kern_shutdown.c revision 164033 
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 164033 2006-11-06 13:42:10Z rwatson $");
39
40#include "opt_kdb.h"
41#include "opt_mac.h"
42#include "opt_panic.h"
43#include "opt_show_busybufs.h"
44#include "opt_sched.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/malloc.h>
57#include <sys/mount.h>
58#include <sys/priv.h>
59#include <sys/proc.h>
60#include <sys/reboot.h>
61#include <sys/resourcevar.h>
62#include <sys/sched.h>
63#include <sys/smp.h>		/* smp_active */
64#include <sys/sysctl.h>
65#include <sys/sysproto.h>
66
67#include <machine/cpu.h>
68#include <machine/pcb.h>
69#include <machine/smp.h>
70
71#include <security/mac/mac_framework.h>
72
73#include <vm/vm.h>
74#include <vm/vm_object.h>
75#include <vm/vm_page.h>
76#include <vm/vm_pager.h>
77#include <vm/swap_pager.h>
78
79#include <sys/signalvar.h>
80
81#ifndef PANIC_REBOOT_WAIT_TIME
82#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
83#endif
84
85/*
86 * Note that stdarg.h and the ANSI style va_start macro is used for both
87 * ANSI and traditional C compilers.
88 */
89#include <machine/stdarg.h>
90
91#ifdef KDB
92#ifdef KDB_UNATTENDED
93int debugger_on_panic = 0;
94#else
95int debugger_on_panic = 1;
96#endif
97SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
98	&debugger_on_panic, 0, "Run debugger on kernel panic");
99
100#ifdef KDB_TRACE
101int trace_on_panic = 1;
102#else
103int trace_on_panic = 0;
104#endif
105SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
106	&trace_on_panic, 0, "Print stack trace on kernel panic");
107#endif /* KDB */
108
109int sync_on_panic = 0;
110SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
111	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
112
113SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
114
115/*
116 * Variable panicstr contains argument to first call to panic; used as flag
117 * to indicate that the kernel has already called panic.
118 */
119const char *panicstr;
120
121int dumping;				/* system is dumping */
122int rebooting;				/* system is rebooting */
123static struct dumperinfo dumper;	/* our selected dumper */
124
125/* Context information for dump-debuggers. */
126static struct pcb dumppcb;		/* Registers. */
127static lwpid_t dumptid;			/* Thread ID. */
128
129static void boot(int) __dead2;
130static void poweroff_wait(void *, int);
131static void shutdown_halt(void *junk, int howto);
132static void shutdown_panic(void *junk, int howto);
133static void shutdown_reset(void *junk, int howto);
134
135/* register various local shutdown events */
136static void
137shutdown_conf(void *unused)
138{
139
140	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
141	    SHUTDOWN_PRI_FIRST);
142	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
143	    SHUTDOWN_PRI_LAST + 100);
144	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
145	    SHUTDOWN_PRI_LAST + 100);
146	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
147	    SHUTDOWN_PRI_LAST + 200);
148}
149
150SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
151
152/*
153 * The system call that results in a reboot
154 *
155 * MPSAFE
156 */
157/* ARGSUSED */
158int
159reboot(struct thread *td, struct reboot_args *uap)
160{
161	int error;
162
163	error = 0;
164#ifdef MAC
165	error = mac_check_system_reboot(td->td_ucred, uap->opt);
166#endif
167	if (error == 0)
168		error = priv_check(td, PRIV_REBOOT);
169	if (error == 0) {
170		mtx_lock(&Giant);
171		boot(uap->opt);
172		mtx_unlock(&Giant);
173	}
174	return (error);
175}
176
177/*
178 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
179 */
180static int shutdown_howto = 0;
181
182void
183shutdown_nice(int howto)
184{
185
186	shutdown_howto = howto;
187
188	/* Send a signal to init(8) and have it shutdown the world */
189	if (initproc != NULL) {
190		PROC_LOCK(initproc);
191		psignal(initproc, SIGINT);
192		PROC_UNLOCK(initproc);
193	} else {
194		/* No init(8) running, so simply reboot */
195		boot(RB_NOSYNC);
196	}
197	return;
198}
199static int	waittime = -1;
200
201static void
202print_uptime(void)
203{
204	int f;
205	struct timespec ts;
206
207	getnanouptime(&ts);
208	printf("Uptime: ");
209	f = 0;
210	if (ts.tv_sec >= 86400) {
211		printf("%ldd", (long)ts.tv_sec / 86400);
212		ts.tv_sec %= 86400;
213		f = 1;
214	}
215	if (f || ts.tv_sec >= 3600) {
216		printf("%ldh", (long)ts.tv_sec / 3600);
217		ts.tv_sec %= 3600;
218		f = 1;
219	}
220	if (f || ts.tv_sec >= 60) {
221		printf("%ldm", (long)ts.tv_sec / 60);
222		ts.tv_sec %= 60;
223		f = 1;
224	}
225	printf("%lds\n", (long)ts.tv_sec);
226}
227
228static void
229doadump(void)
230{
231
232	/*
233	 * Sometimes people have to call this from the kernel debugger.
234	 * (if 'panic' can not dump)
235	 * Give them a clue as to why they can't dump.
236	 */
237	if (dumper.dumper == NULL) {
238		printf("Cannot dump. No dump device defined.\n");
239		return;
240	}
241
242	savectx(&dumppcb);
243	dumptid = curthread->td_tid;
244	dumping++;
245	dumpsys(&dumper);
246}
247
248static int
249isbufbusy(struct buf *bp)
250{
251	if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
252	    BUF_REFCNT(bp) > 0) ||
253	    ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
254		return (1);
255	return (0);
256}
257
258/*
259 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
260 */
261static void
262boot(int howto)
263{
264	static int first_buf_printf = 1;
265
266#if defined(SMP)
267	/*
268	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
269	 * systems don't shutdown properly (i.e., ACPI power off) if we
270	 * run on another processor.
271	 */
272	mtx_lock_spin(&sched_lock);
273	sched_bind(curthread, 0);
274	mtx_unlock_spin(&sched_lock);
275	KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
276#endif
277	/* We're in the process of rebooting. */
278	rebooting = 1;
279
280	/* collect extra flags that shutdown_nice might have set */
281	howto |= shutdown_howto;
282
283	/* We are out of the debugger now. */
284	kdb_active = 0;
285
286	/*
287	 * Do any callouts that should be done BEFORE syncing the filesystems.
288	 */
289	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
290
291	/*
292	 * Now sync filesystems
293	 */
294	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
295		register struct buf *bp;
296		int iter, nbusy, pbusy;
297#ifndef PREEMPTION
298		int subiter;
299#endif
300
301		waittime = 0;
302
303		sync(curthread, NULL);
304
305		/*
306		 * With soft updates, some buffers that are
307		 * written will be remarked as dirty until other
308		 * buffers are written.
309		 */
310		for (iter = pbusy = 0; iter < 20; iter++) {
311			nbusy = 0;
312			for (bp = &buf[nbuf]; --bp >= buf; )
313				if (isbufbusy(bp))
314					nbusy++;
315			if (nbusy == 0) {
316				if (first_buf_printf)
317					printf("All buffers synced.");
318				break;
319			}
320			if (first_buf_printf) {
321				printf("Syncing disks, buffers remaining... ");
322				first_buf_printf = 0;
323			}
324			printf("%d ", nbusy);
325			if (nbusy < pbusy)
326				iter = 0;
327			pbusy = nbusy;
328			sync(curthread, NULL);
329
330#ifdef PREEMPTION
331			/*
332			 * Drop Giant and spin for a while to allow
333			 * interrupt threads to run.
334			 */
335			DROP_GIANT();
336			DELAY(50000 * iter);
337			PICKUP_GIANT();
338#else
339			/*
340			 * Drop Giant and context switch several times to
341			 * allow interrupt threads to run.
342			 */
343			DROP_GIANT();
344			for (subiter = 0; subiter < 50 * iter; subiter++) {
345				mtx_lock_spin(&sched_lock);
346				mi_switch(SW_VOL, NULL);
347				mtx_unlock_spin(&sched_lock);
348				DELAY(1000);
349			}
350			PICKUP_GIANT();
351#endif
352		}
353		printf("\n");
354		/*
355		 * Count only busy local buffers to prevent forcing
356		 * a fsck if we're just a client of a wedged NFS server
357		 */
358		nbusy = 0;
359		for (bp = &buf[nbuf]; --bp >= buf; ) {
360			if (isbufbusy(bp)) {
361#if 0
362/* XXX: This is bogus.  We should probably have a BO_REMOTE flag instead */
363				if (bp->b_dev == NULL) {
364					TAILQ_REMOVE(&mountlist,
365					    bp->b_vp->v_mount, mnt_list);
366					continue;
367				}
368#endif
369				nbusy++;
370#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
371				printf(
372			    "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n",
373				    nbusy, bp->b_bufobj,
374				    bp->b_flags, (long)bp->b_blkno,
375				    (long)bp->b_lblkno);
376#endif
377			}
378		}
379		if (nbusy) {
380			/*
381			 * Failed to sync all blocks. Indicate this and don't
382			 * unmount filesystems (thus forcing an fsck on reboot).
383			 */
384			printf("Giving up on %d buffers\n", nbusy);
385			DELAY(5000000);	/* 5 seconds */
386		} else {
387			if (!first_buf_printf)
388				printf("Final sync complete\n");
389			/*
390			 * Unmount filesystems
391			 */
392			if (panicstr == 0)
393				vfs_unmountall();
394		}
395		swapoff_all();
396		DELAY(100000);		/* wait for console output to finish */
397	}
398
399	print_uptime();
400
401	/*
402	 * Ok, now do things that assume all filesystem activity has
403	 * been completed.
404	 */
405	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
406
407	/* XXX This doesn't disable interrupts any more.  Reconsider? */
408	splhigh();
409
410	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
411		doadump();
412
413	/* Now that we're going to really halt the system... */
414	EVENTHANDLER_INVOKE(shutdown_final, howto);
415
416	for(;;) ;	/* safety against shutdown_reset not working */
417	/* NOTREACHED */
418}
419
420/*
421 * If the shutdown was a clean halt, behave accordingly.
422 */
423static void
424shutdown_halt(void *junk, int howto)
425{
426
427	if (howto & RB_HALT) {
428		printf("\n");
429		printf("The operating system has halted.\n");
430		printf("Please press any key to reboot.\n\n");
431		switch (cngetc()) {
432		case -1:		/* No console, just die */
433			cpu_halt();
434			/* NOTREACHED */
435		default:
436			howto &= ~RB_HALT;
437			break;
438		}
439	}
440}
441
442/*
443 * Check to see if the system paniced, pause and then reboot
444 * according to the specified delay.
445 */
446static void
447shutdown_panic(void *junk, int howto)
448{
449	int loop;
450
451	if (howto & RB_DUMP) {
452		if (PANIC_REBOOT_WAIT_TIME != 0) {
453			if (PANIC_REBOOT_WAIT_TIME != -1) {
454				printf("Automatic reboot in %d seconds - "
455				       "press a key on the console to abort\n",
456					PANIC_REBOOT_WAIT_TIME);
457				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
458				     loop > 0; --loop) {
459					DELAY(1000 * 100); /* 1/10th second */
460					/* Did user type a key? */
461					if (cncheckc() != -1)
462						break;
463				}
464				if (!loop)
465					return;
466			}
467		} else { /* zero time specified - reboot NOW */
468			return;
469		}
470		printf("--> Press a key on the console to reboot,\n");
471		printf("--> or switch off the system now.\n");
472		cngetc();
473	}
474}
475
476/*
477 * Everything done, now reset
478 */
479static void
480shutdown_reset(void *junk, int howto)
481{
482
483	printf("Rebooting...\n");
484	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
485	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
486	cpu_reset();
487	/* NOTREACHED */ /* assuming reset worked */
488}
489
490#ifdef SMP
491static u_int panic_cpu = NOCPU;
492#endif
493
494/*
495 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
496 * and then reboots.  If we are called twice, then we avoid trying to sync
497 * the disks as this often leads to recursive panics.
498 *
499 * MPSAFE
500 */
501void
502panic(const char *fmt, ...)
503{
504	struct thread *td = curthread;
505	int bootopt, newpanic;
506	va_list ap;
507	static char buf[256];
508
509#ifdef SMP
510	/*
511	 * We don't want multiple CPU's to panic at the same time, so we
512	 * use panic_cpu as a simple spinlock.  We have to keep checking
513	 * panic_cpu if we are spinning in case the panic on the first
514	 * CPU is canceled.
515	 */
516	if (panic_cpu != PCPU_GET(cpuid))
517		while (atomic_cmpset_int(&panic_cpu, NOCPU,
518		    PCPU_GET(cpuid)) == 0)
519			while (panic_cpu != NOCPU)
520				; /* nothing */
521#endif
522
523	bootopt = RB_AUTOBOOT | RB_DUMP;
524	newpanic = 0;
525	if (panicstr)
526		bootopt |= RB_NOSYNC;
527	else {
528		panicstr = fmt;
529		newpanic = 1;
530	}
531
532	va_start(ap, fmt);
533	if (newpanic) {
534		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
535		panicstr = buf;
536		printf("panic: %s\n", buf);
537	} else {
538		printf("panic: ");
539		vprintf(fmt, ap);
540		printf("\n");
541	}
542	va_end(ap);
543#ifdef SMP
544	printf("cpuid = %d\n", PCPU_GET(cpuid));
545#endif
546
547#ifdef KDB
548	if (newpanic && trace_on_panic)
549		kdb_backtrace();
550	if (debugger_on_panic)
551		kdb_enter("panic");
552#ifdef RESTARTABLE_PANICS
553	/* See if the user aborted the panic, in which case we continue. */
554	if (panicstr == NULL) {
555#ifdef SMP
556		atomic_store_rel_int(&panic_cpu, NOCPU);
557#endif
558		return;
559	}
560#endif
561#endif
562	mtx_lock_spin(&sched_lock);
563	td->td_flags |= TDF_INPANIC;
564	mtx_unlock_spin(&sched_lock);
565	if (!sync_on_panic)
566		bootopt |= RB_NOSYNC;
567	boot(bootopt);
568}
569
570/*
571 * Support for poweroff delay.
572 */
573#ifndef POWEROFF_DELAY
574# define POWEROFF_DELAY 5000
575#endif
576static int poweroff_delay = POWEROFF_DELAY;
577
578SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
579	&poweroff_delay, 0, "");
580
581static void
582poweroff_wait(void *junk, int howto)
583{
584
585	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
586		return;
587	DELAY(poweroff_delay * 1000);
588}
589
590/*
591 * Some system processes (e.g. syncer) need to be stopped at appropriate
592 * points in their main loops prior to a system shutdown, so that they
593 * won't interfere with the shutdown process (e.g. by holding a disk buf
594 * to cause sync to fail).  For each of these system processes, register
595 * shutdown_kproc() as a handler for one of shutdown events.
596 */
597static int kproc_shutdown_wait = 60;
598SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
599    &kproc_shutdown_wait, 0, "");
600
601void
602kproc_shutdown(void *arg, int howto)
603{
604	struct proc *p;
605	char procname[MAXCOMLEN + 1];
606	int error;
607
608	if (panicstr)
609		return;
610
611	p = (struct proc *)arg;
612	strlcpy(procname, p->p_comm, sizeof(procname));
613	printf("Waiting (max %d seconds) for system process `%s' to stop...",
614	    kproc_shutdown_wait, procname);
615	error = kthread_suspend(p, kproc_shutdown_wait * hz);
616
617	if (error == EWOULDBLOCK)
618		printf("timed out\n");
619	else
620		printf("done\n");
621}
622
623/* Registration of dumpers */
624int
625set_dumper(struct dumperinfo *di)
626{
627
628	if (di == NULL) {
629		bzero(&dumper, sizeof dumper);
630		return (0);
631	}
632	if (dumper.dumper != NULL)
633		return (EBUSY);
634	dumper = *di;
635	return (0);
636}
637
638#if defined(__powerpc__)
639void
640dumpsys(struct dumperinfo *di __unused)
641{
642
643	printf("Kernel dumps not implemented on this architecture\n");
644}
645#endif
646