1 /* $NetBSD: statd.c,v 1.32 2018/01/23 21:06:26 sevan Exp $ */ 2 3 /* 4 * Copyright (c) 1995 5 * A.R. Gordon (andrew.gordon@net-tel.co.uk). All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed for the FreeBSD project 18 * This product includes software developed by Christos Zoulas. 19 * 4. Neither the name of the author nor the names of any co-contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY ANDREW GORDON AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 */ 36 37 #include <sys/cdefs.h> 38 #ifndef lint 39 __RCSID("$NetBSD: statd.c,v 1.32 2018/01/23 21:06:26 sevan Exp $"); 40 #endif 41 42 /* main() function for status monitor daemon. Some of the code in this */ 43 /* file was generated by running rpcgen /usr/include/rpcsvc/sm_inter.x */ 44 /* The actual program logic is in the file procs.c */ 45 46 #include <sys/param.h> 47 #include <sys/wait.h> 48 49 #include <err.h> 50 #include <ctype.h> 51 #include <errno.h> 52 #include <fcntl.h> 53 #include <signal.h> 54 #include <stdio.h> 55 #include <stdlib.h> 56 #include <string.h> 57 #include <syslog.h> 58 #include <unistd.h> 59 #include <util.h> 60 #include <db.h> 61 #include <netconfig.h> 62 63 #include <rpc/rpc.h> 64 65 #include "statd.h" 66 67 struct sigaction sa; 68 int debug = 0; /* Controls syslog() for debug msgs */ 69 int _rpcsvcdirty = 0; /* XXX ??? */ 70 static DB *db; /* Database file */ 71 72 Header status_info; 73 74 static char undefdata[] = "\0\1\2\3\4\5\6\7"; 75 static DBT undefkey = { 76 undefdata, 77 sizeof(undefdata) 78 }; 79 80 81 /* statd.c */ 82 static int walk_one(int (*fun )(DBT *, HostInfo *, void *), DBT *, DBT *, void *); 83 static int walk_db(int (*fun )(DBT *, HostInfo *, void *), void *); 84 static int reset_host(DBT *, HostInfo *, void *); 85 static int check_work(DBT *, HostInfo *, void *); 86 static int unmon_host(DBT *, HostInfo *, void *); 87 static int notify_one(DBT *, HostInfo *, void *); 88 static void init_file(const char *); 89 static int notify_one_host(const char *); 90 static void die(int) __dead; 91 92 int 93 main(int argc, char *argv[]) 94 { 95 int ch; 96 struct sigaction nsa; 97 int maxrec = RPC_MAXDATASIZE; 98 99 while ((ch = getopt(argc, argv, "d")) != (-1)) { 100 switch (ch) { 101 case 'd': 102 debug = 1; 103 break; 104 default: 105 case '?': 106 (void)fprintf(stderr, "usage: %s [-d]\n", 107 getprogname()); 108 exit(1); 109 /* NOTREACHED */ 110 } 111 } 112 (void)rpcb_unset(SM_PROG, SM_VERS, NULL); 113 114 rpc_control(RPC_SVC_CONNMAXREC_SET, &maxrec); 115 116 if (!svc_create(sm_prog_1, SM_PROG, SM_VERS, "udp")) { 117 errx(EXIT_FAILURE, "cannot create udp service."); 118 /* NOTREACHED */ 119 } 120 if (!svc_create(sm_prog_1, SM_PROG, SM_VERS, "tcp")) { 121 errx(EXIT_FAILURE, "cannot create udp service."); 122 /* NOTREACHED */ 123 } 124 125 init_file("/var/db/statd.status"); 126 127 /* 128 * Note that it is NOT sensible to run this program from inetd - the 129 * protocol assumes that it will run immediately at boot time. 130 */ 131 if (!debug) 132 daemon(0, 0); 133 134 sigemptyset(&nsa.sa_mask); 135 nsa.sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT; 136 nsa.sa_handler = SIG_IGN; 137 (void)sigaction(SIGCHLD, &nsa, NULL); 138 139 pidfile(NULL); 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 (void)sigaction(SIGTERM, &sa, NULL); 150 (void)sigaction(SIGQUIT, &sa, NULL); 151 (void)sigaction(SIGHUP, &sa, NULL); 152 (void)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 (void)sigaction(SIGALRM, &sa, NULL); 187 188 now = time(NULL); 189 190 (void) 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[MAXHOSTNAMELEN + 1]; 244 HostInfo h; 245 246 strncpy(hostname, hostnamep, MAXHOSTNAMELEN + 1); 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 } 352 353 354 for (;;) 355 switch ((*db->seq)(db, &key, &data, R_NEXT)) { 356 case -1: 357 goto bad; 358 case 0: 359 if (walk_one(fun, &key, &data, ptr) == -1) 360 return -1; 361 break; 362 case 1: 363 return 0; 364 default: 365 abort(); 366 } 367 bad: 368 syslog(LOG_ERR, "Corrupted database %m"); 369 die(1); 370 } 371 372 /* reset_host ------------------------------------------------------------ */ 373 /* 374 * Purpose: Clean up existing hosts in file. 375 * Returns: Always success 0. 376 * Notes: Clean-up of existing file - monitored hosts will have a 377 * pointer to a list of clients, which refers to memory in 378 * the previous incarnation of the program and so are 379 * meaningless now. These pointers are zeroed and the fact 380 * that the host was previously monitored is recorded by 381 * setting the notifyReqd flag, which will in due course 382 * cause a SM_NOTIFY to be sent. 383 * 384 * Note that if we crash twice in quick succession, some hosts 385 * may already have notifyReqd set, where we didn't manage to 386 * notify them before the second crash occurred. 387 */ 388 static int 389 reset_host(DBT *key, HostInfo *hi, void *ptr) 390 { 391 392 if (hi->monList) { 393 hi->notifyReqd = *(time_t *) ptr; 394 hi->attempts = 0; 395 hi->monList = NULL; 396 change_host((char *)key->data, hi); 397 } 398 return 0; 399 } 400 401 /* check_work ------------------------------------------------------------ */ 402 /* 403 * Purpose: Check if there is work to be done. 404 * Returns: 0 if there is no work to be done -1 if there is. 405 * Notes: 406 */ 407 static int 408 check_work(DBT *key, HostInfo *hi, void *ptr) 409 { 410 return hi->notifyReqd ? -1 : 0; 411 } 412 413 /* unmon_host ------------------------------------------------------------ */ 414 /* 415 * Purpose: Unmonitor a host 416 * Returns: 0 417 * Notes: 418 */ 419 static int 420 unmon_host(DBT *key, HostInfo *hi, void *ptr) 421 { 422 char *name = key->data; 423 424 if (do_unmon(name, hi, ptr)) 425 change_host(name, hi); 426 return 0; 427 } 428 429 /* notify_one ------------------------------------------------------------ */ 430 /* 431 * Purpose: Notify one host. 432 * Returns: 0 if success -1 on failure 433 * Notes: 434 */ 435 static int 436 notify_one(DBT *key, HostInfo *hi, void *ptr) 437 { 438 time_t now = *(time_t *) ptr; 439 char *name = key->data; 440 int error; 441 442 if (hi->notifyReqd == 0 || hi->notifyReqd > now) 443 return 0; 444 445 /* 446 * If one of the initial attempts fails, we wait 447 * for a while and have another go. This is necessary 448 * because when we have crashed, (eg. a power outage) 449 * it is quite possible that we won't be able to 450 * contact all monitored hosts immediately on restart, 451 * either because they crashed too and take longer 452 * to come up (in which case the notification isn't 453 * really required), or more importantly if some 454 * router etc. needed to reach the monitored host 455 * has not come back up yet. In this case, we will 456 * be a bit late in re-establishing locks (after the 457 * grace period) but that is the best we can do. We 458 * try 10 times at 5 sec intervals, 10 more times at 459 * 1 minute intervals, then 24 more times at hourly 460 * intervals, finally giving up altogether if the 461 * host hasn't come back to life after 24 hours. 462 */ 463 if (notify_one_host(name) || hi->attempts++ >= 44) { 464 error = 0; 465 hi->notifyReqd = 0; 466 hi->attempts = 0; 467 } else { 468 error = -1; 469 if (hi->attempts < 10) 470 hi->notifyReqd += 5; 471 else if (hi->attempts < 20) 472 hi->notifyReqd += 60; 473 else 474 hi->notifyReqd += 60 * 60; 475 } 476 change_host(name, hi); 477 return error; 478 } 479 480 /* init_file -------------------------------------------------------------- */ 481 /* 482 * Purpose: Open file, create if necessary, initialise it. 483 * Returns: Nothing - exits on error 484 * Notes: Called before process becomes daemon, hence logs to 485 * stderr rather than syslog. 486 * Opens the file, then mmap()s it for ease of access. 487 * Also performs initial clean-up of the file, zeroing 488 * monitor list pointers, setting the notifyReqd flag in 489 * all hosts that had a monitor list, and incrementing 490 * the state number to the next even value. 491 */ 492 static void 493 init_file(const char *filename) 494 { 495 DBT data; 496 497 db = dbopen(filename, O_RDWR|O_CREAT|O_NDELAY|O_EXLOCK, 0644, DB_HASH, 498 NULL); 499 if (db == NULL) 500 err(EXIT_FAILURE, "Cannot open `%s'", filename); 501 502 switch ((*db->get)(db, &undefkey, &data, 0)) { 503 case 1: 504 /* New database */ 505 (void)memset(&status_info, 0, sizeof(status_info)); 506 sync_file(); 507 return; 508 509 case -1: 510 err(EXIT_FAILURE, "error accessing database (%s)", strerror(errno)); 511 case 0: 512 /* Existing database */ 513 if (data.size != sizeof(status_info)) 514 errx(EXIT_FAILURE, "database corrupted %lu != %lu", 515 (u_long)data.size, (u_long)sizeof(status_info)); 516 memcpy(&status_info, data.data, data.size); 517 break; 518 default: 519 abort(); 520 } 521 522 reset_database(); 523 return; 524 } 525 526 /* reset_database --------------------------------------------------------- */ 527 /* 528 * Purpose: Clears the statd database 529 * Returns: Nothing 530 * Notes: If this is not called on reset, it will leak memory. 531 */ 532 void 533 reset_database() 534 { 535 time_t now = time(NULL); 536 walk_db(reset_host, &now); 537 538 /* Select the next higher even number for the state counter */ 539 status_info.ourState = 540 (status_info.ourState + 2) & 0xfffffffe; 541 status_info.ourState++; /* XXX - ??? */ 542 sync_file(); 543 } 544 545 /* unmon_hosts --------------------------------------------------------- */ 546 /* 547 * Purpose: Unmonitor all the hosts 548 * Returns: Nothing 549 * Notes: 550 */ 551 void 552 unmon_hosts() 553 { 554 time_t now = time(NULL); 555 walk_db(unmon_host, &now); 556 sync_file(); 557 } 558 559 static int 560 notify_one_host(const char *hostname) 561 { 562 struct timeval timeout = {20, 0}; /* 20 secs timeout */ 563 CLIENT *cli; 564 char dummy; 565 stat_chge arg; 566 char our_hostname[MAXHOSTNAMELEN + 1]; 567 568 gethostname(our_hostname, sizeof(our_hostname)); 569 our_hostname[sizeof(our_hostname) - 1] = '\0'; 570 arg.mon_name = our_hostname; 571 arg.state = status_info.ourState; 572 573 if (debug) 574 syslog(LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s", 575 hostname, our_hostname); 576 577 cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp"); 578 if (!cli) { 579 syslog(LOG_ERR, "Failed to contact host %s%s", hostname, 580 clnt_spcreateerror("")); 581 return (FALSE); 582 } 583 if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void, 584 &dummy, timeout) != RPC_SUCCESS) { 585 syslog(LOG_ERR, "Failed to contact rpc.statd at host %s", 586 hostname); 587 clnt_destroy(cli); 588 return (FALSE); 589 } 590 clnt_destroy(cli); 591 return (TRUE); 592 } 593 594 595 static void 596 die(int n) 597 { 598 (*db->close)(db); 599 exit(n); 600 } 601