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_launch.h>
56 #include <rte_eal_memconfig.h>
57 #include <rte_errno.h>
58 #include <rte_per_lcore.h>
59 #include <rte_lcore.h>
60 #include <rte_service_component.h>
62 #include <rte_random.h>
63 #include <rte_cycles.h>
64 #include <rte_string_fns.h>
65 #include <rte_cpuflags.h>
66 #include <rte_interrupts.h>
69 #include <rte_devargs.h>
70 #include <rte_version.h>
71 #include <rte_atomic.h>
72 #include <malloc_heap.h>
74 #include "eal_private.h"
75 #include "eal_thread.h"
76 #include "eal_internal_cfg.h"
77 #include "eal_filesystem.h"
78 #include "eal_hugepages.h"
79 #include "eal_options.h"
81 #define MEMSIZE_IF_NO_HUGE_PAGE (64ULL * 1024ULL * 1024ULL)
83 /* Allow the application to print its usage message too if set */
84 static rte_usage_hook_t rte_application_usage_hook = NULL;
85 /* early configuration structure, when memory config is not mmapped */
86 static struct rte_mem_config early_mem_config;
88 /* define fd variable here, because file needs to be kept open for the
89 * duration of the program, as we hold a write lock on it in the primary proc */
90 static int mem_cfg_fd = -1;
92 static struct flock wr_lock = {
95 .l_start = offsetof(struct rte_mem_config, memseg),
96 .l_len = sizeof(early_mem_config.memseg),
99 /* Address of global and public configuration */
100 static struct rte_config rte_config = {
101 .mem_config = &early_mem_config,
104 /* internal configuration (per-core) */
105 struct lcore_config lcore_config[RTE_MAX_LCORE];
107 /* internal configuration */
108 struct internal_config internal_config;
110 /* used by rte_rdtsc() */
111 int rte_cycles_vmware_tsc_map;
113 /* Return mbuf pool ops name */
115 rte_eal_mbuf_default_mempool_ops(void)
117 return internal_config.mbuf_pool_ops_name;
120 /* Return a pointer to the configuration structure */
122 rte_eal_get_configuration(void)
128 rte_eal_iova_mode(void)
130 return rte_eal_get_configuration()->iova_mode;
133 /* parse a sysfs (or other) file containing one integer value */
135 eal_parse_sysfs_value(const char *filename, unsigned long *val)
141 if ((f = fopen(filename, "r")) == NULL) {
142 RTE_LOG(ERR, EAL, "%s(): cannot open sysfs value %s\n",
147 if (fgets(buf, sizeof(buf), f) == NULL) {
148 RTE_LOG(ERR, EAL, "%s(): cannot read sysfs value %s\n",
153 *val = strtoul(buf, &end, 0);
154 if ((buf[0] == '\0') || (end == NULL) || (*end != '\n')) {
155 RTE_LOG(ERR, EAL, "%s(): cannot parse sysfs value %s\n",
165 /* create memory configuration in shared/mmap memory. Take out
166 * a write lock on the memsegs, so we can auto-detect primary/secondary.
167 * This means we never close the file while running (auto-close on exit).
168 * We also don't lock the whole file, so that in future we can use read-locks
169 * on other parts, e.g. memzones, to detect if there are running secondary
172 rte_eal_config_create(void)
174 void *rte_mem_cfg_addr;
177 const char *pathname = eal_runtime_config_path();
179 if (internal_config.no_shconf)
183 mem_cfg_fd = open(pathname, O_RDWR | O_CREAT, 0660);
185 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
188 retval = ftruncate(mem_cfg_fd, sizeof(*rte_config.mem_config));
191 rte_panic("Cannot resize '%s' for rte_mem_config\n", pathname);
194 retval = fcntl(mem_cfg_fd, F_SETLK, &wr_lock);
197 rte_exit(EXIT_FAILURE, "Cannot create lock on '%s'. Is another primary "
198 "process running?\n", pathname);
201 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
202 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
204 if (rte_mem_cfg_addr == MAP_FAILED){
205 rte_panic("Cannot mmap memory for rte_config\n");
207 memcpy(rte_mem_cfg_addr, &early_mem_config, sizeof(early_mem_config));
208 rte_config.mem_config = rte_mem_cfg_addr;
211 /* attach to an existing shared memory config */
213 rte_eal_config_attach(void)
215 void *rte_mem_cfg_addr;
216 const char *pathname = eal_runtime_config_path();
218 if (internal_config.no_shconf)
222 mem_cfg_fd = open(pathname, O_RDWR);
224 rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
227 rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
228 PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
230 if (rte_mem_cfg_addr == MAP_FAILED)
231 rte_panic("Cannot mmap memory for rte_config\n");
233 rte_config.mem_config = rte_mem_cfg_addr;
236 /* Detect if we are a primary or a secondary process */
238 eal_proc_type_detect(void)
240 enum rte_proc_type_t ptype = RTE_PROC_PRIMARY;
241 const char *pathname = eal_runtime_config_path();
243 /* if we can open the file but not get a write-lock we are a secondary
244 * process. NOTE: if we get a file handle back, we keep that open
245 * and don't close it to prevent a race condition between multiple opens */
246 if (((mem_cfg_fd = open(pathname, O_RDWR)) >= 0) &&
247 (fcntl(mem_cfg_fd, F_SETLK, &wr_lock) < 0))
248 ptype = RTE_PROC_SECONDARY;
250 RTE_LOG(INFO, EAL, "Auto-detected process type: %s\n",
251 ptype == RTE_PROC_PRIMARY ? "PRIMARY" : "SECONDARY");
256 /* Sets up rte_config structure with the pointer to shared memory config.*/
258 rte_config_init(void)
260 rte_config.process_type = internal_config.process_type;
262 switch (rte_config.process_type){
263 case RTE_PROC_PRIMARY:
264 rte_eal_config_create();
266 case RTE_PROC_SECONDARY:
267 rte_eal_config_attach();
268 rte_eal_mcfg_wait_complete(rte_config.mem_config);
271 case RTE_PROC_INVALID:
272 rte_panic("Invalid process type\n");
278 eal_usage(const char *prgname)
280 printf("\nUsage: %s ", prgname);
282 /* Allow the application to print its usage message too if hook is set */
283 if ( rte_application_usage_hook ) {
284 printf("===== Application Usage =====\n\n");
285 rte_application_usage_hook(prgname);
289 /* Set a per-application usage message */
291 rte_set_application_usage_hook( rte_usage_hook_t usage_func )
293 rte_usage_hook_t old_func;
295 /* Will be NULL on the first call to denote the last usage routine. */
296 old_func = rte_application_usage_hook;
297 rte_application_usage_hook = usage_func;
303 eal_get_hugepage_mem_size(void)
308 for (i = 0; i < internal_config.num_hugepage_sizes; i++) {
309 struct hugepage_info *hpi = &internal_config.hugepage_info[i];
310 if (hpi->hugedir != NULL) {
311 for (j = 0; j < RTE_MAX_NUMA_NODES; j++) {
312 size += hpi->hugepage_sz * hpi->num_pages[j];
317 return (size < SIZE_MAX) ? (size_t)(size) : SIZE_MAX;
320 /* Parse the arguments for --log-level only */
322 eal_log_level_parse(int argc, char **argv)
327 const int old_optind = optind;
328 const int old_optopt = optopt;
329 const int old_optreset = optreset;
330 char * const old_optarg = optarg;
336 while ((opt = getopt_long(argc, argvopt, eal_short_options,
337 eal_long_options, &option_index)) != EOF) {
341 /* getopt is not happy, stop right now */
345 ret = (opt == OPT_LOG_LEVEL_NUM) ?
346 eal_parse_common_option(opt, optarg, &internal_config) : 0;
348 /* common parser is not happy */
353 /* restore getopt lib */
356 optreset = old_optreset;
360 /* Parse the argument given in the command line of the application */
362 eal_parse_args(int argc, char **argv)
367 char *prgname = argv[0];
368 const int old_optind = optind;
369 const int old_optopt = optopt;
370 const int old_optreset = optreset;
371 char * const old_optarg = optarg;
377 while ((opt = getopt_long(argc, argvopt, eal_short_options,
378 eal_long_options, &option_index)) != EOF) {
380 /* getopt is not happy, stop right now */
387 ret = eal_parse_common_option(opt, optarg, &internal_config);
388 /* common parser is not happy */
394 /* common parser handled this option */
399 case OPT_MBUF_POOL_OPS_NAME_NUM:
400 internal_config.mbuf_pool_ops_name = optarg;
406 if (opt < OPT_LONG_MIN_NUM && isprint(opt)) {
407 RTE_LOG(ERR, EAL, "Option %c is not supported "
408 "on FreeBSD\n", opt);
409 } else if (opt >= OPT_LONG_MIN_NUM &&
410 opt < OPT_LONG_MAX_NUM) {
411 RTE_LOG(ERR, EAL, "Option %s is not supported "
413 eal_long_options[option_index].name);
415 RTE_LOG(ERR, EAL, "Option %d is not supported "
416 "on FreeBSD\n", opt);
424 if (eal_adjust_config(&internal_config) != 0) {
430 if (eal_check_common_options(&internal_config) != 0) {
437 argv[optind-1] = prgname;
441 /* restore getopt lib */
444 optreset = old_optreset;
451 eal_check_mem_on_local_socket(void)
453 const struct rte_memseg *ms;
456 socket_id = rte_lcore_to_socket_id(rte_config.master_lcore);
458 ms = rte_eal_get_physmem_layout();
460 for (i = 0; i < RTE_MAX_MEMSEG; i++)
461 if (ms[i].socket_id == socket_id &&
465 RTE_LOG(WARNING, EAL, "WARNING: Master core has no "
466 "memory on local socket!\n");
470 sync_func(__attribute__((unused)) void *arg)
476 rte_eal_mcfg_complete(void)
478 /* ALL shared mem_config related INIT DONE */
479 if (rte_config.process_type == RTE_PROC_PRIMARY)
480 rte_config.mem_config->magic = RTE_MAGIC;
483 /* return non-zero if hugepages are enabled. */
484 int rte_eal_has_hugepages(void)
486 return !internal_config.no_hugetlbfs;
489 /* Abstraction for port I/0 privilege */
491 rte_eal_iopl_init(void)
495 fd = open("/dev/io", O_RDWR);
498 /* keep fd open for iopl */
502 static void rte_eal_init_alert(const char *msg)
504 fprintf(stderr, "EAL: FATAL: %s\n", msg);
505 RTE_LOG(ERR, EAL, "%s\n", msg);
508 /* Launch threads, called at application init(). */
510 rte_eal_init(int argc, char **argv)
514 static rte_atomic32_t run_once = RTE_ATOMIC32_INIT(0);
515 char cpuset[RTE_CPU_AFFINITY_STR_LEN];
516 char thread_name[RTE_MAX_THREAD_NAME_LEN];
518 /* checks if the machine is adequate */
519 if (!rte_cpu_is_supported()) {
520 rte_eal_init_alert("unsupported cpu type.");
525 if (!rte_atomic32_test_and_set(&run_once)) {
526 rte_eal_init_alert("already called initialization.");
527 rte_errno = EALREADY;
531 thread_id = pthread_self();
533 eal_reset_internal_config(&internal_config);
535 /* set log level as early as possible */
536 eal_log_level_parse(argc, argv);
538 if (rte_eal_cpu_init() < 0) {
539 rte_eal_init_alert("Cannot detect lcores.");
544 fctret = eal_parse_args(argc, argv);
546 rte_eal_init_alert("Invalid 'command line' arguments.");
548 rte_atomic32_clear(&run_once);
552 if (eal_plugins_init() < 0) {
553 rte_eal_init_alert("Cannot init plugins\n");
555 rte_atomic32_clear(&run_once);
559 if (eal_option_device_parse()) {
561 rte_atomic32_clear(&run_once);
565 if (rte_bus_scan()) {
566 rte_eal_init_alert("Cannot scan the buses for devices\n");
568 rte_atomic32_clear(&run_once);
572 /* autodetect the iova mapping mode (default is iova_pa) */
573 rte_eal_get_configuration()->iova_mode = rte_bus_get_iommu_class();
575 if (internal_config.no_hugetlbfs == 0 &&
576 internal_config.process_type != RTE_PROC_SECONDARY &&
577 eal_hugepage_info_init() < 0) {
578 rte_eal_init_alert("Cannot get hugepage information.");
580 rte_atomic32_clear(&run_once);
584 if (internal_config.memory == 0 && internal_config.force_sockets == 0) {
585 if (internal_config.no_hugetlbfs)
586 internal_config.memory = MEMSIZE_IF_NO_HUGE_PAGE;
588 internal_config.memory = eal_get_hugepage_mem_size();
591 if (internal_config.vmware_tsc_map == 1) {
592 #ifdef RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT
593 rte_cycles_vmware_tsc_map = 1;
594 RTE_LOG (DEBUG, EAL, "Using VMWARE TSC MAP, "
595 "you must have monitor_control.pseudo_perfctr = TRUE\n");
597 RTE_LOG (WARNING, EAL, "Ignoring --vmware-tsc-map because "
598 "RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT is not set\n");
602 rte_srand(rte_rdtsc());
606 if (rte_eal_memory_init() < 0) {
607 rte_eal_init_alert("Cannot init memory\n");
612 if (rte_eal_memzone_init() < 0) {
613 rte_eal_init_alert("Cannot init memzone\n");
618 if (rte_eal_tailqs_init() < 0) {
619 rte_eal_init_alert("Cannot init tail queues for objects\n");
624 if (rte_eal_alarm_init() < 0) {
625 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
626 /* rte_eal_alarm_init sets rte_errno on failure. */
630 if (rte_eal_intr_init() < 0) {
631 rte_eal_init_alert("Cannot init interrupt-handling thread\n");
635 if (rte_eal_timer_init() < 0) {
636 rte_eal_init_alert("Cannot init HPET or TSC timers\n");
641 eal_check_mem_on_local_socket();
643 eal_thread_init_master(rte_config.master_lcore);
645 ret = eal_thread_dump_affinity(cpuset, RTE_CPU_AFFINITY_STR_LEN);
647 RTE_LOG(DEBUG, EAL, "Master lcore %u is ready (tid=%p;cpuset=[%s%s])\n",
648 rte_config.master_lcore, thread_id, cpuset,
649 ret == 0 ? "" : "...");
651 RTE_LCORE_FOREACH_SLAVE(i) {
654 * create communication pipes between master thread
657 if (pipe(lcore_config[i].pipe_master2slave) < 0)
658 rte_panic("Cannot create pipe\n");
659 if (pipe(lcore_config[i].pipe_slave2master) < 0)
660 rte_panic("Cannot create pipe\n");
662 lcore_config[i].state = WAIT;
664 /* create a thread for each lcore */
665 ret = pthread_create(&lcore_config[i].thread_id, NULL,
666 eal_thread_loop, NULL);
668 rte_panic("Cannot create thread\n");
670 /* Set thread_name for aid in debugging. */
671 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN,
672 "lcore-slave-%d", i);
673 rte_thread_setname(lcore_config[i].thread_id, thread_name);
677 * Launch a dummy function on all slave lcores, so that master lcore
678 * knows they are all ready when this function returns.
680 rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MASTER);
681 rte_eal_mp_wait_lcore();
683 /* initialize services so vdevs register service during bus_probe. */
684 ret = rte_service_init();
686 rte_eal_init_alert("rte_service_init() failed\n");
691 /* Probe all the buses and devices/drivers on them */
692 if (rte_bus_probe()) {
693 rte_eal_init_alert("Cannot probe devices\n");
698 /* initialize default service/lcore mappings and start running. Ignore
699 * -ENOTSUP, as it indicates no service coremask passed to EAL.
701 ret = rte_service_start_with_defaults();
702 if (ret < 0 && ret != -ENOTSUP) {
707 rte_eal_mcfg_complete();
713 enum rte_lcore_role_t
714 rte_eal_lcore_role(unsigned lcore_id)
716 return rte_config.lcore_role[lcore_id];
720 rte_eal_process_type(void)
722 return rte_config.process_type;
725 int rte_eal_has_pci(void)
727 return !internal_config.no_pci;
730 int rte_eal_create_uio_dev(void)
732 return internal_config.create_uio_dev;
736 rte_eal_vfio_intr_mode(void)
738 return RTE_INTR_MODE_NONE;
741 /* dummy forward declaration. */
742 struct vfio_device_info;
744 /* dummy prototypes. */
745 int rte_vfio_setup_device(const char *sysfs_base, const char *dev_addr,
746 int *vfio_dev_fd, struct vfio_device_info *device_info);
747 int rte_vfio_release_device(const char *sysfs_base, const char *dev_addr, int fd);
748 int rte_vfio_enable(const char *modname);
749 int rte_vfio_is_enabled(const char *modname);
750 int rte_vfio_noiommu_is_enabled(void);
752 int rte_vfio_setup_device(__rte_unused const char *sysfs_base,
753 __rte_unused const char *dev_addr,
754 __rte_unused int *vfio_dev_fd,
755 __rte_unused struct vfio_device_info *device_info)
760 int rte_vfio_release_device(__rte_unused const char *sysfs_base,
761 __rte_unused const char *dev_addr,
767 int rte_vfio_enable(__rte_unused const char *modname)
772 int rte_vfio_is_enabled(__rte_unused const char *modname)
777 int rte_vfio_noiommu_is_enabled(void)