savecore.c revision 1.21 1 /* $NetBSD: savecore.c,v 1.21 1995/03/18 15:01:02 cgd 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
37 static 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
44 static char sccsid[] = "@(#)savecore.c 8.3 (Berkeley) 1/2/94";
45 #else
46 static char rcsid[] = "$NetBSD: savecore.c,v 1.21 1995/03/18 15:01:02 cgd 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
67 extern FILE *zopen __P((const char *fname, const char *mode, int bits));
68
69 #define ok(number) ((number) - KERNBASE)
70
71 struct 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 };
88 int cursyms[] = { X_DUMPDEV, X_DUMPLO, X_VERSION, X_DUMPMAG, -1 };
89 int dumpsyms[] = { X_TIME, X_DUMPSIZE, X_VERSION, X_PANICSTR, X_DUMPMAG, -1 };
90
91 struct 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. */
103 long dumplo; /* where dump starts on dumpdev */
104 int dumpmag; /* magic number in dump */
105 int dumpsize; /* amount of memory dumped */
106
107 char *kernel;
108 char *dirname; /* directory to save dumps in */
109 char *ddname; /* name of dump device */
110 dev_t dumpdev; /* dump device */
111 int dumpfd; /* read/write descriptor on block dev */
112 time_t now; /* current date */
113 char panic_mesg[1024];
114 int panicstr;
115 char vers[1024];
116
117 int clear, compress, force, verbose; /* flags */
118
119 void check_kmem __P((void));
120 int check_space __P((void));
121 void clear_dump __P((void));
122 int Create __P((char *, int));
123 int dump_exists __P((void));
124 char *find_dev __P((dev_t, int));
125 int get_crashtime __P((void));
126 void kmem_setup __P((void));
127 void log __P((int, char *, ...));
128 void Lseek __P((int, off_t, int));
129 int Open __P((char *, int rw));
130 int Read __P((int, void *, int));
131 char *rawname __P((char *s));
132 void save_core __P((void));
133 void usage __P((void));
134 void Write __P((int, void *, int));
135
136 int
137 main(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
205 void
206 kmem_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 if (verbose)
249 (void)printf("dumplo = %d (%d * %d)\n",
250 dumplo, dumplo/DEV_BSIZE, DEV_BSIZE);
251 Lseek(kmem, (off_t)current_nl[X_DUMPMAG].n_value, L_SET);
252 (void)Read(kmem, &dumpmag, sizeof(dumpmag));
253 dumplo *= DEV_BSIZE;
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
269 void
270 check_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
300 void
301 clear_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
310 int
311 dump_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 if (newdumpmag != dumpmag) {
318 if (verbose)
319 syslog(LOG_WARNING, "magic number mismatch (%x != %x)",
320 newdumpmag, dumpmag);
321 syslog(LOG_WARNING, "no core dump");
322 return (0);
323 }
324 return (1);
325 }
326
327 char buf[1024 * 1024];
328
329 void
330 save_core()
331 {
332 register FILE *fp;
333 register int bounds, ifd, nr, nw, ofd;
334 char *rawp, path[MAXPATHLEN];
335
336 /*
337 * Get the current number and update the bounds file. Do the update
338 * now, because may fail later and don't want to overwrite anything.
339 */
340 (void)snprintf(path, sizeof(path), "%s/bounds", dirname);
341 if ((fp = fopen(path, "r")) == NULL)
342 goto err1;
343 if (fgets(buf, sizeof(buf), fp) == NULL) {
344 if (ferror(fp))
345 err1: syslog(LOG_WARNING, "%s: %s", path, strerror(errno));
346 bounds = 0;
347 } else
348 bounds = atoi(buf);
349 if (fp != NULL)
350 (void)fclose(fp);
351 if ((fp = fopen(path, "w")) == NULL)
352 syslog(LOG_ERR, "%s: %m", path);
353 else {
354 (void)fprintf(fp, "%d\n", bounds + 1);
355 (void)fclose(fp);
356 }
357 (void)fclose(fp);
358
359 /* Create the core file. */
360 (void)snprintf(path, sizeof(path), "%s/netbsd.%d%s.core",
361 dirname, bounds, compress ? ".Z" : "");
362 if (compress) {
363 if ((fp = zopen(path, "w", 0)) == NULL) {
364 syslog(LOG_ERR, "%s: %s", path, strerror(errno));
365 exit(1);
366 }
367 } else
368 ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
369
370 /* Open the raw device. */
371 rawp = rawname(ddname);
372 if ((ifd = open(rawp, O_RDONLY)) == -1) {
373 syslog(LOG_WARNING, "%s: %m; using block device", rawp);
374 ifd = dumpfd;
375 }
376
377 /* Read the dump size. */
378 Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPSIZE].n_value)), L_SET);
379 (void)Read(dumpfd, &dumpsize, sizeof(dumpsize));
380
381 /* Seek to the start of the core. */
382 Lseek(ifd, (off_t)dumplo, L_SET);
383
384 /* Copy the core file. */
385 dumpsize *= getpagesize();
386 syslog(LOG_NOTICE, "writing %score to %s",
387 compress ? "compressed " : "", path);
388 for (; dumpsize > 0; dumpsize -= nr) {
389 (void)printf("%6dK\r", dumpsize / 1024);
390 (void)fflush(stdout);
391 nr = read(ifd, buf, MIN(dumpsize, sizeof(buf)));
392 if (nr <= 0) {
393 if (nr == 0)
394 syslog(LOG_WARNING,
395 "WARNING: EOF on dump device");
396 else
397 syslog(LOG_ERR, "%s: %m", rawp);
398 goto err2;
399 }
400 if (compress)
401 nw = fwrite(buf, 1, nr, fp);
402 else
403 nw = write(ofd, buf, nr);
404 if (nw != nr) {
405 syslog(LOG_ERR, "%s: %s",
406 path, strerror(nw == 0 ? EIO : errno));
407 err2: syslog(LOG_WARNING,
408 "WARNING: core may be incomplete");
409 (void)printf("\n");
410 exit(1);
411 }
412 }
413 (void)printf("\n");
414 (void)close(ifd);
415 if (compress)
416 (void)fclose(fp);
417 else
418 (void)close(ofd);
419
420 /* Copy the kernel. */
421 ifd = Open(kernel ? kernel : _PATH_UNIX, O_RDONLY);
422 (void)snprintf(path, sizeof(path), "%s/netbsd.%d%s",
423 dirname, bounds, compress ? ".Z" : "");
424 if (compress) {
425 if ((fp = zopen(path, "w", 0)) == NULL) {
426 syslog(LOG_ERR, "%s: %s", path, strerror(errno));
427 exit(1);
428 }
429 } else
430 ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
431 syslog(LOG_NOTICE, "writing %skernel to %s",
432 compress ? "compressed " : "", path);
433 while ((nr = read(ifd, buf, sizeof(buf))) > 0) {
434 if (compress)
435 nw = fwrite(buf, 1, nr, fp);
436 else
437 nw = write(ofd, buf, nr);
438 if (nw != nr) {
439 syslog(LOG_ERR, "%s: %s",
440 path, strerror(nw == 0 ? EIO : errno));
441 syslog(LOG_WARNING,
442 "WARNING: kernel may be incomplete");
443 exit(1);
444 }
445 }
446 if (nr < 0) {
447 syslog(LOG_ERR, "%s: %s",
448 kernel ? kernel : _PATH_UNIX, strerror(errno));
449 syslog(LOG_WARNING,
450 "WARNING: kernel may be incomplete");
451 exit(1);
452 }
453 if (compress)
454 (void)fclose(fp);
455 else
456 (void)close(ofd);
457 }
458
459 char *
460 find_dev(dev, type)
461 register dev_t dev;
462 register int type;
463 {
464 register DIR *dfd;
465 struct dirent *dir;
466 struct stat sb;
467 char *dp, devname[MAXPATHLEN + 1];
468
469 if ((dfd = opendir(_PATH_DEV)) == NULL) {
470 syslog(LOG_ERR, "%s: %s", _PATH_DEV, strerror(errno));
471 exit(1);
472 }
473 (void)strcpy(devname, _PATH_DEV);
474 while ((dir = readdir(dfd))) {
475 (void)strcpy(devname + sizeof(_PATH_DEV) - 1, dir->d_name);
476 if (lstat(devname, &sb)) {
477 syslog(LOG_ERR, "%s: %s", devname, strerror(errno));
478 continue;
479 }
480 if ((sb.st_mode & S_IFMT) != type)
481 continue;
482 if (dev == sb.st_rdev) {
483 closedir(dfd);
484 if ((dp = strdup(devname)) == NULL) {
485 syslog(LOG_ERR, "%s", strerror(errno));
486 exit(1);
487 }
488 return (dp);
489 }
490 }
491 closedir(dfd);
492 syslog(LOG_ERR, "can't find device %d/%d", major(dev), minor(dev));
493 exit(1);
494 }
495
496 char *
497 rawname(s)
498 char *s;
499 {
500 char *sl, name[MAXPATHLEN];
501
502 if ((sl = strrchr(s, '/')) == NULL || sl[1] == '0') {
503 syslog(LOG_ERR,
504 "can't make raw dump device name from %s", s);
505 return (s);
506 }
507 (void)snprintf(name, sizeof(name), "%.*s/r%s", sl - s, s, sl + 1);
508 if ((sl = strdup(name)) == NULL) {
509 syslog(LOG_ERR, "%s", strerror(errno));
510 exit(1);
511 }
512 return (sl);
513 }
514
515 int
516 get_crashtime()
517 {
518 time_t dumptime; /* Time the dump was taken. */
519
520 Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_TIME].n_value)), L_SET);
521 (void)Read(dumpfd, &dumptime, sizeof(dumptime));
522 if (dumptime == 0) {
523 if (verbose)
524 syslog(LOG_ERR, "dump time is zero");
525 return (0);
526 }
527 (void)printf("savecore: system went down at %s", ctime(&dumptime));
528 #define LEEWAY (7 * SECSPERDAY)
529 if (dumptime < now - LEEWAY || dumptime > now + LEEWAY) {
530 (void)printf("dump time is unreasonable\n");
531 return (0);
532 }
533 return (1);
534 }
535
536 int
537 check_space()
538 {
539 register FILE *fp;
540 char *tkernel;
541 off_t minfree, spacefree, kernelsize, needed;
542 struct stat st;
543 struct statfs fsbuf;
544 char buf[100], path[MAXPATHLEN];
545
546 tkernel = kernel ? kernel : _PATH_UNIX;
547 if (stat(tkernel, &st) < 0) {
548 syslog(LOG_ERR, "%s: %m", tkernel);
549 exit(1);
550 }
551 kernelsize = st.st_blocks * S_BLKSIZE;
552 if (statfs(dirname, &fsbuf) < 0) {
553 syslog(LOG_ERR, "%s: %m", dirname);
554 exit(1);
555 }
556 spacefree = (fsbuf.f_bavail * fsbuf.f_bsize) / 1024;
557
558 (void)snprintf(path, sizeof(path), "%s/minfree", dirname);
559 if ((fp = fopen(path, "r")) == NULL)
560 minfree = 0;
561 else {
562 if (fgets(buf, sizeof(buf), fp) == NULL)
563 minfree = 0;
564 else
565 minfree = atoi(buf);
566 (void)fclose(fp);
567 }
568
569 needed = (dumpsize + kernelsize) / 1024;
570 if (minfree > 0 && spacefree - needed < minfree) {
571 syslog(LOG_WARNING,
572 "no dump, not enough free space on device");
573 return (0);
574 }
575 if (spacefree - needed < minfree)
576 syslog(LOG_WARNING,
577 "dump performed, but free space threshold crossed");
578 return (1);
579 }
580
581 int
582 Open(name, rw)
583 char *name;
584 int rw;
585 {
586 int fd;
587
588 if ((fd = open(name, rw, 0)) < 0) {
589 syslog(LOG_ERR, "%s: %m", name);
590 exit(1);
591 }
592 return (fd);
593 }
594
595 int
596 Read(fd, bp, size)
597 int fd, size;
598 void *bp;
599 {
600 int nr;
601
602 nr = read(fd, bp, size);
603 if (nr != size) {
604 syslog(LOG_ERR, "read: %m");
605 exit(1);
606 }
607 return (nr);
608 }
609
610 void
611 Lseek(fd, off, flag)
612 int fd, flag;
613 off_t off;
614 {
615 off_t ret;
616
617 ret = lseek(fd, off, flag);
618 if (ret == -1) {
619 syslog(LOG_ERR, "lseek: %m");
620 exit(1);
621 }
622 }
623
624 int
625 Create(file, mode)
626 char *file;
627 int mode;
628 {
629 register int fd;
630
631 fd = open(file, O_WRONLY | O_CREAT | O_TRUNC, mode);
632 if (fd < 0) {
633 syslog(LOG_ERR, "%s: %m", file);
634 exit(1);
635 }
636 return (fd);
637 }
638
639 void
640 Write(fd, bp, size)
641 int fd, size;
642 void *bp;
643 {
644 int n;
645
646 if ((n = write(fd, bp, size)) < size) {
647 syslog(LOG_ERR, "write: %s", strerror(n == -1 ? errno : EIO));
648 exit(1);
649 }
650 }
651
652 void
653 usage()
654 {
655 (void)syslog(LOG_ERR, "usage: savecore [-cfvz] [-N system] directory");
656 exit(1);
657 }
658