kern_shutdown.c revision 73913 
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 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *	This product includes software developed by the University of
21 *	California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 *    may be used to endorse or promote products derived from this software
24 *    without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * 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 73913 2001-03-07 02:50:09Z jhb $
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/mutex.h>
59#include <sys/queue.h>
60#include <sys/sysctl.h>
61#include <sys/conf.h>
62#include <sys/sysproto.h>
63#include <sys/cons.h>
64
65#include <machine/pcb.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
110int dumping;				 /* system is dumping */
111
112static void boot(int) __dead2;
113static void dumpsys(void);
114static void poweroff_wait(void *, int);
115static void shutdown_halt(void *junk, int howto);
116static void shutdown_panic(void *junk, int howto);
117static void shutdown_reset(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(struct proc *p, struct reboot_args *uap)
138{
139	int error;
140
141	if ((error = suser(p)))
142		return (error);
143
144	boot(uap->opt);
145	return (0);
146}
147
148/*
149 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
150 */
151static int shutdown_howto = 0;
152
153void
154shutdown_nice(int howto)
155{
156	shutdown_howto = howto;
157
158	/* Send a signal to init(8) and have it shutdown the world */
159	if (initproc != NULL) {
160		PROC_LOCK(initproc);
161		psignal(initproc, SIGINT);
162		PROC_UNLOCK(initproc);
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(void)
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", (long)ts.tv_sec / 86400);
183		ts.tv_sec %= 86400;
184		f = 1;
185	}
186	if (f || ts.tv_sec >= 3600) {
187		printf("%ldh", (long)ts.tv_sec / 3600);
188		ts.tv_sec %= 3600;
189		f = 1;
190	}
191	if (f || ts.tv_sec >= 60) {
192		printf("%ldm", (long)ts.tv_sec / 60);
193		ts.tv_sec %= 60;
194		f = 1;
195	}
196	printf("%lds\n", (long)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(int howto)
206{
207
208	/* collect extra flags that shutdown_nice might have set */
209	howto |= shutdown_howto;
210
211#ifdef SMP
212	if (smp_active)
213		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
214#endif
215	/*
216	 * Do any callouts that should be done BEFORE syncing the filesystems.
217	 */
218	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
219
220	/*
221	 * Now sync filesystems
222	 */
223	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
224		register struct buf *bp;
225		int iter, nbusy, pbusy;
226		int subiter;
227
228		waittime = 0;
229		printf("\nsyncing disks... ");
230
231		sync(&proc0, NULL);
232
233		/*
234		 * With soft updates, some buffers that are
235		 * written will be remarked as dirty until other
236		 * buffers are written.
237		 */
238		for (iter = pbusy = 0; iter < 20; iter++) {
239			nbusy = 0;
240			for (bp = &buf[nbuf]; --bp >= buf; ) {
241				if ((bp->b_flags & B_INVAL) == 0 &&
242				    BUF_REFCNT(bp) > 0) {
243					nbusy++;
244				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
245						== B_DELWRI) {
246					/* bawrite(bp);*/
247					nbusy++;
248				}
249			}
250			if (nbusy == 0)
251				break;
252			printf("%d ", nbusy);
253			if (nbusy < pbusy)
254				iter = 0;
255			pbusy = nbusy;
256			sync(&proc0, NULL);
257 			if (curproc != NULL) {
258				DROP_GIANT_NOSWITCH();
259   				for (subiter = 0; subiter < 50 * iter; subiter++) {
260     					mtx_lock_spin(&sched_lock);
261     					setrunqueue(curproc);
262     					mi_switch(); /* Allow interrupt threads to run */
263     					mtx_unlock_spin(&sched_lock);
264     					DELAY(1000);
265   				}
266				PICKUP_GIANT();
267 			} else
268			DELAY(50000 * iter);
269		}
270		printf("\n");
271		/*
272		 * Count only busy local buffers to prevent forcing
273		 * a fsck if we're just a client of a wedged NFS server
274		 */
275		nbusy = 0;
276		for (bp = &buf[nbuf]; --bp >= buf; ) {
277			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
278			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
279				if (bp->b_dev == NODEV) {
280					TAILQ_REMOVE(&mountlist,
281					    bp->b_vp->v_mount, mnt_list);
282					continue;
283				}
284				nbusy++;
285#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
286				printf(
287			    "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
288				    nbusy, devtoname(bp->b_dev),
289				    bp->b_flags, (long)bp->b_blkno,
290				    (long)bp->b_lblkno);
291#endif
292			}
293		}
294		if (nbusy) {
295			/*
296			 * Failed to sync all blocks. Indicate this and don't
297			 * unmount filesystems (thus forcing an fsck on reboot).
298			 */
299			printf("giving up on %d buffers\n", nbusy);
300			DELAY(5000000);	/* 5 seconds */
301		} else {
302			printf("done\n");
303			/*
304			 * Unmount filesystems
305			 */
306			if (panicstr == 0)
307				vfs_unmountall();
308		}
309		DELAY(100000);		/* wait for console output to finish */
310	}
311
312	print_uptime();
313
314	/*
315	 * Ok, now do things that assume all filesystem activity has
316	 * been completed.
317	 */
318	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
319	splhigh();
320	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
321		dumpsys();
322
323	/* Now that we're going to really halt the system... */
324	EVENTHANDLER_INVOKE(shutdown_final, howto);
325
326	for(;;) ;	/* safety against shutdown_reset not working */
327	/* NOTREACHED */
328}
329
330/*
331 * If the shutdown was a clean halt, behave accordingly.
332 */
333static void
334shutdown_halt(void *junk, int howto)
335{
336	if (howto & RB_HALT) {
337		printf("\n");
338		printf("The operating system has halted.\n");
339		printf("Please press any key to reboot.\n\n");
340		switch (cngetc()) {
341		case -1:		/* No console, just die */
342			cpu_halt();
343			/* NOTREACHED */
344		default:
345			howto &= ~RB_HALT;
346			break;
347		}
348	}
349}
350
351/*
352 * Check to see if the system paniced, pause and then reboot
353 * according to the specified delay.
354 */
355static void
356shutdown_panic(void *junk, int howto)
357{
358	int loop;
359
360	if (howto & RB_DUMP) {
361		if (PANIC_REBOOT_WAIT_TIME != 0) {
362			if (PANIC_REBOOT_WAIT_TIME != -1) {
363				printf("Automatic reboot in %d seconds - "
364				       "press a key on the console to abort\n",
365					PANIC_REBOOT_WAIT_TIME);
366				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
367				     loop > 0; --loop) {
368					DELAY(1000 * 100); /* 1/10th second */
369					/* Did user type a key? */
370					if (cncheckc() != -1)
371						break;
372				}
373				if (!loop)
374					return;
375			}
376		} else { /* zero time specified - reboot NOW */
377			return;
378		}
379		printf("--> Press a key on the console to reboot <--\n");
380		cngetc();
381	}
382}
383
384/*
385 * Everything done, now reset
386 */
387static void
388shutdown_reset(void *junk, int howto)
389{
390	printf("Rebooting...\n");
391	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
392	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
393	cpu_reset();
394	/* NOTREACHED */ /* assuming reset worked */
395}
396
397/*
398 * Magic number for savecore
399 *
400 * exported (symorder) and used at least by savecore(8)
401 *
402 */
403static u_long const	dumpmag = 0x8fca0101UL;
404
405static int	dumpsize = 0;		/* also for savecore */
406
407static int	dodump = 1;
408
409SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
410    "Try to perform coredump on kernel panic");
411
412static int
413setdumpdev(dev_t dev)
414{
415	int psize;
416	long newdumplo;
417
418	if (dev == NODEV) {
419		dumpdev = dev;
420		return (0);
421	}
422	if (devsw(dev) == NULL)
423		return (ENXIO);		/* XXX is this right? */
424	if (devsw(dev)->d_psize == NULL)
425		return (ENXIO);		/* XXX should be ENODEV ? */
426	psize = devsw(dev)->d_psize(dev);
427	if (psize == -1)
428		return (ENXIO);		/* XXX should be ENODEV ? */
429	/*
430	 * XXX should clean up checking in dumpsys() to be more like this.
431	 */
432	newdumplo = psize - Maxmem * PAGE_SIZE / DEV_BSIZE;
433	if (newdumplo < 0)
434		return (ENOSPC);
435	dumpdev = dev;
436	dumplo = newdumplo;
437	return (0);
438}
439
440
441/* ARGSUSED */
442static void
443dump_conf(void *dummy)
444{
445	if (setdumpdev(dumpdev) != 0)
446		dumpdev = NODEV;
447}
448
449SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
450
451static int
452sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
453{
454	int error;
455	udev_t ndumpdev;
456
457	ndumpdev = dev2udev(dumpdev);
458	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
459	if (error == 0 && req->newptr != NULL)
460		error = setdumpdev(udev2dev(ndumpdev, 0));
461	return (error);
462}
463
464SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
465	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
466
467/*
468 * Doadump comes here after turning off memory management and
469 * getting on the dump stack, either when called above, or by
470 * the auto-restart code.
471 */
472static void
473dumpsys(void)
474{
475	int	error;
476
477	savectx(&dumppcb);
478	if (dumping++) {
479		printf("Dump already in progress, bailing...\n");
480		return;
481	}
482	if (!dodump)
483		return;
484	if (dumpdev == NODEV)
485		return;
486	if (!(devsw(dumpdev)))
487		return;
488	if (!(devsw(dumpdev)->d_dump))
489		return;
490	dumpsize = Maxmem;
491	printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
492	printf("dump ");
493	error = (*devsw(dumpdev)->d_dump)(dumpdev);
494	if (error == 0) {
495		printf("succeeded\n");
496		return;
497	}
498	printf("failed, reason: ");
499	switch (error) {
500	case ENODEV:
501		printf("device doesn't support a dump routine\n");
502		break;
503
504	case ENXIO:
505		printf("device bad\n");
506		break;
507
508	case EFAULT:
509		printf("device not ready\n");
510		break;
511
512	case EINVAL:
513		printf("area improper\n");
514		break;
515
516	case EIO:
517		printf("i/o error\n");
518		break;
519
520	case EINTR:
521		printf("aborted from console\n");
522		break;
523
524	default:
525		printf("unknown, error = %d\n", error);
526		break;
527	}
528}
529
530/*
531 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
532 * and then reboots.  If we are called twice, then we avoid trying to sync
533 * the disks as this often leads to recursive panics.
534 */
535void
536panic(const char *fmt, ...)
537{
538	int bootopt;
539	va_list ap;
540	static char buf[256];
541
542#ifdef SMP
543	/* Only 1 CPU can panic at a time */
544	mtx_lock(&panic_mtx);
545#endif
546
547	bootopt = RB_AUTOBOOT | RB_DUMP;
548	if (panicstr)
549		bootopt |= RB_NOSYNC;
550	else
551		panicstr = fmt;
552
553	va_start(ap, fmt);
554	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
555	if (panicstr == fmt)
556		panicstr = buf;
557	va_end(ap);
558	printf("panic: %s\n", buf);
559#ifdef SMP
560	/* two separate prints in case of an unmapped page and trap */
561	printf("cpuid = %d; ", PCPU_GET(cpuid));
562#ifdef APIC_IO
563	printf("lapic.id = %08x\n", lapic.id);
564#endif
565#endif
566
567#if defined(DDB)
568	if (debugger_on_panic)
569		Debugger ("panic");
570#endif
571	boot(bootopt);
572}
573
574/*
575 * Support for poweroff delay.
576 */
577#ifndef POWEROFF_DELAY
578# define POWEROFF_DELAY 5000
579#endif
580static int poweroff_delay = POWEROFF_DELAY;
581
582SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
583	&poweroff_delay, 0, "");
584
585static void
586poweroff_wait(void *junk, int howto)
587{
588	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
589		return;
590	DELAY(poweroff_delay * 1000);
591}
592
593/*
594 * Some system processes (e.g. syncer) need to be stopped at appropriate
595 * points in their main loops prior to a system shutdown, so that they
596 * won't interfere with the shutdown process (e.g. by holding a disk buf
597 * to cause sync to fail).  For each of these system processes, register
598 * shutdown_kproc() as a handler for one of shutdown events.
599 */
600static int kproc_shutdown_wait = 60;
601SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
602    &kproc_shutdown_wait, 0, "");
603
604void
605kproc_shutdown(void *arg, int howto)
606{
607	struct proc *p;
608	int error;
609
610	if (panicstr)
611		return;
612
613	p = (struct proc *)arg;
614	printf("Waiting (max %d seconds) for system process `%s' to stop...",
615	    kproc_shutdown_wait, p->p_comm);
616	error = kthread_suspend(p, kproc_shutdown_wait * hz);
617
618	if (error == EWOULDBLOCK)
619		printf("timed out\n");
620	else
621		printf("stopped\n");
622}
623