xref: /freebsd/usr.sbin/rpc.statd/file.c (revision daf1cffce2e07931f27c6c6998652e90df6ba87e)
1 /*
2  * Copyright (c) 1995
3  *	A.R. Gordon (andrew.gordon@net-tel.co.uk).  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed for the FreeBSD project
16  * 4. Neither the name of the author nor the names of any co-contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY ANDREW GORDON AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  */
33 
34 #include <err.h>
35 #include <errno.h>
36 #include <fcntl.h>
37 #include <stdio.h>
38 #include <string.h>
39 #include <unistd.h>
40 #include <sys/types.h>
41 #include <sys/mman.h>		/* For mmap()				*/
42 #include <rpc/rpc.h>
43 #include <syslog.h>
44 
45 #include "statd.h"
46 
47 FileLayout *status_info;	/* Pointer to the mmap()ed status file	*/
48 static int status_fd;		/* File descriptor for the open file	*/
49 static off_t status_file_len;	/* Current on-disc length of file	*/
50 
51 /* sync_file --------------------------------------------------------------- */
52 /*
53    Purpose:	Packaged call of msync() to flush changes to mmap()ed file
54    Returns:	Nothing.  Errors to syslog.
55 */
56 
57 void sync_file(void)
58 {
59   if (msync((void *)status_info, 0, 0) < 0)
60   {
61     syslog(LOG_ERR, "msync() failed: %s", strerror(errno));
62   }
63 }
64 
65 /* find_host -------------------------------------------------------------- */
66 /*
67    Purpose:	Find the entry in the status file for a given host
68    Returns:	Pointer to that entry in the mmap() region, or NULL.
69    Notes:	Also creates entries if requested.
70 		Failure to create also returns NULL.
71 */
72 
73 HostInfo *find_host(char *hostname, int create)
74 {
75   HostInfo *hp;
76   HostInfo *spare_slot = NULL;
77   HostInfo *result = NULL;
78   int i;
79 
80   for (i = 0, hp = status_info->hosts; i < status_info->noOfHosts; i++, hp++)
81   {
82     if (!strncasecmp(hostname, hp->hostname, SM_MAXSTRLEN))
83     {
84       result = hp;
85       break;
86     }
87     if (!spare_slot && !hp->monList && !hp->notifyReqd)
88       spare_slot = hp;
89   }
90 
91   /* Return if entry found, or if not asked to create one.		*/
92   if (result || !create) return (result);
93 
94   /* Now create an entry, using the spare slot if one was found or	*/
95   /* adding to the end of the list otherwise, extending file if reqd	*/
96   if (!spare_slot)
97   {
98     off_t desired_size;
99     spare_slot = &status_info->hosts[status_info->noOfHosts];
100     desired_size = ((char*)spare_slot - (char*)status_info) + sizeof(HostInfo);
101     if (desired_size > status_file_len)
102     {
103       /* Extend file by writing 1 byte of junk at the desired end pos	*/
104       lseek(status_fd, desired_size - 1, SEEK_SET);
105       i = write(status_fd, &i, 1);
106       if (i < 1)
107       {
108 	syslog(LOG_ERR, "Unable to extend status file");
109 	return (NULL);
110       }
111       status_file_len = desired_size;
112     }
113     status_info->noOfHosts++;
114   }
115 
116   /* Initialise the spare slot that has been found/created		*/
117   /* Note that we do not msync(), since the caller is presumed to be	*/
118   /* about to modify the entry further					*/
119   memset(spare_slot, 0, sizeof(HostInfo));
120   strncpy(spare_slot->hostname, hostname, SM_MAXSTRLEN);
121   return (spare_slot);
122 }
123 
124 /* init_file -------------------------------------------------------------- */
125 /*
126    Purpose:	Open file, create if necessary, initialise it.
127    Returns:	Nothing - exits on error
128    Notes:	Called before process becomes daemon, hence logs to
129 		stderr rather than syslog.
130 		Opens the file, then mmap()s it for ease of access.
131 		Also performs initial clean-up of the file, zeroing
132 		monitor list pointers, setting the notifyReqd flag in
133 		all hosts that had a monitor list, and incrementing
134 		the state number to the next even value.
135 */
136 
137 void init_file(char *filename)
138 {
139   int new_file = FALSE;
140   char buf[HEADER_LEN];
141   int i;
142 
143   /* try to open existing file - if not present, create one		*/
144   status_fd = open(filename, O_RDWR);
145   if ((status_fd < 0) && (errno == ENOENT))
146   {
147     status_fd = open(filename, O_RDWR | O_CREAT, 0644);
148     new_file = TRUE;
149   }
150   if (status_fd < 0)
151     errx(1, "unable to open status file %s", filename);
152 
153   /* File now open.  mmap() it, with a generous size to allow for	*/
154   /* later growth, where we will extend the file but not re-map it.	*/
155   status_info = (FileLayout *)
156     mmap(NULL, 0x10000000, PROT_READ | PROT_WRITE, MAP_SHARED, status_fd, 0);
157 
158   if (status_info == (FileLayout *) MAP_FAILED)
159     warn("unable to mmap() status file");
160 
161   status_file_len = lseek(status_fd, 0L, SEEK_END);
162 
163   /* If the file was not newly created, validate the contents, and if	*/
164   /* defective, re-create from scratch.					*/
165   if (!new_file)
166   {
167     if ((status_file_len < HEADER_LEN) || (status_file_len
168       < (HEADER_LEN + sizeof(HostInfo) * status_info->noOfHosts)) )
169     {
170       warnx("status file is corrupt");
171       new_file = TRUE;
172     }
173   }
174 
175   /* Initialisation of a new, empty file.				*/
176   if (new_file)
177   {
178     memset(buf, 0, sizeof(buf));
179     lseek(status_fd, 0L, SEEK_SET);
180     write(status_fd, buf, HEADER_LEN);
181     status_file_len = HEADER_LEN;
182   }
183   else
184   {
185     /* Clean-up of existing file - monitored hosts will have a pointer	*/
186     /* to a list of clients, which refers to memory in the previous	*/
187     /* incarnation of the program and so are meaningless now.  These	*/
188     /* pointers are zeroed and the fact that the host was previously	*/
189     /* monitored is recorded by setting the notifyReqd flag, which will	*/
190     /* in due course cause a SM_NOTIFY to be sent.			*/
191     /* Note that if we crash twice in quick succession, some hosts may	*/
192     /* already have notifyReqd set, where we didn't manage to notify	*/
193     /* them before the second crash occurred.				*/
194     for (i = 0; i < status_info->noOfHosts; i++)
195     {
196       HostInfo *this_host = &status_info->hosts[i];
197 
198       if (this_host->monList)
199       {
200 	this_host->notifyReqd = TRUE;
201 	this_host->monList = NULL;
202       }
203     }
204     /* Select the next higher even number for the state counter		*/
205     status_info->ourState = (status_info->ourState + 2) & 0xfffffffe;
206 /*???????******/ status_info->ourState++;
207   }
208 }
209 
210 /* xdr_stat_chge ----------------------------------------------------------- */
211 /*
212    Purpose:	XDR-encode structure of type stat_chge
213    Returns:	TRUE if successful
214    Notes:	This function is missing from librpcsvc, because the
215 		sm_inter.x distributed by Sun omits the SM_NOTIFY
216 		procedure used between co-operating statd's
217 */
218 
219 bool_t xdr_stat_chge(XDR *xdrs, stat_chge *objp)
220 {
221   if (!xdr_string(xdrs, &objp->mon_name, SM_MAXSTRLEN))
222   {
223     return (FALSE);
224   }
225   if (!xdr_int(xdrs, &objp->state))
226   {
227     return (FALSE);
228   }
229   return (TRUE);
230 }
231 
232 
233 /* notify_one_host --------------------------------------------------------- */
234 /*
235    Purpose:	Perform SM_NOTIFY procedure at specified host
236    Returns:	TRUE if success, FALSE if failed.
237 */
238 
239 static int notify_one_host(char *hostname)
240 {
241   struct timeval timeout = { 20, 0 };	/* 20 secs timeout		*/
242   CLIENT *cli;
243   char dummy;
244   stat_chge arg;
245   char our_hostname[SM_MAXSTRLEN+1];
246 
247   gethostname(our_hostname, sizeof(our_hostname));
248   our_hostname[SM_MAXSTRLEN] = '\0';
249   arg.mon_name = our_hostname;
250   arg.state = status_info->ourState;
251 
252   if (debug) syslog (LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s", hostname, our_hostname);
253 
254   cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
255   if (!cli)
256   {
257     syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
258       clnt_spcreateerror(""));
259     return (FALSE);
260   }
261 
262   if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void, &dummy, timeout)
263     != RPC_SUCCESS)
264   {
265     syslog(LOG_ERR, "Failed to contact rpc.statd at host %s", hostname);
266     clnt_destroy(cli);
267     return (FALSE);
268   }
269 
270   clnt_destroy(cli);
271   return (TRUE);
272 }
273 
274 /* notify_hosts ------------------------------------------------------------ */
275 /*
276    Purpose:	Send SM_NOTIFY to all hosts marked as requiring it
277    Returns:	Nothing, immediately - forks a process to do the work.
278    Notes:	Does nothing if there are no monitored hosts.
279 		Called after all the initialisation has been done -
280 		logs to syslog.
281 */
282 
283 void notify_hosts(void)
284 {
285   int i;
286   int attempts;
287   int work_to_do = FALSE;
288   HostInfo *hp;
289   pid_t pid;
290 
291   /* First check if there is in fact any work to do.			*/
292   for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
293   {
294     if (hp->notifyReqd)
295     {
296       work_to_do = TRUE;
297       break;
298     }
299   }
300 
301   if (!work_to_do) return;	/* No work found			*/
302 
303   pid = fork();
304   if (pid == -1)
305   {
306     syslog(LOG_ERR, "Unable to fork notify process - %s", strerror(errno));
307     return;
308   }
309   if (pid) return;
310 
311   /* Here in the child process.  We continue until all the hosts marked	*/
312   /* as requiring notification have been duly notified.			*/
313   /* If one of the initial attempts fails, we sleep for a while and	*/
314   /* have another go.  This is necessary because when we have crashed,	*/
315   /* (eg. a power outage) it is quite possible that we won't be able to	*/
316   /* contact all monitored hosts immediately on restart, either because	*/
317   /* they crashed too and take longer to come up (in which case the	*/
318   /* notification isn't really required), or more importantly if some	*/
319   /* router etc. needed to reach the monitored host has not come back	*/
320   /* up yet.  In this case, we will be a bit late in re-establishing	*/
321   /* locks (after the grace period) but that is the best we can do.	*/
322   /* We try 10 times at 5 sec intervals, 10 more times at 1 minute	*/
323   /* intervals, then 24 more times at hourly intervals, finally		*/
324   /* giving up altogether if the host hasn't come back to life after	*/
325   /* 24 hours.								*/
326 
327   for (attempts = 0; attempts < 44; attempts++)
328   {
329     work_to_do = FALSE;		/* Unless anything fails		*/
330     for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
331     {
332       if (hp->notifyReqd)
333       {
334         if (notify_one_host(hp->hostname))
335 	{
336 	  hp->notifyReqd = FALSE;
337           sync_file();
338 	}
339 	else work_to_do = TRUE;
340       }
341     }
342     if (!work_to_do) break;
343     if (attempts < 10) sleep(5);
344     else if (attempts < 20) sleep(60);
345     else sleep(60*60);
346   }
347   exit(0);
348 }
349 
350 
351