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_compat.h>
52 #include <rte_common.h>
53 #include <rte_debug.h>
54 #include <rte_memory.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>
70 #include <rte_devargs.h>
71 #include <rte_version.h>
72 #include <rte_atomic.h>
73 #include <malloc_heap.h>
75 #include "eal_private.h"
76 #include "eal_thread.h"
77 #include "eal_internal_cfg.h"
78 #include "eal_filesystem.h"
79 #include "eal_hugepages.h"
80 #include "eal_options.h"
82 #define MEMSIZE_IF_NO_HUGE_PAGE (64ULL * 1024ULL * 1024ULL)
84 /* Allow the application to print its usage message too if set */
85 static rte_usage_hook_t rte_application_usage_hook = NULL;
86 /* early configuration structure, when memory config is not mmapped */
87 static struct rte_mem_config early_mem_config;
89 /* define fd variable here, because file needs to be kept open for the
90 * duration of the program, as we hold a write lock on it in the primary proc */
91 static int mem_cfg_fd = -1;
93 static struct flock wr_lock = {
96 .l_start = offsetof(struct rte_mem_config, memseg),
97 .l_len = sizeof(early_mem_config.memseg),
100 /* Address of global and public configuration */
101 static struct rte_config rte_config = {
102 .mem_config = &early_mem_config,
105 /* internal configuration (per-core) */
106 struct lcore_config lcore_config[RTE_MAX_LCORE];
108 /* internal configuration */
109 struct internal_config internal_config;
111 /* used by rte_rdtsc() */
112 int rte_cycles_vmware_tsc_map;
114 /* Return mbuf pool ops name */
116 rte_eal_mbuf_default_mempool_ops(void)
118 return internal_config.user_mbuf_pool_ops_name;
121 /* Return a pointer to the configuration structure */
123 rte_eal_get_configuration(void)
129 rte_eal_iova_mode(void)
131 return rte_eal_get_configuration()->iova_mode;
134 /* parse a sysfs (or other) file containing one integer value */
136 eal_parse_sysfs_value(const char *filename, unsigned long *val)
142 if ((f = fopen(filename, "r")) == NULL) {
143 RTE_LOG(ERR, EAL, "%s(): cannot open sysfs value %s\n",
148 if (fgets(buf, sizeof(buf), f) == NULL) {
149 RTE_LOG(ERR, EAL, "%s(): cannot read sysfs value %s\n",
154 *val = strtoul(buf, &end, 0);
155 if ((buf[0] == '\0') || (end == NULL) || (*end != '\n')) {
156 RTE_LOG(ERR, EAL, "%s(): cannot parse sysfs value %s\n",
166 /* create memory configuration in shared/mmap memory. Take out
167 * a write lock on the memsegs, so we can auto-detect primary/secondary.
168 * This means we never close the file while running (auto-close on exit).
169 * We also don't lock the whole file, so that in future we can use read-locks
170 * on other parts, e.g. memzones, to detect if there are running secondary
173 rte_eal_config_create(void)
175 void *rte_mem_cfg_addr;
178 const char *pathname = eal_runtime_config_path();
180 if (internal_config.no_shconf)
184 mem_cfg_fd = open(pathname, O_RDWR | O_CREAT, 0660);
186 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
189 retval = ftruncate(mem_cfg_fd, sizeof(*rte_config.mem_config));
192 rte_panic("Cannot resize '%s' for rte_mem_config\n", pathname);
195 retval = fcntl(mem_cfg_fd, F_SETLK, &wr_lock);
198 rte_exit(EXIT_FAILURE, "Cannot create lock on '%s'. Is another primary "
199 "process running?\n", pathname);
202 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
203 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
205 if (rte_mem_cfg_addr == MAP_FAILED){
206 rte_panic("Cannot mmap memory for rte_config\n");
208 memcpy(rte_mem_cfg_addr, &early_mem_config, sizeof(early_mem_config));
209 rte_config.mem_config = rte_mem_cfg_addr;
212 /* attach to an existing shared memory config */
214 rte_eal_config_attach(void)
216 void *rte_mem_cfg_addr;
217 const char *pathname = eal_runtime_config_path();
219 if (internal_config.no_shconf)
223 mem_cfg_fd = open(pathname, O_RDWR);
225 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
228 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
229 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
231 if (rte_mem_cfg_addr == MAP_FAILED)
232 rte_panic("Cannot mmap memory for rte_config\n");
234 rte_config.mem_config = rte_mem_cfg_addr;
237 /* Detect if we are a primary or a secondary process */
239 eal_proc_type_detect(void)
241 enum rte_proc_type_t ptype = RTE_PROC_PRIMARY;
242 const char *pathname = eal_runtime_config_path();
244 /* if we can open the file but not get a write-lock we are a secondary
245 * process. NOTE: if we get a file handle back, we keep that open
246 * and don't close it to prevent a race condition between multiple opens */
247 if (((mem_cfg_fd = open(pathname, O_RDWR)) >= 0) &&
248 (fcntl(mem_cfg_fd, F_SETLK, &wr_lock) < 0))
249 ptype = RTE_PROC_SECONDARY;
251 RTE_LOG(INFO, EAL, "Auto-detected process type: %s\n",
252 ptype == RTE_PROC_PRIMARY ? "PRIMARY" : "SECONDARY");
257 /* Sets up rte_config structure with the pointer to shared memory config.*/
259 rte_config_init(void)
261 rte_config.process_type = internal_config.process_type;
263 switch (rte_config.process_type){
264 case RTE_PROC_PRIMARY:
265 rte_eal_config_create();
267 case RTE_PROC_SECONDARY:
268 rte_eal_config_attach();
269 rte_eal_mcfg_wait_complete(rte_config.mem_config);
272 case RTE_PROC_INVALID:
273 rte_panic("Invalid process type\n");
279 eal_usage(const char *prgname)
281 printf("\nUsage: %s ", prgname);
283 /* Allow the application to print its usage message too if hook is set */
284 if ( rte_application_usage_hook ) {
285 printf("===== Application Usage =====\n\n");
286 rte_application_usage_hook(prgname);
290 /* Set a per-application usage message */
292 rte_set_application_usage_hook( rte_usage_hook_t usage_func )
294 rte_usage_hook_t old_func;
296 /* Will be NULL on the first call to denote the last usage routine. */
297 old_func = rte_application_usage_hook;
298 rte_application_usage_hook = usage_func;
304 eal_get_hugepage_mem_size(void)
309 for (i = 0; i < internal_config.num_hugepage_sizes; i++) {
310 struct hugepage_info *hpi = &internal_config.hugepage_info[i];
311 if (hpi->hugedir != NULL) {
312 for (j = 0; j < RTE_MAX_NUMA_NODES; j++) {
313 size += hpi->hugepage_sz * hpi->num_pages[j];
318 return (size < SIZE_MAX) ? (size_t)(size) : SIZE_MAX;
321 /* Parse the arguments for --log-level only */
323 eal_log_level_parse(int argc, char **argv)
328 const int old_optind = optind;
329 const int old_optopt = optopt;
330 const int old_optreset = optreset;
331 char * const old_optarg = optarg;
337 while ((opt = getopt_long(argc, argvopt, eal_short_options,
338 eal_long_options, &option_index)) != EOF) {
342 /* getopt is not happy, stop right now */
346 ret = (opt == OPT_LOG_LEVEL_NUM) ?
347 eal_parse_common_option(opt, optarg, &internal_config) : 0;
349 /* common parser is not happy */
354 /* restore getopt lib */
357 optreset = old_optreset;
361 /* Parse the argument given in the command line of the application */
363 eal_parse_args(int argc, char **argv)
368 char *prgname = argv[0];
369 const int old_optind = optind;
370 const int old_optopt = optopt;
371 const int old_optreset = optreset;
372 char * const old_optarg = optarg;
378 while ((opt = getopt_long(argc, argvopt, eal_short_options,
379 eal_long_options, &option_index)) != EOF) {
381 /* getopt is not happy, stop right now */
388 ret = eal_parse_common_option(opt, optarg, &internal_config);
389 /* common parser is not happy */
395 /* common parser handled this option */
400 case OPT_MBUF_POOL_OPS_NAME_NUM:
401 internal_config.user_mbuf_pool_ops_name = optarg;
407 if (opt < OPT_LONG_MIN_NUM && isprint(opt)) {
408 RTE_LOG(ERR, EAL, "Option %c is not supported "
409 "on FreeBSD\n", opt);
410 } else if (opt >= OPT_LONG_MIN_NUM &&
411 opt < OPT_LONG_MAX_NUM) {
412 RTE_LOG(ERR, EAL, "Option %s is not supported "
414 eal_long_options[option_index].name);
416 RTE_LOG(ERR, EAL, "Option %d is not supported "
417 "on FreeBSD\n", opt);
425 if (eal_adjust_config(&internal_config) != 0) {
431 if (eal_check_common_options(&internal_config) != 0) {
438 argv[optind-1] = prgname;
442 /* restore getopt lib */
445 optreset = old_optreset;
452 eal_check_mem_on_local_socket(void)
454 const struct rte_memseg *ms;
457 socket_id = rte_lcore_to_socket_id(rte_config.master_lcore);
459 ms = rte_eal_get_physmem_layout();
461 for (i = 0; i < RTE_MAX_MEMSEG; i++)
462 if (ms[i].socket_id == socket_id &&
466 RTE_LOG(WARNING, EAL, "WARNING: Master core has no "
467 "memory on local socket!\n");
471 sync_func(__attribute__((unused)) void *arg)
477 rte_eal_mcfg_complete(void)
479 /* ALL shared mem_config related INIT DONE */
480 if (rte_config.process_type == RTE_PROC_PRIMARY)
481 rte_config.mem_config->magic = RTE_MAGIC;
484 /* return non-zero if hugepages are enabled. */
485 int rte_eal_has_hugepages(void)
487 return !internal_config.no_hugetlbfs;
490 /* Abstraction for port I/0 privilege */
492 rte_eal_iopl_init(void)
496 fd = open("/dev/io", O_RDWR);
499 /* keep fd open for iopl */
503 static void rte_eal_init_alert(const char *msg)
505 fprintf(stderr, "EAL: FATAL: %s\n", msg);
506 RTE_LOG(ERR, EAL, "%s\n", msg);
509 /* Launch threads, called at application init(). */
511 rte_eal_init(int argc, char **argv)
515 static rte_atomic32_t run_once = RTE_ATOMIC32_INIT(0);
516 char cpuset[RTE_CPU_AFFINITY_STR_LEN];
517 char thread_name[RTE_MAX_THREAD_NAME_LEN];
519 /* checks if the machine is adequate */
520 if (!rte_cpu_is_supported()) {
521 rte_eal_init_alert("unsupported cpu type.");
526 if (!rte_atomic32_test_and_set(&run_once)) {
527 rte_eal_init_alert("already called initialization.");
528 rte_errno = EALREADY;
532 thread_id = pthread_self();
534 eal_reset_internal_config(&internal_config);
536 /* set log level as early as possible */
537 eal_log_level_parse(argc, argv);
539 if (rte_eal_cpu_init() < 0) {
540 rte_eal_init_alert("Cannot detect lcores.");
545 fctret = eal_parse_args(argc, argv);
547 rte_eal_init_alert("Invalid 'command line' arguments.");
549 rte_atomic32_clear(&run_once);
553 if (eal_plugins_init() < 0) {
554 rte_eal_init_alert("Cannot init plugins\n");
556 rte_atomic32_clear(&run_once);
560 if (eal_option_device_parse()) {
562 rte_atomic32_clear(&run_once);
566 if (rte_bus_scan()) {
567 rte_eal_init_alert("Cannot scan the buses for devices\n");
569 rte_atomic32_clear(&run_once);
573 /* autodetect the iova mapping mode (default is iova_pa) */
574 rte_eal_get_configuration()->iova_mode = rte_bus_get_iommu_class();
576 if (internal_config.no_hugetlbfs == 0 &&
577 internal_config.process_type != RTE_PROC_SECONDARY &&
578 eal_hugepage_info_init() < 0) {
579 rte_eal_init_alert("Cannot get hugepage information.");
581 rte_atomic32_clear(&run_once);
585 if (internal_config.memory == 0 && internal_config.force_sockets == 0) {
586 if (internal_config.no_hugetlbfs)
587 internal_config.memory = MEMSIZE_IF_NO_HUGE_PAGE;
589 internal_config.memory = eal_get_hugepage_mem_size();
592 if (internal_config.vmware_tsc_map == 1) {
593 #ifdef RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT
594 rte_cycles_vmware_tsc_map = 1;
595 RTE_LOG (DEBUG, EAL, "Using VMWARE TSC MAP, "
596 "you must have monitor_control.pseudo_perfctr = TRUE\n");
598 RTE_LOG (WARNING, EAL, "Ignoring --vmware-tsc-map because "
599 "RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT is not set\n");
603 rte_srand(rte_rdtsc());
607 if (rte_eal_memory_init() < 0) {
608 rte_eal_init_alert("Cannot init memory\n");
613 if (rte_eal_memzone_init() < 0) {
614 rte_eal_init_alert("Cannot init memzone\n");
619 if (rte_eal_tailqs_init() < 0) {
620 rte_eal_init_alert("Cannot init tail queues for objects\n");
625 if (rte_eal_alarm_init() < 0) {
626 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
627 /* rte_eal_alarm_init sets rte_errno on failure. */
631 if (rte_eal_intr_init() < 0) {
632 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
636 if (rte_eal_timer_init() < 0) {
637 rte_eal_init_alert("Cannot init HPET or TSC timers\n");
642 eal_check_mem_on_local_socket();
644 eal_thread_init_master(rte_config.master_lcore);
646 ret = eal_thread_dump_affinity(cpuset, RTE_CPU_AFFINITY_STR_LEN);
648 RTE_LOG(DEBUG, EAL, "Master lcore %u is ready (tid=%p;cpuset=[%s%s])\n",
649 rte_config.master_lcore, thread_id, cpuset,
650 ret == 0 ? "" : "...");
652 RTE_LCORE_FOREACH_SLAVE(i) {
655 * create communication pipes between master thread
658 if (pipe(lcore_config[i].pipe_master2slave) < 0)
659 rte_panic("Cannot create pipe\n");
660 if (pipe(lcore_config[i].pipe_slave2master) < 0)
661 rte_panic("Cannot create pipe\n");
663 lcore_config[i].state = WAIT;
665 /* create a thread for each lcore */
666 ret = pthread_create(&lcore_config[i].thread_id, NULL,
667 eal_thread_loop, NULL);
669 rte_panic("Cannot create thread\n");
671 /* Set thread_name for aid in debugging. */
672 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN,
673 "lcore-slave-%d", i);
674 rte_thread_setname(lcore_config[i].thread_id, thread_name);
678 * Launch a dummy function on all slave lcores, so that master lcore
679 * knows they are all ready when this function returns.
681 rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MASTER);
682 rte_eal_mp_wait_lcore();
684 /* initialize services so vdevs register service during bus_probe. */
685 ret = rte_service_init();
687 rte_eal_init_alert("rte_service_init() failed\n");
692 /* Probe all the buses and devices/drivers on them */
693 if (rte_bus_probe()) {
694 rte_eal_init_alert("Cannot probe devices\n");
699 /* initialize default service/lcore mappings and start running. Ignore
700 * -ENOTSUP, as it indicates no service coremask passed to EAL.
702 ret = rte_service_start_with_defaults();
703 if (ret < 0 && ret != -ENOTSUP) {
708 rte_eal_mcfg_complete();
713 int __rte_experimental
714 rte_eal_cleanup(void)
716 rte_service_finalize();
721 enum rte_lcore_role_t
722 rte_eal_lcore_role(unsigned lcore_id)
724 return rte_config.lcore_role[lcore_id];
728 rte_eal_process_type(void)
730 return rte_config.process_type;
733 int rte_eal_has_pci(void)
735 return !internal_config.no_pci;
738 int rte_eal_create_uio_dev(void)
740 return internal_config.create_uio_dev;
744 rte_eal_vfio_intr_mode(void)
746 return RTE_INTR_MODE_NONE;
749 /* dummy forward declaration. */
750 struct vfio_device_info;
752 /* dummy prototypes. */
753 int rte_vfio_setup_device(const char *sysfs_base, const char *dev_addr,
754 int *vfio_dev_fd, struct vfio_device_info *device_info);
755 int rte_vfio_release_device(const char *sysfs_base, const char *dev_addr, int fd);
756 int rte_vfio_enable(const char *modname);
757 int rte_vfio_is_enabled(const char *modname);
758 int rte_vfio_noiommu_is_enabled(void);
759 int rte_vfio_clear_group(int vfio_group_fd);
761 int rte_vfio_setup_device(__rte_unused const char *sysfs_base,
762 __rte_unused const char *dev_addr,
763 __rte_unused int *vfio_dev_fd,
764 __rte_unused struct vfio_device_info *device_info)
769 int rte_vfio_release_device(__rte_unused const char *sysfs_base,
770 __rte_unused const char *dev_addr,
776 int rte_vfio_enable(__rte_unused const char *modname)
781 int rte_vfio_is_enabled(__rte_unused const char *modname)
786 int rte_vfio_noiommu_is_enabled(void)
791 int rte_vfio_clear_group(__rte_unused int vfio_group_fd)