kern_shutdown.c revision 65268
1238901Sandrew/*-
2238901Sandrew * Copyright (c) 1986, 1988, 1991, 1993
3238901Sandrew *	The Regents of the University of California.  All rights reserved.
4238901Sandrew * (c) UNIX System Laboratories, Inc.
5238901Sandrew * All or some portions of this file are derived from material licensed
6238901Sandrew * to the University of California by American Telephone and Telegraph
7238901Sandrew * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8238901Sandrew * the permission of UNIX System Laboratories, Inc.
9238901Sandrew *
10238901Sandrew * Redistribution and use in source and binary forms, with or without
11238901Sandrew * modification, are permitted provided that the following conditions
12238901Sandrew * are met:
13238901Sandrew * 1. Redistributions of source code must retain the above copyright
14238901Sandrew *    notice, this list of conditions and the following disclaimer.
15238901Sandrew * 2. Redistributions in binary form must reproduce the above copyright
16238901Sandrew *    notice, this list of conditions and the following disclaimer in the
17238901Sandrew *    documentation and/or other materials provided with the distribution.
18238901Sandrew * 3. All advertising materials mentioning features or use of this software
19238901Sandrew *    must display the following acknowledgement:
20238901Sandrew *	This product includes software developed by the University of
21238901Sandrew *	California, Berkeley and its contributors.
22238901Sandrew * 4. Neither the name of the University nor the names of its contributors
23238901Sandrew *    may be used to endorse or promote products derived from this software
24238901Sandrew *    without specific prior written permission.
25238901Sandrew *
26238901Sandrew * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27238901Sandrew * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28238901Sandrew * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29238901Sandrew * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30245641Sandrew * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31245641Sandrew * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 *	@(#)kern_shutdown.c	8.3 (Berkeley) 1/21/94
39 * $FreeBSD: head/sys/kern/kern_shutdown.c 65268 2000-08-31 00:08:50Z msmith $
40 */
41
42#include "opt_ddb.h"
43#include "opt_hw_wdog.h"
44#include "opt_panic.h"
45#include "opt_show_busybufs.h"
46
47#include <sys/param.h>
48#include <sys/systm.h>
49#include <sys/eventhandler.h>
50#include <sys/bio.h>
51#include <sys/buf.h>
52#include <sys/reboot.h>
53#include <sys/proc.h>
54#include <sys/vnode.h>
55#include <sys/kernel.h>
56#include <sys/kthread.h>
57#include <sys/mount.h>
58#include <sys/queue.h>
59#include <sys/sysctl.h>
60#include <sys/conf.h>
61#include <sys/sysproto.h>
62#include <sys/cons.h>
63
64#include <machine/pcb.h>
65#include <machine/clock.h>
66#include <machine/md_var.h>
67#include <machine/smp.h>		/* smp_active, cpuid */
68
69#include <sys/signalvar.h>
70
71#ifndef PANIC_REBOOT_WAIT_TIME
72#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
73#endif
74
75/*
76 * Note that stdarg.h and the ANSI style va_start macro is used for both
77 * ANSI and traditional C compilers.
78 */
79#include <machine/stdarg.h>
80
81#ifdef DDB
82#ifdef DDB_UNATTENDED
83int debugger_on_panic = 0;
84#else
85int debugger_on_panic = 1;
86#endif
87SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
88	&debugger_on_panic, 0, "Run debugger on kernel panic");
89#endif
90
91SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
92
93#ifdef	HW_WDOG
94/*
95 * If there is a hardware watchdog, point this at the function needed to
96 * hold it off.
97 * It's needed when the kernel needs to do some lengthy operations.
98 * e.g. in wd.c when dumping core.. It's most annoying to have
99 * your precious core-dump only half written because the wdog kicked in.
100 */
101watchdog_tickle_fn wdog_tickler = NULL;
102#endif	/* HW_WDOG */
103
104/*
105 * Variable panicstr contains argument to first call to panic; used as flag
106 * to indicate that the kernel has already called panic.
107 */
108const char *panicstr;
109
110static void boot __P((int)) __dead2;
111static void dumpsys __P((void));
112static int setdumpdev __P((dev_t dev));
113static void poweroff_wait __P((void *, int));
114static void print_uptime __P((void));
115static void shutdown_halt __P((void *junk, int howto));
116static void shutdown_panic __P((void *junk, int howto));
117static void shutdown_reset __P((void *junk, int howto));
118
119/* register various local shutdown events */
120static void
121shutdown_conf(void *unused)
122{
123	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
124	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
125	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
126	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
127}
128
129SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
130
131/* ARGSUSED */
132
133/*
134 * The system call that results in a reboot
135 */
136int
137reboot(p, uap)
138	struct proc *p;
139	struct reboot_args *uap;
140{
141	int error;
142
143	if ((error = suser(p)))
144		return (error);
145
146	boot(uap->opt);
147	return (0);
148}
149
150/*
151 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
152 */
153static int shutdown_howto = 0;
154
155void
156shutdown_nice(int howto)
157{
158	shutdown_howto = howto;
159
160	/* Send a signal to init(8) and have it shutdown the world */
161	if (initproc != NULL) {
162		psignal(initproc, SIGINT);
163	} else {
164		/* No init(8) running, so simply reboot */
165		boot(RB_NOSYNC);
166	}
167	return;
168}
169static int	waittime = -1;
170static struct pcb dumppcb;
171
172static void
173print_uptime()
174{
175	int f;
176	struct timespec ts;
177
178	getnanouptime(&ts);
179	printf("Uptime: ");
180	f = 0;
181	if (ts.tv_sec >= 86400) {
182		printf("%ldd", ts.tv_sec / 86400);
183		ts.tv_sec %= 86400;
184		f = 1;
185	}
186	if (f || ts.tv_sec >= 3600) {
187		printf("%ldh", ts.tv_sec / 3600);
188		ts.tv_sec %= 3600;
189		f = 1;
190	}
191	if (f || ts.tv_sec >= 60) {
192		printf("%ldm", ts.tv_sec / 60);
193		ts.tv_sec %= 60;
194		f = 1;
195	}
196	printf("%lds\n", ts.tv_sec);
197}
198
199/*
200 *  Go through the rigmarole of shutting down..
201 * this used to be in machdep.c but I'll be dammned if I could see
202 * anything machine dependant in it.
203 */
204static void
205boot(howto)
206	int howto;
207{
208
209	/* collect extra flags that shutdown_nice might have set */
210	howto |= shutdown_howto;
211
212#ifdef SMP
213	if (smp_active) {
214		printf("boot() called on cpu#%d\n", cpuid);
215	}
216#endif
217	/*
218	 * Do any callouts that should be done BEFORE syncing the filesystems.
219	 */
220	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
221
222	/*
223	 * Now sync filesystems
224	 */
225	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
226		register struct buf *bp;
227		int iter, nbusy;
228
229		waittime = 0;
230		printf("\nsyncing disks... ");
231
232		sync(&proc0, NULL);
233
234		/*
235		 * With soft updates, some buffers that are
236		 * written will be remarked as dirty until other
237		 * buffers are written.
238		 */
239		for (iter = 0; iter < 20; iter++) {
240			nbusy = 0;
241			for (bp = &buf[nbuf]; --bp >= buf; ) {
242				if ((bp->b_flags & B_INVAL) == 0 &&
243				    BUF_REFCNT(bp) > 0) {
244					nbusy++;
245				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
246						== B_DELWRI) {
247					/* bawrite(bp);*/
248					nbusy++;
249				}
250			}
251			if (nbusy == 0)
252				break;
253			printf("%d ", nbusy);
254			sync(&proc0, NULL);
255			DELAY(50000 * iter);
256		}
257		printf("\n");
258		/*
259		 * Count only busy local buffers to prevent forcing
260		 * a fsck if we're just a client of a wedged NFS server
261		 */
262		nbusy = 0;
263		for (bp = &buf[nbuf]; --bp >= buf; ) {
264			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
265			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
266				if (bp->b_dev == NODEV) {
267					TAILQ_REMOVE(&mountlist,
268					    bp->b_vp->v_mount, mnt_list);
269					continue;
270				}
271				nbusy++;
272#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
273				printf(
274			    "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
275				    nbusy, devtoname(bp->b_dev),
276				    bp->b_flags, (long)bp->b_blkno,
277				    (long)bp->b_lblkno);
278#endif
279			}
280		}
281		if (nbusy) {
282			/*
283			 * Failed to sync all blocks. Indicate this and don't
284			 * unmount filesystems (thus forcing an fsck on reboot).
285			 */
286			printf("giving up on %d buffers\n", nbusy);
287			DELAY(5000000);	/* 5 seconds */
288		} else {
289			printf("done\n");
290			/*
291			 * Unmount filesystems
292			 */
293			if (panicstr == 0)
294				vfs_unmountall();
295		}
296		DELAY(100000);		/* wait for console output to finish */
297	}
298
299	print_uptime();
300
301	/*
302	 * Ok, now do things that assume all filesystem activity has
303	 * been completed.
304	 */
305	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
306	splhigh();
307	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) {
308		savectx(&dumppcb);
309#ifdef __i386__
310		dumppcb.pcb_cr3 = rcr3();
311#endif
312		dumpsys();
313	}
314
315	/* Now that we're going to really halt the system... */
316	EVENTHANDLER_INVOKE(shutdown_final, howto);
317
318	for(;;) ;	/* safety against shutdown_reset not working */
319	/* NOTREACHED */
320}
321
322/*
323 * If the shutdown was a clean halt, behave accordingly.
324 */
325static void
326shutdown_halt(void *junk, int howto)
327{
328	if (howto & RB_HALT) {
329		printf("\n");
330		printf("The operating system has halted.\n");
331		printf("Please press any key to reboot.\n\n");
332		switch (cngetc()) {
333		case -1:		/* No console, just die */
334			cpu_halt();
335			/* NOTREACHED */
336		default:
337			howto &= ~RB_HALT;
338			break;
339		}
340	}
341}
342
343/*
344 * Check to see if the system paniced, pause and then reboot
345 * according to the specified delay.
346 */
347static void
348shutdown_panic(void *junk, int howto)
349{
350	int loop;
351
352	if (howto & RB_DUMP) {
353		if (PANIC_REBOOT_WAIT_TIME != 0) {
354			if (PANIC_REBOOT_WAIT_TIME != -1) {
355				printf("Automatic reboot in %d seconds - "
356				       "press a key on the console to abort\n",
357					PANIC_REBOOT_WAIT_TIME);
358				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
359				     loop > 0; --loop) {
360					DELAY(1000 * 100); /* 1/10th second */
361					/* Did user type a key? */
362					if (cncheckc() != -1)
363						break;
364				}
365				if (!loop)
366					return;
367			}
368		} else { /* zero time specified - reboot NOW */
369			return;
370		}
371		printf("--> Press a key on the console to reboot <--\n");
372		cngetc();
373	}
374}
375
376/*
377 * Everything done, now reset
378 */
379static void
380shutdown_reset(void *junk, int howto)
381{
382	printf("Rebooting...\n");
383	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
384	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
385	cpu_reset();
386	/* NOTREACHED */ /* assuming reset worked */
387}
388
389/*
390 * Magic number for savecore
391 *
392 * exported (symorder) and used at least by savecore(8)
393 *
394 */
395static u_long const	dumpmag = 0x8fca0101UL;
396
397static int	dumpsize = 0;		/* also for savecore */
398
399static int	dodump = 1;
400
401SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
402    "Try to perform coredump on kernel panic");
403
404static int
405setdumpdev(dev)
406	dev_t dev;
407{
408	int psize;
409	long newdumplo;
410
411	if (dev == NODEV) {
412		dumpdev = dev;
413		return (0);
414	}
415	if (devsw(dev) == NULL)
416		return (ENXIO);		/* XXX is this right? */
417	if (devsw(dev)->d_psize == NULL)
418		return (ENXIO);		/* XXX should be ENODEV ? */
419	psize = devsw(dev)->d_psize(dev);
420	if (psize == -1)
421		return (ENXIO);		/* XXX should be ENODEV ? */
422	/*
423	 * XXX should clean up checking in dumpsys() to be more like this.
424	 */
425	newdumplo = psize - Maxmem * PAGE_SIZE / DEV_BSIZE;
426	if (newdumplo < 0)
427		return (ENOSPC);
428	dumpdev = dev;
429	dumplo = newdumplo;
430	return (0);
431}
432
433
434/* ARGSUSED */
435static void dump_conf __P((void *dummy));
436static void
437dump_conf(dummy)
438	void *dummy;
439{
440	if (setdumpdev(dumpdev) != 0)
441		dumpdev = NODEV;
442}
443
444SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
445
446static int
447sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
448{
449	int error;
450	udev_t ndumpdev;
451
452	ndumpdev = dev2udev(dumpdev);
453	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
454	if (error == 0 && req->newptr != NULL)
455		error = setdumpdev(udev2dev(ndumpdev, 0));
456	return (error);
457}
458
459SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
460	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
461
462/*
463 * Doadump comes here after turning off memory management and
464 * getting on the dump stack, either when called above, or by
465 * the auto-restart code.
466 */
467static void
468dumpsys(void)
469{
470	int	error;
471	static int dumping;
472
473	if (dumping++) {
474		printf("Dump already in progress, bailing...\n");
475		return;
476	}
477	if (!dodump)
478		return;
479	if (dumpdev == NODEV)
480		return;
481	if (!(devsw(dumpdev)))
482		return;
483	if (!(devsw(dumpdev)->d_dump))
484		return;
485	dumpsize = Maxmem;
486	printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
487	printf("dump ");
488	error = (*devsw(dumpdev)->d_dump)(dumpdev);
489	if (error == 0) {
490		printf("succeeded\n");
491		return;
492	}
493	printf("failed, reason: ");
494	switch (error) {
495	case ENODEV:
496		printf("device doesn't support a dump routine\n");
497		break;
498
499	case ENXIO:
500		printf("device bad\n");
501		break;
502
503	case EFAULT:
504		printf("device not ready\n");
505		break;
506
507	case EINVAL:
508		printf("area improper\n");
509		break;
510
511	case EIO:
512		printf("i/o error\n");
513		break;
514
515	case EINTR:
516		printf("aborted from console\n");
517		break;
518
519	default:
520		printf("unknown, error = %d\n", error);
521		break;
522	}
523}
524
525/*
526 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
527 * and then reboots.  If we are called twice, then we avoid trying to sync
528 * the disks as this often leads to recursive panics.
529 */
530void
531panic(const char *fmt, ...)
532{
533	int bootopt;
534	va_list ap;
535	static char buf[256];
536
537	bootopt = RB_AUTOBOOT | RB_DUMP;
538	if (panicstr)
539		bootopt |= RB_NOSYNC;
540	else
541		panicstr = fmt;
542
543	va_start(ap, fmt);
544	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
545	if (panicstr == fmt)
546		panicstr = buf;
547	va_end(ap);
548	printf("panic: %s\n", buf);
549#ifdef SMP
550	/* three seperate prints in case of an unmapped page and trap */
551	printf("mp_lock = %08x; ", mp_lock);
552	printf("cpuid = %d; ", cpuid);
553	printf("lapic.id = %08x\n", lapic.id);
554#endif
555
556#if defined(DDB)
557	if (debugger_on_panic)
558		Debugger ("panic");
559#endif
560	boot(bootopt);
561}
562
563/*
564 * Support for poweroff delay.
565 */
566#ifndef POWEROFF_DELAY
567# define POWEROFF_DELAY 5000
568#endif
569static int poweroff_delay = POWEROFF_DELAY;
570
571SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
572	&poweroff_delay, 0, "");
573
574static void
575poweroff_wait(void *junk, int howto)
576{
577	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
578		return;
579	DELAY(poweroff_delay * 1000);
580}
581
582/*
583 * Some system processes (e.g. syncer) need to be stopped at appropriate
584 * points in their main loops prior to a system shutdown, so that they
585 * won't interfere with the shutdown process (e.g. by holding a disk buf
586 * to cause sync to fail).  For each of these system processes, register
587 * shutdown_kproc() as a handler for one of shutdown events.
588 */
589static int kproc_shutdown_wait = 60;
590SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
591    &kproc_shutdown_wait, 0, "");
592
593void
594shutdown_kproc(void *arg, int howto)
595{
596	struct proc *p;
597	int error;
598
599	if (panicstr)
600		return;
601
602	p = (struct proc *)arg;
603	printf("Waiting (max %d seconds) for system process `%s' to stop...",
604	    kproc_shutdown_wait, p->p_comm);
605	error = suspend_kproc(p, kproc_shutdown_wait * hz);
606
607	if (error == EWOULDBLOCK)
608		printf("timed out\n");
609	else
610		printf("stopped\n");
611}
612