xref: /freebsd/sys/kern/kern_shutdown.c (revision 5129159789cc9d7bc514e4546b88e3427695002d)
1 /*-
2  * Copyright (c) 1986, 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)kern_shutdown.c	8.3 (Berkeley) 1/21/94
39  * $FreeBSD$
40  */
41 
42 #include "opt_ddb.h"
43 #include "opt_hw_wdog.h"
44 #include "opt_panic.h"
45 #include "opt_show_busybufs.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/eventhandler.h>
50 #include <sys/buf.h>
51 #include <sys/reboot.h>
52 #include <sys/proc.h>
53 #include <sys/vnode.h>
54 #include <sys/kernel.h>
55 #include <sys/mount.h>
56 #include <sys/queue.h>
57 #include <sys/sysctl.h>
58 #include <sys/conf.h>
59 #include <sys/sysproto.h>
60 #include <sys/cons.h>
61 
62 #include <machine/pcb.h>
63 #include <machine/clock.h>
64 #include <machine/md_var.h>
65 #ifdef SMP
66 #include <machine/smp.h>		/* smp_active, cpuid */
67 #endif
68 
69 #include <sys/signalvar.h>
70 
71 #ifndef PANIC_REBOOT_WAIT_TIME
72 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
73 #endif
74 
75 /*
76  * Note that stdarg.h and the ANSI style va_start macro is used for both
77  * ANSI and traditional C compilers.
78  */
79 #include <machine/stdarg.h>
80 
81 #ifdef DDB
82 #ifdef DDB_UNATTENDED
83 int debugger_on_panic = 0;
84 #else
85 int debugger_on_panic = 1;
86 #endif
87 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
88 	&debugger_on_panic, 0, "Run debugger on kernel panic");
89 #endif
90 
91 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
92 
93 #ifdef	HW_WDOG
94 /*
95  * If there is a hardware watchdog, point this at the function needed to
96  * hold it off.
97  * It's needed when the kernel needs to do some lengthy operations.
98  * e.g. in wd.c when dumping core.. It's most annoying to have
99  * your precious core-dump only half written because the wdog kicked in.
100  */
101 watchdog_tickle_fn wdog_tickler = NULL;
102 #endif	/* HW_WDOG */
103 
104 /*
105  * Variable panicstr contains argument to first call to panic; used as flag
106  * to indicate that the kernel has already called panic.
107  */
108 const char *panicstr;
109 
110 static void boot __P((int)) __dead2;
111 static void dumpsys __P((void));
112 static int setdumpdev __P((dev_t dev));
113 static void poweroff_wait __P((void *, int));
114 static void print_uptime __P((void));
115 static void shutdown_halt __P((void *junk, int howto));
116 static void shutdown_panic __P((void *junk, int howto));
117 static void shutdown_reset __P((void *junk, int howto));
118 
119 /* register various local shutdown events */
120 static void
121 shutdown_conf(void *unused)
122 {
123 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
124 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
125 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
126 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
127 }
128 
129 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
130 
131 /* ARGSUSED */
132 
133 /*
134  * The system call that results in a reboot
135  */
136 int
137 reboot(p, uap)
138 	struct proc *p;
139 	struct reboot_args *uap;
140 {
141 	int error;
142 
143 	if ((error = suser(p)))
144 		return (error);
145 
146 	boot(uap->opt);
147 	return (0);
148 }
149 
150 /*
151  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
152  */
153 void
154 shutdown_nice()
155 {
156 	/* Send a signal to init(8) and have it shutdown the world */
157 	if (initproc != NULL) {
158 		psignal(initproc, SIGINT);
159 	} else {
160 		/* No init(8) running, so simply reboot */
161 		boot(RB_NOSYNC);
162 	}
163 	return;
164 }
165 static int	waittime = -1;
166 static struct pcb dumppcb;
167 
168 static void
169 print_uptime()
170 {
171 	int f;
172 	struct timespec ts;
173 
174 	getnanouptime(&ts);
175 	printf("Uptime: ");
176 	f = 0;
177 	if (ts.tv_sec >= 86400) {
178 		printf("%ldd", ts.tv_sec / 86400);
179 		ts.tv_sec %= 86400;
180 		f = 1;
181 	}
182 	if (f || ts.tv_sec >= 3600) {
183 		printf("%ldh", ts.tv_sec / 3600);
184 		ts.tv_sec %= 3600;
185 		f = 1;
186 	}
187 	if (f || ts.tv_sec >= 60) {
188 		printf("%ldm", ts.tv_sec / 60);
189 		ts.tv_sec %= 60;
190 		f = 1;
191 	}
192 	printf("%lds\n", ts.tv_sec);
193 }
194 
195 /*
196  *  Go through the rigmarole of shutting down..
197  * this used to be in machdep.c but I'll be dammned if I could see
198  * anything machine dependant in it.
199  */
200 static void
201 boot(howto)
202 	int howto;
203 {
204 
205 #ifdef SMP
206 	if (smp_active) {
207 		printf("boot() called on cpu#%d\n", cpuid);
208 	}
209 #endif
210 	/*
211 	 * Do any callouts that should be done BEFORE syncing the filesystems.
212 	 */
213 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
214 
215 	/*
216 	 * Now sync filesystems
217 	 */
218 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
219 		register struct buf *bp;
220 		int iter, nbusy;
221 
222 		waittime = 0;
223 		printf("\nsyncing disks... ");
224 
225 		sync(&proc0, NULL);
226 
227 		/*
228 		 * With soft updates, some buffers that are
229 		 * written will be remarked as dirty until other
230 		 * buffers are written.
231 		 */
232 		for (iter = 0; iter < 20; iter++) {
233 			nbusy = 0;
234 			for (bp = &buf[nbuf]; --bp >= buf; ) {
235 				if ((bp->b_flags & B_INVAL) == 0 &&
236 				    BUF_REFCNT(bp) > 0) {
237 					nbusy++;
238 				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
239 						== B_DELWRI) {
240 					/* bawrite(bp);*/
241 					nbusy++;
242 				}
243 			}
244 			if (nbusy == 0)
245 				break;
246 			printf("%d ", nbusy);
247 			sync(&proc0, NULL);
248 			DELAY(50000 * iter);
249 		}
250 		printf("\n");
251 		/*
252 		 * Count only busy local buffers to prevent forcing
253 		 * a fsck if we're just a client of a wedged NFS server
254 		 */
255 		nbusy = 0;
256 		for (bp = &buf[nbuf]; --bp >= buf; ) {
257 			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
258 			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
259 				if (bp->b_dev == NODEV) {
260 					TAILQ_REMOVE(&mountlist,
261 					    bp->b_vp->v_mount, mnt_list);
262 					continue;
263 				}
264 				nbusy++;
265 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
266 				printf(
267 			    "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
268 				    nbusy, devtoname(bp->b_dev),
269 				    bp->b_flags, (long)bp->b_blkno,
270 				    (long)bp->b_lblkno);
271 #endif
272 			}
273 		}
274 		if (nbusy) {
275 			/*
276 			 * Failed to sync all blocks. Indicate this and don't
277 			 * unmount filesystems (thus forcing an fsck on reboot).
278 			 */
279 			printf("giving up on %d buffers\n", nbusy);
280 			DELAY(5000000);	/* 5 seconds */
281 		} else {
282 			printf("done\n");
283 			/*
284 			 * Unmount filesystems
285 			 */
286 			if (panicstr == 0)
287 				vfs_unmountall();
288 		}
289 		DELAY(100000);		/* wait for console output to finish */
290 	}
291 
292 	print_uptime();
293 
294 	/*
295 	 * Ok, now do things that assume all filesystem activity has
296 	 * been completed.
297 	 */
298 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
299 	splhigh();
300 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) {
301 		savectx(&dumppcb);
302 #ifdef __i386__
303 		dumppcb.pcb_cr3 = rcr3();
304 #endif
305 		dumpsys();
306 	}
307 
308 	/* Now that we're going to really halt the system... */
309 	EVENTHANDLER_INVOKE(shutdown_final, howto);
310 
311 	for(;;) ;	/* safety against shutdown_reset not working */
312 	/* NOTREACHED */
313 }
314 
315 /*
316  * If the shutdown was a clean halt, behave accordingly.
317  */
318 static void
319 shutdown_halt(void *junk, int howto)
320 {
321 	if (howto & RB_HALT) {
322 		printf("\n");
323 		printf("The operating system has halted.\n");
324 		printf("Please press any key to reboot.\n\n");
325 		switch (cngetc()) {
326 		case -1:		/* No console, just die */
327 			cpu_halt();
328 			/* NOTREACHED */
329 		default:
330 			howto &= ~RB_HALT;
331 			break;
332 		}
333 	}
334 }
335 
336 /*
337  * Check to see if the system paniced, pause and then reboot
338  * according to the specified delay.
339  */
340 static void
341 shutdown_panic(void *junk, int howto)
342 {
343 	int loop;
344 
345 	if (howto & RB_DUMP) {
346 		if (PANIC_REBOOT_WAIT_TIME != 0) {
347 			if (PANIC_REBOOT_WAIT_TIME != -1) {
348 				printf("Automatic reboot in %d seconds - "
349 				       "press a key on the console to abort\n",
350 					PANIC_REBOOT_WAIT_TIME);
351 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
352 				     loop > 0; --loop) {
353 					DELAY(1000 * 100); /* 1/10th second */
354 					/* Did user type a key? */
355 					if (cncheckc() != -1)
356 						break;
357 				}
358 				if (!loop)
359 					return;
360 			}
361 		} else { /* zero time specified - reboot NOW */
362 			return;
363 		}
364 		printf("--> Press a key on the console to reboot <--\n");
365 		cngetc();
366 	}
367 }
368 
369 /*
370  * Everything done, now reset
371  */
372 static void
373 shutdown_reset(void *junk, int howto)
374 {
375 	printf("Rebooting...\n");
376 	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
377 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
378 	cpu_reset();
379 	/* NOTREACHED */ /* assuming reset worked */
380 }
381 
382 /*
383  * Magic number for savecore
384  *
385  * exported (symorder) and used at least by savecore(8)
386  *
387  */
388 static u_long const	dumpmag = 0x8fca0101UL;
389 
390 static int	dumpsize = 0;		/* also for savecore */
391 
392 static int	dodump = 1;
393 
394 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
395     "Try to perform coredump on kernel panic");
396 
397 static int
398 setdumpdev(dev)
399 	dev_t dev;
400 {
401 	int psize;
402 	long newdumplo;
403 
404 	if (dev == NODEV) {
405 		dumpdev = dev;
406 		return (0);
407 	}
408 	if (devsw(dev) == NULL)
409 		return (ENXIO);		/* XXX is this right? */
410 	if (devsw(dev)->d_psize == NULL)
411 		return (ENXIO);		/* XXX should be ENODEV ? */
412 	psize = devsw(dev)->d_psize(dev);
413 	if (psize == -1)
414 		return (ENXIO);		/* XXX should be ENODEV ? */
415 	/*
416 	 * XXX should clean up checking in dumpsys() to be more like this.
417 	 */
418 	newdumplo = psize - Maxmem * PAGE_SIZE / DEV_BSIZE;
419 	if (newdumplo < 0)
420 		return (ENOSPC);
421 	dumpdev = dev;
422 	dumplo = newdumplo;
423 	return (0);
424 }
425 
426 
427 /* ARGSUSED */
428 static void dump_conf __P((void *dummy));
429 static void
430 dump_conf(dummy)
431 	void *dummy;
432 {
433 	if (setdumpdev(dumpdev) != 0)
434 		dumpdev = NODEV;
435 }
436 
437 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
438 
439 static int
440 sysctl_kern_dumpdev SYSCTL_HANDLER_ARGS
441 {
442 	int error;
443 	udev_t ndumpdev;
444 
445 	ndumpdev = dev2udev(dumpdev);
446 	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
447 	if (error == 0 && req->newptr != NULL)
448 		error = setdumpdev(udev2dev(ndumpdev, 0));
449 	return (error);
450 }
451 
452 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
453 	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
454 
455 /*
456  * Doadump comes here after turning off memory management and
457  * getting on the dump stack, either when called above, or by
458  * the auto-restart code.
459  */
460 static void
461 dumpsys(void)
462 {
463 	int	error;
464 	static int dumping;
465 
466 	if (dumping++) {
467 		printf("Dump already in progress, bailing...\n");
468 		return;
469 	}
470 	if (!dodump)
471 		return;
472 	if (dumpdev == NODEV)
473 		return;
474 	if (!(devsw(dumpdev)))
475 		return;
476 	if (!(devsw(dumpdev)->d_dump))
477 		return;
478 	dumpsize = Maxmem;
479 	printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
480 	printf("dump ");
481 	error = (*devsw(dumpdev)->d_dump)(dumpdev);
482 	if (error == 0) {
483 		printf("succeeded\n");
484 		return;
485 	}
486 	printf("failed, reason: ");
487 	switch (error) {
488 	case ENODEV:
489 		printf("device doesn't support a dump routine\n");
490 		break;
491 
492 	case ENXIO:
493 		printf("device bad\n");
494 		break;
495 
496 	case EFAULT:
497 		printf("device not ready\n");
498 		break;
499 
500 	case EINVAL:
501 		printf("area improper\n");
502 		break;
503 
504 	case EIO:
505 		printf("i/o error\n");
506 		break;
507 
508 	case EINTR:
509 		printf("aborted from console\n");
510 		break;
511 
512 	default:
513 		printf("unknown, error = %d\n", error);
514 		break;
515 	}
516 }
517 
518 /*
519  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
520  * and then reboots.  If we are called twice, then we avoid trying to sync
521  * the disks as this often leads to recursive panics.
522  */
523 void
524 panic(const char *fmt, ...)
525 {
526 	int bootopt;
527 	va_list ap;
528 	static char buf[256];
529 
530 	bootopt = RB_AUTOBOOT | RB_DUMP;
531 	if (panicstr)
532 		bootopt |= RB_NOSYNC;
533 	else
534 		panicstr = fmt;
535 
536 	va_start(ap, fmt);
537 	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
538 	if (panicstr == fmt)
539 		panicstr = buf;
540 	va_end(ap);
541 	printf("panic: %s\n", buf);
542 #ifdef SMP
543 	/* three seperate prints in case of an unmapped page and trap */
544 	printf("mp_lock = %08x; ", mp_lock);
545 	printf("cpuid = %d; ", cpuid);
546 	printf("lapic.id = %08x\n", lapic.id);
547 #endif
548 
549 #if defined(DDB)
550 	if (debugger_on_panic)
551 		Debugger ("panic");
552 #endif
553 	boot(bootopt);
554 }
555 
556 /*
557  * Support for poweroff delay.
558  */
559 #ifndef POWEROFF_DELAY
560 # define POWEROFF_DELAY 5000
561 #endif
562 static int poweroff_delay = POWEROFF_DELAY;
563 
564 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
565 	&poweroff_delay, 0, "");
566 
567 static void
568 poweroff_wait(void *junk, int howto)
569 {
570 	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
571 		return;
572 	DELAY(poweroff_delay * 1000);
573 }
574