kern_shutdown.c revision 132506 
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 132506 2004-07-21 16:04:46Z rwatson $");
39
40#include "opt_kdb.h"
41#include "opt_hw_wdog.h"
42#include "opt_mac.h"
43#include "opt_panic.h"
44#include "opt_show_busybufs.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/mac.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/cpu.h>
68#include <machine/pcb.h>
69#include <machine/smp.h>
70
71#include <sys/signalvar.h>
72
73#ifndef PANIC_REBOOT_WAIT_TIME
74#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
75#endif
76
77/*
78 * Note that stdarg.h and the ANSI style va_start macro is used for both
79 * ANSI and traditional C compilers.
80 */
81#include <machine/stdarg.h>
82
83#ifdef KDB
84#ifdef KDB_UNATTENDED
85int debugger_on_panic = 0;
86#else
87int debugger_on_panic = 1;
88#endif
89SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
90	&debugger_on_panic, 0, "Run debugger on kernel panic");
91
92#ifdef KDB_TRACE
93int trace_on_panic = 1;
94#else
95int trace_on_panic = 0;
96#endif
97SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
98	&trace_on_panic, 0, "Print stack trace on kernel panic");
99#endif /* KDB */
100
101int sync_on_panic = 0;
102SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
103	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
104
105SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
106
107#ifdef	HW_WDOG
108/*
109 * If there is a hardware watchdog, point this at the function needed to
110 * hold it off.
111 * It's needed when the kernel needs to do some lengthy operations.
112 * e.g. in wd.c when dumping core.. It's most annoying to have
113 * your precious core-dump only half written because the wdog kicked in.
114 */
115watchdog_tickle_fn wdog_tickler = NULL;
116#endif	/* HW_WDOG */
117
118/*
119 * Variable panicstr contains argument to first call to panic; used as flag
120 * to indicate that the kernel has already called panic.
121 */
122const char *panicstr;
123
124int dumping;				/* system is dumping */
125static struct dumperinfo dumper;	/* our selected dumper */
126
127/* Context information for dump-debuggers. */
128static struct pcb dumppcb;		/* Registers. */
129static lwpid_t dumptid;			/* Thread ID. */
130
131static void boot(int) __dead2;
132static void poweroff_wait(void *, int);
133static void shutdown_halt(void *junk, int howto);
134static void shutdown_panic(void *junk, int howto);
135static void shutdown_reset(void *junk, int howto);
136
137/* register various local shutdown events */
138static void
139shutdown_conf(void *unused)
140{
141
142	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
143	    SHUTDOWN_PRI_FIRST);
144	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
145	    SHUTDOWN_PRI_LAST + 100);
146	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
147	    SHUTDOWN_PRI_LAST + 100);
148	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
149	    SHUTDOWN_PRI_LAST + 200);
150}
151
152SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
153
154/*
155 * The system call that results in a reboot
156 *
157 * MPSAFE
158 */
159/* ARGSUSED */
160int
161reboot(struct thread *td, struct reboot_args *uap)
162{
163	int error;
164
165	error = 0;
166#ifdef MAC
167	error = mac_check_system_reboot(td->td_ucred, uap->opt);
168#endif
169	if (error == 0)
170		error = suser(td);
171	if (error == 0) {
172		mtx_lock(&Giant);
173		boot(uap->opt);
174		mtx_unlock(&Giant);
175	}
176	return (error);
177}
178
179/*
180 * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
181 */
182static int shutdown_howto = 0;
183
184void
185shutdown_nice(int howto)
186{
187
188	shutdown_howto = howto;
189
190	/* Send a signal to init(8) and have it shutdown the world */
191	if (initproc != NULL) {
192		PROC_LOCK(initproc);
193		psignal(initproc, SIGINT);
194		PROC_UNLOCK(initproc);
195	} else {
196		/* No init(8) running, so simply reboot */
197		boot(RB_NOSYNC);
198	}
199	return;
200}
201static int	waittime = -1;
202
203static void
204print_uptime(void)
205{
206	int f;
207	struct timespec ts;
208
209	getnanouptime(&ts);
210	printf("Uptime: ");
211	f = 0;
212	if (ts.tv_sec >= 86400) {
213		printf("%ldd", (long)ts.tv_sec / 86400);
214		ts.tv_sec %= 86400;
215		f = 1;
216	}
217	if (f || ts.tv_sec >= 3600) {
218		printf("%ldh", (long)ts.tv_sec / 3600);
219		ts.tv_sec %= 3600;
220		f = 1;
221	}
222	if (f || ts.tv_sec >= 60) {
223		printf("%ldm", (long)ts.tv_sec / 60);
224		ts.tv_sec %= 60;
225		f = 1;
226	}
227	printf("%lds\n", (long)ts.tv_sec);
228}
229
230static void
231doadump(void)
232{
233
234	/*
235	 * Sometimes people have to call this from the kernel debugger.
236	 * (if 'panic' can not dump)
237	 * Give them a clue as to why they can't dump.
238	 */
239	if (dumper.dumper == NULL) {
240		printf("Cannot dump. No dump device defined.\n");
241		return;
242	}
243
244	savectx(&dumppcb);
245	dumptid = curthread->td_tid;
246	dumping++;
247	dumpsys(&dumper);
248}
249
250/*
251 *  Go through the rigmarole of shutting down..
252 * this used to be in machdep.c but I'll be dammned if I could see
253 * anything machine dependant in it.
254 */
255static void
256boot(int howto)
257{
258	static int first_buf_printf = 1;
259
260	/* collect extra flags that shutdown_nice might have set */
261	howto |= shutdown_howto;
262
263	/* We are out of the debugger now. */
264	kdb_active = 0;
265
266#ifdef SMP
267	if (smp_active)
268		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
269#endif
270	/*
271	 * Do any callouts that should be done BEFORE syncing the filesystems.
272	 */
273	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
274
275	/*
276	 * Now sync filesystems
277	 */
278	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
279		register struct buf *bp;
280		int iter, nbusy, pbusy;
281#ifndef PREEMPTION
282		int subiter;
283#endif
284
285		waittime = 0;
286
287		sync(&thread0, NULL);
288
289		/*
290		 * With soft updates, some buffers that are
291		 * written will be remarked as dirty until other
292		 * buffers are written.
293		 */
294		for (iter = pbusy = 0; iter < 20; iter++) {
295			nbusy = 0;
296			for (bp = &buf[nbuf]; --bp >= buf; ) {
297				if ((bp->b_flags & B_INVAL) == 0 &&
298				    BUF_REFCNT(bp) > 0) {
299					nbusy++;
300				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
301						== B_DELWRI) {
302					/* bawrite(bp);*/
303					nbusy++;
304				}
305			}
306			if (nbusy == 0)
307				break;
308			if (first_buf_printf) {
309				printf("syncing disks, buffers remaining... ");
310				first_buf_printf = 0;
311			}
312			printf("%d ", nbusy);
313			if (nbusy < pbusy)
314				iter = 0;
315			pbusy = nbusy;
316			sync(&thread0, NULL);
317
318#ifdef PREEMPTION
319			/*
320			 * Drop Giant and spin for a while to allow
321			 * interrupt threads to run.
322			 */
323			DROP_GIANT();
324			DELAY(50000 * iter);
325			PICKUP_GIANT();
326#else
327			/*
328			 * Drop Giant and context switch several times to
329			 * allow interrupt threads to run.
330			 */
331			DROP_GIANT();
332			for (subiter = 0; subiter < 50 * iter; subiter++) {
333				mtx_lock_spin(&sched_lock);
334				mi_switch(SW_VOL, NULL);
335				mtx_unlock_spin(&sched_lock);
336				DELAY(1000);
337			}
338			PICKUP_GIANT();
339#endif
340		}
341
342		/*
343		 * Count only busy local buffers to prevent forcing
344		 * a fsck if we're just a client of a wedged NFS server
345		 */
346		nbusy = 0;
347		for (bp = &buf[nbuf]; --bp >= buf; ) {
348			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
349			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
350				if (bp->b_dev == NULL) {
351					TAILQ_REMOVE(&mountlist,
352					    bp->b_vp->v_mount, mnt_list);
353					continue;
354				}
355				nbusy++;
356#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
357				printf(
358			    "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n",
359				    nbusy, devtoname(bp->b_dev),
360				    bp->b_flags, (long)bp->b_blkno,
361				    (long)bp->b_lblkno);
362#endif
363			}
364		}
365		if (nbusy) {
366			/*
367			 * Failed to sync all blocks. Indicate this and don't
368			 * unmount filesystems (thus forcing an fsck on reboot).
369			 */
370			printf("giving up on %d buffers\n", nbusy);
371			DELAY(5000000);	/* 5 seconds */
372		} else {
373			printf("done\n");
374			/*
375			 * Unmount filesystems
376			 */
377			if (panicstr == 0)
378				vfs_unmountall();
379		}
380		DELAY(100000);		/* wait for console output to finish */
381	}
382
383	print_uptime();
384
385	/*
386	 * Ok, now do things that assume all filesystem activity has
387	 * been completed.
388	 */
389	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
390	splhigh();
391	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
392		doadump();
393
394	/* Now that we're going to really halt the system... */
395	EVENTHANDLER_INVOKE(shutdown_final, howto);
396
397	for(;;) ;	/* safety against shutdown_reset not working */
398	/* NOTREACHED */
399}
400
401/*
402 * If the shutdown was a clean halt, behave accordingly.
403 */
404static void
405shutdown_halt(void *junk, int howto)
406{
407
408	if (howto & RB_HALT) {
409		printf("\n");
410		printf("The operating system has halted.\n");
411		printf("Please press any key to reboot.\n\n");
412		switch (cngetc()) {
413		case -1:		/* No console, just die */
414			cpu_halt();
415			/* NOTREACHED */
416		default:
417			howto &= ~RB_HALT;
418			break;
419		}
420	}
421}
422
423/*
424 * Check to see if the system paniced, pause and then reboot
425 * according to the specified delay.
426 */
427static void
428shutdown_panic(void *junk, int howto)
429{
430	int loop;
431
432	if (howto & RB_DUMP) {
433		if (PANIC_REBOOT_WAIT_TIME != 0) {
434			if (PANIC_REBOOT_WAIT_TIME != -1) {
435				printf("Automatic reboot in %d seconds - "
436				       "press a key on the console to abort\n",
437					PANIC_REBOOT_WAIT_TIME);
438				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
439				     loop > 0; --loop) {
440					DELAY(1000 * 100); /* 1/10th second */
441					/* Did user type a key? */
442					if (cncheckc() != -1)
443						break;
444				}
445				if (!loop)
446					return;
447			}
448		} else { /* zero time specified - reboot NOW */
449			return;
450		}
451		printf("--> Press a key on the console to reboot,\n");
452		printf("--> or switch off the system now.\n");
453		cngetc();
454	}
455}
456
457/*
458 * Everything done, now reset
459 */
460static void
461shutdown_reset(void *junk, int howto)
462{
463
464	printf("Rebooting...\n");
465	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
466	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
467	cpu_reset();
468	/* NOTREACHED */ /* assuming reset worked */
469}
470
471#ifdef SMP
472static u_int panic_cpu = NOCPU;
473#endif
474
475/*
476 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
477 * and then reboots.  If we are called twice, then we avoid trying to sync
478 * the disks as this often leads to recursive panics.
479 *
480 * MPSAFE
481 */
482void
483panic(const char *fmt, ...)
484{
485	struct thread *td = curthread;
486	int bootopt, newpanic;
487	va_list ap;
488	static char buf[256];
489
490#ifdef SMP
491	/*
492	 * We don't want multiple CPU's to panic at the same time, so we
493	 * use panic_cpu as a simple spinlock.  We have to keep checking
494	 * panic_cpu if we are spinning in case the panic on the first
495	 * CPU is canceled.
496	 */
497	if (panic_cpu != PCPU_GET(cpuid))
498		while (atomic_cmpset_int(&panic_cpu, NOCPU,
499		    PCPU_GET(cpuid)) == 0)
500			while (panic_cpu != NOCPU)
501				; /* nothing */
502#endif
503
504	bootopt = RB_AUTOBOOT | RB_DUMP;
505	newpanic = 0;
506	if (panicstr)
507		bootopt |= RB_NOSYNC;
508	else {
509		panicstr = fmt;
510		newpanic = 1;
511	}
512
513	va_start(ap, fmt);
514	if (newpanic) {
515		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
516		panicstr = buf;
517		printf("panic: %s\n", buf);
518	} else {
519		printf("panic: ");
520		vprintf(fmt, ap);
521		printf("\n");
522	}
523	va_end(ap);
524#ifdef SMP
525	/* two separate prints in case of an unmapped page and trap */
526	printf("cpuid = %d; ", PCPU_GET(cpuid));
527#ifdef APIC_IO
528	printf("lapic.id = %08x\n", lapic.id);
529#else
530	printf("\n");
531#endif
532#endif
533
534#ifdef KDB
535	if (newpanic && trace_on_panic)
536		kdb_backtrace();
537	if (debugger_on_panic)
538		kdb_enter("panic");
539#ifdef RESTARTABLE_PANICS
540	/* See if the user aborted the panic, in which case we continue. */
541	if (panicstr == NULL) {
542#ifdef SMP
543		atomic_store_rel_int(&panic_cpu, NOCPU);
544#endif
545		return;
546	}
547#endif
548#endif
549	mtx_lock_spin(&sched_lock);
550	td->td_flags |= TDF_INPANIC;
551	mtx_unlock_spin(&sched_lock);
552	if (!sync_on_panic)
553		bootopt |= RB_NOSYNC;
554	boot(bootopt);
555}
556
557/*
558 * Support for poweroff delay.
559 */
560#ifndef POWEROFF_DELAY
561# define POWEROFF_DELAY 5000
562#endif
563static int poweroff_delay = POWEROFF_DELAY;
564
565SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
566	&poweroff_delay, 0, "");
567
568static void
569poweroff_wait(void *junk, int howto)
570{
571
572	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
573		return;
574	DELAY(poweroff_delay * 1000);
575}
576
577/*
578 * Some system processes (e.g. syncer) need to be stopped at appropriate
579 * points in their main loops prior to a system shutdown, so that they
580 * won't interfere with the shutdown process (e.g. by holding a disk buf
581 * to cause sync to fail).  For each of these system processes, register
582 * shutdown_kproc() as a handler for one of shutdown events.
583 */
584static int kproc_shutdown_wait = 60;
585SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
586    &kproc_shutdown_wait, 0, "");
587
588void
589kproc_shutdown(void *arg, int howto)
590{
591	struct proc *p;
592	char procname[MAXCOMLEN + 1];
593	int error;
594
595	if (panicstr)
596		return;
597
598	p = (struct proc *)arg;
599	strlcpy(procname, p->p_comm, sizeof(procname));
600	printf("Waiting (max %d seconds) for system process `%s' to stop...\n",
601	    kproc_shutdown_wait, procname);
602	error = kthread_suspend(p, kproc_shutdown_wait * hz);
603
604	if (error == EWOULDBLOCK)
605		printf("Stop of '%s' timed out.\n", procname);
606	else
607		printf("Process '%s' stopped.\n", procname);
608}
609
610/* Registration of dumpers */
611int
612set_dumper(struct dumperinfo *di)
613{
614
615	if (di == NULL) {
616		bzero(&dumper, sizeof dumper);
617		return (0);
618	}
619	if (dumper.dumper != NULL)
620		return (EBUSY);
621	dumper = *di;
622	return (0);
623}
624
625#if defined(__powerpc__)
626void
627dumpsys(struct dumperinfo *di __unused)
628{
629
630	printf("Kernel dumps not implemented on this architecture\n");
631}
632#endif
633