xref: /freebsd/sbin/savecore/savecore.c (revision 8e6b01171e30297084bb0b4457c4183c2746aacc)
1 /*-
2  * Copyright (c) 1986, 1992, 1993
3  *	The Regents of the University of California.  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 by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #ifndef lint
35 static char copyright[] =
36 "@(#) Copyright (c) 1986, 1992, 1993\n\
37 	The Regents of the University of California.  All rights reserved.\n";
38 #endif /* not lint */
39 
40 #ifndef lint
41 static char sccsid[] = "@(#)savecore.c	8.3 (Berkeley) 1/2/94";
42 #endif /* not lint */
43 
44 #include <sys/param.h>
45 #include <sys/stat.h>
46 #include <sys/mount.h>
47 #include <sys/syslog.h>
48 #include <sys/time.h>
49 
50 #include <vm/vm.h>
51 #include <vm/vm_param.h>
52 
53 #include <dirent.h>
54 #include <errno.h>
55 #include <fcntl.h>
56 #include <nlist.h>
57 #include <paths.h>
58 #include <stdio.h>
59 #include <stdlib.h>
60 #include <string.h>
61 #include <unistd.h>
62 
63 #define ok(number) ((number) - KERNBASE)
64 
65 struct nlist current_nl[] = {	/* Namelist for currently running system. */
66 #define X_DUMPDEV	0
67 	{ "_dumpdev" },
68 #define X_DUMPLO	1
69 	{ "_dumplo" },
70 #define X_TIME		2
71 	{ "_time" },
72 #define	X_DUMPSIZE	3
73 	{ "_dumpsize" },
74 #define X_VERSION	4
75 	{ "_version" },
76 #define X_PANICSTR	5
77 	{ "_panicstr" },
78 #define	X_DUMPMAG	6
79 	{ "_dumpmag" },
80 	{ "" },
81 };
82 int cursyms[] = { X_DUMPDEV, X_DUMPLO, X_VERSION, X_DUMPMAG, -1 };
83 int dumpsyms[] = { X_TIME, X_DUMPSIZE, X_VERSION, X_PANICSTR, X_DUMPMAG, -1 };
84 
85 struct nlist dump_nl[] = {	/* Name list for dumped system. */
86 	{ "_dumpdev" },		/* Entries MUST be the same as */
87 	{ "_dumplo" },		/*	those in current_nl[].  */
88 	{ "_time" },
89 	{ "_dumpsize" },
90 	{ "_version" },
91 	{ "_panicstr" },
92 	{ "_dumpmag" },
93 	{ "" },
94 };
95 
96 /* Types match kernel declarations. */
97 long	dumplo;				/* where dump starts on dumpdev */
98 int	dumpmag;			/* magic number in dump */
99 int	dumpsize;			/* amount of memory dumped */
100 
101 char	*kernel;
102 char	*dirname;			/* directory to save dumps in */
103 char	*ddname;			/* name of dump device */
104 dev_t	dumpdev;			/* dump device */
105 int	dumpfd;				/* read/write descriptor on block dev */
106 time_t	now;				/* current date */
107 char	panic_mesg[1024];
108 int	panicstr;
109 char	vers[1024];
110 
111 int	clear, compress, force, verbose;	/* flags */
112 
113 void	 check_kmem __P((void));
114 int	 check_space __P((void));
115 void	 clear_dump __P((void));
116 int	 Create __P((char *, int));
117 int	 dump_exists __P((void));
118 char	*find_dev __P((dev_t, int));
119 int	 get_crashtime __P((void));
120 void	 kmem_setup __P((void));
121 void	 log __P((int, char *, ...));
122 void	 Lseek __P((int, off_t, int));
123 int	 Open __P((const char *, int rw));
124 int	 Read __P((int, void *, int));
125 char	*rawname __P((char *s));
126 void	 save_core __P((void));
127 void	 usage __P((void));
128 void	 Write __P((int, void *, int));
129 
130 int
131 main(argc, argv)
132 	int argc;
133 	char *argv[];
134 {
135 	int ch;
136 
137 	openlog("savecore", LOG_PERROR, LOG_DAEMON);
138 
139 	while ((ch = getopt(argc, argv, "cdfN:vz")) != EOF)
140 		switch(ch) {
141 		case 'c':
142 			clear = 1;
143 			break;
144 		case 'd':		/* Not documented. */
145 		case 'v':
146 			verbose = 1;
147 			break;
148 		case 'f':
149 			force = 1;
150 			break;
151 		case 'N':
152 			kernel = optarg;
153 			break;
154 		case 'z':
155 			compress = 1;
156 			break;
157 		case '?':
158 		default:
159 			usage();
160 		}
161 	argc -= optind;
162 	argv += optind;
163 
164 	if (!clear) {
165 		if (argc != 1 && argc != 2)
166 			usage();
167 		dirname = argv[0];
168 	}
169 	if (argc == 2)
170 		kernel = argv[1];
171 
172 	(void)time(&now);
173 	kmem_setup();
174 
175 	if (clear) {
176 		clear_dump();
177 		exit(0);
178 	}
179 
180 	if (!dump_exists() && !force)
181 		exit(1);
182 
183 	check_kmem();
184 
185 	if (panicstr)
186 		syslog(LOG_ALERT, "reboot after panic: %s", panic_mesg);
187 	else
188 		syslog(LOG_ALERT, "reboot");
189 
190 	if ((!get_crashtime() || !check_space()) && !force)
191 		exit(1);
192 
193 	save_core();
194 
195 	clear_dump();
196 	exit(0);
197 }
198 
199 void
200 kmem_setup()
201 {
202 	FILE *fp;
203 	int kmem, i;
204 	const char *dump_sys;
205 
206 	/*
207 	 * Some names we need for the currently running system, others for
208 	 * the system that was running when the dump was made.  The values
209 	 * obtained from the current system are used to look for things in
210 	 * /dev/kmem that cannot be found in the dump_sys namelist, but are
211 	 * presumed to be the same (since the disk partitions are probably
212 	 * the same!)
213 	 */
214 	if ((nlist(getbootfile(), current_nl)) == -1)
215 		syslog(LOG_ERR, "%s: nlist: %s", getbootfile(),
216 		       strerror(errno));
217 	for (i = 0; cursyms[i] != -1; i++)
218 		if (current_nl[cursyms[i]].n_value == 0) {
219 			syslog(LOG_ERR, "%s: %s not in namelist",
220 			    getbootfile(), current_nl[cursyms[i]].n_name);
221 			exit(1);
222 		}
223 
224 	dump_sys = kernel ? kernel : getbootfile();
225 	if ((nlist(dump_sys, dump_nl)) == -1)
226 		syslog(LOG_ERR, "%s: nlist: %s", dump_sys, strerror(errno));
227 	for (i = 0; dumpsyms[i] != -1; i++)
228 		if (dump_nl[dumpsyms[i]].n_value == 0) {
229 			syslog(LOG_ERR, "%s: %s not in namelist",
230 			    dump_sys, dump_nl[dumpsyms[i]].n_name);
231 			exit(1);
232 		}
233 
234 	kmem = Open(_PATH_KMEM, O_RDONLY);
235 	Lseek(kmem, (off_t)current_nl[X_DUMPDEV].n_value, L_SET);
236 	(void)Read(kmem, &dumpdev, sizeof(dumpdev));
237 	if (dumpdev == NODEV) {
238 		syslog(LOG_WARNING, "no core dump (no dumpdev)");
239 		exit(1);
240 	}
241 	Lseek(kmem, (off_t)current_nl[X_DUMPLO].n_value, L_SET);
242 	(void)Read(kmem, &dumplo, sizeof(dumplo));
243 	if (verbose)
244 		(void)printf("dumplo = %d (%d * %d)\n",
245 		    dumplo, dumplo/DEV_BSIZE, DEV_BSIZE);
246 	Lseek(kmem, (off_t)current_nl[X_DUMPMAG].n_value, L_SET);
247 	(void)Read(kmem, &dumpmag, sizeof(dumpmag));
248 	dumplo *= DEV_BSIZE;
249 	ddname = find_dev(dumpdev, S_IFBLK);
250 	dumpfd = Open(ddname, O_RDWR);
251 	fp = fdopen(kmem, "r");
252 	if (fp == NULL) {
253 		syslog(LOG_ERR, "%s: fdopen: %m", _PATH_KMEM);
254 		exit(1);
255 	}
256 	if (kernel)
257 		return;
258 	(void)fseek(fp, (off_t)current_nl[X_VERSION].n_value, L_SET);
259 	(void)fgets(vers, sizeof(vers), fp);
260 
261 	/* Don't fclose(fp), we use dumpfd later. */
262 }
263 
264 void
265 check_kmem()
266 {
267 	register char *cp;
268 	FILE *fp;
269 	char core_vers[1024];
270 
271 	fp = fdopen(dumpfd, "r");
272 	if (fp == NULL) {
273 		syslog(LOG_ERR, "%s: fdopen: %m", ddname);
274 		exit(1);
275 	}
276 	fseek(fp, (off_t)(dumplo + ok(dump_nl[X_VERSION].n_value)), L_SET);
277 	fgets(core_vers, sizeof(core_vers), fp);
278 	if (strcmp(vers, core_vers) && kernel == 0)
279 		syslog(LOG_WARNING,
280 		    "warning: %s version mismatch:\n\t%s\nand\t%s\n",
281 		    getbootfile(), vers, core_vers);
282 	(void)fseek(fp,
283 	    (off_t)(dumplo + ok(dump_nl[X_PANICSTR].n_value)), L_SET);
284 	(void)fread(&panicstr, sizeof(panicstr), 1, fp);
285 	if (panicstr) {
286 		(void)fseek(fp, dumplo + ok(panicstr), L_SET);
287 		cp = panic_mesg;
288 		do
289 			*cp = getc(fp);
290 		while (*cp++ && cp < &panic_mesg[sizeof(panic_mesg)]);
291 	}
292 	/* Don't fclose(fp), we use dumpfd later. */
293 }
294 
295 void
296 clear_dump()
297 {
298 	long newdumplo;
299 
300 	newdumplo = 0;
301 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPMAG].n_value)), L_SET);
302 	Write(dumpfd, &newdumplo, sizeof(newdumplo));
303 }
304 
305 int
306 dump_exists()
307 {
308 	int newdumpmag;
309 
310 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPMAG].n_value)), L_SET);
311 	(void)Read(dumpfd, &newdumpmag, sizeof(newdumpmag));
312 	if (newdumpmag != dumpmag) {
313 		if (verbose)
314 			syslog(LOG_WARNING, "magic number mismatch (%x != %x)",
315 			    newdumpmag, dumpmag);
316 		syslog(LOG_WARNING, "no core dump");
317 		return (0);
318 	}
319 	return (1);
320 }
321 
322 char buf[1024 * 1024];
323 
324 void
325 save_core()
326 {
327 	register FILE *fp;
328 	register int bounds, ifd, nr, nw, ofd;
329 	char *rawp, path[MAXPATHLEN];
330 
331 	/*
332 	 * Get the current number and update the bounds file.  Do the update
333 	 * now, because may fail later and don't want to overwrite anything.
334 	 */
335 	(void)snprintf(path, sizeof(path), "%s/bounds", dirname);
336 	if ((fp = fopen(path, "r")) == NULL)
337 		goto err1;
338 	if (fgets(buf, sizeof(buf), fp) == NULL) {
339 		if (ferror(fp))
340 err1:			syslog(LOG_WARNING, "%s: %s", path, strerror(errno));
341 		bounds = 0;
342 	} else
343 		bounds = atoi(buf);
344 	if (fp != NULL)
345 		(void)fclose(fp);
346 	if ((fp = fopen(path, "w")) == NULL)
347 		syslog(LOG_ERR, "%s: %m", path);
348 	else {
349 		(void)fprintf(fp, "%d\n", bounds + 1);
350 		(void)fclose(fp);
351 	}
352 
353 	/* Create the core file. */
354 	(void)snprintf(path, sizeof(path), "%s/vmcore.%d%s",
355 	    dirname, bounds, compress ? ".Z" : "");
356 	if (compress) {
357 		if ((fp = zopen(path, "w", 0)) == NULL) {
358 			syslog(LOG_ERR, "%s: %s", path, strerror(errno));
359 			exit(1);
360 		}
361 	} else
362 		ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
363 
364 	/* Open the raw device. */
365 	rawp = rawname(ddname);
366 	if ((ifd = open(rawp, O_RDONLY)) == -1) {
367 		syslog(LOG_WARNING, "%s: %m; using block device", rawp);
368 		ifd = dumpfd;
369 	}
370 
371 	/* Read the dump size. */
372 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPSIZE].n_value)), L_SET);
373 	(void)Read(dumpfd, &dumpsize, sizeof(dumpsize));
374 
375 	/* Seek to the start of the core. */
376 	Lseek(ifd, (off_t)dumplo, L_SET);
377 
378 	/* Copy the core file. */
379 	dumpsize *= NBPG;
380 	syslog(LOG_NOTICE, "writing %score to %s",
381 	    compress ? "compressed " : "", path);
382 	for (; dumpsize > 0; dumpsize -= nr) {
383 		(void)printf("%6dK\r", dumpsize / 1024);
384 		(void)fflush(stdout);
385 		nr = read(ifd, buf, MIN(dumpsize, sizeof(buf)));
386 		if (nr <= 0) {
387 			if (nr == 0)
388 				syslog(LOG_WARNING,
389 				    "WARNING: EOF on dump device");
390 			else
391 				syslog(LOG_ERR, "%s: %m", rawp);
392 			goto err2;
393 		}
394 		if (compress)
395 			nw = fwrite(buf, 1, nr, fp);
396 		else
397 			nw = write(ofd, buf, nr);
398 		if (nw != nr) {
399 			syslog(LOG_ERR, "%s: %s",
400 			    path, strerror(nw == 0 ? EIO : errno));
401 err2:			syslog(LOG_WARNING,
402 			    "WARNING: vmcore may be incomplete");
403 			(void)printf("\n");
404 			exit(1);
405 		}
406 	}
407 	(void)close(ifd);
408 	if (compress)
409 		(void)fclose(fp);
410 	else
411 		(void)close(ofd);
412 
413 	/* Copy the kernel. */
414 	ifd = Open(kernel ? kernel : getbootfile(), O_RDONLY);
415 	(void)snprintf(path, sizeof(path), "%s/kernel.%d%s",
416 	    dirname, bounds, compress ? ".Z" : "");
417 	if (compress) {
418 		if ((fp = zopen(path, "w", 0)) == NULL) {
419 			syslog(LOG_ERR, "%s: %s", path, strerror(errno));
420 			exit(1);
421 		}
422 	} else
423 		ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
424 	syslog(LOG_NOTICE, "writing %skernel to %s",
425 	    compress ? "compressed " : "", path);
426 	while ((nr = read(ifd, buf, sizeof(buf))) > 0) {
427 		if (compress)
428 			nw = fwrite(buf, 1, nr, fp);
429 		else
430 			nw = write(ofd, buf, nr);
431 		if (nw != nr) {
432 			syslog(LOG_ERR, "%s: %s",
433 			    path, strerror(nw == 0 ? EIO : errno));
434 			syslog(LOG_WARNING,
435 			    "WARNING: kernel may be incomplete");
436 			exit(1);
437 		}
438 	}
439 	if (nr < 0) {
440 		syslog(LOG_ERR, "%s: %s",
441 		    kernel ? kernel : getbootfile(), strerror(errno));
442 		syslog(LOG_WARNING,
443 		    "WARNING: kernel may be incomplete");
444 		exit(1);
445 	}
446 	if (compress)
447 		(void)fclose(fp);
448 	else
449 		(void)close(ofd);
450 }
451 
452 char *
453 find_dev(dev, type)
454 	register dev_t dev;
455 	register int type;
456 {
457 	register DIR *dfd;
458 	struct dirent *dir;
459 	struct stat sb;
460 	char *dp, devname[MAXPATHLEN + 1];
461 
462 	if ((dfd = opendir(_PATH_DEV)) == NULL) {
463 		syslog(LOG_ERR, "%s: %s", _PATH_DEV, strerror(errno));
464 		exit(1);
465 	}
466 	(void)strcpy(devname, _PATH_DEV);
467 	while ((dir = readdir(dfd))) {
468 		(void)strcpy(devname + sizeof(_PATH_DEV) - 1, dir->d_name);
469 		if (lstat(devname, &sb)) {
470 			syslog(LOG_ERR, "%s: %s", devname, strerror(errno));
471 			continue;
472 		}
473 		if ((sb.st_mode & S_IFMT) != type)
474 			continue;
475 		if (dev == sb.st_rdev) {
476 			closedir(dfd);
477 			if ((dp = strdup(devname)) == NULL) {
478 				syslog(LOG_ERR, "%s", strerror(errno));
479 				exit(1);
480 			}
481 			return (dp);
482 		}
483 	}
484 	closedir(dfd);
485 	syslog(LOG_ERR, "can't find device %d/%d", major(dev), minor(dev));
486 	exit(1);
487 }
488 
489 char *
490 rawname(s)
491 	char *s;
492 {
493 	char *sl, name[MAXPATHLEN];
494 
495 	if ((sl = rindex(s, '/')) == NULL || sl[1] == '0') {
496 		syslog(LOG_ERR,
497 		    "can't make raw dump device name from %s", s);
498 		return (s);
499 	}
500 	(void)snprintf(name, sizeof(name), "%.*s/r%s", sl - s, s, sl + 1);
501 	if ((sl = strdup(name)) == NULL) {
502 		syslog(LOG_ERR, "%s", strerror(errno));
503 		exit(1);
504 	}
505 	return (sl);
506 }
507 
508 int
509 get_crashtime()
510 {
511 	time_t dumptime;			/* Time the dump was taken. */
512 
513 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_TIME].n_value)), L_SET);
514 	(void)Read(dumpfd, &dumptime, sizeof(dumptime));
515 	if (dumptime == 0) {
516 		if (verbose)
517 			syslog(LOG_ERR, "dump time is zero");
518 		return (0);
519 	}
520 	(void)printf("savecore: system went down at %s", ctime(&dumptime));
521 #define	LEEWAY	(7 * 86400)
522 	if (dumptime < now - LEEWAY || dumptime > now + LEEWAY) {
523 		(void)printf("dump time is unreasonable\n");
524 		return (0);
525 	}
526 	return (1);
527 }
528 
529 int
530 check_space()
531 {
532 	register FILE *fp;
533 	const char *tkernel;
534 	off_t minfree, spacefree, kernelsize, needed;
535 	struct stat st;
536 	struct statfs fsbuf;
537 	char buf[100], path[MAXPATHLEN];
538 
539 	tkernel = kernel ? kernel : getbootfile();
540 	if (stat(tkernel, &st) < 0) {
541 		syslog(LOG_ERR, "%s: %m", tkernel);
542 		exit(1);
543 	}
544 	kernelsize = st.st_blocks * S_BLKSIZE;
545 	if (statfs(dirname, &fsbuf) < 0) {
546 		syslog(LOG_ERR, "%s: %m", dirname);
547 		exit(1);
548 	}
549  	spacefree = (fsbuf.f_bavail * fsbuf.f_bsize) / 1024;
550 
551 	(void)snprintf(path, sizeof(path), "%s/minfree", dirname);
552 	if ((fp = fopen(path, "r")) == NULL)
553 		minfree = 0;
554 	else {
555 		if (fgets(buf, sizeof(buf), fp) == NULL)
556 			minfree = 0;
557 		else
558 			minfree = atoi(buf);
559 		(void)fclose(fp);
560 	}
561 
562 	needed = (dumpsize + kernelsize) / 1024;
563  	if (minfree > 0 && spacefree - needed < minfree) {
564 		syslog(LOG_WARNING,
565 		    "no dump, not enough free space on device");
566 		return (0);
567 	}
568 	if (spacefree - needed < minfree)
569 		syslog(LOG_WARNING,
570 		    "dump performed, but free space threshold crossed");
571 	return (1);
572 }
573 
574 int
575 Open(name, rw)
576 	const char *name;
577 	int rw;
578 {
579 	int fd;
580 
581 	if ((fd = open(name, rw, 0)) < 0) {
582 		syslog(LOG_ERR, "%s: %m", name);
583 		exit(1);
584 	}
585 	return (fd);
586 }
587 
588 int
589 Read(fd, bp, size)
590 	int fd, size;
591 	void *bp;
592 {
593 	int nr;
594 
595 	nr = read(fd, bp, size);
596 	if (nr != size) {
597 		syslog(LOG_ERR, "read: %m");
598 		exit(1);
599 	}
600 	return (nr);
601 }
602 
603 void
604 Lseek(fd, off, flag)
605 	int fd, flag;
606 	off_t off;
607 {
608 	off_t ret;
609 
610 	ret = lseek(fd, off, flag);
611 	if (ret == -1) {
612 		syslog(LOG_ERR, "lseek: %m");
613 		exit(1);
614 	}
615 }
616 
617 int
618 Create(file, mode)
619 	char *file;
620 	int mode;
621 {
622 	register int fd;
623 
624 	fd = creat(file, mode);
625 	if (fd < 0) {
626 		syslog(LOG_ERR, "%s: %m", file);
627 		exit(1);
628 	}
629 	return (fd);
630 }
631 
632 void
633 Write(fd, bp, size)
634 	int fd, size;
635 	void *bp;
636 {
637 	int n;
638 
639 	if ((n = write(fd, bp, size)) < size) {
640 		syslog(LOG_ERR, "write: %s", strerror(n == -1 ? errno : EIO));
641 		exit(1);
642 	}
643 }
644 
645 void
646 usage()
647 {
648 	(void)syslog(LOG_ERR, "usage: savecore [-cfvz] [-N system] directory");
649 	exit(1);
650 }
651