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) {
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 status = mkdir(dir, S_IRWXU | S_IRWXG | S_IRWXO);
195 if (status != 0) {
196 char errbuf[1024];
197 ERROR("change_basedir: mkdir (%s): %s", dir,
198 sstrerror(errno, errbuf, sizeof(errbuf)));
199 free(dir);
200 return (-1);
201 }
203 status = chdir(dir);
204 if (status != 0) {
205 char errbuf[1024];
206 ERROR("change_basedir: chdir (%s): %s", dir,
207 sstrerror(errno, errbuf, sizeof(errbuf)));
208 free(dir);
209 return (-1);
210 }
212 free(dir);
213 return (0);
214 } /* static int change_basedir (char *dir) */
216 #if HAVE_LIBKSTAT
217 static void update_kstat(void) {
218 if (kc == NULL) {
219 if ((kc = kstat_open()) == NULL)
220 ERROR("Unable to open kstat control structure");
221 } else {
222 kid_t kid;
223 kid = kstat_chain_update(kc);
224 if (kid > 0) {
225 INFO("kstat chain has been updated");
226 plugin_init_all();
227 } else if (kid < 0)
228 ERROR("kstat chain update failed");
229 /* else: everything works as expected */
230 }
232 return;
233 } /* static void update_kstat (void) */
234 #endif /* HAVE_LIBKSTAT */
236 /* TODO
237 * Remove all settings but `-f' and `-C'
238 */
239 __attribute__((noreturn)) static void exit_usage(int status) {
240 printf("Usage: " PACKAGE_NAME " [OPTIONS]\n\n"
242 "Available options:\n"
243 " General:\n"
244 " -C <file> Configuration file.\n"
245 " Default: " CONFIGFILE "\n"
246 " -t Test config and exit.\n"
247 " -T Test plugin read and exit.\n"
248 " -P <file> PID-file.\n"
249 " Default: " PIDFILE "\n"
250 #if COLLECT_DAEMON
251 " -f Don't fork to the background.\n"
252 #endif
253 " -h Display help (this message)\n"
254 "\nBuiltin defaults:\n"
255 " Config file " CONFIGFILE "\n"
256 " PID file " PIDFILE "\n"
257 " Plugin directory " PLUGINDIR "\n"
258 " Data directory " PKGLOCALSTATEDIR "\n"
259 "\n" PACKAGE_NAME " " PACKAGE_VERSION ", http://collectd.org/\n"
260 "by Florian octo Forster <octo@collectd.org>\n"
261 "for contributions see `AUTHORS'\n");
262 exit(status);
263 } /* static void exit_usage (int status) */
265 static int do_init(void) {
266 #if HAVE_SETLOCALE
267 if (setlocale(LC_NUMERIC, COLLECTD_LOCALE) == NULL)
268 WARNING("setlocale (\"%s\") failed.", COLLECTD_LOCALE);
270 /* Update the environment, so that libraries that are calling
271 * setlocale(LC_NUMERIC, "") don't accidentally revert these changes. */
272 unsetenv("LC_ALL");
273 setenv("LC_NUMERIC", COLLECTD_LOCALE, /* overwrite = */ 1);
274 #endif
276 #if HAVE_LIBKSTAT
277 kc = NULL;
278 update_kstat();
279 #endif
281 #if HAVE_LIBSTATGRAB
282 if (sg_init(
283 #if HAVE_LIBSTATGRAB_0_90
284 0
285 #endif
286 )) {
287 ERROR("sg_init: %s", sg_str_error(sg_get_error()));
288 return (-1);
289 }
291 if (sg_drop_privileges()) {
292 ERROR("sg_drop_privileges: %s", sg_str_error(sg_get_error()));
293 return (-1);
294 }
295 #endif
297 return plugin_init_all();
298 } /* int do_init () */
300 static int do_loop(void) {
301 cdtime_t interval = cf_get_default_interval();
302 cdtime_t wait_until;
304 wait_until = cdtime() + interval;
306 while (loop == 0) {
307 cdtime_t now;
309 #if HAVE_LIBKSTAT
310 update_kstat();
311 #endif
313 /* Issue all plugins */
314 plugin_read_all();
316 now = cdtime();
317 if (now >= wait_until) {
318 WARNING("Not sleeping because the next interval is "
319 "%.3f seconds in the past!",
320 CDTIME_T_TO_DOUBLE(now - wait_until));
321 wait_until = now + interval;
322 continue;
323 }
325 struct timespec ts_wait = CDTIME_T_TO_TIMESPEC(wait_until - now);
326 wait_until = wait_until + interval;
328 while ((loop == 0) && (nanosleep(&ts_wait, &ts_wait) != 0)) {
329 if (errno != EINTR) {
330 char errbuf[1024];
331 ERROR("nanosleep failed: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
332 return (-1);
333 }
334 }
335 } /* while (loop == 0) */
337 return (0);
338 } /* int do_loop */
340 static int do_shutdown(void) {
341 return plugin_shutdown_all();
342 } /* int do_shutdown */
344 #if COLLECT_DAEMON
345 static int pidfile_create(void) {
346 FILE *fh;
347 const char *file = global_option_get("PIDFile");
349 if ((fh = fopen(file, "w")) == NULL) {
350 char errbuf[1024];
351 ERROR("fopen (%s): %s", file, sstrerror(errno, errbuf, sizeof(errbuf)));
352 return (1);
353 }
355 fprintf(fh, "%i\n", (int)getpid());
356 fclose(fh);
358 return (0);
359 } /* static int pidfile_create (const char *file) */
361 static int pidfile_remove(void) {
362 const char *file = global_option_get("PIDFile");
363 if (file == NULL)
364 return 0;
366 return (unlink(file));
367 } /* static int pidfile_remove (const char *file) */
368 #endif /* COLLECT_DAEMON */
370 #ifdef KERNEL_LINUX
371 static int notify_upstart(void) {
372 char const *upstart_job = getenv("UPSTART_JOB");
374 if (upstart_job == NULL)
375 return 0;
377 if (strcmp(upstart_job, "collectd") != 0) {
378 WARNING("Environment specifies unexpected UPSTART_JOB=\"%s\", expected "
379 "\"collectd\". Ignoring the variable.",
380 upstart_job);
381 return 0;
382 }
384 NOTICE("Upstart detected, stopping now to signal readyness.");
385 raise(SIGSTOP);
386 unsetenv("UPSTART_JOB");
388 return 1;
389 }
391 static int notify_systemd(void) {
392 int fd;
393 const char *notifysocket;
394 struct sockaddr_un su = {0};
395 size_t su_size;
396 char buffer[] = "READY=1\n";
398 notifysocket = getenv("NOTIFY_SOCKET");
399 if (notifysocket == NULL)
400 return 0;
402 if ((strlen(notifysocket) < 2) ||
403 ((notifysocket[0] != '@') && (notifysocket[0] != '/'))) {
404 ERROR("invalid notification socket NOTIFY_SOCKET=\"%s\": path must be "
405 "absolute",
406 notifysocket);
407 return 0;
408 }
409 NOTICE("Systemd detected, trying to signal readyness.");
411 unsetenv("NOTIFY_SOCKET");
413 #if defined(SOCK_CLOEXEC)
414 fd = socket(AF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC, /* protocol = */ 0);
415 #else
416 fd = socket(AF_UNIX, SOCK_DGRAM, /* protocol = */ 0);
417 #endif
418 if (fd < 0) {
419 char errbuf[1024];
420 ERROR("creating UNIX socket failed: %s",
421 sstrerror(errno, errbuf, sizeof(errbuf)));
422 return 0;
423 }
425 su.sun_family = AF_UNIX;
426 if (notifysocket[0] != '@') {
427 /* regular UNIX socket */
428 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
429 su_size = sizeof(su);
430 } else {
431 /* Linux abstract namespace socket: specify address as "\0foo", i.e.
432 * start with a null byte. Since null bytes have no special meaning in
433 * that case, we have to set su_size correctly to cover only the bytes
434 * that are part of the address. */
435 sstrncpy(su.sun_path, notifysocket, sizeof(su.sun_path));
436 su.sun_path[0] = 0;
437 su_size = sizeof(sa_family_t) + strlen(notifysocket);
438 if (su_size > sizeof(su))
439 su_size = sizeof(su);
440 }
442 if (sendto(fd, buffer, strlen(buffer), MSG_NOSIGNAL, (void *)&su,
443 (socklen_t)su_size) < 0) {
444 char errbuf[1024];
445 ERROR("sendto(\"%s\") failed: %s", notifysocket,
446 sstrerror(errno, errbuf, sizeof(errbuf)));
447 close(fd);
448 return 0;
449 }
451 unsetenv("NOTIFY_SOCKET");
452 close(fd);
453 return 1;
454 }
455 #endif /* KERNEL_LINUX */
457 int main(int argc, char **argv) {
458 const char *configfile = CONFIGFILE;
459 int test_config = 0;
460 int test_readall = 0;
461 const char *basedir;
462 #if COLLECT_DAEMON
463 pid_t pid;
464 int daemonize = 1;
465 #endif
466 int exit_status = 0;
468 /* read options */
469 while (1) {
470 int c;
472 c = getopt(argc, argv, "htTC:"
473 #if COLLECT_DAEMON
474 "fP:"
475 #endif
476 );
478 if (c == -1)
479 break;
481 switch (c) {
482 case 'C':
483 configfile = optarg;
484 break;
485 case 't':
486 test_config = 1;
487 break;
488 case 'T':
489 test_readall = 1;
490 global_option_set("ReadThreads", "-1", 1);
491 #if COLLECT_DAEMON
492 daemonize = 0;
493 #endif /* COLLECT_DAEMON */
494 break;
495 #if COLLECT_DAEMON
496 case 'P':
497 global_option_set("PIDFile", optarg, 1);
498 break;
499 case 'f':
500 daemonize = 0;
501 break;
502 #endif /* COLLECT_DAEMON */
503 case 'h':
504 exit_usage(0);
505 break;
506 default:
507 exit_usage(1);
508 } /* switch (c) */
509 } /* while (1) */
511 if (optind < argc)
512 exit_usage(1);
514 plugin_init_ctx();
516 /*
517 * Read options from the config file, the environment and the command
518 * line (in that order, with later options overwriting previous ones in
519 * general).
520 * Also, this will automatically load modules.
521 */
522 if (cf_read(configfile)) {
523 fprintf(stderr, "Error: Reading the config file failed!\n"
524 "Read the syslog for details.\n");
525 return (1);
526 }
528 /*
529 * Change directory. We do this _after_ reading the config and loading
530 * modules to relative paths work as expected.
531 */
532 if ((basedir = global_option_get("BaseDir")) == NULL) {
533 fprintf(stderr,
534 "Don't have a basedir to use. This should not happen. Ever.");
535 return (1);
536 } else if (change_basedir(basedir)) {
537 fprintf(stderr, "Error: Unable to change to directory `%s'.\n", basedir);
538 return (1);
539 }
541 /*
542 * Set global variables or, if that failes, exit. We cannot run with
543 * them being uninitialized. If nothing is configured, then defaults
544 * are being used. So this means that the user has actually done
545 * something wrong.
546 */
547 if (init_global_variables() != 0)
548 exit(EXIT_FAILURE);
550 if (test_config)
551 return (0);
553 #if COLLECT_DAEMON
554 /*
555 * fork off child
556 */
557 struct sigaction sig_chld_action = {.sa_handler = SIG_IGN};
559 sigaction(SIGCHLD, &sig_chld_action, NULL);
561 /*
562 * Only daemonize if we're not being supervised
563 * by upstart or systemd (when using Linux).
564 */
565 if (daemonize
566 #ifdef KERNEL_LINUX
567 && notify_upstart() == 0 && notify_systemd() == 0
568 #endif
569 ) {
570 int status;
572 if ((pid = fork()) == -1) {
573 /* error */
574 char errbuf[1024];
575 fprintf(stderr, "fork: %s", sstrerror(errno, errbuf, sizeof(errbuf)));
576 return (1);
577 } else if (pid != 0) {
578 /* parent */
579 /* printf ("Running (PID %i)\n", pid); */
580 return (0);
581 }
583 /* Detach from session */
584 setsid();
586 /* Write pidfile */
587 if (pidfile_create())
588 exit(2);
590 /* close standard descriptors */
591 close(2);
592 close(1);
593 close(0);
595 status = open("/dev/null", O_RDWR);
596 if (status != 0) {
597 ERROR("Error: Could not connect `STDIN' to `/dev/null' (status %d)",
598 status);
599 return (1);
600 }
602 status = dup(0);
603 if (status != 1) {
604 ERROR("Error: Could not connect `STDOUT' to `/dev/null' (status %d)",
605 status);
606 return (1);
607 }
609 status = dup(0);
610 if (status != 2) {
611 ERROR("Error: Could not connect `STDERR' to `/dev/null', (status %d)",
612 status);
613 return (1);
614 }
615 } /* if (daemonize) */
616 #endif /* COLLECT_DAEMON */
618 struct sigaction sig_pipe_action = {.sa_handler = SIG_IGN};
620 sigaction(SIGPIPE, &sig_pipe_action, NULL);
622 /*
623 * install signal handlers
624 */
625 struct sigaction sig_int_action = {.sa_handler = sig_int_handler};
627 if (0 != sigaction(SIGINT, &sig_int_action, NULL)) {
628 char errbuf[1024];
629 ERROR("Error: Failed to install a signal handler for signal INT: %s",
630 sstrerror(errno, errbuf, sizeof(errbuf)));
631 return (1);
632 }
634 struct sigaction sig_term_action = {.sa_handler = sig_term_handler};
636 if (0 != sigaction(SIGTERM, &sig_term_action, NULL)) {
637 char errbuf[1024];
638 ERROR("Error: Failed to install a signal handler for signal TERM: %s",
639 sstrerror(errno, errbuf, sizeof(errbuf)));
640 return (1);
641 }
643 struct sigaction sig_usr1_action = {.sa_handler = sig_usr1_handler};
645 if (0 != sigaction(SIGUSR1, &sig_usr1_action, NULL)) {
646 char errbuf[1024];
647 ERROR("Error: Failed to install a signal handler for signal USR1: %s",
648 sstrerror(errno, errbuf, sizeof(errbuf)));
649 return (1);
650 }
652 /*
653 * run the actual loops
654 */
655 if (do_init() != 0) {
656 ERROR("Error: one or more plugin init callbacks failed.");
657 exit_status = 1;
658 }
660 if (test_readall) {
661 if (plugin_read_all_once() != 0) {
662 ERROR("Error: one or more plugin read callbacks failed.");
663 exit_status = 1;
664 }
665 } else {
666 INFO("Initialization complete, entering read-loop.");
667 do_loop();
668 }
670 /* close syslog */
671 INFO("Exiting normally.");
673 if (do_shutdown() != 0) {
674 ERROR("Error: one or more plugin shutdown callbacks failed.");
675 exit_status = 1;
676 }
678 #if COLLECT_DAEMON
679 if (daemonize)
680 pidfile_remove();
681 #endif /* COLLECT_DAEMON */
683 return (exit_status);
684 } /* int main */