys_ports == 0)
579 FATAL_ERROR("No supported Ethernet devices found - check that "
580 "CONFIG_RTE_LIBRTE_IGB_PMD=y and/or "
581 "CONFIG_RTE_LIBRTE_IXGBE_PMD=y in the config file");
582 /* Find highest port set in portmask */
583 for (high_port = (sizeof(ports_mask) * 8) - 1;
584 (high_port != 0) && !(ports_mask & (1 << high_port));
587 if (high_port > nb_sys_ports)
588 FATAL_ERROR("Port mask requires more ports than available");
590 /* Initialise each port */
591 for (port = 0; port < nb_sys_ports; port++) {
592 /* Skip ports that are not enabled */
593 if ((ports_mask & (1 << port)) == 0) {
598 check_all_ports_link_status(nb_sys_ports, ports_mask);
600 /* Launch per-lcore function on every lcore */
601 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
602 RTE_LCORE_FOREACH_SLAVE(i) {
603 if (rte_eal_wait_lcore(i) < 0)