kern_shutdown.c revision 132171 
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: head/sys/kern/kern_shutdown.c 132171 2004-07-15 03:20:08Z njl $");
39
40#include "opt_kdb.h"
41#include "opt_hw_wdog.h"
42#include "opt_mac.h"
43#include "opt_panic.h"
44#include "opt_show_busybufs.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/kdb.h>
54#include <sys/kernel.h>
55#include <sys/kthread.h>
56#include <sys/mac.h>
57#include <sys/malloc.h>
58#include <sys/mount.h>
59#include <sys/proc.h>
60#include <sys/reboot.h>
61#include <sys/resourcevar.h>
62#include <sys/smp.h>		/* smp_active */
63#include <sys/sysctl.h>
64#include <sys/sysproto.h>
65#include <sys/vnode.h>
66
67#include <machine/cpu.h>
68#include <machine/pcb.h>
69#include <machine/smp.h>
70
71#include <sys/signalvar.h>
72
73#ifndef PANIC_REBOOT_WAIT_TIME
74#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
75#endif
76
77/*
78 * Note that stdarg.h and the ANSI style va_start macro is used for both
79 * ANSI and traditional C compilers.
80 */
81#include <machine/stdarg.h>
82
83#ifdef KDB
84#ifdef KDB_UNATTENDED
85int debugger_on_panic = 0;
86#else
87int debugger_on_panic = 1;
88#endif
89SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
90	&debugger_on_panic, 0, "Run debugger on kernel panic");
91
92#ifdef KDB_TRACE
93int trace_on_panic = 1;
94#else
95int trace_on_panic = 0;
96#endif
97SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
98	&trace_on_panic, 0, "Print stack trace on kernel panic");
99#endif /* KDB */
100
101int sync_on_panic = 1;
102SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
103	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
104
105SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
106
107#ifdef	HW_WDOG
108/*
109 * If there is a hardware watchdog, point this at the function needed to
110 * hold it off.
111 * It's needed when the kernel needs to do some lengthy operations.
112 * e.g. in wd.c when dumping core.. It's most annoying to have
113 * your precious core-dump only half written because the wdog kicked in.
114 */
115watchdog_tickle_fn wdog_tickler = NULL;
116#endif	/* HW_WDOG */
117
118/*
119 * Variable panicstr contains argument to first call to panic; used as flag
120 * to indicate that the kernel has already called panic.
121 */
122const char *panicstr;
123
124int dumping;				/* system is dumping */
125static struct dumperinfo dumper;	/* our selected dumper */
126
127/* Context information for dump-debuggers. */
128static struct pcb dumppcb;		/* Registers. */
129static lwpid_t dumptid;			/* Thread ID. */
130
131static void boot(int) __dead2;
132static void poweroff_wait(void *, int);
133static void shutdown_halt(void *junk, int howto);
134static void shutdown_panic(void *junk, int howto);
135static void shutdown_reset(void *junk, int howto);
136
137/* register various local shutdown events */
138static void
139shutdown_conf(void *unused)
140{
141
142	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
143	    SHUTDOWN_PRI_FIRST);
144	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
145	    SHUTDOWN_PRI_LAST + 100);
146	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
147	    SHUTDOWN_PRI_LAST + 100);
148	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
149	    SHUTDOWN_PRI_LAST + 200);
150}
151
152SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
153
154/*
155 * The system call that results in a reboot
156 *
157 * MPSAFE
158 */
159/* ARGSUSED */
160int
161reboot(struct thread *td, struct reboot_args *uap)
162{
163	int error;
164
165	error = 0;
166#ifdef MAC
167	error = mac_check_system_reboot(td->td_ucred, uap->opt);
168#endif
169	if (error == 0)
170		error = suser(td);
171	if (error == 0) {
172		mtx_lock(&Giant);
173		boot(uap->opt);
174		mtx_unlock(&Giant);
175	}
176	return (error);
177}
178
179/*
180 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
181 */
182static int shutdown_howto = 0;
183
184void
185shutdown_nice(int howto)
186{
187
188	shutdown_howto = howto;
189
190	/* Send a signal to init(8) and have it shutdown the world */
191	if (initproc != NULL) {
192		PROC_LOCK(initproc);
193		psignal(initproc, SIGINT);
194		PROC_UNLOCK(initproc);
195	} else {
196		/* No init(8) running, so simply reboot */
197		boot(RB_NOSYNC);
198	}
199	return;
200}
201static int	waittime = -1;
202
203static void
204print_uptime(void)
205{
206	int f;
207	struct timespec ts;
208
209	getnanouptime(&ts);
210	printf("Uptime: ");
211	f = 0;
212	if (ts.tv_sec >= 86400) {
213		printf("%ldd", (long)ts.tv_sec / 86400);
214		ts.tv_sec %= 86400;
215		f = 1;
216	}
217	if (f || ts.tv_sec >= 3600) {
218		printf("%ldh", (long)ts.tv_sec / 3600);
219		ts.tv_sec %= 3600;
220		f = 1;
221	}
222	if (f || ts.tv_sec >= 60) {
223		printf("%ldm", (long)ts.tv_sec / 60);
224		ts.tv_sec %= 60;
225		f = 1;
226	}
227	printf("%lds\n", (long)ts.tv_sec);
228}
229
230static void
231doadump(void)
232{
233
234	savectx(&dumppcb);
235	dumptid = curthread->td_tid;
236	dumping++;
237	dumpsys(&dumper);
238}
239
240/*
241 *  Go through the rigmarole of shutting down..
242 * this used to be in machdep.c but I'll be dammned if I could see
243 * anything machine dependant in it.
244 */
245static void
246boot(int howto)
247{
248
249	/* collect extra flags that shutdown_nice might have set */
250	howto |= shutdown_howto;
251
252	/* We are out of the debugger now. */
253	kdb_active = 0;
254
255#ifdef SMP
256	if (smp_active)
257		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
258#endif
259	/*
260	 * Do any callouts that should be done BEFORE syncing the filesystems.
261	 */
262	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
263
264	/*
265	 * Now sync filesystems
266	 */
267	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
268		register struct buf *bp;
269		int iter, nbusy, pbusy;
270#ifndef PREEMPTION
271		int subiter;
272#endif
273
274		waittime = 0;
275		printf("syncing disks, buffers remaining... ");
276
277		sync(&thread0, NULL);
278
279		/*
280		 * With soft updates, some buffers that are
281		 * written will be remarked as dirty until other
282		 * buffers are written.
283		 */
284		for (iter = pbusy = 0; iter < 20; iter++) {
285			nbusy = 0;
286			for (bp = &buf[nbuf]; --bp >= buf; ) {
287				if ((bp->b_flags & B_INVAL) == 0 &&
288				    BUF_REFCNT(bp) > 0) {
289					nbusy++;
290				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
291						== B_DELWRI) {
292					/* bawrite(bp);*/
293					nbusy++;
294				}
295			}
296			if (nbusy == 0)
297				break;
298			printf("%d ", nbusy);
299			if (nbusy < pbusy)
300				iter = 0;
301			pbusy = nbusy;
302			sync(&thread0, NULL);
303
304#ifdef PREEMPTION
305			/*
306			 * Drop Giant and spin for a while to allow
307			 * interrupt threads to run.
308			 */
309			DROP_GIANT();
310			DELAY(50000 * iter);
311			PICKUP_GIANT();
312#else
313			/*
314			 * Drop Giant and context switch several times to
315			 * allow interrupt threads to run.
316			 */
317			DROP_GIANT();
318			for (subiter = 0; subiter < 50 * iter; subiter++) {
319				mtx_lock_spin(&sched_lock);
320				mi_switch(SW_VOL, NULL);
321				mtx_unlock_spin(&sched_lock);
322				DELAY(1000);
323			}
324			PICKUP_GIANT();
325#endif
326		}
327
328		/*
329		 * Count only busy local buffers to prevent forcing
330		 * a fsck if we're just a client of a wedged NFS server
331		 */
332		nbusy = 0;
333		for (bp = &buf[nbuf]; --bp >= buf; ) {
334			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
335			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
336				if (bp->b_dev == NULL) {
337					TAILQ_REMOVE(&mountlist,
338					    bp->b_vp->v_mount, mnt_list);
339					continue;
340				}
341				nbusy++;
342#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
343				printf(
344			    "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n",
345				    nbusy, devtoname(bp->b_dev),
346				    bp->b_flags, (long)bp->b_blkno,
347				    (long)bp->b_lblkno);
348#endif
349			}
350		}
351		if (nbusy) {
352			/*
353			 * Failed to sync all blocks. Indicate this and don't
354			 * unmount filesystems (thus forcing an fsck on reboot).
355			 */
356			printf("giving up on %d buffers\n", nbusy);
357			DELAY(5000000);	/* 5 seconds */
358		} else {
359			printf("done\n");
360			/*
361			 * Unmount filesystems
362			 */
363			if (panicstr == 0)
364				vfs_unmountall();
365		}
366		DELAY(100000);		/* wait for console output to finish */
367	}
368
369	print_uptime();
370
371	/*
372	 * Ok, now do things that assume all filesystem activity has
373	 * been completed.
374	 */
375	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
376	splhigh();
377	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP &&
378	    !cold && dumper.dumper != NULL && !dumping)
379		doadump();
380
381	/* Now that we're going to really halt the system... */
382	EVENTHANDLER_INVOKE(shutdown_final, howto);
383
384	for(;;) ;	/* safety against shutdown_reset not working */
385	/* NOTREACHED */
386}
387
388/*
389 * If the shutdown was a clean halt, behave accordingly.
390 */
391static void
392shutdown_halt(void *junk, int howto)
393{
394
395	if (howto & RB_HALT) {
396		printf("\n");
397		printf("The operating system has halted.\n");
398		printf("Please press any key to reboot.\n\n");
399		switch (cngetc()) {
400		case -1:		/* No console, just die */
401			cpu_halt();
402			/* NOTREACHED */
403		default:
404			howto &= ~RB_HALT;
405			break;
406		}
407	}
408}
409
410/*
411 * Check to see if the system paniced, pause and then reboot
412 * according to the specified delay.
413 */
414static void
415shutdown_panic(void *junk, int howto)
416{
417	int loop;
418
419	if (howto & RB_DUMP) {
420		if (PANIC_REBOOT_WAIT_TIME != 0) {
421			if (PANIC_REBOOT_WAIT_TIME != -1) {
422				printf("Automatic reboot in %d seconds - "
423				       "press a key on the console to abort\n",
424					PANIC_REBOOT_WAIT_TIME);
425				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
426				     loop > 0; --loop) {
427					DELAY(1000 * 100); /* 1/10th second */
428					/* Did user type a key? */
429					if (cncheckc() != -1)
430						break;
431				}
432				if (!loop)
433					return;
434			}
435		} else { /* zero time specified - reboot NOW */
436			return;
437		}
438		printf("--> Press a key on the console to reboot,\n");
439		printf("--> or switch off the system now.\n");
440		cngetc();
441	}
442}
443
444/*
445 * Everything done, now reset
446 */
447static void
448shutdown_reset(void *junk, int howto)
449{
450
451	printf("Rebooting...\n");
452	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
453	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
454	cpu_reset();
455	/* NOTREACHED */ /* assuming reset worked */
456}
457
458#ifdef SMP
459static u_int panic_cpu = NOCPU;
460#endif
461
462/*
463 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
464 * and then reboots.  If we are called twice, then we avoid trying to sync
465 * the disks as this often leads to recursive panics.
466 *
467 * MPSAFE
468 */
469void
470panic(const char *fmt, ...)
471{
472	struct thread *td = curthread;
473	int bootopt, newpanic;
474	va_list ap;
475	static char buf[256];
476
477#ifdef SMP
478	/*
479	 * We don't want multiple CPU's to panic at the same time, so we
480	 * use panic_cpu as a simple spinlock.  We have to keep checking
481	 * panic_cpu if we are spinning in case the panic on the first
482	 * CPU is canceled.
483	 */
484	if (panic_cpu != PCPU_GET(cpuid))
485		while (atomic_cmpset_int(&panic_cpu, NOCPU,
486		    PCPU_GET(cpuid)) == 0)
487			while (panic_cpu != NOCPU)
488				; /* nothing */
489#endif
490
491	bootopt = RB_AUTOBOOT | RB_DUMP;
492	newpanic = 0;
493	if (panicstr)
494		bootopt |= RB_NOSYNC;
495	else {
496		panicstr = fmt;
497		newpanic = 1;
498	}
499
500	va_start(ap, fmt);
501	if (newpanic) {
502		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
503		panicstr = buf;
504		printf("panic: %s\n", buf);
505	} else {
506		printf("panic: ");
507		vprintf(fmt, ap);
508		printf("\n");
509	}
510	va_end(ap);
511#ifdef SMP
512	/* two separate prints in case of an unmapped page and trap */
513	printf("cpuid = %d; ", PCPU_GET(cpuid));
514#ifdef APIC_IO
515	printf("lapic.id = %08x\n", lapic.id);
516#else
517	printf("\n");
518#endif
519#endif
520
521#ifdef KDB
522	if (newpanic && trace_on_panic)
523		kdb_backtrace();
524	if (debugger_on_panic)
525		kdb_enter("panic");
526#ifdef RESTARTABLE_PANICS
527	/* See if the user aborted the panic, in which case we continue. */
528	if (panicstr == NULL) {
529#ifdef SMP
530		atomic_store_rel_int(&panic_cpu, NOCPU);
531#endif
532		return;
533	}
534#endif
535#endif
536	mtx_lock_spin(&sched_lock);
537	td->td_flags |= TDF_INPANIC;
538	mtx_unlock_spin(&sched_lock);
539	if (!sync_on_panic)
540		bootopt |= RB_NOSYNC;
541	boot(bootopt);
542}
543
544/*
545 * Support for poweroff delay.
546 */
547#ifndef POWEROFF_DELAY
548# define POWEROFF_DELAY 5000
549#endif
550static int poweroff_delay = POWEROFF_DELAY;
551
552SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
553	&poweroff_delay, 0, "");
554
555static void
556poweroff_wait(void *junk, int howto)
557{
558
559	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
560		return;
561	DELAY(poweroff_delay * 1000);
562}
563
564/*
565 * Some system processes (e.g. syncer) need to be stopped at appropriate
566 * points in their main loops prior to a system shutdown, so that they
567 * won't interfere with the shutdown process (e.g. by holding a disk buf
568 * to cause sync to fail).  For each of these system processes, register
569 * shutdown_kproc() as a handler for one of shutdown events.
570 */
571static int kproc_shutdown_wait = 60;
572SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
573    &kproc_shutdown_wait, 0, "");
574
575void
576kproc_shutdown(void *arg, int howto)
577{
578	struct proc *p;
579	int error;
580
581	if (panicstr)
582		return;
583
584	p = (struct proc *)arg;
585	printf("Waiting (max %d seconds) for system process `%s' to stop...",
586	    kproc_shutdown_wait, p->p_comm);
587	error = kthread_suspend(p, kproc_shutdown_wait * hz);
588
589	if (error == EWOULDBLOCK)
590		printf("timed out\n");
591	else
592		printf("stopped\n");
593}
594
595/* Registration of dumpers */
596int
597set_dumper(struct dumperinfo *di)
598{
599
600	if (di == NULL) {
601		bzero(&dumper, sizeof dumper);
602		return (0);
603	}
604	if (dumper.dumper != NULL)
605		return (EBUSY);
606	dumper = *di;
607	return (0);
608}
609
610#if defined(__powerpc__)
611void
612dumpsys(struct dumperinfo *di __unused)
613{
614
615	printf("Kernel dumps not implemented on this architecture\n");
616}
617#endif
618