791199ffb6302a8968749a23bedb677cd1d93f12
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];
89 hostname_set(hostname);
91 str = global_option_get("Hostname");
92 if ((str != NULL) && (str[0] != 0)) {
93 hostname_set(str);
94 return 0;
95 }
97 if (gethostname(hostname, hostname_len) != 0) {
98 fprintf(stderr, "`gethostname' failed and no "
99 "hostname was configured.\n");
100 return -1;
101 }
103 str = global_option_get("FQDNLookup");
104 if (IS_FALSE(str))
105 return 0;
107 struct addrinfo ai_hints = {.ai_flags = AI_CANONNAME};
109 status = getaddrinfo(hostname, NULL, &ai_hints, &ai_list);
110 if (status != 0) {
111 ERROR("Looking up \"%s\" failed. You have set the "
112 "\"FQDNLookup\" option, but I cannot resolve "
113 "my hostname to a fully qualified domain "
114 "name. Please fix the network "
115 "configuration.",
116 hostname);
117 return -1;
118 }
120 for (struct addrinfo *ai_ptr = ai_list; ai_ptr != NULL;
121 ai_ptr = ai_ptr->ai_next) {
122 if (ai_ptr->ai_canonname == NULL)
123 continue;
125 hostname_set(ai_ptr->ai_canonname);
126 break;
127 }
129 freeaddrinfo(ai_list);
130 return 0;
131 } /* int init_hostname */
133 static int init_global_variables(void) {
134 char const *str;
136 interval_g = cf_get_default_interval();
137 assert(interval_g > 0);
138 DEBUG("interval_g = %.3f;", CDTIME_T_TO_DOUBLE(interval_g));
140 str = global_option_get("Timeout");
141 if (str == NULL)
142 str = "2";
143 timeout_g = atoi(str);
144 if (timeout_g <= 1) {
145 fprintf(stderr, "Cannot set the timeout to a correct value.\n"
146 "Please check your settings.\n");
147 return -1;
148 }
149 DEBUG("timeout_g = %i;", timeout_g);
151 if (init_hostname() != 0)
152 return -1;
153 DEBUG("hostname_g = %s;", hostname_g);
155 return 0;
156 } /* int init_global_variables */
158 static int change_basedir(const char *orig_dir, _Bool create) {
159 char *dir;
160 size_t dirlen;
161 int status;
163 dir = strdup(orig_dir);
164 if (dir == NULL) {
165 char errbuf[1024];
166 ERROR("strdup failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
167 return -1;
168 }
170 dirlen = strlen(dir);
171 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
172 dir[--dirlen] = '\0';
174 if (dirlen == 0) {
175 free(dir);
176 return -1;
177 }
179 status = chdir(dir);
180 if (status == 0) {
181 free(dir);
182 return 0;
183 } else if (!create || (errno != ENOENT)) {
184 char errbuf[1024];
185 ERROR("change_basedir: chdir (%s): %s", dir,
186 sstrerror(errno, errbuf, sizeof(errbuf)));
187 free(dir);
188 return -1;
189 }
191 status = mkdir(dir, S_IRWXU | S_IRWXG | S_IRWXO);
192 if (status != 0) {
193 char errbuf[1024];
194 ERROR("change_basedir: mkdir (%s): %s", dir,
195 sstrerror(errno, errbuf, sizeof(errbuf)));
196 free(dir);
197 return -1;
198 }
200 status = chdir(dir);
201 if (status != 0) {
202 char errbuf[1024];
203 ERROR("change_basedir: chdir (%s): %s", dir,
204 sstrerror(errno, errbuf, sizeof(errbuf)));
205 free(dir);
206 return -1;
207 }
209 free(dir);
210 return 0;
211 } /* static int change_basedir (char *dir) */
213 #if HAVE_LIBKSTAT
214 static void update_kstat(void) {
215 if (kc == NULL) {
216 if ((kc = kstat_open()) == NULL)
217 ERROR("Unable to open kstat control structure");
218 } else {
219 kid_t kid;
220 kid = kstat_chain_update(kc);
221 if (kid > 0) {
222 INFO("kstat chain has been updated");
223 plugin_init_all();
224 } else if (kid < 0)
225 ERROR("kstat chain update failed");
226 /* else: everything works as expected */
227 }
229 return;
230 } /* static void update_kstat (void) */
231 #endif /* HAVE_LIBKSTAT */
233 /* TODO
234 * Remove all settings but `-f' and `-C'
235 */
236 __attribute__((noreturn)) static void exit_usage(int status) {
237 printf("Usage: " PACKAGE_NAME " [OPTIONS]\n\n"
239 "Available options:\n"
240 " General:\n"
241 " -C <file> Configuration file.\n"
242 " Default: " CONFIGFILE "\n"
243 " -t Test config and exit.\n"
244 " -T Test plugin read and exit.\n"
245 " -P <file> PID-file.\n"
246 " Default: " PIDFILE "\n"
247 #if COLLECT_DAEMON
248 " -f Don't fork to the background.\n"
249 #endif
250 " -B Don't create the BaseDir\n"
251 " -h Display help (this message)\n"
252 "\nBuiltin defaults:\n"
253 " Config file " CONFIGFILE "\n"
254 " PID file " PIDFILE "\n"
255 " Plugin directory " PLUGINDIR "\n"
256 " Data directory " PKGLOCALSTATEDIR "\n"
257 "\n" PACKAGE_NAME " " PACKAGE_VERSION ", http://collectd.org/\n"
258 "by Florian octo Forster <octo@collectd.org>\n"
259 "for contributions see `AUTHORS'\n");
260 exit(status);
261 } /* static void exit_usage (int status) */
263 static int do_init(void) {
264 #if HAVE_SETLOCALE
265 if (setlocale(LC_NUMERIC, COLLECTD_LOCALE) == NULL)
266 WARNING("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
268 /* Update the environment, so that libraries that are calling
269 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
270 unsetenv("LC_ALL");
271 setenv("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
272 #endif
274 #if HAVE_LIBKSTAT
275 kc = NULL;
276 update_kstat();
277 #endif
279 #if HAVE_LIBSTATGRAB
280 if (sg_init(
281 #if HAVE_LIBSTATGRAB_0_90
282 0
283 #endif
284 )) {
285 ERROR("sg_init: %s", sg_str_error(sg_get_error()));
286 return -1;
287 }
289 if (sg_drop_privileges()) {
290 ERROR("sg_drop_privileges: %s", sg_str_error(sg_get_error()));
291 return -1;
292 }
293 #endif
295 return plugin_init_all();
296 } /* int do_init () */
298 static int do_loop(void) {
299 cdtime_t interval = cf_get_default_interval();
300 cdtime_t wait_until;
302 wait_until = cdtime() + interval;
304 while (loop == 0) {
305 cdtime_t now;
307 #if HAVE_LIBKSTAT
308 update_kstat();
309 #endif
311 /* Issue all plugins */
312 plugin_read_all();
314 now = cdtime();
315 if (now >= wait_until) {
316 WARNING("Not sleeping because the next interval is "
317 "%.3f seconds in the past!",
318 CDTIME_T_TO_DOUBLE(now - wait_until));
319 wait_until = now + interval;
320 continue;
321 }
323 struct timespec ts_wait = CDTIME_T_TO_TIMESPEC(wait_until - now);
324 wait_until = wait_until + interval;
326 while ((loop == 0) && (nanosleep(&ts_wait, &ts_wait) != 0)) {
327 if (errno != EINTR) {
328 char errbuf[1024];
329 ERROR("nanosleep failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
330 return -1;
331 }
332 }
333 } /* while (loop == 0) */
335 return 0;
336 } /* int do_loop */
338 static int do_shutdown(void) {
339 return plugin_shutdown_all();
340 } /* int do_shutdown */
342 #if COLLECT_DAEMON
343 static int pidfile_create(void) {
344 FILE *fh;
345 const char *file = global_option_get("PIDFile");
347 if ((fh = fopen(file, "w")) == NULL) {
348 char errbuf[1024];
349 ERROR("fopen (%s): %s", file, sstrerror(errno, errbuf, sizeof(errbuf)));
350 return 1;
351 }
353 fprintf(fh, "%i\n", (int)getpid());
354 fclose(fh);
356 return 0;
357 } /* static int pidfile_create (const char *file) */
359 static int pidfile_remove(void) {
360 const char *file = global_option_get("PIDFile");
361 if (file == NULL)
362 return 0;
364 return unlink(file);
365 } /* static int pidfile_remove (const char *file) */
366 #endif /* COLLECT_DAEMON */
368 #ifdef KERNEL_LINUX
369 static int notify_upstart(void) {
370 char const *upstart_job = getenv("UPSTART_JOB");
372 if (upstart_job == NULL)
373 return 0;
375 if (strcmp(upstart_job, "collectd") != 0) {
376 WARNING("Environment specifies unexpected UPSTART_JOB=\"%s\", expected "
377 "\"collectd\". Ignoring the variable.",
378 upstart_job);
379 return 0;
380 }
382 NOTICE("Upstart detected, stopping now to signal readyness.");
383 raise(SIGSTOP);
384 unsetenv("UPSTART_JOB");
386 return 1;
387 }
389 static int notify_systemd(void) {
390 int fd;
391 const char *notifysocket;
392 struct sockaddr_un su = {0};
393 size_t su_size;
394 char buffer[] = "READY=1\n";
396 notifysocket = getenv("NOTIFY_SOCKET");
397 if (notifysocket == NULL)
398 return 0;
400 if ((strlen(notifysocket) < 2) ||
401 ((notifysocket[0] != '@') && (notifysocket[0] != '/'))) {
402 ERROR("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be "
403 "absolute",
404 notifysocket);
405 return 0;
406 }
407 NOTICE("Systemd detected, trying to signal readyness.");
409 unsetenv("NOTIFY_SOCKET");
411 #if defined(SOCK_CLOEXEC)
412 fd = socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, /* protocol = */ 0);
413 #else
414 fd = socket(AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
415 #endif
416 if (fd < 0) {
417 char errbuf[1024];
418 ERROR("creating UNIX socket failed: %s",
419 sstrerror(errno, errbuf, sizeof(errbuf)));
420 return 0;
421 }
423 su.sun_family = AF_UNIX;
424 if (notifysocket[0] != '@') {
425 /* regular UNIX socket */
426 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
427 su_size = sizeof(su);
428 } else {
429 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
430 * start with a null byte. Since null bytes have no special meaning in
431 * that case, we have to set su_size correctly to cover only the bytes
432 * that are part of the address. */
433 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
434 su.sun_path[0] = 0;
435 su_size = sizeof(sa_family_t) + strlen(notifysocket);
436 if (su_size > sizeof(su))
437 su_size = sizeof(su);
438 }
440 if (sendto(fd, buffer, strlen(buffer), MSG_NOSIGNAL, (void *)&su,
441 (socklen_t)su_size) < 0) {
442 char errbuf[1024];
443 ERROR("sendto(\"%s\") failed: %s", notifysocket,
444 sstrerror(errno, errbuf, sizeof(errbuf)));
445 close(fd);
446 return 0;
447 }
449 unsetenv("NOTIFY_SOCKET");
450 close(fd);
451 return 1;
452 }
453 #endif /* KERNEL_LINUX */
455 struct cmdline_config {
456 _Bool test_config;
457 _Bool test_readall;
458 _Bool create_basedir;
459 const char *configfile;
460 _Bool daemonize;
461 };
463 void read_cmdline(int argc, char **argv, struct cmdline_config *config) {
464 /* read options */
465 while (1) {
466 int c;
467 c = getopt(argc, argv,
468 "htTC:"
469 #if COLLECT_DAEMON
470 "fP:"
471 #endif
472 );
474 if (c == -1)
475 break;
477 switch (c) {
478 case 'B':
479 config->create_basedir = 0;
480 break;
481 case 'C':
482 config->configfile = optarg;
483 break;
484 case 't':
485 config->test_config = 1;
486 break;
487 case 'T':
488 config->test_readall = 1;
489 global_option_set("ReadThreads", "-1", 1);
490 #if COLLECT_DAEMON
491 config->daemonize = 0;
492 #endif /* COLLECT_DAEMON */
493 break;
494 #if COLLECT_DAEMON
495 case 'P':
496 global_option_set("PIDFile", optarg, 1);
497 break;
498 case 'f':
499 config->daemonize = 0;
500 break;
501 #endif /* COLLECT_DAEMON */
502 case 'h':
503 exit_usage(0);
504 break;
505 default:
506 exit_usage(1);
507 } /* switch (c) */
508 } /* while (1) */
509 }
511 int configure_collectd(struct cmdline_config *config) {
512 const char *basedir;
513 /*
514 * Read options from the config file, the environment and the command
515 * line (in that order, with later options overwriting previous ones in
516 * general).
517 * Also, this will automatically load modules.
518 */
519 if (cf_read(config->configfile)) {
520 fprintf(stderr, "Error: Reading the config file failed!\n"
521 "Read the logs for details.\n");
522 return 1;
523 }
525 /*
526 * Change directory. We do this _after_ reading the config and loading
527 * modules to relative paths work as expected.
528 */
529 if ((basedir = global_option_get("BaseDir")) == NULL) {
530 fprintf(stderr,
531 "Don't have a basedir to use. This should not happen. Ever.");
532 return 1;
533 } else if (change_basedir(basedir, config->create_basedir)) {
534 fprintf(stderr, "Error: Unable to change to directory `%s'.\n", basedir);
535 return 1;
536 }
538 /*
539 * Set global variables or, if that fails, exit. We cannot run with
540 * them being uninitialized. If nothing is configured, then defaults
541 * are being used. So this means that the user has actually done
542 * something wrong.
543 */
544 if (init_global_variables() != 0)
545 return 1;
547 return 0;
548 }
550 int main(int argc, char **argv) {
551 #if COLLECT_DAEMON
552 pid_t pid;
553 #endif
554 int exit_status = 0;
556 struct cmdline_config config = {
557 .daemonize = 1,
558 .create_basedir = 1,
559 .configfile = CONFIGFILE,
560 };
562 read_cmdline(argc, argv, &config);
564 if (config.test_config)
565 return 0;
567 if (optind < argc)
568 exit_usage(1);
570 plugin_init_ctx();
572 int status;
573 if ((status = configure_collectd(&config)) != 0)
574 exit(EXIT_FAILURE);
576 #if COLLECT_DAEMON
577 /*
578 * fork off child
579 */
580 struct sigaction sig_chld_action = {.sa_handler = SIG_IGN};
582 sigaction(SIGCHLD, &sig_chld_action, NULL);
584 /*
585 * Only daemonize if we're not being supervised
586 * by upstart or systemd (when using Linux).
587 */
588 if (config.daemonize
589 #ifdef KERNEL_LINUX
590 && notify_upstart() == 0 && notify_systemd() == 0
591 #endif
592 ) {
593 int status;
595 if ((pid = fork()) == -1) {
596 /* error */
597 char errbuf[1024];
598 fprintf(stderr, "fork: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
599 return 1;
600 } else if (pid != 0) {
601 /* parent */
602 /* printf ("Running (PID %i)\n", pid); */
603 return 0;
604 }
606 /* Detach from session */
607 setsid();
609 /* Write pidfile */
610 if (pidfile_create())
611 exit(2);
613 /* close standard descriptors */
614 close(2);
615 close(1);
616 close(0);
618 status = open("/dev/null", O_RDWR);
619 if (status != 0) {
620 ERROR("Error: Could not connect `STDIN' to `/dev/null' (status %d)",
621 status);
622 return 1;
623 }
625 status = dup(0);
626 if (status != 1) {
627 ERROR("Error: Could not connect `STDOUT' to `/dev/null' (status %d)",
628 status);
629 return 1;
630 }
632 status = dup(0);
633 if (status != 2) {
634 ERROR("Error: Could not connect `STDERR' to `/dev/null', (status %d)",
635 status);
636 return 1;
637 }
638 } /* if (config.daemonize) */
639 #endif /* COLLECT_DAEMON */
641 struct sigaction sig_pipe_action = {.sa_handler = SIG_IGN};
643 sigaction(SIGPIPE, &sig_pipe_action, NULL);
645 /*
646 * install signal handlers
647 */
648 struct sigaction sig_int_action = {.sa_handler = sig_int_handler};
650 if (0 != sigaction(SIGINT, &sig_int_action, NULL)) {
651 char errbuf[1024];
652 ERROR("Error: Failed to install a signal handler for signal INT: %s",
653 sstrerror(errno, errbuf, sizeof(errbuf)));
654 return 1;
655 }
657 struct sigaction sig_term_action = {.sa_handler = sig_term_handler};
659 if (0 != sigaction(SIGTERM, &sig_term_action, NULL)) {
660 char errbuf[1024];
661 ERROR("Error: Failed to install a signal handler for signal TERM: %s",
662 sstrerror(errno, errbuf, sizeof(errbuf)));
663 return 1;
664 }
666 struct sigaction sig_usr1_action = {.sa_handler = sig_usr1_handler};
668 if (0 != sigaction(SIGUSR1, &sig_usr1_action, NULL)) {
669 char errbuf[1024];
670 ERROR("Error: Failed to install a signal handler for signal USR1: %s",
671 sstrerror(errno, errbuf, sizeof(errbuf)));
672 return 1;
673 }
675 /*
676 * run the actual loops
677 */
678 if (do_init() != 0) {
679 ERROR("Error: one or more plugin init callbacks failed.");
680 exit_status = 1;
681 }
683 if (config.test_readall) {
684 if (plugin_read_all_once() != 0) {
685 ERROR("Error: one or more plugin read callbacks failed.");
686 exit_status = 1;
687 }
688 } else {
689 INFO("Initialization complete, entering read-loop.");
690 do_loop();
691 }
693 /* close syslog */
694 INFO("Exiting normally.");
696 if (do_shutdown() != 0) {
697 ERROR("Error: one or more plugin shutdown callbacks failed.");
698 exit_status = 1;
699 }
701 #if COLLECT_DAEMON
702 if (config.daemonize)
703 pidfile_remove();
704 #endif /* COLLECT_DAEMON */
706 return exit_status;
707 } /* int main */