xref: /freebsd/sys/kern/kern_shutdown.c (revision 23f282aa31e9b6fceacd449020e936e98d6f2298)
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/kthread.h>
56 #include <sys/mount.h>
57 #include <sys/queue.h>
58 #include <sys/sysctl.h>
59 #include <sys/conf.h>
60 #include <sys/sysproto.h>
61 #include <sys/cons.h>
62 
63 #include <machine/pcb.h>
64 #include <machine/clock.h>
65 #include <machine/md_var.h>
66 #include <machine/smp.h>		/* smp_active, cpuid */
67 
68 #include <sys/signalvar.h>
69 
70 #ifndef PANIC_REBOOT_WAIT_TIME
71 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
72 #endif
73 
74 /*
75  * Note that stdarg.h and the ANSI style va_start macro is used for both
76  * ANSI and traditional C compilers.
77  */
78 #include <machine/stdarg.h>
79 
80 #ifdef DDB
81 #ifdef DDB_UNATTENDED
82 int debugger_on_panic = 0;
83 #else
84 int debugger_on_panic = 1;
85 #endif
86 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
87 	&debugger_on_panic, 0, "Run debugger on kernel panic");
88 #endif
89 
90 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
91 
92 #ifdef	HW_WDOG
93 /*
94  * If there is a hardware watchdog, point this at the function needed to
95  * hold it off.
96  * It's needed when the kernel needs to do some lengthy operations.
97  * e.g. in wd.c when dumping core.. It's most annoying to have
98  * your precious core-dump only half written because the wdog kicked in.
99  */
100 watchdog_tickle_fn wdog_tickler = NULL;
101 #endif	/* HW_WDOG */
102 
103 /*
104  * Variable panicstr contains argument to first call to panic; used as flag
105  * to indicate that the kernel has already called panic.
106  */
107 const char *panicstr;
108 
109 static void boot __P((int)) __dead2;
110 static void dumpsys __P((void));
111 static int setdumpdev __P((dev_t dev));
112 static void poweroff_wait __P((void *, int));
113 static void print_uptime __P((void));
114 static void shutdown_halt __P((void *junk, int howto));
115 static void shutdown_panic __P((void *junk, int howto));
116 static void shutdown_reset __P((void *junk, int howto));
117 
118 /* register various local shutdown events */
119 static void
120 shutdown_conf(void *unused)
121 {
122 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
123 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
124 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
125 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
126 }
127 
128 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
129 
130 /* ARGSUSED */
131 
132 /*
133  * The system call that results in a reboot
134  */
135 int
136 reboot(p, uap)
137 	struct proc *p;
138 	struct reboot_args *uap;
139 {
140 	int error;
141 
142 	if ((error = suser(p)))
143 		return (error);
144 
145 	boot(uap->opt);
146 	return (0);
147 }
148 
149 /*
150  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
151  */
152 void
153 shutdown_nice()
154 {
155 	/* Send a signal to init(8) and have it shutdown the world */
156 	if (initproc != NULL) {
157 		psignal(initproc, SIGINT);
158 	} else {
159 		/* No init(8) running, so simply reboot */
160 		boot(RB_NOSYNC);
161 	}
162 	return;
163 }
164 static int	waittime = -1;
165 static struct pcb dumppcb;
166 
167 static void
168 print_uptime()
169 {
170 	int f;
171 	struct timespec ts;
172 
173 	getnanouptime(&ts);
174 	printf("Uptime: ");
175 	f = 0;
176 	if (ts.tv_sec >= 86400) {
177 		printf("%ldd", ts.tv_sec / 86400);
178 		ts.tv_sec %= 86400;
179 		f = 1;
180 	}
181 	if (f || ts.tv_sec >= 3600) {
182 		printf("%ldh", ts.tv_sec / 3600);
183 		ts.tv_sec %= 3600;
184 		f = 1;
185 	}
186 	if (f || ts.tv_sec >= 60) {
187 		printf("%ldm", ts.tv_sec / 60);
188 		ts.tv_sec %= 60;
189 		f = 1;
190 	}
191 	printf("%lds\n", ts.tv_sec);
192 }
193 
194 /*
195  *  Go through the rigmarole of shutting down..
196  * this used to be in machdep.c but I'll be dammned if I could see
197  * anything machine dependant in it.
198  */
199 static void
200 boot(howto)
201 	int howto;
202 {
203 
204 #ifdef SMP
205 	if (smp_active) {
206 		printf("boot() called on cpu#%d\n", cpuid);
207 	}
208 #endif
209 	/*
210 	 * Do any callouts that should be done BEFORE syncing the filesystems.
211 	 */
212 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
213 
214 	/*
215 	 * Now sync filesystems
216 	 */
217 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
218 		register struct buf *bp;
219 		int iter, nbusy;
220 
221 		waittime = 0;
222 		printf("\nsyncing disks... ");
223 
224 		sync(&proc0, NULL);
225 
226 		/*
227 		 * With soft updates, some buffers that are
228 		 * written will be remarked as dirty until other
229 		 * buffers are written.
230 		 */
231 		for (iter = 0; iter < 20; iter++) {
232 			nbusy = 0;
233 			for (bp = &buf[nbuf]; --bp >= buf; ) {
234 				if ((bp->b_flags & B_INVAL) == 0 &&
235 				    BUF_REFCNT(bp) > 0) {
236 					nbusy++;
237 				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
238 						== B_DELWRI) {
239 					/* bawrite(bp);*/
240 					nbusy++;
241 				}
242 			}
243 			if (nbusy == 0)
244 				break;
245 			printf("%d ", nbusy);
246 			sync(&proc0, NULL);
247 			DELAY(50000 * iter);
248 		}
249 		printf("\n");
250 		/*
251 		 * Count only busy local buffers to prevent forcing
252 		 * a fsck if we're just a client of a wedged NFS server
253 		 */
254 		nbusy = 0;
255 		for (bp = &buf[nbuf]; --bp >= buf; ) {
256 			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
257 			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
258 				if (bp->b_dev == NODEV) {
259 					TAILQ_REMOVE(&mountlist,
260 					    bp->b_vp->v_mount, mnt_list);
261 					continue;
262 				}
263 				nbusy++;
264 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
265 				printf(
266 			    "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
267 				    nbusy, devtoname(bp->b_dev),
268 				    bp->b_flags, (long)bp->b_blkno,
269 				    (long)bp->b_lblkno);
270 #endif
271 			}
272 		}
273 		if (nbusy) {
274 			/*
275 			 * Failed to sync all blocks. Indicate this and don't
276 			 * unmount filesystems (thus forcing an fsck on reboot).
277 			 */
278 			printf("giving up on %d buffers\n", nbusy);
279 			DELAY(5000000);	/* 5 seconds */
280 		} else {
281 			printf("done\n");
282 			/*
283 			 * Unmount filesystems
284 			 */
285 			if (panicstr == 0)
286 				vfs_unmountall();
287 		}
288 		DELAY(100000);		/* wait for console output to finish */
289 	}
290 
291 	print_uptime();
292 
293 	/*
294 	 * Ok, now do things that assume all filesystem activity has
295 	 * been completed.
296 	 */
297 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
298 	splhigh();
299 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) {
300 		savectx(&dumppcb);
301 #ifdef __i386__
302 		dumppcb.pcb_cr3 = rcr3();
303 #endif
304 		dumpsys();
305 	}
306 
307 	/* Now that we're going to really halt the system... */
308 	EVENTHANDLER_INVOKE(shutdown_final, howto);
309 
310 	for(;;) ;	/* safety against shutdown_reset not working */
311 	/* NOTREACHED */
312 }
313 
314 /*
315  * If the shutdown was a clean halt, behave accordingly.
316  */
317 static void
318 shutdown_halt(void *junk, int howto)
319 {
320 	if (howto & RB_HALT) {
321 		printf("\n");
322 		printf("The operating system has halted.\n");
323 		printf("Please press any key to reboot.\n\n");
324 		switch (cngetc()) {
325 		case -1:		/* No console, just die */
326 			cpu_halt();
327 			/* NOTREACHED */
328 		default:
329 			howto &= ~RB_HALT;
330 			break;
331 		}
332 	}
333 }
334 
335 /*
336  * Check to see if the system paniced, pause and then reboot
337  * according to the specified delay.
338  */
339 static void
340 shutdown_panic(void *junk, int howto)
341 {
342 	int loop;
343 
344 	if (howto & RB_DUMP) {
345 		if (PANIC_REBOOT_WAIT_TIME != 0) {
346 			if (PANIC_REBOOT_WAIT_TIME != -1) {
347 				printf("Automatic reboot in %d seconds - "
348 				       "press a key on the console to abort\n",
349 					PANIC_REBOOT_WAIT_TIME);
350 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
351 				     loop > 0; --loop) {
352 					DELAY(1000 * 100); /* 1/10th second */
353 					/* Did user type a key? */
354 					if (cncheckc() != -1)
355 						break;
356 				}
357 				if (!loop)
358 					return;
359 			}
360 		} else { /* zero time specified - reboot NOW */
361 			return;
362 		}
363 		printf("--> Press a key on the console to reboot <--\n");
364 		cngetc();
365 	}
366 }
367 
368 /*
369  * Everything done, now reset
370  */
371 static void
372 shutdown_reset(void *junk, int howto)
373 {
374 	printf("Rebooting...\n");
375 	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
376 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
377 	cpu_reset();
378 	/* NOTREACHED */ /* assuming reset worked */
379 }
380 
381 /*
382  * Magic number for savecore
383  *
384  * exported (symorder) and used at least by savecore(8)
385  *
386  */
387 static u_long const	dumpmag = 0x8fca0101UL;
388 
389 static int	dumpsize = 0;		/* also for savecore */
390 
391 static int	dodump = 1;
392 
393 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
394     "Try to perform coredump on kernel panic");
395 
396 static int
397 setdumpdev(dev)
398 	dev_t dev;
399 {
400 	int psize;
401 	long newdumplo;
402 
403 	if (dev == NODEV) {
404 		dumpdev = dev;
405 		return (0);
406 	}
407 	if (devsw(dev) == NULL)
408 		return (ENXIO);		/* XXX is this right? */
409 	if (devsw(dev)->d_psize == NULL)
410 		return (ENXIO);		/* XXX should be ENODEV ? */
411 	psize = devsw(dev)->d_psize(dev);
412 	if (psize == -1)
413 		return (ENXIO);		/* XXX should be ENODEV ? */
414 	/*
415 	 * XXX should clean up checking in dumpsys() to be more like this.
416 	 */
417 	newdumplo = psize - Maxmem * PAGE_SIZE / DEV_BSIZE;
418 	if (newdumplo < 0)
419 		return (ENOSPC);
420 	dumpdev = dev;
421 	dumplo = newdumplo;
422 	return (0);
423 }
424 
425 
426 /* ARGSUSED */
427 static void dump_conf __P((void *dummy));
428 static void
429 dump_conf(dummy)
430 	void *dummy;
431 {
432 	if (setdumpdev(dumpdev) != 0)
433 		dumpdev = NODEV;
434 }
435 
436 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
437 
438 static int
439 sysctl_kern_dumpdev SYSCTL_HANDLER_ARGS
440 {
441 	int error;
442 	udev_t ndumpdev;
443 
444 	ndumpdev = dev2udev(dumpdev);
445 	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
446 	if (error == 0 && req->newptr != NULL)
447 		error = setdumpdev(udev2dev(ndumpdev, 0));
448 	return (error);
449 }
450 
451 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
452 	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
453 
454 /*
455  * Doadump comes here after turning off memory management and
456  * getting on the dump stack, either when called above, or by
457  * the auto-restart code.
458  */
459 static void
460 dumpsys(void)
461 {
462 	int	error;
463 	static int dumping;
464 
465 	if (dumping++) {
466 		printf("Dump already in progress, bailing...\n");
467 		return;
468 	}
469 	if (!dodump)
470 		return;
471 	if (dumpdev == NODEV)
472 		return;
473 	if (!(devsw(dumpdev)))
474 		return;
475 	if (!(devsw(dumpdev)->d_dump))
476 		return;
477 	dumpsize = Maxmem;
478 	printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
479 	printf("dump ");
480 	error = (*devsw(dumpdev)->d_dump)(dumpdev);
481 	if (error == 0) {
482 		printf("succeeded\n");
483 		return;
484 	}
485 	printf("failed, reason: ");
486 	switch (error) {
487 	case ENODEV:
488 		printf("device doesn't support a dump routine\n");
489 		break;
490 
491 	case ENXIO:
492 		printf("device bad\n");
493 		break;
494 
495 	case EFAULT:
496 		printf("device not ready\n");
497 		break;
498 
499 	case EINVAL:
500 		printf("area improper\n");
501 		break;
502 
503 	case EIO:
504 		printf("i/o error\n");
505 		break;
506 
507 	case EINTR:
508 		printf("aborted from console\n");
509 		break;
510 
511 	default:
512 		printf("unknown, error = %d\n", error);
513 		break;
514 	}
515 }
516 
517 /*
518  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
519  * and then reboots.  If we are called twice, then we avoid trying to sync
520  * the disks as this often leads to recursive panics.
521  */
522 void
523 panic(const char *fmt, ...)
524 {
525 	int bootopt;
526 	va_list ap;
527 	static char buf[256];
528 
529 	bootopt = RB_AUTOBOOT | RB_DUMP;
530 	if (panicstr)
531 		bootopt |= RB_NOSYNC;
532 	else
533 		panicstr = fmt;
534 
535 	va_start(ap, fmt);
536 	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
537 	if (panicstr == fmt)
538 		panicstr = buf;
539 	va_end(ap);
540 	printf("panic: %s\n", buf);
541 #ifdef SMP
542 	/* three seperate prints in case of an unmapped page and trap */
543 	printf("mp_lock = %08x; ", mp_lock);
544 	printf("cpuid = %d; ", cpuid);
545 	printf("lapic.id = %08x\n", lapic.id);
546 #endif
547 
548 #if defined(DDB)
549 	if (debugger_on_panic)
550 		Debugger ("panic");
551 #endif
552 	boot(bootopt);
553 }
554 
555 /*
556  * Support for poweroff delay.
557  */
558 #ifndef POWEROFF_DELAY
559 # define POWEROFF_DELAY 5000
560 #endif
561 static int poweroff_delay = POWEROFF_DELAY;
562 
563 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
564 	&poweroff_delay, 0, "");
565 
566 static void
567 poweroff_wait(void *junk, int howto)
568 {
569 	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
570 		return;
571 	DELAY(poweroff_delay * 1000);
572 }
573 
574 /*
575  * Some system processes (e.g. syncer) need to be stopped at appropriate
576  * points in their main loops prior to a system shutdown, so that they
577  * won't interfere with the shutdown process (e.g. by holding a disk buf
578  * to cause sync to fail).  For each of these system processes, register
579  * shutdown_kproc() as a handler for one of shutdown events.
580  */
581 static int kproc_shutdown_wait = 60;
582 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
583     &kproc_shutdown_wait, 0, "");
584 
585 void
586 shutdown_kproc(void *arg, int howto)
587 {
588 	struct proc *p;
589 	int error;
590 
591 	if (panicstr)
592 		return;
593 
594 	p = (struct proc *)arg;
595 	printf("Waiting (max %d seconds) for system process `%s' to stop...",
596 	    kproc_shutdown_wait, p->p_comm);
597 	error = suspend_kproc(p, kproc_shutdown_wait * hz);
598 
599 	if (error == EWOULDBLOCK)
600 		printf("timed out\n");
601 	else
602 		printf("stopped\n");
603 }
604