kern_shutdown.c revision 283676
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: stable/10/sys/kern/kern_shutdown.c 283676 2015-05-29 04:01:39Z markj $");
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/pcb.h>
77#include <machine/smp.h>
78
79#include <security/mac/mac_framework.h>
80
81#include <vm/vm.h>
82#include <vm/vm_object.h>
83#include <vm/vm_page.h>
84#include <vm/vm_pager.h>
85#include <vm/swap_pager.h>
86
87#include <sys/signalvar.h>
88
89#ifndef PANIC_REBOOT_WAIT_TIME
90#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
91#endif
92static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
93SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RW | CTLFLAG_TUN,
94    &panic_reboot_wait_time, 0,
95    "Seconds to wait before rebooting after a panic");
96TUNABLE_INT("kern.panic_reboot_wait_time", &panic_reboot_wait_time);
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
106int debugger_on_panic = 0;
107#else
108int debugger_on_panic = 1;
109#endif
110SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
111    CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN,
112    &debugger_on_panic, 0, "Run debugger on kernel panic");
113TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
114
115#ifdef KDB_TRACE
116static int trace_on_panic = 1;
117#else
118static int trace_on_panic = 0;
119#endif
120SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
121    CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN,
122    &trace_on_panic, 0, "Print stack trace on kernel panic");
123TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
124#endif /* KDB */
125
126static int sync_on_panic = 0;
127SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
128	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
129TUNABLE_INT("kern.sync_on_panic", &sync_on_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
142/*
143 * Variable panicstr contains argument to first call to panic; used as flag
144 * to indicate that the kernel has already called panic.
145 */
146const char *panicstr;
147
148int dumping;				/* system is dumping */
149int rebooting;				/* system is rebooting */
150static struct dumperinfo dumper;	/* our selected dumper */
151
152/* Context information for dump-debuggers. */
153static struct pcb dumppcb;		/* Registers. */
154lwpid_t dumptid;			/* Thread ID. */
155
156static void poweroff_wait(void *, int);
157static void shutdown_halt(void *junk, int howto);
158static void shutdown_panic(void *junk, int howto);
159static void shutdown_reset(void *junk, int howto);
160
161/* register various local shutdown events */
162static void
163shutdown_conf(void *unused)
164{
165
166	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
167	    SHUTDOWN_PRI_FIRST);
168	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
169	    SHUTDOWN_PRI_LAST + 100);
170	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
171	    SHUTDOWN_PRI_LAST + 100);
172	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
173	    SHUTDOWN_PRI_LAST + 200);
174}
175
176SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
177
178/*
179 * The system call that results in a reboot.
180 */
181/* ARGSUSED */
182int
183sys_reboot(struct thread *td, struct reboot_args *uap)
184{
185	int error;
186
187	error = 0;
188#ifdef MAC
189	error = mac_system_check_reboot(td->td_ucred, uap->opt);
190#endif
191	if (error == 0)
192		error = priv_check(td, PRIV_REBOOT);
193	if (error == 0) {
194		mtx_lock(&Giant);
195		kern_reboot(uap->opt);
196		mtx_unlock(&Giant);
197	}
198	return (error);
199}
200
201/*
202 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
203 */
204static int shutdown_howto = 0;
205
206void
207shutdown_nice(int howto)
208{
209
210	shutdown_howto = howto;
211
212	/* Send a signal to init(8) and have it shutdown the world */
213	if (initproc != NULL) {
214		PROC_LOCK(initproc);
215		kern_psignal(initproc, SIGINT);
216		PROC_UNLOCK(initproc);
217	} else {
218		/* No init(8) running, so simply reboot */
219		kern_reboot(RB_NOSYNC);
220	}
221	return;
222}
223static int	waittime = -1;
224
225static void
226print_uptime(void)
227{
228	int f;
229	struct timespec ts;
230
231	getnanouptime(&ts);
232	printf("Uptime: ");
233	f = 0;
234	if (ts.tv_sec >= 86400) {
235		printf("%ldd", (long)ts.tv_sec / 86400);
236		ts.tv_sec %= 86400;
237		f = 1;
238	}
239	if (f || ts.tv_sec >= 3600) {
240		printf("%ldh", (long)ts.tv_sec / 3600);
241		ts.tv_sec %= 3600;
242		f = 1;
243	}
244	if (f || ts.tv_sec >= 60) {
245		printf("%ldm", (long)ts.tv_sec / 60);
246		ts.tv_sec %= 60;
247		f = 1;
248	}
249	printf("%lds\n", (long)ts.tv_sec);
250}
251
252int
253doadump(boolean_t textdump)
254{
255	boolean_t coredump;
256
257	if (dumping)
258		return (EBUSY);
259	if (dumper.dumper == NULL)
260		return (ENXIO);
261
262	savectx(&dumppcb);
263	dumptid = curthread->td_tid;
264	dumping++;
265
266	coredump = TRUE;
267#ifdef DDB
268	if (textdump && textdump_pending) {
269		coredump = FALSE;
270		textdump_dumpsys(&dumper);
271	}
272#endif
273	if (coredump)
274		dumpsys(&dumper);
275
276	dumping--;
277	return (0);
278}
279
280static int
281isbufbusy(struct buf *bp)
282{
283	if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
284	    BUF_ISLOCKED(bp)) ||
285	    ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
286		return (1);
287	return (0);
288}
289
290/*
291 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
292 */
293void
294kern_reboot(int howto)
295{
296	static int first_buf_printf = 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	/* collect extra flags that shutdown_nice might have set */
315	howto |= shutdown_howto;
316
317	/* We are out of the debugger now. */
318	kdb_active = 0;
319
320	/*
321	 * Do any callouts that should be done BEFORE syncing the filesystems.
322	 */
323	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
324
325	/*
326	 * Now sync filesystems
327	 */
328	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
329		register struct buf *bp;
330		int iter, nbusy, pbusy;
331#ifndef PREEMPTION
332		int subiter;
333#endif
334
335		waittime = 0;
336
337		wdog_kern_pat(WD_LASTVAL);
338		sys_sync(curthread, NULL);
339
340		/*
341		 * With soft updates, some buffers that are
342		 * written will be remarked as dirty until other
343		 * buffers are written.
344		 */
345		for (iter = pbusy = 0; iter < 20; iter++) {
346			nbusy = 0;
347			for (bp = &buf[nbuf]; --bp >= buf; )
348				if (isbufbusy(bp))
349					nbusy++;
350			if (nbusy == 0) {
351				if (first_buf_printf)
352					printf("All buffers synced.");
353				break;
354			}
355			if (first_buf_printf) {
356				printf("Syncing disks, buffers remaining... ");
357				first_buf_printf = 0;
358			}
359			printf("%d ", nbusy);
360			if (nbusy < pbusy)
361				iter = 0;
362			pbusy = nbusy;
363
364			wdog_kern_pat(WD_LASTVAL);
365			sys_sync(curthread, NULL);
366
367#ifdef PREEMPTION
368			/*
369			 * Drop Giant and spin for a while to allow
370			 * interrupt threads to run.
371			 */
372			DROP_GIANT();
373			DELAY(50000 * iter);
374			PICKUP_GIANT();
375#else
376			/*
377			 * Drop Giant and context switch several times to
378			 * allow interrupt threads to run.
379			 */
380			DROP_GIANT();
381			for (subiter = 0; subiter < 50 * iter; subiter++) {
382				thread_lock(curthread);
383				mi_switch(SW_VOL, NULL);
384				thread_unlock(curthread);
385				DELAY(1000);
386			}
387			PICKUP_GIANT();
388#endif
389		}
390		printf("\n");
391		/*
392		 * Count only busy local buffers to prevent forcing
393		 * a fsck if we're just a client of a wedged NFS server
394		 */
395		nbusy = 0;
396		for (bp = &buf[nbuf]; --bp >= buf; ) {
397			if (isbufbusy(bp)) {
398#if 0
399/* XXX: This is bogus.  We should probably have a BO_REMOTE flag instead */
400				if (bp->b_dev == NULL) {
401					TAILQ_REMOVE(&mountlist,
402					    bp->b_vp->v_mount, mnt_list);
403					continue;
404				}
405#endif
406				nbusy++;
407				if (show_busybufs > 0) {
408					printf(
409	    "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
410					    nbusy, bp, bp->b_vp, bp->b_flags,
411					    (intmax_t)bp->b_blkno,
412					    (intmax_t)bp->b_lblkno);
413					BUF_LOCKPRINTINFO(bp);
414					if (show_busybufs > 1)
415						vn_printf(bp->b_vp,
416						    "vnode content: ");
417				}
418			}
419		}
420		if (nbusy) {
421			/*
422			 * Failed to sync all blocks. Indicate this and don't
423			 * unmount filesystems (thus forcing an fsck on reboot).
424			 */
425			printf("Giving up on %d buffers\n", nbusy);
426			DELAY(5000000);	/* 5 seconds */
427		} else {
428			if (!first_buf_printf)
429				printf("Final sync complete\n");
430			/*
431			 * Unmount filesystems
432			 */
433			if (panicstr == 0)
434				vfs_unmountall();
435		}
436		swapoff_all();
437		DELAY(100000);		/* wait for console output to finish */
438	}
439
440	print_uptime();
441
442	cngrab();
443
444	/*
445	 * Ok, now do things that assume all filesystem activity has
446	 * been completed.
447	 */
448	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
449
450	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
451		doadump(TRUE);
452
453	/* Now that we're going to really halt the system... */
454	EVENTHANDLER_INVOKE(shutdown_final, howto);
455
456	for(;;) ;	/* safety against shutdown_reset not working */
457	/* NOTREACHED */
458}
459
460/*
461 * If the shutdown was a clean halt, behave accordingly.
462 */
463static void
464shutdown_halt(void *junk, int howto)
465{
466
467	if (howto & RB_HALT) {
468		printf("\n");
469		printf("The operating system has halted.\n");
470		printf("Please press any key to reboot.\n\n");
471		switch (cngetc()) {
472		case -1:		/* No console, just die */
473			cpu_halt();
474			/* NOTREACHED */
475		default:
476			howto &= ~RB_HALT;
477			break;
478		}
479	}
480}
481
482/*
483 * Check to see if the system paniced, pause and then reboot
484 * according to the specified delay.
485 */
486static void
487shutdown_panic(void *junk, int howto)
488{
489	int loop;
490
491	if (howto & RB_DUMP) {
492		if (panic_reboot_wait_time != 0) {
493			if (panic_reboot_wait_time != -1) {
494				printf("Automatic reboot in %d seconds - "
495				       "press a key on the console to abort\n",
496					panic_reboot_wait_time);
497				for (loop = panic_reboot_wait_time * 10;
498				     loop > 0; --loop) {
499					DELAY(1000 * 100); /* 1/10th second */
500					/* Did user type a key? */
501					if (cncheckc() != -1)
502						break;
503				}
504				if (!loop)
505					return;
506			}
507		} else { /* zero time specified - reboot NOW */
508			return;
509		}
510		printf("--> Press a key on the console to reboot,\n");
511		printf("--> or switch off the system now.\n");
512		cngetc();
513	}
514}
515
516/*
517 * Everything done, now reset
518 */
519static void
520shutdown_reset(void *junk, int howto)
521{
522
523	printf("Rebooting...\n");
524	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
525
526	/*
527	 * Acquiring smp_ipi_mtx here has a double effect:
528	 * - it disables interrupts avoiding CPU0 preemption
529	 *   by fast handlers (thus deadlocking  against other CPUs)
530	 * - it avoids deadlocks against smp_rendezvous() or, more
531	 *   generally, threads busy-waiting, with this spinlock held,
532	 *   and waiting for responses by threads on other CPUs
533	 *   (ie. smp_tlb_shootdown()).
534	 *
535	 * For the !SMP case it just needs to handle the former problem.
536	 */
537#ifdef SMP
538	mtx_lock_spin(&smp_ipi_mtx);
539#else
540	spinlock_enter();
541#endif
542
543	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
544	cpu_reset();
545	/* NOTREACHED */ /* assuming reset worked */
546}
547
548#if defined(WITNESS) || defined(INVARIANTS)
549static int kassert_warn_only = 0;
550#ifdef KDB
551static int kassert_do_kdb = 0;
552#endif
553#ifdef KTR
554static int kassert_do_ktr = 0;
555#endif
556static int kassert_do_log = 1;
557static int kassert_log_pps_limit = 4;
558static int kassert_log_mute_at = 0;
559static int kassert_log_panic_at = 0;
560static int kassert_warnings = 0;
561
562SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
563
564SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RW | CTLFLAG_TUN,
565    &kassert_warn_only, 0,
566    "KASSERT triggers a panic (1) or just a warning (0)");
567TUNABLE_INT("debug.kassert.warn_only", &kassert_warn_only);
568
569#ifdef KDB
570SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RW | CTLFLAG_TUN,
571    &kassert_do_kdb, 0, "KASSERT will enter the debugger");
572TUNABLE_INT("debug.kassert.do_kdb", &kassert_do_kdb);
573#endif
574
575#ifdef KTR
576SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RW | CTLFLAG_TUN,
577    &kassert_do_ktr, 0,
578    "KASSERT does a KTR, set this to the KTRMASK you want");
579TUNABLE_INT("debug.kassert.do_ktr", &kassert_do_ktr);
580#endif
581
582SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RW | CTLFLAG_TUN,
583    &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
584TUNABLE_INT("debug.kassert.do_log", &kassert_do_log);
585
586SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RW | CTLFLAG_TUN,
587    &kassert_warnings, 0, "number of KASSERTs that have been triggered");
588TUNABLE_INT("debug.kassert.warnings", &kassert_warnings);
589
590SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RW | CTLFLAG_TUN,
591    &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
592TUNABLE_INT("debug.kassert.log_panic_at", &kassert_log_panic_at);
593
594SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RW | CTLFLAG_TUN,
595    &kassert_log_pps_limit, 0, "limit number of log messages per second");
596TUNABLE_INT("debug.kassert.log_pps_limit", &kassert_log_pps_limit);
597
598SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RW | CTLFLAG_TUN,
599    &kassert_log_mute_at, 0, "max number of KASSERTS to log");
600TUNABLE_INT("debug.kassert.log_mute_at", &kassert_log_mute_at);
601
602static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
603
604SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
605    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
606    kassert_sysctl_kassert, "I", "set to trigger a test kassert");
607
608static int
609kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
610{
611	int error, i;
612
613	error = sysctl_wire_old_buffer(req, sizeof(int));
614	if (error == 0) {
615		i = 0;
616		error = sysctl_handle_int(oidp, &i, 0, req);
617	}
618	if (error != 0 || req->newptr == NULL)
619		return (error);
620	KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
621	return (0);
622}
623
624/*
625 * Called by KASSERT, this decides if we will panic
626 * or if we will log via printf and/or ktr.
627 */
628void
629kassert_panic(const char *fmt, ...)
630{
631	static char buf[256];
632	va_list ap;
633
634	va_start(ap, fmt);
635	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
636	va_end(ap);
637
638	/*
639	 * panic if we're not just warning, or if we've exceeded
640	 * kassert_log_panic_at warnings.
641	 */
642	if (!kassert_warn_only ||
643	    (kassert_log_panic_at > 0 &&
644	     kassert_warnings >= kassert_log_panic_at)) {
645		va_start(ap, fmt);
646		vpanic(fmt, ap);
647		/* NORETURN */
648	}
649#ifdef KTR
650	if (kassert_do_ktr)
651		CTR0(ktr_mask, buf);
652#endif /* KTR */
653	/*
654	 * log if we've not yet met the mute limit.
655	 */
656	if (kassert_do_log &&
657	    (kassert_log_mute_at == 0 ||
658	     kassert_warnings < kassert_log_mute_at)) {
659		static  struct timeval lasterr;
660		static  int curerr;
661
662		if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
663			printf("KASSERT failed: %s\n", buf);
664			kdb_backtrace();
665		}
666	}
667#ifdef KDB
668	if (kassert_do_kdb) {
669		kdb_enter(KDB_WHY_KASSERT, buf);
670	}
671#endif
672	atomic_add_int(&kassert_warnings, 1);
673}
674#endif
675
676/*
677 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
678 * and then reboots.  If we are called twice, then we avoid trying to sync
679 * the disks as this often leads to recursive panics.
680 */
681void
682panic(const char *fmt, ...)
683{
684	va_list ap;
685
686	va_start(ap, fmt);
687	vpanic(fmt, ap);
688}
689
690void
691vpanic(const char *fmt, va_list ap)
692{
693#ifdef SMP
694	cpuset_t other_cpus;
695#endif
696	struct thread *td = curthread;
697	int bootopt, newpanic;
698	static char buf[256];
699
700	spinlock_enter();
701
702#ifdef SMP
703	/*
704	 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
705	 * concurrently entering panic.  Only the winner will proceed
706	 * further.
707	 */
708	if (panicstr == NULL && !kdb_active) {
709		other_cpus = all_cpus;
710		CPU_CLR(PCPU_GET(cpuid), &other_cpus);
711		stop_cpus_hard(other_cpus);
712	}
713
714	/*
715	 * We set stop_scheduler here and not in the block above,
716	 * because we want to ensure that if panic has been called and
717	 * stop_scheduler_on_panic is true, then stop_scheduler will
718	 * always be set.  Even if panic has been entered from kdb.
719	 */
720	td->td_stopsched = 1;
721#endif
722
723	bootopt = RB_AUTOBOOT;
724	newpanic = 0;
725	if (panicstr)
726		bootopt |= RB_NOSYNC;
727	else {
728		bootopt |= RB_DUMP;
729		panicstr = fmt;
730		newpanic = 1;
731	}
732
733	if (newpanic) {
734		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
735		panicstr = buf;
736		cngrab();
737		printf("panic: %s\n", buf);
738	} else {
739		printf("panic: ");
740		vprintf(fmt, ap);
741		printf("\n");
742	}
743#ifdef SMP
744	printf("cpuid = %d\n", PCPU_GET(cpuid));
745#endif
746
747#ifdef KDB
748	if (newpanic && trace_on_panic)
749		kdb_backtrace();
750	if (debugger_on_panic)
751		kdb_enter(KDB_WHY_PANIC, "panic");
752#endif
753	/*thread_lock(td); */
754	td->td_flags |= TDF_INPANIC;
755	/* thread_unlock(td); */
756	if (!sync_on_panic)
757		bootopt |= RB_NOSYNC;
758	kern_reboot(bootopt);
759}
760
761/*
762 * Support for poweroff delay.
763 *
764 * Please note that setting this delay too short might power off your machine
765 * before the write cache on your hard disk has been flushed, leading to
766 * soft-updates inconsistencies.
767 */
768#ifndef POWEROFF_DELAY
769# define POWEROFF_DELAY 5000
770#endif
771static int poweroff_delay = POWEROFF_DELAY;
772
773SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
774    &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
775
776static void
777poweroff_wait(void *junk, int howto)
778{
779
780	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
781		return;
782	DELAY(poweroff_delay * 1000);
783}
784
785/*
786 * Some system processes (e.g. syncer) need to be stopped at appropriate
787 * points in their main loops prior to a system shutdown, so that they
788 * won't interfere with the shutdown process (e.g. by holding a disk buf
789 * to cause sync to fail).  For each of these system processes, register
790 * shutdown_kproc() as a handler for one of shutdown events.
791 */
792static int kproc_shutdown_wait = 60;
793SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
794    &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
795
796void
797kproc_shutdown(void *arg, int howto)
798{
799	struct proc *p;
800	int error;
801
802	if (panicstr)
803		return;
804
805	p = (struct proc *)arg;
806	printf("Waiting (max %d seconds) for system process `%s' to stop...",
807	    kproc_shutdown_wait, p->p_comm);
808	error = kproc_suspend(p, kproc_shutdown_wait * hz);
809
810	if (error == EWOULDBLOCK)
811		printf("timed out\n");
812	else
813		printf("done\n");
814}
815
816void
817kthread_shutdown(void *arg, int howto)
818{
819	struct thread *td;
820	int error;
821
822	if (panicstr)
823		return;
824
825	td = (struct thread *)arg;
826	printf("Waiting (max %d seconds) for system thread `%s' to stop...",
827	    kproc_shutdown_wait, td->td_name);
828	error = kthread_suspend(td, kproc_shutdown_wait * hz);
829
830	if (error == EWOULDBLOCK)
831		printf("timed out\n");
832	else
833		printf("done\n");
834}
835
836static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
837SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
838    dumpdevname, 0, "Device for kernel dumps");
839
840/* Registration of dumpers */
841int
842set_dumper(struct dumperinfo *di, const char *devname)
843{
844	size_t wantcopy;
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. */
863int
864dump_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
879void
880mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
881    uint64_t dumplen, uint32_t blksz)
882{
883
884	bzero(kdh, sizeof(*kdh));
885	strncpy(kdh->magic, magic, sizeof(kdh->magic));
886	strncpy(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	strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
893	strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
894	if (panicstr != NULL)
895		strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
896	kdh->parity = kerneldump_parity(kdh);
897}
898