xref: /netbsd/usr.sbin/rpc.statd/statd.c (revision 6464b8bb)
1 /*	$NetBSD: statd.c,v 1.34 2019/12/02 19:23:53 christos 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 #include <sys/cdefs.h>
37 #ifndef lint
38 __RCSID("$NetBSD: statd.c,v 1.34 2019/12/02 19:23:53 christos Exp $");
39 #endif
40 
41 /* main() function for status monitor daemon.  Some of the code in this	*/
42 /* file was generated by running rpcgen /usr/include/rpcsvc/sm_inter.x	*/
43 /* The actual program logic is in the file procs.c			*/
44 
45 #include <sys/param.h>
46 #include <sys/wait.h>
47 
48 #include <err.h>
49 #include <ctype.h>
50 #include <errno.h>
51 #include <fcntl.h>
52 #include <signal.h>
53 #include <stdio.h>
54 #include <stdlib.h>
55 #include <string.h>
56 #include <syslog.h>
57 #include <unistd.h>
58 #include <util.h>
59 #include <db.h>
60 #include <netconfig.h>
61 
62 #include <rpc/rpc.h>
63 
64 #include "statd.h"
65 
66 struct sigaction sa;
67 int     	debug = 0;		/* Controls syslog() for debug msgs */
68 extern int     	_rpcsvcdirty;
69 static DB	*db;			/* Database file */
70 
71 Header		 status_info;
72 
73 static char undefdata[] = "\0\1\2\3\4\5\6\7";
74 static DBT undefkey = {
75 	undefdata,
76 	sizeof(undefdata)
77 };
78 
79 
80 /* statd.c */
81 static int walk_one(int (*fun )(DBT *, HostInfo *, void *), DBT *, DBT *, void *);
82 static int walk_db(int (*fun )(DBT *, HostInfo *, void *), void *);
83 static int reset_host(DBT *, HostInfo *, void *);
84 static int check_work(DBT *, HostInfo *, void *);
85 static int unmon_host(DBT *, HostInfo *, void *);
86 static int notify_one(DBT *, HostInfo *, void *);
87 static void init_file(const char *);
88 static int notify_one_host(const char *);
89 static void die(int) __dead;
90 
91 int
main(int argc,char * argv[])92 main(int argc, char *argv[])
93 {
94 	int ch;
95 	struct sigaction nsa;
96 	int maxrec = RPC_MAXDATASIZE;
97 
98 	while ((ch = getopt(argc, argv, "d")) != (-1)) {
99 		switch (ch) {
100 		case 'd':
101 			debug = 1;
102 			break;
103 		default:
104 		case '?':
105 			(void)fprintf(stderr, "usage: %s [-d]\n",
106 			    getprogname());
107 			exit(1);
108 			/* NOTREACHED */
109 		}
110 	}
111 	(void)rpcb_unset(SM_PROG, SM_VERS, NULL);
112 
113 	rpc_control(RPC_SVC_CONNMAXREC_SET, &maxrec);
114 
115 	if (!svc_create(sm_prog_1, SM_PROG, SM_VERS, "udp")) {
116 		errx(EXIT_FAILURE, "cannot create udp service.");
117 		/* NOTREACHED */
118 	}
119 	if (!svc_create(sm_prog_1, SM_PROG, SM_VERS, "tcp")) {
120 		errx(EXIT_FAILURE, "cannot create udp service.");
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 	if (!debug)
131 		daemon(0, 0);
132 
133 	sigemptyset(&nsa.sa_mask);
134 	nsa.sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT;
135 	nsa.sa_handler = SIG_IGN;
136 	(void)sigaction(SIGCHLD, &nsa, NULL);
137 
138 	pidfile(NULL);
139 	openlog("rpc.statd", 0, LOG_DAEMON);
140 	if (debug)
141 		syslog(LOG_INFO, "Starting - debug enabled");
142 	else
143 		syslog(LOG_INFO, "Starting");
144 
145 	sa.sa_handler = die;
146 	sa.sa_flags = 0;
147 	sigemptyset(&sa.sa_mask);
148 	(void)sigaction(SIGTERM, &sa, NULL);
149 	(void)sigaction(SIGQUIT, &sa, NULL);
150 	(void)sigaction(SIGHUP, &sa, NULL);
151 	(void)sigaction(SIGINT, &sa, NULL);
152 
153 	sa.sa_handler = SIG_IGN;
154 	sa.sa_flags = SA_RESTART;
155 	sigemptyset(&sa.sa_mask);
156 	sigaddset(&sa.sa_mask, SIGALRM);
157 
158 	/* Initialisation now complete - start operating */
159 
160 	/* Notify hosts that need it */
161 	notify_handler(0);
162 
163 	while (1)
164 		svc_run();		/* Should never return */
165 	die(0);
166 }
167 
168 /* notify_handler ---------------------------------------------------------- */
169 /*
170  * Purpose:	Catch SIGALRM and collect process status
171  * Returns:	Nothing.
172  * Notes:	No special action required, other than to collect the
173  *		process status and hence allow the child to die:
174  *		we only use child processes for asynchronous transmission
175  *		of SM_NOTIFY to other systems, so it is normal for the
176  *		children to exit when they have done their work.
177  */
178 void
notify_handler(int sig)179 notify_handler(int sig)
180 {
181 	time_t now;
182 
183 	NO_ALARM;
184 	sa.sa_handler = SIG_IGN;
185 	(void)sigaction(SIGALRM, &sa, NULL);
186 
187 	now = time(NULL);
188 
189 	(void) walk_db(notify_one, &now);
190 
191 	if (walk_db(check_work, &now) == 0) {
192 		/*
193 		 * No more work to be done.
194 		 */
195 		CLR_ALARM;
196 		return;
197 	}
198 	sync_file();
199 	ALARM;
200 	alarm(5);
201 }
202 
203 /* sync_file --------------------------------------------------------------- */
204 /*
205  * Purpose:	Packaged call of msync() to flush changes to mmap()ed file
206  * Returns:	Nothing.  Errors to syslog.
207  */
208 void
sync_file()209 sync_file()
210 {
211 	DBT data;
212 
213 	data.data = &status_info;
214 	data.size = sizeof(status_info);
215 	switch ((*db->put)(db, &undefkey, &data, 0)) {
216 	case 0:
217 		return;
218 	case -1:
219 		goto bad;
220 	default:
221 		abort();
222 	}
223 	if ((*db->sync)(db, 0) == -1) {
224 bad:
225 		syslog(LOG_ERR, "database corrupted %m");
226 		die(1);
227 	}
228 }
229 
230 /* change_host -------------------------------------------------------------- */
231 /*
232  * Purpose:	Update/Create an entry for host
233  * Returns:	Nothing
234  * Notes:
235  *
236  */
237 void
change_host(char * hostnamep,HostInfo * hp)238 change_host(char *hostnamep, HostInfo *hp)
239 {
240 	DBT key, data;
241 	char *ptr;
242 	char hostname[MAXHOSTNAMELEN + 1];
243 	HostInfo h;
244 
245 	strncpy(hostname, hostnamep, MAXHOSTNAMELEN + 1);
246 	h = *hp;
247 
248 	for (ptr = hostname; *ptr; ptr++)
249 		if (isupper((unsigned char) *ptr))
250 			*ptr = tolower((unsigned char) *ptr);
251 
252 	key.data = hostname;
253 	key.size = ptr - hostname + 1;
254 	data.data = &h;
255 	data.size = sizeof(h);
256 
257 	switch ((*db->put)(db, &key, &data, 0)) {
258 	case -1:
259 		syslog(LOG_ERR, "database corrupted %m");
260 		die(1);
261 	case 0:
262 		return;
263 	default:
264 		abort();
265 	}
266 }
267 
268 
269 /* find_host -------------------------------------------------------------- */
270 /*
271  * Purpose:	Find the entry in the status file for a given host
272  * Returns:	Copy of entry in hd, or NULL
273  * Notes:
274  *
275  */
276 HostInfo *
find_host(char * hostname,HostInfo * hp)277 find_host(char *hostname, HostInfo *hp)
278 {
279 	DBT key, data;
280 	char *ptr;
281 
282 	for (ptr = hostname; *ptr; ptr++)
283 		if (isupper((unsigned char) *ptr))
284 			*ptr = tolower((unsigned char) *ptr);
285 
286 	key.data = hostname;
287 	key.size = ptr - hostname + 1;
288 	switch ((*db->get)(db, &key, &data, 0)) {
289 	case 0:
290 		if (data.size != sizeof(*hp))
291 			goto bad;
292 		return memcpy(hp, data.data, sizeof(*hp));
293 	case 1:
294 		return NULL;
295 	case -1:
296 		goto bad;
297 	default:
298 		abort();
299 	}
300 
301 bad:
302 	syslog(LOG_ERR, "Database corrupted %m");
303 	return NULL;
304 }
305 
306 /* walk_one ------------------------------------------------------------- */
307 /*
308  * Purpose:	Call the given function if the element is valid
309  * Returns:	Nothing - exits on error
310  * Notes:
311  */
312 static int
walk_one(int (* fun)(DBT *,HostInfo *,void *),DBT * key,DBT * data,void * ptr)313 walk_one(int (*fun)(DBT *, HostInfo *, void *), DBT *key, DBT *data, void *ptr)
314 {
315 	HostInfo h;
316 	if (key->size == undefkey.size &&
317 	    memcmp(key->data, undefkey.data, key->size) == 0)
318 		return 0;
319 	if (data->size != sizeof(HostInfo)) {
320 		syslog(LOG_ERR, "Bad data in database");
321 		die(1);
322 	}
323 	memcpy(&h, data->data, sizeof(h));
324 	return (*fun)(key, &h, ptr);
325 }
326 
327 /* walk_db -------------------------------------------------------------- */
328 /*
329  * Purpose:	Iterate over all elements calling the given function
330  * Returns:	-1 if function failed, 0 on success
331  * Notes:
332  */
333 static int
walk_db(int (* fun)(DBT *,HostInfo *,void *),void * ptr)334 walk_db(int (*fun)(DBT *, HostInfo *, void *), void *ptr)
335 {
336 	DBT key, data;
337 
338 	switch ((*db->seq)(db, &key, &data, R_FIRST)) {
339 	case -1:
340 		goto bad;
341 	case 1:
342 		/* We should have at least the magic entry at this point */
343 		abort();
344 	case 0:
345 		if (walk_one(fun, &key, &data, ptr) == -1)
346 			return -1;
347 		break;
348 	default:
349 		abort();
350 	}
351 
352 
353 	for (;;)
354 		switch ((*db->seq)(db, &key, &data, R_NEXT)) {
355 		case -1:
356 			goto bad;
357 		case 0:
358 			if (walk_one(fun, &key, &data, ptr) == -1)
359 				return -1;
360 			break;
361 		case 1:
362 			return 0;
363 		default:
364 			abort();
365 		}
366 bad:
367 	syslog(LOG_ERR, "Corrupted database %m");
368 	die(1);
369 }
370 
371 /* reset_host ------------------------------------------------------------ */
372 /*
373  * Purpose:	Clean up existing hosts in file.
374  * Returns:	Always success 0.
375  * Notes:	Clean-up of existing file - monitored hosts will have a
376  *		pointer to a list of clients, which refers to memory in
377  *		the previous incarnation of the program and so are
378  *		meaningless now.  These pointers are zeroed and the fact
379  *		that the host was previously monitored is recorded by
380  *		setting the notifyReqd flag, which will in due course
381  *		cause a SM_NOTIFY to be sent.
382  *
383  *		Note that if we crash twice in quick succession, some hosts
384  *		may already have notifyReqd set, where we didn't manage to
385  *		notify them before the second crash occurred.
386  */
387 static int
reset_host(DBT * key,HostInfo * hi,void * ptr)388 reset_host(DBT *key, HostInfo *hi, void *ptr)
389 {
390 
391 	if (hi->monList) {
392 		hi->notifyReqd = *(time_t *) ptr;
393 		hi->attempts = 0;
394 		hi->monList = NULL;
395 		change_host((char *)key->data, hi);
396 	}
397 	return 0;
398 }
399 
400 /* check_work ------------------------------------------------------------ */
401 /*
402  * Purpose:	Check if there is work to be done.
403  * Returns:	0 if there is no work to be done -1 if there is.
404  * Notes:
405  */
406 static int
check_work(DBT * key,HostInfo * hi,void * ptr)407 check_work(DBT *key, HostInfo *hi, void *ptr)
408 {
409 	return hi->notifyReqd ? -1 : 0;
410 }
411 
412 /* unmon_host ------------------------------------------------------------ */
413 /*
414  * Purpose:	Unmonitor a host
415  * Returns:	0
416  * Notes:
417  */
418 static int
unmon_host(DBT * key,HostInfo * hi,void * ptr)419 unmon_host(DBT *key, HostInfo *hi, void *ptr)
420 {
421 	char *name = key->data;
422 
423 	if (do_unmon(name, hi, ptr))
424 		change_host(name, hi);
425 	return 0;
426 }
427 
428 /* notify_one ------------------------------------------------------------ */
429 /*
430  * Purpose:	Notify one host.
431  * Returns:	0 if success -1 on failure
432  * Notes:
433  */
434 static int
notify_one(DBT * key,HostInfo * hi,void * ptr)435 notify_one(DBT *key, HostInfo *hi, void *ptr)
436 {
437 	time_t now = *(time_t *) ptr;
438 	char *name = key->data;
439 	int error;
440 
441 	if (hi->notifyReqd == 0 || hi->notifyReqd > now)
442 		return 0;
443 
444 	/*
445 	 * If one of the initial attempts fails, we wait
446 	 * for a while and have another go.  This is necessary
447 	 * because when we have crashed, (eg. a power outage)
448 	 * it is quite possible that we won't be able to
449 	 * contact all monitored hosts immediately on restart,
450 	 * either because they crashed too and take longer
451 	 * to come up (in which case the notification isn't
452 	 * really required), or more importantly if some
453 	 * router etc. needed to reach the monitored host
454 	 * has not come back up yet.  In this case, we will
455 	 * be a bit late in re-establishing locks (after the
456 	 * grace period) but that is the best we can do.  We
457 	 * try 10 times at 5 sec intervals, 10 more times at
458 	 * 1 minute intervals, then 24 more times at hourly
459 	 * intervals, finally giving up altogether if the
460 	 * host hasn't come back to life after 24 hours.
461 	 */
462 	if (notify_one_host(name) || hi->attempts++ >= 44) {
463 		error = 0;
464 		hi->notifyReqd = 0;
465 		hi->attempts = 0;
466 	} else {
467 		error = -1;
468 		if (hi->attempts < 10)
469 			hi->notifyReqd += 5;
470 		else if (hi->attempts < 20)
471 			hi->notifyReqd += 60;
472 		else
473 			hi->notifyReqd += 60 * 60;
474 	}
475 	change_host(name, hi);
476 	return error;
477 }
478 
479 /* init_file -------------------------------------------------------------- */
480 /*
481  * Purpose:	Open file, create if necessary, initialise it.
482  * Returns:	Nothing - exits on error
483  * Notes:	Called before process becomes daemon, hence logs to
484  *		stderr rather than syslog.
485  *		Opens the file, then mmap()s it for ease of access.
486  *		Also performs initial clean-up of the file, zeroing
487  *		monitor list pointers, setting the notifyReqd flag in
488  *		all hosts that had a monitor list, and incrementing
489  *		the state number to the next even value.
490  */
491 static void
init_file(const char * filename)492 init_file(const char *filename)
493 {
494 	DBT data;
495 
496 	db = dbopen(filename, O_RDWR|O_CREAT|O_NDELAY|O_EXLOCK, 0644, DB_HASH,
497 	    NULL);
498 	if (db == NULL)
499 		err(EXIT_FAILURE, "Cannot open `%s'", filename);
500 
501 	switch ((*db->get)(db, &undefkey, &data, 0)) {
502 	case 1:
503 		/* New database */
504 		(void)memset(&status_info, 0, sizeof(status_info));
505 		sync_file();
506 		return;
507 
508 	case -1:
509 		err(EXIT_FAILURE, "error accessing database (%s)", strerror(errno));
510 	case 0:
511 		/* Existing database */
512 		if (data.size != sizeof(status_info))
513 			errx(EXIT_FAILURE, "database corrupted %lu != %lu",
514 			    (u_long)data.size, (u_long)sizeof(status_info));
515 		memcpy(&status_info, data.data, data.size);
516 		break;
517 	default:
518 		abort();
519 	}
520 
521 	reset_database();
522 	return;
523 }
524 
525 /* reset_database --------------------------------------------------------- */
526 /*
527  * Purpose:	Clears the statd database
528  * Returns:	Nothing
529  * Notes:	If this is not called on reset, it will leak memory.
530  */
531 void
reset_database()532 reset_database()
533 {
534 	time_t now = time(NULL);
535 	walk_db(reset_host, &now);
536 
537 	/* Select the next higher even number for the state counter */
538 	status_info.ourState =
539 	    (status_info.ourState + 2) & 0xfffffffe;
540 	status_info.ourState++;	/* XXX - ??? */
541 	sync_file();
542 }
543 
544 /* unmon_hosts --------------------------------------------------------- */
545 /*
546  * Purpose:	Unmonitor all the hosts
547  * Returns:	Nothing
548  * Notes:
549  */
550 void
unmon_hosts()551 unmon_hosts()
552 {
553 	time_t now = time(NULL);
554 	walk_db(unmon_host, &now);
555 	sync_file();
556 }
557 
558 static int
notify_one_host(const char * hostname)559 notify_one_host(const char *hostname)
560 {
561 	struct timeval timeout = {20, 0};	/* 20 secs timeout */
562 	CLIENT *cli;
563 	char dummy;
564 	stat_chge arg;
565 	char our_hostname[MAXHOSTNAMELEN + 1];
566 
567 	gethostname(our_hostname, sizeof(our_hostname));
568 	our_hostname[sizeof(our_hostname) - 1] = '\0';
569 	arg.mon_name = our_hostname;
570 	arg.state = status_info.ourState;
571 
572 	if (debug)
573 		syslog(LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s",
574 		    hostname, our_hostname);
575 
576 	cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
577 	if (!cli) {
578 		syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
579 		    clnt_spcreateerror(""));
580 		return (FALSE);
581 	}
582 	if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void,
583 	    &dummy, timeout) != RPC_SUCCESS) {
584 		syslog(LOG_ERR, "Failed to contact rpc.statd at host %s",
585 		    hostname);
586 		clnt_destroy(cli);
587 		return (FALSE);
588 	}
589 	clnt_destroy(cli);
590 	return (TRUE);
591 }
592 
593 
594 static void
die(int n)595 die(int n)
596 {
597 	(*db->close)(db);
598 	exit(n);
599 }
600