xref: /freebsd/sys/kern/kern_shutdown.c (revision a9fcb51fbb8d2c5b47a35bcae5b4d5fb771c5985)
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/filedesc.h>
54 #include <sys/jail.h>
55 #include <sys/kdb.h>
56 #include <sys/kernel.h>
57 #include <sys/kerneldump.h>
58 #include <sys/kthread.h>
59 #include <sys/ktr.h>
60 #include <sys/malloc.h>
61 #include <sys/mount.h>
62 #include <sys/priv.h>
63 #include <sys/proc.h>
64 #include <sys/reboot.h>
65 #include <sys/resourcevar.h>
66 #include <sys/rwlock.h>
67 #include <sys/sched.h>
68 #include <sys/smp.h>
69 #include <sys/sysctl.h>
70 #include <sys/sysproto.h>
71 #include <sys/vnode.h>
72 #include <sys/watchdog.h>
73 
74 #include <ddb/ddb.h>
75 
76 #include <machine/cpu.h>
77 #include <machine/dump.h>
78 #include <machine/pcb.h>
79 #include <machine/smp.h>
80 
81 #include <security/mac/mac_framework.h>
82 
83 #include <vm/vm.h>
84 #include <vm/vm_object.h>
85 #include <vm/vm_page.h>
86 #include <vm/vm_pager.h>
87 #include <vm/swap_pager.h>
88 
89 #include <sys/signalvar.h>
90 
91 #ifndef PANIC_REBOOT_WAIT_TIME
92 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
93 #endif
94 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
95 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
96     &panic_reboot_wait_time, 0,
97     "Seconds to wait before rebooting after a panic");
98 
99 /*
100  * Note that stdarg.h and the ANSI style va_start macro is used for both
101  * ANSI and traditional C compilers.
102  */
103 #include <machine/stdarg.h>
104 
105 #ifdef KDB
106 #ifdef KDB_UNATTENDED
107 int debugger_on_panic = 0;
108 #else
109 int debugger_on_panic = 1;
110 #endif
111 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
112     CTLFLAG_RWTUN | CTLFLAG_SECURE,
113     &debugger_on_panic, 0, "Run debugger on kernel panic");
114 
115 #ifdef KDB_TRACE
116 static int trace_on_panic = 1;
117 #else
118 static int trace_on_panic = 0;
119 #endif
120 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
121     CTLFLAG_RWTUN | CTLFLAG_SECURE,
122     &trace_on_panic, 0, "Print stack trace on kernel panic");
123 #endif /* KDB */
124 
125 static int sync_on_panic = 0;
126 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
127 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
128 
129 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
130     "Shutdown environment");
131 
132 #ifndef DIAGNOSTIC
133 static int show_busybufs;
134 #else
135 static int show_busybufs = 1;
136 #endif
137 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
138 	&show_busybufs, 0, "");
139 
140 /*
141  * Variable panicstr contains argument to first call to panic; used as flag
142  * to indicate that the kernel has already called panic.
143  */
144 const char *panicstr;
145 
146 int dumping;				/* system is dumping */
147 int rebooting;				/* system is rebooting */
148 static struct dumperinfo dumper;	/* our selected dumper */
149 
150 /* Context information for dump-debuggers. */
151 static struct pcb dumppcb;		/* Registers. */
152 lwpid_t dumptid;			/* Thread ID. */
153 
154 static struct cdevsw reroot_cdevsw = {
155      .d_version = D_VERSION,
156      .d_name    = "reroot",
157 };
158 
159 static void poweroff_wait(void *, int);
160 static void shutdown_halt(void *junk, int howto);
161 static void shutdown_panic(void *junk, int howto);
162 static void shutdown_reset(void *junk, int howto);
163 static int kern_reroot(void);
164 
165 /* register various local shutdown events */
166 static void
167 shutdown_conf(void *unused)
168 {
169 
170 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
171 	    SHUTDOWN_PRI_FIRST);
172 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
173 	    SHUTDOWN_PRI_LAST + 100);
174 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
175 	    SHUTDOWN_PRI_LAST + 100);
176 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
177 	    SHUTDOWN_PRI_LAST + 200);
178 }
179 
180 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
181 
182 /*
183  * The only reason this exists is to create the /dev/reroot/ directory,
184  * used by reroot code in init(8) as a mountpoint for tmpfs.
185  */
186 static void
187 reroot_conf(void *unused)
188 {
189 	int error;
190 	struct cdev *cdev;
191 
192 	error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev,
193 	    &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot");
194 	if (error != 0) {
195 		printf("%s: failed to create device node, error %d",
196 		    __func__, error);
197 	}
198 }
199 
200 SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL);
201 
202 /*
203  * The system call that results in a reboot.
204  */
205 /* ARGSUSED */
206 int
207 sys_reboot(struct thread *td, struct reboot_args *uap)
208 {
209 	int error;
210 
211 	error = 0;
212 #ifdef MAC
213 	error = mac_system_check_reboot(td->td_ucred, uap->opt);
214 #endif
215 	if (error == 0)
216 		error = priv_check(td, PRIV_REBOOT);
217 	if (error == 0) {
218 		if (uap->opt & RB_REROOT) {
219 			error = kern_reroot();
220 		} else {
221 			mtx_lock(&Giant);
222 			kern_reboot(uap->opt);
223 			mtx_unlock(&Giant);
224 		}
225 	}
226 	return (error);
227 }
228 
229 /*
230  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
231  */
232 void
233 shutdown_nice(int howto)
234 {
235 
236 	if (initproc != NULL) {
237 		/* Send a signal to init(8) and have it shutdown the world. */
238 		PROC_LOCK(initproc);
239 		if (howto & RB_POWEROFF)
240 			kern_psignal(initproc, SIGUSR2);
241 		else if (howto & RB_HALT)
242 			kern_psignal(initproc, SIGUSR1);
243 		else
244 			kern_psignal(initproc, SIGINT);
245 		PROC_UNLOCK(initproc);
246 	} else {
247 		/* No init(8) running, so simply reboot. */
248 		kern_reboot(howto | RB_NOSYNC);
249 	}
250 }
251 
252 static void
253 print_uptime(void)
254 {
255 	int f;
256 	struct timespec ts;
257 
258 	getnanouptime(&ts);
259 	printf("Uptime: ");
260 	f = 0;
261 	if (ts.tv_sec >= 86400) {
262 		printf("%ldd", (long)ts.tv_sec / 86400);
263 		ts.tv_sec %= 86400;
264 		f = 1;
265 	}
266 	if (f || ts.tv_sec >= 3600) {
267 		printf("%ldh", (long)ts.tv_sec / 3600);
268 		ts.tv_sec %= 3600;
269 		f = 1;
270 	}
271 	if (f || ts.tv_sec >= 60) {
272 		printf("%ldm", (long)ts.tv_sec / 60);
273 		ts.tv_sec %= 60;
274 		f = 1;
275 	}
276 	printf("%lds\n", (long)ts.tv_sec);
277 }
278 
279 int
280 doadump(boolean_t textdump)
281 {
282 	boolean_t coredump;
283 	int error;
284 
285 	error = 0;
286 	if (dumping)
287 		return (EBUSY);
288 	if (dumper.dumper == NULL)
289 		return (ENXIO);
290 
291 	savectx(&dumppcb);
292 	dumptid = curthread->td_tid;
293 	dumping++;
294 
295 	coredump = TRUE;
296 #ifdef DDB
297 	if (textdump && textdump_pending) {
298 		coredump = FALSE;
299 		textdump_dumpsys(&dumper);
300 	}
301 #endif
302 	if (coredump)
303 		error = dumpsys(&dumper);
304 
305 	dumping--;
306 	return (error);
307 }
308 
309 /*
310  * Shutdown the system cleanly to prepare for reboot, halt, or power off.
311  */
312 void
313 kern_reboot(int howto)
314 {
315 	static int once = 0;
316 
317 #if defined(SMP)
318 	/*
319 	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
320 	 * systems don't shutdown properly (i.e., ACPI power off) if we
321 	 * run on another processor.
322 	 */
323 	if (!SCHEDULER_STOPPED()) {
324 		thread_lock(curthread);
325 		sched_bind(curthread, 0);
326 		thread_unlock(curthread);
327 		KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
328 	}
329 #endif
330 	/* We're in the process of rebooting. */
331 	rebooting = 1;
332 
333 	/* We are out of the debugger now. */
334 	kdb_active = 0;
335 
336 	/*
337 	 * Do any callouts that should be done BEFORE syncing the filesystems.
338 	 */
339 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
340 
341 	/*
342 	 * Now sync filesystems
343 	 */
344 	if (!cold && (howto & RB_NOSYNC) == 0 && once == 0) {
345 		once = 1;
346 		bufshutdown(show_busybufs);
347 	}
348 
349 	print_uptime();
350 
351 	cngrab();
352 
353 	/*
354 	 * Ok, now do things that assume all filesystem activity has
355 	 * been completed.
356 	 */
357 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
358 
359 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
360 		doadump(TRUE);
361 
362 	/* Now that we're going to really halt the system... */
363 	EVENTHANDLER_INVOKE(shutdown_final, howto);
364 
365 	for(;;) ;	/* safety against shutdown_reset not working */
366 	/* NOTREACHED */
367 }
368 
369 /*
370  * The system call that results in changing the rootfs.
371  */
372 static int
373 kern_reroot(void)
374 {
375 	struct vnode *oldrootvnode, *vp;
376 	struct mount *mp, *devmp;
377 	int error;
378 
379 	if (curproc != initproc)
380 		return (EPERM);
381 
382 	/*
383 	 * Mark the filesystem containing currently-running executable
384 	 * (the temporary copy of init(8)) busy.
385 	 */
386 	vp = curproc->p_textvp;
387 	error = vn_lock(vp, LK_SHARED);
388 	if (error != 0)
389 		return (error);
390 	mp = vp->v_mount;
391 	error = vfs_busy(mp, MBF_NOWAIT);
392 	if (error != 0) {
393 		vfs_ref(mp);
394 		VOP_UNLOCK(vp, 0);
395 		error = vfs_busy(mp, 0);
396 		vn_lock(vp, LK_SHARED | LK_RETRY);
397 		vfs_rel(mp);
398 		if (error != 0) {
399 			VOP_UNLOCK(vp, 0);
400 			return (ENOENT);
401 		}
402 		if (vp->v_iflag & VI_DOOMED) {
403 			VOP_UNLOCK(vp, 0);
404 			vfs_unbusy(mp);
405 			return (ENOENT);
406 		}
407 	}
408 	VOP_UNLOCK(vp, 0);
409 
410 	/*
411 	 * Remove the filesystem containing currently-running executable
412 	 * from the mount list, to prevent it from being unmounted
413 	 * by vfs_unmountall(), and to avoid confusing vfs_mountroot().
414 	 *
415 	 * Also preserve /dev - forcibly unmounting it could cause driver
416 	 * reinitialization.
417 	 */
418 
419 	vfs_ref(rootdevmp);
420 	devmp = rootdevmp;
421 	rootdevmp = NULL;
422 
423 	mtx_lock(&mountlist_mtx);
424 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
425 	TAILQ_REMOVE(&mountlist, devmp, mnt_list);
426 	mtx_unlock(&mountlist_mtx);
427 
428 	oldrootvnode = rootvnode;
429 
430 	/*
431 	 * Unmount everything except for the two filesystems preserved above.
432 	 */
433 	vfs_unmountall();
434 
435 	/*
436 	 * Add /dev back; vfs_mountroot() will move it into its new place.
437 	 */
438 	mtx_lock(&mountlist_mtx);
439 	TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list);
440 	mtx_unlock(&mountlist_mtx);
441 	rootdevmp = devmp;
442 	vfs_rel(rootdevmp);
443 
444 	/*
445 	 * Mount the new rootfs.
446 	 */
447 	vfs_mountroot();
448 
449 	/*
450 	 * Update all references to the old rootvnode.
451 	 */
452 	mountcheckdirs(oldrootvnode, rootvnode);
453 
454 	/*
455 	 * Add the temporary filesystem back and unbusy it.
456 	 */
457 	mtx_lock(&mountlist_mtx);
458 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
459 	mtx_unlock(&mountlist_mtx);
460 	vfs_unbusy(mp);
461 
462 	return (0);
463 }
464 
465 /*
466  * If the shutdown was a clean halt, behave accordingly.
467  */
468 static void
469 shutdown_halt(void *junk, int howto)
470 {
471 
472 	if (howto & RB_HALT) {
473 		printf("\n");
474 		printf("The operating system has halted.\n");
475 		printf("Please press any key to reboot.\n\n");
476 		switch (cngetc()) {
477 		case -1:		/* No console, just die */
478 			cpu_halt();
479 			/* NOTREACHED */
480 		default:
481 			howto &= ~RB_HALT;
482 			break;
483 		}
484 	}
485 }
486 
487 /*
488  * Check to see if the system paniced, pause and then reboot
489  * according to the specified delay.
490  */
491 static void
492 shutdown_panic(void *junk, int howto)
493 {
494 	int loop;
495 
496 	if (howto & RB_DUMP) {
497 		if (panic_reboot_wait_time != 0) {
498 			if (panic_reboot_wait_time != -1) {
499 				printf("Automatic reboot in %d seconds - "
500 				       "press a key on the console to abort\n",
501 					panic_reboot_wait_time);
502 				for (loop = panic_reboot_wait_time * 10;
503 				     loop > 0; --loop) {
504 					DELAY(1000 * 100); /* 1/10th second */
505 					/* Did user type a key? */
506 					if (cncheckc() != -1)
507 						break;
508 				}
509 				if (!loop)
510 					return;
511 			}
512 		} else { /* zero time specified - reboot NOW */
513 			return;
514 		}
515 		printf("--> Press a key on the console to reboot,\n");
516 		printf("--> or switch off the system now.\n");
517 		cngetc();
518 	}
519 }
520 
521 /*
522  * Everything done, now reset
523  */
524 static void
525 shutdown_reset(void *junk, int howto)
526 {
527 
528 	printf("Rebooting...\n");
529 	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
530 
531 	/*
532 	 * Acquiring smp_ipi_mtx here has a double effect:
533 	 * - it disables interrupts avoiding CPU0 preemption
534 	 *   by fast handlers (thus deadlocking  against other CPUs)
535 	 * - it avoids deadlocks against smp_rendezvous() or, more
536 	 *   generally, threads busy-waiting, with this spinlock held,
537 	 *   and waiting for responses by threads on other CPUs
538 	 *   (ie. smp_tlb_shootdown()).
539 	 *
540 	 * For the !SMP case it just needs to handle the former problem.
541 	 */
542 #ifdef SMP
543 	mtx_lock_spin(&smp_ipi_mtx);
544 #else
545 	spinlock_enter();
546 #endif
547 
548 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
549 	cpu_reset();
550 	/* NOTREACHED */ /* assuming reset worked */
551 }
552 
553 #if defined(WITNESS) || defined(INVARIANTS)
554 static int kassert_warn_only = 0;
555 #ifdef KDB
556 static int kassert_do_kdb = 0;
557 #endif
558 #ifdef KTR
559 static int kassert_do_ktr = 0;
560 #endif
561 static int kassert_do_log = 1;
562 static int kassert_log_pps_limit = 4;
563 static int kassert_log_mute_at = 0;
564 static int kassert_log_panic_at = 0;
565 static int kassert_warnings = 0;
566 
567 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
568 
569 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
570     &kassert_warn_only, 0,
571     "KASSERT triggers a panic (1) or just a warning (0)");
572 
573 #ifdef KDB
574 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
575     &kassert_do_kdb, 0, "KASSERT will enter the debugger");
576 #endif
577 
578 #ifdef KTR
579 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
580     &kassert_do_ktr, 0,
581     "KASSERT does a KTR, set this to the KTRMASK you want");
582 #endif
583 
584 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
585     &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
586 
587 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
588     &kassert_warnings, 0, "number of KASSERTs that have been triggered");
589 
590 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
591     &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
592 
593 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
594     &kassert_log_pps_limit, 0, "limit number of log messages per second");
595 
596 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
597     &kassert_log_mute_at, 0, "max number of KASSERTS to log");
598 
599 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
600 
601 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
602     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
603     kassert_sysctl_kassert, "I", "set to trigger a test kassert");
604 
605 static int
606 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
607 {
608 	int error, i;
609 
610 	error = sysctl_wire_old_buffer(req, sizeof(int));
611 	if (error == 0) {
612 		i = 0;
613 		error = sysctl_handle_int(oidp, &i, 0, req);
614 	}
615 	if (error != 0 || req->newptr == NULL)
616 		return (error);
617 	KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
618 	return (0);
619 }
620 
621 /*
622  * Called by KASSERT, this decides if we will panic
623  * or if we will log via printf and/or ktr.
624  */
625 void
626 kassert_panic(const char *fmt, ...)
627 {
628 	static char buf[256];
629 	va_list ap;
630 
631 	va_start(ap, fmt);
632 	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
633 	va_end(ap);
634 
635 	/*
636 	 * panic if we're not just warning, or if we've exceeded
637 	 * kassert_log_panic_at warnings.
638 	 */
639 	if (!kassert_warn_only ||
640 	    (kassert_log_panic_at > 0 &&
641 	     kassert_warnings >= kassert_log_panic_at)) {
642 		va_start(ap, fmt);
643 		vpanic(fmt, ap);
644 		/* NORETURN */
645 	}
646 #ifdef KTR
647 	if (kassert_do_ktr)
648 		CTR0(ktr_mask, buf);
649 #endif /* KTR */
650 	/*
651 	 * log if we've not yet met the mute limit.
652 	 */
653 	if (kassert_do_log &&
654 	    (kassert_log_mute_at == 0 ||
655 	     kassert_warnings < kassert_log_mute_at)) {
656 		static  struct timeval lasterr;
657 		static  int curerr;
658 
659 		if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
660 			printf("KASSERT failed: %s\n", buf);
661 			kdb_backtrace();
662 		}
663 	}
664 #ifdef KDB
665 	if (kassert_do_kdb) {
666 		kdb_enter(KDB_WHY_KASSERT, buf);
667 	}
668 #endif
669 	atomic_add_int(&kassert_warnings, 1);
670 }
671 #endif
672 
673 /*
674  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
675  * and then reboots.  If we are called twice, then we avoid trying to sync
676  * the disks as this often leads to recursive panics.
677  */
678 void
679 panic(const char *fmt, ...)
680 {
681 	va_list ap;
682 
683 	va_start(ap, fmt);
684 	vpanic(fmt, ap);
685 }
686 
687 void
688 vpanic(const char *fmt, va_list ap)
689 {
690 #ifdef SMP
691 	cpuset_t other_cpus;
692 #endif
693 	struct thread *td = curthread;
694 	int bootopt, newpanic;
695 	static char buf[256];
696 
697 	spinlock_enter();
698 
699 #ifdef SMP
700 	/*
701 	 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
702 	 * concurrently entering panic.  Only the winner will proceed
703 	 * further.
704 	 */
705 	if (panicstr == NULL && !kdb_active) {
706 		other_cpus = all_cpus;
707 		CPU_CLR(PCPU_GET(cpuid), &other_cpus);
708 		stop_cpus_hard(other_cpus);
709 	}
710 
711 	/*
712 	 * Ensure that the scheduler is stopped while panicking, even if panic
713 	 * has been entered from kdb.
714 	 */
715 	td->td_stopsched = 1;
716 #endif
717 
718 	bootopt = RB_AUTOBOOT;
719 	newpanic = 0;
720 	if (panicstr)
721 		bootopt |= RB_NOSYNC;
722 	else {
723 		bootopt |= RB_DUMP;
724 		panicstr = fmt;
725 		newpanic = 1;
726 	}
727 
728 	if (newpanic) {
729 		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
730 		panicstr = buf;
731 		cngrab();
732 		printf("panic: %s\n", buf);
733 	} else {
734 		printf("panic: ");
735 		vprintf(fmt, ap);
736 		printf("\n");
737 	}
738 #ifdef SMP
739 	printf("cpuid = %d\n", PCPU_GET(cpuid));
740 #endif
741 
742 #ifdef KDB
743 	if (newpanic && trace_on_panic)
744 		kdb_backtrace();
745 	if (debugger_on_panic)
746 		kdb_enter(KDB_WHY_PANIC, "panic");
747 #endif
748 	/*thread_lock(td); */
749 	td->td_flags |= TDF_INPANIC;
750 	/* thread_unlock(td); */
751 	if (!sync_on_panic)
752 		bootopt |= RB_NOSYNC;
753 	kern_reboot(bootopt);
754 }
755 
756 /*
757  * Support for poweroff delay.
758  *
759  * Please note that setting this delay too short might power off your machine
760  * before the write cache on your hard disk has been flushed, leading to
761  * soft-updates inconsistencies.
762  */
763 #ifndef POWEROFF_DELAY
764 # define POWEROFF_DELAY 5000
765 #endif
766 static int poweroff_delay = POWEROFF_DELAY;
767 
768 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
769     &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
770 
771 static void
772 poweroff_wait(void *junk, int howto)
773 {
774 
775 	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
776 		return;
777 	DELAY(poweroff_delay * 1000);
778 }
779 
780 /*
781  * Some system processes (e.g. syncer) need to be stopped at appropriate
782  * points in their main loops prior to a system shutdown, so that they
783  * won't interfere with the shutdown process (e.g. by holding a disk buf
784  * to cause sync to fail).  For each of these system processes, register
785  * shutdown_kproc() as a handler for one of shutdown events.
786  */
787 static int kproc_shutdown_wait = 60;
788 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
789     &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
790 
791 void
792 kproc_shutdown(void *arg, int howto)
793 {
794 	struct proc *p;
795 	int error;
796 
797 	if (panicstr)
798 		return;
799 
800 	p = (struct proc *)arg;
801 	printf("Waiting (max %d seconds) for system process `%s' to stop...",
802 	    kproc_shutdown_wait, p->p_comm);
803 	error = kproc_suspend(p, kproc_shutdown_wait * hz);
804 
805 	if (error == EWOULDBLOCK)
806 		printf("timed out\n");
807 	else
808 		printf("done\n");
809 }
810 
811 void
812 kthread_shutdown(void *arg, int howto)
813 {
814 	struct thread *td;
815 	int error;
816 
817 	if (panicstr)
818 		return;
819 
820 	td = (struct thread *)arg;
821 	printf("Waiting (max %d seconds) for system thread `%s' to stop...",
822 	    kproc_shutdown_wait, td->td_name);
823 	error = kthread_suspend(td, kproc_shutdown_wait * hz);
824 
825 	if (error == EWOULDBLOCK)
826 		printf("timed out\n");
827 	else
828 		printf("done\n");
829 }
830 
831 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
832 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
833     dumpdevname, 0, "Device for kernel dumps");
834 
835 /* Registration of dumpers */
836 int
837 set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
838 {
839 	size_t wantcopy;
840 	int error;
841 
842 	error = priv_check(td, PRIV_SETDUMPER);
843 	if (error != 0)
844 		return (error);
845 
846 	if (di == NULL) {
847 		bzero(&dumper, sizeof dumper);
848 		dumpdevname[0] = '\0';
849 		return (0);
850 	}
851 	if (dumper.dumper != NULL)
852 		return (EBUSY);
853 	dumper = *di;
854 	wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
855 	if (wantcopy >= sizeof(dumpdevname)) {
856 		printf("set_dumper: device name truncated from '%s' -> '%s'\n",
857 			devname, dumpdevname);
858 	}
859 	return (0);
860 }
861 
862 /* Call dumper with bounds checking. */
863 int
864 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
865     off_t offset, size_t length)
866 {
867 
868 	if (length != 0 && (offset < di->mediaoffset ||
869 	    offset - di->mediaoffset + length > di->mediasize)) {
870 		printf("Attempt to write outside dump device boundaries.\n"
871 	    "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
872 		    (intmax_t)offset, (intmax_t)di->mediaoffset,
873 		    (uintmax_t)length, (intmax_t)di->mediasize);
874 		return (ENOSPC);
875 	}
876 	return (di->dumper(di->priv, virtual, physical, offset, length));
877 }
878 
879 void
880 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
881     uint64_t dumplen, uint32_t blksz)
882 {
883 
884 	bzero(kdh, sizeof(*kdh));
885 	strlcpy(kdh->magic, magic, sizeof(kdh->magic));
886 	strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
887 	kdh->version = htod32(KERNELDUMPVERSION);
888 	kdh->architectureversion = htod32(archver);
889 	kdh->dumplength = htod64(dumplen);
890 	kdh->dumptime = htod64(time_second);
891 	kdh->blocksize = htod32(blksz);
892 	strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
893 	strlcpy(kdh->versionstring, version, sizeof(kdh->versionstring));
894 	if (panicstr != NULL)
895 		strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
896 	kdh->parity = kerneldump_parity(kdh);
897 }
898