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 /*
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 INFO("Flushing all data.");
64 plugin_flush(/* plugin = */ NULL,
65 /* timeout = */ 0,
66 /* ident = */ NULL);
67 INFO("Finished flushing all data.");
68 pthread_exit(NULL);
69 return NULL;
70 }
72 static void sig_int_handler(int __attribute__((unused)) signal) { loop++; }
74 static void sig_term_handler(int __attribute__((unused)) signal) { loop++; }
76 static void sig_usr1_handler(int __attribute__((unused)) signal) {
77 pthread_t thread;
78 pthread_attr_t attr;
80 /* flushing the data might take a while,
81 * so it should be done asynchronously */
82 pthread_attr_init(&attr);
83 pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
84 pthread_create(&thread, &attr, do_flush, NULL);
85 pthread_attr_destroy(&attr);
86 }
88 static int init_hostname(void) {
89 const char *str;
91 struct addrinfo *ai_list;
92 int status;
94 str = global_option_get("Hostname");
95 if ((str != NULL) && (str[0] != 0)) {
96 sstrncpy(hostname_g, str, sizeof(hostname_g));
97 return 0;
98 }
100 if (gethostname(hostname_g, sizeof(hostname_g)) != 0) {
101 fprintf(stderr, "`gethostname' failed and no "
102 "hostname was configured.\n");
103 return -1;
104 }
106 str = global_option_get("FQDNLookup");
107 if (IS_FALSE(str))
108 return 0;
110 struct addrinfo ai_hints = {.ai_flags = AI_CANONNAME};
112 status = getaddrinfo(hostname_g, NULL, &ai_hints, &ai_list);
113 if (status != 0) {
114 ERROR("Looking up \"%s\" failed. You have set the "
115 "\"FQDNLookup\" option, but I cannot resolve "
116 "my hostname to a fully qualified domain "
117 "name. Please fix the network "
118 "configuration.",
119 hostname_g);
120 return -1;
121 }
123 for (struct addrinfo *ai_ptr = ai_list; ai_ptr != NULL;
124 ai_ptr = ai_ptr->ai_next) {
125 if (ai_ptr->ai_canonname == NULL)
126 continue;
128 sstrncpy(hostname_g, ai_ptr->ai_canonname, sizeof(hostname_g));
129 break;
130 }
132 freeaddrinfo(ai_list);
133 return 0;
134 } /* int init_hostname */
136 static int init_global_variables(void) {
137 char const *str;
139 interval_g = cf_get_default_interval();
140 assert(interval_g > 0);
141 DEBUG("interval_g = %.3f;", CDTIME_T_TO_DOUBLE(interval_g));
143 str = global_option_get("Timeout");
144 if (str == NULL)
145 str = "2";
146 timeout_g = atoi(str);
147 if (timeout_g <= 1) {
148 fprintf(stderr, "Cannot set the timeout to a correct value.\n"
149 "Please check your settings.\n");
150 return -1;
151 }
152 DEBUG("timeout_g = %i;", timeout_g);
154 if (init_hostname() != 0)
155 return -1;
156 DEBUG("hostname_g = %s;", hostname_g);
158 return 0;
159 } /* int init_global_variables */
161 static int change_basedir(const char *orig_dir, int nocreate) {
162 char *dir;
163 size_t dirlen;
164 int status;
166 dir = strdup(orig_dir);
167 if (dir == NULL) {
168 char errbuf[1024];
169 ERROR("strdup failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
170 return -1;
171 }
173 dirlen = strlen(dir);
174 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
175 dir[--dirlen] = '\0';
177 if (dirlen == 0) {
178 free(dir);
179 return -1;
180 }
182 status = chdir(dir);
183 if (status == 0) {
184 free(dir);
185 return 0;
186 } else if (errno != ENOENT) {
187 char errbuf[1024];
188 ERROR("change_basedir: chdir (%s): %s", dir,
189 sstrerror(errno, errbuf, sizeof(errbuf)));
190 free(dir);
191 return -1;
192 }
194 if (nocreate == 0) {
195 status = mkdir(dir, S_IRWXU | S_IRWXG | S_IRWXO);
196 if (status != 0) {
197 char errbuf[1024];
198 ERROR("change_basedir: mkdir (%s): %s", dir,
199 sstrerror(errno, errbuf, sizeof(errbuf)));
200 free(dir);
201 return -1;
202 }
203 }
205 status = chdir(dir);
206 if (status != 0) {
207 char errbuf[1024];
208 ERROR("change_basedir: chdir (%s): %s", dir,
209 sstrerror(errno, errbuf, sizeof(errbuf)));
210 free(dir);
211 return -1;
212 }
214 free(dir);
215 return 0;
216 } /* static int change_basedir (char *dir) */
218 #if HAVE_LIBKSTAT
219 static void update_kstat(void) {
220 if (kc == NULL) {
221 if ((kc = kstat_open()) == NULL)
222 ERROR("Unable to open kstat control structure");
223 } else {
224 kid_t kid;
225 kid = kstat_chain_update(kc);
226 if (kid > 0) {
227 INFO("kstat chain has been updated");
228 plugin_init_all();
229 } else if (kid < 0)
230 ERROR("kstat chain update failed");
231 /* else: everything works as expected */
232 }
234 return;
235 } /* static void update_kstat (void) */
236 #endif /* HAVE_LIBKSTAT */
238 /* TODO
239 * Remove all settings but `-f' and `-C'
240 */
241 __attribute__((noreturn)) static void exit_usage(int status) {
242 printf("Usage: " PACKAGE_NAME " [OPTIONS]\n\n"
244 "Available options:\n"
245 " General:\n"
246 " -C <file> Configuration file.\n"
247 " Default: " CONFIGFILE "\n"
248 " -t Test config and exit.\n"
249 " -T Test plugin read and exit.\n"
250 " -P <file> PID-file.\n"
251 " Default: " PIDFILE "\n"
252 #if COLLECT_DAEMON
253 " -f Don't fork to the background.\n"
254 #endif
255 " -B Don't create the BaseDir\n"
256 " -h Display help (this message)\n"
257 "\nBuiltin defaults:\n"
258 " Config file " CONFIGFILE "\n"
259 " PID file " PIDFILE "\n"
260 " Plugin directory " PLUGINDIR "\n"
261 " Data directory " PKGLOCALSTATEDIR "\n"
262 "\n" PACKAGE_NAME " " PACKAGE_VERSION ", http://collectd.org/\n"
263 "by Florian octo Forster <octo@collectd.org>\n"
264 "for contributions see `AUTHORS'\n");
265 exit(status);
266 } /* static void exit_usage (int status) */
268 static int do_init(void) {
269 #if HAVE_SETLOCALE
270 if (setlocale(LC_NUMERIC, COLLECTD_LOCALE) == NULL)
271 WARNING("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
273 /* Update the environment, so that libraries that are calling
274 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
275 unsetenv("LC_ALL");
276 setenv("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
277 #endif
279 #if HAVE_LIBKSTAT
280 kc = NULL;
281 update_kstat();
282 #endif
284 #if HAVE_LIBSTATGRAB
285 if (sg_init(
286 #if HAVE_LIBSTATGRAB_0_90
287 0
288 #endif
289 )) {
290 ERROR("sg_init: %s", sg_str_error(sg_get_error()));
291 return -1;
292 }
294 if (sg_drop_privileges()) {
295 ERROR("sg_drop_privileges: %s", sg_str_error(sg_get_error()));
296 return -1;
297 }
298 #endif
300 return plugin_init_all();
301 } /* int do_init () */
303 static int do_loop(void) {
304 cdtime_t interval = cf_get_default_interval();
305 cdtime_t wait_until;
307 wait_until = cdtime() + interval;
309 while (loop == 0) {
310 cdtime_t now;
312 #if HAVE_LIBKSTAT
313 update_kstat();
314 #endif
316 /* Issue all plugins */
317 plugin_read_all();
319 now = cdtime();
320 if (now >= wait_until) {
321 WARNING("Not sleeping because the next interval is "
322 "%.3f seconds in the past!",
323 CDTIME_T_TO_DOUBLE(now - wait_until));
324 wait_until = now + interval;
325 continue;
326 }
328 struct timespec ts_wait = CDTIME_T_TO_TIMESPEC(wait_until - now);
329 wait_until = wait_until + interval;
331 while ((loop == 0) && (nanosleep(&ts_wait, &ts_wait) != 0)) {
332 if (errno != EINTR) {
333 char errbuf[1024];
334 ERROR("nanosleep failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
335 return -1;
336 }
337 }
338 } /* while (loop == 0) */
340 return 0;
341 } /* int do_loop */
343 static int do_shutdown(void) {
344 return plugin_shutdown_all();
345 } /* int do_shutdown */
347 #if COLLECT_DAEMON
348 static int pidfile_create(void) {
349 FILE *fh;
350 const char *file = global_option_get("PIDFile");
352 if ((fh = fopen(file, "w")) == NULL) {
353 char errbuf[1024];
354 ERROR("fopen (%s): %s", file, sstrerror(errno, errbuf, sizeof(errbuf)));
355 return 1;
356 }
358 fprintf(fh, "%i\n", (int)getpid());
359 fclose(fh);
361 return 0;
362 } /* static int pidfile_create (const char *file) */
364 static int pidfile_remove(void) {
365 const char *file = global_option_get("PIDFile");
366 if (file == NULL)
367 return 0;
369 return unlink(file);
370 } /* static int pidfile_remove (const char *file) */
371 #endif /* COLLECT_DAEMON */
373 #ifdef KERNEL_LINUX
374 static int notify_upstart(void) {
375 char const *upstart_job = getenv("UPSTART_JOB");
377 if (upstart_job == NULL)
378 return 0;
380 if (strcmp(upstart_job, "collectd") != 0) {
381 WARNING("Environment specifies unexpected UPSTART_JOB=\"%s\", expected "
382 "\"collectd\". Ignoring the variable.",
383 upstart_job);
384 return 0;
385 }
387 NOTICE("Upstart detected, stopping now to signal readyness.");
388 raise(SIGSTOP);
389 unsetenv("UPSTART_JOB");
391 return 1;
392 }
394 static int notify_systemd(void) {
395 int fd;
396 const char *notifysocket;
397 struct sockaddr_un su = {0};
398 size_t su_size;
399 char buffer[] = "READY=1\n";
401 notifysocket = getenv("NOTIFY_SOCKET");
402 if (notifysocket == NULL)
403 return 0;
405 if ((strlen(notifysocket) < 2) ||
406 ((notifysocket[0] != '@') && (notifysocket[0] != '/'))) {
407 ERROR("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be "
408 "absolute",
409 notifysocket);
410 return 0;
411 }
412 NOTICE("Systemd detected, trying to signal readyness.");
414 unsetenv("NOTIFY_SOCKET");
416 #if defined(SOCK_CLOEXEC)
417 fd = socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, /* protocol = */ 0);
418 #else
419 fd = socket(AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
420 #endif
421 if (fd < 0) {
422 char errbuf[1024];
423 ERROR("creating UNIX socket failed: %s",
424 sstrerror(errno, errbuf, sizeof(errbuf)));
425 return 0;
426 }
428 su.sun_family = AF_UNIX;
429 if (notifysocket[0] != '@') {
430 /* regular UNIX socket */
431 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
432 su_size = sizeof(su);
433 } else {
434 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
435 * start with a null byte. Since null bytes have no special meaning in
436 * that case, we have to set su_size correctly to cover only the bytes
437 * that are part of the address. */
438 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
439 su.sun_path[0] = 0;
440 su_size = sizeof(sa_family_t) + strlen(notifysocket);
441 if (su_size > sizeof(su))
442 su_size = sizeof(su);
443 }
445 if (sendto(fd, buffer, strlen(buffer), MSG_NOSIGNAL, (void *)&su,
446 (socklen_t)su_size) < 0) {
447 char errbuf[1024];
448 ERROR("sendto(\"%s\") failed: %s", notifysocket,
449 sstrerror(errno, errbuf, sizeof(errbuf)));
450 close(fd);
451 return 0;
452 }
454 unsetenv("NOTIFY_SOCKET");
455 close(fd);
456 return 1;
457 }
458 #endif /* KERNEL_LINUX */
460 int main(int argc, char **argv) {
461 const char *configfile = CONFIGFILE;
462 int test_config = 0;
463 int test_readall = 0;
464 const char *basedir;
465 int basedir_nocreate = 0;
466 #if COLLECT_DAEMON
467 pid_t pid;
468 int daemonize = 1;
469 #endif
470 int exit_status = 0;
472 /* read options */
473 while (1) {
474 int c;
476 c = getopt(argc, argv, "BhtTC:"
477 #if COLLECT_DAEMON
478 "fP:"
479 #endif
480 );
482 if (c == -1)
483 break;
485 switch (c) {
486 case 'B':
487 basedir_nocreate = 1;
488 break;
489 case 'C':
490 configfile = optarg;
491 break;
492 case 't':
493 test_config = 1;
494 break;
495 case 'T':
496 test_readall = 1;
497 global_option_set("ReadThreads", "-1", 1);
498 #if COLLECT_DAEMON
499 daemonize = 0;
500 #endif /* COLLECT_DAEMON */
501 break;
502 #if COLLECT_DAEMON
503 case 'P':
504 global_option_set("PIDFile", optarg, 1);
505 break;
506 case 'f':
507 daemonize = 0;
508 break;
509 #endif /* COLLECT_DAEMON */
510 case 'h':
511 exit_usage(0);
512 break;
513 default:
514 exit_usage(1);
515 } /* switch (c) */
516 } /* while (1) */
518 if (optind < argc)
519 exit_usage(1);
521 plugin_init_ctx();
523 /*
524 * Read options from the config file, the environment and the command
525 * line (in that order, with later options overwriting previous ones in
526 * general).
527 * Also, this will automatically load modules.
528 */
529 if (cf_read(configfile)) {
530 fprintf(stderr, "Error: Reading the config file failed!\n"
531 "Read the logs for details.\n");
532 return 1;
533 }
535 /*
536 * Change directory. We do this _after_ reading the config and loading
537 * modules to relative paths work as expected.
538 */
539 if ((basedir = global_option_get("BaseDir")) == NULL) {
540 fprintf(stderr,
541 "Don't have a basedir to use. This should not happen. Ever.");
542 return 1;
543 } else if (change_basedir(basedir, basedir_nocreate)) {
544 fprintf(stderr, "Error: Unable to change to directory `%s'.\n", basedir);
545 return 1;
546 }
548 /*
549 * Set global variables or, if that failes, exit. We cannot run with
550 * them being uninitialized. If nothing is configured, then defaults
551 * are being used. So this means that the user has actually done
552 * something wrong.
553 */
554 if (init_global_variables() != 0)
555 exit(EXIT_FAILURE);
557 if (test_config)
558 return 0;
560 #if COLLECT_DAEMON
561 /*
562 * fork off child
563 */
564 struct sigaction sig_chld_action = {.sa_handler = SIG_IGN};
566 sigaction(SIGCHLD, &sig_chld_action, NULL);
568 /*
569 * Only daemonize if we're not being supervised
570 * by upstart or systemd (when using Linux).
571 */
572 if (daemonize
573 #ifdef KERNEL_LINUX
574 && notify_upstart() == 0 && notify_systemd() == 0
575 #endif
576 ) {
577 int status;
579 if ((pid = fork()) == -1) {
580 /* error */
581 char errbuf[1024];
582 fprintf(stderr, "fork: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
583 return 1;
584 } else if (pid != 0) {
585 /* parent */
586 /* printf ("Running (PID %i)\n", pid); */
587 return 0;
588 }
590 /* Detach from session */
591 setsid();
593 /* Write pidfile */
594 if (pidfile_create())
595 exit(2);
597 /* close standard descriptors */
598 close(2);
599 close(1);
600 close(0);
602 status = open("/dev/null", O_RDWR);
603 if (status != 0) {
604 ERROR("Error: Could not connect `STDIN' to `/dev/null' (status %d)",
605 status);
606 return 1;
607 }
609 status = dup(0);
610 if (status != 1) {
611 ERROR("Error: Could not connect `STDOUT' to `/dev/null' (status %d)",
612 status);
613 return 1;
614 }
616 status = dup(0);
617 if (status != 2) {
618 ERROR("Error: Could not connect `STDERR' to `/dev/null', (status %d)",
619 status);
620 return 1;
621 }
622 } /* if (daemonize) */
623 #endif /* COLLECT_DAEMON */
625 struct sigaction sig_pipe_action = {.sa_handler = SIG_IGN};
627 sigaction(SIGPIPE, &sig_pipe_action, NULL);
629 /*
630 * install signal handlers
631 */
632 struct sigaction sig_int_action = {.sa_handler = sig_int_handler};
634 if (0 != sigaction(SIGINT, &sig_int_action, NULL)) {
635 char errbuf[1024];
636 ERROR("Error: Failed to install a signal handler for signal INT: %s",
637 sstrerror(errno, errbuf, sizeof(errbuf)));
638 return 1;
639 }
641 struct sigaction sig_term_action = {.sa_handler = sig_term_handler};
643 if (0 != sigaction(SIGTERM, &sig_term_action, NULL)) {
644 char errbuf[1024];
645 ERROR("Error: Failed to install a signal handler for signal TERM: %s",
646 sstrerror(errno, errbuf, sizeof(errbuf)));
647 return 1;
648 }
650 struct sigaction sig_usr1_action = {.sa_handler = sig_usr1_handler};
652 if (0 != sigaction(SIGUSR1, &sig_usr1_action, NULL)) {
653 char errbuf[1024];
654 ERROR("Error: Failed to install a signal handler for signal USR1: %s",
655 sstrerror(errno, errbuf, sizeof(errbuf)));
656 return 1;
657 }
659 /*
660 * run the actual loops
661 */
662 if (do_init() != 0) {
663 ERROR("Error: one or more plugin init callbacks failed.");
664 exit_status = 1;
665 }
667 if (test_readall) {
668 if (plugin_read_all_once() != 0) {
669 ERROR("Error: one or more plugin read callbacks failed.");
670 exit_status = 1;
671 }
672 } else {
673 INFO("Initialization complete, entering read-loop.");
674 do_loop();
675 }
677 /* close syslog */
678 INFO("Exiting normally.");
680 if (do_shutdown() != 0) {
681 ERROR("Error: one or more plugin shutdown callbacks failed.");
682 exit_status = 1;
683 }
685 #if COLLECT_DAEMON
686 if (daemonize)
687 pidfile_remove();
688 #endif /* COLLECT_DAEMON */
690 return exit_status;
691 } /* int main */