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