xref: /freebsd/sys/kern/kern_shutdown.c (revision fcb560670601b2a4d87bb31d7531c8dcc37ee71b)
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_panic.h"
43 #include "opt_sched.h"
44 #include "opt_watchdog.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/bio.h>
49 #include <sys/buf.h>
50 #include <sys/conf.h>
51 #include <sys/cons.h>
52 #include <sys/eventhandler.h>
53 #include <sys/jail.h>
54 #include <sys/kdb.h>
55 #include <sys/kernel.h>
56 #include <sys/kerneldump.h>
57 #include <sys/kthread.h>
58 #include <sys/ktr.h>
59 #include <sys/malloc.h>
60 #include <sys/mount.h>
61 #include <sys/priv.h>
62 #include <sys/proc.h>
63 #include <sys/reboot.h>
64 #include <sys/resourcevar.h>
65 #include <sys/rwlock.h>
66 #include <sys/sched.h>
67 #include <sys/smp.h>
68 #include <sys/sysctl.h>
69 #include <sys/sysproto.h>
70 #include <sys/vnode.h>
71 #include <sys/watchdog.h>
72 
73 #include <ddb/ddb.h>
74 
75 #include <machine/cpu.h>
76 #include <machine/dump.h>
77 #include <machine/pcb.h>
78 #include <machine/smp.h>
79 
80 #include <security/mac/mac_framework.h>
81 
82 #include <vm/vm.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/vm_pager.h>
86 #include <vm/swap_pager.h>
87 
88 #include <sys/signalvar.h>
89 
90 #ifndef PANIC_REBOOT_WAIT_TIME
91 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
92 #endif
93 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
94 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
95     &panic_reboot_wait_time, 0,
96     "Seconds to wait before rebooting after a panic");
97 
98 /*
99  * Note that stdarg.h and the ANSI style va_start macro is used for both
100  * ANSI and traditional C compilers.
101  */
102 #include <machine/stdarg.h>
103 
104 #ifdef KDB
105 #ifdef KDB_UNATTENDED
106 int debugger_on_panic = 0;
107 #else
108 int debugger_on_panic = 1;
109 #endif
110 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
111     CTLFLAG_RWTUN | CTLFLAG_SECURE,
112     &debugger_on_panic, 0, "Run debugger on kernel panic");
113 
114 #ifdef KDB_TRACE
115 static int trace_on_panic = 1;
116 #else
117 static int trace_on_panic = 0;
118 #endif
119 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
120     CTLFLAG_RWTUN | CTLFLAG_SECURE,
121     &trace_on_panic, 0, "Print stack trace on kernel panic");
122 #endif /* KDB */
123 
124 static int sync_on_panic = 0;
125 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
126 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
127 
128 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
129     "Shutdown environment");
130 
131 #ifndef DIAGNOSTIC
132 static int show_busybufs;
133 #else
134 static int show_busybufs = 1;
135 #endif
136 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
137 	&show_busybufs, 0, "");
138 
139 /*
140  * Variable panicstr contains argument to first call to panic; used as flag
141  * to indicate that the kernel has already called panic.
142  */
143 const char *panicstr;
144 
145 int dumping;				/* system is dumping */
146 int rebooting;				/* system is rebooting */
147 static struct dumperinfo dumper;	/* our selected dumper */
148 
149 /* Context information for dump-debuggers. */
150 static struct pcb dumppcb;		/* Registers. */
151 lwpid_t dumptid;			/* Thread ID. */
152 
153 static void poweroff_wait(void *, int);
154 static void shutdown_halt(void *junk, int howto);
155 static void shutdown_panic(void *junk, int howto);
156 static void shutdown_reset(void *junk, int howto);
157 static void vpanic(const char *fmt, va_list ap) __dead2;
158 
159 /* register various local shutdown events */
160 static void
161 shutdown_conf(void *unused)
162 {
163 
164 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
165 	    SHUTDOWN_PRI_FIRST);
166 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
167 	    SHUTDOWN_PRI_LAST + 100);
168 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
169 	    SHUTDOWN_PRI_LAST + 100);
170 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
171 	    SHUTDOWN_PRI_LAST + 200);
172 }
173 
174 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
175 
176 /*
177  * The system call that results in a reboot.
178  */
179 /* ARGSUSED */
180 int
181 sys_reboot(struct thread *td, struct reboot_args *uap)
182 {
183 	int error;
184 
185 	error = 0;
186 #ifdef MAC
187 	error = mac_system_check_reboot(td->td_ucred, uap->opt);
188 #endif
189 	if (error == 0)
190 		error = priv_check(td, PRIV_REBOOT);
191 	if (error == 0) {
192 		mtx_lock(&Giant);
193 		kern_reboot(uap->opt);
194 		mtx_unlock(&Giant);
195 	}
196 	return (error);
197 }
198 
199 /*
200  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
201  */
202 void
203 shutdown_nice(int howto)
204 {
205 
206 	if (initproc != NULL) {
207 		/* Send a signal to init(8) and have it shutdown the world. */
208 		PROC_LOCK(initproc);
209 		if (howto & RB_POWEROFF)
210 			kern_psignal(initproc, SIGUSR2);
211 		else if (howto & RB_HALT)
212 			kern_psignal(initproc, SIGUSR1);
213 		else
214 			kern_psignal(initproc, SIGINT);
215 		PROC_UNLOCK(initproc);
216 	} else {
217 		/* No init(8) running, so simply reboot. */
218 		kern_reboot(howto | RB_NOSYNC);
219 	}
220 }
221 
222 static void
223 print_uptime(void)
224 {
225 	int f;
226 	struct timespec ts;
227 
228 	getnanouptime(&ts);
229 	printf("Uptime: ");
230 	f = 0;
231 	if (ts.tv_sec >= 86400) {
232 		printf("%ldd", (long)ts.tv_sec / 86400);
233 		ts.tv_sec %= 86400;
234 		f = 1;
235 	}
236 	if (f || ts.tv_sec >= 3600) {
237 		printf("%ldh", (long)ts.tv_sec / 3600);
238 		ts.tv_sec %= 3600;
239 		f = 1;
240 	}
241 	if (f || ts.tv_sec >= 60) {
242 		printf("%ldm", (long)ts.tv_sec / 60);
243 		ts.tv_sec %= 60;
244 		f = 1;
245 	}
246 	printf("%lds\n", (long)ts.tv_sec);
247 }
248 
249 int
250 doadump(boolean_t textdump)
251 {
252 	boolean_t coredump;
253 	int error;
254 
255 	error = 0;
256 	if (dumping)
257 		return (EBUSY);
258 	if (dumper.dumper == NULL)
259 		return (ENXIO);
260 
261 	savectx(&dumppcb);
262 	dumptid = curthread->td_tid;
263 	dumping++;
264 
265 	coredump = TRUE;
266 #ifdef DDB
267 	if (textdump && textdump_pending) {
268 		coredump = FALSE;
269 		textdump_dumpsys(&dumper);
270 	}
271 #endif
272 	if (coredump)
273 		error = dumpsys(&dumper);
274 
275 	dumping--;
276 	return (error);
277 }
278 
279 static int
280 isbufbusy(struct buf *bp)
281 {
282 	if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
283 	    BUF_ISLOCKED(bp)) ||
284 	    ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
285 		return (1);
286 	return (0);
287 }
288 
289 /*
290  * Shutdown the system cleanly to prepare for reboot, halt, or power off.
291  */
292 void
293 kern_reboot(int howto)
294 {
295 	static int first_buf_printf = 1;
296 	static int waittime = -1;
297 
298 #if defined(SMP)
299 	/*
300 	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
301 	 * systems don't shutdown properly (i.e., ACPI power off) if we
302 	 * run on another processor.
303 	 */
304 	if (!SCHEDULER_STOPPED()) {
305 		thread_lock(curthread);
306 		sched_bind(curthread, 0);
307 		thread_unlock(curthread);
308 		KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
309 	}
310 #endif
311 	/* We're in the process of rebooting. */
312 	rebooting = 1;
313 
314 	/* We are out of the debugger now. */
315 	kdb_active = 0;
316 
317 	/*
318 	 * Do any callouts that should be done BEFORE syncing the filesystems.
319 	 */
320 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
321 
322 	/*
323 	 * Now sync filesystems
324 	 */
325 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
326 		register struct buf *bp;
327 		int iter, nbusy, pbusy;
328 #ifndef PREEMPTION
329 		int subiter;
330 #endif
331 
332 		waittime = 0;
333 
334 		wdog_kern_pat(WD_LASTVAL);
335 		sys_sync(curthread, NULL);
336 
337 		/*
338 		 * With soft updates, some buffers that are
339 		 * written will be remarked as dirty until other
340 		 * buffers are written.
341 		 */
342 		for (iter = pbusy = 0; iter < 20; iter++) {
343 			nbusy = 0;
344 			for (bp = &buf[nbuf]; --bp >= buf; )
345 				if (isbufbusy(bp))
346 					nbusy++;
347 			if (nbusy == 0) {
348 				if (first_buf_printf)
349 					printf("All buffers synced.");
350 				break;
351 			}
352 			if (first_buf_printf) {
353 				printf("Syncing disks, buffers remaining... ");
354 				first_buf_printf = 0;
355 			}
356 			printf("%d ", nbusy);
357 			if (nbusy < pbusy)
358 				iter = 0;
359 			pbusy = nbusy;
360 
361 			wdog_kern_pat(WD_LASTVAL);
362 			sys_sync(curthread, NULL);
363 
364 #ifdef PREEMPTION
365 			/*
366 			 * Drop Giant and spin for a while to allow
367 			 * interrupt threads to run.
368 			 */
369 			DROP_GIANT();
370 			DELAY(50000 * iter);
371 			PICKUP_GIANT();
372 #else
373 			/*
374 			 * Drop Giant and context switch several times to
375 			 * allow interrupt threads to run.
376 			 */
377 			DROP_GIANT();
378 			for (subiter = 0; subiter < 50 * iter; subiter++) {
379 				thread_lock(curthread);
380 				mi_switch(SW_VOL, NULL);
381 				thread_unlock(curthread);
382 				DELAY(1000);
383 			}
384 			PICKUP_GIANT();
385 #endif
386 		}
387 		printf("\n");
388 		/*
389 		 * Count only busy local buffers to prevent forcing
390 		 * a fsck if we're just a client of a wedged NFS server
391 		 */
392 		nbusy = 0;
393 		for (bp = &buf[nbuf]; --bp >= buf; ) {
394 			if (isbufbusy(bp)) {
395 #if 0
396 /* XXX: This is bogus.  We should probably have a BO_REMOTE flag instead */
397 				if (bp->b_dev == NULL) {
398 					TAILQ_REMOVE(&mountlist,
399 					    bp->b_vp->v_mount, mnt_list);
400 					continue;
401 				}
402 #endif
403 				nbusy++;
404 				if (show_busybufs > 0) {
405 					printf(
406 	    "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
407 					    nbusy, bp, bp->b_vp, bp->b_flags,
408 					    (intmax_t)bp->b_blkno,
409 					    (intmax_t)bp->b_lblkno);
410 					BUF_LOCKPRINTINFO(bp);
411 					if (show_busybufs > 1)
412 						vn_printf(bp->b_vp,
413 						    "vnode content: ");
414 				}
415 			}
416 		}
417 		if (nbusy) {
418 			/*
419 			 * Failed to sync all blocks. Indicate this and don't
420 			 * unmount filesystems (thus forcing an fsck on reboot).
421 			 */
422 			printf("Giving up on %d buffers\n", nbusy);
423 			DELAY(5000000);	/* 5 seconds */
424 		} else {
425 			if (!first_buf_printf)
426 				printf("Final sync complete\n");
427 			/*
428 			 * Unmount filesystems
429 			 */
430 			if (panicstr == 0)
431 				vfs_unmountall();
432 		}
433 		swapoff_all();
434 		DELAY(100000);		/* wait for console output to finish */
435 	}
436 
437 	print_uptime();
438 
439 	cngrab();
440 
441 	/*
442 	 * Ok, now do things that assume all filesystem activity has
443 	 * been completed.
444 	 */
445 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
446 
447 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
448 		doadump(TRUE);
449 
450 	/* Now that we're going to really halt the system... */
451 	EVENTHANDLER_INVOKE(shutdown_final, howto);
452 
453 	for(;;) ;	/* safety against shutdown_reset not working */
454 	/* NOTREACHED */
455 }
456 
457 /*
458  * If the shutdown was a clean halt, behave accordingly.
459  */
460 static void
461 shutdown_halt(void *junk, int howto)
462 {
463 
464 	if (howto & RB_HALT) {
465 		printf("\n");
466 		printf("The operating system has halted.\n");
467 		printf("Please press any key to reboot.\n\n");
468 		switch (cngetc()) {
469 		case -1:		/* No console, just die */
470 			cpu_halt();
471 			/* NOTREACHED */
472 		default:
473 			howto &= ~RB_HALT;
474 			break;
475 		}
476 	}
477 }
478 
479 /*
480  * Check to see if the system paniced, pause and then reboot
481  * according to the specified delay.
482  */
483 static void
484 shutdown_panic(void *junk, int howto)
485 {
486 	int loop;
487 
488 	if (howto & RB_DUMP) {
489 		if (panic_reboot_wait_time != 0) {
490 			if (panic_reboot_wait_time != -1) {
491 				printf("Automatic reboot in %d seconds - "
492 				       "press a key on the console to abort\n",
493 					panic_reboot_wait_time);
494 				for (loop = panic_reboot_wait_time * 10;
495 				     loop > 0; --loop) {
496 					DELAY(1000 * 100); /* 1/10th second */
497 					/* Did user type a key? */
498 					if (cncheckc() != -1)
499 						break;
500 				}
501 				if (!loop)
502 					return;
503 			}
504 		} else { /* zero time specified - reboot NOW */
505 			return;
506 		}
507 		printf("--> Press a key on the console to reboot,\n");
508 		printf("--> or switch off the system now.\n");
509 		cngetc();
510 	}
511 }
512 
513 /*
514  * Everything done, now reset
515  */
516 static void
517 shutdown_reset(void *junk, int howto)
518 {
519 
520 	printf("Rebooting...\n");
521 	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
522 
523 	/*
524 	 * Acquiring smp_ipi_mtx here has a double effect:
525 	 * - it disables interrupts avoiding CPU0 preemption
526 	 *   by fast handlers (thus deadlocking  against other CPUs)
527 	 * - it avoids deadlocks against smp_rendezvous() or, more
528 	 *   generally, threads busy-waiting, with this spinlock held,
529 	 *   and waiting for responses by threads on other CPUs
530 	 *   (ie. smp_tlb_shootdown()).
531 	 *
532 	 * For the !SMP case it just needs to handle the former problem.
533 	 */
534 #ifdef SMP
535 	mtx_lock_spin(&smp_ipi_mtx);
536 #else
537 	spinlock_enter();
538 #endif
539 
540 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
541 	cpu_reset();
542 	/* NOTREACHED */ /* assuming reset worked */
543 }
544 
545 #if defined(WITNESS) || defined(INVARIANTS)
546 static int kassert_warn_only = 0;
547 #ifdef KDB
548 static int kassert_do_kdb = 0;
549 #endif
550 #ifdef KTR
551 static int kassert_do_ktr = 0;
552 #endif
553 static int kassert_do_log = 1;
554 static int kassert_log_pps_limit = 4;
555 static int kassert_log_mute_at = 0;
556 static int kassert_log_panic_at = 0;
557 static int kassert_warnings = 0;
558 
559 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
560 
561 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
562     &kassert_warn_only, 0,
563     "KASSERT triggers a panic (1) or just a warning (0)");
564 
565 #ifdef KDB
566 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
567     &kassert_do_kdb, 0, "KASSERT will enter the debugger");
568 #endif
569 
570 #ifdef KTR
571 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
572     &kassert_do_ktr, 0,
573     "KASSERT does a KTR, set this to the KTRMASK you want");
574 #endif
575 
576 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
577     &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
578 
579 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
580     &kassert_warnings, 0, "number of KASSERTs that have been triggered");
581 
582 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
583     &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
584 
585 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
586     &kassert_log_pps_limit, 0, "limit number of log messages per second");
587 
588 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
589     &kassert_log_mute_at, 0, "max number of KASSERTS to log");
590 
591 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
592 
593 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
594     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
595     kassert_sysctl_kassert, "I", "set to trigger a test kassert");
596 
597 static int
598 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
599 {
600 	int error, i;
601 
602 	error = sysctl_wire_old_buffer(req, sizeof(int));
603 	if (error == 0) {
604 		i = 0;
605 		error = sysctl_handle_int(oidp, &i, 0, req);
606 	}
607 	if (error != 0 || req->newptr == NULL)
608 		return (error);
609 	KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
610 	return (0);
611 }
612 
613 /*
614  * Called by KASSERT, this decides if we will panic
615  * or if we will log via printf and/or ktr.
616  */
617 void
618 kassert_panic(const char *fmt, ...)
619 {
620 	static char buf[256];
621 	va_list ap;
622 
623 	va_start(ap, fmt);
624 	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
625 	va_end(ap);
626 
627 	/*
628 	 * panic if we're not just warning, or if we've exceeded
629 	 * kassert_log_panic_at warnings.
630 	 */
631 	if (!kassert_warn_only ||
632 	    (kassert_log_panic_at > 0 &&
633 	     kassert_warnings >= kassert_log_panic_at)) {
634 		va_start(ap, fmt);
635 		vpanic(fmt, ap);
636 		/* NORETURN */
637 	}
638 #ifdef KTR
639 	if (kassert_do_ktr)
640 		CTR0(ktr_mask, buf);
641 #endif /* KTR */
642 	/*
643 	 * log if we've not yet met the mute limit.
644 	 */
645 	if (kassert_do_log &&
646 	    (kassert_log_mute_at == 0 ||
647 	     kassert_warnings < kassert_log_mute_at)) {
648 		static  struct timeval lasterr;
649 		static  int curerr;
650 
651 		if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
652 			printf("KASSERT failed: %s\n", buf);
653 			kdb_backtrace();
654 		}
655 	}
656 #ifdef KDB
657 	if (kassert_do_kdb) {
658 		kdb_enter(KDB_WHY_KASSERT, buf);
659 	}
660 #endif
661 	atomic_add_int(&kassert_warnings, 1);
662 }
663 #endif
664 
665 /*
666  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
667  * and then reboots.  If we are called twice, then we avoid trying to sync
668  * the disks as this often leads to recursive panics.
669  */
670 void
671 panic(const char *fmt, ...)
672 {
673 	va_list ap;
674 
675 	va_start(ap, fmt);
676 	vpanic(fmt, ap);
677 }
678 
679 static void
680 vpanic(const char *fmt, va_list ap)
681 {
682 #ifdef SMP
683 	cpuset_t other_cpus;
684 #endif
685 	struct thread *td = curthread;
686 	int bootopt, newpanic;
687 	static char buf[256];
688 
689 	spinlock_enter();
690 
691 #ifdef SMP
692 	/*
693 	 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
694 	 * concurrently entering panic.  Only the winner will proceed
695 	 * further.
696 	 */
697 	if (panicstr == NULL && !kdb_active) {
698 		other_cpus = all_cpus;
699 		CPU_CLR(PCPU_GET(cpuid), &other_cpus);
700 		stop_cpus_hard(other_cpus);
701 	}
702 
703 	/*
704 	 * We set stop_scheduler here and not in the block above,
705 	 * because we want to ensure that if panic has been called and
706 	 * stop_scheduler_on_panic is true, then stop_scheduler will
707 	 * always be set.  Even if panic has been entered from kdb.
708 	 */
709 	td->td_stopsched = 1;
710 #endif
711 
712 	bootopt = RB_AUTOBOOT;
713 	newpanic = 0;
714 	if (panicstr)
715 		bootopt |= RB_NOSYNC;
716 	else {
717 		bootopt |= RB_DUMP;
718 		panicstr = fmt;
719 		newpanic = 1;
720 	}
721 
722 	if (newpanic) {
723 		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
724 		panicstr = buf;
725 		cngrab();
726 		printf("panic: %s\n", buf);
727 	} else {
728 		printf("panic: ");
729 		vprintf(fmt, ap);
730 		printf("\n");
731 	}
732 #ifdef SMP
733 	printf("cpuid = %d\n", PCPU_GET(cpuid));
734 #endif
735 
736 #ifdef KDB
737 	if (newpanic && trace_on_panic)
738 		kdb_backtrace();
739 	if (debugger_on_panic)
740 		kdb_enter(KDB_WHY_PANIC, "panic");
741 #endif
742 	/*thread_lock(td); */
743 	td->td_flags |= TDF_INPANIC;
744 	/* thread_unlock(td); */
745 	if (!sync_on_panic)
746 		bootopt |= RB_NOSYNC;
747 	kern_reboot(bootopt);
748 }
749 
750 /*
751  * Support for poweroff delay.
752  *
753  * Please note that setting this delay too short might power off your machine
754  * before the write cache on your hard disk has been flushed, leading to
755  * soft-updates inconsistencies.
756  */
757 #ifndef POWEROFF_DELAY
758 # define POWEROFF_DELAY 5000
759 #endif
760 static int poweroff_delay = POWEROFF_DELAY;
761 
762 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
763     &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
764 
765 static void
766 poweroff_wait(void *junk, int howto)
767 {
768 
769 	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
770 		return;
771 	DELAY(poweroff_delay * 1000);
772 }
773 
774 /*
775  * Some system processes (e.g. syncer) need to be stopped at appropriate
776  * points in their main loops prior to a system shutdown, so that they
777  * won't interfere with the shutdown process (e.g. by holding a disk buf
778  * to cause sync to fail).  For each of these system processes, register
779  * shutdown_kproc() as a handler for one of shutdown events.
780  */
781 static int kproc_shutdown_wait = 60;
782 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
783     &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
784 
785 void
786 kproc_shutdown(void *arg, int howto)
787 {
788 	struct proc *p;
789 	int error;
790 
791 	if (panicstr)
792 		return;
793 
794 	p = (struct proc *)arg;
795 	printf("Waiting (max %d seconds) for system process `%s' to stop...",
796 	    kproc_shutdown_wait, p->p_comm);
797 	error = kproc_suspend(p, kproc_shutdown_wait * hz);
798 
799 	if (error == EWOULDBLOCK)
800 		printf("timed out\n");
801 	else
802 		printf("done\n");
803 }
804 
805 void
806 kthread_shutdown(void *arg, int howto)
807 {
808 	struct thread *td;
809 	int error;
810 
811 	if (panicstr)
812 		return;
813 
814 	td = (struct thread *)arg;
815 	printf("Waiting (max %d seconds) for system thread `%s' to stop...",
816 	    kproc_shutdown_wait, td->td_name);
817 	error = kthread_suspend(td, kproc_shutdown_wait * hz);
818 
819 	if (error == EWOULDBLOCK)
820 		printf("timed out\n");
821 	else
822 		printf("done\n");
823 }
824 
825 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
826 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
827     dumpdevname, 0, "Device for kernel dumps");
828 
829 /* Registration of dumpers */
830 int
831 set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
832 {
833 	size_t wantcopy;
834 	int error;
835 
836 	error = priv_check(td, PRIV_SETDUMPER);
837 	if (error != 0)
838 		return (error);
839 
840 	if (di == NULL) {
841 		bzero(&dumper, sizeof dumper);
842 		dumpdevname[0] = '\0';
843 		return (0);
844 	}
845 	if (dumper.dumper != NULL)
846 		return (EBUSY);
847 	dumper = *di;
848 	wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
849 	if (wantcopy >= sizeof(dumpdevname)) {
850 		printf("set_dumper: device name truncated from '%s' -> '%s'\n",
851 			devname, dumpdevname);
852 	}
853 	return (0);
854 }
855 
856 /* Call dumper with bounds checking. */
857 int
858 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
859     off_t offset, size_t length)
860 {
861 
862 	if (length != 0 && (offset < di->mediaoffset ||
863 	    offset - di->mediaoffset + length > di->mediasize)) {
864 		printf("Attempt to write outside dump device boundaries.\n"
865 	    "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
866 		    (intmax_t)offset, (intmax_t)di->mediaoffset,
867 		    (uintmax_t)length, (intmax_t)di->mediasize);
868 		return (ENOSPC);
869 	}
870 	return (di->dumper(di->priv, virtual, physical, offset, length));
871 }
872 
873 void
874 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
875     uint64_t dumplen, uint32_t blksz)
876 {
877 
878 	bzero(kdh, sizeof(*kdh));
879 	strncpy(kdh->magic, magic, sizeof(kdh->magic));
880 	strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
881 	kdh->version = htod32(KERNELDUMPVERSION);
882 	kdh->architectureversion = htod32(archver);
883 	kdh->dumplength = htod64(dumplen);
884 	kdh->dumptime = htod64(time_second);
885 	kdh->blocksize = htod32(blksz);
886 	strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
887 	strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
888 	if (panicstr != NULL)
889 		strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
890 	kdh->parity = kerneldump_parity(kdh);
891 }
892