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 #include <pthread.h>
40 #if HAVE_LOCALE_H
41 # include <locale.h>
42 #endif
44 #if HAVE_STATGRAB_H
45 # include <statgrab.h>
46 #endif
48 #ifndef COLLECTD_LOCALE
49 # define COLLECTD_LOCALE "C"
50 #endif
52 /*
53 * Global variables
54 */
55 char hostname_g[DATA_MAX_NAME_LEN];
56 cdtime_t interval_g;
57 int pidfile_from_cli = 0;
58 int timeout_g;
59 #if HAVE_LIBKSTAT
60 kstat_ctl_t *kc;
61 #endif /* HAVE_LIBKSTAT */
63 static int loop = 0;
65 static void *do_flush (void __attribute__((unused)) *arg)
66 {
67 INFO ("Flushing all data.");
68 plugin_flush (/* plugin = */ NULL,
69 /* timeout = */ 0,
70 /* ident = */ NULL);
71 INFO ("Finished flushing all data.");
72 pthread_exit (NULL);
73 return NULL;
74 }
76 static void sig_int_handler (int __attribute__((unused)) signal)
77 {
78 loop++;
79 }
81 static void sig_term_handler (int __attribute__((unused)) signal)
82 {
83 loop++;
84 }
86 static void sig_usr1_handler (int __attribute__((unused)) signal)
87 {
88 pthread_t thread;
89 pthread_attr_t attr;
91 /* flushing the data might take a while,
92 * so it should be done asynchronously */
93 pthread_attr_init (&attr);
94 pthread_attr_setdetachstate (&attr, PTHREAD_CREATE_DETACHED);
95 pthread_create (&thread, &attr, do_flush, NULL);
96 pthread_attr_destroy (&attr);
97 }
99 static int init_hostname (void)
100 {
101 const char *str;
103 struct addrinfo ai_hints;
104 struct addrinfo *ai_list;
105 struct addrinfo *ai_ptr;
106 int status;
108 str = global_option_get ("Hostname");
109 if (str != NULL)
110 {
111 sstrncpy (hostname_g, str, sizeof (hostname_g));
112 return (0);
113 }
115 if (gethostname (hostname_g, sizeof (hostname_g)) != 0)
116 {
117 fprintf (stderr, "`gethostname' failed and no "
118 "hostname was configured.\n");
119 return (-1);
120 }
122 str = global_option_get ("FQDNLookup");
123 if (IS_FALSE (str))
124 return (0);
126 memset (&ai_hints, '\0', sizeof (ai_hints));
127 ai_hints.ai_flags = AI_CANONNAME;
129 status = getaddrinfo (hostname_g, NULL, &ai_hints, &ai_list);
130 if (status != 0)
131 {
132 ERROR ("Looking up \"%s\" failed. You have set the "
133 "\"FQDNLookup\" option, but I cannot resolve "
134 "my hostname to a fully qualified domain "
135 "name. Please fix the network "
136 "configuration.", hostname_g);
137 return (-1);
138 }
140 for (ai_ptr = ai_list; ai_ptr != NULL; ai_ptr = ai_ptr->ai_next)
141 {
142 if (ai_ptr->ai_canonname == NULL)
143 continue;
145 sstrncpy (hostname_g, ai_ptr->ai_canonname, sizeof (hostname_g));
146 break;
147 }
149 freeaddrinfo (ai_list);
150 return (0);
151 } /* int init_hostname */
153 static int init_global_variables (void)
154 {
155 char const *str;
157 interval_g = cf_get_default_interval ();
158 assert (interval_g > 0);
159 DEBUG ("interval_g = %.3f;", CDTIME_T_TO_DOUBLE (interval_g));
161 str = global_option_get ("Timeout");
162 if (str == NULL)
163 str = "2";
164 timeout_g = atoi (str);
165 if (timeout_g <= 1)
166 {
167 fprintf (stderr, "Cannot set the timeout to a correct value.\n"
168 "Please check your settings.\n");
169 return (-1);
170 }
171 DEBUG ("timeout_g = %i;", timeout_g);
173 if (init_hostname () != 0)
174 return (-1);
175 DEBUG ("hostname_g = %s;", hostname_g);
177 return (0);
178 } /* int init_global_variables */
180 static int change_basedir (const char *orig_dir)
181 {
182 char *dir;
183 size_t dirlen;
184 int status;
186 dir = strdup (orig_dir);
187 if (dir == NULL)
188 {
189 char errbuf[1024];
190 ERROR ("strdup failed: %s",
191 sstrerror (errno, errbuf, sizeof (errbuf)));
192 return (-1);
193 }
195 dirlen = strlen (dir);
196 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
197 dir[--dirlen] = '\0';
199 if (dirlen == 0) {
200 free (dir);
201 return (-1);
202 }
204 status = chdir (dir);
205 if (status == 0)
206 {
207 free (dir);
208 return (0);
209 }
210 else if (errno != ENOENT)
211 {
212 char errbuf[1024];
213 ERROR ("change_basedir: chdir (%s): %s", dir,
214 sstrerror (errno, errbuf, sizeof (errbuf)));
215 free (dir);
216 return (-1);
217 }
219 status = mkdir (dir, S_IRWXU | S_IRWXG | S_IRWXO);
220 if (status != 0)
221 {
222 char errbuf[1024];
223 ERROR ("change_basedir: mkdir (%s): %s", dir,
224 sstrerror (errno, errbuf, sizeof (errbuf)));
225 free (dir);
226 return (-1);
227 }
229 status = chdir (dir);
230 if (status != 0)
231 {
232 char errbuf[1024];
233 ERROR ("change_basedir: chdir (%s): %s", dir,
234 sstrerror (errno, errbuf, sizeof (errbuf)));
235 free (dir);
236 return (-1);
237 }
239 free (dir);
240 return (0);
241 } /* static int change_basedir (char *dir) */
243 #if HAVE_LIBKSTAT
244 static void update_kstat (void)
245 {
246 if (kc == NULL)
247 {
248 if ((kc = kstat_open ()) == NULL)
249 ERROR ("Unable to open kstat control structure");
250 }
251 else
252 {
253 kid_t kid;
254 kid = kstat_chain_update (kc);
255 if (kid > 0)
256 {
257 INFO ("kstat chain has been updated");
258 plugin_init_all ();
259 }
260 else if (kid < 0)
261 ERROR ("kstat chain update failed");
262 /* else: everything works as expected */
263 }
265 return;
266 } /* static void update_kstat (void) */
267 #endif /* HAVE_LIBKSTAT */
269 /* TODO
270 * Remove all settings but `-f' and `-C'
271 */
272 __attribute__((noreturn))
273 static void exit_usage (int status)
274 {
275 printf ("Usage: "PACKAGE_NAME" [OPTIONS]\n\n"
277 "Available options:\n"
278 " General:\n"
279 " -C <file> Configuration file.\n"
280 " Default: "CONFIGFILE"\n"
281 " -t Test config and exit.\n"
282 " -T Test plugin read and exit.\n"
283 " -P <file> PID-file.\n"
284 " Default: "PIDFILE"\n"
285 #if COLLECT_DAEMON
286 " -f Don't fork to the background.\n"
287 #endif
288 " -h Display help (this message)\n"
289 "\nBuiltin defaults:\n"
290 " Config file "CONFIGFILE"\n"
291 " PID file "PIDFILE"\n"
292 " Plugin directory "PLUGINDIR"\n"
293 " Data directory "PKGLOCALSTATEDIR"\n"
294 "\n"PACKAGE_NAME" "PACKAGE_VERSION", http://collectd.org/\n"
295 "by Florian octo Forster <octo@collectd.org>\n"
296 "for contributions see `AUTHORS'\n");
297 exit (status);
298 } /* static void exit_usage (int status) */
300 static int do_init (void)
301 {
302 #if HAVE_SETLOCALE
303 if (setlocale (LC_NUMERIC, COLLECTD_LOCALE) == NULL)
304 WARNING ("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
306 /* Update the environment, so that libraries that are calling
307 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
308 unsetenv ("LC_ALL");
309 setenv ("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
310 #endif
312 #if HAVE_LIBKSTAT
313 kc = NULL;
314 update_kstat ();
315 #endif
317 #if HAVE_LIBSTATGRAB
318 if (sg_init (
319 # if HAVE_LIBSTATGRAB_0_90
320 0
321 # endif
322 ))
323 {
324 ERROR ("sg_init: %s", sg_str_error (sg_get_error ()));
325 return (-1);
326 }
328 if (sg_drop_privileges ())
329 {
330 ERROR ("sg_drop_privileges: %s", sg_str_error (sg_get_error ()));
331 return (-1);
332 }
333 #endif
335 plugin_init_all ();
337 return (0);
338 } /* int do_init () */
341 static int do_loop (void)
342 {
343 cdtime_t interval = cf_get_default_interval ();
344 cdtime_t wait_until;
346 wait_until = cdtime () + interval;
348 while (loop == 0)
349 {
350 struct timespec ts_wait = { 0, 0 };
351 cdtime_t now;
353 #if HAVE_LIBKSTAT
354 update_kstat ();
355 #endif
357 /* Issue all plugins */
358 plugin_read_all ();
360 now = cdtime ();
361 if (now >= wait_until)
362 {
363 WARNING ("Not sleeping because the next interval is "
364 "%.3f seconds in the past!",
365 CDTIME_T_TO_DOUBLE (now - wait_until));
366 wait_until = now + interval;
367 continue;
368 }
370 CDTIME_T_TO_TIMESPEC (wait_until - now, &ts_wait);
371 wait_until = wait_until + interval;
373 while ((loop == 0) && (nanosleep (&ts_wait, &ts_wait) != 0))
374 {
375 if (errno != EINTR)
376 {
377 char errbuf[1024];
378 ERROR ("nanosleep failed: %s",
379 sstrerror (errno, errbuf,
380 sizeof (errbuf)));
381 return (-1);
382 }
383 }
384 } /* while (loop == 0) */
386 return (0);
387 } /* int do_loop */
389 static int do_shutdown (void)
390 {
391 plugin_shutdown_all ();
392 return (0);
393 } /* int do_shutdown */
395 #if COLLECT_DAEMON
396 static int pidfile_create (void)
397 {
398 FILE *fh;
399 const char *file = global_option_get ("PIDFile");
401 if ((fh = fopen (file, "w")) == NULL)
402 {
403 char errbuf[1024];
404 ERROR ("fopen (%s): %s", file,
405 sstrerror (errno, errbuf, sizeof (errbuf)));
406 return (1);
407 }
409 fprintf (fh, "%i\n", (int) getpid ());
410 fclose(fh);
412 return (0);
413 } /* static int pidfile_create (const char *file) */
415 static int pidfile_remove (void)
416 {
417 const char *file = global_option_get ("PIDFile");
418 if (file == NULL)
419 return 0;
421 return (unlink (file));
422 } /* static int pidfile_remove (const char *file) */
423 #endif /* COLLECT_DAEMON */
425 #ifdef KERNEL_LINUX
426 static int notify_upstart (void)
427 {
428 char const *upstart_job = getenv("UPSTART_JOB");
430 if (upstart_job == NULL)
431 return 0;
433 if (strcmp(upstart_job, "collectd") != 0)
434 {
435 WARNING ("Environment specifies unexpected UPSTART_JOB=\"%s\", expected \"collectd\". Ignoring the variable.", upstart_job);
436 return 0;
437 }
439 NOTICE("Upstart detected, stopping now to signal readyness.");
440 raise(SIGSTOP);
441 unsetenv("UPSTART_JOB");
443 return 1;
444 }
446 static int notify_systemd (void)
447 {
448 int fd;
449 const char *notifysocket;
450 struct sockaddr_un su;
451 size_t su_size;
452 char buffer[] = "READY=1\n";
454 notifysocket = getenv ("NOTIFY_SOCKET");
455 if (notifysocket == NULL)
456 return 0;
458 if ((strlen (notifysocket) < 2)
459 || ((notifysocket[0] != '@') && (notifysocket[0] != '/')))
460 {
461 ERROR ("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be absolute", notifysocket);
462 return 0;
463 }
464 NOTICE ("Systemd detected, trying to signal readyness.");
466 unsetenv ("NOTIFY_SOCKET");
468 #if defined(SOCK_CLOEXEC)
469 fd = socket (AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, /* protocol = */ 0);
470 #else
471 fd = socket (AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
472 #endif
473 if (fd < 0) {
474 char errbuf[1024];
475 ERROR ("creating UNIX socket failed: %s",
476 sstrerror (errno, errbuf, sizeof (errbuf)));
477 return 0;
478 }
480 memset (&su, 0, sizeof (su));
481 su.sun_family = AF_UNIX;
482 if (notifysocket[0] != '@')
483 {
484 /* regular UNIX socket */
485 sstrncpy (su.sun_path, notifysocket, sizeof (su.sun_path));
486 su_size = sizeof (su);
487 }
488 else
489 {
490 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
491 * start with a null byte. Since null bytes have no special meaning in
492 * that case, we have to set su_size correctly to cover only the bytes
493 * that are part of the address. */
494 sstrncpy (su.sun_path, notifysocket, sizeof (su.sun_path));
495 su.sun_path[0] = 0;
496 su_size = sizeof (sa_family_t) + strlen (notifysocket);
497 if (su_size > sizeof (su))
498 su_size = sizeof (su);
499 }
501 if (sendto (fd, buffer, strlen (buffer), MSG_NOSIGNAL, (void *) &su, (socklen_t) su_size) < 0)
502 {
503 char errbuf[1024];
504 ERROR ("sendto(\"%s\") failed: %s", notifysocket,
505 sstrerror (errno, errbuf, sizeof (errbuf)));
506 close(fd);
507 return 0;
508 }
510 unsetenv ("NOTIFY_SOCKET");
511 close(fd);
512 return 1;
513 }
514 #endif /* KERNEL_LINUX */
516 int main (int argc, char **argv)
517 {
518 struct sigaction sig_int_action;
519 struct sigaction sig_term_action;
520 struct sigaction sig_usr1_action;
521 struct sigaction sig_pipe_action;
522 const char *configfile = CONFIGFILE;
523 int test_config = 0;
524 int test_readall = 0;
525 const char *basedir;
526 #if COLLECT_DAEMON
527 struct sigaction sig_chld_action;
528 pid_t pid;
529 int daemonize = 1;
530 #endif
531 int exit_status = 0;
533 /* read options */
534 while (1)
535 {
536 int c;
538 c = getopt (argc, argv, "htTC:"
539 #if COLLECT_DAEMON
540 "fP:"
541 #endif
542 );
544 if (c == -1)
545 break;
547 switch (c)
548 {
549 case 'C':
550 configfile = optarg;
551 break;
552 case 't':
553 test_config = 1;
554 break;
555 case 'T':
556 test_readall = 1;
557 global_option_set ("ReadThreads", "-1");
558 #if COLLECT_DAEMON
559 daemonize = 0;
560 #endif /* COLLECT_DAEMON */
561 break;
562 #if COLLECT_DAEMON
563 case 'P':
564 global_option_set ("PIDFile", optarg);
565 pidfile_from_cli = 1;
566 break;
567 case 'f':
568 daemonize = 0;
569 break;
570 #endif /* COLLECT_DAEMON */
571 case 'h':
572 exit_usage (0);
573 break;
574 default:
575 exit_usage (1);
576 } /* switch (c) */
577 } /* while (1) */
579 if (optind < argc)
580 exit_usage (1);
582 plugin_init_ctx ();
584 /*
585 * Read options from the config file, the environment and the command
586 * line (in that order, with later options overwriting previous ones in
587 * general).
588 * Also, this will automatically load modules.
589 */
590 if (cf_read (configfile))
591 {
592 fprintf (stderr, "Error: Reading the config file failed!\n"
593 "Read the syslog for details.\n");
594 return (1);
595 }
597 /*
598 * Change directory. We do this _after_ reading the config and loading
599 * modules to relative paths work as expected.
600 */
601 if ((basedir = global_option_get ("BaseDir")) == NULL)
602 {
603 fprintf (stderr, "Don't have a basedir to use. This should not happen. Ever.");
604 return (1);
605 }
606 else if (change_basedir (basedir))
607 {
608 fprintf (stderr, "Error: Unable to change to directory `%s'.\n", basedir);
609 return (1);
610 }
612 /*
613 * Set global variables or, if that failes, exit. We cannot run with
614 * them being uninitialized. If nothing is configured, then defaults
615 * are being used. So this means that the user has actually done
616 * something wrong.
617 */
618 if (init_global_variables () != 0)
619 return (1);
621 if (test_config)
622 return (0);
624 #if COLLECT_DAEMON
625 /*
626 * fork off child
627 */
628 memset (&sig_chld_action, '\0', sizeof (sig_chld_action));
629 sig_chld_action.sa_handler = SIG_IGN;
630 sigaction (SIGCHLD, &sig_chld_action, NULL);
632 /*
633 * Only daemonize if we're not being supervised
634 * by upstart or systemd (when using Linux).
635 */
636 if (daemonize
637 #ifdef KERNEL_LINUX
638 && notify_upstart() == 0 && notify_systemd() == 0
639 #endif
640 )
641 {
642 int status;
644 if ((pid = fork ()) == -1)
645 {
646 /* error */
647 char errbuf[1024];
648 fprintf (stderr, "fork: %s",
649 sstrerror (errno, errbuf,
650 sizeof (errbuf)));
651 return (1);
652 }
653 else if (pid != 0)
654 {
655 /* parent */
656 /* printf ("Running (PID %i)\n", pid); */
657 return (0);
658 }
660 /* Detach from session */
661 setsid ();
663 /* Write pidfile */
664 if (pidfile_create ())
665 exit (2);
667 /* close standard descriptors */
668 close (2);
669 close (1);
670 close (0);
672 status = open ("/dev/null", O_RDWR);
673 if (status != 0)
674 {
675 ERROR ("Error: Could not connect `STDIN' to `/dev/null' (status %d)", status);
676 return (1);
677 }
679 status = dup (0);
680 if (status != 1)
681 {
682 ERROR ("Error: Could not connect `STDOUT' to `/dev/null' (status %d)", status);
683 return (1);
684 }
686 status = dup (0);
687 if (status != 2)
688 {
689 ERROR ("Error: Could not connect `STDERR' to `/dev/null', (status %d)", status);
690 return (1);
691 }
692 } /* if (daemonize) */
693 #endif /* COLLECT_DAEMON */
695 memset (&sig_pipe_action, '\0', sizeof (sig_pipe_action));
696 sig_pipe_action.sa_handler = SIG_IGN;
697 sigaction (SIGPIPE, &sig_pipe_action, NULL);
699 /*
700 * install signal handlers
701 */
702 memset (&sig_int_action, '\0', sizeof (sig_int_action));
703 sig_int_action.sa_handler = sig_int_handler;
704 if (0 != sigaction (SIGINT, &sig_int_action, NULL)) {
705 char errbuf[1024];
706 ERROR ("Error: Failed to install a signal handler for signal INT: %s",
707 sstrerror (errno, errbuf, sizeof (errbuf)));
708 return (1);
709 }
711 memset (&sig_term_action, '\0', sizeof (sig_term_action));
712 sig_term_action.sa_handler = sig_term_handler;
713 if (0 != sigaction (SIGTERM, &sig_term_action, NULL)) {
714 char errbuf[1024];
715 ERROR ("Error: Failed to install a signal handler for signal TERM: %s",
716 sstrerror (errno, errbuf, sizeof (errbuf)));
717 return (1);
718 }
720 memset (&sig_usr1_action, '\0', sizeof (sig_usr1_action));
721 sig_usr1_action.sa_handler = sig_usr1_handler;
722 if (0 != sigaction (SIGUSR1, &sig_usr1_action, NULL)) {
723 char errbuf[1024];
724 ERROR ("Error: Failed to install a signal handler for signal USR1: %s",
725 sstrerror (errno, errbuf, sizeof (errbuf)));
726 return (1);
727 }
729 /*
730 * run the actual loops
731 */
732 do_init ();
734 if (test_readall)
735 {
736 if (plugin_read_all_once () != 0)
737 exit_status = 1;
738 }
739 else
740 {
741 INFO ("Initialization complete, entering read-loop.");
742 do_loop ();
743 }
745 /* close syslog */
746 INFO ("Exiting normally.");
748 do_shutdown ();
750 #if COLLECT_DAEMON
751 if (daemonize)
752 pidfile_remove ();
753 #endif /* COLLECT_DAEMON */
755 return (exit_status);
756 } /* int main */