kern_shutdown.c revision 90361 
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 90361 2002-02-07 20:58:47Z julian $
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/bio.h>
50#include <sys/buf.h>
51#include <sys/conf.h>
52#include <sys/cons.h>
53#include <sys/disklabel.h>
54#include <sys/eventhandler.h>
55#include <sys/kernel.h>
56#include <sys/kthread.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/pcb.h>
68#include <machine/md_var.h>
69
70#include <sys/signalvar.h>
71#ifdef DDB
72#include <ddb/ddb.h>
73#endif
74
75#ifndef PANIC_REBOOT_WAIT_TIME
76#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
77#endif
78
79/*
80 * Note that stdarg.h and the ANSI style va_start macro is used for both
81 * ANSI and traditional C compilers.
82 */
83#include <machine/stdarg.h>
84
85#ifdef DDB
86#ifdef DDB_UNATTENDED
87int debugger_on_panic = 0;
88#else
89int debugger_on_panic = 1;
90#endif
91SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
92	&debugger_on_panic, 0, "Run debugger on kernel panic");
93#endif
94
95int sync_on_panic = 1;
96SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
97	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
98
99SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
100
101#ifdef	HW_WDOG
102/*
103 * If there is a hardware watchdog, point this at the function needed to
104 * hold it off.
105 * It's needed when the kernel needs to do some lengthy operations.
106 * e.g. in wd.c when dumping core.. It's most annoying to have
107 * your precious core-dump only half written because the wdog kicked in.
108 */
109watchdog_tickle_fn wdog_tickler = NULL;
110#endif	/* HW_WDOG */
111
112/*
113 * Variable panicstr contains argument to first call to panic; used as flag
114 * to indicate that the kernel has already called panic.
115 */
116const char *panicstr;
117
118int dumping;				 /* system is dumping */
119
120static void boot(int) __dead2;
121static void dumpsys(void);
122static void poweroff_wait(void *, int);
123static void shutdown_halt(void *junk, int howto);
124static void shutdown_panic(void *junk, int howto);
125static void shutdown_reset(void *junk, int howto);
126
127/* register various local shutdown events */
128static void
129shutdown_conf(void *unused)
130{
131	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
132	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
133	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
134	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
135}
136
137SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
138
139/*
140 * The system call that results in a reboot
141 *
142 * MPSAFE
143 */
144/* ARGSUSED */
145int
146reboot(struct thread *td, struct reboot_args *uap)
147{
148	int error;
149
150	mtx_lock(&Giant);
151	if ((error = suser_td(td)) == 0)
152		boot(uap->opt);
153	mtx_unlock(&Giant);
154	return (error);
155}
156
157/*
158 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
159 */
160static int shutdown_howto = 0;
161
162void
163shutdown_nice(int howto)
164{
165	shutdown_howto = howto;
166
167	/* Send a signal to init(8) and have it shutdown the world */
168	if (initproc != NULL) {
169		PROC_LOCK(initproc);
170		psignal(initproc, SIGINT);
171		PROC_UNLOCK(initproc);
172	} else {
173		/* No init(8) running, so simply reboot */
174		boot(RB_NOSYNC);
175	}
176	return;
177}
178static int	waittime = -1;
179static struct pcb dumppcb;
180
181static void
182print_uptime(void)
183{
184	int f;
185	struct timespec ts;
186
187	getnanouptime(&ts);
188	printf("Uptime: ");
189	f = 0;
190	if (ts.tv_sec >= 86400) {
191		printf("%ldd", (long)ts.tv_sec / 86400);
192		ts.tv_sec %= 86400;
193		f = 1;
194	}
195	if (f || ts.tv_sec >= 3600) {
196		printf("%ldh", (long)ts.tv_sec / 3600);
197		ts.tv_sec %= 3600;
198		f = 1;
199	}
200	if (f || ts.tv_sec >= 60) {
201		printf("%ldm", (long)ts.tv_sec / 60);
202		ts.tv_sec %= 60;
203		f = 1;
204	}
205	printf("%lds\n", (long)ts.tv_sec);
206}
207
208/*
209 *  Go through the rigmarole of shutting down..
210 * this used to be in machdep.c but I'll be dammned if I could see
211 * anything machine dependant in it.
212 */
213static void
214boot(int howto)
215{
216
217	/* collect extra flags that shutdown_nice might have set */
218	howto |= shutdown_howto;
219
220#ifdef DDB
221	/* We are out of the debugger now. */
222	db_active = 0;
223#endif
224
225#ifdef SMP
226	if (smp_active)
227		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
228#endif
229	/*
230	 * Do any callouts that should be done BEFORE syncing the filesystems.
231	 */
232	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
233
234	/*
235	 * Now sync filesystems
236	 */
237	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
238		register struct buf *bp;
239		int iter, nbusy, pbusy;
240		int subiter;
241
242		waittime = 0;
243		printf("\nsyncing disks... ");
244
245		sync(&thread0, NULL);
246
247		/*
248		 * With soft updates, some buffers that are
249		 * written will be remarked as dirty until other
250		 * buffers are written.
251		 */
252		for (iter = pbusy = 0; iter < 20; iter++) {
253			nbusy = 0;
254			for (bp = &buf[nbuf]; --bp >= buf; ) {
255				if ((bp->b_flags & B_INVAL) == 0 &&
256				    BUF_REFCNT(bp) > 0) {
257					nbusy++;
258				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
259						== B_DELWRI) {
260					/* bawrite(bp);*/
261					nbusy++;
262				}
263			}
264			if (nbusy == 0)
265				break;
266			printf("%d ", nbusy);
267			if (nbusy < pbusy)
268				iter = 0;
269			pbusy = nbusy;
270			sync(&thread0, NULL);
271 			if (curthread != NULL) {
272				DROP_GIANT();
273   				for (subiter = 0; subiter < 50 * iter; subiter++) {
274     					mtx_lock_spin(&sched_lock);
275					curthread->td_proc->p_stats->p_ru.ru_nvcsw++;
276     					mi_switch(); /* Allow interrupt threads to run */
277     					mtx_unlock_spin(&sched_lock);
278     					DELAY(1000);
279   				}
280				PICKUP_GIANT();
281 			} else
282			DELAY(50000 * iter);
283		}
284		printf("\n");
285		/*
286		 * Count only busy local buffers to prevent forcing
287		 * a fsck if we're just a client of a wedged NFS server
288		 */
289		nbusy = 0;
290		for (bp = &buf[nbuf]; --bp >= buf; ) {
291			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
292			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
293				if (bp->b_dev == NODEV) {
294					TAILQ_REMOVE(&mountlist,
295					    bp->b_vp->v_mount, mnt_list);
296					continue;
297				}
298				nbusy++;
299#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
300				printf(
301			    "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
302				    nbusy, devtoname(bp->b_dev),
303				    bp->b_flags, (long)bp->b_blkno,
304				    (long)bp->b_lblkno);
305#endif
306			}
307		}
308		if (nbusy) {
309			/*
310			 * Failed to sync all blocks. Indicate this and don't
311			 * unmount filesystems (thus forcing an fsck on reboot).
312			 */
313			printf("giving up on %d buffers\n", nbusy);
314			DELAY(5000000);	/* 5 seconds */
315		} else {
316			printf("done\n");
317			/*
318			 * Unmount filesystems
319			 */
320			if (panicstr == 0)
321				vfs_unmountall();
322		}
323		DELAY(100000);		/* wait for console output to finish */
324	}
325
326	print_uptime();
327
328	/*
329	 * Ok, now do things that assume all filesystem activity has
330	 * been completed.
331	 */
332	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
333	splhigh();
334	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
335		dumpsys();
336
337	/* Now that we're going to really halt the system... */
338	EVENTHANDLER_INVOKE(shutdown_final, howto);
339
340	for(;;) ;	/* safety against shutdown_reset not working */
341	/* NOTREACHED */
342}
343
344/*
345 * If the shutdown was a clean halt, behave accordingly.
346 */
347static void
348shutdown_halt(void *junk, int howto)
349{
350	if (howto & RB_HALT) {
351		printf("\n");
352		printf("The operating system has halted.\n");
353		printf("Please press any key to reboot.\n\n");
354		switch (cngetc()) {
355		case -1:		/* No console, just die */
356			cpu_halt();
357			/* NOTREACHED */
358		default:
359			howto &= ~RB_HALT;
360			break;
361		}
362	}
363}
364
365/*
366 * Check to see if the system paniced, pause and then reboot
367 * according to the specified delay.
368 */
369static void
370shutdown_panic(void *junk, int howto)
371{
372	int loop;
373
374	if (howto & RB_DUMP) {
375		if (PANIC_REBOOT_WAIT_TIME != 0) {
376			if (PANIC_REBOOT_WAIT_TIME != -1) {
377				printf("Automatic reboot in %d seconds - "
378				       "press a key on the console to abort\n",
379					PANIC_REBOOT_WAIT_TIME);
380				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
381				     loop > 0; --loop) {
382					DELAY(1000 * 100); /* 1/10th second */
383					/* Did user type a key? */
384					if (cncheckc() != -1)
385						break;
386				}
387				if (!loop)
388					return;
389			}
390		} else { /* zero time specified - reboot NOW */
391			return;
392		}
393		printf("--> Press a key on the console to reboot,\n");
394		printf("--> or switch off the system now.\n");
395		cngetc();
396	}
397}
398
399/*
400 * Everything done, now reset
401 */
402static void
403shutdown_reset(void *junk, int howto)
404{
405	printf("Rebooting...\n");
406	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
407	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
408	cpu_reset();
409	/* NOTREACHED */ /* assuming reset worked */
410}
411
412/*
413 * Magic number for savecore
414 *
415 * exported (symorder) and used at least by savecore(8)
416 *
417 */
418static u_long const	dumpmag = 0x8fca0101UL;
419
420static int	dumpsize = 0;		/* also for savecore */
421
422static int	dodump = 1;
423
424SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
425    "Try to perform coredump on kernel panic");
426
427static int
428setdumpdev(dev_t dev)
429{
430	int psize;
431	long newdumplo;
432
433	if (dev == NODEV) {
434		dumpdev = dev;
435		return (0);
436	}
437	if (devsw(dev) == NULL)
438		return (ENXIO);		/* XXX is this right? */
439	if (devsw(dev)->d_psize == NULL)
440		return (ENXIO);		/* XXX should be ENODEV ? */
441	psize = devsw(dev)->d_psize(dev);
442	if (psize == -1)
443		return (ENXIO);		/* XXX should be ENODEV ? */
444	/*
445	 * XXX should clean up checking in dumpsys() to be more like this.
446	 */
447	newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
448	if (newdumplo <= LABELSECTOR)
449		return (ENOSPC);
450	dumpdev = dev;
451	dumplo = newdumplo;
452	return (0);
453}
454
455
456/* ARGSUSED */
457static void
458dump_conf(void *dummy)
459{
460	char *path;
461	dev_t dev;
462
463	path = malloc(MNAMELEN, M_TEMP, M_WAITOK);
464	if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
465		dev = getdiskbyname(path);
466		if (dev != NODEV)
467			dumpdev = dev;
468	}
469	free(path, M_TEMP);
470	if (setdumpdev(dumpdev) != 0)
471		dumpdev = NODEV;
472}
473
474SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
475
476static int
477sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
478{
479	int error;
480	udev_t ndumpdev;
481
482	ndumpdev = dev2udev(dumpdev);
483	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
484	if (error == 0 && req->newptr != NULL)
485		error = setdumpdev(udev2dev(ndumpdev, 0));
486	return (error);
487}
488
489SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
490	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
491
492/*
493 * Doadump comes here after turning off memory management and
494 * getting on the dump stack, either when called above, or by
495 * the auto-restart code.
496 */
497static void
498dumpsys(void)
499{
500	int	error;
501
502	savectx(&dumppcb);
503	if (!dodump)
504		return;
505	if (dumpdev == NODEV)
506		return;
507	if (!(devsw(dumpdev)))
508		return;
509	if (!(devsw(dumpdev)->d_dump))
510		return;
511	if (dumping++) {
512		dumping--;
513		printf("Dump already in progress, bailing...\n");
514		return;
515	}
516	dumpsize = Maxmem;
517	printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
518	printf("dump ");
519	error = (*devsw(dumpdev)->d_dump)(dumpdev);
520	dumping--;
521	if (error == 0) {
522		printf("succeeded\n");
523		return;
524	}
525	printf("failed, reason: ");
526	switch (error) {
527	case ENODEV:
528		printf("device doesn't support a dump routine\n");
529		break;
530
531	case ENXIO:
532		printf("device bad\n");
533		break;
534
535	case EFAULT:
536		printf("device not ready\n");
537		break;
538
539	case EINVAL:
540		printf("area improper\n");
541		break;
542
543	case EIO:
544		printf("i/o error\n");
545		break;
546
547	case EINTR:
548		printf("aborted from console\n");
549		break;
550
551	default:
552		printf("unknown, error = %d\n", error);
553		break;
554	}
555}
556
557int
558dumpstatus(vm_offset_t addr, off_t count)
559{
560	int c;
561
562	if (addr % (1024 * 1024) == 0) {
563#ifdef HW_WDOG
564		if (wdog_tickler)
565			(*wdog_tickler)();
566#endif
567		printf("%ld ", (long)(count / (1024 * 1024)));
568	}
569
570	if ((c = cncheckc()) == 0x03)
571		return -1;
572	else if (c != -1)
573		printf("[CTRL-C to abort] ");
574
575	return 0;
576}
577
578#ifdef SMP
579static u_int panic_cpu = NOCPU;
580#endif
581
582/*
583 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
584 * and then reboots.  If we are called twice, then we avoid trying to sync
585 * the disks as this often leads to recursive panics.
586 *
587 * MPSAFE
588 */
589void
590panic(const char *fmt, ...)
591{
592	int bootopt;
593	va_list ap;
594	static char buf[256];
595
596#ifdef SMP
597	/*
598	 * We don't want multiple CPU's to panic at the same time, so we
599	 * use panic_cpu as a simple spinlock.  We have to keep checking
600	 * panic_cpu if we are spinning in case the panic on the first
601	 * CPU is canceled.
602	 */
603	if (panic_cpu != PCPU_GET(cpuid))
604		while (atomic_cmpset_int(&panic_cpu, NOCPU,
605		    PCPU_GET(cpuid)) == 0)
606			while (panic_cpu != NOCPU)
607				; /* nothing */
608#endif
609
610	bootopt = RB_AUTOBOOT | RB_DUMP;
611	if (panicstr)
612		bootopt |= RB_NOSYNC;
613	else
614		panicstr = fmt;
615
616	va_start(ap, fmt);
617	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
618	if (panicstr == fmt)
619		panicstr = buf;
620	va_end(ap);
621	printf("panic: %s\n", buf);
622#ifdef SMP
623	/* two separate prints in case of an unmapped page and trap */
624	printf("cpuid = %d; ", PCPU_GET(cpuid));
625#ifdef APIC_IO
626	printf("lapic.id = %08x\n", lapic.id);
627#endif
628#endif
629
630#if defined(DDB)
631	if (debugger_on_panic)
632		Debugger ("panic");
633#ifdef RESTARTABLE_PANICS
634	/* See if the user aborted the panic, in which case we continue. */
635	if (panicstr == NULL) {
636#ifdef SMP
637		atomic_store_rel_int(&panic_cpu, NOCPU);
638#endif
639		return;
640	}
641#endif
642#endif
643	if (!sync_on_panic)
644		bootopt |= RB_NOSYNC;
645	boot(bootopt);
646}
647
648/*
649 * Support for poweroff delay.
650 */
651#ifndef POWEROFF_DELAY
652# define POWEROFF_DELAY 5000
653#endif
654static int poweroff_delay = POWEROFF_DELAY;
655
656SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
657	&poweroff_delay, 0, "");
658
659static void
660poweroff_wait(void *junk, int howto)
661{
662	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
663		return;
664	DELAY(poweroff_delay * 1000);
665}
666
667/*
668 * Some system processes (e.g. syncer) need to be stopped at appropriate
669 * points in their main loops prior to a system shutdown, so that they
670 * won't interfere with the shutdown process (e.g. by holding a disk buf
671 * to cause sync to fail).  For each of these system processes, register
672 * shutdown_kproc() as a handler for one of shutdown events.
673 */
674static int kproc_shutdown_wait = 60;
675SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
676    &kproc_shutdown_wait, 0, "");
677
678void
679kproc_shutdown(void *arg, int howto)
680{
681	struct proc *p;
682	int error;
683
684	if (panicstr)
685		return;
686
687	p = (struct proc *)arg;
688	printf("Waiting (max %d seconds) for system process `%s' to stop...",
689	    kproc_shutdown_wait, p->p_comm);
690	error = kthread_suspend(p, kproc_shutdown_wait * hz);
691
692	if (error == EWOULDBLOCK)
693		printf("timed out\n");
694	else
695		printf("stopped\n");
696}
697