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: releng/11.0/sys/kern/kern_shutdown.c 302349 2016-07-05 18:34:34Z glebius $");
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
91static MALLOC_DEFINE(M_DUMPER, "dumper", "dumper block buffer");
92
93#ifndef PANIC_REBOOT_WAIT_TIME
94#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
95#endif
96static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
97SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
98    &panic_reboot_wait_time, 0,
99    "Seconds to wait before rebooting after a panic");
100
101/*
102 * Note that stdarg.h and the ANSI style va_start macro is used for both
103 * ANSI and traditional C compilers.
104 */
105#include <machine/stdarg.h>
106
107#ifdef KDB
108#ifdef KDB_UNATTENDED
109int debugger_on_panic = 0;
110#else
111int debugger_on_panic = 1;
112#endif
113SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
114    CTLFLAG_RWTUN | CTLFLAG_SECURE,
115    &debugger_on_panic, 0, "Run debugger on kernel panic");
116
117#ifdef KDB_TRACE
118static int trace_on_panic = 1;
119#else
120static int trace_on_panic = 0;
121#endif
122SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
123    CTLFLAG_RWTUN | CTLFLAG_SECURE,
124    &trace_on_panic, 0, "Print stack trace on kernel panic");
125#endif /* KDB */
126
127static int sync_on_panic = 0;
128SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
129	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
130
131static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
132    "Shutdown environment");
133
134#ifndef DIAGNOSTIC
135static int show_busybufs;
136#else
137static int show_busybufs = 1;
138#endif
139SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
140	&show_busybufs, 0, "");
141
142int suspend_blocked = 0;
143SYSCTL_INT(_kern, OID_AUTO, suspend_blocked, CTLFLAG_RW,
144	&suspend_blocked, 0, "Block suspend due to a pending shutdown");
145
146/*
147 * Variable panicstr contains argument to first call to panic; used as flag
148 * to indicate that the kernel has already called panic.
149 */
150const char *panicstr;
151
152int dumping;				/* system is dumping */
153int rebooting;				/* system is rebooting */
154static struct dumperinfo dumper;	/* our selected dumper */
155
156/* Context information for dump-debuggers. */
157static struct pcb dumppcb;		/* Registers. */
158lwpid_t dumptid;			/* Thread ID. */
159
160static struct cdevsw reroot_cdevsw = {
161     .d_version = D_VERSION,
162     .d_name    = "reroot",
163};
164
165static void poweroff_wait(void *, int);
166static void shutdown_halt(void *junk, int howto);
167static void shutdown_panic(void *junk, int howto);
168static void shutdown_reset(void *junk, int howto);
169static int kern_reroot(void);
170
171/* register various local shutdown events */
172static void
173shutdown_conf(void *unused)
174{
175
176	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
177	    SHUTDOWN_PRI_FIRST);
178	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
179	    SHUTDOWN_PRI_LAST + 100);
180	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
181	    SHUTDOWN_PRI_LAST + 100);
182	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
183	    SHUTDOWN_PRI_LAST + 200);
184}
185
186SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
187
188/*
189 * The only reason this exists is to create the /dev/reroot/ directory,
190 * used by reroot code in init(8) as a mountpoint for tmpfs.
191 */
192static void
193reroot_conf(void *unused)
194{
195	int error;
196	struct cdev *cdev;
197
198	error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev,
199	    &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot");
200	if (error != 0) {
201		printf("%s: failed to create device node, error %d",
202		    __func__, error);
203	}
204}
205
206SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL);
207
208/*
209 * The system call that results in a reboot.
210 */
211/* ARGSUSED */
212int
213sys_reboot(struct thread *td, struct reboot_args *uap)
214{
215	int error;
216
217	error = 0;
218#ifdef MAC
219	error = mac_system_check_reboot(td->td_ucred, uap->opt);
220#endif
221	if (error == 0)
222		error = priv_check(td, PRIV_REBOOT);
223	if (error == 0) {
224		if (uap->opt & RB_REROOT) {
225			error = kern_reroot();
226		} else {
227			mtx_lock(&Giant);
228			kern_reboot(uap->opt);
229			mtx_unlock(&Giant);
230		}
231	}
232	return (error);
233}
234
235/*
236 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
237 */
238void
239shutdown_nice(int howto)
240{
241
242	if (initproc != NULL) {
243		/* Send a signal to init(8) and have it shutdown the world. */
244		PROC_LOCK(initproc);
245		if (howto & RB_POWEROFF)
246			kern_psignal(initproc, SIGUSR2);
247		else if (howto & RB_HALT)
248			kern_psignal(initproc, SIGUSR1);
249		else
250			kern_psignal(initproc, SIGINT);
251		PROC_UNLOCK(initproc);
252	} else {
253		/* No init(8) running, so simply reboot. */
254		kern_reboot(howto | RB_NOSYNC);
255	}
256}
257
258static void
259print_uptime(void)
260{
261	int f;
262	struct timespec ts;
263
264	getnanouptime(&ts);
265	printf("Uptime: ");
266	f = 0;
267	if (ts.tv_sec >= 86400) {
268		printf("%ldd", (long)ts.tv_sec / 86400);
269		ts.tv_sec %= 86400;
270		f = 1;
271	}
272	if (f || ts.tv_sec >= 3600) {
273		printf("%ldh", (long)ts.tv_sec / 3600);
274		ts.tv_sec %= 3600;
275		f = 1;
276	}
277	if (f || ts.tv_sec >= 60) {
278		printf("%ldm", (long)ts.tv_sec / 60);
279		ts.tv_sec %= 60;
280		f = 1;
281	}
282	printf("%lds\n", (long)ts.tv_sec);
283}
284
285int
286doadump(boolean_t textdump)
287{
288	boolean_t coredump;
289	int error;
290
291	error = 0;
292	if (dumping)
293		return (EBUSY);
294	if (dumper.dumper == NULL)
295		return (ENXIO);
296
297	savectx(&dumppcb);
298	dumptid = curthread->td_tid;
299	dumping++;
300
301	coredump = TRUE;
302#ifdef DDB
303	if (textdump && textdump_pending) {
304		coredump = FALSE;
305		textdump_dumpsys(&dumper);
306	}
307#endif
308	if (coredump)
309		error = dumpsys(&dumper);
310
311	dumping--;
312	return (error);
313}
314
315/*
316 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
317 */
318void
319kern_reboot(int howto)
320{
321	static int once = 0;
322
323#if defined(SMP)
324	/*
325	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
326	 * systems don't shutdown properly (i.e., ACPI power off) if we
327	 * run on another processor.
328	 */
329	if (!SCHEDULER_STOPPED()) {
330		thread_lock(curthread);
331		sched_bind(curthread, 0);
332		thread_unlock(curthread);
333		KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
334	}
335#endif
336	/* We're in the process of rebooting. */
337	rebooting = 1;
338
339	/* We are out of the debugger now. */
340	kdb_active = 0;
341
342	/*
343	 * Do any callouts that should be done BEFORE syncing the filesystems.
344	 */
345	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
346
347	/*
348	 * Now sync filesystems
349	 */
350	if (!cold && (howto & RB_NOSYNC) == 0 && once == 0) {
351		once = 1;
352		bufshutdown(show_busybufs);
353	}
354
355	print_uptime();
356
357	cngrab();
358
359	/*
360	 * Ok, now do things that assume all filesystem activity has
361	 * been completed.
362	 */
363	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
364
365	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
366		doadump(TRUE);
367
368	/* Now that we're going to really halt the system... */
369	EVENTHANDLER_INVOKE(shutdown_final, howto);
370
371	for(;;) ;	/* safety against shutdown_reset not working */
372	/* NOTREACHED */
373}
374
375/*
376 * The system call that results in changing the rootfs.
377 */
378static int
379kern_reroot(void)
380{
381	struct vnode *oldrootvnode, *vp;
382	struct mount *mp, *devmp;
383	int error;
384
385	if (curproc != initproc)
386		return (EPERM);
387
388	/*
389	 * Mark the filesystem containing currently-running executable
390	 * (the temporary copy of init(8)) busy.
391	 */
392	vp = curproc->p_textvp;
393	error = vn_lock(vp, LK_SHARED);
394	if (error != 0)
395		return (error);
396	mp = vp->v_mount;
397	error = vfs_busy(mp, MBF_NOWAIT);
398	if (error != 0) {
399		vfs_ref(mp);
400		VOP_UNLOCK(vp, 0);
401		error = vfs_busy(mp, 0);
402		vn_lock(vp, LK_SHARED | LK_RETRY);
403		vfs_rel(mp);
404		if (error != 0) {
405			VOP_UNLOCK(vp, 0);
406			return (ENOENT);
407		}
408		if (vp->v_iflag & VI_DOOMED) {
409			VOP_UNLOCK(vp, 0);
410			vfs_unbusy(mp);
411			return (ENOENT);
412		}
413	}
414	VOP_UNLOCK(vp, 0);
415
416	/*
417	 * Remove the filesystem containing currently-running executable
418	 * from the mount list, to prevent it from being unmounted
419	 * by vfs_unmountall(), and to avoid confusing vfs_mountroot().
420	 *
421	 * Also preserve /dev - forcibly unmounting it could cause driver
422	 * reinitialization.
423	 */
424
425	vfs_ref(rootdevmp);
426	devmp = rootdevmp;
427	rootdevmp = NULL;
428
429	mtx_lock(&mountlist_mtx);
430	TAILQ_REMOVE(&mountlist, mp, mnt_list);
431	TAILQ_REMOVE(&mountlist, devmp, mnt_list);
432	mtx_unlock(&mountlist_mtx);
433
434	oldrootvnode = rootvnode;
435
436	/*
437	 * Unmount everything except for the two filesystems preserved above.
438	 */
439	vfs_unmountall();
440
441	/*
442	 * Add /dev back; vfs_mountroot() will move it into its new place.
443	 */
444	mtx_lock(&mountlist_mtx);
445	TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list);
446	mtx_unlock(&mountlist_mtx);
447	rootdevmp = devmp;
448	vfs_rel(rootdevmp);
449
450	/*
451	 * Mount the new rootfs.
452	 */
453	vfs_mountroot();
454
455	/*
456	 * Update all references to the old rootvnode.
457	 */
458	mountcheckdirs(oldrootvnode, rootvnode);
459
460	/*
461	 * Add the temporary filesystem back and unbusy it.
462	 */
463	mtx_lock(&mountlist_mtx);
464	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
465	mtx_unlock(&mountlist_mtx);
466	vfs_unbusy(mp);
467
468	return (0);
469}
470
471/*
472 * If the shutdown was a clean halt, behave accordingly.
473 */
474static void
475shutdown_halt(void *junk, int howto)
476{
477
478	if (howto & RB_HALT) {
479		printf("\n");
480		printf("The operating system has halted.\n");
481		printf("Please press any key to reboot.\n\n");
482		switch (cngetc()) {
483		case -1:		/* No console, just die */
484			cpu_halt();
485			/* NOTREACHED */
486		default:
487			howto &= ~RB_HALT;
488			break;
489		}
490	}
491}
492
493/*
494 * Check to see if the system paniced, pause and then reboot
495 * according to the specified delay.
496 */
497static void
498shutdown_panic(void *junk, int howto)
499{
500	int loop;
501
502	if (howto & RB_DUMP) {
503		if (panic_reboot_wait_time != 0) {
504			if (panic_reboot_wait_time != -1) {
505				printf("Automatic reboot in %d seconds - "
506				       "press a key on the console to abort\n",
507					panic_reboot_wait_time);
508				for (loop = panic_reboot_wait_time * 10;
509				     loop > 0; --loop) {
510					DELAY(1000 * 100); /* 1/10th second */
511					/* Did user type a key? */
512					if (cncheckc() != -1)
513						break;
514				}
515				if (!loop)
516					return;
517			}
518		} else { /* zero time specified - reboot NOW */
519			return;
520		}
521		printf("--> Press a key on the console to reboot,\n");
522		printf("--> or switch off the system now.\n");
523		cngetc();
524	}
525}
526
527/*
528 * Everything done, now reset
529 */
530static void
531shutdown_reset(void *junk, int howto)
532{
533
534	printf("Rebooting...\n");
535	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
536
537	/*
538	 * Acquiring smp_ipi_mtx here has a double effect:
539	 * - it disables interrupts avoiding CPU0 preemption
540	 *   by fast handlers (thus deadlocking  against other CPUs)
541	 * - it avoids deadlocks against smp_rendezvous() or, more
542	 *   generally, threads busy-waiting, with this spinlock held,
543	 *   and waiting for responses by threads on other CPUs
544	 *   (ie. smp_tlb_shootdown()).
545	 *
546	 * For the !SMP case it just needs to handle the former problem.
547	 */
548#ifdef SMP
549	mtx_lock_spin(&smp_ipi_mtx);
550#else
551	spinlock_enter();
552#endif
553
554	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
555	cpu_reset();
556	/* NOTREACHED */ /* assuming reset worked */
557}
558
559#if defined(WITNESS) || defined(INVARIANT_SUPPORT)
560static int kassert_warn_only = 0;
561#ifdef KDB
562static int kassert_do_kdb = 0;
563#endif
564#ifdef KTR
565static int kassert_do_ktr = 0;
566#endif
567static int kassert_do_log = 1;
568static int kassert_log_pps_limit = 4;
569static int kassert_log_mute_at = 0;
570static int kassert_log_panic_at = 0;
571static int kassert_warnings = 0;
572
573SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
574
575SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
576    &kassert_warn_only, 0,
577    "KASSERT triggers a panic (1) or just a warning (0)");
578
579#ifdef KDB
580SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
581    &kassert_do_kdb, 0, "KASSERT will enter the debugger");
582#endif
583
584#ifdef KTR
585SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
586    &kassert_do_ktr, 0,
587    "KASSERT does a KTR, set this to the KTRMASK you want");
588#endif
589
590SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
591    &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
592
593SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
594    &kassert_warnings, 0, "number of KASSERTs that have been triggered");
595
596SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
597    &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
598
599SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
600    &kassert_log_pps_limit, 0, "limit number of log messages per second");
601
602SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
603    &kassert_log_mute_at, 0, "max number of KASSERTS to log");
604
605static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
606
607SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
608    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
609    kassert_sysctl_kassert, "I", "set to trigger a test kassert");
610
611static int
612kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
613{
614	int error, i;
615
616	error = sysctl_wire_old_buffer(req, sizeof(int));
617	if (error == 0) {
618		i = 0;
619		error = sysctl_handle_int(oidp, &i, 0, req);
620	}
621	if (error != 0 || req->newptr == NULL)
622		return (error);
623	KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
624	return (0);
625}
626
627/*
628 * Called by KASSERT, this decides if we will panic
629 * or if we will log via printf and/or ktr.
630 */
631void
632kassert_panic(const char *fmt, ...)
633{
634	static char buf[256];
635	va_list ap;
636
637	va_start(ap, fmt);
638	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
639	va_end(ap);
640
641	/*
642	 * panic if we're not just warning, or if we've exceeded
643	 * kassert_log_panic_at warnings.
644	 */
645	if (!kassert_warn_only ||
646	    (kassert_log_panic_at > 0 &&
647	     kassert_warnings >= kassert_log_panic_at)) {
648		va_start(ap, fmt);
649		vpanic(fmt, ap);
650		/* NORETURN */
651	}
652#ifdef KTR
653	if (kassert_do_ktr)
654		CTR0(ktr_mask, buf);
655#endif /* KTR */
656	/*
657	 * log if we've not yet met the mute limit.
658	 */
659	if (kassert_do_log &&
660	    (kassert_log_mute_at == 0 ||
661	     kassert_warnings < kassert_log_mute_at)) {
662		static  struct timeval lasterr;
663		static  int curerr;
664
665		if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
666			printf("KASSERT failed: %s\n", buf);
667			kdb_backtrace();
668		}
669	}
670#ifdef KDB
671	if (kassert_do_kdb) {
672		kdb_enter(KDB_WHY_KASSERT, buf);
673	}
674#endif
675	atomic_add_int(&kassert_warnings, 1);
676}
677#endif
678
679/*
680 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
681 * and then reboots.  If we are called twice, then we avoid trying to sync
682 * the disks as this often leads to recursive panics.
683 */
684void
685panic(const char *fmt, ...)
686{
687	va_list ap;
688
689	va_start(ap, fmt);
690	vpanic(fmt, ap);
691}
692
693void
694vpanic(const char *fmt, va_list ap)
695{
696#ifdef SMP
697	cpuset_t other_cpus;
698#endif
699	struct thread *td = curthread;
700	int bootopt, newpanic;
701	static char buf[256];
702
703	spinlock_enter();
704
705#ifdef SMP
706	/*
707	 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
708	 * concurrently entering panic.  Only the winner will proceed
709	 * further.
710	 */
711	if (panicstr == NULL && !kdb_active) {
712		other_cpus = all_cpus;
713		CPU_CLR(PCPU_GET(cpuid), &other_cpus);
714		stop_cpus_hard(other_cpus);
715	}
716
717	/*
718	 * Ensure that the scheduler is stopped while panicking, even if panic
719	 * has been entered from kdb.
720	 */
721	td->td_stopsched = 1;
722#endif
723
724	bootopt = RB_AUTOBOOT;
725	newpanic = 0;
726	if (panicstr)
727		bootopt |= RB_NOSYNC;
728	else {
729		bootopt |= RB_DUMP;
730		panicstr = fmt;
731		newpanic = 1;
732	}
733
734	if (newpanic) {
735		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
736		panicstr = buf;
737		cngrab();
738		printf("panic: %s\n", buf);
739	} else {
740		printf("panic: ");
741		vprintf(fmt, ap);
742		printf("\n");
743	}
744#ifdef SMP
745	printf("cpuid = %d\n", PCPU_GET(cpuid));
746#endif
747
748#ifdef KDB
749	if (newpanic && trace_on_panic)
750		kdb_backtrace();
751	if (debugger_on_panic)
752		kdb_enter(KDB_WHY_PANIC, "panic");
753#endif
754	/*thread_lock(td); */
755	td->td_flags |= TDF_INPANIC;
756	/* thread_unlock(td); */
757	if (!sync_on_panic)
758		bootopt |= RB_NOSYNC;
759	kern_reboot(bootopt);
760}
761
762/*
763 * Support for poweroff delay.
764 *
765 * Please note that setting this delay too short might power off your machine
766 * before the write cache on your hard disk has been flushed, leading to
767 * soft-updates inconsistencies.
768 */
769#ifndef POWEROFF_DELAY
770# define POWEROFF_DELAY 5000
771#endif
772static int poweroff_delay = POWEROFF_DELAY;
773
774SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
775    &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
776
777static void
778poweroff_wait(void *junk, int howto)
779{
780
781	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
782		return;
783	DELAY(poweroff_delay * 1000);
784}
785
786/*
787 * Some system processes (e.g. syncer) need to be stopped at appropriate
788 * points in their main loops prior to a system shutdown, so that they
789 * won't interfere with the shutdown process (e.g. by holding a disk buf
790 * to cause sync to fail).  For each of these system processes, register
791 * shutdown_kproc() as a handler for one of shutdown events.
792 */
793static int kproc_shutdown_wait = 60;
794SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
795    &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
796
797void
798kproc_shutdown(void *arg, int howto)
799{
800	struct proc *p;
801	int error;
802
803	if (panicstr)
804		return;
805
806	p = (struct proc *)arg;
807	printf("Waiting (max %d seconds) for system process `%s' to stop... ",
808	    kproc_shutdown_wait, p->p_comm);
809	error = kproc_suspend(p, kproc_shutdown_wait * hz);
810
811	if (error == EWOULDBLOCK)
812		printf("timed out\n");
813	else
814		printf("done\n");
815}
816
817void
818kthread_shutdown(void *arg, int howto)
819{
820	struct thread *td;
821	int error;
822
823	if (panicstr)
824		return;
825
826	td = (struct thread *)arg;
827	printf("Waiting (max %d seconds) for system thread `%s' to stop... ",
828	    kproc_shutdown_wait, td->td_name);
829	error = kthread_suspend(td, kproc_shutdown_wait * hz);
830
831	if (error == EWOULDBLOCK)
832		printf("timed out\n");
833	else
834		printf("done\n");
835}
836
837static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
838SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
839    dumpdevname, 0, "Device for kernel dumps");
840
841/* Registration of dumpers */
842int
843set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
844{
845	size_t wantcopy;
846	int error;
847
848	error = priv_check(td, PRIV_SETDUMPER);
849	if (error != 0)
850		return (error);
851
852	if (di == NULL) {
853		if (dumper.blockbuf != NULL)
854			free(dumper.blockbuf, M_DUMPER);
855		bzero(&dumper, sizeof(dumper));
856		dumpdevname[0] = '\0';
857		return (0);
858	}
859	if (dumper.dumper != NULL)
860		return (EBUSY);
861	dumper = *di;
862	wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
863	if (wantcopy >= sizeof(dumpdevname)) {
864		printf("set_dumper: device name truncated from '%s' -> '%s'\n",
865			devname, dumpdevname);
866	}
867	dumper.blockbuf = malloc(di->blocksize, M_DUMPER, M_WAITOK | M_ZERO);
868	return (0);
869}
870
871/* Call dumper with bounds checking. */
872int
873dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
874    off_t offset, size_t length)
875{
876
877	if (length != 0 && (offset < di->mediaoffset ||
878	    offset - di->mediaoffset + length > di->mediasize)) {
879		printf("Attempt to write outside dump device boundaries.\n"
880	    "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
881		    (intmax_t)offset, (intmax_t)di->mediaoffset,
882		    (uintmax_t)length, (intmax_t)di->mediasize);
883		return (ENOSPC);
884	}
885	return (di->dumper(di->priv, virtual, physical, offset, length));
886}
887
888/* Call dumper with bounds checking. */
889int
890dump_write_pad(struct dumperinfo *di, void *virtual, vm_offset_t physical,
891    off_t offset, size_t length, size_t *size)
892{
893	char *temp;
894	int ret;
895
896	if (length > di->blocksize)
897		return (ENOMEM);
898
899	*size = di->blocksize;
900	if (length == di->blocksize)
901		temp = virtual;
902	else {
903		temp = di->blockbuf;
904		memset(temp + length, 0, di->blocksize - length);
905		memcpy(temp, virtual, length);
906	}
907	ret = dump_write(di, temp, physical, offset, *size);
908
909	return (ret);
910}
911
912
913void
914mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
915    uint64_t dumplen, uint32_t blksz)
916{
917
918	bzero(kdh, sizeof(*kdh));
919	strlcpy(kdh->magic, magic, sizeof(kdh->magic));
920	strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
921	kdh->version = htod32(KERNELDUMPVERSION);
922	kdh->architectureversion = htod32(archver);
923	kdh->dumplength = htod64(dumplen);
924	kdh->dumptime = htod64(time_second);
925	kdh->blocksize = htod32(blksz);
926	strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
927	strlcpy(kdh->versionstring, version, sizeof(kdh->versionstring));
928	if (panicstr != NULL)
929		strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
930	kdh->parity = kerneldump_parity(kdh);
931}
932
933#ifdef DDB
934DB_SHOW_COMMAND(panic, db_show_panic)
935{
936
937	if (panicstr == NULL)
938		db_printf("panicstr not set\n");
939	else
940		db_printf("panic: %s\n", panicstr);
941}
942#endif
943