dd9b12f85669cfc2e490320308b3bd3b0adaf37e
1 /**
2 * collectd - src/collectd.c
3 * Copyright (C) 2005-2007 Florian octo Forster
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice shall be included in
13 * all copies or substantial portions of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
21 * DEALINGS IN THE SOFTWARE.
22 *
23 * Authors:
24 * Florian octo Forster <octo at collectd.org>
25 * Alvaro Barcellos <alvaro.barcellos at gmail.com>
26 **/
28 #include "collectd.h"
30 #include "common.h"
31 #include "configfile.h"
32 #include "plugin.h"
34 #include <netdb.h>
35 #include <sys/types.h>
36 #include <sys/un.h>
38 #if HAVE_LOCALE_H
39 #include <locale.h>
40 #endif
42 #if HAVE_STATGRAB_H
43 #include <statgrab.h>
44 #endif
46 #ifndef COLLECTD_LOCALE
47 #define COLLECTD_LOCALE "C"
48 #endif
50 static int loop = 0;
52 static void *do_flush(void __attribute__((unused)) * arg) {
53 INFO("Flushing all data.");
54 plugin_flush(/* plugin = */ NULL,
55 /* timeout = */ 0,
56 /* ident = */ NULL);
57 INFO("Finished flushing all data.");
58 pthread_exit(NULL);
59 return NULL;
60 }
62 static void sig_int_handler(int __attribute__((unused)) signal) { loop++; }
64 static void sig_term_handler(int __attribute__((unused)) signal) { loop++; }
66 static void sig_usr1_handler(int __attribute__((unused)) signal) {
67 pthread_t thread;
68 pthread_attr_t attr;
70 /* flushing the data might take a while,
71 * so it should be done asynchronously */
72 pthread_attr_init(&attr);
73 pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
74 pthread_create(&thread, &attr, do_flush, NULL);
75 pthread_attr_destroy(&attr);
76 }
78 static int init_hostname(void) {
79 const char *str;
81 struct addrinfo *ai_list;
82 int status;
84 long hostname_len = sysconf(_SC_HOST_NAME_MAX);
85 if (hostname_len == -1) {
86 hostname_len = NI_MAXHOST;
87 }
88 char hostname[hostname_len];
90 str = global_option_get("Hostname");
91 if ((str != NULL) && (str[0] != 0)) {
92 hostname_set(str);
93 return 0;
94 }
96 if (gethostname(hostname, hostname_len) != 0) {
97 fprintf(stderr, "`gethostname' failed and no "
98 "hostname was configured.\n");
99 return -1;
100 }
102 str = global_option_get("FQDNLookup");
103 if (IS_FALSE(str))
104 return 0;
106 struct addrinfo ai_hints = {.ai_flags = AI_CANONNAME};
108 status = getaddrinfo(hostname, NULL, &ai_hints, &ai_list);
109 if (status != 0) {
110 ERROR("Looking up \"%s\" failed. You have set the "
111 "\"FQDNLookup\" option, but I cannot resolve "
112 "my hostname to a fully qualified domain "
113 "name. Please fix the network "
114 "configuration.",
115 hostname);
116 return -1;
117 }
119 for (struct addrinfo *ai_ptr = ai_list; ai_ptr != NULL;
120 ai_ptr = ai_ptr->ai_next) {
121 if (ai_ptr->ai_canonname == NULL)
122 continue;
124 hostname_set(ai_ptr->ai_canonname);
125 break;
126 }
128 freeaddrinfo(ai_list);
129 return 0;
130 } /* int init_hostname */
132 static int init_global_variables(void) {
133 char const *str;
135 interval_g = cf_get_default_interval();
136 assert(interval_g > 0);
137 DEBUG("interval_g = %.3f;", CDTIME_T_TO_DOUBLE(interval_g));
139 str = global_option_get("Timeout");
140 if (str == NULL)
141 str = "2";
142 timeout_g = atoi(str);
143 if (timeout_g <= 1) {
144 fprintf(stderr, "Cannot set the timeout to a correct value.\n"
145 "Please check your settings.\n");
146 return -1;
147 }
148 DEBUG("timeout_g = %i;", timeout_g);
150 if (init_hostname() != 0)
151 return -1;
152 DEBUG("hostname_g = %s;", hostname_g);
154 return 0;
155 } /* int init_global_variables */
157 static int change_basedir(const char *orig_dir, _Bool create) {
158 char *dir;
159 size_t dirlen;
160 int status;
162 dir = strdup(orig_dir);
163 if (dir == NULL) {
164 char errbuf[1024];
165 ERROR("strdup failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
166 return -1;
167 }
169 dirlen = strlen(dir);
170 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
171 dir[--dirlen] = '\0';
173 if (dirlen == 0) {
174 free(dir);
175 return -1;
176 }
178 status = chdir(dir);
179 if (status == 0) {
180 free(dir);
181 return 0;
182 } else if (!create || (errno != ENOENT)) {
183 char errbuf[1024];
184 ERROR("change_basedir: chdir (%s): %s", dir,
185 sstrerror(errno, errbuf, sizeof(errbuf)));
186 free(dir);
187 return -1;
188 }
190 status = mkdir(dir, S_IRWXU | S_IRWXG | S_IRWXO);
191 if (status != 0) {
192 char errbuf[1024];
193 ERROR("change_basedir: mkdir (%s): %s", dir,
194 sstrerror(errno, errbuf, sizeof(errbuf)));
195 free(dir);
196 return -1;
197 }
199 status = chdir(dir);
200 if (status != 0) {
201 char errbuf[1024];
202 ERROR("change_basedir: chdir (%s): %s", dir,
203 sstrerror(errno, errbuf, sizeof(errbuf)));
204 free(dir);
205 return -1;
206 }
208 free(dir);
209 return 0;
210 } /* static int change_basedir (char *dir) */
212 #if HAVE_LIBKSTAT
213 static void update_kstat(void) {
214 if (kc == NULL) {
215 if ((kc = kstat_open()) == NULL)
216 ERROR("Unable to open kstat control structure");
217 } else {
218 kid_t kid;
219 kid = kstat_chain_update(kc);
220 if (kid > 0) {
221 INFO("kstat chain has been updated");
222 plugin_init_all();
223 } else if (kid < 0)
224 ERROR("kstat chain update failed");
225 /* else: everything works as expected */
226 }
228 return;
229 } /* static void update_kstat (void) */
230 #endif /* HAVE_LIBKSTAT */
232 /* TODO
233 * Remove all settings but `-f' and `-C'
234 */
235 __attribute__((noreturn)) static void exit_usage(int status) {
236 printf("Usage: " PACKAGE_NAME " [OPTIONS]\n\n"
238 "Available options:\n"
239 " General:\n"
240 " -C <file> Configuration file.\n"
241 " Default: " CONFIGFILE "\n"
242 " -t Test config and exit.\n"
243 " -T Test plugin read and exit.\n"
244 " -P <file> PID-file.\n"
245 " Default: " PIDFILE "\n"
246 #if COLLECT_DAEMON
247 " -f Don't fork to the background.\n"
248 #endif
249 " -B Don't create the BaseDir\n"
250 " -h Display help (this message)\n"
251 "\nBuiltin defaults:\n"
252 " Config file " CONFIGFILE "\n"
253 " PID file " PIDFILE "\n"
254 " Plugin directory " PLUGINDIR "\n"
255 " Data directory " PKGLOCALSTATEDIR "\n"
256 "\n" PACKAGE_NAME " " PACKAGE_VERSION ", http://collectd.org/\n"
257 "by Florian octo Forster <octo@collectd.org>\n"
258 "for contributions see `AUTHORS'\n");
259 exit(status);
260 } /* static void exit_usage (int status) */
262 static int do_init(void) {
263 #if HAVE_SETLOCALE
264 if (setlocale(LC_NUMERIC, COLLECTD_LOCALE) == NULL)
265 WARNING("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
267 /* Update the environment, so that libraries that are calling
268 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
269 unsetenv("LC_ALL");
270 setenv("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
271 #endif
273 #if HAVE_LIBKSTAT
274 kc = NULL;
275 update_kstat();
276 #endif
278 #if HAVE_LIBSTATGRAB
279 if (sg_init(
280 #if HAVE_LIBSTATGRAB_0_90
281 0
282 #endif
283 )) {
284 ERROR("sg_init: %s", sg_str_error(sg_get_error()));
285 return -1;
286 }
288 if (sg_drop_privileges()) {
289 ERROR("sg_drop_privileges: %s", sg_str_error(sg_get_error()));
290 return -1;
291 }
292 #endif
294 return plugin_init_all();
295 } /* int do_init () */
297 static int do_loop(void) {
298 cdtime_t interval = cf_get_default_interval();
299 cdtime_t wait_until;
301 wait_until = cdtime() + interval;
303 while (loop == 0) {
304 cdtime_t now;
306 #if HAVE_LIBKSTAT
307 update_kstat();
308 #endif
310 /* Issue all plugins */
311 plugin_read_all();
313 now = cdtime();
314 if (now >= wait_until) {
315 WARNING("Not sleeping because the next interval is "
316 "%.3f seconds in the past!",
317 CDTIME_T_TO_DOUBLE(now - wait_until));
318 wait_until = now + interval;
319 continue;
320 }
322 struct timespec ts_wait = CDTIME_T_TO_TIMESPEC(wait_until - now);
323 wait_until = wait_until + interval;
325 while ((loop == 0) && (nanosleep(&ts_wait, &ts_wait) != 0)) {
326 if (errno != EINTR) {
327 char errbuf[1024];
328 ERROR("nanosleep failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
329 return -1;
330 }
331 }
332 } /* while (loop == 0) */
334 return 0;
335 } /* int do_loop */
337 static int do_shutdown(void) {
338 return plugin_shutdown_all();
339 } /* int do_shutdown */
341 #if COLLECT_DAEMON
342 static int pidfile_create(void) {
343 FILE *fh;
344 const char *file = global_option_get("PIDFile");
346 if ((fh = fopen(file, "w")) == NULL) {
347 char errbuf[1024];
348 ERROR("fopen (%s): %s", file, sstrerror(errno, errbuf, sizeof(errbuf)));
349 return 1;
350 }
352 fprintf(fh, "%i\n", (int)getpid());
353 fclose(fh);
355 return 0;
356 } /* static int pidfile_create (const char *file) */
358 static int pidfile_remove(void) {
359 const char *file = global_option_get("PIDFile");
360 if (file == NULL)
361 return 0;
363 return unlink(file);
364 } /* static int pidfile_remove (const char *file) */
365 #endif /* COLLECT_DAEMON */
367 #ifdef KERNEL_LINUX
368 static int notify_upstart(void) {
369 char const *upstart_job = getenv("UPSTART_JOB");
371 if (upstart_job == NULL)
372 return 0;
374 if (strcmp(upstart_job, "collectd") != 0) {
375 WARNING("Environment specifies unexpected UPSTART_JOB=\"%s\", expected "
376 "\"collectd\". Ignoring the variable.",
377 upstart_job);
378 return 0;
379 }
381 NOTICE("Upstart detected, stopping now to signal readyness.");
382 raise(SIGSTOP);
383 unsetenv("UPSTART_JOB");
385 return 1;
386 }
388 static int notify_systemd(void) {
389 int fd;
390 const char *notifysocket;
391 struct sockaddr_un su = {0};
392 size_t su_size;
393 char buffer[] = "READY=1\n";
395 notifysocket = getenv("NOTIFY_SOCKET");
396 if (notifysocket == NULL)
397 return 0;
399 if ((strlen(notifysocket) < 2) ||
400 ((notifysocket[0] != '@') && (notifysocket[0] != '/'))) {
401 ERROR("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be "
402 "absolute",
403 notifysocket);
404 return 0;
405 }
406 NOTICE("Systemd detected, trying to signal readyness.");
408 unsetenv("NOTIFY_SOCKET");
410 #if defined(SOCK_CLOEXEC)
411 fd = socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, /* protocol = */ 0);
412 #else
413 fd = socket(AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
414 #endif
415 if (fd < 0) {
416 char errbuf[1024];
417 ERROR("creating UNIX socket failed: %s",
418 sstrerror(errno, errbuf, sizeof(errbuf)));
419 return 0;
420 }
422 su.sun_family = AF_UNIX;
423 if (notifysocket[0] != '@') {
424 /* regular UNIX socket */
425 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
426 su_size = sizeof(su);
427 } else {
428 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
429 * start with a null byte. Since null bytes have no special meaning in
430 * that case, we have to set su_size correctly to cover only the bytes
431 * that are part of the address. */
432 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
433 su.sun_path[0] = 0;
434 su_size = sizeof(sa_family_t) + strlen(notifysocket);
435 if (su_size > sizeof(su))
436 su_size = sizeof(su);
437 }
439 if (sendto(fd, buffer, strlen(buffer), MSG_NOSIGNAL, (void *)&su,
440 (socklen_t)su_size) < 0) {
441 char errbuf[1024];
442 ERROR("sendto(\"%s\") failed: %s", notifysocket,
443 sstrerror(errno, errbuf, sizeof(errbuf)));
444 close(fd);
445 return 0;
446 }
448 unsetenv("NOTIFY_SOCKET");
449 close(fd);
450 return 1;
451 }
452 #endif /* KERNEL_LINUX */
454 struct cmdline_config {
455 _Bool test_config;
456 _Bool test_readall;
457 _Bool create_basedir;
458 const char *configfile;
459 _Bool daemonize;
460 };
462 void read_cmdline(int argc, char **argv, struct cmdline_config *config) {
463 /* read options */
464 while (1) {
465 int c;
466 c = getopt(argc, argv, "htTC:"
467 #if COLLECT_DAEMON
468 "fP:"
469 #endif
470 );
472 if (c == -1)
473 break;
475 switch (c) {
476 case 'B':
477 config->create_basedir = 0;
478 break;
479 case 'C':
480 config->configfile = optarg;
481 break;
482 case 't':
483 config->test_config = 1;
484 break;
485 case 'T':
486 config->test_readall = 1;
487 global_option_set("ReadThreads", "-1", 1);
488 #if COLLECT_DAEMON
489 config->daemonize = 0;
490 #endif /* COLLECT_DAEMON */
491 break;
492 #if COLLECT_DAEMON
493 case 'P':
494 global_option_set("PIDFile", optarg, 1);
495 break;
496 case 'f':
497 config->daemonize = 0;
498 break;
499 #endif /* COLLECT_DAEMON */
500 case 'h':
501 exit_usage(0);
502 break;
503 default:
504 exit_usage(1);
505 } /* switch (c) */
506 } /* while (1) */
507 }
509 int configure_collectd(struct cmdline_config *config) {
510 const char *basedir;
511 /*
512 * Read options from the config file, the environment and the command
513 * line (in that order, with later options overwriting previous ones in
514 * general).
515 * Also, this will automatically load modules.
516 */
517 if (cf_read(config->configfile)) {
518 fprintf(stderr, "Error: Reading the config file failed!\n"
519 "Read the logs for details.\n");
520 return 1;
521 }
523 /*
524 * Change directory. We do this _after_ reading the config and loading
525 * modules to relative paths work as expected.
526 */
527 if ((basedir = global_option_get("BaseDir")) == NULL) {
528 fprintf(stderr,
529 "Don't have a basedir to use. This should not happen. Ever.");
530 return 1;
531 } else if (change_basedir(basedir, config->create_basedir)) {
532 fprintf(stderr, "Error: Unable to change to directory `%s'.\n", basedir);
533 return 1;
534 }
536 /*
537 * Set global variables or, if that fails, exit. We cannot run with
538 * them being uninitialized. If nothing is configured, then defaults
539 * are being used. So this means that the user has actually done
540 * something wrong.
541 */
542 if (init_global_variables() != 0)
543 return 1;
545 return 0;
546 }
548 int main(int argc, char **argv) {
549 #if COLLECT_DAEMON
550 pid_t pid;
551 #endif
552 int exit_status = 0;
554 struct cmdline_config config = {
555 .daemonize = 1, .create_basedir = 1, .configfile = CONFIGFILE,
556 };
558 read_cmdline(argc, argv, &config);
560 if (config.test_config)
561 return 0;
563 if (optind < argc)
564 exit_usage(1);
566 plugin_init_ctx();
568 int status;
569 if ((status = configure_collectd(&config)) != 0)
570 exit(EXIT_FAILURE);
572 #if COLLECT_DAEMON
573 /*
574 * fork off child
575 */
576 struct sigaction sig_chld_action = {.sa_handler = SIG_IGN};
578 sigaction(SIGCHLD, &sig_chld_action, NULL);
580 /*
581 * Only daemonize if we're not being supervised
582 * by upstart or systemd (when using Linux).
583 */
584 if (config.daemonize
585 #ifdef KERNEL_LINUX
586 && notify_upstart() == 0 && notify_systemd() == 0
587 #endif
588 ) {
589 int status;
591 if ((pid = fork()) == -1) {
592 /* error */
593 char errbuf[1024];
594 fprintf(stderr, "fork: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
595 return 1;
596 } else if (pid != 0) {
597 /* parent */
598 /* printf ("Running (PID %i)\n", pid); */
599 return 0;
600 }
602 /* Detach from session */
603 setsid();
605 /* Write pidfile */
606 if (pidfile_create())
607 exit(2);
609 /* close standard descriptors */
610 close(2);
611 close(1);
612 close(0);
614 status = open("/dev/null", O_RDWR);
615 if (status != 0) {
616 ERROR("Error: Could not connect `STDIN' to `/dev/null' (status %d)",
617 status);
618 return 1;
619 }
621 status = dup(0);
622 if (status != 1) {
623 ERROR("Error: Could not connect `STDOUT' to `/dev/null' (status %d)",
624 status);
625 return 1;
626 }
628 status = dup(0);
629 if (status != 2) {
630 ERROR("Error: Could not connect `STDERR' to `/dev/null', (status %d)",
631 status);
632 return 1;
633 }
634 } /* if (config.daemonize) */
635 #endif /* COLLECT_DAEMON */
637 struct sigaction sig_pipe_action = {.sa_handler = SIG_IGN};
639 sigaction(SIGPIPE, &sig_pipe_action, NULL);
641 /*
642 * install signal handlers
643 */
644 struct sigaction sig_int_action = {.sa_handler = sig_int_handler};
646 if (0 != sigaction(SIGINT, &sig_int_action, NULL)) {
647 char errbuf[1024];
648 ERROR("Error: Failed to install a signal handler for signal INT: %s",
649 sstrerror(errno, errbuf, sizeof(errbuf)));
650 return 1;
651 }
653 struct sigaction sig_term_action = {.sa_handler = sig_term_handler};
655 if (0 != sigaction(SIGTERM, &sig_term_action, NULL)) {
656 char errbuf[1024];
657 ERROR("Error: Failed to install a signal handler for signal TERM: %s",
658 sstrerror(errno, errbuf, sizeof(errbuf)));
659 return 1;
660 }
662 struct sigaction sig_usr1_action = {.sa_handler = sig_usr1_handler};
664 if (0 != sigaction(SIGUSR1, &sig_usr1_action, NULL)) {
665 char errbuf[1024];
666 ERROR("Error: Failed to install a signal handler for signal USR1: %s",
667 sstrerror(errno, errbuf, sizeof(errbuf)));
668 return 1;
669 }
671 /*
672 * run the actual loops
673 */
674 if (do_init() != 0) {
675 ERROR("Error: one or more plugin init callbacks failed.");
676 exit_status = 1;
677 }
679 if (config.test_readall) {
680 if (plugin_read_all_once() != 0) {
681 ERROR("Error: one or more plugin read callbacks failed.");
682 exit_status = 1;
683 }
684 } else {
685 INFO("Initialization complete, entering read-loop.");
686 do_loop();
687 }
689 /* close syslog */
690 INFO("Exiting normally.");
692 if (do_shutdown() != 0) {
693 ERROR("Error: one or more plugin shutdown callbacks failed.");
694 exit_status = 1;
695 }
697 #if COLLECT_DAEMON
698 if (config.daemonize)
699 pidfile_remove();
700 #endif /* COLLECT_DAEMON */
702 return exit_status;
703 } /* int main */