xref: /freebsd/sys/kern/kern_shutdown.c (revision a3e8fd0b7f663db7eafff527d5c3ca3bcfa8a537)
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_ddb_trace.h"
44 #include "opt_ddb_unattended.h"
45 #include "opt_hw_wdog.h"
46 #include "opt_panic.h"
47 #include "opt_show_busybufs.h"
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/bio.h>
52 #include <sys/buf.h>
53 #include <sys/conf.h>
54 #include <sys/cons.h>
55 #include <sys/disklabel.h>
56 #include <sys/eventhandler.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/malloc.h>
60 #include <sys/mount.h>
61 #include <sys/proc.h>
62 #include <sys/reboot.h>
63 #include <sys/resourcevar.h>
64 #include <sys/smp.h>		/* smp_active */
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/vnode.h>
68 
69 #include <machine/pcb.h>
70 #include <machine/md_var.h>
71 #include <machine/smp.h>
72 
73 #include <sys/signalvar.h>
74 #ifdef DDB
75 #include <ddb/ddb.h>
76 #endif
77 
78 #ifndef PANIC_REBOOT_WAIT_TIME
79 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
80 #endif
81 
82 /*
83  * Note that stdarg.h and the ANSI style va_start macro is used for both
84  * ANSI and traditional C compilers.
85  */
86 #include <machine/stdarg.h>
87 
88 #ifdef DDB
89 #ifdef DDB_UNATTENDED
90 int debugger_on_panic = 0;
91 #else
92 int debugger_on_panic = 1;
93 #endif
94 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
95 	&debugger_on_panic, 0, "Run debugger on kernel panic");
96 
97 #ifdef DDB_TRACE
98 int trace_on_panic = 1;
99 #else
100 int trace_on_panic = 0;
101 #endif
102 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
103 	&trace_on_panic, 0, "Print stack trace on kernel panic");
104 #endif
105 
106 int sync_on_panic = 1;
107 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
108 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
109 
110 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
111 
112 #ifdef	HW_WDOG
113 /*
114  * If there is a hardware watchdog, point this at the function needed to
115  * hold it off.
116  * It's needed when the kernel needs to do some lengthy operations.
117  * e.g. in wd.c when dumping core.. It's most annoying to have
118  * your precious core-dump only half written because the wdog kicked in.
119  */
120 watchdog_tickle_fn wdog_tickler = NULL;
121 #endif	/* HW_WDOG */
122 
123 /*
124  * Variable panicstr contains argument to first call to panic; used as flag
125  * to indicate that the kernel has already called panic.
126  */
127 const char *panicstr;
128 
129 int dumping;				/* system is dumping */
130 static struct dumperinfo dumper;	/* our selected dumper */
131 static struct pcb dumppcb;		/* "You Are Here" sign for dump-debuggers */
132 
133 static void boot(int) __dead2;
134 static void poweroff_wait(void *, int);
135 static void shutdown_halt(void *junk, int howto);
136 static void shutdown_panic(void *junk, int howto);
137 static void shutdown_reset(void *junk, int howto);
138 
139 /* register various local shutdown events */
140 static void
141 shutdown_conf(void *unused)
142 {
143 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
144 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
145 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
146 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
147 }
148 
149 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
150 
151 /*
152  * The system call that results in a reboot
153  *
154  * MPSAFE
155  */
156 /* ARGSUSED */
157 int
158 reboot(struct thread *td, struct reboot_args *uap)
159 {
160 	int error;
161 
162 	mtx_lock(&Giant);
163 	if ((error = suser(td)) == 0)
164 		boot(uap->opt);
165 	mtx_unlock(&Giant);
166 	return (error);
167 }
168 
169 /*
170  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
171  */
172 static int shutdown_howto = 0;
173 
174 void
175 shutdown_nice(int howto)
176 {
177 	shutdown_howto = howto;
178 
179 	/* Send a signal to init(8) and have it shutdown the world */
180 	if (initproc != NULL) {
181 		PROC_LOCK(initproc);
182 		psignal(initproc, SIGINT);
183 		PROC_UNLOCK(initproc);
184 	} else {
185 		/* No init(8) running, so simply reboot */
186 		boot(RB_NOSYNC);
187 	}
188 	return;
189 }
190 static int	waittime = -1;
191 
192 static void
193 print_uptime(void)
194 {
195 	int f;
196 	struct timespec ts;
197 
198 	getnanouptime(&ts);
199 	printf("Uptime: ");
200 	f = 0;
201 	if (ts.tv_sec >= 86400) {
202 		printf("%ldd", (long)ts.tv_sec / 86400);
203 		ts.tv_sec %= 86400;
204 		f = 1;
205 	}
206 	if (f || ts.tv_sec >= 3600) {
207 		printf("%ldh", (long)ts.tv_sec / 3600);
208 		ts.tv_sec %= 3600;
209 		f = 1;
210 	}
211 	if (f || ts.tv_sec >= 60) {
212 		printf("%ldm", (long)ts.tv_sec / 60);
213 		ts.tv_sec %= 60;
214 		f = 1;
215 	}
216 	printf("%lds\n", (long)ts.tv_sec);
217 }
218 
219 static void
220 doadump(void)
221 {
222 	savectx(&dumppcb);
223 	dumping++;
224 	dumpsys(&dumper);
225 }
226 
227 /*
228  *  Go through the rigmarole of shutting down..
229  * this used to be in machdep.c but I'll be dammned if I could see
230  * anything machine dependant in it.
231  */
232 static void
233 boot(int howto)
234 {
235 
236 	/* collect extra flags that shutdown_nice might have set */
237 	howto |= shutdown_howto;
238 
239 #ifdef DDB
240 	/* We are out of the debugger now. */
241 	db_active = 0;
242 #endif
243 
244 #ifdef SMP
245 	if (smp_active)
246 		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
247 #endif
248 	/*
249 	 * Do any callouts that should be done BEFORE syncing the filesystems.
250 	 */
251 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
252 
253 	/*
254 	 * Now sync filesystems
255 	 */
256 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
257 		register struct buf *bp;
258 		int iter, nbusy, pbusy;
259 		int subiter;
260 
261 		waittime = 0;
262 		printf("\nsyncing disks... ");
263 
264 		sync(&thread0, NULL);
265 
266 		/*
267 		 * With soft updates, some buffers that are
268 		 * written will be remarked as dirty until other
269 		 * buffers are written.
270 		 */
271 		for (iter = pbusy = 0; iter < 20; iter++) {
272 			nbusy = 0;
273 			for (bp = &buf[nbuf]; --bp >= buf; ) {
274 				if ((bp->b_flags & B_INVAL) == 0 &&
275 				    BUF_REFCNT(bp) > 0) {
276 					nbusy++;
277 				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
278 						== B_DELWRI) {
279 					/* bawrite(bp);*/
280 					nbusy++;
281 				}
282 			}
283 			if (nbusy == 0)
284 				break;
285 			printf("%d ", nbusy);
286 			if (nbusy < pbusy)
287 				iter = 0;
288 			pbusy = nbusy;
289 			sync(&thread0, NULL);
290  			if (curthread != NULL) {
291 				DROP_GIANT();
292    				for (subiter = 0; subiter < 50 * iter; subiter++) {
293      					mtx_lock_spin(&sched_lock);
294 					curthread->td_proc->p_stats->p_ru.ru_nvcsw++;
295      					mi_switch(); /* Allow interrupt threads to run */
296      					mtx_unlock_spin(&sched_lock);
297      					DELAY(1000);
298    				}
299 				PICKUP_GIANT();
300  			} else
301 			DELAY(50000 * iter);
302 		}
303 		printf("\n");
304 		/*
305 		 * Count only busy local buffers to prevent forcing
306 		 * a fsck if we're just a client of a wedged NFS server
307 		 */
308 		nbusy = 0;
309 		for (bp = &buf[nbuf]; --bp >= buf; ) {
310 			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
311 			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
312 				if (bp->b_dev == NODEV) {
313 					TAILQ_REMOVE(&mountlist,
314 					    bp->b_vp->v_mount, mnt_list);
315 					continue;
316 				}
317 				nbusy++;
318 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
319 				printf(
320 			    "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
321 				    nbusy, devtoname(bp->b_dev),
322 				    bp->b_flags, (long)bp->b_blkno,
323 				    (long)bp->b_lblkno);
324 #endif
325 			}
326 		}
327 		if (nbusy) {
328 			/*
329 			 * Failed to sync all blocks. Indicate this and don't
330 			 * unmount filesystems (thus forcing an fsck on reboot).
331 			 */
332 			printf("giving up on %d buffers\n", nbusy);
333 			DELAY(5000000);	/* 5 seconds */
334 		} else {
335 			printf("done\n");
336 			/*
337 			 * Unmount filesystems
338 			 */
339 			if (panicstr == 0)
340 				vfs_unmountall();
341 		}
342 		DELAY(100000);		/* wait for console output to finish */
343 	}
344 
345 	print_uptime();
346 
347 	/*
348 	 * Ok, now do things that assume all filesystem activity has
349 	 * been completed.
350 	 */
351 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
352 	splhigh();
353 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP &&
354 	    !cold && dumper.dumper != NULL && !dumping)
355 		doadump();
356 
357 	/* Now that we're going to really halt the system... */
358 	EVENTHANDLER_INVOKE(shutdown_final, howto);
359 
360 	for(;;) ;	/* safety against shutdown_reset not working */
361 	/* NOTREACHED */
362 }
363 
364 /*
365  * If the shutdown was a clean halt, behave accordingly.
366  */
367 static void
368 shutdown_halt(void *junk, int howto)
369 {
370 	if (howto & RB_HALT) {
371 		printf("\n");
372 		printf("The operating system has halted.\n");
373 		printf("Please press any key to reboot.\n\n");
374 		switch (cngetc()) {
375 		case -1:		/* No console, just die */
376 			cpu_halt();
377 			/* NOTREACHED */
378 		default:
379 			howto &= ~RB_HALT;
380 			break;
381 		}
382 	}
383 }
384 
385 /*
386  * Check to see if the system paniced, pause and then reboot
387  * according to the specified delay.
388  */
389 static void
390 shutdown_panic(void *junk, int howto)
391 {
392 	int loop;
393 
394 	if (howto & RB_DUMP) {
395 		if (PANIC_REBOOT_WAIT_TIME != 0) {
396 			if (PANIC_REBOOT_WAIT_TIME != -1) {
397 				printf("Automatic reboot in %d seconds - "
398 				       "press a key on the console to abort\n",
399 					PANIC_REBOOT_WAIT_TIME);
400 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
401 				     loop > 0; --loop) {
402 					DELAY(1000 * 100); /* 1/10th second */
403 					/* Did user type a key? */
404 					if (cncheckc() != -1)
405 						break;
406 				}
407 				if (!loop)
408 					return;
409 			}
410 		} else { /* zero time specified - reboot NOW */
411 			return;
412 		}
413 		printf("--> Press a key on the console to reboot,\n");
414 		printf("--> or switch off the system now.\n");
415 		cngetc();
416 	}
417 }
418 
419 /*
420  * Everything done, now reset
421  */
422 static void
423 shutdown_reset(void *junk, int howto)
424 {
425 	printf("Rebooting...\n");
426 	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
427 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
428 	cpu_reset();
429 	/* NOTREACHED */ /* assuming reset worked */
430 }
431 
432 #ifdef SMP
433 static u_int panic_cpu = NOCPU;
434 #endif
435 
436 /*
437  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
438  * and then reboots.  If we are called twice, then we avoid trying to sync
439  * the disks as this often leads to recursive panics.
440  *
441  * MPSAFE
442  */
443 void
444 panic(const char *fmt, ...)
445 {
446 	struct thread *td = curthread;
447 	int bootopt, newpanic;
448 	va_list ap;
449 	static char buf[256];
450 
451 #ifdef SMP
452 	/*
453 	 * We don't want multiple CPU's to panic at the same time, so we
454 	 * use panic_cpu as a simple spinlock.  We have to keep checking
455 	 * panic_cpu if we are spinning in case the panic on the first
456 	 * CPU is canceled.
457 	 */
458 	if (panic_cpu != PCPU_GET(cpuid))
459 		while (atomic_cmpset_int(&panic_cpu, NOCPU,
460 		    PCPU_GET(cpuid)) == 0)
461 			while (panic_cpu != NOCPU)
462 				; /* nothing */
463 #endif
464 
465 	bootopt = RB_AUTOBOOT | RB_DUMP;
466 	newpanic = 0;
467 	if (panicstr)
468 		bootopt |= RB_NOSYNC;
469 	else {
470 		panicstr = fmt;
471 		newpanic = 1;
472 	}
473 
474 	va_start(ap, fmt);
475 	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
476 	if (panicstr == fmt)
477 		panicstr = buf;
478 	va_end(ap);
479 	printf("panic: %s\n", buf);
480 #ifdef SMP
481 	/* two separate prints in case of an unmapped page and trap */
482 	printf("cpuid = %d; ", PCPU_GET(cpuid));
483 #ifdef APIC_IO
484 	printf("lapic.id = %08x\n", lapic.id);
485 #else
486 	printf("\n");
487 #endif
488 #endif
489 
490 #if defined(DDB)
491 	if (newpanic && trace_on_panic)
492 		db_print_backtrace();
493 	if (debugger_on_panic)
494 		Debugger ("panic");
495 #ifdef RESTARTABLE_PANICS
496 	/* See if the user aborted the panic, in which case we continue. */
497 	if (panicstr == NULL) {
498 #ifdef SMP
499 		atomic_store_rel_int(&panic_cpu, NOCPU);
500 #endif
501 		return;
502 	}
503 #endif
504 #endif
505 	td->td_flags |= TDF_INPANIC;
506 	if (!sync_on_panic)
507 		bootopt |= RB_NOSYNC;
508 	boot(bootopt);
509 }
510 
511 /*
512  * Support for poweroff delay.
513  */
514 #ifndef POWEROFF_DELAY
515 # define POWEROFF_DELAY 5000
516 #endif
517 static int poweroff_delay = POWEROFF_DELAY;
518 
519 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
520 	&poweroff_delay, 0, "");
521 
522 static void
523 poweroff_wait(void *junk, int howto)
524 {
525 	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
526 		return;
527 	DELAY(poweroff_delay * 1000);
528 }
529 
530 /*
531  * Some system processes (e.g. syncer) need to be stopped at appropriate
532  * points in their main loops prior to a system shutdown, so that they
533  * won't interfere with the shutdown process (e.g. by holding a disk buf
534  * to cause sync to fail).  For each of these system processes, register
535  * shutdown_kproc() as a handler for one of shutdown events.
536  */
537 static int kproc_shutdown_wait = 60;
538 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
539     &kproc_shutdown_wait, 0, "");
540 
541 void
542 kproc_shutdown(void *arg, int howto)
543 {
544 	struct proc *p;
545 	int error;
546 
547 	if (panicstr)
548 		return;
549 
550 	p = (struct proc *)arg;
551 	printf("Waiting (max %d seconds) for system process `%s' to stop...",
552 	    kproc_shutdown_wait, p->p_comm);
553 	error = kthread_suspend(p, kproc_shutdown_wait * hz);
554 
555 	if (error == EWOULDBLOCK)
556 		printf("timed out\n");
557 	else
558 		printf("stopped\n");
559 }
560 
561 /* Registration of dumpers */
562 int
563 set_dumper(struct dumperinfo *di)
564 {
565 	if (di == NULL) {
566 		bzero(&dumper, sizeof dumper);
567 		return (0);
568 	}
569 	if (dumper.dumper != NULL)
570 		return (EBUSY);
571 	dumper = *di;
572 	return (0);
573 }
574 
575 #if defined(__powerpc__)
576 void
577 dumpsys(struct dumperinfo *di __unused)
578 {
579 
580 	printf("Kernel dumps not implemented on this architecture\n");
581 }
582 #endif
583