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.
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 /* Max ports than can be used (each port is associated with two lcores) */
83 #define MAX_PORTS (RTE_MAX_LCORE / 2)
85 /* Max size of a single packet */
86 #define MAX_PACKET_SZ 2048
88 /* Number of bytes needed for each mbuf */
90 (MAX_PACKET_SZ + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
92 /* Number of mbufs in mempool that is created */
95 /* How many packets to attempt to read from NIC in one go */
96 #define PKT_BURST_SZ 32
98 /* How many objects (mbufs) to keep in per-lcore mempool cache */
99 #define MEMPOOL_CACHE_SZ PKT_BURST_SZ
101 /* Number of RX ring descriptors */
104 /* Number of TX ring descriptors */
108 * RX and TX Prefetch, Host, and Write-back threshold values should be
109 * carefully set for optimal performance. Consult the network
110 * controller's datasheet and supporting DPDK documentation for guidance
111 * on how these parameters should be set.
113 /* RX ring configuration */
114 static const struct rte_eth_rxconf rx_conf = {
116 .pthresh = 8, /* Ring prefetch threshold */
117 .hthresh = 8, /* Ring host threshold */
118 .wthresh = 4, /* Ring writeback threshold */
120 .rx_free_thresh = 0, /* Immediately free RX descriptors */
124 * These default values are optimized for use with the Intel(R) 82599 10 GbE
125 * Controller and the DPDK ixgbe PMD. Consider using other values for other
126 * network controllers and/or network drivers.
128 /* TX ring configuration */
129 static const struct rte_eth_txconf tx_conf = {
131 .pthresh = 36, /* Ring prefetch threshold */
132 .hthresh = 0, /* Ring host threshold */
133 .wthresh = 0, /* Ring writeback threshold */
135 .tx_free_thresh = 0, /* Use PMD default values */
136 .tx_rs_thresh = 0, /* Use PMD default values */
139 /* Options for configuring ethernet port */
140 static const struct rte_eth_conf port_conf = {
142 .header_split = 0, /* Header Split disabled */
143 .hw_ip_checksum = 0, /* IP checksum offload disabled */
144 .hw_vlan_filter = 0, /* VLAN filtering disabled */
145 .jumbo_frame = 0, /* Jumbo Frame Support disabled */
146 .hw_strip_crc = 0, /* CRC stripped by hardware */
149 .mq_mode = ETH_MQ_TX_NONE,
153 /* Mempool for mbufs */
154 static struct rte_mempool * pktmbuf_pool = NULL;
156 /* Mask of enabled ports */
157 static uint32_t ports_mask = 0;
159 /* Mask of cores that read from NIC and write to tap */
160 static uint32_t input_cores_mask = 0;
162 /* Mask of cores that read from tap and write to NIC */
163 static uint32_t output_cores_mask = 0;
165 /* Array storing port_id that is associated with each lcore */
166 static uint8_t port_ids[RTE_MAX_LCORE];
168 /* Structure type for recording lcore-specific stats */
175 /* Array of lcore-specific stats */
176 static struct stats lcore_stats[RTE_MAX_LCORE];
178 /* Print out statistics on packets handled */
184 printf("\n**Exception-Path example application statistics**\n"
185 "======= ====== ============ ============ ===============\n"
186 " Lcore Port RX TX Dropped on TX\n"
187 "------- ------ ------------ ------------ ---------------\n");
188 RTE_LCORE_FOREACH(i) {
189 printf("%6u %7u %13"PRIu64" %13"PRIu64" %16"PRIu64"\n",
190 i, (unsigned)port_ids[i],
191 lcore_stats[i].rx, lcore_stats[i].tx,
192 lcore_stats[i].dropped);
194 printf("======= ====== ============ ============ ===============\n");
197 /* Custom handling of signals to handle stats */
199 signal_handler(int signum)
201 /* When we receive a USR1 signal, print stats */
202 if (signum == SIGUSR1) {
206 /* When we receive a USR2 signal, reset stats */
207 if (signum == SIGUSR2) {
208 memset(&lcore_stats, 0, sizeof(lcore_stats));
209 printf("\n**Statistics have been reset**\n");
215 * Create a tap network interface, or use existing one with same name.
216 * If name[0]='\0' then a name is automatically assigned and returned in name.
218 static int tap_create(char *name)
223 fd = open("/dev/net/tun", O_RDWR);
227 memset(&ifr, 0, sizeof(ifr));
229 /* TAP device without packet information */
230 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
233 rte_snprintf(ifr.ifr_name, IFNAMSIZ, name);
235 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
242 rte_snprintf(name, IFNAMSIZ, ifr.ifr_name);
247 /* Main processing loop */
248 static __attribute__((noreturn)) int
249 main_loop(__attribute__((unused)) void *arg)
251 const unsigned lcore_id = rte_lcore_id();
252 char tap_name[IFNAMSIZ];
255 /* Create new tap interface */
256 rte_snprintf(tap_name, IFNAMSIZ, "tap_dpdk_%.2u", lcore_id);
257 tap_fd = tap_create(tap_name);
259 FATAL_ERROR("Could not create tap interface \"%s\" (%d)",
262 if ((1 << lcore_id) & input_cores_mask) {
263 PRINT_INFO("Lcore %u is reading from port %u and writing to %s",
264 lcore_id, (unsigned)port_ids[lcore_id], tap_name);
266 /* Loop forever reading from NIC and writing to tap */
268 struct rte_mbuf *pkts_burst[PKT_BURST_SZ];
270 const unsigned nb_rx =
271 rte_eth_rx_burst(port_ids[lcore_id], 0,
272 pkts_burst, PKT_BURST_SZ);
273 lcore_stats[lcore_id].rx += nb_rx;
274 for (i = 0; likely(i < nb_rx); i++) {
275 struct rte_mbuf *m = pkts_burst[i];
276 /* Ignore return val from write() */
277 int ret = write(tap_fd,
278 rte_pktmbuf_mtod(m, void*),
279 rte_pktmbuf_data_len(m));
281 if (unlikely(ret < 0))
282 lcore_stats[lcore_id].dropped++;
284 lcore_stats[lcore_id].tx++;
288 else if ((1 << lcore_id) & output_cores_mask) {
289 PRINT_INFO("Lcore %u is reading from %s and writing to port %u",
290 lcore_id, tap_name, (unsigned)port_ids[lcore_id]);
292 /* Loop forever reading from tap and writing to NIC */
295 struct rte_mbuf *m = rte_pktmbuf_alloc(pktmbuf_pool);
299 ret = read(tap_fd, m->pkt.data, MAX_PACKET_SZ);
300 lcore_stats[lcore_id].rx++;
301 if (unlikely(ret < 0)) {
302 FATAL_ERROR("Reading from %s interface failed",
307 m->pkt.pkt_len = (uint16_t)ret;
308 m->pkt.data_len = (uint16_t)ret;
309 ret = rte_eth_tx_burst(port_ids[lcore_id], 0, &m, 1);
310 if (unlikely(ret < 1)) {
312 lcore_stats[lcore_id].dropped++;
315 lcore_stats[lcore_id].tx++;
320 PRINT_INFO("Lcore %u has nothing to do", lcore_id);
322 ; /* loop doing nothing */
325 * Tap file is closed automatically when program exits. Putting close()
326 * here will cause the compiler to give an error about unreachable code.
330 /* Display usage instructions */
332 print_usage(const char *prgname)
334 PRINT_INFO("\nUsage: %s [EAL options] -- -p PORTMASK -i IN_CORES -o OUT_CORES\n"
335 " -p PORTMASK: hex bitmask of ports to use\n"
336 " -i IN_CORES: hex bitmask of cores which read from NIC\n"
337 " -o OUT_CORES: hex bitmask of cores which write to NIC",
341 /* Convert string to unsigned number. 0 is returned if error occurs */
343 parse_unsigned(const char *portmask)
348 num = strtoul(portmask, &end, 16);
349 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
352 return (uint32_t)num;
355 /* Record affinities between ports and lcores in global port_ids[] array */
357 setup_port_lcore_affinities(void)
363 /* Setup port_ids[] array, and check masks were ok */
364 RTE_LCORE_FOREACH(i) {
365 if (input_cores_mask & (1 << i)) {
366 /* Skip ports that are not enabled */
367 while ((ports_mask & (1 << rx_port)) == 0) {
369 if (rx_port > (sizeof(ports_mask) * 8))
370 goto fail; /* not enough ports */
373 port_ids[i] = rx_port++;
375 else if (output_cores_mask & (1 << i)) {
376 /* Skip ports that are not enabled */
377 while ((ports_mask & (1 << tx_port)) == 0) {
379 if (tx_port > (sizeof(ports_mask) * 8))
380 goto fail; /* not enough ports */
383 port_ids[i] = tx_port++;
387 if (rx_port != tx_port)
388 goto fail; /* uneven number of cores in masks */
390 if (ports_mask & (~((1 << rx_port) - 1)))
391 goto fail; /* unused ports */
395 FATAL_ERROR("Invalid core/port masks specified on command line");
398 /* Parse the arguments given in the command line of the application */
400 parse_args(int argc, char **argv)
403 const char *prgname = argv[0];
405 /* Disable printing messages within getopt() */
408 /* Parse command line */
409 while ((opt = getopt(argc, argv, "i:o:p:")) != EOF) {
412 input_cores_mask = parse_unsigned(optarg);
415 output_cores_mask = parse_unsigned(optarg);
418 ports_mask = parse_unsigned(optarg);
421 print_usage(prgname);
422 FATAL_ERROR("Invalid option specified");
426 /* Check that options were parsed ok */
427 if (input_cores_mask == 0) {
428 print_usage(prgname);
429 FATAL_ERROR("IN_CORES not specified correctly");
431 if (output_cores_mask == 0) {
432 print_usage(prgname);
433 FATAL_ERROR("OUT_CORES not specified correctly");
435 if (ports_mask == 0) {
436 print_usage(prgname);
437 FATAL_ERROR("PORTMASK not specified correctly");
440 setup_port_lcore_affinities();
443 /* Initialise a single port on an Ethernet device */
445 init_port(uint8_t port)
449 /* Initialise device and RX/TX queues */
450 PRINT_INFO("Initialising port %u ...", (unsigned)port);
452 ret = rte_eth_dev_configure(port, 1, 1, &port_conf);
454 FATAL_ERROR("Could not configure port%u (%d)",
455 (unsigned)port, ret);
457 ret = rte_eth_rx_queue_setup(port, 0, NB_RXD, rte_eth_dev_socket_id(port),
458 &rx_conf, pktmbuf_pool);
460 FATAL_ERROR("Could not setup up RX queue for port%u (%d)",
461 (unsigned)port, ret);
463 ret = rte_eth_tx_queue_setup(port, 0, NB_TXD, rte_eth_dev_socket_id(port),
466 FATAL_ERROR("Could not setup up TX queue for port%u (%d)",
467 (unsigned)port, ret);
469 ret = rte_eth_dev_start(port);
471 FATAL_ERROR("Could not start port%u (%d)", (unsigned)port, ret);
473 rte_eth_promiscuous_enable(port);
476 /* Check the link status of all ports in up to 9s, and print them finally */
478 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
480 #define CHECK_INTERVAL 100 /* 100ms */
481 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
482 uint8_t portid, count, all_ports_up, print_flag = 0;
483 struct rte_eth_link link;
485 printf("\nChecking link status");
487 for (count = 0; count <= MAX_CHECK_TIME; count++) {
489 for (portid = 0; portid < port_num; portid++) {
490 if ((port_mask & (1 << portid)) == 0)
492 memset(&link, 0, sizeof(link));
493 rte_eth_link_get_nowait(portid, &link);
494 /* print link status if flag set */
495 if (print_flag == 1) {
496 if (link.link_status)
497 printf("Port %d Link Up - speed %u "
498 "Mbps - %s\n", (uint8_t)portid,
499 (unsigned)link.link_speed,
500 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
501 ("full-duplex") : ("half-duplex\n"));
503 printf("Port %d Link Down\n",
507 /* clear all_ports_up flag if any link down */
508 if (link.link_status == 0) {
513 /* after finally printing all link status, get out */
517 if (all_ports_up == 0) {
520 rte_delay_ms(CHECK_INTERVAL);
523 /* set the print_flag if all ports up or timeout */
524 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
531 /* Initialise ports/queues etc. and start main loop on each core */
533 main(int argc, char** argv)
536 unsigned i,high_port;
537 uint8_t nb_sys_ports, port;
539 /* Associate signal_hanlder function with USR signals */
540 signal(SIGUSR1, signal_handler);
541 signal(SIGUSR2, signal_handler);
544 ret = rte_eal_init(argc, argv);
546 FATAL_ERROR("Could not initialise EAL (%d)", ret);
550 /* Parse application arguments (after the EAL ones) */
551 parse_args(argc, argv);
553 /* Create the mbuf pool */
554 pktmbuf_pool = rte_mempool_create("mbuf_pool", NB_MBUF, MBUF_SZ,
556 sizeof(struct rte_pktmbuf_pool_private),
557 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
559 if (pktmbuf_pool == NULL) {
560 FATAL_ERROR("Could not initialise mbuf pool");
564 /* Initialise PMD driver(s) */
565 ret = rte_pmd_init_all();
567 FATAL_ERROR("Could not probe PMD (%d)", ret);
569 /* Scan PCI bus for recognised devices */
570 ret = rte_eal_pci_probe();
572 FATAL_ERROR("Could not probe PCI (%d)", ret);
574 /* Get number of ports found in scan */
575 nb_sys_ports = rte_eth_dev_count();
576 if (nb_sys_ports == 0)
577 FATAL_ERROR("No supported Ethernet devices found - check that "
578 "CONFIG_RTE_LIBRTE_IGB_PMD=y and/or "
579 "CONFIG_RTE_LIBRTE_IXGBE_PMD=y in the config file");
580 /* Find highest port set in portmask */
581 for (high_port = (sizeof(ports_mask) * 8) - 1;
582 (high_port != 0) && !(ports_mask & (1 << high_port));
585 if (high_port > nb_sys_ports)
586 FATAL_ERROR("Port mask requires more ports than available");
588 /* Initialise each port */
589 for (port = 0; port < nb_sys_ports; port++) {
590 /* Skip ports that are not enabled */
591 if ((ports_mask & (1 << port)) == 0) {
596 check_all_ports_link_status(nb_sys_ports, ports_mask);
598 /* Launch per-lcore function on every lcore */
599 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
600 RTE_LCORE_FOREACH_SLAVE(i) {
601 if (rte_eal_wait_lcore(i) < 0)