1 /* $OpenBSD: statd.c,v 1.6 2023/01/04 14:42:46 jsg Exp $ */
2
3 /*
4 * Copyright (c) 1995
5 * A.R. Gordon (andrew.gordon@net-tel.co.uk). All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed for the FreeBSD project
18 * 4. Neither the name of the author nor the names of any co-contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY ANDREW GORDON AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 */
35
36 /* main() function for status monitor daemon. Some of the code in this */
37 /* file was generated by running rpcgen /usr/include/rpcsvc/sm_inter.x */
38 /* The actual program logic is in the file procs.c */
39
40 #include <sys/wait.h>
41
42 #include <err.h>
43 #include <ctype.h>
44 #include <errno.h>
45 #include <fcntl.h>
46 #include <signal.h>
47 #include <stdio.h>
48 #include <stdlib.h>
49 #include <string.h>
50 #include <syslog.h>
51 #include <unistd.h>
52 #include <db.h>
53
54 #include <rpc/rpc.h>
55
56 #include "statd.h"
57
58 struct sigaction sa;
59 int debug = 0; /* Controls syslog() for debug msgs */
60 int _rpcsvcdirty = 0; /* XXX ??? */
61 static DB *db; /* Database file */
62
63 Header status_info;
64
65 static char undefdata[] = "\0\1\2\3\4\5\6\7";
66 static DBT undefkey = {
67 undefdata,
68 sizeof(undefdata)
69 };
70 extern char *__progname;
71
72 /* statd.c */
73 static int walk_one(int (*fun )(DBT *, HostInfo *, void *), DBT *, DBT *, void *);
74 static int walk_db(int (*fun )(DBT *, HostInfo *, void *), void *);
75 static int reset_host(DBT *, HostInfo *, void *);
76 static int check_work(DBT *, HostInfo *, void *);
77 static int unmon_host(DBT *, HostInfo *, void *);
78 static int notify_one(DBT *, HostInfo *, void *);
79 static void init_file(char *);
80 static int notify_one_host(char *);
81 static __dead void die(int);
82
83 int main(int, char **);
84
85 int
main(int argc,char ** argv)86 main(int argc, char **argv)
87 {
88 SVCXPRT *transp;
89 int ch;
90 struct sigaction nsa;
91
92 while ((ch = getopt(argc, argv, "d")) != (-1)) {
93 switch (ch) {
94 case 'd':
95 debug = 1;
96 break;
97 default:
98 fprintf(stderr, "usage: %s [-d]\n", __progname);
99 exit(1);
100 /* NOTREACHED */
101 }
102 }
103 pmap_unset(SM_PROG, SM_VERS);
104
105 transp = svcudp_create(RPC_ANYSOCK);
106 if (transp == NULL) {
107 errx(1, "cannot create udp service.");
108 /* NOTREACHED */
109 }
110 if (!svc_register(transp, SM_PROG, SM_VERS, sm_prog_1, IPPROTO_UDP)) {
111 errx(1, "unable to register (SM_PROG, SM_VERS, udp).");
112 /* NOTREACHED */
113 }
114 transp = svctcp_create(RPC_ANYSOCK, 0, 0);
115 if (transp == NULL) {
116 errx(1, "cannot create tcp service.");
117 /* NOTREACHED */
118 }
119 if (!svc_register(transp, SM_PROG, SM_VERS, sm_prog_1, IPPROTO_TCP)) {
120 errx(1, "unable to register (SM_PROG, SM_VERS, tcp).");
121 /* NOTREACHED */
122 }
123
124 init_file("/var/db/statd.status");
125
126 /*
127 * Note that it is NOT sensible to run this program from inetd - the
128 * protocol assumes that it will run immediately at boot time.
129 */
130 daemon(0, 0);
131
132 sigemptyset(&nsa.sa_mask);
133 nsa.sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT;
134 nsa.sa_handler = SIG_IGN;
135 sigaction(SIGCHLD, &nsa, NULL);
136
137 openlog("rpc.statd", 0, LOG_DAEMON);
138 if (debug)
139 syslog(LOG_INFO, "Starting - debug enabled");
140 else
141 syslog(LOG_INFO, "Starting");
142
143 sa.sa_handler = die;
144 sa.sa_flags = 0;
145 sigemptyset(&sa.sa_mask);
146 sigaction(SIGTERM, &sa, NULL);
147 sigaction(SIGQUIT, &sa, NULL);
148 sigaction(SIGHUP, &sa, NULL);
149 sigaction(SIGINT, &sa, NULL);
150
151 sa.sa_handler = SIG_IGN;
152 sa.sa_flags = SA_RESTART;
153 sigemptyset(&sa.sa_mask);
154 sigaddset(&sa.sa_mask, SIGALRM);
155
156 /* Initialisation now complete - start operating */
157
158 /* Notify hosts that need it */
159 notify_handler(0);
160
161 while (1)
162 svc_run(); /* Should never return */
163 die(0);
164 }
165
166 /* notify_handler ---------------------------------------------------------- */
167 /*
168 * Purpose: Catch SIGALRM and collect process status
169 * Returns: Nothing.
170 * Notes: No special action required, other than to collect the
171 * process status and hence allow the child to die:
172 * we only use child processes for asynchronous transmission
173 * of SM_NOTIFY to other systems, so it is normal for the
174 * children to exit when they have done their work.
175 */
176 void
notify_handler(int sig)177 notify_handler(int sig)
178 {
179 time_t now;
180
181 NO_ALARM;
182 sa.sa_handler = SIG_IGN;
183 sigaction(SIGALRM, &sa, NULL);
184
185 now = time(NULL);
186
187 walk_db(notify_one, &now);
188
189 if (walk_db(check_work, &now) == 0) {
190 /*
191 * No more work to be done.
192 */
193 CLR_ALARM;
194 return;
195 }
196 sync_file();
197 ALARM;
198 alarm(5);
199 }
200
201 /* sync_file --------------------------------------------------------------- */
202 /*
203 * Purpose: Packaged call of msync() to flush changes to mmap()ed file
204 * Returns: Nothing. Errors to syslog.
205 */
206 void
sync_file()207 sync_file()
208 {
209 DBT data;
210
211 data.data = &status_info;
212 data.size = sizeof(status_info);
213 switch ((*db->put)(db, &undefkey, &data, 0)) {
214 case 0:
215 return;
216 case -1:
217 goto bad;
218 default:
219 abort();
220 }
221 if ((*db->sync)(db, 0) == -1) {
222 bad:
223 syslog(LOG_ERR, "database corrupted %m");
224 die(1);
225 }
226 }
227
228 /* change_host -------------------------------------------------------------- */
229 /*
230 * Purpose: Update/Create an entry for host
231 * Returns: Nothing
232 * Notes:
233 *
234 */
235 void
change_host(char * hostnamep,HostInfo * hp)236 change_host(char *hostnamep, HostInfo *hp)
237 {
238 DBT key, data;
239 char *ptr;
240 char hostname[HOST_NAME_MAX+1 + 1];
241 HostInfo h;
242
243 strlcpy(hostname, hostnamep, sizeof(hostname));
244 h = *hp;
245
246 for (ptr = hostname; *ptr; ptr++)
247 if (isupper((unsigned char) *ptr))
248 *ptr = tolower((unsigned char) *ptr);
249
250 key.data = hostname;
251 key.size = ptr - hostname + 1;
252 data.data = &h;
253 data.size = sizeof(h);
254
255 switch ((*db->put)(db, &key, &data, 0)) {
256 case -1:
257 syslog(LOG_ERR, "database corrupted %m");
258 die(1);
259 case 0:
260 return;
261 default:
262 abort();
263 }
264 }
265
266
267 /* find_host -------------------------------------------------------------- */
268 /*
269 * Purpose: Find the entry in the status file for a given host
270 * Returns: Copy of entry in hd, or NULL
271 * Notes:
272 *
273 */
274 HostInfo *
find_host(char * hostname,HostInfo * hp)275 find_host(char *hostname, HostInfo *hp)
276 {
277 DBT key, data;
278 char *ptr;
279
280 for (ptr = hostname; *ptr; ptr++)
281 if (isupper((unsigned char) *ptr))
282 *ptr = tolower((unsigned char) *ptr);
283
284 key.data = hostname;
285 key.size = ptr - hostname + 1;
286 switch ((*db->get)(db, &key, &data, 0)) {
287 case 0:
288 if (data.size != sizeof(*hp))
289 goto bad;
290 return memcpy(hp, data.data, sizeof(*hp));
291 case 1:
292 return NULL;
293 case -1:
294 goto bad;
295 default:
296 abort();
297 }
298
299 bad:
300 syslog(LOG_ERR, "Database corrupted %m");
301 return NULL;
302 }
303
304 /* walk_one ------------------------------------------------------------- */
305 /*
306 * Purpose: Call the given function if the element is valid
307 * Returns: Nothing - exits on error
308 * Notes:
309 */
310 static int
walk_one(int (* fun)(DBT *,HostInfo *,void *),DBT * key,DBT * data,void * ptr)311 walk_one(int (*fun)(DBT *, HostInfo *, void *), DBT *key, DBT *data, void *ptr)
312 {
313 HostInfo h;
314 if (key->size == undefkey.size &&
315 memcmp(key->data, undefkey.data, key->size) == 0)
316 return 0;
317 if (data->size != sizeof(HostInfo)) {
318 syslog(LOG_ERR, "Bad data in database");
319 die(1);
320 }
321 memcpy(&h, data->data, sizeof(h));
322 return (*fun)(key, &h, ptr);
323 }
324
325 /* walk_db -------------------------------------------------------------- */
326 /*
327 * Purpose: Iterate over all elements calling the given function
328 * Returns: -1 if function failed, 0 on success
329 * Notes:
330 */
331 static int
walk_db(int (* fun)(DBT *,HostInfo *,void *),void * ptr)332 walk_db(int (*fun)(DBT *, HostInfo *, void *), void *ptr)
333 {
334 DBT key, data;
335
336 switch ((*db->seq)(db, &key, &data, R_FIRST)) {
337 case -1:
338 goto bad;
339 case 1:
340 /* We should have at least the magic entry at this point */
341 abort();
342 case 0:
343 if (walk_one(fun, &key, &data, ptr) == -1)
344 return -1;
345 break;
346 default:
347 abort();
348 }
349
350 for (;;)
351 switch ((*db->seq)(db, &key, &data, R_NEXT)) {
352 case -1:
353 goto bad;
354 case 0:
355 if (walk_one(fun, &key, &data, ptr) == -1)
356 return -1;
357 break;
358 case 1:
359 return 0;
360 default:
361 abort();
362 }
363 bad:
364 syslog(LOG_ERR, "Corrupted database %m");
365 die(1);
366 }
367
368 /* reset_host ------------------------------------------------------------ */
369 /*
370 * Purpose: Clean up existing hosts in file.
371 * Returns: Always success 0.
372 * Notes: Clean-up of existing file - monitored hosts will have a
373 * pointer to a list of clients, which refers to memory in
374 * the previous incarnation of the program and so are
375 * meaningless now. These pointers are zeroed and the fact
376 * that the host was previously monitored is recorded by
377 * setting the notifyReqd flag, which will in due course
378 * cause a SM_NOTIFY to be sent.
379 *
380 * Note that if we crash twice in quick succession, some hosts
381 * may already have notifyReqd set, where we didn't manage to
382 * notify them before the second crash occurred.
383 */
384 static int
reset_host(DBT * key,HostInfo * hi,void * ptr)385 reset_host(DBT *key, HostInfo *hi, void *ptr)
386 {
387 if (hi->monList) {
388 hi->notifyReqd = *(time_t *) ptr;
389 hi->attempts = 0;
390 hi->monList = NULL;
391 change_host((char *)key->data, hi);
392 }
393 return 0;
394 }
395
396 /* check_work ------------------------------------------------------------ */
397 /*
398 * Purpose: Check if there is work to be done.
399 * Returns: 0 if there is no work to be done -1 if there is.
400 * Notes:
401 */
402 static int
check_work(DBT * key,HostInfo * hi,void * ptr)403 check_work(DBT *key, HostInfo *hi, void *ptr)
404 {
405 return hi->notifyReqd ? -1 : 0;
406 }
407
408 /* unmon_host ------------------------------------------------------------ */
409 /*
410 * Purpose: Unmonitor a host
411 * Returns: 0
412 * Notes:
413 */
414 static int
unmon_host(DBT * key,HostInfo * hi,void * ptr)415 unmon_host(DBT *key, HostInfo *hi, void *ptr)
416 {
417 char *name = key->data;
418
419 if (do_unmon(name, hi, ptr))
420 change_host(name, hi);
421 return 0;
422 }
423
424 /* notify_one ------------------------------------------------------------ */
425 /*
426 * Purpose: Notify one host.
427 * Returns: 0 if success -1 on failure
428 * Notes:
429 */
430 static int
notify_one(DBT * key,HostInfo * hi,void * ptr)431 notify_one(DBT *key, HostInfo *hi, void *ptr)
432 {
433 time_t now = *(time_t *) ptr;
434 char *name = key->data;
435 int error;
436
437 if (hi->notifyReqd == 0 || hi->notifyReqd > now)
438 return 0;
439
440 /*
441 * If one of the initial attempts fails, we wait
442 * for a while and have another go. This is necessary
443 * because when we have crashed, (eg. a power outage)
444 * it is quite possible that we won't be able to
445 * contact all monitored hosts immediately on restart,
446 * either because they crashed too and take longer
447 * to come up (in which case the notification isn't
448 * really required), or more importantly if some
449 * router etc. needed to reach the monitored host
450 * has not come back up yet. In this case, we will
451 * be a bit late in re-establishing locks (after the
452 * grace period) but that is the best we can do. We
453 * try 10 times at 5 sec intervals, 10 more times at
454 * 1 minute intervals, then 24 more times at hourly
455 * intervals, finally giving up altogether if the
456 * host hasn't come back to life after 24 hours.
457 */
458 if (notify_one_host(name) || hi->attempts++ >= 44) {
459 error = 0;
460 hi->notifyReqd = 0;
461 hi->attempts = 0;
462 } else {
463 error = -1;
464 if (hi->attempts < 10)
465 hi->notifyReqd += 5;
466 else if (hi->attempts < 20)
467 hi->notifyReqd += 60;
468 else
469 hi->notifyReqd += 60 * 60;
470 }
471 change_host(name, hi);
472 return error;
473 }
474
475 /* init_file -------------------------------------------------------------- */
476 /*
477 * Purpose: Open file, create if necessary, initialise it.
478 * Returns: Nothing - exits on error
479 * Notes: Called before process becomes daemon, hence logs to
480 * stderr rather than syslog.
481 * Opens the file, then mmap()s it for ease of access.
482 * Also performs initial clean-up of the file, zeroing
483 * monitor list pointers, setting the notifyReqd flag in
484 * all hosts that had a monitor list, and incrementing
485 * the state number to the next even value.
486 */
487 static void
init_file(char * filename)488 init_file(char *filename)
489 {
490 DBT data;
491
492 db = dbopen(filename, O_RDWR|O_CREAT|O_NDELAY|O_EXLOCK, 0644, DB_HASH,
493 NULL);
494 if (db == NULL)
495 err(1, "Cannot open `%s'", filename);
496
497 switch ((*db->get)(db, &undefkey, &data, 0)) {
498 case 1:
499 /* New database */
500 memset(&status_info, 0, sizeof(status_info));
501 sync_file();
502 return;
503 case -1:
504 err(1, "error accessing database (%m)");
505 case 0:
506 /* Existing database */
507 if (data.size != sizeof(status_info))
508 errx(1, "database corrupted %lu != %lu",
509 (u_long)data.size, (u_long)sizeof(status_info));
510 memcpy(&status_info, data.data, data.size);
511 break;
512 default:
513 abort();
514 }
515
516 reset_database();
517 return;
518 }
519
520 /* reset_database --------------------------------------------------------- */
521 /*
522 * Purpose: Clears the statd database
523 * Returns: Nothing
524 * Notes: If this is not called on reset, it will leak memory.
525 */
526 void
reset_database(void)527 reset_database(void)
528 {
529 time_t now = time(NULL);
530 walk_db(reset_host, &now);
531
532 /* Select the next higher even number for the state counter */
533 status_info.ourState =
534 (status_info.ourState + 2) & 0xfffffffe;
535 status_info.ourState++; /* XXX - ??? */
536 sync_file();
537 }
538
539 /* unmon_hosts --------------------------------------------------------- */
540 /*
541 * Purpose: Unmonitor all the hosts
542 * Returns: Nothing
543 * Notes:
544 */
545 void
unmon_hosts(void)546 unmon_hosts(void)
547 {
548 time_t now = time(NULL);
549 walk_db(unmon_host, &now);
550 sync_file();
551 }
552
553 static int
notify_one_host(char * hostname)554 notify_one_host(char *hostname)
555 {
556 struct timeval timeout = {20, 0}; /* 20 secs timeout */
557 CLIENT *cli;
558 char dummy;
559 stat_chge arg;
560 char our_hostname[HOST_NAME_MAX+1 + 1];
561
562 gethostname(our_hostname, sizeof(our_hostname));
563 our_hostname[sizeof(our_hostname) - 1] = '\0';
564 arg.mon_name = our_hostname;
565 arg.state = status_info.ourState;
566
567 if (debug)
568 syslog(LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s",
569 hostname, our_hostname);
570
571 cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
572 if (!cli) {
573 syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
574 clnt_spcreateerror(""));
575 return (FALSE);
576 }
577 if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void,
578 &dummy, timeout) != RPC_SUCCESS) {
579 syslog(LOG_ERR, "Failed to contact rpc.statd at host %s",
580 hostname);
581 clnt_destroy(cli);
582 return (FALSE);
583 }
584 clnt_destroy(cli);
585 return (TRUE);
586 }
587
588 static __dead void
die(int n)589 die(int n)
590 {
591 (*db->close)(db);
592 exit(n);
593 }
594