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 "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 int status;
103 str = global_option_get ("Hostname");
104 if ((str != NULL) && (str[0] != 0))
105 {
106 sstrncpy (hostname_g, str, sizeof (hostname_g));
107 return (0);
108 }
110 if (gethostname (hostname_g, sizeof (hostname_g)) != 0)
111 {
112 fprintf (stderr, "`gethostname' failed and no "
113 "hostname was configured.\n");
114 return (-1);
115 }
117 str = global_option_get ("FQDNLookup");
118 if (IS_FALSE (str))
119 return (0);
121 struct addrinfo ai_hints = {
122 .ai_flags = AI_CANONNAME
123 };
125 status = getaddrinfo (hostname_g, NULL, &ai_hints, &ai_list);
126 if (status != 0)
127 {
128 ERROR ("Looking up \"%s\" failed. You have set the "
129 "\"FQDNLookup\" option, but I cannot resolve "
130 "my hostname to a fully qualified domain "
131 "name. Please fix the network "
132 "configuration.", hostname_g);
133 return (-1);
134 }
136 for (struct addrinfo *ai_ptr = ai_list; ai_ptr != NULL; ai_ptr = ai_ptr->ai_next)
137 {
138 if (ai_ptr->ai_canonname == NULL)
139 continue;
141 sstrncpy (hostname_g, ai_ptr->ai_canonname, sizeof (hostname_g));
142 break;
143 }
145 freeaddrinfo (ai_list);
146 return (0);
147 } /* int init_hostname */
149 static int init_global_variables (void)
150 {
151 char const *str;
153 interval_g = cf_get_default_interval ();
154 assert (interval_g > 0);
155 DEBUG ("interval_g = %.3f;", CDTIME_T_TO_DOUBLE (interval_g));
157 str = global_option_get ("Timeout");
158 if (str == NULL)
159 str = "2";
160 timeout_g = atoi (str);
161 if (timeout_g <= 1)
162 {
163 fprintf (stderr, "Cannot set the timeout to a correct value.\n"
164 "Please check your settings.\n");
165 return (-1);
166 }
167 DEBUG ("timeout_g = %i;", timeout_g);
169 if (init_hostname () != 0)
170 return (-1);
171 DEBUG ("hostname_g = %s;", hostname_g);
173 return (0);
174 } /* int init_global_variables */
176 static int change_basedir (const char *orig_dir)
177 {
178 char *dir;
179 size_t dirlen;
180 int status;
182 dir = strdup (orig_dir);
183 if (dir == NULL)
184 {
185 char errbuf[1024];
186 ERROR ("strdup failed: %s",
187 sstrerror (errno, errbuf, sizeof (errbuf)));
188 return (-1);
189 }
191 dirlen = strlen (dir);
192 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
193 dir[--dirlen] = '\0';
195 if (dirlen == 0) {
196 free (dir);
197 return (-1);
198 }
200 status = chdir (dir);
201 if (status == 0)
202 {
203 free (dir);
204 return (0);
205 }
206 else if (errno != ENOENT)
207 {
208 char errbuf[1024];
209 ERROR ("change_basedir: chdir (%s): %s", dir,
210 sstrerror (errno, errbuf, sizeof (errbuf)));
211 free (dir);
212 return (-1);
213 }
215 status = mkdir (dir, S_IRWXU | S_IRWXG | S_IRWXO);
216 if (status != 0)
217 {
218 char errbuf[1024];
219 ERROR ("change_basedir: mkdir (%s): %s", dir,
220 sstrerror (errno, errbuf, sizeof (errbuf)));
221 free (dir);
222 return (-1);
223 }
225 status = chdir (dir);
226 if (status != 0)
227 {
228 char errbuf[1024];
229 ERROR ("change_basedir: chdir (%s): %s", dir,
230 sstrerror (errno, errbuf, sizeof (errbuf)));
231 free (dir);
232 return (-1);
233 }
235 free (dir);
236 return (0);
237 } /* static int change_basedir (char *dir) */
239 #if HAVE_LIBKSTAT
240 static void update_kstat (void)
241 {
242 if (kc == NULL)
243 {
244 if ((kc = kstat_open ()) == NULL)
245 ERROR ("Unable to open kstat control structure");
246 }
247 else
248 {
249 kid_t kid;
250 kid = kstat_chain_update (kc);
251 if (kid > 0)
252 {
253 INFO ("kstat chain has been updated");
254 plugin_init_all ();
255 }
256 else if (kid < 0)
257 ERROR ("kstat chain update failed");
258 /* else: everything works as expected */
259 }
261 return;
262 } /* static void update_kstat (void) */
263 #endif /* HAVE_LIBKSTAT */
265 /* TODO
266 * Remove all settings but `-f' and `-C'
267 */
268 __attribute__((noreturn))
269 static void exit_usage (int status)
270 {
271 printf ("Usage: "PACKAGE_NAME" [OPTIONS]\n\n"
273 "Available options:\n"
274 " General:\n"
275 " -C <file> Configuration file.\n"
276 " Default: "CONFIGFILE"\n"
277 " -t Test config and exit.\n"
278 " -T Test plugin read and exit.\n"
279 " -P <file> PID-file.\n"
280 " Default: "PIDFILE"\n"
281 #if COLLECT_DAEMON
282 " -f Don't fork to the background.\n"
283 #endif
284 " -h Display help (this message)\n"
285 "\nBuiltin defaults:\n"
286 " Config file "CONFIGFILE"\n"
287 " PID file "PIDFILE"\n"
288 " Plugin directory "PLUGINDIR"\n"
289 " Data directory "PKGLOCALSTATEDIR"\n"
290 "\n"PACKAGE_NAME" "PACKAGE_VERSION", http://collectd.org/\n"
291 "by Florian octo Forster <octo@collectd.org>\n"
292 "for contributions see `AUTHORS'\n");
293 exit (status);
294 } /* static void exit_usage (int status) */
296 static int do_init (void)
297 {
298 #if HAVE_SETLOCALE
299 if (setlocale (LC_NUMERIC, COLLECTD_LOCALE) == NULL)
300 WARNING ("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
302 /* Update the environment, so that libraries that are calling
303 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
304 unsetenv ("LC_ALL");
305 setenv ("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
306 #endif
308 #if HAVE_LIBKSTAT
309 kc = NULL;
310 update_kstat ();
311 #endif
313 #if HAVE_LIBSTATGRAB
314 if (sg_init (
315 # if HAVE_LIBSTATGRAB_0_90
316 0
317 # endif
318 ))
319 {
320 ERROR ("sg_init: %s", sg_str_error (sg_get_error ()));
321 return (-1);
322 }
324 if (sg_drop_privileges ())
325 {
326 ERROR ("sg_drop_privileges: %s", sg_str_error (sg_get_error ()));
327 return (-1);
328 }
329 #endif
331 return plugin_init_all ();
332 } /* int do_init () */
335 static int do_loop (void)
336 {
337 cdtime_t interval = cf_get_default_interval ();
338 cdtime_t wait_until;
340 wait_until = cdtime () + interval;
342 while (loop == 0)
343 {
344 cdtime_t now;
346 #if HAVE_LIBKSTAT
347 update_kstat ();
348 #endif
350 /* Issue all plugins */
351 plugin_read_all ();
353 now = cdtime ();
354 if (now >= wait_until)
355 {
356 WARNING ("Not sleeping because the next interval is "
357 "%.3f seconds in the past!",
358 CDTIME_T_TO_DOUBLE (now - wait_until));
359 wait_until = now + interval;
360 continue;
361 }
363 struct timespec ts_wait = CDTIME_T_TO_TIMESPEC (wait_until - now);
364 wait_until = wait_until + interval;
366 while ((loop == 0) && (nanosleep (&ts_wait, &ts_wait) != 0))
367 {
368 if (errno != EINTR)
369 {
370 char errbuf[1024];
371 ERROR ("nanosleep failed: %s",
372 sstrerror (errno, errbuf,
373 sizeof (errbuf)));
374 return (-1);
375 }
376 }
377 } /* while (loop == 0) */
379 return (0);
380 } /* int do_loop */
382 static int do_shutdown (void)
383 {
384 return plugin_shutdown_all ();
385 } /* int do_shutdown */
387 #if COLLECT_DAEMON
388 static int pidfile_create (void)
389 {
390 FILE *fh;
391 const char *file = global_option_get ("PIDFile");
393 if ((fh = fopen (file, "w")) == NULL)
394 {
395 char errbuf[1024];
396 ERROR ("fopen (%s): %s", file,
397 sstrerror (errno, errbuf, sizeof (errbuf)));
398 return (1);
399 }
401 fprintf (fh, "%i\n", (int) getpid ());
402 fclose(fh);
404 return (0);
405 } /* static int pidfile_create (const char *file) */
407 static int pidfile_remove (void)
408 {
409 const char *file = global_option_get ("PIDFile");
410 if (file == NULL)
411 return 0;
413 return (unlink (file));
414 } /* static int pidfile_remove (const char *file) */
415 #endif /* COLLECT_DAEMON */
417 #ifdef KERNEL_LINUX
418 static int notify_upstart (void)
419 {
420 char const *upstart_job = getenv("UPSTART_JOB");
422 if (upstart_job == NULL)
423 return 0;
425 if (strcmp(upstart_job, "collectd") != 0)
426 {
427 WARNING ("Environment specifies unexpected UPSTART_JOB=\"%s\", expected \"collectd\". Ignoring the variable.", upstart_job);
428 return 0;
429 }
431 NOTICE("Upstart detected, stopping now to signal readyness.");
432 raise(SIGSTOP);
433 unsetenv("UPSTART_JOB");
435 return 1;
436 }
438 static int notify_systemd (void)
439 {
440 int fd;
441 const char *notifysocket;
442 struct sockaddr_un su = { 0 };
443 size_t su_size;
444 char buffer[] = "READY=1\n";
446 notifysocket = getenv ("NOTIFY_SOCKET");
447 if (notifysocket == NULL)
448 return 0;
450 if ((strlen (notifysocket) < 2)
451 || ((notifysocket[0] != '@') && (notifysocket[0] != '/')))
452 {
453 ERROR ("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be absolute", notifysocket);
454 return 0;
455 }
456 NOTICE ("Systemd detected, trying to signal readyness.");
458 unsetenv ("NOTIFY_SOCKET");
460 #if defined(SOCK_CLOEXEC)
461 fd = socket (AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, /* protocol = */ 0);
462 #else
463 fd = socket (AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
464 #endif
465 if (fd < 0) {
466 char errbuf[1024];
467 ERROR ("creating UNIX socket failed: %s",
468 sstrerror (errno, errbuf, sizeof (errbuf)));
469 return 0;
470 }
472 su.sun_family = AF_UNIX;
473 if (notifysocket[0] != '@')
474 {
475 /* regular UNIX socket */
476 sstrncpy (su.sun_path, notifysocket, sizeof (su.sun_path));
477 su_size = sizeof (su);
478 }
479 else
480 {
481 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
482 * start with a null byte. Since null bytes have no special meaning in
483 * that case, we have to set su_size correctly to cover only the bytes
484 * that are part of the address. */
485 sstrncpy (su.sun_path, notifysocket, sizeof (su.sun_path));
486 su.sun_path[0] = 0;
487 su_size = sizeof (sa_family_t) + strlen (notifysocket);
488 if (su_size > sizeof (su))
489 su_size = sizeof (su);
490 }
492 if (sendto (fd, buffer, strlen (buffer), MSG_NOSIGNAL, (void *) &su, (socklen_t) su_size) < 0)
493 {
494 char errbuf[1024];
495 ERROR ("sendto(\"%s\") failed: %s", notifysocket,
496 sstrerror (errno, errbuf, sizeof (errbuf)));
497 close(fd);
498 return 0;
499 }
501 unsetenv ("NOTIFY_SOCKET");
502 close(fd);
503 return 1;
504 }
505 #endif /* KERNEL_LINUX */
507 int main (int argc, char **argv)
508 {
509 const char *configfile = CONFIGFILE;
510 int test_config = 0;
511 int test_readall = 0;
512 const char *basedir;
513 #if COLLECT_DAEMON
514 pid_t pid;
515 int daemonize = 1;
516 #endif
517 int exit_status = 0;
519 /* read options */
520 while (1)
521 {
522 int c;
524 c = getopt (argc, argv, "htTC:"
525 #if COLLECT_DAEMON
526 "fP:"
527 #endif
528 );
530 if (c == -1)
531 break;
533 switch (c)
534 {
535 case 'C':
536 configfile = optarg;
537 break;
538 case 't':
539 test_config = 1;
540 break;
541 case 'T':
542 test_readall = 1;
543 global_option_set ("ReadThreads", "-1", 1);
544 #if COLLECT_DAEMON
545 daemonize = 0;
546 #endif /* COLLECT_DAEMON */
547 break;
548 #if COLLECT_DAEMON
549 case 'P':
550 global_option_set ("PIDFile", optarg, 1);
551 break;
552 case 'f':
553 daemonize = 0;
554 break;
555 #endif /* COLLECT_DAEMON */
556 case 'h':
557 exit_usage (0);
558 break;
559 default:
560 exit_usage (1);
561 } /* switch (c) */
562 } /* while (1) */
564 if (optind < argc)
565 exit_usage (1);
567 plugin_init_ctx ();
569 /*
570 * Read options from the config file, the environment and the command
571 * line (in that order, with later options overwriting previous ones in
572 * general).
573 * Also, this will automatically load modules.
574 */
575 if (cf_read (configfile))
576 {
577 fprintf (stderr, "Error: Reading the config file failed!\n"
578 "Read the syslog for details.\n");
579 return (1);
580 }
582 /*
583 * Change directory. We do this _after_ reading the config and loading
584 * modules to relative paths work as expected.
585 */
586 if ((basedir = global_option_get ("BaseDir")) == NULL)
587 {
588 fprintf (stderr, "Don't have a basedir to use. This should not happen. Ever.");
589 return (1);
590 }
591 else if (change_basedir (basedir))
592 {
593 fprintf (stderr, "Error: Unable to change to directory `%s'.\n", basedir);
594 return (1);
595 }
597 /*
598 * Set global variables or, if that failes, exit. We cannot run with
599 * them being uninitialized. If nothing is configured, then defaults
600 * are being used. So this means that the user has actually done
601 * something wrong.
602 */
603 if (init_global_variables () != 0)
604 exit (EXIT_FAILURE);
606 if (test_config)
607 return (0);
609 #if COLLECT_DAEMON
610 /*
611 * fork off child
612 */
613 struct sigaction sig_chld_action = {
614 .sa_handler = SIG_IGN
615 };
617 sigaction (SIGCHLD, &sig_chld_action, NULL);
619 /*
620 * Only daemonize if we're not being supervised
621 * by upstart or systemd (when using Linux).
622 */
623 if (daemonize
624 #ifdef KERNEL_LINUX
625 && notify_upstart() == 0 && notify_systemd() == 0
626 #endif
627 )
628 {
629 int status;
631 if ((pid = fork ()) == -1)
632 {
633 /* error */
634 char errbuf[1024];
635 fprintf (stderr, "fork: %s",
636 sstrerror (errno, errbuf,
637 sizeof (errbuf)));
638 return (1);
639 }
640 else if (pid != 0)
641 {
642 /* parent */
643 /* printf ("Running (PID %i)\n", pid); */
644 return (0);
645 }
647 /* Detach from session */
648 setsid ();
650 /* Write pidfile */
651 if (pidfile_create ())
652 exit (2);
654 /* close standard descriptors */
655 close (2);
656 close (1);
657 close (0);
659 status = open ("/dev/null", O_RDWR);
660 if (status != 0)
661 {
662 ERROR ("Error: Could not connect `STDIN' to `/dev/null' (status %d)", status);
663 return (1);
664 }
666 status = dup (0);
667 if (status != 1)
668 {
669 ERROR ("Error: Could not connect `STDOUT' to `/dev/null' (status %d)", status);
670 return (1);
671 }
673 status = dup (0);
674 if (status != 2)
675 {
676 ERROR ("Error: Could not connect `STDERR' to `/dev/null', (status %d)", status);
677 return (1);
678 }
679 } /* if (daemonize) */
680 #endif /* COLLECT_DAEMON */
682 struct sigaction sig_pipe_action = {
683 .sa_handler = SIG_IGN
684 };
686 sigaction (SIGPIPE, &sig_pipe_action, NULL);
688 /*
689 * install signal handlers
690 */
691 struct sigaction sig_int_action = {
692 .sa_handler = sig_int_handler
693 };
695 if (0 != sigaction (SIGINT, &sig_int_action, NULL)) {
696 char errbuf[1024];
697 ERROR ("Error: Failed to install a signal handler for signal INT: %s",
698 sstrerror (errno, errbuf, sizeof (errbuf)));
699 return (1);
700 }
702 struct sigaction sig_term_action = {
703 .sa_handler = sig_term_handler
704 };
706 if (0 != sigaction (SIGTERM, &sig_term_action, NULL)) {
707 char errbuf[1024];
708 ERROR ("Error: Failed to install a signal handler for signal TERM: %s",
709 sstrerror (errno, errbuf, sizeof (errbuf)));
710 return (1);
711 }
713 struct sigaction sig_usr1_action = {
714 .sa_handler = sig_usr1_handler
715 };
717 if (0 != sigaction (SIGUSR1, &sig_usr1_action, NULL)) {
718 char errbuf[1024];
719 ERROR ("Error: Failed to install a signal handler for signal USR1: %s",
720 sstrerror (errno, errbuf, sizeof (errbuf)));
721 return (1);
722 }
724 /*
725 * run the actual loops
726 */
727 if (do_init () != 0)
728 {
729 ERROR ("Error: one or more plugin init callbacks failed.");
730 exit_status = 1;
731 }
733 if (test_readall)
734 {
735 if (plugin_read_all_once () != 0)
736 {
737 ERROR ("Error: one or more plugin read callbacks failed.");
738 exit_status = 1;
739 }
740 }
741 else
742 {
743 INFO ("Initialization complete, entering read-loop.");
744 do_loop ();
745 }
747 /* close syslog */
748 INFO ("Exiting normally.");
750 if (do_shutdown () != 0)
751 {
752 ERROR ("Error: one or more plugin shutdown callbacks failed.");
753 exit_status = 1;
754 }
756 #if COLLECT_DAEMON
757 if (daemonize)
758 pidfile_remove ();
759 #endif /* COLLECT_DAEMON */
761 return (exit_status);
762 } /* int main */