kern_shutdown.c revision 113633 
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 113633 2003-04-17 22:29:23Z jhb $
40 */
41
42#include "opt_ddb.h"
43#include "opt_ddb_trace.h"
44#include "opt_ddb_unattended.h"
45#include "opt_hw_wdog.h"
46#include "opt_mac.h"
47#include "opt_panic.h"
48#include "opt_show_busybufs.h"
49
50#include <sys/param.h>
51#include <sys/systm.h>
52#include <sys/bio.h>
53#include <sys/buf.h>
54#include <sys/conf.h>
55#include <sys/cons.h>
56#include <sys/eventhandler.h>
57#include <sys/kernel.h>
58#include <sys/kthread.h>
59#include <sys/mac.h>
60#include <sys/malloc.h>
61#include <sys/mount.h>
62#include <sys/proc.h>
63#include <sys/reboot.h>
64#include <sys/resourcevar.h>
65#include <sys/smp.h>		/* smp_active */
66#include <sys/sysctl.h>
67#include <sys/sysproto.h>
68#include <sys/vnode.h>
69
70#include <machine/pcb.h>
71#include <machine/md_var.h>
72#include <machine/smp.h>
73
74#include <sys/signalvar.h>
75#ifdef DDB
76#include <ddb/ddb.h>
77#endif
78
79#ifndef PANIC_REBOOT_WAIT_TIME
80#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
81#endif
82
83/*
84 * Note that stdarg.h and the ANSI style va_start macro is used for both
85 * ANSI and traditional C compilers.
86 */
87#include <machine/stdarg.h>
88
89#ifdef DDB
90#ifdef DDB_UNATTENDED
91int debugger_on_panic = 0;
92#else
93int debugger_on_panic = 1;
94#endif
95SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
96	&debugger_on_panic, 0, "Run debugger on kernel panic");
97
98#ifdef DDB_TRACE
99int trace_on_panic = 1;
100#else
101int trace_on_panic = 0;
102#endif
103SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
104	&trace_on_panic, 0, "Print stack trace on kernel panic");
105#endif
106
107int sync_on_panic = 1;
108SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
109	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
110
111SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
112
113#ifdef	HW_WDOG
114/*
115 * If there is a hardware watchdog, point this at the function needed to
116 * hold it off.
117 * It's needed when the kernel needs to do some lengthy operations.
118 * e.g. in wd.c when dumping core.. It's most annoying to have
119 * your precious core-dump only half written because the wdog kicked in.
120 */
121watchdog_tickle_fn wdog_tickler = NULL;
122#endif	/* HW_WDOG */
123
124/*
125 * Variable panicstr contains argument to first call to panic; used as flag
126 * to indicate that the kernel has already called panic.
127 */
128const char *panicstr;
129
130int dumping;				/* system is dumping */
131static struct dumperinfo dumper;	/* our selected dumper */
132static struct pcb dumppcb;		/* "You Are Here" sign for dump-debuggers */
133
134static void boot(int) __dead2;
135static void poweroff_wait(void *, int);
136static void shutdown_halt(void *junk, int howto);
137static void shutdown_panic(void *junk, int howto);
138static void shutdown_reset(void *junk, int howto);
139
140/* register various local shutdown events */
141static void
142shutdown_conf(void *unused)
143{
144
145	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
146	    SHUTDOWN_PRI_FIRST);
147	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
148	    SHUTDOWN_PRI_LAST + 100);
149	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
150	    SHUTDOWN_PRI_LAST + 100);
151	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
152	    SHUTDOWN_PRI_LAST + 200);
153}
154
155SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
156
157/*
158 * The system call that results in a reboot
159 *
160 * MPSAFE
161 */
162/* ARGSUSED */
163int
164reboot(struct thread *td, struct reboot_args *uap)
165{
166	int error;
167
168	error = 0;
169#ifdef MAC
170	error = mac_check_system_reboot(td->td_ucred, uap->opt);
171#endif
172	if (error == 0)
173		error = suser(td);
174	if (error == 0) {
175		mtx_lock(&Giant);
176		boot(uap->opt);
177		mtx_unlock(&Giant);
178	}
179	return (error);
180}
181
182/*
183 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
184 */
185static int shutdown_howto = 0;
186
187void
188shutdown_nice(int howto)
189{
190
191	shutdown_howto = howto;
192
193	/* Send a signal to init(8) and have it shutdown the world */
194	if (initproc != NULL) {
195		PROC_LOCK(initproc);
196		psignal(initproc, SIGINT);
197		PROC_UNLOCK(initproc);
198	} else {
199		/* No init(8) running, so simply reboot */
200		boot(RB_NOSYNC);
201	}
202	return;
203}
204static int	waittime = -1;
205
206static void
207print_uptime(void)
208{
209	int f;
210	struct timespec ts;
211
212	getnanouptime(&ts);
213	printf("Uptime: ");
214	f = 0;
215	if (ts.tv_sec >= 86400) {
216		printf("%ldd", (long)ts.tv_sec / 86400);
217		ts.tv_sec %= 86400;
218		f = 1;
219	}
220	if (f || ts.tv_sec >= 3600) {
221		printf("%ldh", (long)ts.tv_sec / 3600);
222		ts.tv_sec %= 3600;
223		f = 1;
224	}
225	if (f || ts.tv_sec >= 60) {
226		printf("%ldm", (long)ts.tv_sec / 60);
227		ts.tv_sec %= 60;
228		f = 1;
229	}
230	printf("%lds\n", (long)ts.tv_sec);
231}
232
233static void
234doadump(void)
235{
236
237	savectx(&dumppcb);
238	dumping++;
239	dumpsys(&dumper);
240}
241
242/*
243 *  Go through the rigmarole of shutting down..
244 * this used to be in machdep.c but I'll be dammned if I could see
245 * anything machine dependant in it.
246 */
247static void
248boot(int howto)
249{
250
251	/* collect extra flags that shutdown_nice might have set */
252	howto |= shutdown_howto;
253
254#ifdef DDB
255	/* We are out of the debugger now. */
256	db_active = 0;
257#endif
258
259#ifdef SMP
260	if (smp_active)
261		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
262#endif
263	/*
264	 * Do any callouts that should be done BEFORE syncing the filesystems.
265	 */
266	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
267
268	/*
269	 * Now sync filesystems
270	 */
271	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
272		register struct buf *bp;
273		int iter, nbusy, pbusy;
274		int subiter;
275
276		waittime = 0;
277		printf("\nsyncing disks, buffers remaining... ");
278
279		sync(&thread0, NULL);
280
281		/*
282		 * With soft updates, some buffers that are
283		 * written will be remarked as dirty until other
284		 * buffers are written.
285		 */
286		for (iter = pbusy = 0; iter < 20; iter++) {
287			nbusy = 0;
288			for (bp = &buf[nbuf]; --bp >= buf; ) {
289				if ((bp->b_flags & B_INVAL) == 0 &&
290				    BUF_REFCNT(bp) > 0) {
291					nbusy++;
292				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
293						== B_DELWRI) {
294					/* bawrite(bp);*/
295					nbusy++;
296				}
297			}
298			if (nbusy == 0)
299				break;
300			printf("%d ", nbusy);
301			if (nbusy < pbusy)
302				iter = 0;
303			pbusy = nbusy;
304			sync(&thread0, NULL);
305 			if (curthread != NULL) {
306				DROP_GIANT();
307   				for (subiter = 0; subiter < 50 * iter; subiter++) {
308     					mtx_lock_spin(&sched_lock);
309					curthread->td_proc->p_stats->p_ru.ru_nvcsw++;
310     					mi_switch(); /* Allow interrupt threads to run */
311     					mtx_unlock_spin(&sched_lock);
312     					DELAY(1000);
313   				}
314				PICKUP_GIANT();
315 			} else
316			DELAY(50000 * iter);
317		}
318		printf("\n");
319		/*
320		 * Count only busy local buffers to prevent forcing
321		 * a fsck if we're just a client of a wedged NFS server
322		 */
323		nbusy = 0;
324		for (bp = &buf[nbuf]; --bp >= buf; ) {
325			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
326			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
327				if (bp->b_dev == NODEV) {
328					TAILQ_REMOVE(&mountlist,
329					    bp->b_vp->v_mount, mnt_list);
330					continue;
331				}
332				nbusy++;
333#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
334				printf(
335			    "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n",
336				    nbusy, devtoname(bp->b_dev),
337				    bp->b_flags, (long)bp->b_blkno,
338				    (long)bp->b_lblkno);
339#endif
340			}
341		}
342		if (nbusy) {
343			/*
344			 * Failed to sync all blocks. Indicate this and don't
345			 * unmount filesystems (thus forcing an fsck on reboot).
346			 */
347			printf("giving up on %d buffers\n", nbusy);
348			DELAY(5000000);	/* 5 seconds */
349		} else {
350			printf("done\n");
351			/*
352			 * Unmount filesystems
353			 */
354			if (panicstr == 0)
355				vfs_unmountall();
356		}
357		DELAY(100000);		/* wait for console output to finish */
358	}
359
360	print_uptime();
361
362	/*
363	 * Ok, now do things that assume all filesystem activity has
364	 * been completed.
365	 */
366	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
367	splhigh();
368	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP &&
369	    !cold && dumper.dumper != NULL && !dumping)
370		doadump();
371
372	/* Now that we're going to really halt the system... */
373	EVENTHANDLER_INVOKE(shutdown_final, howto);
374
375	for(;;) ;	/* safety against shutdown_reset not working */
376	/* NOTREACHED */
377}
378
379/*
380 * If the shutdown was a clean halt, behave accordingly.
381 */
382static void
383shutdown_halt(void *junk, int howto)
384{
385
386	if (howto & RB_HALT) {
387		printf("\n");
388		printf("The operating system has halted.\n");
389		printf("Please press any key to reboot.\n\n");
390		switch (cngetc()) {
391		case -1:		/* No console, just die */
392			cpu_halt();
393			/* NOTREACHED */
394		default:
395			howto &= ~RB_HALT;
396			break;
397		}
398	}
399}
400
401/*
402 * Check to see if the system paniced, pause and then reboot
403 * according to the specified delay.
404 */
405static void
406shutdown_panic(void *junk, int howto)
407{
408	int loop;
409
410	if (howto & RB_DUMP) {
411		if (PANIC_REBOOT_WAIT_TIME != 0) {
412			if (PANIC_REBOOT_WAIT_TIME != -1) {
413				printf("Automatic reboot in %d seconds - "
414				       "press a key on the console to abort\n",
415					PANIC_REBOOT_WAIT_TIME);
416				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
417				     loop > 0; --loop) {
418					DELAY(1000 * 100); /* 1/10th second */
419					/* Did user type a key? */
420					if (cncheckc() != -1)
421						break;
422				}
423				if (!loop)
424					return;
425			}
426		} else { /* zero time specified - reboot NOW */
427			return;
428		}
429		printf("--> Press a key on the console to reboot,\n");
430		printf("--> or switch off the system now.\n");
431		cngetc();
432	}
433}
434
435/*
436 * Everything done, now reset
437 */
438static void
439shutdown_reset(void *junk, int howto)
440{
441
442	printf("Rebooting...\n");
443	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
444	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
445	cpu_reset();
446	/* NOTREACHED */ /* assuming reset worked */
447}
448
449/*
450 * Print a backtrace if we can.
451 */
452
453void
454backtrace(void)
455{
456
457#ifdef DDB
458	printf("Stack backtrace:\n");
459	db_print_backtrace();
460#else
461	printf("Sorry, need DDB option to print backtrace");
462#endif
463}
464
465#ifdef SMP
466static u_int panic_cpu = NOCPU;
467#endif
468
469/*
470 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
471 * and then reboots.  If we are called twice, then we avoid trying to sync
472 * the disks as this often leads to recursive panics.
473 *
474 * MPSAFE
475 */
476void
477panic(const char *fmt, ...)
478{
479	struct thread *td = curthread;
480	int bootopt, newpanic;
481	va_list ap;
482	static char buf[256];
483
484#ifdef SMP
485	/*
486	 * We don't want multiple CPU's to panic at the same time, so we
487	 * use panic_cpu as a simple spinlock.  We have to keep checking
488	 * panic_cpu if we are spinning in case the panic on the first
489	 * CPU is canceled.
490	 */
491	if (panic_cpu != PCPU_GET(cpuid))
492		while (atomic_cmpset_int(&panic_cpu, NOCPU,
493		    PCPU_GET(cpuid)) == 0)
494			while (panic_cpu != NOCPU)
495				; /* nothing */
496#endif
497
498	bootopt = RB_AUTOBOOT | RB_DUMP;
499	newpanic = 0;
500	if (panicstr)
501		bootopt |= RB_NOSYNC;
502	else {
503		panicstr = fmt;
504		newpanic = 1;
505	}
506
507	va_start(ap, fmt);
508	(void)vsnprintf(buf, sizeof(buf), fmt, ap);
509	if (panicstr == fmt)
510		panicstr = buf;
511	va_end(ap);
512	printf("panic: %s\n", buf);
513#ifdef SMP
514	/* two separate prints in case of an unmapped page and trap */
515	printf("cpuid = %d; ", PCPU_GET(cpuid));
516#ifdef APIC_IO
517	printf("lapic.id = %08x\n", lapic.id);
518#else
519	printf("\n");
520#endif
521#endif
522
523#if defined(DDB)
524	if (newpanic && trace_on_panic)
525		backtrace();
526	if (debugger_on_panic)
527		Debugger ("panic");
528#ifdef RESTARTABLE_PANICS
529	/* See if the user aborted the panic, in which case we continue. */
530	if (panicstr == NULL) {
531#ifdef SMP
532		atomic_store_rel_int(&panic_cpu, NOCPU);
533#endif
534		return;
535	}
536#endif
537#endif
538	mtx_lock_spin(&sched_lock);
539	td->td_flags |= TDF_INPANIC;
540	mtx_unlock_spin(&sched_lock);
541	if (!sync_on_panic)
542		bootopt |= RB_NOSYNC;
543	boot(bootopt);
544}
545
546/*
547 * Support for poweroff delay.
548 */
549#ifndef POWEROFF_DELAY
550# define POWEROFF_DELAY 5000
551#endif
552static int poweroff_delay = POWEROFF_DELAY;
553
554SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
555	&poweroff_delay, 0, "");
556
557static void
558poweroff_wait(void *junk, int howto)
559{
560
561	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
562		return;
563	DELAY(poweroff_delay * 1000);
564}
565
566/*
567 * Some system processes (e.g. syncer) need to be stopped at appropriate
568 * points in their main loops prior to a system shutdown, so that they
569 * won't interfere with the shutdown process (e.g. by holding a disk buf
570 * to cause sync to fail).  For each of these system processes, register
571 * shutdown_kproc() as a handler for one of shutdown events.
572 */
573static int kproc_shutdown_wait = 60;
574SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
575    &kproc_shutdown_wait, 0, "");
576
577void
578kproc_shutdown(void *arg, int howto)
579{
580	struct proc *p;
581	int error;
582
583	if (panicstr)
584		return;
585
586	p = (struct proc *)arg;
587	printf("Waiting (max %d seconds) for system process `%s' to stop...",
588	    kproc_shutdown_wait, p->p_comm);
589	error = kthread_suspend(p, kproc_shutdown_wait * hz);
590
591	if (error == EWOULDBLOCK)
592		printf("timed out\n");
593	else
594		printf("stopped\n");
595}
596
597/* Registration of dumpers */
598int
599set_dumper(struct dumperinfo *di)
600{
601
602	if (di == NULL) {
603		bzero(&dumper, sizeof dumper);
604		return (0);
605	}
606	if (dumper.dumper != NULL)
607		return (EBUSY);
608	dumper = *di;
609	return (0);
610}
611
612#if defined(__powerpc__)
613void
614dumpsys(struct dumperinfo *di __unused)
615{
616
617	printf("Kernel dumps not implemented on this architecture\n");
618}
619#endif
620