kern_shutdown.c revision 132177 
1130803Smarcel/*-
2130803Smarcel * Copyright (c) 1986, 1988, 1991, 1993
3130803Smarcel *	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 132177 2004-07-15 04:29:48Z alfred $");
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 = 1;
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	savectx(&dumppcb);
235	dumptid = curthread->td_tid;
236	dumping++;
237	dumpsys(&dumper);
238}
239
240/*
241 *  Go through the rigmarole of shutting down..
242 * this used to be in machdep.c but I'll be dammned if I could see
243 * anything machine dependant in it.
244 */
245static void
246boot(int howto)
247{
248	int first_buf_printf;
249
250	first_buf_printf = 1;
251
252	/* collect extra flags that shutdown_nice might have set */
253	howto |= shutdown_howto;
254
255	/* We are out of the debugger now. */
256	kdb_active = 0;
257
258#ifdef SMP
259	if (smp_active)
260		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
261#endif
262	/*
263	 * Do any callouts that should be done BEFORE syncing the filesystems.
264	 */
265	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
266
267	/*
268	 * Now sync filesystems
269	 */
270	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
271		register struct buf *bp;
272		int iter, nbusy, pbusy;
273#ifndef PREEMPTION
274		int subiter;
275#endif
276
277		waittime = 0;
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			if (first_buf_printf) {
301				printf("syncing disks, buffers remaining... ");
302				first_buf_printf = 0;
303			}
304			printf("%d ", nbusy);
305			if (nbusy < pbusy)
306				iter = 0;
307			pbusy = nbusy;
308			sync(&thread0, NULL);
309
310#ifdef PREEMPTION
311			/*
312			 * Drop Giant and spin for a while to allow
313			 * interrupt threads to run.
314			 */
315			DROP_GIANT();
316			DELAY(50000 * iter);
317			PICKUP_GIANT();
318#else
319			/*
320			 * Drop Giant and context switch several times to
321			 * allow interrupt threads to run.
322			 */
323			DROP_GIANT();
324			for (subiter = 0; subiter < 50 * iter; subiter++) {
325				mtx_lock_spin(&sched_lock);
326				mi_switch(SW_VOL, NULL);
327				mtx_unlock_spin(&sched_lock);
328				DELAY(1000);
329			}
330			PICKUP_GIANT();
331#endif
332		}
333
334		/*
335		 * Count only busy local buffers to prevent forcing
336		 * a fsck if we're just a client of a wedged NFS server
337		 */
338		nbusy = 0;
339		for (bp = &buf[nbuf]; --bp >= buf; ) {
340			if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
341			    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
342				if (bp->b_dev == NULL) {
343					TAILQ_REMOVE(&mountlist,
344					    bp->b_vp->v_mount, mnt_list);
345					continue;
346				}
347				nbusy++;
348#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
349				printf(
350			    "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n",
351				    nbusy, devtoname(bp->b_dev),
352				    bp->b_flags, (long)bp->b_blkno,
353				    (long)bp->b_lblkno);
354#endif
355			}
356		}
357		if (nbusy) {
358			/*
359			 * Failed to sync all blocks. Indicate this and don't
360			 * unmount filesystems (thus forcing an fsck on reboot).
361			 */
362			printf("giving up on %d buffers\n", nbusy);
363			DELAY(5000000);	/* 5 seconds */
364		} else {
365			printf("done\n");
366			/*
367			 * Unmount filesystems
368			 */
369			if (panicstr == 0)
370				vfs_unmountall();
371		}
372		DELAY(100000);		/* wait for console output to finish */
373	}
374
375	print_uptime();
376
377	/*
378	 * Ok, now do things that assume all filesystem activity has
379	 * been completed.
380	 */
381	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
382	splhigh();
383	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP &&
384	    !cold && dumper.dumper != NULL && !dumping)
385		doadump();
386
387	/* Now that we're going to really halt the system... */
388	EVENTHANDLER_INVOKE(shutdown_final, howto);
389
390	for(;;) ;	/* safety against shutdown_reset not working */
391	/* NOTREACHED */
392}
393
394/*
395 * If the shutdown was a clean halt, behave accordingly.
396 */
397static void
398shutdown_halt(void *junk, int howto)
399{
400
401	if (howto & RB_HALT) {
402		printf("\n");
403		printf("The operating system has halted.\n");
404		printf("Please press any key to reboot.\n\n");
405		switch (cngetc()) {
406		case -1:		/* No console, just die */
407			cpu_halt();
408			/* NOTREACHED */
409		default:
410			howto &= ~RB_HALT;
411			break;
412		}
413	}
414}
415
416/*
417 * Check to see if the system paniced, pause and then reboot
418 * according to the specified delay.
419 */
420static void
421shutdown_panic(void *junk, int howto)
422{
423	int loop;
424
425	if (howto & RB_DUMP) {
426		if (PANIC_REBOOT_WAIT_TIME != 0) {
427			if (PANIC_REBOOT_WAIT_TIME != -1) {
428				printf("Automatic reboot in %d seconds - "
429				       "press a key on the console to abort\n",
430					PANIC_REBOOT_WAIT_TIME);
431				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
432				     loop > 0; --loop) {
433					DELAY(1000 * 100); /* 1/10th second */
434					/* Did user type a key? */
435					if (cncheckc() != -1)
436						break;
437				}
438				if (!loop)
439					return;
440			}
441		} else { /* zero time specified - reboot NOW */
442			return;
443		}
444		printf("--> Press a key on the console to reboot,\n");
445		printf("--> or switch off the system now.\n");
446		cngetc();
447	}
448}
449
450/*
451 * Everything done, now reset
452 */
453static void
454shutdown_reset(void *junk, int howto)
455{
456
457	printf("Rebooting...\n");
458	DELAY(1000000);	/* wait 1 sec for printf's to complete and be read */
459	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
460	cpu_reset();
461	/* NOTREACHED */ /* assuming reset worked */
462}
463
464#ifdef SMP
465static u_int panic_cpu = NOCPU;
466#endif
467
468/*
469 * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
470 * and then reboots.  If we are called twice, then we avoid trying to sync
471 * the disks as this often leads to recursive panics.
472 *
473 * MPSAFE
474 */
475void
476panic(const char *fmt, ...)
477{
478	struct thread *td = curthread;
479	int bootopt, newpanic;
480	va_list ap;
481	static char buf[256];
482
483#ifdef SMP
484	/*
485	 * We don't want multiple CPU's to panic at the same time, so we
486	 * use panic_cpu as a simple spinlock.  We have to keep checking
487	 * panic_cpu if we are spinning in case the panic on the first
488	 * CPU is canceled.
489	 */
490	if (panic_cpu != PCPU_GET(cpuid))
491		while (atomic_cmpset_int(&panic_cpu, NOCPU,
492		    PCPU_GET(cpuid)) == 0)
493			while (panic_cpu != NOCPU)
494				; /* nothing */
495#endif
496
497	bootopt = RB_AUTOBOOT | RB_DUMP;
498	newpanic = 0;
499	if (panicstr)
500		bootopt |= RB_NOSYNC;
501	else {
502		panicstr = fmt;
503		newpanic = 1;
504	}
505
506	va_start(ap, fmt);
507	if (newpanic) {
508		(void)vsnprintf(buf, sizeof(buf), fmt, ap);
509		panicstr = buf;
510		printf("panic: %s\n", buf);
511	} else {
512		printf("panic: ");
513		vprintf(fmt, ap);
514		printf("\n");
515	}
516	va_end(ap);
517#ifdef SMP
518	/* two separate prints in case of an unmapped page and trap */
519	printf("cpuid = %d; ", PCPU_GET(cpuid));
520#ifdef APIC_IO
521	printf("lapic.id = %08x\n", lapic.id);
522#else
523	printf("\n");
524#endif
525#endif
526
527#ifdef KDB
528	if (newpanic && trace_on_panic)
529		kdb_backtrace();
530	if (debugger_on_panic)
531		kdb_enter("panic");
532#ifdef RESTARTABLE_PANICS
533	/* See if the user aborted the panic, in which case we continue. */
534	if (panicstr == NULL) {
535#ifdef SMP
536		atomic_store_rel_int(&panic_cpu, NOCPU);
537#endif
538		return;
539	}
540#endif
541#endif
542	mtx_lock_spin(&sched_lock);
543	td->td_flags |= TDF_INPANIC;
544	mtx_unlock_spin(&sched_lock);
545	if (!sync_on_panic)
546		bootopt |= RB_NOSYNC;
547	boot(bootopt);
548}
549
550/*
551 * Support for poweroff delay.
552 */
553#ifndef POWEROFF_DELAY
554# define POWEROFF_DELAY 5000
555#endif
556static int poweroff_delay = POWEROFF_DELAY;
557
558SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
559	&poweroff_delay, 0, "");
560
561static void
562poweroff_wait(void *junk, int howto)
563{
564
565	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
566		return;
567	DELAY(poweroff_delay * 1000);
568}
569
570/*
571 * Some system processes (e.g. syncer) need to be stopped at appropriate
572 * points in their main loops prior to a system shutdown, so that they
573 * won't interfere with the shutdown process (e.g. by holding a disk buf
574 * to cause sync to fail).  For each of these system processes, register
575 * shutdown_kproc() as a handler for one of shutdown events.
576 */
577static int kproc_shutdown_wait = 60;
578SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
579    &kproc_shutdown_wait, 0, "");
580
581void
582kproc_shutdown(void *arg, int howto)
583{
584	struct proc *p;
585	char procname[MAXCOMLEN + 1];
586	int error;
587
588	if (panicstr)
589		return;
590
591	p = (struct proc *)arg;
592	strlcpy(procname, p->p_comm, sizeof(procname));
593	printf("Waiting (max %d seconds) for system process `%s' to stop...\n",
594	    kproc_shutdown_wait, procname);
595	error = kthread_suspend(p, kproc_shutdown_wait * hz);
596
597	if (error == EWOULDBLOCK)
598		printf("Stop of '%s' timed out\n", procname);
599	else
600		printf("Process '%s' stopped\n", procname);
601}
602
603/* Registration of dumpers */
604int
605set_dumper(struct dumperinfo *di)
606{
607
608	if (di == NULL) {
609		bzero(&dumper, sizeof dumper);
610		return (0);
611	}
612	if (dumper.dumper != NULL)
613		return (EBUSY);
614	dumper = *di;
615	return (0);
616}
617
618#if defined(__powerpc__)
619void
620dumpsys(struct dumperinfo *di __unused)
621{
622
623	printf("Kernel dumps not implemented on this architecture\n");
624}
625#endif
626