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/pcb.h>
78#include <machine/smp.h>
79
80#include <security/mac/mac_framework.h>
81
82#include <vm/vm.h>
83#include <vm/vm_object.h>
84#include <vm/vm_page.h>
85#include <vm/vm_pager.h>
86#include <vm/swap_pager.h>
87
88#include <sys/signalvar.h>
89
90#ifndef PANIC_REBOOT_WAIT_TIME
91#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
92#endif
93static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
94SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RW | CTLFLAG_TUN,
95    &panic_reboot_wait_time, 0,
96    "Seconds to wait before rebooting after a panic");
97TUNABLE_INT("kern.panic_reboot_wait_time", &panic_reboot_wait_time);
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
107int debugger_on_panic = 0;
108#else
109int debugger_on_panic = 1;
110#endif
111SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
112    CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN,
113    &debugger_on_panic, 0, "Run debugger on kernel panic");
114TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
115
116#ifdef KDB_TRACE
117static int trace_on_panic = 1;
118#else
119static int trace_on_panic = 0;
120#endif
121SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
122    CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN,
123    &trace_on_panic, 0, "Print stack trace on kernel panic");
124TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
125#endif /* KDB */
126
127static int sync_on_panic = 0;
128SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
129	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
130TUNABLE_INT("kern.sync_on_panic", &sync_on_panic);
131
132static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
133    "Shutdown environment");
134
135#ifndef DIAGNOSTIC
136static int show_busybufs;
137#else
138static int show_busybufs = 1;
139#endif
140SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
141	&show_busybufs, 0, "");
142
143int suspend_blocked = 0;
144SYSCTL_INT(_kern, OID_AUTO, suspend_blocked, CTLFLAG_RW,
145	&suspend_blocked, 0, "Block suspend due to a pending shutdown");
146
147/*
148 * Variable panicstr contains argument to first call to panic; used as flag
149 * to indicate that the kernel has already called panic.
150 */
151const char *panicstr;
152
153int dumping;				/* system is dumping */
154int rebooting;				/* system is rebooting */
155static struct dumperinfo dumper;	/* our selected dumper */
156
157/* Context information for dump-debuggers. */
158static struct pcb dumppcb;		/* Registers. */
159lwpid_t dumptid;			/* Thread ID. */
160
161static struct cdevsw reroot_cdevsw = {
162     .d_version = D_VERSION,
163     .d_name    = "reroot",
164};
165
166static void poweroff_wait(void *, int);
167static void shutdown_halt(void *junk, int howto);
168static void shutdown_panic(void *junk, int howto);
169static void shutdown_reset(void *junk, int howto);
170static int kern_reroot(void);
171
172/* register various local shutdown events */
173static void
174shutdown_conf(void *unused)
175{
176
177	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
178	    SHUTDOWN_PRI_FIRST);
179	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
180	    SHUTDOWN_PRI_LAST + 100);
181	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
182	    SHUTDOWN_PRI_LAST + 100);
183	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
184	    SHUTDOWN_PRI_LAST + 200);
185}
186
187SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
188
189/*
190 * The only reason this exists is to create the /dev/reroot/ directory,
191 * used by reroot code in init(8) as a mountpoint for tmpfs.
192 */
193static void
194reroot_conf(void *unused)
195{
196	int error;
197	struct cdev *cdev;
198
199	error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev,
200	    &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot");
201	if (error != 0) {
202		printf("%s: failed to create device node, error %d",
203		    __func__, error);
204	}
205}
206
207SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL);
208
209/*
210 * The system call that results in a reboot.
211 */
212/* ARGSUSED */
213int
214sys_reboot(struct thread *td, struct reboot_args *uap)
215{
216	int error;
217
218	error = 0;
219#ifdef MAC
220	error = mac_system_check_reboot(td->td_ucred, uap->opt);
221#endif
222	if (error == 0)
223		error = priv_check(td, PRIV_REBOOT);
224	if (error == 0) {
225		if (uap->opt & RB_REROOT) {
226			error = kern_reroot();
227		} else {
228			mtx_lock(&Giant);
229			kern_reboot(uap->opt);
230			mtx_unlock(&Giant);
231		}
232	}
233	return (error);
234}
235
236/*
237 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
238 */
239static int shutdown_howto = 0;
240
241void
242shutdown_nice(int howto)
243{
244
245	shutdown_howto = howto;
246
247	/* Send a signal to init(8) and have it shutdown the world */
248	if (initproc != NULL) {
249		PROC_LOCK(initproc);
250		kern_psignal(initproc, SIGINT);
251		PROC_UNLOCK(initproc);
252	} else {
253		/* No init(8) running, so simply reboot */
254		kern_reboot(RB_NOSYNC);
255	}
256	return;
257}
258static int	waittime = -1;
259
260static void
261print_uptime(void)
262{
263	int f;
264	struct timespec ts;
265
266	getnanouptime(&ts);
267	printf("Uptime: ");
268	f = 0;
269	if (ts.tv_sec >= 86400) {
270		printf("%ldd", (long)ts.tv_sec / 86400);
271		ts.tv_sec %= 86400;
272		f = 1;
273	}
274	if (f || ts.tv_sec >= 3600) {
275		printf("%ldh", (long)ts.tv_sec / 3600);
276		ts.tv_sec %= 3600;
277		f = 1;
278	}
279	if (f || ts.tv_sec >= 60) {
280		printf("%ldm", (long)ts.tv_sec / 60);
281		ts.tv_sec %= 60;
282		f = 1;
283	}
284	printf("%lds\n", (long)ts.tv_sec);
285}
286
287int
288doadump(boolean_t textdump)
289{
290	boolean_t coredump;
291
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		dumpsys(&dumper);
310
311	dumping--;
312	return (0);
313}
314
315static int
316isbufbusy(struct buf *bp)
317{
318	if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
319	    BUF_ISLOCKED(bp)) ||
320	    ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
321		return (1);
322	return (0);
323}
324
325/*
326 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
327 */
328void
329kern_reboot(int howto)
330{
331	static int first_buf_printf = 1;
332
333#if defined(SMP)
334	/*
335	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
336	 * systems don't shutdown properly (i.e., ACPI power off) if we
337	 * run on another processor.
338	 */
339	if (!SCHEDULER_STOPPED()) {
340		thread_lock(curthread);
341		sched_bind(curthread, 0);
342		thread_unlock(curthread);
343		KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
344	}
345#endif
346	/* We're in the process of rebooting. */
347	rebooting = 1;
348
349	/* collect extra flags that shutdown_nice might have set */
350	howto |= shutdown_howto;
351
352	/* We are out of the debugger now. */
353	kdb_active = 0;
354
355	/*
356	 * Do any callouts that should be done BEFORE syncing the filesystems.
357	 */
358	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
359
360	/*
361	 * Now sync filesystems
362	 */
363	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
364		register struct buf *bp;
365		int iter, nbusy, pbusy;
366#ifndef PREEMPTION
367		int subiter;
368#endif
369
370		waittime = 0;
371
372		wdog_kern_pat(WD_LASTVAL);
373		sys_sync(curthread, NULL);
374
375		/*
376		 * With soft updates, some buffers that are
377		 * written will be remarked as dirty until other
378		 * buffers are written.
379		 */
380		for (iter = pbusy = 0; iter < 20; iter++) {
381			nbusy = 0;
382			for (bp = &buf[nbuf]; --bp >= buf; )
383				if (isbufbusy(bp))
384					nbusy++;
385			if (nbusy == 0) {
386				if (first_buf_printf)
387					printf("All buffers synced.");
388				break;
389			}
390			if (first_buf_printf) {
391				printf("Syncing disks, buffers remaining... ");
392				first_buf_printf = 0;
393			}
394			printf("%d ", nbusy);
395			if (nbusy < pbusy)
396				iter = 0;
397			pbusy = nbusy;
398
399			wdog_kern_pat(WD_LASTVAL);
400			sys_sync(curthread, NULL);
401
402#ifdef PREEMPTION
403			/*
404			 * Drop Giant and spin for a while to allow
405			 * interrupt threads to run.
406			 */
407			DROP_GIANT();
408			DELAY(50000 * iter);
409			PICKUP_GIANT();
410#else
411			/*
412			 * Drop Giant and context switch several times to
413			 * allow interrupt threads to run.
414			 */
415			DROP_GIANT();
416			for (subiter = 0; subiter < 50 * iter; subiter++) {
417				thread_lock(curthread);
418				mi_switch(SW_VOL, NULL);
419				thread_unlock(curthread);
420				DELAY(1000);
421			}
422			PICKUP_GIANT();
423#endif
424		}
425		printf("\n");
426		/*
427		 * Count only busy local buffers to prevent forcing
428		 * a fsck if we're just a client of a wedged NFS server
429		 */
430		nbusy = 0;
431		for (bp = &buf[nbuf]; --bp >= buf; ) {
432			if (isbufbusy(bp)) {
433#if 0
434/* XXX: This is bogus.  We should probably have a BO_REMOTE flag instead */
435				if (bp->b_dev == NULL) {
436					TAILQ_REMOVE(&mountlist,
437					    bp->b_vp->v_mount, mnt_list);
438					continue;
439				}
440#endif
441				nbusy++;
442				if (show_busybufs > 0) {
443					printf(
444	    "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
445					    nbusy, bp, bp->b_vp, bp->b_flags,
446					    (intmax_t)bp->b_blkno,
447					    (intmax_t)bp->b_lblkno);
448					BUF_LOCKPRINTINFO(bp);
449					if (show_busybufs > 1)
450						vn_printf(bp->b_vp,
451						    "vnode content: ");
452				}
453			}
454		}
455		if (nbusy) {
456			/*
457			 * Failed to sync all blocks. Indicate this and don't
458			 * unmount filesystems (thus forcing an fsck on reboot).
459			 */
460			printf("Giving up on %d buffers\n", nbusy);
461			DELAY(5000000);	/* 5 seconds */
462		} else {
463			if (!first_buf_printf)
464				printf("Final sync complete\n");
465			/*
466			 * Unmount filesystems
467			 */
468			if (panicstr == 0)
469				vfs_unmountall();
470		}
471		swapoff_all();
472		DELAY(100000);		/* wait for console output to finish */
473	}
474
475	print_uptime();
476
477	cngrab();
478
479	/*
480	 * Ok, now do things that assume all filesystem activity has
481	 * been completed.
482	 */
483	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
484
485	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
486		doadump(TRUE);
487
488	/* Now that we're going to really halt the system... */
489	EVENTHANDLER_INVOKE(shutdown_final, howto);
490
491	for(;;) ;	/* safety against shutdown_reset not working */
492	/* NOTREACHED */
493}
494
495/*
496 * The system call that results in changing the rootfs.
497 */
498static int
499kern_reroot(void)
500{
501	struct vnode *oldrootvnode, *vp;
502	struct mount *mp, *devmp;
503	int error;
504
505	if (curproc != initproc)
506		return (EPERM);
507
508	/*
509	 * Mark the filesystem containing currently-running executable
510	 * (the temporary copy of init(8)) busy.
511	 */
512	vp = curproc->p_textvp;
513	error = vn_lock(vp, LK_SHARED);
514	if (error != 0)
515		return (error);
516	mp = vp->v_mount;
517	error = vfs_busy(mp, MBF_NOWAIT);
518	if (error != 0) {
519		vfs_ref(mp);
520		VOP_UNLOCK(vp, 0);
521		error = vfs_busy(mp, 0);
522		vn_lock(vp, LK_SHARED | LK_RETRY);
523		vfs_rel(mp);
524		if (error != 0) {
525			VOP_UNLOCK(vp, 0);
526			return (ENOENT);
527		}
528		if (vp->v_iflag & VI_DOOMED) {
529			VOP_UNLOCK(vp, 0);
530			vfs_unbusy(mp);
531			return (ENOENT);
532		}
533	}
534	VOP_UNLOCK(vp, 0);
535
536	/*
537	 * Remove the filesystem containing currently-running executable
538	 * from the mount list, to prevent it from being unmounted
539	 * by vfs_unmountall(), and to avoid confusing vfs_mountroot().
540	 *
541	 * Also preserve /dev - forcibly unmounting it could cause driver
542	 * reinitialization.
543	 */
544
545	vfs_ref(rootdevmp);
546	devmp = rootdevmp;
547	rootdevmp = NULL;
548
549	mtx_lock(&mountlist_mtx);
550	TAILQ_REMOVE(&mountlist, mp, mnt_list);
551	TAILQ_REMOVE(&mountlist, devmp, mnt_list);
552	mtx_unlock(&mountlist_mtx);
553
554	oldrootvnode = rootvnode;
555
556	/*
557	 * Unmount everything except for the two filesystems preserved above.
558	 */
559	vfs_unmountall();
560
561	/*
562	 * Add /dev back; vfs_mountroot() will move it into its new place.
563	 */
564	mtx_lock(&mountlist_mtx);
565	TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list);
566	mtx_unlock(&mountlist_mtx);
567	rootdevmp = devmp;
568	vfs_rel(rootdevmp);
569
570	/*
571	 * Mount the new rootfs.
572	 */
573	vfs_mountroot();
574
575	/*
576	 * Update all references to the old rootvnode.
577	 */
578	mountcheckdirs(oldrootvnode, rootvnode);
579
580	/*
581	 * Add the temporary filesystem back and unbusy it.
582	 */
583	mtx_lock(&mountlist_mtx);
584	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
585	mtx_unlock(&mountlist_mtx);
586	vfs_unbusy(mp);
587
588	return (0);
589}
590
591/*
592 * If the shutdown was a clean halt, behave accordingly.
593 */
594static void
595shutdown_halt(void *junk, int howto)
596{
597
598	if (howto & RB_HALT) {
599		printf("\n");
600		printf("The operating system has halted.\n");
601		printf("Please press any key to reboot.\n\n");
602		switch (cngetc()) {
603		case -1:		/* No console, just die */
604			cpu_halt();
605			/* NOTREACHED */
606		default:
607			howto &= ~RB_HALT;
608			break;
609		}
610	}
611}
612
613/*
614 * Check to see if the system paniced, pause and then reboot
615 * according to the specified delay.
616 */
617static void
618shutdown_panic(void *junk, int howto)
619{
620	int loop;
621
622	if (howto & RB_DUMP) {
623		if (panic_reboot_wait_time != 0) {
624			if (panic_reboot_wait_time != -1) {
625				printf("Automatic reboot in %d seconds - "
626				       "press a key on the console to abort\n",
627					panic_reboot_wait_time);
628				for (loop = panic_reboot_wait_time * 10;
629				     loop > 0; --loop) {
630					DELAY(1000 * 100); /* 1/10th second */
631					/* Did user type a key? */
632					if (cncheckc() != -1)
633						break;
634				}
635				if (!loop)
636					return;
637			}
638		} else { /* zero time specified - reboot NOW */
639			return;
640		}
641		printf("--> Press a key on the console to reboot,\n");
642		printf("--> or switch off the system now.\n");
643		cngetc();
644	}
645}
646
647/*
648 * Everything done, now reset
649 */
650static void
651shutdown_reset(void *junk, int howto)
652{
653
654	printf("Rebooting...\n");
655	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
656
657	/*
658	 * Acquiring smp_ipi_mtx here has a double effect:
659	 * - it disables interrupts avoiding CPU0 preemption
660	 *   by fast handlers (thus deadlocking  against other CPUs)
661	 * - it avoids deadlocks against smp_rendezvous() or, more
662	 *   generally, threads busy-waiting, with this spinlock held,
663	 *   and waiting for responses by threads on other CPUs
664	 *   (ie. smp_tlb_shootdown()).
665	 *
666	 * For the !SMP case it just needs to handle the former problem.
667	 */
668#ifdef SMP
669	mtx_lock_spin(&smp_ipi_mtx);
670#else
671	spinlock_enter();
672#endif
673
674	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
675	cpu_reset();
676	/* NOTREACHED */ /* assuming reset worked */
677}
678
679#if defined(WITNESS) || defined(INVARIANTS)
680static int kassert_warn_only = 0;
681#ifdef KDB
682static int kassert_do_kdb = 0;
683#endif
684#ifdef KTR
685static int kassert_do_ktr = 0;
686#endif
687static int kassert_do_log = 1;
688static int kassert_log_pps_limit = 4;
689static int kassert_log_mute_at = 0;
690static int kassert_log_panic_at = 0;
691static int kassert_warnings = 0;
692
693SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
694
695SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RW | CTLFLAG_TUN,
696    &kassert_warn_only, 0,
697    "KASSERT triggers a panic (1) or just a warning (0)");
698TUNABLE_INT("debug.kassert.warn_only", &kassert_warn_only);
699
700#ifdef KDB
701SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RW | CTLFLAG_TUN,
702    &kassert_do_kdb, 0, "KASSERT will enter the debugger");
703TUNABLE_INT("debug.kassert.do_kdb", &kassert_do_kdb);
704#endif
705
706#ifdef KTR
707SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RW | CTLFLAG_TUN,
708    &kassert_do_ktr, 0,
709    "KASSERT does a KTR, set this to the KTRMASK you want");
710TUNABLE_INT("debug.kassert.do_ktr", &kassert_do_ktr);
711#endif
712
713SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RW | CTLFLAG_TUN,
714    &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
715TUNABLE_INT("debug.kassert.do_log", &kassert_do_log);
716
717SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RW | CTLFLAG_TUN,
718    &kassert_warnings, 0, "number of KASSERTs that have been triggered");
719TUNABLE_INT("debug.kassert.warnings", &kassert_warnings);
720
721SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RW | CTLFLAG_TUN,
722    &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
723TUNABLE_INT("debug.kassert.log_panic_at", &kassert_log_panic_at);
724
725SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RW | CTLFLAG_TUN,
726    &kassert_log_pps_limit, 0, "limit number of log messages per second");
727TUNABLE_INT("debug.kassert.log_pps_limit", &kassert_log_pps_limit);
728
729SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RW | CTLFLAG_TUN,
730    &kassert_log_mute_at, 0, "max number of KASSERTS to log");
731TUNABLE_INT("debug.kassert.log_mute_at", &kassert_log_mute_at);
732
733static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
734
735SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
736    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
737    kassert_sysctl_kassert, "I", "set to trigger a test kassert");
738
739static int
740kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
741{
742	int error, i;
743
744	error = sysctl_wire_old_buffer(req, sizeof(int));
745	if (error == 0) {
746		i = 0;
747		error = sysctl_handle_int(oidp, &i, 0, req);
748	}
749	if (error != 0 || req->newptr == NULL)
750		return (error);
751	KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
752	return (0);
753}
754
755/*
756 * Called by KASSERT, this decides if we will panic
757 * or if we will log via printf and/or ktr.
758 */
759void
760kassert_panic(const char *fmt, ...)
761{
762	static char buf[256];
763	va_list ap;
764
765	va_start(ap, fmt);
766	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
767	va_end(ap);
768
769	/*
770	 * panic if we're not just warning, or if we've exceeded
771	 * kassert_log_panic_at warnings.
772	 */
773	if (!kassert_warn_only ||
774	    (kassert_log_panic_at > 0 &&
775	     kassert_warnings >= kassert_log_panic_at)) {
776		va_start(ap, fmt);
777		vpanic(fmt, ap);
778		/* NORETURN */
779	}
780#ifdef KTR
781	if (kassert_do_ktr)
782		CTR0(ktr_mask, buf);
783#endif /* KTR */
784	/*
785	 * log if we've not yet met the mute limit.
786	 */
787	if (kassert_do_log &&
788	    (kassert_log_mute_at == 0 ||
789	     kassert_warnings < kassert_log_mute_at)) {
790		static  struct timeval lasterr;
791		static  int curerr;
792
793		if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
794			printf("KASSERT failed: %s\n", buf);
795			kdb_backtrace();
796		}
797	}
798#ifdef KDB
799	if (kassert_do_kdb) {
800		kdb_enter(KDB_WHY_KASSERT, buf);
801	}
802#endif
803	atomic_add_int(&kassert_warnings, 1);
804}
805#endif
806
807/*
808 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
809 * and then reboots.  If we are called twice, then we avoid trying to sync
810 * the disks as this often leads to recursive panics.
811 */
812void
813panic(const char *fmt, ...)
814{
815	va_list ap;
816
817	va_start(ap, fmt);
818	vpanic(fmt, ap);
819}
820
821void
822vpanic(const char *fmt, va_list ap)
823{
824#ifdef SMP
825	cpuset_t other_cpus;
826#endif
827	struct thread *td = curthread;
828	int bootopt, newpanic;
829	static char buf[256];
830
831	spinlock_enter();
832
833#ifdef SMP
834	/*
835	 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
836	 * concurrently entering panic.  Only the winner will proceed
837	 * further.
838	 */
839	if (panicstr == NULL && !kdb_active) {
840		other_cpus = all_cpus;
841		CPU_CLR(PCPU_GET(cpuid), &other_cpus);
842		stop_cpus_hard(other_cpus);
843	}
844
845	/*
846	 * We set stop_scheduler here and not in the block above,
847	 * because we want to ensure that if panic has been called and
848	 * stop_scheduler_on_panic is true, then stop_scheduler will
849	 * always be set.  Even if panic has been entered from kdb.
850	 */
851	td->td_stopsched = 1;
852#endif
853
854	bootopt = RB_AUTOBOOT;
855	newpanic = 0;
856	if (panicstr)
857		bootopt |= RB_NOSYNC;
858	else {
859		bootopt |= RB_DUMP;
860		panicstr = fmt;
861		newpanic = 1;
862	}
863
864	if (newpanic) {
865		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
866		panicstr = buf;
867		cngrab();
868		printf("panic: %s\n", buf);
869	} else {
870		printf("panic: ");
871		vprintf(fmt, ap);
872		printf("\n");
873	}
874#ifdef SMP
875	printf("cpuid = %d\n", PCPU_GET(cpuid));
876#endif
877
878#ifdef KDB
879	if (newpanic && trace_on_panic)
880		kdb_backtrace();
881	if (debugger_on_panic)
882		kdb_enter(KDB_WHY_PANIC, "panic");
883#endif
884	/*thread_lock(td); */
885	td->td_flags |= TDF_INPANIC;
886	/* thread_unlock(td); */
887	if (!sync_on_panic)
888		bootopt |= RB_NOSYNC;
889	kern_reboot(bootopt);
890}
891
892/*
893 * Support for poweroff delay.
894 *
895 * Please note that setting this delay too short might power off your machine
896 * before the write cache on your hard disk has been flushed, leading to
897 * soft-updates inconsistencies.
898 */
899#ifndef POWEROFF_DELAY
900# define POWEROFF_DELAY 5000
901#endif
902static int poweroff_delay = POWEROFF_DELAY;
903
904SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
905    &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
906
907static void
908poweroff_wait(void *junk, int howto)
909{
910
911	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
912		return;
913	DELAY(poweroff_delay * 1000);
914}
915
916/*
917 * Some system processes (e.g. syncer) need to be stopped at appropriate
918 * points in their main loops prior to a system shutdown, so that they
919 * won't interfere with the shutdown process (e.g. by holding a disk buf
920 * to cause sync to fail).  For each of these system processes, register
921 * shutdown_kproc() as a handler for one of shutdown events.
922 */
923static int kproc_shutdown_wait = 60;
924SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
925    &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
926
927void
928kproc_shutdown(void *arg, int howto)
929{
930	struct proc *p;
931	int error;
932
933	if (panicstr)
934		return;
935
936	p = (struct proc *)arg;
937	printf("Waiting (max %d seconds) for system process `%s' to stop...",
938	    kproc_shutdown_wait, p->p_comm);
939	error = kproc_suspend(p, kproc_shutdown_wait * hz);
940
941	if (error == EWOULDBLOCK)
942		printf("timed out\n");
943	else
944		printf("done\n");
945}
946
947void
948kthread_shutdown(void *arg, int howto)
949{
950	struct thread *td;
951	int error;
952
953	if (panicstr)
954		return;
955
956	td = (struct thread *)arg;
957	printf("Waiting (max %d seconds) for system thread `%s' to stop...",
958	    kproc_shutdown_wait, td->td_name);
959	error = kthread_suspend(td, kproc_shutdown_wait * hz);
960
961	if (error == EWOULDBLOCK)
962		printf("timed out\n");
963	else
964		printf("done\n");
965}
966
967static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
968SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
969    dumpdevname, 0, "Device for kernel dumps");
970
971/* Registration of dumpers */
972int
973set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
974{
975	size_t wantcopy;
976	int error;
977
978	error = priv_check(td, PRIV_SETDUMPER);
979	if (error != 0)
980		return (error);
981
982	if (di == NULL) {
983		bzero(&dumper, sizeof dumper);
984		dumpdevname[0] = '\0';
985		return (0);
986	}
987	if (dumper.dumper != NULL)
988		return (EBUSY);
989	dumper = *di;
990	wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
991	if (wantcopy >= sizeof(dumpdevname)) {
992		printf("set_dumper: device name truncated from '%s' -> '%s'\n",
993			devname, dumpdevname);
994	}
995	return (0);
996}
997
998/* Call dumper with bounds checking. */
999int
1000dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
1001    off_t offset, size_t length)
1002{
1003
1004	if (length != 0 && (offset < di->mediaoffset ||
1005	    offset - di->mediaoffset + length > di->mediasize)) {
1006		printf("Attempt to write outside dump device boundaries.\n"
1007	    "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
1008		    (intmax_t)offset, (intmax_t)di->mediaoffset,
1009		    (uintmax_t)length, (intmax_t)di->mediasize);
1010		return (ENOSPC);
1011	}
1012	return (di->dumper(di->priv, virtual, physical, offset, length));
1013}
1014
1015void
1016mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
1017    uint64_t dumplen, uint32_t blksz)
1018{
1019
1020	bzero(kdh, sizeof(*kdh));
1021	strlcpy(kdh->magic, magic, sizeof(kdh->magic));
1022	strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
1023	kdh->version = htod32(KERNELDUMPVERSION);
1024	kdh->architectureversion = htod32(archver);
1025	kdh->dumplength = htod64(dumplen);
1026	kdh->dumptime = htod64(time_second);
1027	kdh->blocksize = htod32(blksz);
1028	strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
1029	strlcpy(kdh->versionstring, version, sizeof(kdh->versionstring));
1030	if (panicstr != NULL)
1031		strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
1032	kdh->parity = kerneldump_parity(kdh);
1033}
1034
1035#ifdef DDB
1036DB_SHOW_COMMAND(panic, db_show_panic)
1037{
1038
1039	if (panicstr == NULL)
1040		db_printf("panicstr not set\n");
1041	else
1042		db_printf("panic: %s\n", panicstr);
1043}
1044#endif
1045