xref: /freebsd/usr.sbin/rpc.statd/file.c (revision 1669d8afc64812c8d2d1d147ae1fd42ff441e1b1)
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  * $FreeBSD$
33  *
34  */
35 
36 #include <err.h>
37 #include <errno.h>
38 #include <fcntl.h>
39 #include <stdio.h>
40 #include <string.h>
41 #include <unistd.h>
42 #include <sys/types.h>
43 #include <sys/mman.h>		/* For mmap()				*/
44 #include <rpc/rpc.h>
45 #include <syslog.h>
46 #include <stdlib.h>
47 
48 #include "statd.h"
49 
50 FileLayout *status_info;	/* Pointer to the mmap()ed status file	*/
51 static int status_fd;		/* File descriptor for the open file	*/
52 static off_t status_file_len;	/* Current on-disc length of file	*/
53 
54 /* sync_file --------------------------------------------------------------- */
55 /*
56    Purpose:	Packaged call of msync() to flush changes to mmap()ed file
57    Returns:	Nothing.  Errors to syslog.
58 */
59 
60 void sync_file(void)
61 {
62   if (msync((void *)status_info, 0, 0) < 0)
63   {
64     syslog(LOG_ERR, "msync() failed: %s", strerror(errno));
65   }
66 }
67 
68 /* find_host -------------------------------------------------------------- */
69 /*
70    Purpose:	Find the entry in the status file for a given host
71    Returns:	Pointer to that entry in the mmap() region, or NULL.
72    Notes:	Also creates entries if requested.
73 		Failure to create also returns NULL.
74 */
75 
76 HostInfo *find_host(char *hostname, int create)
77 {
78   HostInfo *hp;
79   HostInfo *spare_slot = NULL;
80   HostInfo *result = NULL;
81   int i;
82 
83   for (i = 0, hp = status_info->hosts; i < status_info->noOfHosts; i++, hp++)
84   {
85     if (!strncasecmp(hostname, hp->hostname, SM_MAXSTRLEN))
86     {
87       result = hp;
88       break;
89     }
90     if (!spare_slot && !hp->monList && !hp->notifyReqd)
91       spare_slot = hp;
92   }
93 
94   /* Return if entry found, or if not asked to create one.		*/
95   if (result || !create) return (result);
96 
97   /* Now create an entry, using the spare slot if one was found or	*/
98   /* adding to the end of the list otherwise, extending file if reqd	*/
99   if (!spare_slot)
100   {
101     off_t desired_size;
102     spare_slot = &status_info->hosts[status_info->noOfHosts];
103     desired_size = ((char*)spare_slot - (char*)status_info) + sizeof(HostInfo);
104     if (desired_size > status_file_len)
105     {
106       /* Extend file by writing 1 byte of junk at the desired end pos	*/
107       lseek(status_fd, desired_size - 1, SEEK_SET);
108       i = write(status_fd, &i, 1);
109       if (i < 1)
110       {
111 	syslog(LOG_ERR, "Unable to extend status file");
112 	return (NULL);
113       }
114       status_file_len = desired_size;
115     }
116     status_info->noOfHosts++;
117   }
118 
119   /* Initialise the spare slot that has been found/created		*/
120   /* Note that we do not msync(), since the caller is presumed to be	*/
121   /* about to modify the entry further					*/
122   memset(spare_slot, 0, sizeof(HostInfo));
123   strncpy(spare_slot->hostname, hostname, SM_MAXSTRLEN);
124   return (spare_slot);
125 }
126 
127 /* init_file -------------------------------------------------------------- */
128 /*
129    Purpose:	Open file, create if necessary, initialise it.
130    Returns:	Nothing - exits on error
131    Notes:	Called before process becomes daemon, hence logs to
132 		stderr rather than syslog.
133 		Opens the file, then mmap()s it for ease of access.
134 		Also performs initial clean-up of the file, zeroing
135 		monitor list pointers, setting the notifyReqd flag in
136 		all hosts that had a monitor list, and incrementing
137 		the state number to the next even value.
138 */
139 
140 void init_file(const char *filename)
141 {
142   int new_file = FALSE;
143   char buf[HEADER_LEN];
144   int i;
145 
146   /* try to open existing file - if not present, create one		*/
147   status_fd = open(filename, O_RDWR);
148   if ((status_fd < 0) && (errno == ENOENT))
149   {
150     status_fd = open(filename, O_RDWR | O_CREAT, 0644);
151     new_file = TRUE;
152   }
153   if (status_fd < 0)
154     errx(1, "unable to open status file %s", filename);
155 
156   /* File now open.  mmap() it, with a generous size to allow for	*/
157   /* later growth, where we will extend the file but not re-map it.	*/
158   status_info = (FileLayout *)
159     mmap(NULL, 0x10000000, PROT_READ | PROT_WRITE, MAP_SHARED, status_fd, 0);
160 
161   if (status_info == (FileLayout *) MAP_FAILED)
162     err(1, "unable to mmap() status file");
163 
164   status_file_len = lseek(status_fd, 0L, SEEK_END);
165 
166   /* If the file was not newly created, validate the contents, and if	*/
167   /* defective, re-create from scratch.					*/
168   if (!new_file)
169   {
170     if ((status_file_len < HEADER_LEN) || (status_file_len
171       < (HEADER_LEN + sizeof(HostInfo) * status_info->noOfHosts)) )
172     {
173       warnx("status file is corrupt");
174       new_file = TRUE;
175     }
176   }
177 
178   /* Initialisation of a new, empty file.				*/
179   if (new_file)
180   {
181     memset(buf, 0, sizeof(buf));
182     lseek(status_fd, 0L, SEEK_SET);
183     write(status_fd, buf, HEADER_LEN);
184     status_file_len = HEADER_LEN;
185   }
186   else
187   {
188     /* Clean-up of existing file - monitored hosts will have a pointer	*/
189     /* to a list of clients, which refers to memory in the previous	*/
190     /* incarnation of the program and so are meaningless now.  These	*/
191     /* pointers are zeroed and the fact that the host was previously	*/
192     /* monitored is recorded by setting the notifyReqd flag, which will	*/
193     /* in due course cause a SM_NOTIFY to be sent.			*/
194     /* Note that if we crash twice in quick succession, some hosts may	*/
195     /* already have notifyReqd set, where we didn't manage to notify	*/
196     /* them before the second crash occurred.				*/
197     for (i = 0; i < status_info->noOfHosts; i++)
198     {
199       HostInfo *this_host = &status_info->hosts[i];
200 
201       if (this_host->monList)
202       {
203 	this_host->notifyReqd = TRUE;
204 	this_host->monList = NULL;
205       }
206     }
207     /* Select the next higher even number for the state counter		*/
208     status_info->ourState = (status_info->ourState + 2) & 0xfffffffe;
209 /*???????******/ status_info->ourState++;
210   }
211 }
212 
213 /* notify_one_host --------------------------------------------------------- */
214 /*
215    Purpose:	Perform SM_NOTIFY procedure at specified host
216    Returns:	TRUE if success, FALSE if failed.
217 */
218 
219 static int notify_one_host(char *hostname)
220 {
221   struct timeval timeout = { 20, 0 };	/* 20 secs timeout		*/
222   CLIENT *cli;
223   char dummy;
224   stat_chge arg;
225   char our_hostname[SM_MAXSTRLEN+1];
226 
227   gethostname(our_hostname, sizeof(our_hostname));
228   our_hostname[SM_MAXSTRLEN] = '\0';
229   arg.mon_name = our_hostname;
230   arg.state = status_info->ourState;
231 
232   if (debug) syslog (LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s", hostname, our_hostname);
233 
234   cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
235   if (!cli)
236   {
237     syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
238       clnt_spcreateerror(""));
239     return (FALSE);
240   }
241 
242   if (clnt_call(cli, SM_NOTIFY, (xdrproc_t)xdr_stat_chge, &arg,
243       (xdrproc_t)xdr_void, &dummy, timeout)
244     != RPC_SUCCESS)
245   {
246     syslog(LOG_ERR, "Failed to contact rpc.statd at host %s", hostname);
247     clnt_destroy(cli);
248     return (FALSE);
249   }
250 
251   clnt_destroy(cli);
252   return (TRUE);
253 }
254 
255 /* notify_hosts ------------------------------------------------------------ */
256 /*
257    Purpose:	Send SM_NOTIFY to all hosts marked as requiring it
258    Returns:	Nothing, immediately - forks a process to do the work.
259    Notes:	Does nothing if there are no monitored hosts.
260 		Called after all the initialisation has been done -
261 		logs to syslog.
262 */
263 
264 void notify_hosts(void)
265 {
266   int i;
267   int attempts;
268   int work_to_do = FALSE;
269   HostInfo *hp;
270   pid_t pid;
271 
272   /* First check if there is in fact any work to do.			*/
273   for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
274   {
275     if (hp->notifyReqd)
276     {
277       work_to_do = TRUE;
278       break;
279     }
280   }
281 
282   if (!work_to_do) return;	/* No work found			*/
283 
284   pid = fork();
285   if (pid == -1)
286   {
287     syslog(LOG_ERR, "Unable to fork notify process - %s", strerror(errno));
288     return;
289   }
290   if (pid) return;
291 
292   /* Here in the child process.  We continue until all the hosts marked	*/
293   /* as requiring notification have been duly notified.			*/
294   /* If one of the initial attempts fails, we sleep for a while and	*/
295   /* have another go.  This is necessary because when we have crashed,	*/
296   /* (eg. a power outage) it is quite possible that we won't be able to	*/
297   /* contact all monitored hosts immediately on restart, either because	*/
298   /* they crashed too and take longer to come up (in which case the	*/
299   /* notification isn't really required), or more importantly if some	*/
300   /* router etc. needed to reach the monitored host has not come back	*/
301   /* up yet.  In this case, we will be a bit late in re-establishing	*/
302   /* locks (after the grace period) but that is the best we can do.	*/
303   /* We try 10 times at 5 sec intervals, 10 more times at 1 minute	*/
304   /* intervals, then 24 more times at hourly intervals, finally		*/
305   /* giving up altogether if the host hasn't come back to life after	*/
306   /* 24 hours.								*/
307 
308   for (attempts = 0; attempts < 44; attempts++)
309   {
310     work_to_do = FALSE;		/* Unless anything fails		*/
311     for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
312     {
313       if (hp->notifyReqd)
314       {
315         if (notify_one_host(hp->hostname))
316 	{
317 	  hp->notifyReqd = FALSE;
318           sync_file();
319 	}
320 	else work_to_do = TRUE;
321       }
322     }
323     if (!work_to_do) break;
324     if (attempts < 10) sleep(5);
325     else if (attempts < 20) sleep(60);
326     else sleep(60*60);
327   }
328   exit(0);
329 }
330 
331 
332