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 mbuf pool ops name */
117 rte_eal_mbuf_default_mempool_ops(void)
119 return internal_config.mbuf_pool_ops_name;
122 /* Return a pointer to the configuration structure */
124 rte_eal_get_configuration(void)
130 rte_eal_iova_mode(void)
132 return rte_eal_get_configuration()->iova_mode;
135 /* parse a sysfs (or other) file containing one integer value */
137 eal_parse_sysfs_value(const char *filename, unsigned long *val)
143 if ((f = fopen(filename, "r")) == NULL) {
144 RTE_LOG(ERR, EAL, "%s(): cannot open sysfs value %s\n",
149 if (fgets(buf, sizeof(buf), f) == NULL) {
150 RTE_LOG(ERR, EAL, "%s(): cannot read sysfs value %s\n",
155 *val = strtoul(buf, &end, 0);
156 if ((buf[0] == '\0') || (end == NULL) || (*end != '\n')) {
157 RTE_LOG(ERR, EAL, "%s(): cannot parse sysfs value %s\n",
167 /* create memory configuration in shared/mmap memory. Take out
168 * a write lock on the memsegs, so we can auto-detect primary/secondary.
169 * This means we never close the file while running (auto-close on exit).
170 * We also don't lock the whole file, so that in future we can use read-locks
171 * on other parts, e.g. memzones, to detect if there are running secondary
174 rte_eal_config_create(void)
176 void *rte_mem_cfg_addr;
179 const char *pathname = eal_runtime_config_path();
181 if (internal_config.no_shconf)
185 mem_cfg_fd = open(pathname, O_RDWR | O_CREAT, 0660);
187 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
190 retval = ftruncate(mem_cfg_fd, sizeof(*rte_config.mem_config));
193 rte_panic("Cannot resize '%s' for rte_mem_config\n", pathname);
196 retval = fcntl(mem_cfg_fd, F_SETLK, &wr_lock);
199 rte_exit(EXIT_FAILURE, "Cannot create lock on '%s'. Is another primary "
200 "process running?\n", pathname);
203 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
204 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
206 if (rte_mem_cfg_addr == MAP_FAILED){
207 rte_panic("Cannot mmap memory for rte_config\n");
209 memcpy(rte_mem_cfg_addr, &early_mem_config, sizeof(early_mem_config));
210 rte_config.mem_config = rte_mem_cfg_addr;
213 /* attach to an existing shared memory config */
215 rte_eal_config_attach(void)
217 void *rte_mem_cfg_addr;
218 const char *pathname = eal_runtime_config_path();
220 if (internal_config.no_shconf)
224 mem_cfg_fd = open(pathname, O_RDWR);
226 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
229 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
230 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
232 if (rte_mem_cfg_addr == MAP_FAILED)
233 rte_panic("Cannot mmap memory for rte_config\n");
235 rte_config.mem_config = rte_mem_cfg_addr;
238 /* Detect if we are a primary or a secondary process */
240 eal_proc_type_detect(void)
242 enum rte_proc_type_t ptype = RTE_PROC_PRIMARY;
243 const char *pathname = eal_runtime_config_path();
245 /* if we can open the file but not get a write-lock we are a secondary
246 * process. NOTE: if we get a file handle back, we keep that open
247 * and don't close it to prevent a race condition between multiple opens */
248 if (((mem_cfg_fd = open(pathname, O_RDWR)) >= 0) &&
249 (fcntl(mem_cfg_fd, F_SETLK, &wr_lock) < 0))
250 ptype = RTE_PROC_SECONDARY;
252 RTE_LOG(INFO, EAL, "Auto-detected process type: %s\n",
253 ptype == RTE_PROC_PRIMARY ? "PRIMARY" : "SECONDARY");
258 /* Sets up rte_config structure with the pointer to shared memory config.*/
260 rte_config_init(void)
262 rte_config.process_type = internal_config.process_type;
264 switch (rte_config.process_type){
265 case RTE_PROC_PRIMARY:
266 rte_eal_config_create();
268 case RTE_PROC_SECONDARY:
269 rte_eal_config_attach();
270 rte_eal_mcfg_wait_complete(rte_config.mem_config);
273 case RTE_PROC_INVALID:
274 rte_panic("Invalid process type\n");
280 eal_usage(const char *prgname)
282 printf("\nUsage: %s ", prgname);
284 /* Allow the application to print its usage message too if hook is set */
285 if ( rte_application_usage_hook ) {
286 printf("===== Application Usage =====\n\n");
287 rte_application_usage_hook(prgname);
291 /* Set a per-application usage message */
293 rte_set_application_usage_hook( rte_usage_hook_t usage_func )
295 rte_usage_hook_t old_func;
297 /* Will be NULL on the first call to denote the last usage routine. */
298 old_func = rte_application_usage_hook;
299 rte_application_usage_hook = usage_func;
305 eal_get_hugepage_mem_size(void)
310 for (i = 0; i < internal_config.num_hugepage_sizes; i++) {
311 struct hugepage_info *hpi = &internal_config.hugepage_info[i];
312 if (hpi->hugedir != NULL) {
313 for (j = 0; j < RTE_MAX_NUMA_NODES; j++) {
314 size += hpi->hugepage_sz * hpi->num_pages[j];
319 return (size < SIZE_MAX) ? (size_t)(size) : SIZE_MAX;
322 /* Parse the arguments for --log-level only */
324 eal_log_level_parse(int argc, char **argv)
329 const int old_optind = optind;
330 const int old_optopt = optopt;
331 const int old_optreset = optreset;
332 char * const old_optarg = optarg;
338 while ((opt = getopt_long(argc, argvopt, eal_short_options,
339 eal_long_options, &option_index)) != EOF) {
343 /* getopt is not happy, stop right now */
347 ret = (opt == OPT_LOG_LEVEL_NUM) ?
348 eal_parse_common_option(opt, optarg, &internal_config) : 0;
350 /* common parser is not happy */
355 /* restore getopt lib */
358 optreset = old_optreset;
362 /* Parse the argument given in the command line of the application */
364 eal_parse_args(int argc, char **argv)
369 char *prgname = argv[0];
370 const int old_optind = optind;
371 const int old_optopt = optopt;
372 const int old_optreset = optreset;
373 char * const old_optarg = optarg;
379 while ((opt = getopt_long(argc, argvopt, eal_short_options,
380 eal_long_options, &option_index)) != EOF) {
382 /* getopt is not happy, stop right now */
389 ret = eal_parse_common_option(opt, optarg, &internal_config);
390 /* common parser is not happy */
396 /* common parser handled this option */
401 case OPT_MBUF_POOL_OPS_NAME_NUM:
402 internal_config.mbuf_pool_ops_name = optarg;
408 if (opt < OPT_LONG_MIN_NUM && isprint(opt)) {
409 RTE_LOG(ERR, EAL, "Option %c is not supported "
410 "on FreeBSD\n", opt);
411 } else if (opt >= OPT_LONG_MIN_NUM &&
412 opt < OPT_LONG_MAX_NUM) {
413 RTE_LOG(ERR, EAL, "Option %s is not supported "
415 eal_long_options[option_index].name);
417 RTE_LOG(ERR, EAL, "Option %d is not supported "
418 "on FreeBSD\n", opt);
426 if (eal_adjust_config(&internal_config) != 0) {
432 if (eal_check_common_options(&internal_config) != 0) {
439 argv[optind-1] = prgname;
443 /* restore getopt lib */
446 optreset = old_optreset;
453 eal_check_mem_on_local_socket(void)
455 const struct rte_memseg *ms;
458 socket_id = rte_lcore_to_socket_id(rte_config.master_lcore);
460 ms = rte_eal_get_physmem_layout();
462 for (i = 0; i < RTE_MAX_MEMSEG; i++)
463 if (ms[i].socket_id == socket_id &&
467 RTE_LOG(WARNING, EAL, "WARNING: Master core has no "
468 "memory on local socket!\n");
472 sync_func(__attribute__((unused)) void *arg)
478 rte_eal_mcfg_complete(void)
480 /* ALL shared mem_config related INIT DONE */
481 if (rte_config.process_type == RTE_PROC_PRIMARY)
482 rte_config.mem_config->magic = RTE_MAGIC;
485 /* return non-zero if hugepages are enabled. */
486 int rte_eal_has_hugepages(void)
488 return !internal_config.no_hugetlbfs;
491 /* Abstraction for port I/0 privilege */
493 rte_eal_iopl_init(void)
497 fd = open("/dev/io", O_RDWR);
500 /* keep fd open for iopl */
504 static void rte_eal_init_alert(const char *msg)
506 fprintf(stderr, "EAL: FATAL: %s\n", msg);
507 RTE_LOG(ERR, EAL, "%s\n", msg);
510 /* Launch threads, called at application init(). */
512 rte_eal_init(int argc, char **argv)
516 static rte_atomic32_t run_once = RTE_ATOMIC32_INIT(0);
517 char cpuset[RTE_CPU_AFFINITY_STR_LEN];
518 char thread_name[RTE_MAX_THREAD_NAME_LEN];
520 /* checks if the machine is adequate */
521 if (!rte_cpu_is_supported()) {
522 rte_eal_init_alert("unsupported cpu type.");
527 if (!rte_atomic32_test_and_set(&run_once)) {
528 rte_eal_init_alert("already called initialization.");
529 rte_errno = EALREADY;
533 thread_id = pthread_self();
535 eal_reset_internal_config(&internal_config);
537 /* set log level as early as possible */
538 eal_log_level_parse(argc, argv);
540 if (rte_eal_cpu_init() < 0) {
541 rte_eal_init_alert("Cannot detect lcores.");
546 fctret = eal_parse_args(argc, argv);
548 rte_eal_init_alert("Invalid 'command line' arguments.");
550 rte_atomic32_clear(&run_once);
554 if (eal_plugins_init() < 0) {
555 rte_eal_init_alert("Cannot init plugins\n");
557 rte_atomic32_clear(&run_once);
561 if (eal_option_device_parse()) {
563 rte_atomic32_clear(&run_once);
567 if (rte_bus_scan()) {
568 rte_eal_init_alert("Cannot scan the buses for devices\n");
570 rte_atomic32_clear(&run_once);
574 /* autodetect the iova mapping mode (default is iova_pa) */
575 rte_eal_get_configuration()->iova_mode = rte_bus_get_iommu_class();
577 if (internal_config.no_hugetlbfs == 0 &&
578 internal_config.process_type != RTE_PROC_SECONDARY &&
579 eal_hugepage_info_init() < 0) {
580 rte_eal_init_alert("Cannot get hugepage information.");
582 rte_atomic32_clear(&run_once);
586 if (internal_config.memory == 0 && internal_config.force_sockets == 0) {
587 if (internal_config.no_hugetlbfs)
588 internal_config.memory = MEMSIZE_IF_NO_HUGE_PAGE;
590 internal_config.memory = eal_get_hugepage_mem_size();
593 if (internal_config.vmware_tsc_map == 1) {
594 #ifdef RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT
595 rte_cycles_vmware_tsc_map = 1;
596 RTE_LOG (DEBUG, EAL, "Using VMWARE TSC MAP, "
597 "you must have monitor_control.pseudo_perfctr = TRUE\n");
599 RTE_LOG (WARNING, EAL, "Ignoring --vmware-tsc-map because "
600 "RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT is not set\n");
604 rte_srand(rte_rdtsc());
608 if (rte_eal_memory_init() < 0) {
609 rte_eal_init_alert("Cannot init memory\n");
614 if (rte_eal_memzone_init() < 0) {
615 rte_eal_init_alert("Cannot init memzone\n");
620 if (rte_eal_tailqs_init() < 0) {
621 rte_eal_init_alert("Cannot init tail queues for objects\n");
626 if (rte_eal_alarm_init() < 0) {
627 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
628 /* rte_eal_alarm_init sets rte_errno on failure. */
632 if (rte_eal_intr_init() < 0) {
633 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
637 if (rte_eal_timer_init() < 0) {
638 rte_eal_init_alert("Cannot init HPET or TSC timers\n");
643 eal_check_mem_on_local_socket();
645 eal_thread_init_master(rte_config.master_lcore);
647 ret = eal_thread_dump_affinity(cpuset, RTE_CPU_AFFINITY_STR_LEN);
649 RTE_LOG(DEBUG, EAL, "Master lcore %u is ready (tid=%p;cpuset=[%s%s])\n",
650 rte_config.master_lcore, thread_id, cpuset,
651 ret == 0 ? "" : "...");
653 RTE_LCORE_FOREACH_SLAVE(i) {
656 * create communication pipes between master thread
659 if (pipe(lcore_config[i].pipe_master2slave) < 0)
660 rte_panic("Cannot create pipe\n");
661 if (pipe(lcore_config[i].pipe_slave2master) < 0)
662 rte_panic("Cannot create pipe\n");
664 lcore_config[i].state = WAIT;
666 /* create a thread for each lcore */
667 ret = pthread_create(&lcore_config[i].thread_id, NULL,
668 eal_thread_loop, NULL);
670 rte_panic("Cannot create thread\n");
672 /* Set thread_name for aid in debugging. */
673 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN,
674 "lcore-slave-%d", i);
675 rte_thread_setname(lcore_config[i].thread_id, thread_name);
679 * Launch a dummy function on all slave lcores, so that master lcore
680 * knows they are all ready when this function returns.
682 rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MASTER);
683 rte_eal_mp_wait_lcore();
685 /* initialize services so vdevs register service during bus_probe. */
686 ret = rte_service_init();
688 rte_eal_init_alert("rte_service_init() failed\n");
693 /* Probe all the buses and devices/drivers on them */
694 if (rte_bus_probe()) {
695 rte_eal_init_alert("Cannot probe devices\n");
700 /* initialize default service/lcore mappings and start running. Ignore
701 * -ENOTSUP, as it indicates no service coremask passed to EAL.
703 ret = rte_service_start_with_defaults();
704 if (ret < 0 && ret != -ENOTSUP) {
709 rte_eal_mcfg_complete();
715 enum rte_lcore_role_t
716 rte_eal_lcore_role(unsigned lcore_id)
718 return rte_config.lcore_role[lcore_id];
722 rte_eal_process_type(void)
724 return rte_config.process_type;
727 int rte_eal_has_pci(void)
729 return !internal_config.no_pci;
732 int rte_eal_create_uio_dev(void)
734 return internal_config.create_uio_dev;
738 rte_eal_vfio_intr_mode(void)
740 return RTE_INTR_MODE_NONE;
743 /* dummy forward declaration. */
744 struct vfio_device_info;
746 /* dummy prototypes. */
747 int vfio_setup_device(const char *sysfs_base, const char *dev_addr,
748 int *vfio_dev_fd, struct vfio_device_info *device_info);
749 int vfio_release_device(const char *sysfs_base, const char *dev_addr, int fd);
750 int vfio_enable(const char *modname);
751 int vfio_is_enabled(const char *modname);
752 int vfio_noiommu_is_enabled(void);
754 int vfio_setup_device(__rte_unused const char *sysfs_base,
755 __rte_unused const char *dev_addr,
756 __rte_unused int *vfio_dev_fd,
757 __rte_unused struct vfio_device_info *device_info)
762 int vfio_release_device(__rte_unused const char *sysfs_base,
763 __rte_unused const char *dev_addr,
769 int vfio_enable(__rte_unused const char *modname)
774 int vfio_is_enabled(__rte_unused const char *modname)
779 int vfio_noiommu_is_enabled(void)