kern_shutdown.c revision 344905 
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/11/sys/kern/kern_shutdown.c 344905 2019-03-08 00:20:37Z jhb $");
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/taskqueue.h>
72#include <sys/vnode.h>
73#include <sys/watchdog.h>
74
75#include <ddb/ddb.h>
76
77#include <machine/cpu.h>
78#include <machine/dump.h>
79#include <machine/pcb.h>
80#include <machine/smp.h>
81
82#include <security/mac/mac_framework.h>
83
84#include <vm/vm.h>
85#include <vm/vm_object.h>
86#include <vm/vm_page.h>
87#include <vm/vm_pager.h>
88#include <vm/swap_pager.h>
89
90#include <sys/signalvar.h>
91
92static MALLOC_DEFINE(M_DUMPER, "dumper", "dumper block buffer");
93
94#ifndef PANIC_REBOOT_WAIT_TIME
95#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
96#endif
97static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
98SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
99    &panic_reboot_wait_time, 0,
100    "Seconds to wait before rebooting after a panic");
101
102/*
103 * Note that stdarg.h and the ANSI style va_start macro is used for both
104 * ANSI and traditional C compilers.
105 */
106#include <machine/stdarg.h>
107
108#ifdef KDB
109#ifdef KDB_UNATTENDED
110static int debugger_on_panic = 0;
111#else
112static int debugger_on_panic = 1;
113#endif
114SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
115    CTLFLAG_RWTUN | CTLFLAG_SECURE,
116    &debugger_on_panic, 0, "Run debugger on kernel panic");
117
118int debugger_on_trap = 0;
119SYSCTL_INT(_debug, OID_AUTO, debugger_on_trap,
120    CTLFLAG_RWTUN | CTLFLAG_SECURE,
121    &debugger_on_trap, 0, "Run debugger on kernel trap before panic");
122
123#ifdef KDB_TRACE
124static int trace_on_panic = 1;
125#else
126static int trace_on_panic = 0;
127#endif
128SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
129    CTLFLAG_RWTUN | CTLFLAG_SECURE,
130    &trace_on_panic, 0, "Print stack trace on kernel panic");
131#endif /* KDB */
132
133static int sync_on_panic = 0;
134SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
135	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
136
137static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
138    "Shutdown environment");
139
140#ifndef DIAGNOSTIC
141static int show_busybufs;
142#else
143static int show_busybufs = 1;
144#endif
145SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
146	&show_busybufs, 0, "");
147
148int suspend_blocked = 0;
149SYSCTL_INT(_kern, OID_AUTO, suspend_blocked, CTLFLAG_RW,
150	&suspend_blocked, 0, "Block suspend due to a pending shutdown");
151
152/*
153 * Variable panicstr contains argument to first call to panic; used as flag
154 * to indicate that the kernel has already called panic.
155 */
156const char *panicstr;
157
158int dumping;				/* system is dumping */
159int rebooting;				/* system is rebooting */
160static struct dumperinfo dumper;	/* our selected dumper */
161
162/* Context information for dump-debuggers. */
163static struct pcb dumppcb;		/* Registers. */
164lwpid_t dumptid;			/* Thread ID. */
165
166static struct cdevsw reroot_cdevsw = {
167     .d_version = D_VERSION,
168     .d_name    = "reroot",
169};
170
171static void poweroff_wait(void *, int);
172static void shutdown_halt(void *junk, int howto);
173static void shutdown_panic(void *junk, int howto);
174static void shutdown_reset(void *junk, int howto);
175static int kern_reroot(void);
176
177/* register various local shutdown events */
178static void
179shutdown_conf(void *unused)
180{
181
182	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
183	    SHUTDOWN_PRI_FIRST);
184	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
185	    SHUTDOWN_PRI_LAST + 100);
186	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
187	    SHUTDOWN_PRI_LAST + 100);
188	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
189	    SHUTDOWN_PRI_LAST + 200);
190}
191
192SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
193
194/*
195 * The only reason this exists is to create the /dev/reroot/ directory,
196 * used by reroot code in init(8) as a mountpoint for tmpfs.
197 */
198static void
199reroot_conf(void *unused)
200{
201	int error;
202	struct cdev *cdev;
203
204	error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev,
205	    &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot");
206	if (error != 0) {
207		printf("%s: failed to create device node, error %d",
208		    __func__, error);
209	}
210}
211
212SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL);
213
214/*
215 * The system call that results in a reboot.
216 */
217/* ARGSUSED */
218int
219sys_reboot(struct thread *td, struct reboot_args *uap)
220{
221	int error;
222
223	error = 0;
224#ifdef MAC
225	error = mac_system_check_reboot(td->td_ucred, uap->opt);
226#endif
227	if (error == 0)
228		error = priv_check(td, PRIV_REBOOT);
229	if (error == 0) {
230		if (uap->opt & RB_REROOT) {
231			error = kern_reroot();
232		} else {
233			mtx_lock(&Giant);
234			kern_reboot(uap->opt);
235			mtx_unlock(&Giant);
236		}
237	}
238	return (error);
239}
240
241static void
242shutdown_nice_task_fn(void *arg, int pending __unused)
243{
244	int howto;
245
246	howto = (uintptr_t)arg;
247	/* Send a signal to init(8) and have it shutdown the world. */
248	PROC_LOCK(initproc);
249	if (howto & RB_POWEROFF)
250		kern_psignal(initproc, SIGUSR2);
251	else if (howto & RB_HALT)
252		kern_psignal(initproc, SIGUSR1);
253	else
254		kern_psignal(initproc, SIGINT);
255	PROC_UNLOCK(initproc);
256}
257
258static struct task shutdown_nice_task = TASK_INITIALIZER(0,
259    &shutdown_nice_task_fn, NULL);
260
261/*
262 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
263 */
264void
265shutdown_nice(int howto)
266{
267
268	if (initproc != NULL && !SCHEDULER_STOPPED()) {
269		shutdown_nice_task.ta_context = (void *)(uintptr_t)howto;
270		taskqueue_enqueue(taskqueue_fast, &shutdown_nice_task);
271	} else {
272		/*
273		 * No init(8) running, or scheduler would not allow it
274		 * to run, so simply reboot.
275		 */
276		kern_reboot(howto | RB_NOSYNC);
277	}
278}
279
280static void
281print_uptime(void)
282{
283	int f;
284	struct timespec ts;
285
286	getnanouptime(&ts);
287	printf("Uptime: ");
288	f = 0;
289	if (ts.tv_sec >= 86400) {
290		printf("%ldd", (long)ts.tv_sec / 86400);
291		ts.tv_sec %= 86400;
292		f = 1;
293	}
294	if (f || ts.tv_sec >= 3600) {
295		printf("%ldh", (long)ts.tv_sec / 3600);
296		ts.tv_sec %= 3600;
297		f = 1;
298	}
299	if (f || ts.tv_sec >= 60) {
300		printf("%ldm", (long)ts.tv_sec / 60);
301		ts.tv_sec %= 60;
302		f = 1;
303	}
304	printf("%lds\n", (long)ts.tv_sec);
305}
306
307int
308doadump(boolean_t textdump)
309{
310	boolean_t coredump;
311	int error;
312
313	error = 0;
314	if (dumping)
315		return (EBUSY);
316	if (dumper.dumper == NULL)
317		return (ENXIO);
318
319	savectx(&dumppcb);
320	dumptid = curthread->td_tid;
321	dumping++;
322
323	coredump = TRUE;
324#ifdef DDB
325	if (textdump && textdump_pending) {
326		coredump = FALSE;
327		textdump_dumpsys(&dumper);
328	}
329#endif
330	if (coredump)
331		error = dumpsys(&dumper);
332
333	dumping--;
334	return (error);
335}
336
337/*
338 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
339 */
340void
341kern_reboot(int howto)
342{
343	static int once = 0;
344
345#if defined(SMP)
346	/*
347	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
348	 * systems don't shutdown properly (i.e., ACPI power off) if we
349	 * run on another processor.
350	 */
351	if (!SCHEDULER_STOPPED()) {
352		thread_lock(curthread);
353		sched_bind(curthread, 0);
354		thread_unlock(curthread);
355		KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
356	}
357#endif
358	/* We're in the process of rebooting. */
359	rebooting = 1;
360
361	/* We are out of the debugger now. */
362	kdb_active = 0;
363
364	/*
365	 * Do any callouts that should be done BEFORE syncing the filesystems.
366	 */
367	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
368
369	/*
370	 * Now sync filesystems
371	 */
372	if (!cold && (howto & RB_NOSYNC) == 0 && once == 0) {
373		once = 1;
374		bufshutdown(show_busybufs);
375	}
376
377	print_uptime();
378
379	cngrab();
380
381	/*
382	 * Ok, now do things that assume all filesystem activity has
383	 * been completed.
384	 */
385	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
386
387	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
388		doadump(TRUE);
389
390	/* Now that we're going to really halt the system... */
391	EVENTHANDLER_INVOKE(shutdown_final, howto);
392
393	for(;;) ;	/* safety against shutdown_reset not working */
394	/* NOTREACHED */
395}
396
397/*
398 * The system call that results in changing the rootfs.
399 */
400static int
401kern_reroot(void)
402{
403	struct vnode *oldrootvnode, *vp;
404	struct mount *mp, *devmp;
405	int error;
406
407	if (curproc != initproc)
408		return (EPERM);
409
410	/*
411	 * Mark the filesystem containing currently-running executable
412	 * (the temporary copy of init(8)) busy.
413	 */
414	vp = curproc->p_textvp;
415	error = vn_lock(vp, LK_SHARED);
416	if (error != 0)
417		return (error);
418	mp = vp->v_mount;
419	error = vfs_busy(mp, MBF_NOWAIT);
420	if (error != 0) {
421		vfs_ref(mp);
422		VOP_UNLOCK(vp, 0);
423		error = vfs_busy(mp, 0);
424		vn_lock(vp, LK_SHARED | LK_RETRY);
425		vfs_rel(mp);
426		if (error != 0) {
427			VOP_UNLOCK(vp, 0);
428			return (ENOENT);
429		}
430		if (vp->v_iflag & VI_DOOMED) {
431			VOP_UNLOCK(vp, 0);
432			vfs_unbusy(mp);
433			return (ENOENT);
434		}
435	}
436	VOP_UNLOCK(vp, 0);
437
438	/*
439	 * Remove the filesystem containing currently-running executable
440	 * from the mount list, to prevent it from being unmounted
441	 * by vfs_unmountall(), and to avoid confusing vfs_mountroot().
442	 *
443	 * Also preserve /dev - forcibly unmounting it could cause driver
444	 * reinitialization.
445	 */
446
447	vfs_ref(rootdevmp);
448	devmp = rootdevmp;
449	rootdevmp = NULL;
450
451	mtx_lock(&mountlist_mtx);
452	TAILQ_REMOVE(&mountlist, mp, mnt_list);
453	TAILQ_REMOVE(&mountlist, devmp, mnt_list);
454	mtx_unlock(&mountlist_mtx);
455
456	oldrootvnode = rootvnode;
457
458	/*
459	 * Unmount everything except for the two filesystems preserved above.
460	 */
461	vfs_unmountall();
462
463	/*
464	 * Add /dev back; vfs_mountroot() will move it into its new place.
465	 */
466	mtx_lock(&mountlist_mtx);
467	TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list);
468	mtx_unlock(&mountlist_mtx);
469	rootdevmp = devmp;
470	vfs_rel(rootdevmp);
471
472	/*
473	 * Mount the new rootfs.
474	 */
475	vfs_mountroot();
476
477	/*
478	 * Update all references to the old rootvnode.
479	 */
480	mountcheckdirs(oldrootvnode, rootvnode);
481
482	/*
483	 * Add the temporary filesystem back and unbusy it.
484	 */
485	mtx_lock(&mountlist_mtx);
486	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
487	mtx_unlock(&mountlist_mtx);
488	vfs_unbusy(mp);
489
490	return (0);
491}
492
493/*
494 * If the shutdown was a clean halt, behave accordingly.
495 */
496static void
497shutdown_halt(void *junk, int howto)
498{
499
500	if (howto & RB_HALT) {
501		printf("\n");
502		printf("The operating system has halted.\n");
503		printf("Please press any key to reboot.\n\n");
504		switch (cngetc()) {
505		case -1:		/* No console, just die */
506			cpu_halt();
507			/* NOTREACHED */
508		default:
509			howto &= ~RB_HALT;
510			break;
511		}
512	}
513}
514
515/*
516 * Check to see if the system paniced, pause and then reboot
517 * according to the specified delay.
518 */
519static void
520shutdown_panic(void *junk, int howto)
521{
522	int loop;
523
524	if (howto & RB_DUMP) {
525		if (panic_reboot_wait_time != 0) {
526			if (panic_reboot_wait_time != -1) {
527				printf("Automatic reboot in %d seconds - "
528				       "press a key on the console to abort\n",
529					panic_reboot_wait_time);
530				for (loop = panic_reboot_wait_time * 10;
531				     loop > 0; --loop) {
532					DELAY(1000 * 100); /* 1/10th second */
533					/* Did user type a key? */
534					if (cncheckc() != -1)
535						break;
536				}
537				if (!loop)
538					return;
539			}
540		} else { /* zero time specified - reboot NOW */
541			return;
542		}
543		printf("--> Press a key on the console to reboot,\n");
544		printf("--> or switch off the system now.\n");
545		cngetc();
546	}
547}
548
549/*
550 * Everything done, now reset
551 */
552static void
553shutdown_reset(void *junk, int howto)
554{
555
556	printf("Rebooting...\n");
557	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
558
559	/*
560	 * Acquiring smp_ipi_mtx here has a double effect:
561	 * - it disables interrupts avoiding CPU0 preemption
562	 *   by fast handlers (thus deadlocking  against other CPUs)
563	 * - it avoids deadlocks against smp_rendezvous() or, more
564	 *   generally, threads busy-waiting, with this spinlock held,
565	 *   and waiting for responses by threads on other CPUs
566	 *   (ie. smp_tlb_shootdown()).
567	 *
568	 * For the !SMP case it just needs to handle the former problem.
569	 */
570#ifdef SMP
571	mtx_lock_spin(&smp_ipi_mtx);
572#else
573	spinlock_enter();
574#endif
575
576	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
577	cpu_reset();
578	/* NOTREACHED */ /* assuming reset worked */
579}
580
581#if defined(WITNESS) || defined(INVARIANT_SUPPORT)
582static int kassert_warn_only = 0;
583#ifdef KDB
584static int kassert_do_kdb = 0;
585#endif
586#ifdef KTR
587static int kassert_do_ktr = 0;
588#endif
589static int kassert_do_log = 1;
590static int kassert_log_pps_limit = 4;
591static int kassert_log_mute_at = 0;
592static int kassert_log_panic_at = 0;
593static int kassert_warnings = 0;
594
595SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
596
597SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
598    &kassert_warn_only, 0,
599    "KASSERT triggers a panic (1) or just a warning (0)");
600
601#ifdef KDB
602SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
603    &kassert_do_kdb, 0, "KASSERT will enter the debugger");
604#endif
605
606#ifdef KTR
607SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
608    &kassert_do_ktr, 0,
609    "KASSERT does a KTR, set this to the KTRMASK you want");
610#endif
611
612SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
613    &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
614
615SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
616    &kassert_warnings, 0, "number of KASSERTs that have been triggered");
617
618SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
619    &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
620
621SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
622    &kassert_log_pps_limit, 0, "limit number of log messages per second");
623
624SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
625    &kassert_log_mute_at, 0, "max number of KASSERTS to log");
626
627static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
628
629SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
630    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
631    kassert_sysctl_kassert, "I", "set to trigger a test kassert");
632
633static int
634kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
635{
636	int error, i;
637
638	error = sysctl_wire_old_buffer(req, sizeof(int));
639	if (error == 0) {
640		i = 0;
641		error = sysctl_handle_int(oidp, &i, 0, req);
642	}
643	if (error != 0 || req->newptr == NULL)
644		return (error);
645	KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
646	return (0);
647}
648
649/*
650 * Called by KASSERT, this decides if we will panic
651 * or if we will log via printf and/or ktr.
652 */
653void
654kassert_panic(const char *fmt, ...)
655{
656	static char buf[256];
657	va_list ap;
658
659	va_start(ap, fmt);
660	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
661	va_end(ap);
662
663	/*
664	 * panic if we're not just warning, or if we've exceeded
665	 * kassert_log_panic_at warnings.
666	 */
667	if (!kassert_warn_only ||
668	    (kassert_log_panic_at > 0 &&
669	     kassert_warnings >= kassert_log_panic_at)) {
670		va_start(ap, fmt);
671		vpanic(fmt, ap);
672		/* NORETURN */
673	}
674#ifdef KTR
675	if (kassert_do_ktr)
676		CTR0(ktr_mask, buf);
677#endif /* KTR */
678	/*
679	 * log if we've not yet met the mute limit.
680	 */
681	if (kassert_do_log &&
682	    (kassert_log_mute_at == 0 ||
683	     kassert_warnings < kassert_log_mute_at)) {
684		static  struct timeval lasterr;
685		static  int curerr;
686
687		if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
688			printf("KASSERT failed: %s\n", buf);
689			kdb_backtrace();
690		}
691	}
692#ifdef KDB
693	if (kassert_do_kdb) {
694		kdb_enter(KDB_WHY_KASSERT, buf);
695	}
696#endif
697	atomic_add_int(&kassert_warnings, 1);
698}
699#endif
700
701/*
702 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
703 * and then reboots.  If we are called twice, then we avoid trying to sync
704 * the disks as this often leads to recursive panics.
705 */
706void
707panic(const char *fmt, ...)
708{
709	va_list ap;
710
711	va_start(ap, fmt);
712	vpanic(fmt, ap);
713}
714
715void
716vpanic(const char *fmt, va_list ap)
717{
718#ifdef SMP
719	cpuset_t other_cpus;
720#endif
721	struct thread *td = curthread;
722	int bootopt, newpanic;
723	static char buf[256];
724
725	spinlock_enter();
726
727#ifdef SMP
728	/*
729	 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
730	 * concurrently entering panic.  Only the winner will proceed
731	 * further.
732	 */
733	if (panicstr == NULL && !kdb_active) {
734		other_cpus = all_cpus;
735		CPU_CLR(PCPU_GET(cpuid), &other_cpus);
736		stop_cpus_hard(other_cpus);
737	}
738#endif
739
740	/*
741	 * Ensure that the scheduler is stopped while panicking, even if panic
742	 * has been entered from kdb.
743	 */
744	td->td_stopsched = 1;
745
746	bootopt = RB_AUTOBOOT;
747	newpanic = 0;
748	if (panicstr)
749		bootopt |= RB_NOSYNC;
750	else {
751		bootopt |= RB_DUMP;
752		panicstr = fmt;
753		newpanic = 1;
754	}
755
756	if (newpanic) {
757		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
758		panicstr = buf;
759		cngrab();
760		printf("panic: %s\n", buf);
761	} else {
762		printf("panic: ");
763		vprintf(fmt, ap);
764		printf("\n");
765	}
766#ifdef SMP
767	printf("cpuid = %d\n", PCPU_GET(cpuid));
768#endif
769
770#ifdef KDB
771	if (newpanic && trace_on_panic)
772		kdb_backtrace();
773	if (debugger_on_panic)
774		kdb_enter(KDB_WHY_PANIC, "panic");
775#endif
776	/*thread_lock(td); */
777	td->td_flags |= TDF_INPANIC;
778	/* thread_unlock(td); */
779	if (!sync_on_panic)
780		bootopt |= RB_NOSYNC;
781	kern_reboot(bootopt);
782}
783
784/*
785 * Support for poweroff delay.
786 *
787 * Please note that setting this delay too short might power off your machine
788 * before the write cache on your hard disk has been flushed, leading to
789 * soft-updates inconsistencies.
790 */
791#ifndef POWEROFF_DELAY
792# define POWEROFF_DELAY 5000
793#endif
794static int poweroff_delay = POWEROFF_DELAY;
795
796SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
797    &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
798
799static void
800poweroff_wait(void *junk, int howto)
801{
802
803	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
804		return;
805	DELAY(poweroff_delay * 1000);
806}
807
808/*
809 * Some system processes (e.g. syncer) need to be stopped at appropriate
810 * points in their main loops prior to a system shutdown, so that they
811 * won't interfere with the shutdown process (e.g. by holding a disk buf
812 * to cause sync to fail).  For each of these system processes, register
813 * shutdown_kproc() as a handler for one of shutdown events.
814 */
815static int kproc_shutdown_wait = 60;
816SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
817    &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
818
819void
820kproc_shutdown(void *arg, int howto)
821{
822	struct proc *p;
823	int error;
824
825	if (panicstr)
826		return;
827
828	p = (struct proc *)arg;
829	printf("Waiting (max %d seconds) for system process `%s' to stop... ",
830	    kproc_shutdown_wait, p->p_comm);
831	error = kproc_suspend(p, kproc_shutdown_wait * hz);
832
833	if (error == EWOULDBLOCK)
834		printf("timed out\n");
835	else
836		printf("done\n");
837}
838
839void
840kthread_shutdown(void *arg, int howto)
841{
842	struct thread *td;
843	int error;
844
845	if (panicstr)
846		return;
847
848	td = (struct thread *)arg;
849	printf("Waiting (max %d seconds) for system thread `%s' to stop... ",
850	    kproc_shutdown_wait, td->td_name);
851	error = kthread_suspend(td, kproc_shutdown_wait * hz);
852
853	if (error == EWOULDBLOCK)
854		printf("timed out\n");
855	else
856		printf("done\n");
857}
858
859static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
860SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
861    dumpdevname, 0, "Device for kernel dumps");
862
863/* Registration of dumpers */
864int
865set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
866{
867	size_t wantcopy;
868	int error;
869
870	error = priv_check(td, PRIV_SETDUMPER);
871	if (error != 0)
872		return (error);
873
874	if (di == NULL) {
875		if (dumper.blockbuf != NULL)
876			free(dumper.blockbuf, M_DUMPER);
877		bzero(&dumper, sizeof(dumper));
878		dumpdevname[0] = '\0';
879		return (0);
880	}
881	if (dumper.dumper != NULL)
882		return (EBUSY);
883	dumper = *di;
884	wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
885	if (wantcopy >= sizeof(dumpdevname)) {
886		printf("set_dumper: device name truncated from '%s' -> '%s'\n",
887			devname, dumpdevname);
888	}
889	dumper.blockbuf = malloc(di->blocksize, M_DUMPER, M_WAITOK | M_ZERO);
890	return (0);
891}
892
893/* Call dumper with bounds checking. */
894int
895dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
896    off_t offset, size_t length)
897{
898
899	if (length != 0 && (offset < di->mediaoffset ||
900	    offset - di->mediaoffset + length > di->mediasize)) {
901		printf("Attempt to write outside dump device boundaries.\n"
902	    "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
903		    (intmax_t)offset, (intmax_t)di->mediaoffset,
904		    (uintmax_t)length, (intmax_t)di->mediasize);
905		return (ENOSPC);
906	}
907	return (di->dumper(di->priv, virtual, physical, offset, length));
908}
909
910/* Call dumper with bounds checking. */
911int
912dump_write_pad(struct dumperinfo *di, void *virtual, vm_offset_t physical,
913    off_t offset, size_t length, size_t *size)
914{
915	char *temp;
916	int ret;
917
918	if (length > di->blocksize)
919		return (ENOMEM);
920
921	*size = di->blocksize;
922	if (length == di->blocksize)
923		temp = virtual;
924	else {
925		temp = di->blockbuf;
926		memset(temp + length, 0, di->blocksize - length);
927		memcpy(temp, virtual, length);
928	}
929	ret = dump_write(di, temp, physical, offset, *size);
930
931	return (ret);
932}
933
934
935void
936mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
937    uint64_t dumplen, uint32_t blksz)
938{
939	size_t dstsize;
940
941	bzero(kdh, sizeof(*kdh));
942	strlcpy(kdh->magic, magic, sizeof(kdh->magic));
943	strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
944	kdh->version = htod32(KERNELDUMPVERSION);
945	kdh->architectureversion = htod32(archver);
946	kdh->dumplength = htod64(dumplen);
947	kdh->dumptime = htod64(time_second);
948	kdh->blocksize = htod32(blksz);
949	strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
950	dstsize = sizeof(kdh->versionstring);
951	if (strlcpy(kdh->versionstring, version, dstsize) >= dstsize)
952		kdh->versionstring[dstsize - 2] = '\n';
953	if (panicstr != NULL)
954		strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
955	kdh->parity = kerneldump_parity(kdh);
956}
957
958#ifdef DDB
959DB_SHOW_COMMAND(panic, db_show_panic)
960{
961
962	if (panicstr == NULL)
963		db_printf("panicstr not set\n");
964	else
965		db_printf("panic: %s\n", panicstr);
966}
967#endif
968