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 <netdb.h> 40 #include <stdio.h> 41 #include <string.h> 42 #include <unistd.h> 43 #include <sys/types.h> 44 #include <sys/mman.h> /* For mmap() */ 45 #include <rpc/rpc.h> 46 #include <syslog.h> 47 #include <stdlib.h> 48 49 #include "statd.h" 50 51 FileLayout *status_info; /* Pointer to the mmap()ed status file */ 52 static int status_fd; /* File descriptor for the open file */ 53 static off_t status_file_len; /* Current on-disc length of file */ 54 55 /* sync_file --------------------------------------------------------------- */ 56 /* 57 Purpose: Packaged call of msync() to flush changes to mmap()ed file 58 Returns: Nothing. Errors to syslog. 59 */ 60 61 void sync_file(void) 62 { 63 if (msync((void *)status_info, 0, 0) < 0) 64 { 65 syslog(LOG_ERR, "msync() failed: %s", strerror(errno)); 66 } 67 } 68 69 /* find_host -------------------------------------------------------------- */ 70 /* 71 Purpose: Find the entry in the status file for a given host 72 Returns: Pointer to that entry in the mmap() region, or NULL. 73 Notes: Also creates entries if requested. 74 Failure to create also returns NULL. 75 */ 76 77 HostInfo *find_host(char *hostname, int create) 78 { 79 HostInfo *hp; 80 HostInfo *spare_slot = NULL; 81 HostInfo *result = NULL; 82 struct addrinfo *ai1, *ai2; 83 int i; 84 85 if (getaddrinfo(hostname, NULL, NULL, &ai1) != 0) 86 ai1 = NULL; 87 for (i = 0, hp = status_info->hosts; i < status_info->noOfHosts; i++, hp++) 88 { 89 if (!strncasecmp(hostname, hp->hostname, SM_MAXSTRLEN)) 90 { 91 result = hp; 92 break; 93 } 94 if (hp->hostname[0] && 95 getaddrinfo(hp->hostname, NULL, NULL, &ai2) != 0) 96 ai2 = NULL; 97 if (ai1 && ai2) 98 { 99 struct addrinfo *p1, *p2; 100 for (p1 = ai1; !result && p1; p1 = p1->ai_next) 101 { 102 for (p2 = ai2; !result && p2; p2 = p2->ai_next) 103 { 104 if (p1->ai_family == p2->ai_family 105 && p1->ai_addrlen == p2->ai_addrlen 106 && !memcmp(p1->ai_addr, p2->ai_addr, p1->ai_addrlen)) 107 { 108 result = hp; 109 break; 110 } 111 } 112 } 113 if (result) 114 break; 115 } 116 if (ai2) 117 freeaddrinfo(ai2); 118 if (!spare_slot && !hp->monList && !hp->notifyReqd) 119 spare_slot = hp; 120 } 121 if (ai1) 122 freeaddrinfo(ai1); 123 124 /* Return if entry found, or if not asked to create one. */ 125 if (result || !create) return (result); 126 127 /* Now create an entry, using the spare slot if one was found or */ 128 /* adding to the end of the list otherwise, extending file if reqd */ 129 if (!spare_slot) 130 { 131 off_t desired_size; 132 spare_slot = &status_info->hosts[status_info->noOfHosts]; 133 desired_size = ((char*)spare_slot - (char*)status_info) + sizeof(HostInfo); 134 if (desired_size > status_file_len) 135 { 136 /* Extend file by writing 1 byte of junk at the desired end pos */ 137 lseek(status_fd, desired_size - 1, SEEK_SET); 138 i = write(status_fd, &i, 1); 139 if (i < 1) 140 { 141 syslog(LOG_ERR, "Unable to extend status file"); 142 return (NULL); 143 } 144 status_file_len = desired_size; 145 } 146 status_info->noOfHosts++; 147 } 148 149 /* Initialise the spare slot that has been found/created */ 150 /* Note that we do not msync(), since the caller is presumed to be */ 151 /* about to modify the entry further */ 152 memset(spare_slot, 0, sizeof(HostInfo)); 153 strncpy(spare_slot->hostname, hostname, SM_MAXSTRLEN); 154 return (spare_slot); 155 } 156 157 /* init_file -------------------------------------------------------------- */ 158 /* 159 Purpose: Open file, create if necessary, initialise it. 160 Returns: Nothing - exits on error 161 Notes: Called before process becomes daemon, hence logs to 162 stderr rather than syslog. 163 Opens the file, then mmap()s it for ease of access. 164 Also performs initial clean-up of the file, zeroing 165 monitor list pointers, setting the notifyReqd flag in 166 all hosts that had a monitor list, and incrementing 167 the state number to the next even value. 168 */ 169 170 void init_file(const char *filename) 171 { 172 int new_file = FALSE; 173 char buf[HEADER_LEN]; 174 int i; 175 176 /* try to open existing file - if not present, create one */ 177 status_fd = open(filename, O_RDWR); 178 if ((status_fd < 0) && (errno == ENOENT)) 179 { 180 status_fd = open(filename, O_RDWR | O_CREAT, 0644); 181 new_file = TRUE; 182 } 183 if (status_fd < 0) 184 errx(1, "unable to open status file %s", filename); 185 186 /* File now open. mmap() it, with a generous size to allow for */ 187 /* later growth, where we will extend the file but not re-map it. */ 188 status_info = (FileLayout *) 189 mmap(NULL, 0x10000000, PROT_READ | PROT_WRITE, MAP_SHARED, status_fd, 0); 190 191 if (status_info == (FileLayout *) MAP_FAILED) 192 err(1, "unable to mmap() status file"); 193 194 status_file_len = lseek(status_fd, 0L, SEEK_END); 195 196 /* If the file was not newly created, validate the contents, and if */ 197 /* defective, re-create from scratch. */ 198 if (!new_file) 199 { 200 if ((status_file_len < HEADER_LEN) || (status_file_len 201 < (HEADER_LEN + sizeof(HostInfo) * status_info->noOfHosts)) ) 202 { 203 warnx("status file is corrupt"); 204 new_file = TRUE; 205 } 206 } 207 208 /* Initialisation of a new, empty file. */ 209 if (new_file) 210 { 211 memset(buf, 0, sizeof(buf)); 212 lseek(status_fd, 0L, SEEK_SET); 213 write(status_fd, buf, HEADER_LEN); 214 status_file_len = HEADER_LEN; 215 } 216 else 217 { 218 /* Clean-up of existing file - monitored hosts will have a pointer */ 219 /* to a list of clients, which refers to memory in the previous */ 220 /* incarnation of the program and so are meaningless now. These */ 221 /* pointers are zeroed and the fact that the host was previously */ 222 /* monitored is recorded by setting the notifyReqd flag, which will */ 223 /* in due course cause a SM_NOTIFY to be sent. */ 224 /* Note that if we crash twice in quick succession, some hosts may */ 225 /* already have notifyReqd set, where we didn't manage to notify */ 226 /* them before the second crash occurred. */ 227 for (i = 0; i < status_info->noOfHosts; i++) 228 { 229 HostInfo *this_host = &status_info->hosts[i]; 230 231 if (this_host->monList) 232 { 233 this_host->notifyReqd = TRUE; 234 this_host->monList = NULL; 235 } 236 } 237 /* Select the next higher even number for the state counter */ 238 status_info->ourState = (status_info->ourState + 2) & 0xfffffffe; 239 /*???????******/ status_info->ourState++; 240 } 241 } 242 243 /* notify_one_host --------------------------------------------------------- */ 244 /* 245 Purpose: Perform SM_NOTIFY procedure at specified host 246 Returns: TRUE if success, FALSE if failed. 247 */ 248 249 static int notify_one_host(char *hostname) 250 { 251 struct timeval timeout = { 20, 0 }; /* 20 secs timeout */ 252 CLIENT *cli; 253 char dummy; 254 stat_chge arg; 255 char our_hostname[SM_MAXSTRLEN+1]; 256 257 gethostname(our_hostname, sizeof(our_hostname)); 258 our_hostname[SM_MAXSTRLEN] = '\0'; 259 arg.mon_name = our_hostname; 260 arg.state = status_info->ourState; 261 262 if (debug) syslog (LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s", hostname, our_hostname); 263 264 cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp"); 265 if (!cli) 266 { 267 syslog(LOG_ERR, "Failed to contact host %s%s", hostname, 268 clnt_spcreateerror("")); 269 return (FALSE); 270 } 271 272 if (clnt_call(cli, SM_NOTIFY, (xdrproc_t)xdr_stat_chge, &arg, 273 (xdrproc_t)xdr_void, &dummy, timeout) 274 != RPC_SUCCESS) 275 { 276 syslog(LOG_ERR, "Failed to contact rpc.statd at host %s", hostname); 277 clnt_destroy(cli); 278 return (FALSE); 279 } 280 281 clnt_destroy(cli); 282 return (TRUE); 283 } 284 285 /* notify_hosts ------------------------------------------------------------ */ 286 /* 287 Purpose: Send SM_NOTIFY to all hosts marked as requiring it 288 Returns: Nothing, immediately - forks a process to do the work. 289 Notes: Does nothing if there are no monitored hosts. 290 Called after all the initialisation has been done - 291 logs to syslog. 292 */ 293 294 void notify_hosts(void) 295 { 296 int i; 297 int attempts; 298 int work_to_do = FALSE; 299 HostInfo *hp; 300 pid_t pid; 301 302 /* First check if there is in fact any work to do. */ 303 for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++) 304 { 305 if (hp->notifyReqd) 306 { 307 work_to_do = TRUE; 308 break; 309 } 310 } 311 312 if (!work_to_do) return; /* No work found */ 313 314 pid = fork(); 315 if (pid == -1) 316 { 317 syslog(LOG_ERR, "Unable to fork notify process - %s", strerror(errno)); 318 return; 319 } 320 if (pid) return; 321 322 /* Here in the child process. We continue until all the hosts marked */ 323 /* as requiring notification have been duly notified. */ 324 /* If one of the initial attempts fails, we sleep for a while and */ 325 /* have another go. This is necessary because when we have crashed, */ 326 /* (eg. a power outage) it is quite possible that we won't be able to */ 327 /* contact all monitored hosts immediately on restart, either because */ 328 /* they crashed too and take longer to come up (in which case the */ 329 /* notification isn't really required), or more importantly if some */ 330 /* router etc. needed to reach the monitored host has not come back */ 331 /* up yet. In this case, we will be a bit late in re-establishing */ 332 /* locks (after the grace period) but that is the best we can do. */ 333 /* We try 10 times at 5 sec intervals, 10 more times at 1 minute */ 334 /* intervals, then 24 more times at hourly intervals, finally */ 335 /* giving up altogether if the host hasn't come back to life after */ 336 /* 24 hours. */ 337 338 for (attempts = 0; attempts < 44; attempts++) 339 { 340 work_to_do = FALSE; /* Unless anything fails */ 341 for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++) 342 { 343 if (hp->notifyReqd) 344 { 345 if (notify_one_host(hp->hostname)) 346 { 347 hp->notifyReqd = FALSE; 348 sync_file(); 349 } 350 else work_to_do = TRUE; 351 } 352 } 353 if (!work_to_do) break; 354 if (attempts < 10) sleep(5); 355 else if (attempts < 20) sleep(60); 356 else sleep(60*60); 357 } 358 exit(0); 359 } 360 361 362