savecore.c revision 1.24
1/* $NetBSD: savecore.c,v 1.24 1996/03/08 08:52:12 mycroft Exp $ */ 2 3/*- 4 * Copyright (c) 1986, 1992, 1993 5 * The Regents of the University of California. 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 by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its 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 THE REGENTS 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 REGENTS 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#ifndef lint 37static char copyright[] = 38"@(#) Copyright (c) 1986, 1992, 1993\n\ 39 The Regents of the University of California. All rights reserved.\n"; 40#endif /* not lint */ 41 42#ifndef lint 43#if 0 44static char sccsid[] = "@(#)savecore.c 8.3 (Berkeley) 1/2/94"; 45#else 46static char rcsid[] = "$NetBSD: savecore.c,v 1.24 1996/03/08 08:52:12 mycroft Exp $"; 47#endif 48#endif /* not lint */ 49 50#include <sys/param.h> 51#include <sys/stat.h> 52#include <sys/mount.h> 53#include <sys/syslog.h> 54#include <sys/time.h> 55 56#include <dirent.h> 57#include <errno.h> 58#include <fcntl.h> 59#include <nlist.h> 60#include <paths.h> 61#include <stdio.h> 62#include <stdlib.h> 63#include <string.h> 64#include <tzfile.h> 65#include <unistd.h> 66 67extern FILE *zopen __P((const char *fname, const char *mode, int bits)); 68 69#define ok(number) ((number) - KERNBASE) 70 71struct nlist current_nl[] = { /* Namelist for currently running system. */ 72#define X_DUMPDEV 0 73 { "_dumpdev" }, 74#define X_DUMPLO 1 75 { "_dumplo" }, 76#define X_TIME 2 77 { "_time" }, 78#define X_DUMPSIZE 3 79 { "_dumpsize" }, 80#define X_VERSION 4 81 { "_version" }, 82#define X_PANICSTR 5 83 { "_panicstr" }, 84#define X_DUMPMAG 6 85 { "_dumpmag" }, 86 { "" }, 87}; 88int cursyms[] = { X_DUMPDEV, X_DUMPLO, X_VERSION, X_DUMPMAG, -1 }; 89int dumpsyms[] = { X_TIME, X_DUMPSIZE, X_VERSION, X_PANICSTR, X_DUMPMAG, -1 }; 90 91struct nlist dump_nl[] = { /* Name list for dumped system. */ 92 { "_dumpdev" }, /* Entries MUST be the same as */ 93 { "_dumplo" }, /* those in current_nl[]. */ 94 { "_time" }, 95 { "_dumpsize" }, 96 { "_version" }, 97 { "_panicstr" }, 98 { "_dumpmag" }, 99 { "" }, 100}; 101 102/* Types match kernel declarations. */ 103long dumplo; /* where dump starts on dumpdev */ 104int dumpmag; /* magic number in dump */ 105int dumpsize; /* amount of memory dumped */ 106 107char *kernel; 108char *dirname; /* directory to save dumps in */ 109char *ddname; /* name of dump device */ 110dev_t dumpdev; /* dump device */ 111int dumpfd; /* read/write descriptor on block dev */ 112time_t now; /* current date */ 113char panic_mesg[1024]; 114int panicstr; 115char vers[1024]; 116 117int clear, compress, force, verbose; /* flags */ 118 119void check_kmem __P((void)); 120int check_space __P((void)); 121void clear_dump __P((void)); 122int Create __P((char *, int)); 123int dump_exists __P((void)); 124char *find_dev __P((dev_t, int)); 125int get_crashtime __P((void)); 126void kmem_setup __P((void)); 127void log __P((int, char *, ...)); 128void Lseek __P((int, off_t, int)); 129int Open __P((char *, int rw)); 130int Read __P((int, void *, int)); 131char *rawname __P((char *s)); 132void save_core __P((void)); 133void usage __P((void)); 134void Write __P((int, void *, int)); 135 136int 137main(argc, argv) 138 int argc; 139 char *argv[]; 140{ 141 int ch; 142 143 openlog("savecore", LOG_PERROR, LOG_DAEMON); 144 145 while ((ch = getopt(argc, argv, "cdfN:vz")) != -1) 146 switch(ch) { 147 case 'c': 148 clear = 1; 149 break; 150 case 'd': /* Not documented. */ 151 case 'v': 152 verbose = 1; 153 break; 154 case 'f': 155 force = 1; 156 break; 157 case 'N': 158 kernel = optarg; 159 break; 160 case 'z': 161 compress = 1; 162 break; 163 case '?': 164 default: 165 usage(); 166 } 167 argc -= optind; 168 argv += optind; 169 170 if (!clear) { 171 if (argc != 1 && argc != 2) 172 usage(); 173 dirname = argv[0]; 174 } 175 if (argc == 2) 176 kernel = argv[1]; 177 178 (void)time(&now); 179 kmem_setup(); 180 181 if (clear) { 182 clear_dump(); 183 exit(0); 184 } 185 186 if (!dump_exists() && !force) 187 exit(1); 188 189 check_kmem(); 190 191 if (panicstr) 192 syslog(LOG_ALERT, "reboot after panic: %s", panic_mesg); 193 else 194 syslog(LOG_ALERT, "reboot"); 195 196 if ((!get_crashtime() || !check_space()) && !force) 197 exit(1); 198 199 save_core(); 200 201 clear_dump(); 202 exit(0); 203} 204 205void 206kmem_setup() 207{ 208 FILE *fp; 209 int kmem, i; 210 char *dump_sys; 211 212 /* 213 * Some names we need for the currently running system, others for 214 * the system that was running when the dump was made. The values 215 * obtained from the current system are used to look for things in 216 * /dev/kmem that cannot be found in the dump_sys namelist, but are 217 * presumed to be the same (since the disk partitions are probably 218 * the same!) 219 */ 220 if ((nlist(_PATH_UNIX, current_nl)) == -1) 221 syslog(LOG_ERR, "%s: nlist: %s", _PATH_UNIX, strerror(errno)); 222 for (i = 0; cursyms[i] != -1; i++) 223 if (current_nl[cursyms[i]].n_value == 0) { 224 syslog(LOG_ERR, "%s: %s not in namelist", 225 _PATH_UNIX, current_nl[cursyms[i]].n_name); 226 exit(1); 227 } 228 229 dump_sys = kernel ? kernel : _PATH_UNIX; 230 if ((nlist(dump_sys, dump_nl)) == -1) 231 syslog(LOG_ERR, "%s: nlist: %s", dump_sys, strerror(errno)); 232 for (i = 0; dumpsyms[i] != -1; i++) 233 if (dump_nl[dumpsyms[i]].n_value == 0) { 234 syslog(LOG_ERR, "%s: %s not in namelist", 235 dump_sys, dump_nl[dumpsyms[i]].n_name); 236 exit(1); 237 } 238 239 kmem = Open(_PATH_KMEM, O_RDONLY); 240 Lseek(kmem, (off_t)current_nl[X_DUMPDEV].n_value, L_SET); 241 (void)Read(kmem, &dumpdev, sizeof(dumpdev)); 242 if (dumpdev == NODEV) { 243 syslog(LOG_WARNING, "no core dump (no dumpdev)"); 244 exit(1); 245 } 246 Lseek(kmem, (off_t)current_nl[X_DUMPLO].n_value, L_SET); 247 (void)Read(kmem, &dumplo, sizeof(dumplo)); 248 dumplo *= DEV_BSIZE; 249 if (verbose) 250 (void)printf("dumplo = %d (%d * %d)\n", 251 dumplo, dumplo / DEV_BSIZE, DEV_BSIZE); 252 Lseek(kmem, (off_t)current_nl[X_DUMPMAG].n_value, L_SET); 253 (void)Read(kmem, &dumpmag, sizeof(dumpmag)); 254 ddname = find_dev(dumpdev, S_IFBLK); 255 dumpfd = Open(ddname, O_RDWR); 256 fp = fdopen(kmem, "r"); 257 if (fp == NULL) { 258 syslog(LOG_ERR, "%s: fdopen: %m", _PATH_KMEM); 259 exit(1); 260 } 261 if (kernel) 262 return; 263 (void)fseek(fp, (off_t)current_nl[X_VERSION].n_value, L_SET); 264 (void)fgets(vers, sizeof(vers), fp); 265 266 /* Don't fclose(fp), we use dumpfd later. */ 267} 268 269void 270check_kmem() 271{ 272 register char *cp; 273 FILE *fp; 274 char core_vers[1024]; 275 276 fp = fdopen(dumpfd, "r"); 277 if (fp == NULL) { 278 syslog(LOG_ERR, "%s: fdopen: %m", ddname); 279 exit(1); 280 } 281 fseek(fp, (off_t)(dumplo + ok(dump_nl[X_VERSION].n_value)), L_SET); 282 fgets(core_vers, sizeof(core_vers), fp); 283 if (strcmp(vers, core_vers) && kernel == 0) 284 syslog(LOG_WARNING, 285 "warning: %s version mismatch:\n\t%s\nand\t%s\n", 286 _PATH_UNIX, vers, core_vers); 287 (void)fseek(fp, 288 (off_t)(dumplo + ok(dump_nl[X_PANICSTR].n_value)), L_SET); 289 (void)fread(&panicstr, sizeof(panicstr), 1, fp); 290 if (panicstr) { 291 (void)fseek(fp, dumplo + ok(panicstr), L_SET); 292 cp = panic_mesg; 293 do 294 *cp = getc(fp); 295 while (*cp++ && cp < &panic_mesg[sizeof(panic_mesg)]); 296 } 297 /* Don't fclose(fp), we use dumpfd later. */ 298} 299 300void 301clear_dump() 302{ 303 long newdumplo; 304 305 newdumplo = 0; 306 Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPMAG].n_value)), L_SET); 307 Write(dumpfd, &newdumplo, sizeof(newdumplo)); 308} 309 310int 311dump_exists() 312{ 313 int newdumpmag; 314 315 Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPMAG].n_value)), L_SET); 316 (void)Read(dumpfd, &newdumpmag, sizeof(newdumpmag)); 317 318 /* Read the dump size. */ 319 Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPSIZE].n_value)), L_SET); 320 (void)Read(dumpfd, &dumpsize, sizeof(dumpsize)); 321 dumpsize *= getpagesize(); 322 323 /* 324 * Return zero if core dump doesn't seem to be there, and note 325 * it for syslog. This check and return happens after the dump size 326 * is read, so dumpsize is whether or not the core is valid (for -f). 327 */ 328 if (newdumpmag != dumpmag) { 329 if (verbose) 330 syslog(LOG_WARNING, "magic number mismatch (%x != %x)", 331 newdumpmag, dumpmag); 332 syslog(LOG_WARNING, "no core dump"); 333 return (0); 334 } 335 return (1); 336} 337 338char buf[1024 * 1024]; 339 340void 341save_core() 342{ 343 register FILE *fp; 344 register int bounds, ifd, nr, nw, ofd; 345 char *rawp, path[MAXPATHLEN]; 346 347 /* 348 * Get the current number and update the bounds file. Do the update 349 * now, because may fail later and don't want to overwrite anything. 350 */ 351 (void)snprintf(path, sizeof(path), "%s/bounds", dirname); 352 if ((fp = fopen(path, "r")) == NULL) 353 goto err1; 354 if (fgets(buf, sizeof(buf), fp) == NULL) { 355 if (ferror(fp)) 356err1: syslog(LOG_WARNING, "%s: %s", path, strerror(errno)); 357 bounds = 0; 358 } else 359 bounds = atoi(buf); 360 if (fp != NULL) 361 (void)fclose(fp); 362 if ((fp = fopen(path, "w")) == NULL) 363 syslog(LOG_ERR, "%s: %m", path); 364 else { 365 (void)fprintf(fp, "%d\n", bounds + 1); 366 (void)fclose(fp); 367 } 368 (void)fclose(fp); 369 370 /* Create the core file. */ 371 (void)snprintf(path, sizeof(path), "%s/netbsd.%d.core%s", 372 dirname, bounds, compress ? ".Z" : ""); 373 if (compress) { 374 if ((fp = zopen(path, "w", 0)) == NULL) { 375 syslog(LOG_ERR, "%s: %s", path, strerror(errno)); 376 exit(1); 377 } 378 } else 379 ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH); 380 381 /* Open the raw device. */ 382 rawp = rawname(ddname); 383 if ((ifd = open(rawp, O_RDONLY)) == -1) { 384 syslog(LOG_WARNING, "%s: %m; using block device", rawp); 385 ifd = dumpfd; 386 } 387 388 /* Seek to the start of the core. */ 389 Lseek(ifd, (off_t)dumplo, L_SET); 390 391 /* Copy the core file. */ 392 syslog(LOG_NOTICE, "writing %score to %s", 393 compress ? "compressed " : "", path); 394 for (; dumpsize > 0; dumpsize -= nr) { 395 (void)printf("%6dK\r", dumpsize / 1024); 396 (void)fflush(stdout); 397 nr = read(ifd, buf, MIN(dumpsize, sizeof(buf))); 398 if (nr <= 0) { 399 if (nr == 0) 400 syslog(LOG_WARNING, 401 "WARNING: EOF on dump device"); 402 else 403 syslog(LOG_ERR, "%s: %m", rawp); 404 goto err2; 405 } 406 if (compress) 407 nw = fwrite(buf, 1, nr, fp); 408 else 409 nw = write(ofd, buf, nr); 410 if (nw != nr) { 411 syslog(LOG_ERR, "%s: %s", 412 path, strerror(nw == 0 ? EIO : errno)); 413err2: syslog(LOG_WARNING, 414 "WARNING: core may be incomplete"); 415 (void)printf("\n"); 416 exit(1); 417 } 418 } 419 (void)close(ifd); 420 if (compress) 421 (void)fclose(fp); 422 else 423 (void)close(ofd); 424 425 /* Copy the kernel. */ 426 ifd = Open(kernel ? kernel : _PATH_UNIX, O_RDONLY); 427 (void)snprintf(path, sizeof(path), "%s/netbsd.%d%s", 428 dirname, bounds, compress ? ".Z" : ""); 429 if (compress) { 430 if ((fp = zopen(path, "w", 0)) == NULL) { 431 syslog(LOG_ERR, "%s: %s", path, strerror(errno)); 432 exit(1); 433 } 434 } else 435 ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH); 436 syslog(LOG_NOTICE, "writing %skernel to %s", 437 compress ? "compressed " : "", path); 438 while ((nr = read(ifd, buf, sizeof(buf))) > 0) { 439 if (compress) 440 nw = fwrite(buf, 1, nr, fp); 441 else 442 nw = write(ofd, buf, nr); 443 if (nw != nr) { 444 syslog(LOG_ERR, "%s: %s", 445 path, strerror(nw == 0 ? EIO : errno)); 446 syslog(LOG_WARNING, 447 "WARNING: kernel may be incomplete"); 448 exit(1); 449 } 450 } 451 if (nr < 0) { 452 syslog(LOG_ERR, "%s: %s", 453 kernel ? kernel : _PATH_UNIX, strerror(errno)); 454 syslog(LOG_WARNING, 455 "WARNING: kernel may be incomplete"); 456 exit(1); 457 } 458 if (compress) 459 (void)fclose(fp); 460 else 461 (void)close(ofd); 462} 463 464char * 465find_dev(dev, type) 466 register dev_t dev; 467 register int type; 468{ 469 register DIR *dfd; 470 struct dirent *dir; 471 struct stat sb; 472 char *dp, devname[MAXPATHLEN + 1]; 473 474 if ((dfd = opendir(_PATH_DEV)) == NULL) { 475 syslog(LOG_ERR, "%s: %s", _PATH_DEV, strerror(errno)); 476 exit(1); 477 } 478 (void)strcpy(devname, _PATH_DEV); 479 while ((dir = readdir(dfd))) { 480 (void)strcpy(devname + sizeof(_PATH_DEV) - 1, dir->d_name); 481 if (lstat(devname, &sb)) { 482 syslog(LOG_ERR, "%s: %s", devname, strerror(errno)); 483 continue; 484 } 485 if ((sb.st_mode & S_IFMT) != type) 486 continue; 487 if (dev == sb.st_rdev) { 488 closedir(dfd); 489 if ((dp = strdup(devname)) == NULL) { 490 syslog(LOG_ERR, "%s", strerror(errno)); 491 exit(1); 492 } 493 return (dp); 494 } 495 } 496 closedir(dfd); 497 syslog(LOG_ERR, "can't find device %d/%d", major(dev), minor(dev)); 498 exit(1); 499} 500 501char * 502rawname(s) 503 char *s; 504{ 505 char *sl, name[MAXPATHLEN]; 506 507 if ((sl = strrchr(s, '/')) == NULL || sl[1] == '0') { 508 syslog(LOG_ERR, 509 "can't make raw dump device name from %s", s); 510 return (s); 511 } 512 (void)snprintf(name, sizeof(name), "%.*s/r%s", sl - s, s, sl + 1); 513 if ((sl = strdup(name)) == NULL) { 514 syslog(LOG_ERR, "%s", strerror(errno)); 515 exit(1); 516 } 517 return (sl); 518} 519 520int 521get_crashtime() 522{ 523 time_t dumptime; /* Time the dump was taken. */ 524 525 Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_TIME].n_value)), L_SET); 526 (void)Read(dumpfd, &dumptime, sizeof(dumptime)); 527 if (dumptime == 0) { 528 if (verbose) 529 syslog(LOG_ERR, "dump time is zero"); 530 return (0); 531 } 532 (void)printf("savecore: system went down at %s", ctime(&dumptime)); 533#define LEEWAY (7 * SECSPERDAY) 534 if (dumptime < now - LEEWAY || dumptime > now + LEEWAY) { 535 (void)printf("dump time is unreasonable\n"); 536 return (0); 537 } 538 return (1); 539} 540 541int 542check_space() 543{ 544 register FILE *fp; 545 char *tkernel; 546 off_t minfree, spacefree, kernelsize, needed; 547 struct stat st; 548 struct statfs fsbuf; 549 char buf[100], path[MAXPATHLEN]; 550 551 tkernel = kernel ? kernel : _PATH_UNIX; 552 if (stat(tkernel, &st) < 0) { 553 syslog(LOG_ERR, "%s: %m", tkernel); 554 exit(1); 555 } 556 kernelsize = st.st_blocks * S_BLKSIZE; 557 if (statfs(dirname, &fsbuf) < 0) { 558 syslog(LOG_ERR, "%s: %m", dirname); 559 exit(1); 560 } 561 spacefree = (fsbuf.f_bavail * fsbuf.f_bsize) / 1024; 562 563 (void)snprintf(path, sizeof(path), "%s/minfree", dirname); 564 if ((fp = fopen(path, "r")) == NULL) 565 minfree = 0; 566 else { 567 if (fgets(buf, sizeof(buf), fp) == NULL) 568 minfree = 0; 569 else 570 minfree = atoi(buf); 571 (void)fclose(fp); 572 } 573 574 needed = (dumpsize + kernelsize) / 1024; 575 if (minfree > 0 && spacefree - needed < minfree) { 576 syslog(LOG_WARNING, 577 "no dump, not enough free space on device"); 578 return (0); 579 } 580 if (spacefree - needed < minfree) 581 syslog(LOG_WARNING, 582 "dump performed, but free space threshold crossed"); 583 return (1); 584} 585 586int 587Open(name, rw) 588 char *name; 589 int rw; 590{ 591 int fd; 592 593 if ((fd = open(name, rw, 0)) < 0) { 594 syslog(LOG_ERR, "%s: %m", name); 595 exit(1); 596 } 597 return (fd); 598} 599 600int 601Read(fd, bp, size) 602 int fd, size; 603 void *bp; 604{ 605 int nr; 606 607 nr = read(fd, bp, size); 608 if (nr != size) { 609 syslog(LOG_ERR, "read: %m"); 610 exit(1); 611 } 612 return (nr); 613} 614 615void 616Lseek(fd, off, flag) 617 int fd, flag; 618 off_t off; 619{ 620 off_t ret; 621 622 ret = lseek(fd, off, flag); 623 if (ret == -1) { 624 syslog(LOG_ERR, "lseek: %m"); 625 exit(1); 626 } 627} 628 629int 630Create(file, mode) 631 char *file; 632 int mode; 633{ 634 register int fd; 635 636 fd = open(file, O_WRONLY | O_CREAT | O_TRUNC, mode); 637 if (fd < 0) { 638 syslog(LOG_ERR, "%s: %m", file); 639 exit(1); 640 } 641 return (fd); 642} 643 644void 645Write(fd, bp, size) 646 int fd, size; 647 void *bp; 648{ 649 int n; 650 651 if ((n = write(fd, bp, size)) < size) { 652 syslog(LOG_ERR, "write: %s", strerror(n == -1 ? errno : EIO)); 653 exit(1); 654 } 655} 656 657void 658usage() 659{ 660 (void)syslog(LOG_ERR, "usage: savecore [-cfvz] [-N system] directory"); 661 exit(1); 662} 663