xref: /freebsd/sbin/savecore/savecore.c (revision 05c7a37afb48ddd5ee1bd921a5d46fe59cc70b15)
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 #include <vm/pmap.h>
53 
54 #include <dirent.h>
55 #include <errno.h>
56 #include <fcntl.h>
57 #include <nlist.h>
58 #include <paths.h>
59 #include <stdio.h>
60 #include <stdlib.h>
61 #include <string.h>
62 #include <unistd.h>
63 
64 #define ok(number) ((number) - KERNBASE)
65 
66 struct nlist current_nl[] = {	/* Namelist for currently running system. */
67 #define X_DUMPDEV	0
68 	{ "_dumpdev" },
69 #define X_DUMPLO	1
70 	{ "_dumplo" },
71 #define X_TIME		2
72 	{ "_time" },
73 #define	X_DUMPSIZE	3
74 	{ "_dumpsize" },
75 #define X_VERSION	4
76 	{ "_version" },
77 #define X_PANICSTR	5
78 	{ "_panicstr" },
79 #define	X_DUMPMAG	6
80 	{ "_dumpmag" },
81 	{ "" },
82 };
83 int cursyms[] = { X_DUMPDEV, X_DUMPLO, X_VERSION, X_DUMPMAG, -1 };
84 int dumpsyms[] = { X_TIME, X_DUMPSIZE, X_VERSION, X_PANICSTR, X_DUMPMAG, -1 };
85 
86 struct nlist dump_nl[] = {	/* Name list for dumped system. */
87 	{ "_dumpdev" },		/* Entries MUST be the same as */
88 	{ "_dumplo" },		/*	those in current_nl[].  */
89 	{ "_time" },
90 	{ "_dumpsize" },
91 	{ "_version" },
92 	{ "_panicstr" },
93 	{ "_dumpmag" },
94 	{ "" },
95 };
96 
97 /* Types match kernel declarations. */
98 long	dumplo;				/* where dump starts on dumpdev */
99 int	dumpmag;			/* magic number in dump */
100 int	dumpsize;			/* amount of memory dumped */
101 
102 char	*kernel;
103 char	*dirname;			/* directory to save dumps in */
104 char	*ddname;			/* name of dump device */
105 dev_t	dumpdev;			/* dump device */
106 int	dumpfd;				/* read/write descriptor on block dev */
107 time_t	now;				/* current date */
108 char	panic_mesg[1024];
109 int	panicstr;
110 char	vers[1024];
111 
112 int	clear, compress, force, verbose;	/* flags */
113 
114 void	 check_kmem __P((void));
115 int	 check_space __P((void));
116 void	 clear_dump __P((void));
117 int	 Create __P((char *, int));
118 int	 dump_exists __P((void));
119 char	*find_dev __P((dev_t, int));
120 int	 get_crashtime __P((void));
121 void	 kmem_setup __P((void));
122 void	 log __P((int, char *, ...));
123 void	 Lseek __P((int, off_t, int));
124 int	 Open __P((const char *, int rw));
125 int	 Read __P((int, void *, int));
126 char	*rawname __P((char *s));
127 void	 save_core __P((void));
128 void	 usage __P((void));
129 void	 Write __P((int, void *, int));
130 
131 int
132 main(argc, argv)
133 	int argc;
134 	char *argv[];
135 {
136 	int ch;
137 
138 	openlog("savecore", LOG_PERROR, LOG_DAEMON);
139 
140 	while ((ch = getopt(argc, argv, "cdfN:vz")) != EOF)
141 		switch(ch) {
142 		case 'c':
143 			clear = 1;
144 			break;
145 		case 'd':		/* Not documented. */
146 		case 'v':
147 			verbose = 1;
148 			break;
149 		case 'f':
150 			force = 1;
151 			break;
152 		case 'N':
153 			kernel = optarg;
154 			break;
155 		case 'z':
156 			compress = 1;
157 			break;
158 		case '?':
159 		default:
160 			usage();
161 		}
162 	argc -= optind;
163 	argv += optind;
164 
165 	if (!clear) {
166 		if (argc != 1 && argc != 2)
167 			usage();
168 		dirname = argv[0];
169 	}
170 	if (argc == 2)
171 		kernel = argv[1];
172 
173 	(void)time(&now);
174 	kmem_setup();
175 
176 	if (clear) {
177 		clear_dump();
178 		exit(0);
179 	}
180 
181 	if (!dump_exists() && !force)
182 		exit(1);
183 
184 	check_kmem();
185 
186 	if (panicstr)
187 		syslog(LOG_ALERT, "reboot after panic: %s", panic_mesg);
188 	else
189 		syslog(LOG_ALERT, "reboot");
190 
191 	if ((!get_crashtime() || !check_space()) && !force)
192 		exit(1);
193 
194 	save_core();
195 
196 	clear_dump();
197 	exit(0);
198 }
199 
200 void
201 kmem_setup()
202 {
203 	FILE *fp;
204 	int kmem, i;
205 	const char *dump_sys;
206 
207 	/*
208 	 * Some names we need for the currently running system, others for
209 	 * the system that was running when the dump was made.  The values
210 	 * obtained from the current system are used to look for things in
211 	 * /dev/kmem that cannot be found in the dump_sys namelist, but are
212 	 * presumed to be the same (since the disk partitions are probably
213 	 * the same!)
214 	 */
215 	if ((nlist(getbootfile(), current_nl)) == -1)
216 		syslog(LOG_ERR, "%s: nlist: %s", getbootfile(),
217 		       strerror(errno));
218 	for (i = 0; cursyms[i] != -1; i++)
219 		if (current_nl[cursyms[i]].n_value == 0) {
220 			syslog(LOG_ERR, "%s: %s not in namelist",
221 			    getbootfile(), current_nl[cursyms[i]].n_name);
222 			exit(1);
223 		}
224 
225 	dump_sys = kernel ? kernel : getbootfile();
226 	if ((nlist(dump_sys, dump_nl)) == -1)
227 		syslog(LOG_ERR, "%s: nlist: %s", dump_sys, strerror(errno));
228 	for (i = 0; dumpsyms[i] != -1; i++)
229 		if (dump_nl[dumpsyms[i]].n_value == 0) {
230 			syslog(LOG_ERR, "%s: %s not in namelist",
231 			    dump_sys, dump_nl[dumpsyms[i]].n_name);
232 			exit(1);
233 		}
234 
235 	kmem = Open(_PATH_KMEM, O_RDONLY);
236 	Lseek(kmem, (off_t)current_nl[X_DUMPDEV].n_value, L_SET);
237 	(void)Read(kmem, &dumpdev, sizeof(dumpdev));
238 	if (dumpdev == NODEV) {
239 		syslog(LOG_WARNING, "no core dump (no dumpdev)");
240 		exit(1);
241 	}
242 	Lseek(kmem, (off_t)current_nl[X_DUMPLO].n_value, L_SET);
243 	(void)Read(kmem, &dumplo, sizeof(dumplo));
244 	if (verbose)
245 		(void)printf("dumplo = %d (%d * %d)\n",
246 		    dumplo, dumplo/DEV_BSIZE, DEV_BSIZE);
247 	Lseek(kmem, (off_t)current_nl[X_DUMPMAG].n_value, L_SET);
248 	(void)Read(kmem, &dumpmag, sizeof(dumpmag));
249 	dumplo *= DEV_BSIZE;
250 	ddname = find_dev(dumpdev, S_IFBLK);
251 	dumpfd = Open(ddname, O_RDWR);
252 	fp = fdopen(kmem, "r");
253 	if (fp == NULL) {
254 		syslog(LOG_ERR, "%s: fdopen: %m", _PATH_KMEM);
255 		exit(1);
256 	}
257 	if (kernel)
258 		return;
259 	(void)fseek(fp, (off_t)current_nl[X_VERSION].n_value, L_SET);
260 	(void)fgets(vers, sizeof(vers), fp);
261 
262 	/* Don't fclose(fp), we use dumpfd later. */
263 }
264 
265 void
266 check_kmem()
267 {
268 	register char *cp;
269 	FILE *fp;
270 	char core_vers[1024];
271 
272 	fp = fdopen(dumpfd, "r");
273 	if (fp == NULL) {
274 		syslog(LOG_ERR, "%s: fdopen: %m", ddname);
275 		exit(1);
276 	}
277 	fseek(fp, (off_t)(dumplo + ok(dump_nl[X_VERSION].n_value)), L_SET);
278 	fgets(core_vers, sizeof(core_vers), fp);
279 	if (strcmp(vers, core_vers) && kernel == 0)
280 		syslog(LOG_WARNING,
281 		    "warning: %s version mismatch:\n\t%s\nand\t%s\n",
282 		    getbootfile(), vers, core_vers);
283 	(void)fseek(fp,
284 	    (off_t)(dumplo + ok(dump_nl[X_PANICSTR].n_value)), L_SET);
285 	(void)fread(&panicstr, sizeof(panicstr), 1, fp);
286 	if (panicstr) {
287 		(void)fseek(fp, dumplo + ok(panicstr), L_SET);
288 		cp = panic_mesg;
289 		do
290 			*cp = getc(fp);
291 		while (*cp++ && cp < &panic_mesg[sizeof(panic_mesg)]);
292 	}
293 	/* Don't fclose(fp), we use dumpfd later. */
294 }
295 
296 void
297 clear_dump()
298 {
299 	long newdumplo;
300 
301 	newdumplo = 0;
302 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPMAG].n_value)), L_SET);
303 	Write(dumpfd, &newdumplo, sizeof(newdumplo));
304 }
305 
306 int
307 dump_exists()
308 {
309 	int newdumpmag;
310 
311 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPMAG].n_value)), L_SET);
312 	(void)Read(dumpfd, &newdumpmag, sizeof(newdumpmag));
313 	if (newdumpmag != dumpmag) {
314 		if (verbose)
315 			syslog(LOG_WARNING, "magic number mismatch (%x != %x)",
316 			    newdumpmag, dumpmag);
317 		syslog(LOG_WARNING, "no core dump");
318 		return (0);
319 	}
320 	return (1);
321 }
322 
323 char buf[1024 * 1024];
324 
325 void
326 save_core()
327 {
328 	register FILE *fp;
329 	register int bounds, ifd, nr, nw, ofd;
330 	char *rawp, path[MAXPATHLEN];
331 
332 	/*
333 	 * Get the current number and update the bounds file.  Do the update
334 	 * now, because may fail later and don't want to overwrite anything.
335 	 */
336 	(void)snprintf(path, sizeof(path), "%s/bounds", dirname);
337 	if ((fp = fopen(path, "r")) == NULL)
338 		goto err1;
339 	if (fgets(buf, sizeof(buf), fp) == NULL) {
340 		if (ferror(fp))
341 err1:			syslog(LOG_WARNING, "%s: %s", path, strerror(errno));
342 		bounds = 0;
343 	} else
344 		bounds = atoi(buf);
345 	if (fp != NULL)
346 		(void)fclose(fp);
347 	if ((fp = fopen(path, "w")) == NULL)
348 		syslog(LOG_ERR, "%s: %m", path);
349 	else {
350 		(void)fprintf(fp, "%d\n", bounds + 1);
351 		(void)fclose(fp);
352 	}
353 
354 	/* Create the core file. */
355 	(void)snprintf(path, sizeof(path), "%s/vmcore.%d%s",
356 	    dirname, bounds, compress ? ".Z" : "");
357 	if (compress) {
358 		if ((fp = zopen(path, "w", 0)) == NULL) {
359 			syslog(LOG_ERR, "%s: %s", path, strerror(errno));
360 			exit(1);
361 		}
362 	} else
363 		ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
364 
365 	/* Open the raw device. */
366 	rawp = rawname(ddname);
367 	if ((ifd = open(rawp, O_RDONLY)) == -1) {
368 		syslog(LOG_WARNING, "%s: %m; using block device", rawp);
369 		ifd = dumpfd;
370 	}
371 
372 	/* Read the dump size. */
373 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_DUMPSIZE].n_value)), L_SET);
374 	(void)Read(dumpfd, &dumpsize, sizeof(dumpsize));
375 
376 	/* Seek to the start of the core. */
377 	Lseek(ifd, (off_t)dumplo, L_SET);
378 
379 	/* Copy the core file. */
380 	dumpsize *= NBPG;
381 	syslog(LOG_NOTICE, "writing %score to %s",
382 	    compress ? "compressed " : "", path);
383 	for (; dumpsize > 0; dumpsize -= nr) {
384 		(void)printf("%6dK\r", dumpsize / 1024);
385 		(void)fflush(stdout);
386 		nr = read(ifd, buf, MIN(dumpsize, sizeof(buf)));
387 		if (nr <= 0) {
388 			if (nr == 0)
389 				syslog(LOG_WARNING,
390 				    "WARNING: EOF on dump device");
391 			else
392 				syslog(LOG_ERR, "%s: %m", rawp);
393 			goto err2;
394 		}
395 		if (compress)
396 			nw = fwrite(buf, 1, nr, fp);
397 		else
398 			nw = write(ofd, buf, nr);
399 		if (nw != nr) {
400 			syslog(LOG_ERR, "%s: %s",
401 			    path, strerror(nw == 0 ? EIO : errno));
402 err2:			syslog(LOG_WARNING,
403 			    "WARNING: vmcore may be incomplete");
404 			(void)printf("\n");
405 			exit(1);
406 		}
407 	}
408 	(void)close(ifd);
409 	if (compress)
410 		(void)fclose(fp);
411 	else
412 		(void)close(ofd);
413 
414 	/* Copy the kernel. */
415 	ifd = Open(kernel ? kernel : getbootfile(), O_RDONLY);
416 	(void)snprintf(path, sizeof(path), "%s/kernel.%d%s",
417 	    dirname, bounds, compress ? ".Z" : "");
418 	if (compress) {
419 		if ((fp = zopen(path, "w", 0)) == NULL) {
420 			syslog(LOG_ERR, "%s: %s", path, strerror(errno));
421 			exit(1);
422 		}
423 	} else
424 		ofd = Create(path, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
425 	syslog(LOG_NOTICE, "writing %skernel to %s",
426 	    compress ? "compressed " : "", path);
427 	while ((nr = read(ifd, buf, sizeof(buf))) > 0) {
428 		if (compress)
429 			nw = fwrite(buf, 1, nr, fp);
430 		else
431 			nw = write(ofd, buf, nr);
432 		if (nw != nr) {
433 			syslog(LOG_ERR, "%s: %s",
434 			    path, strerror(nw == 0 ? EIO : errno));
435 			syslog(LOG_WARNING,
436 			    "WARNING: kernel may be incomplete");
437 			exit(1);
438 		}
439 	}
440 	if (nr < 0) {
441 		syslog(LOG_ERR, "%s: %s",
442 		    kernel ? kernel : getbootfile(), strerror(errno));
443 		syslog(LOG_WARNING,
444 		    "WARNING: kernel may be incomplete");
445 		exit(1);
446 	}
447 	if (compress)
448 		(void)fclose(fp);
449 	else
450 		(void)close(ofd);
451 }
452 
453 char *
454 find_dev(dev, type)
455 	register dev_t dev;
456 	register int type;
457 {
458 	register DIR *dfd;
459 	struct dirent *dir;
460 	struct stat sb;
461 	char *dp, devname[MAXPATHLEN + 1];
462 
463 	if ((dfd = opendir(_PATH_DEV)) == NULL) {
464 		syslog(LOG_ERR, "%s: %s", _PATH_DEV, strerror(errno));
465 		exit(1);
466 	}
467 	(void)strcpy(devname, _PATH_DEV);
468 	while ((dir = readdir(dfd))) {
469 		(void)strcpy(devname + sizeof(_PATH_DEV) - 1, dir->d_name);
470 		if (lstat(devname, &sb)) {
471 			syslog(LOG_ERR, "%s: %s", devname, strerror(errno));
472 			continue;
473 		}
474 		if ((sb.st_mode & S_IFMT) != type)
475 			continue;
476 		if (dev == sb.st_rdev) {
477 			closedir(dfd);
478 			if ((dp = strdup(devname)) == NULL) {
479 				syslog(LOG_ERR, "%s", strerror(errno));
480 				exit(1);
481 			}
482 			return (dp);
483 		}
484 	}
485 	closedir(dfd);
486 	syslog(LOG_ERR, "can't find device %d/%d", major(dev), minor(dev));
487 	exit(1);
488 }
489 
490 char *
491 rawname(s)
492 	char *s;
493 {
494 	char *sl, name[MAXPATHLEN];
495 
496 	if ((sl = rindex(s, '/')) == NULL || sl[1] == '0') {
497 		syslog(LOG_ERR,
498 		    "can't make raw dump device name from %s", s);
499 		return (s);
500 	}
501 	(void)snprintf(name, sizeof(name), "%.*s/r%s", sl - s, s, sl + 1);
502 	if ((sl = strdup(name)) == NULL) {
503 		syslog(LOG_ERR, "%s", strerror(errno));
504 		exit(1);
505 	}
506 	return (sl);
507 }
508 
509 int
510 get_crashtime()
511 {
512 	time_t dumptime;			/* Time the dump was taken. */
513 
514 	Lseek(dumpfd, (off_t)(dumplo + ok(dump_nl[X_TIME].n_value)), L_SET);
515 	(void)Read(dumpfd, &dumptime, sizeof(dumptime));
516 	if (dumptime == 0) {
517 		if (verbose)
518 			syslog(LOG_ERR, "dump time is zero");
519 		return (0);
520 	}
521 	(void)printf("savecore: system went down at %s", ctime(&dumptime));
522 #define	LEEWAY	(7 * 86400)
523 	if (dumptime < now - LEEWAY || dumptime > now + LEEWAY) {
524 		(void)printf("dump time is unreasonable\n");
525 		return (0);
526 	}
527 	return (1);
528 }
529 
530 int
531 check_space()
532 {
533 	register FILE *fp;
534 	const char *tkernel;
535 	off_t minfree, spacefree, kernelsize, needed;
536 	struct stat st;
537 	struct statfs fsbuf;
538 	char buf[100], path[MAXPATHLEN];
539 
540 	tkernel = kernel ? kernel : getbootfile();
541 	if (stat(tkernel, &st) < 0) {
542 		syslog(LOG_ERR, "%s: %m", tkernel);
543 		exit(1);
544 	}
545 	kernelsize = st.st_blocks * S_BLKSIZE;
546 	if (statfs(dirname, &fsbuf) < 0) {
547 		syslog(LOG_ERR, "%s: %m", dirname);
548 		exit(1);
549 	}
550  	spacefree = (fsbuf.f_bavail * fsbuf.f_bsize) / 1024;
551 
552 	(void)snprintf(path, sizeof(path), "%s/minfree", dirname);
553 	if ((fp = fopen(path, "r")) == NULL)
554 		minfree = 0;
555 	else {
556 		if (fgets(buf, sizeof(buf), fp) == NULL)
557 			minfree = 0;
558 		else
559 			minfree = atoi(buf);
560 		(void)fclose(fp);
561 	}
562 
563 	needed = (dumpsize + kernelsize) / 1024;
564  	if (minfree > 0 && spacefree - needed < minfree) {
565 		syslog(LOG_WARNING,
566 		    "no dump, not enough free space on device");
567 		return (0);
568 	}
569 	if (spacefree - needed < minfree)
570 		syslog(LOG_WARNING,
571 		    "dump performed, but free space threshold crossed");
572 	return (1);
573 }
574 
575 int
576 Open(name, rw)
577 	const char *name;
578 	int rw;
579 {
580 	int fd;
581 
582 	if ((fd = open(name, rw, 0)) < 0) {
583 		syslog(LOG_ERR, "%s: %m", name);
584 		exit(1);
585 	}
586 	return (fd);
587 }
588 
589 int
590 Read(fd, bp, size)
591 	int fd, size;
592 	void *bp;
593 {
594 	int nr;
595 
596 	nr = read(fd, bp, size);
597 	if (nr != size) {
598 		syslog(LOG_ERR, "read: %m");
599 		exit(1);
600 	}
601 	return (nr);
602 }
603 
604 void
605 Lseek(fd, off, flag)
606 	int fd, flag;
607 	off_t off;
608 {
609 	off_t ret;
610 
611 	ret = lseek(fd, off, flag);
612 	if (ret == -1) {
613 		syslog(LOG_ERR, "lseek: %m");
614 		exit(1);
615 	}
616 }
617 
618 int
619 Create(file, mode)
620 	char *file;
621 	int mode;
622 {
623 	register int fd;
624 
625 	fd = creat(file, mode);
626 	if (fd < 0) {
627 		syslog(LOG_ERR, "%s: %m", file);
628 		exit(1);
629 	}
630 	return (fd);
631 }
632 
633 void
634 Write(fd, bp, size)
635 	int fd, size;
636 	void *bp;
637 {
638 	int n;
639 
640 	if ((n = write(fd, bp, size)) < size) {
641 		syslog(LOG_ERR, "write: %s", strerror(n == -1 ? errno : EIO));
642 		exit(1);
643 	}
644 }
645 
646 void
647 usage()
648 {
649 	(void)syslog(LOG_ERR, "usage: savecore [-cfvz] [-N system] directory");
650 	exit(1);
651 }
652