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"
29 #include "common.h"
31 #include "plugin.h"
32 #include "configfile.h"
34 #include <sys/types.h>
35 #include <sys/un.h>
36 #include <netdb.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 /*
51 * Global variables
52 */
53 char hostname_g[DATA_MAX_NAME_LEN];
54 cdtime_t interval_g;
55 int timeout_g;
56 #if HAVE_LIBKSTAT
57 kstat_ctl_t *kc;
58 #endif /* HAVE_LIBKSTAT */
60 static int loop = 0;
62 static void *do_flush (void __attribute__((unused)) *arg)
63 {
64 INFO ("Flushing all data.");
65 plugin_flush (/* plugin = */ NULL,
66 /* timeout = */ 0,
67 /* ident = */ NULL);
68 INFO ("Finished flushing all data.");
69 pthread_exit (NULL);
70 return NULL;
71 }
73 static void sig_int_handler (int __attribute__((unused)) signal)
74 {
75 loop++;
76 }
78 static void sig_term_handler (int __attribute__((unused)) signal)
79 {
80 loop++;
81 }
83 static void sig_usr1_handler (int __attribute__((unused)) signal)
84 {
85 pthread_t thread;
86 pthread_attr_t attr;
88 /* flushing the data might take a while,
89 * so it should be done asynchronously */
90 pthread_attr_init (&attr);
91 pthread_attr_setdetachstate (&attr, PTHREAD_CREATE_DETACHED);
92 pthread_create (&thread, &attr, do_flush, NULL);
93 pthread_attr_destroy (&attr);
94 }
96 static int init_hostname (void)
97 {
98 const char *str;
100 struct addrinfo *ai_list;
101 struct addrinfo *ai_ptr;
102 int status;
104 str = global_option_get ("Hostname");
105 if (str != NULL)
106 {
107 sstrncpy (hostname_g, str, sizeof (hostname_g));
108 return (0);
109 }
111 if (gethostname (hostname_g, sizeof (hostname_g)) != 0)
112 {
113 fprintf (stderr, "`gethostname' failed and no "
114 "hostname was configured.\n");
115 return (-1);
116 }
118 str = global_option_get ("FQDNLookup");
119 if (IS_FALSE (str))
120 return (0);
122 struct addrinfo ai_hints = {
123 .ai_flags = AI_CANONNAME
124 };
126 status = getaddrinfo (hostname_g, NULL, &ai_hints, &ai_list);
127 if (status != 0)
128 {
129 ERROR ("Looking up \"%s\" failed. You have set the "
130 "\"FQDNLookup\" option, but I cannot resolve "
131 "my hostname to a fully qualified domain "
132 "name. Please fix the network "
133 "configuration.", hostname_g);
134 return (-1);
135 }
137 for (ai_ptr = ai_list; ai_ptr != NULL; ai_ptr = ai_ptr->ai_next)
138 {
139 if (ai_ptr->ai_canonname == NULL)
140 continue;
142 sstrncpy (hostname_g, ai_ptr->ai_canonname, sizeof (hostname_g));
143 break;
144 }
146 freeaddrinfo (ai_list);
147 return (0);
148 } /* int init_hostname */
150 static int init_global_variables (void)
151 {
152 char const *str;
154 interval_g = cf_get_default_interval ();
155 assert (interval_g > 0);
156 DEBUG ("interval_g = %.3f;", CDTIME_T_TO_DOUBLE (interval_g));
158 str = global_option_get ("Timeout");
159 if (str == NULL)
160 str = "2";
161 timeout_g = atoi (str);
162 if (timeout_g <= 1)
163 {
164 fprintf (stderr, "Cannot set the timeout to a correct value.\n"
165 "Please check your settings.\n");
166 return (-1);
167 }
168 DEBUG ("timeout_g = %i;", timeout_g);
170 if (init_hostname () != 0)
171 return (-1);
172 DEBUG ("hostname_g = %s;", hostname_g);
174 return (0);
175 } /* int init_global_variables */
177 static int change_basedir (const char *orig_dir)
178 {
179 char *dir;
180 size_t dirlen;
181 int status;
183 dir = strdup (orig_dir);
184 if (dir == NULL)
185 {
186 char errbuf[1024];
187 ERROR ("strdup failed: %s",
188 sstrerror (errno, errbuf, sizeof (errbuf)));
189 return (-1);
190 }
192 dirlen = strlen (dir);
193 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
194 dir[--dirlen] = '\0';
196 if (dirlen == 0) {
197 free (dir);
198 return (-1);
199 }
201 status = chdir (dir);
202 if (status == 0)
203 {
204 free (dir);
205 return (0);
206 }
207 else if (errno != ENOENT)
208 {
209 char errbuf[1024];
210 ERROR ("change_basedir: chdir (%s): %s", dir,
211 sstrerror (errno, errbuf, sizeof (errbuf)));
212 free (dir);
213 return (-1);
214 }
216 status = mkdir (dir, S_IRWXU | S_IRWXG | S_IRWXO);
217 if (status != 0)
218 {
219 char errbuf[1024];
220 ERROR ("change_basedir: mkdir (%s): %s", dir,
221 sstrerror (errno, errbuf, sizeof (errbuf)));
222 free (dir);
223 return (-1);
224 }
226 status = chdir (dir);
227 if (status != 0)
228 {
229 char errbuf[1024];
230 ERROR ("change_basedir: chdir (%s): %s", dir,
231 sstrerror (errno, errbuf, sizeof (errbuf)));
232 free (dir);
233 return (-1);
234 }
236 free (dir);
237 return (0);
238 } /* static int change_basedir (char *dir) */
240 #if HAVE_LIBKSTAT
241 static void update_kstat (void)
242 {
243 if (kc == NULL)
244 {
245 if ((kc = kstat_open ()) == NULL)
246 ERROR ("Unable to open kstat control structure");
247 }
248 else
249 {
250 kid_t kid;
251 kid = kstat_chain_update (kc);
252 if (kid > 0)
253 {
254 INFO ("kstat chain has been updated");
255 plugin_init_all ();
256 }
257 else if (kid < 0)
258 ERROR ("kstat chain update failed");
259 /* else: everything works as expected */
260 }
262 return;
263 } /* static void update_kstat (void) */
264 #endif /* HAVE_LIBKSTAT */
266 /* TODO
267 * Remove all settings but `-f' and `-C'
268 */
269 __attribute__((noreturn))
270 static void exit_usage (int status)
271 {
272 printf ("Usage: "PACKAGE_NAME" [OPTIONS]\n\n"
274 "Available options:\n"
275 " General:\n"
276 " -C <file> Configuration file.\n"
277 " Default: "CONFIGFILE"\n"
278 " -t Test config and exit.\n"
279 " -T Test plugin read and exit.\n"
280 " -P <file> PID-file.\n"
281 " Default: "PIDFILE"\n"
282 #if COLLECT_DAEMON
283 " -f Don't fork to the background.\n"
284 #endif
285 " -h Display help (this message)\n"
286 "\nBuiltin defaults:\n"
287 " Config file "CONFIGFILE"\n"
288 " PID file "PIDFILE"\n"
289 " Plugin directory "PLUGINDIR"\n"
290 " Data directory "PKGLOCALSTATEDIR"\n"
291 "\n"PACKAGE_NAME" "PACKAGE_VERSION", http://collectd.org/\n"
292 "by Florian octo Forster <octo@collectd.org>\n"
293 "for contributions see `AUTHORS'\n");
294 exit (status);
295 } /* static void exit_usage (int status) */
297 static int do_init (void)
298 {
299 #if HAVE_SETLOCALE
300 if (setlocale (LC_NUMERIC, COLLECTD_LOCALE) == NULL)
301 WARNING ("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
303 /* Update the environment, so that libraries that are calling
304 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
305 unsetenv ("LC_ALL");
306 setenv ("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
307 #endif
309 #if HAVE_LIBKSTAT
310 kc = NULL;
311 update_kstat ();
312 #endif
314 #if HAVE_LIBSTATGRAB
315 if (sg_init (
316 # if HAVE_LIBSTATGRAB_0_90
317 0
318 # endif
319 ))
320 {
321 ERROR ("sg_init: %s", sg_str_error (sg_get_error ()));
322 return (-1);
323 }
325 if (sg_drop_privileges ())
326 {
327 ERROR ("sg_drop_privileges: %s", sg_str_error (sg_get_error ()));
328 return (-1);
329 }
330 #endif
332 return plugin_init_all ();
333 } /* int do_init () */
336 static int do_loop (void)
337 {
338 cdtime_t interval = cf_get_default_interval ();
339 cdtime_t wait_until;
341 wait_until = cdtime () + interval;
343 while (loop == 0)
344 {
345 struct timespec ts_wait = { 0, 0 };
346 cdtime_t now;
348 #if HAVE_LIBKSTAT
349 update_kstat ();
350 #endif
352 /* Issue all plugins */
353 plugin_read_all ();
355 now = cdtime ();
356 if (now >= wait_until)
357 {
358 WARNING ("Not sleeping because the next interval is "
359 "%.3f seconds in the past!",
360 CDTIME_T_TO_DOUBLE (now - wait_until));
361 wait_until = now + interval;
362 continue;
363 }
365 CDTIME_T_TO_TIMESPEC (wait_until - now, &ts_wait);
366 wait_until = wait_until + interval;
368 while ((loop == 0) && (nanosleep (&ts_wait, &ts_wait) != 0))
369 {
370 if (errno != EINTR)
371 {
372 char errbuf[1024];
373 ERROR ("nanosleep failed: %s",
374 sstrerror (errno, errbuf,
375 sizeof (errbuf)));
376 return (-1);
377 }
378 }
379 } /* while (loop == 0) */
381 return (0);
382 } /* int do_loop */
384 static int do_shutdown (void)
385 {
386 return plugin_shutdown_all ();
387 } /* int do_shutdown */
389 #if COLLECT_DAEMON
390 static int pidfile_create (void)
391 {
392 FILE *fh;
393 const char *file = global_option_get ("PIDFile");
395 if ((fh = fopen (file, "w")) == NULL)
396 {
397 char errbuf[1024];
398 ERROR ("fopen (%s): %s", file,
399 sstrerror (errno, errbuf, sizeof (errbuf)));
400 return (1);
401 }
403 fprintf (fh, "%i\n", (int) getpid ());
404 fclose(fh);
406 return (0);
407 } /* static int pidfile_create (const char *file) */
409 static int pidfile_remove (void)
410 {
411 const char *file = global_option_get ("PIDFile");
412 if (file == NULL)
413 return 0;
415 return (unlink (file));
416 } /* static int pidfile_remove (const char *file) */
417 #endif /* COLLECT_DAEMON */
419 #ifdef KERNEL_LINUX
420 static int notify_upstart (void)
421 {
422 char const *upstart_job = getenv("UPSTART_JOB");
424 if (upstart_job == NULL)
425 return 0;
427 if (strcmp(upstart_job, "collectd") != 0)
428 {
429 WARNING ("Environment specifies unexpected UPSTART_JOB=\"%s\", expected \"collectd\". Ignoring the variable.", upstart_job);
430 return 0;
431 }
433 NOTICE("Upstart detected, stopping now to signal readyness.");
434 raise(SIGSTOP);
435 unsetenv("UPSTART_JOB");
437 return 1;
438 }
440 static int notify_systemd (void)
441 {
442 int fd;
443 const char *notifysocket;
444 struct sockaddr_un su = { 0 };
445 size_t su_size;
446 char buffer[] = "READY=1\n";
448 notifysocket = getenv ("NOTIFY_SOCKET");
449 if (notifysocket == NULL)
450 return 0;
452 if ((strlen (notifysocket) < 2)
453 || ((notifysocket[0] != '@') && (notifysocket[0] != '/')))
454 {
455 ERROR ("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be absolute", notifysocket);
456 return 0;
457 }
458 NOTICE ("Systemd detected, trying to signal readyness.");
460 unsetenv ("NOTIFY_SOCKET");
462 #if defined(SOCK_CLOEXEC)
463 fd = socket (AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, /* protocol = */ 0);
464 #else
465 fd = socket (AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
466 #endif
467 if (fd < 0) {
468 char errbuf[1024];
469 ERROR ("creating UNIX socket failed: %s",
470 sstrerror (errno, errbuf, sizeof (errbuf)));
471 return 0;
472 }
474 su.sun_family = AF_UNIX;
475 if (notifysocket[0] != '@')
476 {
477 /* regular UNIX socket */
478 sstrncpy (su.sun_path, notifysocket, sizeof (su.sun_path));
479 su_size = sizeof (su);
480 }
481 else
482 {
483 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
484 * start with a null byte. Since null bytes have no special meaning in
485 * that case, we have to set su_size correctly to cover only the bytes
486 * that are part of the address. */
487 sstrncpy (su.sun_path, notifysocket, sizeof (su.sun_path));
488 su.sun_path[0] = 0;
489 su_size = sizeof (sa_family_t) + strlen (notifysocket);
490 if (su_size > sizeof (su))
491 su_size = sizeof (su);
492 }
494 if (sendto (fd, buffer, strlen (buffer), MSG_NOSIGNAL, (void *) &su, (socklen_t) su_size) < 0)
495 {
496 char errbuf[1024];
497 ERROR ("sendto(\"%s\") failed: %s", notifysocket,
498 sstrerror (errno, errbuf, sizeof (errbuf)));
499 close(fd);
500 return 0;
501 }
503 unsetenv ("NOTIFY_SOCKET");
504 close(fd);
505 return 1;
506 }
507 #endif /* KERNEL_LINUX */
509 int main (int argc, char **argv)
510 {
511 const char *configfile = CONFIGFILE;
512 int test_config = 0;
513 int test_readall = 0;
514 const char *basedir;
515 #if COLLECT_DAEMON
516 pid_t pid;
517 int daemonize = 1;
518 #endif
519 int exit_status = 0;
521 /* read options */
522 while (1)
523 {
524 int c;
526 c = getopt (argc, argv, "htTC:"
527 #if COLLECT_DAEMON
528 "fP:"
529 #endif
530 );
532 if (c == -1)
533 break;
535 switch (c)
536 {
537 case 'C':
538 configfile = optarg;
539 break;
540 case 't':
541 test_config = 1;
542 break;
543 case 'T':
544 test_readall = 1;
545 global_option_set ("ReadThreads", "-1", 1);
546 #if COLLECT_DAEMON
547 daemonize = 0;
548 #endif /* COLLECT_DAEMON */
549 break;
550 #if COLLECT_DAEMON
551 case 'P':
552 global_option_set ("PIDFile", optarg, 1);
553 break;
554 case 'f':
555 daemonize = 0;
556 break;
557 #endif /* COLLECT_DAEMON */
558 case 'h':
559 exit_usage (0);
560 break;
561 default:
562 exit_usage (1);
563 } /* switch (c) */
564 } /* while (1) */
566 if (optind < argc)
567 exit_usage (1);
569 plugin_init_ctx ();
571 /*
572 * Read options from the config file, the environment and the command
573 * line (in that order, with later options overwriting previous ones in
574 * general).
575 * Also, this will automatically load modules.
576 */
577 if (cf_read (configfile))
578 {
579 fprintf (stderr, "Error: Reading the config file failed!\n"
580 "Read the syslog for details.\n");
581 return (1);
582 }
584 /*
585 * Change directory. We do this _after_ reading the config and loading
586 * modules to relative paths work as expected.
587 */
588 if ((basedir = global_option_get ("BaseDir")) == NULL)
589 {
590 fprintf (stderr, "Don't have a basedir to use. This should not happen. Ever.");
591 return (1);
592 }
593 else if (change_basedir (basedir))
594 {
595 fprintf (stderr, "Error: Unable to change to directory `%s'.\n", basedir);
596 return (1);
597 }
599 /*
600 * Set global variables or, if that failes, exit. We cannot run with
601 * them being uninitialized. If nothing is configured, then defaults
602 * are being used. So this means that the user has actually done
603 * something wrong.
604 */
605 if (init_global_variables () != 0)
606 return (1);
608 if (test_config)
609 return (0);
611 #if COLLECT_DAEMON
612 /*
613 * fork off child
614 */
615 struct sigaction sig_chld_action = {
616 .sa_handler = SIG_IGN
617 };
619 sigaction (SIGCHLD, &sig_chld_action, NULL);
621 /*
622 * Only daemonize if we're not being supervised
623 * by upstart or systemd (when using Linux).
624 */
625 if (daemonize
626 #ifdef KERNEL_LINUX
627 && notify_upstart() == 0 && notify_systemd() == 0
628 #endif
629 )
630 {
631 int status;
633 if ((pid = fork ()) == -1)
634 {
635 /* error */
636 char errbuf[1024];
637 fprintf (stderr, "fork: %s",
638 sstrerror (errno, errbuf,
639 sizeof (errbuf)));
640 return (1);
641 }
642 else if (pid != 0)
643 {
644 /* parent */
645 /* printf ("Running (PID %i)\n", pid); */
646 return (0);
647 }
649 /* Detach from session */
650 setsid ();
652 /* Write pidfile */
653 if (pidfile_create ())
654 exit (2);
656 /* close standard descriptors */
657 close (2);
658 close (1);
659 close (0);
661 status = open ("/dev/null", O_RDWR);
662 if (status != 0)
663 {
664 ERROR ("Error: Could not connect `STDIN' to `/dev/null' (status %d)", status);
665 return (1);
666 }
668 status = dup (0);
669 if (status != 1)
670 {
671 ERROR ("Error: Could not connect `STDOUT' to `/dev/null' (status %d)", status);
672 return (1);
673 }
675 status = dup (0);
676 if (status != 2)
677 {
678 ERROR ("Error: Could not connect `STDERR' to `/dev/null', (status %d)", status);
679 return (1);
680 }
681 } /* if (daemonize) */
682 #endif /* COLLECT_DAEMON */
684 struct sigaction sig_pipe_action = {
685 .sa_handler = SIG_IGN
686 };
688 sigaction (SIGPIPE, &sig_pipe_action, NULL);
690 /*
691 * install signal handlers
692 */
693 struct sigaction sig_int_action = {
694 .sa_handler = sig_int_handler
695 };
697 if (0 != sigaction (SIGINT, &sig_int_action, NULL)) {
698 char errbuf[1024];
699 ERROR ("Error: Failed to install a signal handler for signal INT: %s",
700 sstrerror (errno, errbuf, sizeof (errbuf)));
701 return (1);
702 }
704 struct sigaction sig_term_action = {
705 .sa_handler = sig_term_handler
706 };
708 if (0 != sigaction (SIGTERM, &sig_term_action, NULL)) {
709 char errbuf[1024];
710 ERROR ("Error: Failed to install a signal handler for signal TERM: %s",
711 sstrerror (errno, errbuf, sizeof (errbuf)));
712 return (1);
713 }
715 struct sigaction sig_usr1_action = {
716 .sa_handler = sig_usr1_handler
717 };
719 if (0 != sigaction (SIGUSR1, &sig_usr1_action, NULL)) {
720 char errbuf[1024];
721 ERROR ("Error: Failed to install a signal handler for signal USR1: %s",
722 sstrerror (errno, errbuf, sizeof (errbuf)));
723 return (1);
724 }
726 /*
727 * run the actual loops
728 */
729 if (do_init () != 0)
730 {
731 ERROR ("Error: one or more plugin init callbacks failed.");
732 exit_status = 1;
733 }
735 if (test_readall)
736 {
737 if (plugin_read_all_once () != 0)
738 {
739 ERROR ("Error: one or more plugin read callbacks failed.");
740 exit_status = 1;
741 }
742 }
743 else
744 {
745 INFO ("Initialization complete, entering read-loop.");
746 do_loop ();
747 }
749 /* close syslog */
750 INFO ("Exiting normally.");
752 if (do_shutdown () != 0)
753 {
754 ERROR ("Error: one or more plugin shutdown callbacks failed.");
755 exit_status = 1;
756 }
758 #if COLLECT_DAEMON
759 if (daemonize)
760 pidfile_remove ();
761 #endif /* COLLECT_DAEMON */
763 return (exit_status);
764 } /* int main */