xref: /freebsd/sys/kern/kern_shutdown.c (revision 9336e0699bda8a301cd2bfa37106b6ec5e32012e)
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  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)kern_shutdown.c	8.3 (Berkeley) 1/21/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_ddb.h"
41 #include "opt_kdb.h"
42 #include "opt_mac.h"
43 #include "opt_panic.h"
44 #include "opt_show_busybufs.h"
45 #include "opt_sched.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/bio.h>
50 #include <sys/buf.h>
51 #include <sys/conf.h>
52 #include <sys/cons.h>
53 #include <sys/eventhandler.h>
54 #include <sys/kdb.h>
55 #include <sys/kernel.h>
56 #include <sys/kthread.h>
57 #include <sys/malloc.h>
58 #include <sys/mount.h>
59 #include <sys/priv.h>
60 #include <sys/proc.h>
61 #include <sys/reboot.h>
62 #include <sys/resourcevar.h>
63 #include <sys/sched.h>
64 #include <sys/smp.h>		/* smp_active */
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 
68 #include <ddb/ddb.h>
69 
70 #include <machine/cpu.h>
71 #include <machine/pcb.h>
72 #include <machine/smp.h>
73 
74 #include <security/mac/mac_framework.h>
75 
76 #include <vm/vm.h>
77 #include <vm/vm_object.h>
78 #include <vm/vm_page.h>
79 #include <vm/vm_pager.h>
80 #include <vm/swap_pager.h>
81 
82 #include <sys/signalvar.h>
83 
84 #ifndef PANIC_REBOOT_WAIT_TIME
85 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
86 #endif
87 
88 /*
89  * Note that stdarg.h and the ANSI style va_start macro is used for both
90  * ANSI and traditional C compilers.
91  */
92 #include <machine/stdarg.h>
93 
94 #ifdef KDB
95 #ifdef KDB_UNATTENDED
96 int debugger_on_panic = 0;
97 #else
98 int debugger_on_panic = 1;
99 #endif
100 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
101 	&debugger_on_panic, 0, "Run debugger on kernel panic");
102 
103 #ifdef KDB_TRACE
104 int trace_on_panic = 1;
105 #else
106 int trace_on_panic = 0;
107 #endif
108 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
109 	&trace_on_panic, 0, "Print stack trace on kernel panic");
110 #endif /* KDB */
111 
112 int sync_on_panic = 0;
113 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
114 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
115 
116 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
117 
118 /*
119  * Variable panicstr contains argument to first call to panic; used as flag
120  * to indicate that the kernel has already called panic.
121  */
122 const char *panicstr;
123 
124 int dumping;				/* system is dumping */
125 int rebooting;				/* system is rebooting */
126 static struct dumperinfo dumper;	/* our selected dumper */
127 
128 /* Context information for dump-debuggers. */
129 static struct pcb dumppcb;		/* Registers. */
130 static lwpid_t dumptid;			/* Thread ID. */
131 
132 static void boot(int) __dead2;
133 static void poweroff_wait(void *, int);
134 static void shutdown_halt(void *junk, int howto);
135 static void shutdown_panic(void *junk, int howto);
136 static void shutdown_reset(void *junk, int howto);
137 
138 /* register various local shutdown events */
139 static void
140 shutdown_conf(void *unused)
141 {
142 
143 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
144 	    SHUTDOWN_PRI_FIRST);
145 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
146 	    SHUTDOWN_PRI_LAST + 100);
147 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
148 	    SHUTDOWN_PRI_LAST + 100);
149 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
150 	    SHUTDOWN_PRI_LAST + 200);
151 }
152 
153 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
154 
155 /*
156  * The system call that results in a reboot.
157  */
158 /* ARGSUSED */
159 int
160 reboot(struct thread *td, struct reboot_args *uap)
161 {
162 	int error;
163 
164 	error = 0;
165 #ifdef MAC
166 	error = mac_system_check_reboot(td->td_ucred, uap->opt);
167 #endif
168 	if (error == 0)
169 		error = priv_check(td, PRIV_REBOOT);
170 	if (error == 0) {
171 		mtx_lock(&Giant);
172 		boot(uap->opt);
173 		mtx_unlock(&Giant);
174 	}
175 	return (error);
176 }
177 
178 /*
179  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
180  */
181 static int shutdown_howto = 0;
182 
183 void
184 shutdown_nice(int howto)
185 {
186 
187 	shutdown_howto = howto;
188 
189 	/* Send a signal to init(8) and have it shutdown the world */
190 	if (initproc != NULL) {
191 		PROC_LOCK(initproc);
192 		psignal(initproc, SIGINT);
193 		PROC_UNLOCK(initproc);
194 	} else {
195 		/* No init(8) running, so simply reboot */
196 		boot(RB_NOSYNC);
197 	}
198 	return;
199 }
200 static int	waittime = -1;
201 
202 static void
203 print_uptime(void)
204 {
205 	int f;
206 	struct timespec ts;
207 
208 	getnanouptime(&ts);
209 	printf("Uptime: ");
210 	f = 0;
211 	if (ts.tv_sec >= 86400) {
212 		printf("%ldd", (long)ts.tv_sec / 86400);
213 		ts.tv_sec %= 86400;
214 		f = 1;
215 	}
216 	if (f || ts.tv_sec >= 3600) {
217 		printf("%ldh", (long)ts.tv_sec / 3600);
218 		ts.tv_sec %= 3600;
219 		f = 1;
220 	}
221 	if (f || ts.tv_sec >= 60) {
222 		printf("%ldm", (long)ts.tv_sec / 60);
223 		ts.tv_sec %= 60;
224 		f = 1;
225 	}
226 	printf("%lds\n", (long)ts.tv_sec);
227 }
228 
229 static void
230 doadump(void)
231 {
232 
233 	/*
234 	 * Sometimes people have to call this from the kernel debugger.
235 	 * (if 'panic' can not dump)
236 	 * Give them a clue as to why they can't dump.
237 	 */
238 	if (dumper.dumper == NULL) {
239 		printf("Cannot dump. No dump device defined.\n");
240 		return;
241 	}
242 
243 	savectx(&dumppcb);
244 	dumptid = curthread->td_tid;
245 	dumping++;
246 #ifdef DDB
247 	if (textdump_pending)
248 		textdump_dumpsys(&dumper);
249 	else
250 #endif
251 		dumpsys(&dumper);
252 }
253 
254 static int
255 isbufbusy(struct buf *bp)
256 {
257 	if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
258 	    BUF_REFCNT(bp) > 0) ||
259 	    ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
260 		return (1);
261 	return (0);
262 }
263 
264 /*
265  * Shutdown the system cleanly to prepare for reboot, halt, or power off.
266  */
267 static void
268 boot(int howto)
269 {
270 	static int first_buf_printf = 1;
271 
272 #if defined(SMP)
273 	/*
274 	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
275 	 * systems don't shutdown properly (i.e., ACPI power off) if we
276 	 * run on another processor.
277 	 */
278 	thread_lock(curthread);
279 	sched_bind(curthread, 0);
280 	thread_unlock(curthread);
281 	KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
282 #endif
283 	/* We're in the process of rebooting. */
284 	rebooting = 1;
285 
286 	/* collect extra flags that shutdown_nice might have set */
287 	howto |= shutdown_howto;
288 
289 	/* We are out of the debugger now. */
290 	kdb_active = 0;
291 
292 	/*
293 	 * Do any callouts that should be done BEFORE syncing the filesystems.
294 	 */
295 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
296 
297 	/*
298 	 * Now sync filesystems
299 	 */
300 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
301 		register struct buf *bp;
302 		int iter, nbusy, pbusy;
303 #ifndef PREEMPTION
304 		int subiter;
305 #endif
306 
307 		waittime = 0;
308 
309 		sync(curthread, NULL);
310 
311 		/*
312 		 * With soft updates, some buffers that are
313 		 * written will be remarked as dirty until other
314 		 * buffers are written.
315 		 */
316 		for (iter = pbusy = 0; iter < 20; iter++) {
317 			nbusy = 0;
318 			for (bp = &buf[nbuf]; --bp >= buf; )
319 				if (isbufbusy(bp))
320 					nbusy++;
321 			if (nbusy == 0) {
322 				if (first_buf_printf)
323 					printf("All buffers synced.");
324 				break;
325 			}
326 			if (first_buf_printf) {
327 				printf("Syncing disks, buffers remaining... ");
328 				first_buf_printf = 0;
329 			}
330 			printf("%d ", nbusy);
331 			if (nbusy < pbusy)
332 				iter = 0;
333 			pbusy = nbusy;
334 			sync(curthread, NULL);
335 
336 #ifdef PREEMPTION
337 			/*
338 			 * Drop Giant and spin for a while to allow
339 			 * interrupt threads to run.
340 			 */
341 			DROP_GIANT();
342 			DELAY(50000 * iter);
343 			PICKUP_GIANT();
344 #else
345 			/*
346 			 * Drop Giant and context switch several times to
347 			 * allow interrupt threads to run.
348 			 */
349 			DROP_GIANT();
350 			for (subiter = 0; subiter < 50 * iter; subiter++) {
351 				thread_lock(curthread);
352 				mi_switch(SW_VOL, NULL);
353 				thread_unlock(curthread);
354 				DELAY(1000);
355 			}
356 			PICKUP_GIANT();
357 #endif
358 		}
359 		printf("\n");
360 		/*
361 		 * Count only busy local buffers to prevent forcing
362 		 * a fsck if we're just a client of a wedged NFS server
363 		 */
364 		nbusy = 0;
365 		for (bp = &buf[nbuf]; --bp >= buf; ) {
366 			if (isbufbusy(bp)) {
367 #if 0
368 /* XXX: This is bogus.  We should probably have a BO_REMOTE flag instead */
369 				if (bp->b_dev == NULL) {
370 					TAILQ_REMOVE(&mountlist,
371 					    bp->b_vp->v_mount, mnt_list);
372 					continue;
373 				}
374 #endif
375 				nbusy++;
376 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
377 				printf(
378 			    "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n",
379 				    nbusy, bp->b_bufobj,
380 				    bp->b_flags, (long)bp->b_blkno,
381 				    (long)bp->b_lblkno);
382 #endif
383 			}
384 		}
385 		if (nbusy) {
386 			/*
387 			 * Failed to sync all blocks. Indicate this and don't
388 			 * unmount filesystems (thus forcing an fsck on reboot).
389 			 */
390 			printf("Giving up on %d buffers\n", nbusy);
391 			DELAY(5000000);	/* 5 seconds */
392 		} else {
393 			if (!first_buf_printf)
394 				printf("Final sync complete\n");
395 			/*
396 			 * Unmount filesystems
397 			 */
398 			if (panicstr == 0)
399 				vfs_unmountall();
400 		}
401 		swapoff_all();
402 		DELAY(100000);		/* wait for console output to finish */
403 	}
404 
405 	print_uptime();
406 
407 	/*
408 	 * Ok, now do things that assume all filesystem activity has
409 	 * been completed.
410 	 */
411 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
412 
413 	/* XXX This doesn't disable interrupts any more.  Reconsider? */
414 	splhigh();
415 
416 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
417 		doadump();
418 
419 	/* Now that we're going to really halt the system... */
420 	EVENTHANDLER_INVOKE(shutdown_final, howto);
421 
422 	for(;;) ;	/* safety against shutdown_reset not working */
423 	/* NOTREACHED */
424 }
425 
426 /*
427  * If the shutdown was a clean halt, behave accordingly.
428  */
429 static void
430 shutdown_halt(void *junk, int howto)
431 {
432 
433 	if (howto & RB_HALT) {
434 		printf("\n");
435 		printf("The operating system has halted.\n");
436 		printf("Please press any key to reboot.\n\n");
437 		switch (cngetc()) {
438 		case -1:		/* No console, just die */
439 			cpu_halt();
440 			/* NOTREACHED */
441 		default:
442 			howto &= ~RB_HALT;
443 			break;
444 		}
445 	}
446 }
447 
448 /*
449  * Check to see if the system paniced, pause and then reboot
450  * according to the specified delay.
451  */
452 static void
453 shutdown_panic(void *junk, int howto)
454 {
455 	int loop;
456 
457 	if (howto & RB_DUMP) {
458 		if (PANIC_REBOOT_WAIT_TIME != 0) {
459 			if (PANIC_REBOOT_WAIT_TIME != -1) {
460 				printf("Automatic reboot in %d seconds - "
461 				       "press a key on the console to abort\n",
462 					PANIC_REBOOT_WAIT_TIME);
463 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
464 				     loop > 0; --loop) {
465 					DELAY(1000 * 100); /* 1/10th second */
466 					/* Did user type a key? */
467 					if (cncheckc() != -1)
468 						break;
469 				}
470 				if (!loop)
471 					return;
472 			}
473 		} else { /* zero time specified - reboot NOW */
474 			return;
475 		}
476 		printf("--> Press a key on the console to reboot,\n");
477 		printf("--> or switch off the system now.\n");
478 		cngetc();
479 	}
480 }
481 
482 /*
483  * Everything done, now reset
484  */
485 static void
486 shutdown_reset(void *junk, int howto)
487 {
488 
489 	printf("Rebooting...\n");
490 	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
491 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
492 	cpu_reset();
493 	/* NOTREACHED */ /* assuming reset worked */
494 }
495 
496 #ifdef SMP
497 static u_int panic_cpu = NOCPU;
498 #endif
499 
500 /*
501  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
502  * and then reboots.  If we are called twice, then we avoid trying to sync
503  * the disks as this often leads to recursive panics.
504  */
505 void
506 panic(const char *fmt, ...)
507 {
508 	struct thread *td = curthread;
509 	int bootopt, newpanic;
510 	va_list ap;
511 	static char buf[256];
512 
513 #ifdef SMP
514 	/*
515 	 * We don't want multiple CPU's to panic at the same time, so we
516 	 * use panic_cpu as a simple spinlock.  We have to keep checking
517 	 * panic_cpu if we are spinning in case the panic on the first
518 	 * CPU is canceled.
519 	 */
520 	if (panic_cpu != PCPU_GET(cpuid))
521 		while (atomic_cmpset_int(&panic_cpu, NOCPU,
522 		    PCPU_GET(cpuid)) == 0)
523 			while (panic_cpu != NOCPU)
524 				; /* nothing */
525 #endif
526 
527 	bootopt = RB_AUTOBOOT | RB_DUMP;
528 	newpanic = 0;
529 	if (panicstr)
530 		bootopt |= RB_NOSYNC;
531 	else {
532 		panicstr = fmt;
533 		newpanic = 1;
534 	}
535 
536 	va_start(ap, fmt);
537 	if (newpanic) {
538 		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
539 		panicstr = buf;
540 		printf("panic: %s\n", buf);
541 	} else {
542 		printf("panic: ");
543 		vprintf(fmt, ap);
544 		printf("\n");
545 	}
546 	va_end(ap);
547 #ifdef SMP
548 	printf("cpuid = %d\n", PCPU_GET(cpuid));
549 #endif
550 
551 #ifdef KDB
552 	if (newpanic && trace_on_panic)
553 		kdb_backtrace();
554 	if (debugger_on_panic)
555 		kdb_enter(KDB_WHY_PANIC, "panic");
556 #ifdef RESTARTABLE_PANICS
557 	/* See if the user aborted the panic, in which case we continue. */
558 	if (panicstr == NULL) {
559 #ifdef SMP
560 		atomic_store_rel_int(&panic_cpu, NOCPU);
561 #endif
562 		return;
563 	}
564 #endif
565 #endif
566 	/*thread_lock(td); */
567 	td->td_flags |= TDF_INPANIC;
568 	/* thread_unlock(td); */
569 	if (!sync_on_panic)
570 		bootopt |= RB_NOSYNC;
571 	boot(bootopt);
572 }
573 
574 /*
575  * Support for poweroff delay.
576  */
577 #ifndef POWEROFF_DELAY
578 # define POWEROFF_DELAY 5000
579 #endif
580 static int poweroff_delay = POWEROFF_DELAY;
581 
582 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
583 	&poweroff_delay, 0, "");
584 
585 static void
586 poweroff_wait(void *junk, int howto)
587 {
588 
589 	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
590 		return;
591 	DELAY(poweroff_delay * 1000);
592 }
593 
594 /*
595  * Some system processes (e.g. syncer) need to be stopped at appropriate
596  * points in their main loops prior to a system shutdown, so that they
597  * won't interfere with the shutdown process (e.g. by holding a disk buf
598  * to cause sync to fail).  For each of these system processes, register
599  * shutdown_kproc() as a handler for one of shutdown events.
600  */
601 static int kproc_shutdown_wait = 60;
602 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
603     &kproc_shutdown_wait, 0, "");
604 
605 void
606 kproc_shutdown(void *arg, int howto)
607 {
608 	struct proc *p;
609 	char procname[MAXCOMLEN + 1];
610 	int error;
611 
612 	if (panicstr)
613 		return;
614 
615 	p = (struct proc *)arg;
616 	strlcpy(procname, p->p_comm, sizeof(procname));
617 	printf("Waiting (max %d seconds) for system process `%s' to stop...",
618 	    kproc_shutdown_wait, procname);
619 	error = kproc_suspend(p, kproc_shutdown_wait * hz);
620 
621 	if (error == EWOULDBLOCK)
622 		printf("timed out\n");
623 	else
624 		printf("done\n");
625 }
626 
627 void
628 kthread_shutdown(void *arg, int howto)
629 {
630 	struct thread *td;
631 	char procname[MAXCOMLEN + 1];
632 	int error;
633 
634 	if (panicstr)
635 		return;
636 
637 	td = (struct thread *)arg;
638 	strlcpy(procname, td->td_name, sizeof(procname));
639 	printf("Waiting (max %d seconds) for system thread `%s' to stop...",
640 	    kproc_shutdown_wait, procname);
641 	error = kthread_suspend(td, kproc_shutdown_wait * hz);
642 
643 	if (error == EWOULDBLOCK)
644 		printf("timed out\n");
645 	else
646 		printf("done\n");
647 }
648 
649 /* Registration of dumpers */
650 int
651 set_dumper(struct dumperinfo *di)
652 {
653 
654 	if (di == NULL) {
655 		bzero(&dumper, sizeof dumper);
656 		return (0);
657 	}
658 	if (dumper.dumper != NULL)
659 		return (EBUSY);
660 	dumper = *di;
661 	return (0);
662 }
663 
664 #if defined(__powerpc__)
665 void
666 dumpsys(struct dumperinfo *di __unused)
667 {
668 
669 	printf("Kernel dumps not implemented on this architecture\n");
670 }
671 #endif
672