4 * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
5 * Copyright(c) 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
49 #include <sys/queue.h>
51 #include <rte_common.h>
52 #include <rte_debug.h>
53 #include <rte_memory.h>
54 #include <rte_memzone.h>
55 #include <rte_launch.h>
57 #include <rte_eal_memconfig.h>
58 #include <rte_errno.h>
59 #include <rte_per_lcore.h>
60 #include <rte_lcore.h>
61 #include <rte_service_component.h>
63 #include <rte_random.h>
64 #include <rte_cycles.h>
65 #include <rte_string_fns.h>
66 #include <rte_cpuflags.h>
67 #include <rte_interrupts.h>
71 #include <rte_devargs.h>
72 #include <rte_version.h>
73 #include <rte_atomic.h>
74 #include <malloc_heap.h>
76 #include "eal_private.h"
77 #include "eal_thread.h"
78 #include "eal_internal_cfg.h"
79 #include "eal_filesystem.h"
80 #include "eal_hugepages.h"
81 #include "eal_options.h"
83 #define MEMSIZE_IF_NO_HUGE_PAGE (64ULL * 1024ULL * 1024ULL)
85 /* Allow the application to print its usage message too if set */
86 static rte_usage_hook_t rte_application_usage_hook = NULL;
87 /* early configuration structure, when memory config is not mmapped */
88 static struct rte_mem_config early_mem_config;
90 /* define fd variable here, because file needs to be kept open for the
91 * duration of the program, as we hold a write lock on it in the primary proc */
92 static int mem_cfg_fd = -1;
94 static struct flock wr_lock = {
97 .l_start = offsetof(struct rte_mem_config, memseg),
98 .l_len = sizeof(early_mem_config.memseg),
101 /* Address of global and public configuration */
102 static struct rte_config rte_config = {
103 .mem_config = &early_mem_config,
106 /* internal configuration (per-core) */
107 struct lcore_config lcore_config[RTE_MAX_LCORE];
109 /* internal configuration */
110 struct internal_config internal_config;
112 /* used by rte_rdtsc() */
113 int rte_cycles_vmware_tsc_map;
115 /* Return a pointer to the configuration structure */
117 rte_eal_get_configuration(void)
123 rte_eal_iova_mode(void)
125 return rte_eal_get_configuration()->iova_mode;
128 /* parse a sysfs (or other) file containing one integer value */
130 eal_parse_sysfs_value(const char *filename, unsigned long *val)
136 if ((f = fopen(filename, "r")) == NULL) {
137 RTE_LOG(ERR, EAL, "%s(): cannot open sysfs value %s\n",
142 if (fgets(buf, sizeof(buf), f) == NULL) {
143 RTE_LOG(ERR, EAL, "%s(): cannot read sysfs value %s\n",
148 *val = strtoul(buf, &end, 0);
149 if ((buf[0] == '\0') || (end == NULL) || (*end != '\n')) {
150 RTE_LOG(ERR, EAL, "%s(): cannot parse sysfs value %s\n",
160 /* create memory configuration in shared/mmap memory. Take out
161 * a write lock on the memsegs, so we can auto-detect primary/secondary.
162 * This means we never close the file while running (auto-close on exit).
163 * We also don't lock the whole file, so that in future we can use read-locks
164 * on other parts, e.g. memzones, to detect if there are running secondary
167 rte_eal_config_create(void)
169 void *rte_mem_cfg_addr;
172 const char *pathname = eal_runtime_config_path();
174 if (internal_config.no_shconf)
178 mem_cfg_fd = open(pathname, O_RDWR | O_CREAT, 0660);
180 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
183 retval = ftruncate(mem_cfg_fd, sizeof(*rte_config.mem_config));
186 rte_panic("Cannot resize '%s' for rte_mem_config\n", pathname);
189 retval = fcntl(mem_cfg_fd, F_SETLK, &wr_lock);
192 rte_exit(EXIT_FAILURE, "Cannot create lock on '%s'. Is another primary "
193 "process running?\n", pathname);
196 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
197 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
199 if (rte_mem_cfg_addr == MAP_FAILED){
200 rte_panic("Cannot mmap memory for rte_config\n");
202 memcpy(rte_mem_cfg_addr, &early_mem_config, sizeof(early_mem_config));
203 rte_config.mem_config = rte_mem_cfg_addr;
206 /* attach to an existing shared memory config */
208 rte_eal_config_attach(void)
210 void *rte_mem_cfg_addr;
211 const char *pathname = eal_runtime_config_path();
213 if (internal_config.no_shconf)
217 mem_cfg_fd = open(pathname, O_RDWR);
219 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
222 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
223 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
225 if (rte_mem_cfg_addr == MAP_FAILED)
226 rte_panic("Cannot mmap memory for rte_config\n");
228 rte_config.mem_config = rte_mem_cfg_addr;
231 /* Detect if we are a primary or a secondary process */
233 eal_proc_type_detect(void)
235 enum rte_proc_type_t ptype = RTE_PROC_PRIMARY;
236 const char *pathname = eal_runtime_config_path();
238 /* if we can open the file but not get a write-lock we are a secondary
239 * process. NOTE: if we get a file handle back, we keep that open
240 * and don't close it to prevent a race condition between multiple opens */
241 if (((mem_cfg_fd = open(pathname, O_RDWR)) >= 0) &&
242 (fcntl(mem_cfg_fd, F_SETLK, &wr_lock) < 0))
243 ptype = RTE_PROC_SECONDARY;
245 RTE_LOG(INFO, EAL, "Auto-detected process type: %s\n",
246 ptype == RTE_PROC_PRIMARY ? "PRIMARY" : "SECONDARY");
251 /* Sets up rte_config structure with the pointer to shared memory config.*/
253 rte_config_init(void)
255 rte_config.process_type = internal_config.process_type;
257 switch (rte_config.process_type){
258 case RTE_PROC_PRIMARY:
259 rte_eal_config_create();
261 case RTE_PROC_SECONDARY:
262 rte_eal_config_attach();
263 rte_eal_mcfg_wait_complete(rte_config.mem_config);
266 case RTE_PROC_INVALID:
267 rte_panic("Invalid process type\n");
273 eal_usage(const char *prgname)
275 printf("\nUsage: %s ", prgname);
277 /* Allow the application to print its usage message too if hook is set */
278 if ( rte_application_usage_hook ) {
279 printf("===== Application Usage =====\n\n");
280 rte_application_usage_hook(prgname);
284 /* Set a per-application usage message */
286 rte_set_application_usage_hook( rte_usage_hook_t usage_func )
288 rte_usage_hook_t old_func;
290 /* Will be NULL on the first call to denote the last usage routine. */
291 old_func = rte_application_usage_hook;
292 rte_application_usage_hook = usage_func;
298 eal_get_hugepage_mem_size(void)
303 for (i = 0; i < internal_config.num_hugepage_sizes; i++) {
304 struct hugepage_info *hpi = &internal_config.hugepage_info[i];
305 if (hpi->hugedir != NULL) {
306 for (j = 0; j < RTE_MAX_NUMA_NODES; j++) {
307 size += hpi->hugepage_sz * hpi->num_pages[j];
312 return (size < SIZE_MAX) ? (size_t)(size) : SIZE_MAX;
315 /* Parse the arguments for --log-level only */
317 eal_log_level_parse(int argc, char **argv)
322 const int old_optind = optind;
323 const int old_optopt = optopt;
324 const int old_optreset = optreset;
325 char * const old_optarg = optarg;
331 while ((opt = getopt_long(argc, argvopt, eal_short_options,
332 eal_long_options, &option_index)) != EOF) {
336 /* getopt is not happy, stop right now */
340 ret = (opt == OPT_LOG_LEVEL_NUM) ?
341 eal_parse_common_option(opt, optarg, &internal_config) : 0;
343 /* common parser is not happy */
348 /* restore getopt lib */
351 optreset = old_optreset;
355 /* Parse the argument given in the command line of the application */
357 eal_parse_args(int argc, char **argv)
362 char *prgname = argv[0];
363 const int old_optind = optind;
364 const int old_optopt = optopt;
365 const int old_optreset = optreset;
366 char * const old_optarg = optarg;
372 while ((opt = getopt_long(argc, argvopt, eal_short_options,
373 eal_long_options, &option_index)) != EOF) {
375 /* getopt is not happy, stop right now */
382 ret = eal_parse_common_option(opt, optarg, &internal_config);
383 /* common parser is not happy */
389 /* common parser handled this option */
398 if (opt < OPT_LONG_MIN_NUM && isprint(opt)) {
399 RTE_LOG(ERR, EAL, "Option %c is not supported "
400 "on FreeBSD\n", opt);
401 } else if (opt >= OPT_LONG_MIN_NUM &&
402 opt < OPT_LONG_MAX_NUM) {
403 RTE_LOG(ERR, EAL, "Option %s is not supported "
405 eal_long_options[option_index].name);
407 RTE_LOG(ERR, EAL, "Option %d is not supported "
408 "on FreeBSD\n", opt);
416 if (eal_adjust_config(&internal_config) != 0) {
422 if (eal_check_common_options(&internal_config) != 0) {
429 argv[optind-1] = prgname;
433 /* restore getopt lib */
436 optreset = old_optreset;
443 eal_check_mem_on_local_socket(void)
445 const struct rte_memseg *ms;
448 socket_id = rte_lcore_to_socket_id(rte_config.master_lcore);
450 ms = rte_eal_get_physmem_layout();
452 for (i = 0; i < RTE_MAX_MEMSEG; i++)
453 if (ms[i].socket_id == socket_id &&
457 RTE_LOG(WARNING, EAL, "WARNING: Master core has no "
458 "memory on local socket!\n");
462 sync_func(__attribute__((unused)) void *arg)
468 rte_eal_mcfg_complete(void)
470 /* ALL shared mem_config related INIT DONE */
471 if (rte_config.process_type == RTE_PROC_PRIMARY)
472 rte_config.mem_config->magic = RTE_MAGIC;
475 /* return non-zero if hugepages are enabled. */
476 int rte_eal_has_hugepages(void)
478 return !internal_config.no_hugetlbfs;
481 /* Abstraction for port I/0 privilege */
483 rte_eal_iopl_init(void)
487 fd = open("/dev/io", O_RDWR);
490 /* keep fd open for iopl */
494 static void rte_eal_init_alert(const char *msg)
496 fprintf(stderr, "EAL: FATAL: %s\n", msg);
497 RTE_LOG(ERR, EAL, "%s\n", msg);
500 /* Launch threads, called at application init(). */
502 rte_eal_init(int argc, char **argv)
506 static rte_atomic32_t run_once = RTE_ATOMIC32_INIT(0);
507 char cpuset[RTE_CPU_AFFINITY_STR_LEN];
508 char thread_name[RTE_MAX_THREAD_NAME_LEN];
510 /* checks if the machine is adequate */
511 if (!rte_cpu_is_supported()) {
512 rte_eal_init_alert("unsupported cpu type.");
517 if (!rte_atomic32_test_and_set(&run_once)) {
518 rte_eal_init_alert("already called initialization.");
519 rte_errno = EALREADY;
523 thread_id = pthread_self();
525 eal_reset_internal_config(&internal_config);
527 /* set log level as early as possible */
528 eal_log_level_parse(argc, argv);
530 if (rte_eal_cpu_init() < 0) {
531 rte_eal_init_alert("Cannot detect lcores.");
536 fctret = eal_parse_args(argc, argv);
538 rte_eal_init_alert("Invalid 'command line' arguments.");
540 rte_atomic32_clear(&run_once);
544 if (internal_config.no_hugetlbfs == 0 &&
545 internal_config.process_type != RTE_PROC_SECONDARY &&
546 eal_hugepage_info_init() < 0) {
547 rte_eal_init_alert("Cannot get hugepage information.");
549 rte_atomic32_clear(&run_once);
553 if (internal_config.memory == 0 && internal_config.force_sockets == 0) {
554 if (internal_config.no_hugetlbfs)
555 internal_config.memory = MEMSIZE_IF_NO_HUGE_PAGE;
557 internal_config.memory = eal_get_hugepage_mem_size();
560 if (internal_config.vmware_tsc_map == 1) {
561 #ifdef RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT
562 rte_cycles_vmware_tsc_map = 1;
563 RTE_LOG (DEBUG, EAL, "Using VMWARE TSC MAP, "
564 "you must have monitor_control.pseudo_perfctr = TRUE\n");
566 RTE_LOG (WARNING, EAL, "Ignoring --vmware-tsc-map because "
567 "RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT is not set\n");
571 rte_srand(rte_rdtsc());
575 if (rte_eal_memory_init() < 0) {
576 rte_eal_init_alert("Cannot init memory\n");
581 if (rte_eal_memzone_init() < 0) {
582 rte_eal_init_alert("Cannot init memzone\n");
587 if (rte_eal_tailqs_init() < 0) {
588 rte_eal_init_alert("Cannot init tail queues for objects\n");
593 if (rte_eal_alarm_init() < 0) {
594 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
595 /* rte_eal_alarm_init sets rte_errno on failure. */
599 if (rte_eal_intr_init() < 0) {
600 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
604 if (rte_eal_timer_init() < 0) {
605 rte_eal_init_alert("Cannot init HPET or TSC timers\n");
610 eal_check_mem_on_local_socket();
612 if (eal_plugins_init() < 0)
613 rte_eal_init_alert("Cannot init plugins\n");
615 eal_thread_init_master(rte_config.master_lcore);
617 ret = eal_thread_dump_affinity(cpuset, RTE_CPU_AFFINITY_STR_LEN);
619 RTE_LOG(DEBUG, EAL, "Master lcore %u is ready (tid=%p;cpuset=[%s%s])\n",
620 rte_config.master_lcore, thread_id, cpuset,
621 ret == 0 ? "" : "...");
623 if (eal_option_device_parse()) {
628 if (rte_bus_scan()) {
629 rte_eal_init_alert("Cannot scan the buses for devices\n");
634 RTE_LCORE_FOREACH_SLAVE(i) {
637 * create communication pipes between master thread
640 if (pipe(lcore_config[i].pipe_master2slave) < 0)
641 rte_panic("Cannot create pipe\n");
642 if (pipe(lcore_config[i].pipe_slave2master) < 0)
643 rte_panic("Cannot create pipe\n");
645 lcore_config[i].state = WAIT;
647 /* create a thread for each lcore */
648 ret = pthread_create(&lcore_config[i].thread_id, NULL,
649 eal_thread_loop, NULL);
651 rte_panic("Cannot create thread\n");
653 /* Set thread_name for aid in debugging. */
654 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN,
655 "lcore-slave-%d", i);
656 rte_thread_setname(lcore_config[i].thread_id, thread_name);
660 * Launch a dummy function on all slave lcores, so that master lcore
661 * knows they are all ready when this function returns.
663 rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MASTER);
664 rte_eal_mp_wait_lcore();
666 /* initialize services so vdevs register service during bus_probe. */
667 ret = rte_service_init();
669 rte_eal_init_alert("rte_service_init() failed\n");
674 /* Probe all the buses and devices/drivers on them */
675 if (rte_bus_probe()) {
676 rte_eal_init_alert("Cannot probe devices\n");
681 /* initialize default service/lcore mappings and start running. Ignore
682 * -ENOTSUP, as it indicates no service coremask passed to EAL.
684 ret = rte_service_start_with_defaults();
685 if (ret < 0 && ret != -ENOTSUP) {
690 rte_eal_mcfg_complete();
696 enum rte_lcore_role_t
697 rte_eal_lcore_role(unsigned lcore_id)
699 return rte_config.lcore_role[lcore_id];
703 rte_eal_process_type(void)
705 return rte_config.process_type;