Head

GCC Code Coverage Report

Directory:	./		Exec	Total	Coverage
File:	usr.sbin/rpc.statd/statd.c	Lines:	0	182	0.0 %
Date:	2017-11-07	Branches:	0	88	0.0 %


/*	$OpenBSD: statd.c,v 1.4 2017/01/21 08:33:51 krw Exp $	*/

/*
 * Copyright (c) 1997 Christos Zoulas. All rights reserved.
 * Copyright (c) 1995
 *	A.R. Gordon (andrew.gordon@net-tel.co.uk).  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed for the FreeBSD project
 *	This product includes software developed by Christos Zoulas.
 * 4. Neither the name of the author nor the names of any co-contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY ANDREW GORDON AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 */

/* main() function for status monitor daemon.  Some of the code in this	*/
/* file was generated by running rpcgen /usr/include/rpcsvc/sm_inter.x	*/
/* The actual program logic is in the file procs.c			*/

#include <sys/wait.h>

#include <err.h>
#include <ctype.h>
#include <errno.h>
#include <fcntl.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <syslog.h>
#include <unistd.h>
#include <db.h>

#include <rpc/rpc.h>

#include "statd.h"

struct sigaction sa;
int		debug = 0;		/* Controls syslog() for debug msgs */
int		_rpcsvcdirty = 0;	/* XXX ??? */
static DB	*db;			/* Database file */

Header		 status_info;

static char undefdata[] = "\0\1\2\3\4\5\6\7";
static DBT undefkey = {
	undefdata,
	sizeof(undefdata)
};
extern char *__progname;

/* statd.c */
static int walk_one(int (*fun )(DBT *, HostInfo *, void *), DBT *, DBT *, void *);
static int walk_db(int (*fun )(DBT *, HostInfo *, void *), void *);
static int reset_host(DBT *, HostInfo *, void *);
static int check_work(DBT *, HostInfo *, void *);
static int unmon_host(DBT *, HostInfo *, void *);
static int notify_one(DBT *, HostInfo *, void *);
static void init_file(char *);
static int notify_one_host(char *);
static __dead void die(int);

int main(int, char **);

int
main(int argc, char **argv)
{
	SVCXPRT *transp;
	int ch;
	struct sigaction nsa;

	while ((ch = getopt(argc, argv, "d")) != (-1)) {
		switch (ch) {
		case 'd':
			debug = 1;
			break;
		default:
		case '?':
			fprintf(stderr, "usage: %s [-d]\n", __progname);
			exit(1);
			/* NOTREACHED */
		}
	}
	pmap_unset(SM_PROG, SM_VERS);

	transp = svcudp_create(RPC_ANYSOCK);
	if (transp == NULL) {
		errx(1, "cannot create udp service.");
		/* NOTREACHED */
	}
	if (!svc_register(transp, SM_PROG, SM_VERS, sm_prog_1, IPPROTO_UDP)) {
		errx(1, "unable to register (SM_PROG, SM_VERS, udp).");
		/* NOTREACHED */
	}
	transp = svctcp_create(RPC_ANYSOCK, 0, 0);
	if (transp == NULL) {
		errx(1, "cannot create tcp service.");
		/* NOTREACHED */
	}
	if (!svc_register(transp, SM_PROG, SM_VERS, sm_prog_1, IPPROTO_TCP)) {
		errx(1, "unable to register (SM_PROG, SM_VERS, tcp).");
		/* NOTREACHED */
	}

	init_file("/var/db/statd.status");

	/*
	 * Note that it is NOT sensible to run this program from inetd - the
	 * protocol assumes that it will run immediately at boot time.
	 */
	daemon(0, 0);

	sigemptyset(&nsa.sa_mask);
	nsa.sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT;
	nsa.sa_handler = SIG_IGN;
	sigaction(SIGCHLD, &nsa, NULL);

	openlog("rpc.statd", 0, LOG_DAEMON);
	if (debug)
		syslog(LOG_INFO, "Starting - debug enabled");
	else
		syslog(LOG_INFO, "Starting");

	sa.sa_handler = die;
	sa.sa_flags = 0;
	sigemptyset(&sa.sa_mask);
	sigaction(SIGTERM, &sa, NULL);
	sigaction(SIGQUIT, &sa, NULL);
	sigaction(SIGHUP, &sa, NULL);
	sigaction(SIGINT, &sa, NULL);

	sa.sa_handler = SIG_IGN;
	sa.sa_flags = SA_RESTART;
	sigemptyset(&sa.sa_mask);
	sigaddset(&sa.sa_mask, SIGALRM);

	/* Initialisation now complete - start operating */

	/* Notify hosts that need it */
	notify_handler(0);

	while (1)
		svc_run();		/* Should never return */
	die(0);
}

/* notify_handler ---------------------------------------------------------- */
/*
 * Purpose:	Catch SIGALRM and collect process status
 * Returns:	Nothing.
 * Notes:	No special action required, other than to collect the
 *		process status and hence allow the child to die:
 *		we only use child processes for asynchronous transmission
 *		of SM_NOTIFY to other systems, so it is normal for the
 *		children to exit when they have done their work.
 */
void
notify_handler(int sig)
{
	time_t now;

	NO_ALARM;
	sa.sa_handler = SIG_IGN;
	sigaction(SIGALRM, &sa, NULL);

	now = time(NULL);

	walk_db(notify_one, &now);

	if (walk_db(check_work, &now) == 0) {
		/*
		 * No more work to be done.
		 */
		CLR_ALARM;
		return;
	}
	sync_file();
	ALARM;
	alarm(5);
}

/* sync_file --------------------------------------------------------------- */
/*
 * Purpose:	Packaged call of msync() to flush changes to mmap()ed file
 * Returns:	Nothing.  Errors to syslog.
 */
void
sync_file()
{
	DBT data;

	data.data = &status_info;
	data.size = sizeof(status_info);
	switch ((*db->put)(db, &undefkey, &data, 0)) {
	case 0:
		return;
	case -1:
		goto bad;
	default:
		abort();
	}
	if ((*db->sync)(db, 0) == -1) {
bad:
		syslog(LOG_ERR, "database corrupted %m");
		die(1);
	}
}

/* change_host -------------------------------------------------------------- */
/*
 * Purpose:	Update/Create an entry for host
 * Returns:	Nothing
 * Notes:
 *
 */
void
change_host(char *hostnamep, HostInfo *hp)
{
	DBT key, data;
	char *ptr;
	char hostname[HOST_NAME_MAX+1 + 1];
	HostInfo h;

	strlcpy(hostname, hostnamep, sizeof(hostname));
	h = *hp;

	for (ptr = hostname; *ptr; ptr++)
		if (isupper((unsigned char) *ptr))
			*ptr = tolower((unsigned char) *ptr);

	key.data = hostname;
	key.size = ptr - hostname + 1;
	data.data = &h;
	data.size = sizeof(h);

	switch ((*db->put)(db, &key, &data, 0)) {
	case -1:
		syslog(LOG_ERR, "database corrupted %m");
		die(1);
	case 0:
		return;
	default:
		abort();
	}
}


/* find_host -------------------------------------------------------------- */
/*
 * Purpose:	Find the entry in the status file for a given host
 * Returns:	Copy of entry in hd, or NULL
 * Notes:
 *
 */
HostInfo *
find_host(char *hostname, HostInfo *hp)
{
	DBT key, data;
	char *ptr;

	for (ptr = hostname; *ptr; ptr++)
		if (isupper((unsigned char) *ptr))
			*ptr = tolower((unsigned char) *ptr);

	key.data = hostname;
	key.size = ptr - hostname + 1;
	switch ((*db->get)(db, &key, &data, 0)) {
	case 0:
		if (data.size != sizeof(*hp))
			goto bad;
		return memcpy(hp, data.data, sizeof(*hp));
	case 1:
		return NULL;
	case -1:
		goto bad;
	default:
		abort();
	}

bad:
	syslog(LOG_ERR, "Database corrupted %m");
	return NULL;
}

/* walk_one ------------------------------------------------------------- */
/*
 * Purpose:	Call the given function if the element is valid
 * Returns:	Nothing - exits on error
 * Notes:
 */
static int
walk_one(int (*fun)(DBT *, HostInfo *, void *), DBT *key, DBT *data, void *ptr)
{
	HostInfo h;
	if (key->size == undefkey.size &&
	    memcmp(key->data, undefkey.data, key->size) == 0)
		return 0;
	if (data->size != sizeof(HostInfo)) {
		syslog(LOG_ERR, "Bad data in database");
		die(1);
	}
	memcpy(&h, data->data, sizeof(h));
	return (*fun)(key, &h, ptr);
}

/* walk_db -------------------------------------------------------------- */
/*
 * Purpose:	Iterate over all elements calling the given function
 * Returns:	-1 if function failed, 0 on success
 * Notes:
 */
static int
walk_db(int (*fun)(DBT *, HostInfo *, void *), void *ptr)
{
	DBT key, data;

	switch ((*db->seq)(db, &key, &data, R_FIRST)) {
	case -1:
		goto bad;
	case 1:
		/* We should have at least the magic entry at this point */
		abort();
	case 0:
		if (walk_one(fun, &key, &data, ptr) == -1)
			return -1;
		break;
	default:
		abort();
	}

	for (;;)
		switch ((*db->seq)(db, &key, &data, R_NEXT)) {
		case -1:
			goto bad;
		case 0:
			if (walk_one(fun, &key, &data, ptr) == -1)
				return -1;
			break;
		case 1:
			return 0;
		default:
			abort();
		}
bad:
	syslog(LOG_ERR, "Corrupted database %m");
	die(1);
}

/* reset_host ------------------------------------------------------------ */
/*
 * Purpose:	Clean up existing hosts in file.
 * Returns:	Always success 0.
 * Notes:	Clean-up of existing file - monitored hosts will have a
 *		pointer to a list of clients, which refers to memory in
 *		the previous incarnation of the program and so are
 *		meaningless now.  These pointers are zeroed and the fact
 *		that the host was previously monitored is recorded by
 *		setting the notifyReqd flag, which will in due course
 *		cause a SM_NOTIFY to be sent.
 *
 *		Note that if we crash twice in quick succession, some hosts
 *		may already have notifyReqd set, where we didn't manage to
 *		notify them before the second crash occurred.
 */
static int
reset_host(DBT *key, HostInfo *hi, void *ptr)
{
	if (hi->monList) {
		hi->notifyReqd = *(time_t *) ptr;
		hi->attempts = 0;
		hi->monList = NULL;
		change_host((char *)key->data, hi);
	}
	return 0;
}

/* check_work ------------------------------------------------------------ */
/*
 * Purpose:	Check if there is work to be done.
 * Returns:	0 if there is no work to be done -1 if there is.
 * Notes:
 */
static int
check_work(DBT *key, HostInfo *hi, void *ptr)
{
	return hi->notifyReqd ? -1 : 0;
}

/* unmon_host ------------------------------------------------------------ */
/*
 * Purpose:	Unmonitor a host
 * Returns:	0
 * Notes:
 */
static int
unmon_host(DBT *key, HostInfo *hi, void *ptr)
{
	char *name = key->data;

	if (do_unmon(name, hi, ptr))
		change_host(name, hi);
	return 0;
}

/* notify_one ------------------------------------------------------------ */
/*
 * Purpose:	Notify one host.
 * Returns:	0 if success -1 on failure
 * Notes:
 */
static int
notify_one(DBT *key, HostInfo *hi, void *ptr)
{
	time_t now = *(time_t *) ptr;
	char *name = key->data;
	int error;

	if (hi->notifyReqd == 0 || hi->notifyReqd > now)
		return 0;

	/*
	 * If one of the initial attempts fails, we wait
	 * for a while and have another go.  This is necessary
	 * because when we have crashed, (eg. a power outage)
	 * it is quite possible that we won't be able to
	 * contact all monitored hosts immediately on restart,
	 * either because they crashed too and take longer
	 * to come up (in which case the notification isn't
	 * really required), or more importantly if some
	 * router etc. needed to reach the monitored host
	 * has not come back up yet.  In this case, we will
	 * be a bit late in re-establishing locks (after the
	 * grace period) but that is the best we can do.  We
	 * try 10 times at 5 sec intervals, 10 more times at
	 * 1 minute intervals, then 24 more times at hourly
	 * intervals, finally giving up altogether if the
	 * host hasn't come back to life after 24 hours.
	 */
	if (notify_one_host(name) || hi->attempts++ >= 44) {
		error = 0;
		hi->notifyReqd = 0;
		hi->attempts = 0;
	} else {
		error = -1;
		if (hi->attempts < 10)
			hi->notifyReqd += 5;
		else if (hi->attempts < 20)
			hi->notifyReqd += 60;
		else
			hi->notifyReqd += 60 * 60;
	}
	change_host(name, hi);
	return error;
}

/* init_file -------------------------------------------------------------- */
/*
 * Purpose:	Open file, create if necessary, initialise it.
 * Returns:	Nothing - exits on error
 * Notes:	Called before process becomes daemon, hence logs to
 *		stderr rather than syslog.
 *		Opens the file, then mmap()s it for ease of access.
 *		Also performs initial clean-up of the file, zeroing
 *		monitor list pointers, setting the notifyReqd flag in
 *		all hosts that had a monitor list, and incrementing
 *		the state number to the next even value.
 */
static void
init_file(char *filename)
{
	DBT data;

	db = dbopen(filename, O_RDWR|O_CREAT|O_NDELAY|O_EXLOCK, 0644, DB_HASH,
	    NULL);
	if (db == NULL)
		err(1, "Cannot open `%s'", filename);

	switch ((*db->get)(db, &undefkey, &data, 0)) {
	case 1:
		/* New database */
		memset(&status_info, 0, sizeof(status_info));
		sync_file();
		return;
	case -1:
		err(1, "error accessing database (%m)");
	case 0:
		/* Existing database */
		if (data.size != sizeof(status_info))
			errx(1, "database corrupted %lu != %lu",
			    (u_long)data.size, (u_long)sizeof(status_info));
		memcpy(&status_info, data.data, data.size);
		break;
	default:
		abort();
	}

	reset_database();
	return;
}

/* reset_database --------------------------------------------------------- */
/*
 * Purpose:	Clears the statd database
 * Returns:	Nothing
 * Notes:	If this is not called on reset, it will leak memory.
 */
void
reset_database(void)
{
	time_t now = time(NULL);
	walk_db(reset_host, &now);

	/* Select the next higher even number for the state counter */
	status_info.ourState =
	    (status_info.ourState + 2) & 0xfffffffe;
	status_info.ourState++;	/* XXX - ??? */
	sync_file();
}

/* unmon_hosts --------------------------------------------------------- */
/*
 * Purpose:	Unmonitor all the hosts
 * Returns:	Nothing
 * Notes:
 */
void
unmon_hosts(void)
{
	time_t now = time(NULL);
	walk_db(unmon_host, &now);
	sync_file();
}

static int
notify_one_host(char *hostname)
{
	struct timeval timeout = {20, 0};	/* 20 secs timeout */
	CLIENT *cli;
	char dummy;
	stat_chge arg;
	char our_hostname[HOST_NAME_MAX+1 + 1];

	gethostname(our_hostname, sizeof(our_hostname));
	our_hostname[sizeof(our_hostname) - 1] = '\0';
	arg.mon_name = our_hostname;
	arg.state = status_info.ourState;

	if (debug)
		syslog(LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s",
		    hostname, our_hostname);

	cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
	if (!cli) {
		syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
		    clnt_spcreateerror(""));
		return (FALSE);
	}
	if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void,
	    &dummy, timeout) != RPC_SUCCESS) {
		syslog(LOG_ERR, "Failed to contact rpc.statd at host %s",
		    hostname);
		clnt_destroy(cli);
		return (FALSE);
	}
	clnt_destroy(cli);
	return (TRUE);
}

static __dead void
die(int n)
{
	(*db->close)(db);
	exit(n);
}


Generated by: GCOVR (Version 3.3)

Line	Branch	Exec	Source
1			/* $OpenBSD: statd.c,v 1.4 2017/01/21 08:33:51 krw Exp $ */
2
3			/*
4			* Copyright (c) 1997 Christos Zoulas. All rights reserved.
5			* Copyright (c) 1995
6			* A.R. Gordon (andrew.gordon@net-tel.co.uk). All rights reserved.
7			*
8			* Redistribution and use in source and binary forms, with or without
9			* modification, are permitted provided that the following conditions
10			* are met:
11			* 1. Redistributions of source code must retain the above copyright
12			* notice, this list of conditions and the following disclaimer.
13			* 2. Redistributions in binary form must reproduce the above copyright
14			* notice, this list of conditions and the following disclaimer in the
15			* documentation and/or other materials provided with the distribution.
16			* 3. All advertising materials mentioning features or use of this software
17			* must display the following acknowledgement:
18			* This product includes software developed for the FreeBSD project
19			* This product includes software developed by Christos Zoulas.
20			* 4. Neither the name of the author nor the names of any co-contributors
21			* may be used to endorse or promote products derived from this software
22			* without specific prior written permission.
23			*
24			* THIS SOFTWARE IS PROVIDED BY ANDREW GORDON AND CONTRIBUTORS ``AS IS'' AND
25			* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26			* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27			* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28			* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29			* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30			* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31			* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32			* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33			* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34			* SUCH DAMAGE.
35			*
36			*/
37
38			/* main() function for status monitor daemon. Some of the code in this */
39			/* file was generated by running rpcgen /usr/include/rpcsvc/sm_inter.x */
40			/* The actual program logic is in the file procs.c */
41
42			#include <sys/wait.h>
43
44			#include <err.h>
45			#include <ctype.h>
46			#include <errno.h>
47			#include <fcntl.h>
48			#include <signal.h>
49			#include <stdio.h>
50			#include <stdlib.h>
51			#include <string.h>
52			#include <syslog.h>
53			#include <unistd.h>
54			#include <db.h>
55
56			#include <rpc/rpc.h>
57
58			#include "statd.h"
59
60			struct sigaction sa;
61			int debug = 0; /* Controls syslog() for debug msgs */
62			int _rpcsvcdirty = 0; /* XXX ??? */
63			static DB db; / Database file */
64
65			Header status_info;
66
67			static char undefdata[] = "\0\1\2\3\4\5\6\7";
68			static DBT undefkey = {
69			undefdata,
70			sizeof(undefdata)
71			};
72			extern char *__progname;
73
74			/* statd.c */
75			static int walk_one(int (fun )(DBT , HostInfo , void ), DBT , DBT , void *);
76			static int walk_db(int (fun )(DBT , HostInfo , void ), void *);
77			static int reset_host(DBT , HostInfo , void *);
78			static int check_work(DBT , HostInfo , void *);
79			static int unmon_host(DBT , HostInfo , void *);
80			static int notify_one(DBT , HostInfo , void *);
81			static void init_file(char *);
82			static int notify_one_host(char *);
83			static __dead void die(int);
84
85			int main(int, char **);
86
87			int
88			main(int argc, char **argv)
89			{
90			SVCXPRT *transp;
91			int ch;
92			struct sigaction nsa;
93
94			while ((ch = getopt(argc, argv, "d")) != (-1)) {
95			switch (ch) {
96			case 'd':
97			debug = 1;
98			break;
99			default:
100			case '?':
101			fprintf(stderr, "usage: %s [-d]\n", __progname);
102			exit(1);
103			/* NOTREACHED */
104			}
105			}
106			pmap_unset(SM_PROG, SM_VERS);
107
108			transp = svcudp_create(RPC_ANYSOCK);
109			if (transp == NULL) {
110			errx(1, "cannot create udp service.");
111			/* NOTREACHED */
112			}
113			if (!svc_register(transp, SM_PROG, SM_VERS, sm_prog_1, IPPROTO_UDP)) {
114			errx(1, "unable to register (SM_PROG, SM_VERS, udp).");
115			/* NOTREACHED */
116			}
117			transp = svctcp_create(RPC_ANYSOCK, 0, 0);
118			if (transp == NULL) {
119			errx(1, "cannot create tcp service.");
120			/* NOTREACHED */
121			}
122			if (!svc_register(transp, SM_PROG, SM_VERS, sm_prog_1, IPPROTO_TCP)) {
123			errx(1, "unable to register (SM_PROG, SM_VERS, tcp).");
124			/* NOTREACHED */
125			}
126
127			init_file("/var/db/statd.status");
128
129			/*
130			* Note that it is NOT sensible to run this program from inetd - the
131			* protocol assumes that it will run immediately at boot time.
132			*/
133			daemon(0, 0);
134
135			sigemptyset(&nsa.sa_mask);
136			nsa.sa_flags = SA_NOCLDSTOP\|SA_NOCLDWAIT;
137			nsa.sa_handler = SIG_IGN;
138			sigaction(SIGCHLD, &nsa, NULL);
139
140			openlog("rpc.statd", 0, LOG_DAEMON);
141			if (debug)
142			syslog(LOG_INFO, "Starting - debug enabled");
143			else
144			syslog(LOG_INFO, "Starting");
145
146			sa.sa_handler = die;
147			sa.sa_flags = 0;
148			sigemptyset(&sa.sa_mask);
149			sigaction(SIGTERM, &sa, NULL);
150			sigaction(SIGQUIT, &sa, NULL);
151			sigaction(SIGHUP, &sa, NULL);
152			sigaction(SIGINT, &sa, NULL);
153
154			sa.sa_handler = SIG_IGN;
155			sa.sa_flags = SA_RESTART;
156			sigemptyset(&sa.sa_mask);
157			sigaddset(&sa.sa_mask, SIGALRM);
158
159			/* Initialisation now complete - start operating */
160
161			/* Notify hosts that need it */
162			notify_handler(0);
163
164			while (1)
165			svc_run(); /* Should never return */
166			die(0);
167			}
168
169			/* notify_handler ---------------------------------------------------------- */
170			/*
171			* Purpose: Catch SIGALRM and collect process status
172			* Returns: Nothing.
173			* Notes: No special action required, other than to collect the
174			* process status and hence allow the child to die:
175			* we only use child processes for asynchronous transmission
176			* of SM_NOTIFY to other systems, so it is normal for the
177			* children to exit when they have done their work.
178			*/
179			void
180			notify_handler(int sig)
181			{
182			time_t now;
183
184			NO_ALARM;
185			sa.sa_handler = SIG_IGN;
186			sigaction(SIGALRM, &sa, NULL);
187
188			now = time(NULL);
189
190			walk_db(notify_one, &now);
191
192			if (walk_db(check_work, &now) == 0) {
193			/*
194			* No more work to be done.
195			*/
196			CLR_ALARM;
197			return;
198			}
199			sync_file();
200			ALARM;
201			alarm(5);
202			}
203
204			/* sync_file --------------------------------------------------------------- */
205			/*
206			* Purpose: Packaged call of msync() to flush changes to mmap()ed file
207			* Returns: Nothing. Errors to syslog.
208			*/
209			void
210			sync_file()
211			{
212			DBT data;
213
214			data.data = &status_info;
215			data.size = sizeof(status_info);
216			switch ((*db->put)(db, &undefkey, &data, 0)) {
217			case 0:
218			return;
219			case -1:
220			goto bad;
221			default:
222			abort();
223			}
224			if ((*db->sync)(db, 0) == -1) {
225			bad:
226			syslog(LOG_ERR, "database corrupted %m");
227			die(1);
228			}
229			}
230
231			/* change_host -------------------------------------------------------------- */
232			/*
233			* Purpose: Update/Create an entry for host
234			* Returns: Nothing
235			* Notes:
236			*
237			*/
238			void
239			change_host(char hostnamep, HostInfo hp)
240			{
241			DBT key, data;
242			char *ptr;
243			char hostname[HOST_NAME_MAX+1 + 1];
244			HostInfo h;
245
246			strlcpy(hostname, hostnamep, sizeof(hostname));
247			h = *hp;
248
249			for (ptr = hostname; *ptr; ptr++)
250			if (isupper((unsigned char) *ptr))
251			ptr = tolower((unsigned char) ptr);
252
253			key.data = hostname;
254			key.size = ptr - hostname + 1;
255			data.data = &h;
256			data.size = sizeof(h);
257
258			switch ((*db->put)(db, &key, &data, 0)) {
259			case -1:
260			syslog(LOG_ERR, "database corrupted %m");
261			die(1);
262			case 0:
263			return;
264			default:
265			abort();
266			}
267			}
268
269
270			/* find_host -------------------------------------------------------------- */
271			/*
272			* Purpose: Find the entry in the status file for a given host
273			* Returns: Copy of entry in hd, or NULL
274			* Notes:
275			*
276			*/
277			HostInfo *
278			find_host(char hostname, HostInfo hp)
279			{
280			DBT key, data;
281			char *ptr;
282
283			for (ptr = hostname; *ptr; ptr++)
284			if (isupper((unsigned char) *ptr))
285			ptr = tolower((unsigned char) ptr);
286
287			key.data = hostname;
288			key.size = ptr - hostname + 1;
289			switch ((*db->get)(db, &key, &data, 0)) {
290			case 0:
291			if (data.size != sizeof(*hp))
292			goto bad;
293			return memcpy(hp, data.data, sizeof(*hp));
294			case 1:
295			return NULL;
296			case -1:
297			goto bad;
298			default:
299			abort();
300			}
301
302			bad:
303			syslog(LOG_ERR, "Database corrupted %m");
304			return NULL;
305			}
306
307			/* walk_one ------------------------------------------------------------- */
308			/*
309			* Purpose: Call the given function if the element is valid
310			* Returns: Nothing - exits on error
311			* Notes:
312			*/
313			static int
314			walk_one(int (fun)(DBT , HostInfo , void ), DBT key, DBT data, void *ptr)
315			{
316			HostInfo h;
317			if (key->size == undefkey.size &&
318			memcmp(key->data, undefkey.data, key->size) == 0)
319			return 0;
320			if (data->size != sizeof(HostInfo)) {
321			syslog(LOG_ERR, "Bad data in database");
322			die(1);
323			}
324			memcpy(&h, data->data, sizeof(h));
325			return (*fun)(key, &h, ptr);
326			}
327
328			/* walk_db -------------------------------------------------------------- */
329			/*
330			* Purpose: Iterate over all elements calling the given function
331			* Returns: -1 if function failed, 0 on success
332			* Notes:
333			*/
334			static int
335			walk_db(int (fun)(DBT , HostInfo , void ), void *ptr)
336			{
337			DBT key, data;
338
339			switch ((*db->seq)(db, &key, &data, R_FIRST)) {
340			case -1:
341			goto bad;
342			case 1:
343			/* We should have at least the magic entry at this point */
344			abort();
345			case 0:
346			if (walk_one(fun, &key, &data, ptr) == -1)
347			return -1;
348			break;
349			default:
350			abort();
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
388			reset_host(DBT key, HostInfo hi, void *ptr)
389			{
390			if (hi->monList) {
391			hi->notifyReqd = (time_t ) ptr;
392			hi->attempts = 0;
393			hi->monList = NULL;
394			change_host((char *)key->data, hi);
395			}
396			return 0;
397			}
398
399			/* check_work ------------------------------------------------------------ */
400			/*
401			* Purpose: Check if there is work to be done.
402			* Returns: 0 if there is no work to be done -1 if there is.
403			* Notes:
404			*/
405			static int
406			check_work(DBT key, HostInfo hi, void *ptr)
407			{
408			return hi->notifyReqd ? -1 : 0;
409			}
410
411			/* unmon_host ------------------------------------------------------------ */
412			/*
413			* Purpose: Unmonitor a host
414			* Returns: 0
415			* Notes:
416			*/
417			static int
418			unmon_host(DBT key, HostInfo hi, void *ptr)
419			{
420			char *name = key->data;
421
422			if (do_unmon(name, hi, ptr))
423			change_host(name, hi);
424			return 0;
425			}
426
427			/* notify_one ------------------------------------------------------------ */
428			/*
429			* Purpose: Notify one host.
430			* Returns: 0 if success -1 on failure
431			* Notes:
432			*/
433			static int
434			notify_one(DBT key, HostInfo hi, void *ptr)
435			{
436			time_t now = (time_t ) ptr;
437			char *name = key->data;
438			int error;
439
440			if (hi->notifyReqd == 0 \|\| hi->notifyReqd > now)
441			return 0;
442
443			/*
444			* If one of the initial attempts fails, we wait
445			* for a while and have another go. This is necessary
446			* because when we have crashed, (eg. a power outage)
447			* it is quite possible that we won't be able to
448			* contact all monitored hosts immediately on restart,
449			* either because they crashed too and take longer
450			* to come up (in which case the notification isn't
451			* really required), or more importantly if some
452			* router etc. needed to reach the monitored host
453			* has not come back up yet. In this case, we will
454			* be a bit late in re-establishing locks (after the
455			* grace period) but that is the best we can do. We
456			* try 10 times at 5 sec intervals, 10 more times at
457			* 1 minute intervals, then 24 more times at hourly
458			* intervals, finally giving up altogether if the
459			* host hasn't come back to life after 24 hours.
460			*/
461			if (notify_one_host(name) \|\| hi->attempts++ >= 44) {
462			error = 0;
463			hi->notifyReqd = 0;
464			hi->attempts = 0;
465			} else {
466			error = -1;
467			if (hi->attempts < 10)
468			hi->notifyReqd += 5;
469			else if (hi->attempts < 20)
470			hi->notifyReqd += 60;
471			else
472			hi->notifyReqd += 60 * 60;
473			}
474			change_host(name, hi);
475			return error;
476			}
477
478			/* init_file -------------------------------------------------------------- */
479			/*
480			* Purpose: Open file, create if necessary, initialise it.
481			* Returns: Nothing - exits on error
482			* Notes: Called before process becomes daemon, hence logs to
483			* stderr rather than syslog.
484			* Opens the file, then mmap()s it for ease of access.
485			* Also performs initial clean-up of the file, zeroing
486			* monitor list pointers, setting the notifyReqd flag in
487			* all hosts that had a monitor list, and incrementing
488			* the state number to the next even value.
489			*/
490			static void
491			init_file(char *filename)
492			{
493			DBT data;
494
495			db = dbopen(filename, O_RDWR\|O_CREAT\|O_NDELAY\|O_EXLOCK, 0644, DB_HASH,
496			NULL);
497			if (db == NULL)
498			err(1, "Cannot open `%s'", filename);
499
500			switch ((*db->get)(db, &undefkey, &data, 0)) {
501			case 1:
502			/* New database */
503			memset(&status_info, 0, sizeof(status_info));
504			sync_file();
505			return;
506			case -1:
507			err(1, "error accessing database (%m)");
508			case 0:
509			/* Existing database */
510			if (data.size != sizeof(status_info))
511			errx(1, "database corrupted %lu != %lu",
512			(u_long)data.size, (u_long)sizeof(status_info));
513			memcpy(&status_info, data.data, data.size);
514			break;
515			default:
516			abort();
517			}
518
519			reset_database();
520			return;
521			}
522
523			/* reset_database --------------------------------------------------------- */
524			/*
525			* Purpose: Clears the statd database
526			* Returns: Nothing
527			* Notes: If this is not called on reset, it will leak memory.
528			*/
529			void
530			reset_database(void)
531			{
532			time_t now = time(NULL);
533			walk_db(reset_host, &now);
534
535			/* Select the next higher even number for the state counter */
536			status_info.ourState =
537			(status_info.ourState + 2) & 0xfffffffe;
538			status_info.ourState++; /* XXX - ??? */
539			sync_file();
540			}
541
542			/* unmon_hosts --------------------------------------------------------- */
543			/*
544			* Purpose: Unmonitor all the hosts
545			* Returns: Nothing
546			* Notes:
547			*/
548			void
549			unmon_hosts(void)
550			{
551			time_t now = time(NULL);
552			walk_db(unmon_host, &now);
553			sync_file();
554			}
555
556			static int
557			notify_one_host(char *hostname)
558			{
559			struct timeval timeout = {20, 0}; /* 20 secs timeout */
560			CLIENT *cli;
561			char dummy;
562			stat_chge arg;
563			char our_hostname[HOST_NAME_MAX+1 + 1];
564
565			gethostname(our_hostname, sizeof(our_hostname));
566			our_hostname[sizeof(our_hostname) - 1] = '\0';
567			arg.mon_name = our_hostname;
568			arg.state = status_info.ourState;
569
570			if (debug)
571			syslog(LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s",
572			hostname, our_hostname);
573
574			cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
575			if (!cli) {
576			syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
577			clnt_spcreateerror(""));
578			return (FALSE);
579			}
580			if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void,
581			&dummy, timeout) != RPC_SUCCESS) {
582			syslog(LOG_ERR, "Failed to contact rpc.statd at host %s",
583			hostname);
584			clnt_destroy(cli);
585			return (FALSE);
586			}
587			clnt_destroy(cli);
588			return (TRUE);
589			}
590
591			static __dead void
592			die(int n)
593			{
594			(*db->close)(db);
595			exit(n);
596			}