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 str = global_option_get("Hostname");
85 if ((str != NULL) && (str[0] != 0)) {
86 sstrncpy(hostname_g, str, hostname_g_size);
87 return 0;
88 }
90 if (gethostname(hostname_g, hostname_g_size) != 0) {
91 fprintf(stderr, "`gethostname' failed and no "
92 "hostname was configured.\n");
93 return -1;
94 }
96 str = global_option_get("FQDNLookup");
97 if (IS_FALSE(str))
98 return 0;
100 struct addrinfo ai_hints = {.ai_flags = AI_CANONNAME};
102 status = getaddrinfo(hostname_g, NULL, &ai_hints, &ai_list);
103 if (status != 0) {
104 ERROR("Looking up \"%s\" failed. You have set the "
105 "\"FQDNLookup\" option, but I cannot resolve "
106 "my hostname to a fully qualified domain "
107 "name. Please fix the network "
108 "configuration.",
109 hostname_g);
110 return -1;
111 }
113 for (struct addrinfo *ai_ptr = ai_list; ai_ptr != NULL;
114 ai_ptr = ai_ptr->ai_next) {
115 if (ai_ptr->ai_canonname == NULL)
116 continue;
118 sstrncpy(hostname_g, ai_ptr->ai_canonname, hostname_g_size);
119 break;
120 }
122 freeaddrinfo(ai_list);
123 return 0;
124 } /* int init_hostname */
126 static int init_global_variables(void) {
127 char const *str;
129 interval_g = cf_get_default_interval();
130 assert(interval_g > 0);
131 DEBUG("interval_g = %.3f;", CDTIME_T_TO_DOUBLE(interval_g));
133 str = global_option_get("Timeout");
134 if (str == NULL)
135 str = "2";
136 timeout_g = atoi(str);
137 if (timeout_g <= 1) {
138 fprintf(stderr, "Cannot set the timeout to a correct value.\n"
139 "Please check your settings.\n");
140 return -1;
141 }
142 DEBUG("timeout_g = %i;", timeout_g);
144 if (init_hostname() != 0)
145 return -1;
146 DEBUG("hostname_g = %s;", hostname_g);
148 return 0;
149 } /* int init_global_variables */
151 static int change_basedir(const char *orig_dir, _Bool create) {
152 char *dir;
153 size_t dirlen;
154 int status;
156 dir = strdup(orig_dir);
157 if (dir == NULL) {
158 char errbuf[1024];
159 ERROR("strdup failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
160 return -1;
161 }
163 dirlen = strlen(dir);
164 while ((dirlen > 0) && (dir[dirlen - 1] == '/'))
165 dir[--dirlen] = '\0';
167 if (dirlen == 0) {
168 free(dir);
169 return -1;
170 }
172 status = chdir(dir);
173 if (status == 0) {
174 free(dir);
175 return 0;
176 } else if (!create || (errno != ENOENT)) {
177 char errbuf[1024];
178 ERROR("change_basedir: chdir (%s): %s", dir,
179 sstrerror(errno, errbuf, sizeof(errbuf)));
180 free(dir);
181 return -1;
182 }
184 status = mkdir(dir, S_IRWXU | S_IRWXG | S_IRWXO);
185 if (status != 0) {
186 char errbuf[1024];
187 ERROR("change_basedir: mkdir (%s): %s", dir,
188 sstrerror(errno, errbuf, sizeof(errbuf)));
189 free(dir);
190 return -1;
191 }
193 status = chdir(dir);
194 if (status != 0) {
195 char errbuf[1024];
196 ERROR("change_basedir: chdir (%s): %s", dir,
197 sstrerror(errno, errbuf, sizeof(errbuf)));
198 free(dir);
199 return -1;
200 }
202 free(dir);
203 return 0;
204 } /* static int change_basedir (char *dir) */
206 #if HAVE_LIBKSTAT
207 static void update_kstat(void) {
208 if (kc == NULL) {
209 if ((kc = kstat_open()) == NULL)
210 ERROR("Unable to open kstat control structure");
211 } else {
212 kid_t kid;
213 kid = kstat_chain_update(kc);
214 if (kid > 0) {
215 INFO("kstat chain has been updated");
216 plugin_init_all();
217 } else if (kid < 0)
218 ERROR("kstat chain update failed");
219 /* else: everything works as expected */
220 }
222 return;
223 } /* static void update_kstat (void) */
224 #endif /* HAVE_LIBKSTAT */
226 /* TODO
227 * Remove all settings but `-f' and `-C'
228 */
229 __attribute__((noreturn)) static void exit_usage(int status) {
230 printf("Usage: " PACKAGE_NAME " [OPTIONS]\n\n"
232 "Available options:\n"
233 " General:\n"
234 " -C <file> Configuration file.\n"
235 " Default: " CONFIGFILE "\n"
236 " -t Test config and exit.\n"
237 " -T Test plugin read and exit.\n"
238 " -P <file> PID-file.\n"
239 " Default: " PIDFILE "\n"
240 #if COLLECT_DAEMON
241 " -f Don't fork to the background.\n"
242 #endif
243 " -B Don't create the BaseDir\n"
244 " -h Display help (this message)\n"
245 "\nBuiltin defaults:\n"
246 " Config file " CONFIGFILE "\n"
247 " PID file " PIDFILE "\n"
248 " Plugin directory " PLUGINDIR "\n"
249 " Data directory " PKGLOCALSTATEDIR "\n"
250 "\n" PACKAGE_NAME " " PACKAGE_VERSION ", http://collectd.org/\n"
251 "by Florian octo Forster <octo@collectd.org>\n"
252 "for contributions see `AUTHORS'\n");
253 exit(status);
254 } /* static void exit_usage (int status) */
256 static int do_init(void) {
257 #if HAVE_SETLOCALE
258 if (setlocale(LC_NUMERIC, COLLECTD_LOCALE) == NULL)
259 WARNING("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
261 /* Update the environment, so that libraries that are calling
262 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
263 unsetenv("LC_ALL");
264 setenv("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
265 #endif
267 #if HAVE_LIBKSTAT
268 kc = NULL;
269 update_kstat();
270 #endif
272 #if HAVE_LIBSTATGRAB
273 if (sg_init(
274 #if HAVE_LIBSTATGRAB_0_90
275 0
276 #endif
277 )) {
278 ERROR("sg_init: %s", sg_str_error(sg_get_error()));
279 return -1;
280 }
282 if (sg_drop_privileges()) {
283 ERROR("sg_drop_privileges: %s", sg_str_error(sg_get_error()));
284 return -1;
285 }
286 #endif
288 return plugin_init_all();
289 } /* int do_init () */
291 static int do_loop(void) {
292 cdtime_t interval = cf_get_default_interval();
293 cdtime_t wait_until;
295 wait_until = cdtime() + interval;
297 while (loop == 0) {
298 cdtime_t now;
300 #if HAVE_LIBKSTAT
301 update_kstat();
302 #endif
304 /* Issue all plugins */
305 plugin_read_all();
307 now = cdtime();
308 if (now >= wait_until) {
309 WARNING("Not sleeping because the next interval is "
310 "%.3f seconds in the past!",
311 CDTIME_T_TO_DOUBLE(now - wait_until));
312 wait_until = now + interval;
313 continue;
314 }
316 struct timespec ts_wait = CDTIME_T_TO_TIMESPEC(wait_until - now);
317 wait_until = wait_until + interval;
319 while ((loop == 0) && (nanosleep(&ts_wait, &ts_wait) != 0)) {
320 if (errno != EINTR) {
321 char errbuf[1024];
322 ERROR("nanosleep failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
323 return -1;
324 }
325 }
326 } /* while (loop == 0) */
328 return 0;
329 } /* int do_loop */
331 static int do_shutdown(void) {
332 return plugin_shutdown_all();
333 } /* int do_shutdown */
335 #if COLLECT_DAEMON
336 static int pidfile_create(void) {
337 FILE *fh;
338 const char *file = global_option_get("PIDFile");
340 if ((fh = fopen(file, "w")) == NULL) {
341 char errbuf[1024];
342 ERROR("fopen (%s): %s", file, sstrerror(errno, errbuf, sizeof(errbuf)));
343 return 1;
344 }
346 fprintf(fh, "%i\n", (int)getpid());
347 fclose(fh);
349 return 0;
350 } /* static int pidfile_create (const char *file) */
352 static int pidfile_remove(void) {
353 const char *file = global_option_get("PIDFile");
354 if (file == NULL)
355 return 0;
357 return unlink(file);
358 } /* static int pidfile_remove (const char *file) */
359 #endif /* COLLECT_DAEMON */
361 #ifdef KERNEL_LINUX
362 static int notify_upstart(void) {
363 char const *upstart_job = getenv("UPSTART_JOB");
365 if (upstart_job == NULL)
366 return 0;
368 if (strcmp(upstart_job, "collectd") != 0) {
369 WARNING("Environment specifies unexpected UPSTART_JOB=\"%s\", expected "
370 "\"collectd\". Ignoring the variable.",
371 upstart_job);
372 return 0;
373 }
375 NOTICE("Upstart detected, stopping now to signal readyness.");
376 raise(SIGSTOP);
377 unsetenv("UPSTART_JOB");
379 return 1;
380 }
382 static int notify_systemd(void) {
383 int fd;
384 const char *notifysocket;
385 struct sockaddr_un su = {0};
386 size_t su_size;
387 char buffer[] = "READY=1\n";
389 notifysocket = getenv("NOTIFY_SOCKET");
390 if (notifysocket == NULL)
391 return 0;
393 if ((strlen(notifysocket) < 2) ||
394 ((notifysocket[0] != '@') && (notifysocket[0] != '/'))) {
395 ERROR("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be "
396 "absolute",
397 notifysocket);
398 return 0;
399 }
400 NOTICE("Systemd detected, trying to signal readyness.");
402 unsetenv("NOTIFY_SOCKET");
404 #if defined(SOCK_CLOEXEC)
405 fd = socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, /* protocol = */ 0);
406 #else
407 fd = socket(AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
408 #endif
409 if (fd < 0) {
410 char errbuf[1024];
411 ERROR("creating UNIX socket failed: %s",
412 sstrerror(errno, errbuf, sizeof(errbuf)));
413 return 0;
414 }
416 su.sun_family = AF_UNIX;
417 if (notifysocket[0] != '@') {
418 /* regular UNIX socket */
419 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
420 su_size = sizeof(su);
421 } else {
422 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
423 * start with a null byte. Since null bytes have no special meaning in
424 * that case, we have to set su_size correctly to cover only the bytes
425 * that are part of the address. */
426 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
427 su.sun_path[0] = 0;
428 su_size = sizeof(sa_family_t) + strlen(notifysocket);
429 if (su_size > sizeof(su))
430 su_size = sizeof(su);
431 }
433 if (sendto(fd, buffer, strlen(buffer), MSG_NOSIGNAL, (void *)&su,
434 (socklen_t)su_size) < 0) {
435 char errbuf[1024];
436 ERROR("sendto(\"%s\") failed: %s", notifysocket,
437 sstrerror(errno, errbuf, sizeof(errbuf)));
438 close(fd);
439 return 0;
440 }
442 unsetenv("NOTIFY_SOCKET");
443 close(fd);
444 return 1;
445 }
446 #endif /* KERNEL_LINUX */
448 struct cmdline_config {
449 _Bool test_config;
450 _Bool test_readall;
451 _Bool create_basedir;
452 const char *configfile;
453 _Bool daemonize;
454 };
456 void read_cmdline(int argc, char **argv, struct cmdline_config *config) {
457 /* read options */
458 while (1) {
459 int c;
460 c = getopt(argc, argv, "htTC:"
461 #if COLLECT_DAEMON
462 "fP:"
463 #endif
464 );
466 if (c == -1)
467 break;
469 switch (c) {
470 case 'B':
471 config->create_basedir = 0;
472 break;
473 case 'C':
474 config->configfile = optarg;
475 break;
476 case 't':
477 config->test_config = 1;
478 break;
479 case 'T':
480 config->test_readall = 1;
481 global_option_set("ReadThreads", "-1", 1);
482 #if COLLECT_DAEMON
483 config->daemonize = 0;
484 #endif /* COLLECT_DAEMON */
485 break;
486 #if COLLECT_DAEMON
487 case 'P':
488 global_option_set("PIDFile", optarg, 1);
489 break;
490 case 'f':
491 config->daemonize = 0;
492 break;
493 #endif /* COLLECT_DAEMON */
494 case 'h':
495 exit_usage(0);
496 break;
497 default:
498 exit_usage(1);
499 } /* switch (c) */
500 } /* while (1) */
501 }
503 int configure_collectd(struct cmdline_config *config) {
504 const char *basedir;
505 /*
506 * Read options from the config file, the environment and the command
507 * line (in that order, with later options overwriting previous ones in
508 * general).
509 * Also, this will automatically load modules.
510 */
511 if (cf_read(config->configfile)) {
512 fprintf(stderr, "Error: Reading the config file failed!\n"
513 "Read the logs for details.\n");
514 return 1;
515 }
517 /*
518 * Change directory. We do this _after_ reading the config and loading
519 * modules to relative paths work as expected.
520 */
521 if ((basedir = global_option_get("BaseDir")) == NULL) {
522 fprintf(stderr,
523 "Don't have a basedir to use. This should not happen. Ever.");
524 return 1;
525 } else if (change_basedir(basedir, config->create_basedir)) {
526 fprintf(stderr, "Error: Unable to change to directory `%s'.\n", basedir);
527 return 1;
528 }
530 /*
531 * Set global variables or, if that fails, exit. We cannot run with
532 * them being uninitialized. If nothing is configured, then defaults
533 * are being used. So this means that the user has actually done
534 * something wrong.
535 */
536 if (init_global_variables() != 0)
537 return 1;
539 return 0;
540 }
542 int main(int argc, char **argv) {
543 #if COLLECT_DAEMON
544 pid_t pid;
545 #endif
546 int exit_status = 0;
548 struct cmdline_config config = {
549 .daemonize = 1, .create_basedir = 1, .configfile = CONFIGFILE,
550 };
552 read_cmdline(argc, argv, &config);
554 if (config.test_config)
555 return 0;
557 if (optind < argc)
558 exit_usage(1);
560 plugin_init_ctx();
562 int status;
563 if ((status = configure_collectd(&config)) != 0)
564 exit(EXIT_FAILURE);
566 #if COLLECT_DAEMON
567 /*
568 * fork off child
569 */
570 struct sigaction sig_chld_action = {.sa_handler = SIG_IGN};
572 sigaction(SIGCHLD, &sig_chld_action, NULL);
574 /*
575 * Only daemonize if we're not being supervised
576 * by upstart or systemd (when using Linux).
577 */
578 if (config.daemonize
579 #ifdef KERNEL_LINUX
580 && notify_upstart() == 0 && notify_systemd() == 0
581 #endif
582 ) {
583 int status;
585 if ((pid = fork()) == -1) {
586 /* error */
587 char errbuf[1024];
588 fprintf(stderr, "fork: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
589 return 1;
590 } else if (pid != 0) {
591 /* parent */
592 /* printf ("Running (PID %i)\n", pid); */
593 return 0;
594 }
596 /* Detach from session */
597 setsid();
599 /* Write pidfile */
600 if (pidfile_create())
601 exit(2);
603 /* close standard descriptors */
604 close(2);
605 close(1);
606 close(0);
608 status = open("/dev/null", O_RDWR);
609 if (status != 0) {
610 ERROR("Error: Could not connect `STDIN' to `/dev/null' (status %d)",
611 status);
612 return 1;
613 }
615 status = dup(0);
616 if (status != 1) {
617 ERROR("Error: Could not connect `STDOUT' to `/dev/null' (status %d)",
618 status);
619 return 1;
620 }
622 status = dup(0);
623 if (status != 2) {
624 ERROR("Error: Could not connect `STDERR' to `/dev/null', (status %d)",
625 status);
626 return 1;
627 }
628 } /* if (config.daemonize) */
629 #endif /* COLLECT_DAEMON */
631 struct sigaction sig_pipe_action = {.sa_handler = SIG_IGN};
633 sigaction(SIGPIPE, &sig_pipe_action, NULL);
635 /*
636 * install signal handlers
637 */
638 struct sigaction sig_int_action = {.sa_handler = sig_int_handler};
640 if (0 != sigaction(SIGINT, &sig_int_action, NULL)) {
641 char errbuf[1024];
642 ERROR("Error: Failed to install a signal handler for signal INT: %s",
643 sstrerror(errno, errbuf, sizeof(errbuf)));
644 return 1;
645 }
647 struct sigaction sig_term_action = {.sa_handler = sig_term_handler};
649 if (0 != sigaction(SIGTERM, &sig_term_action, NULL)) {
650 char errbuf[1024];
651 ERROR("Error: Failed to install a signal handler for signal TERM: %s",
652 sstrerror(errno, errbuf, sizeof(errbuf)));
653 return 1;
654 }
656 struct sigaction sig_usr1_action = {.sa_handler = sig_usr1_handler};
658 if (0 != sigaction(SIGUSR1, &sig_usr1_action, NULL)) {
659 char errbuf[1024];
660 ERROR("Error: Failed to install a signal handler for signal USR1: %s",
661 sstrerror(errno, errbuf, sizeof(errbuf)));
662 return 1;
663 }
665 /*
666 * run the actual loops
667 */
668 if (do_init() != 0) {
669 ERROR("Error: one or more plugin init callbacks failed.");
670 exit_status = 1;
671 }
673 if (config.test_readall) {
674 if (plugin_read_all_once() != 0) {
675 ERROR("Error: one or more plugin read callbacks failed.");
676 exit_status = 1;
677 }
678 } else {
679 INFO("Initialization complete, entering read-loop.");
680 do_loop();
681 }
683 /* close syslog */
684 INFO("Exiting normally.");
686 if (do_shutdown() != 0) {
687 ERROR("Error: one or more plugin shutdown callbacks failed.");
688 exit_status = 1;
689 }
691 #if COLLECT_DAEMON
692 if (config.daemonize)
693 pidfile_remove();
694 #endif /* COLLECT_DAEMON */
696 return exit_status;
697 } /* int main */