kern_shutdown.c revision 172930 
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 172930 2007-10-24 19:04:04Z 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/* ARGSUSED */
156int
157reboot(struct thread *td, struct reboot_args *uap)
158{
159	int error;
160
161	error = 0;
162#ifdef MAC
163	error = mac_system_check_reboot(td->td_ucred, uap->opt);
164#endif
165	if (error == 0)
166		error = priv_check(td, PRIV_REBOOT);
167	if (error == 0) {
168		mtx_lock(&Giant);
169		boot(uap->opt);
170		mtx_unlock(&Giant);
171	}
172	return (error);
173}
174
175/*
176 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
177 */
178static int shutdown_howto = 0;
179
180void
181shutdown_nice(int howto)
182{
183
184	shutdown_howto = howto;
185
186	/* Send a signal to init(8) and have it shutdown the world */
187	if (initproc != NULL) {
188		PROC_LOCK(initproc);
189		psignal(initproc, SIGINT);
190		PROC_UNLOCK(initproc);
191	} else {
192		/* No init(8) running, so simply reboot */
193		boot(RB_NOSYNC);
194	}
195	return;
196}
197static int	waittime = -1;
198
199static void
200print_uptime(void)
201{
202	int f;
203	struct timespec ts;
204
205	getnanouptime(&ts);
206	printf("Uptime: ");
207	f = 0;
208	if (ts.tv_sec >= 86400) {
209		printf("%ldd", (long)ts.tv_sec / 86400);
210		ts.tv_sec %= 86400;
211		f = 1;
212	}
213	if (f || ts.tv_sec >= 3600) {
214		printf("%ldh", (long)ts.tv_sec / 3600);
215		ts.tv_sec %= 3600;
216		f = 1;
217	}
218	if (f || ts.tv_sec >= 60) {
219		printf("%ldm", (long)ts.tv_sec / 60);
220		ts.tv_sec %= 60;
221		f = 1;
222	}
223	printf("%lds\n", (long)ts.tv_sec);
224}
225
226static void
227doadump(void)
228{
229
230	/*
231	 * Sometimes people have to call this from the kernel debugger.
232	 * (if 'panic' can not dump)
233	 * Give them a clue as to why they can't dump.
234	 */
235	if (dumper.dumper == NULL) {
236		printf("Cannot dump. No dump device defined.\n");
237		return;
238	}
239
240	savectx(&dumppcb);
241	dumptid = curthread->td_tid;
242	dumping++;
243	dumpsys(&dumper);
244}
245
246static int
247isbufbusy(struct buf *bp)
248{
249	if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
250	    BUF_REFCNT(bp) > 0) ||
251	    ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
252		return (1);
253	return (0);
254}
255
256/*
257 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
258 */
259static void
260boot(int howto)
261{
262	static int first_buf_printf = 1;
263
264#if defined(SMP)
265	/*
266	 * Bind us to CPU 0 so that all shutdown code runs there.  Some
267	 * systems don't shutdown properly (i.e., ACPI power off) if we
268	 * run on another processor.
269	 */
270	thread_lock(curthread);
271	sched_bind(curthread, 0);
272	thread_unlock(curthread);
273	KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
274#endif
275	/* We're in the process of rebooting. */
276	rebooting = 1;
277
278	/* collect extra flags that shutdown_nice might have set */
279	howto |= shutdown_howto;
280
281	/* We are out of the debugger now. */
282	kdb_active = 0;
283
284	/*
285	 * Do any callouts that should be done BEFORE syncing the filesystems.
286	 */
287	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
288
289	/*
290	 * Now sync filesystems
291	 */
292	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
293		register struct buf *bp;
294		int iter, nbusy, pbusy;
295#ifndef PREEMPTION
296		int subiter;
297#endif
298
299		waittime = 0;
300
301		sync(curthread, NULL);
302
303		/*
304		 * With soft updates, some buffers that are
305		 * written will be remarked as dirty until other
306		 * buffers are written.
307		 */
308		for (iter = pbusy = 0; iter < 20; iter++) {
309			nbusy = 0;
310			for (bp = &buf[nbuf]; --bp >= buf; )
311				if (isbufbusy(bp))
312					nbusy++;
313			if (nbusy == 0) {
314				if (first_buf_printf)
315					printf("All buffers synced.");
316				break;
317			}
318			if (first_buf_printf) {
319				printf("Syncing disks, buffers remaining... ");
320				first_buf_printf = 0;
321			}
322			printf("%d ", nbusy);
323			if (nbusy < pbusy)
324				iter = 0;
325			pbusy = nbusy;
326			sync(curthread, NULL);
327
328#ifdef PREEMPTION
329			/*
330			 * Drop Giant and spin for a while to allow
331			 * interrupt threads to run.
332			 */
333			DROP_GIANT();
334			DELAY(50000 * iter);
335			PICKUP_GIANT();
336#else
337			/*
338			 * Drop Giant and context switch several times to
339			 * allow interrupt threads to run.
340			 */
341			DROP_GIANT();
342			for (subiter = 0; subiter < 50 * iter; subiter++) {
343				thread_lock(curthread);
344				mi_switch(SW_VOL, NULL);
345				thread_unlock(curthread);
346				DELAY(1000);
347			}
348			PICKUP_GIANT();
349#endif
350		}
351		printf("\n");
352		/*
353		 * Count only busy local buffers to prevent forcing
354		 * a fsck if we're just a client of a wedged NFS server
355		 */
356		nbusy = 0;
357		for (bp = &buf[nbuf]; --bp >= buf; ) {
358			if (isbufbusy(bp)) {
359#if 0
360/* XXX: This is bogus.  We should probably have a BO_REMOTE flag instead */
361				if (bp->b_dev == NULL) {
362					TAILQ_REMOVE(&mountlist,
363					    bp->b_vp->v_mount, mnt_list);
364					continue;
365				}
366#endif
367				nbusy++;
368#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
369				printf(
370			    "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n",
371				    nbusy, bp->b_bufobj,
372				    bp->b_flags, (long)bp->b_blkno,
373				    (long)bp->b_lblkno);
374#endif
375			}
376		}
377		if (nbusy) {
378			/*
379			 * Failed to sync all blocks. Indicate this and don't
380			 * unmount filesystems (thus forcing an fsck on reboot).
381			 */
382			printf("Giving up on %d buffers\n", nbusy);
383			DELAY(5000000);	/* 5 seconds */
384		} else {
385			if (!first_buf_printf)
386				printf("Final sync complete\n");
387			/*
388			 * Unmount filesystems
389			 */
390			if (panicstr == 0)
391				vfs_unmountall();
392		}
393		swapoff_all();
394		DELAY(100000);		/* wait for console output to finish */
395	}
396
397	print_uptime();
398
399	/*
400	 * Ok, now do things that assume all filesystem activity has
401	 * been completed.
402	 */
403	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
404
405	/* XXX This doesn't disable interrupts any more.  Reconsider? */
406	splhigh();
407
408	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
409		doadump();
410
411	/* Now that we're going to really halt the system... */
412	EVENTHANDLER_INVOKE(shutdown_final, howto);
413
414	for(;;) ;	/* safety against shutdown_reset not working */
415	/* NOTREACHED */
416}
417
418/*
419 * If the shutdown was a clean halt, behave accordingly.
420 */
421static void
422shutdown_halt(void *junk, int howto)
423{
424
425	if (howto & RB_HALT) {
426		printf("\n");
427		printf("The operating system has halted.\n");
428		printf("Please press any key to reboot.\n\n");
429		switch (cngetc()) {
430		case -1:		/* No console, just die */
431			cpu_halt();
432			/* NOTREACHED */
433		default:
434			howto &= ~RB_HALT;
435			break;
436		}
437	}
438}
439
440/*
441 * Check to see if the system paniced, pause and then reboot
442 * according to the specified delay.
443 */
444static void
445shutdown_panic(void *junk, int howto)
446{
447	int loop;
448
449	if (howto & RB_DUMP) {
450		if (PANIC_REBOOT_WAIT_TIME != 0) {
451			if (PANIC_REBOOT_WAIT_TIME != -1) {
452				printf("Automatic reboot in %d seconds - "
453				       "press a key on the console to abort\n",
454					PANIC_REBOOT_WAIT_TIME);
455				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
456				     loop > 0; --loop) {
457					DELAY(1000 * 100); /* 1/10th second */
458					/* Did user type a key? */
459					if (cncheckc() != -1)
460						break;
461				}
462				if (!loop)
463					return;
464			}
465		} else { /* zero time specified - reboot NOW */
466			return;
467		}
468		printf("--> Press a key on the console to reboot,\n");
469		printf("--> or switch off the system now.\n");
470		cngetc();
471	}
472}
473
474/*
475 * Everything done, now reset
476 */
477static void
478shutdown_reset(void *junk, int howto)
479{
480
481	printf("Rebooting...\n");
482	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
483	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
484	cpu_reset();
485	/* NOTREACHED */ /* assuming reset worked */
486}
487
488#ifdef SMP
489static u_int panic_cpu = NOCPU;
490#endif
491
492/*
493 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
494 * and then reboots.  If we are called twice, then we avoid trying to sync
495 * the disks as this often leads to recursive panics.
496 */
497void
498panic(const char *fmt, ...)
499{
500	struct thread *td = curthread;
501	int bootopt, newpanic;
502	va_list ap;
503	static char buf[256];
504
505#ifdef SMP
506	/*
507	 * We don't want multiple CPU's to panic at the same time, so we
508	 * use panic_cpu as a simple spinlock.  We have to keep checking
509	 * panic_cpu if we are spinning in case the panic on the first
510	 * CPU is canceled.
511	 */
512	if (panic_cpu != PCPU_GET(cpuid))
513		while (atomic_cmpset_int(&panic_cpu, NOCPU,
514		    PCPU_GET(cpuid)) == 0)
515			while (panic_cpu != NOCPU)
516				; /* nothing */
517#endif
518
519	bootopt = RB_AUTOBOOT | RB_DUMP;
520	newpanic = 0;
521	if (panicstr)
522		bootopt |= RB_NOSYNC;
523	else {
524		panicstr = fmt;
525		newpanic = 1;
526	}
527
528	va_start(ap, fmt);
529	if (newpanic) {
530		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
531		panicstr = buf;
532		printf("panic: %s\n", buf);
533	} else {
534		printf("panic: ");
535		vprintf(fmt, ap);
536		printf("\n");
537	}
538	va_end(ap);
539#ifdef SMP
540	printf("cpuid = %d\n", PCPU_GET(cpuid));
541#endif
542
543#ifdef KDB
544	if (newpanic && trace_on_panic)
545		kdb_backtrace();
546	if (debugger_on_panic)
547		kdb_enter("panic");
548#ifdef RESTARTABLE_PANICS
549	/* See if the user aborted the panic, in which case we continue. */
550	if (panicstr == NULL) {
551#ifdef SMP
552		atomic_store_rel_int(&panic_cpu, NOCPU);
553#endif
554		return;
555	}
556#endif
557#endif
558	/*thread_lock(td); */
559	td->td_flags |= TDF_INPANIC;
560	/* thread_unlock(td); */
561	if (!sync_on_panic)
562		bootopt |= RB_NOSYNC;
563	boot(bootopt);
564}
565
566/*
567 * Support for poweroff delay.
568 */
569#ifndef POWEROFF_DELAY
570# define POWEROFF_DELAY 5000
571#endif
572static int poweroff_delay = POWEROFF_DELAY;
573
574SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
575	&poweroff_delay, 0, "");
576
577static void
578poweroff_wait(void *junk, int howto)
579{
580
581	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
582		return;
583	DELAY(poweroff_delay * 1000);
584}
585
586/*
587 * Some system processes (e.g. syncer) need to be stopped at appropriate
588 * points in their main loops prior to a system shutdown, so that they
589 * won't interfere with the shutdown process (e.g. by holding a disk buf
590 * to cause sync to fail).  For each of these system processes, register
591 * shutdown_kproc() as a handler for one of shutdown events.
592 */
593static int kproc_shutdown_wait = 60;
594SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
595    &kproc_shutdown_wait, 0, "");
596
597void
598kproc_shutdown(void *arg, int howto)
599{
600	struct proc *p;
601	char procname[MAXCOMLEN + 1];
602	int error;
603
604	if (panicstr)
605		return;
606
607	p = (struct proc *)arg;
608	strlcpy(procname, p->p_comm, sizeof(procname));
609	printf("Waiting (max %d seconds) for system process `%s' to stop...",
610	    kproc_shutdown_wait, procname);
611	error = kproc_suspend(p, kproc_shutdown_wait * hz);
612
613	if (error == EWOULDBLOCK)
614		printf("timed out\n");
615	else
616		printf("done\n");
617}
618
619/* Registration of dumpers */
620int
621set_dumper(struct dumperinfo *di)
622{
623
624	if (di == NULL) {
625		bzero(&dumper, sizeof dumper);
626		return (0);
627	}
628	if (dumper.dumper != NULL)
629		return (EBUSY);
630	dumper = *di;
631	return (0);
632}
633
634#if defined(__powerpc__)
635void
636dumpsys(struct dumperinfo *di __unused)
637{
638
639	printf("Kernel dumps not implemented on this architecture\n");
640}
641#endif
642