xref: /dflybsd-src/sys/kern/kern_shutdown.c (revision 558a398b19696474173810daa238fb5df32df27b)
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: src/sys/kern/kern_shutdown.c,v 1.72.2.12 2002/02/21 19:15:10 dillon Exp $
40  * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.46 2006/12/28 18:29:03 dillon Exp $
41  */
42 
43 #include "opt_ddb.h"
44 #include "opt_ddb_trace.h"
45 #include "opt_hw_wdog.h"
46 #include "opt_panic.h"
47 #include "opt_show_busybufs.h"
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/eventhandler.h>
52 #include <sys/buf.h>
53 #include <sys/disklabel.h>
54 #include <sys/reboot.h>
55 #include <sys/proc.h>
56 #include <sys/vnode.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/malloc.h>
60 #include <sys/mount.h>
61 #include <sys/queue.h>
62 #include <sys/sysctl.h>
63 #include <sys/vkernel.h>
64 #include <sys/conf.h>
65 #include <sys/sysproto.h>
66 #include <sys/device.h>
67 #include <sys/cons.h>
68 #include <sys/shm.h>
69 #include <sys/kern_syscall.h>
70 #include <vm/vm_map.h>
71 #include <vm/pmap.h>
72 
73 #include <sys/thread2.h>
74 #include <sys/buf2.h>
75 
76 #include <machine/pcb.h>
77 #include <machine/clock.h>
78 #include <machine/md_var.h>
79 #include <machine/smp.h>		/* smp_active_mask, cpuid */
80 #include <machine/vmparam.h>
81 
82 #include <sys/signalvar.h>
83 
84 #ifndef PANIC_REBOOT_WAIT_TIME
85 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
86 #endif
87 
88 /*
89  * Note that stdarg.h and the ANSI style va_start macro is used for both
90  * ANSI and traditional C compilers.  We use the machine version to stay
91  * within the confines of the kernel header files.
92  */
93 #include <machine/stdarg.h>
94 
95 #ifdef DDB
96 #ifdef DDB_UNATTENDED
97 int debugger_on_panic = 0;
98 #else
99 int debugger_on_panic = 1;
100 #endif
101 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
102 	&debugger_on_panic, 0, "Run debugger on kernel panic");
103 
104 extern void db_print_backtrace(void);
105 
106 #ifdef DDB_TRACE
107 int trace_on_panic = 1;
108 #else
109 int trace_on_panic = 0;
110 #endif
111 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
112 	&trace_on_panic, 0, "Print stack trace on kernel panic");
113 #endif
114 
115 static int sync_on_panic = 1;
116 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
117 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
118 
119 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
120 
121 #ifdef	HW_WDOG
122 /*
123  * If there is a hardware watchdog, point this at the function needed to
124  * hold it off.
125  * It's needed when the kernel needs to do some lengthy operations.
126  * e.g. in wd.c when dumping core.. It's most annoying to have
127  * your precious core-dump only half written because the wdog kicked in.
128  */
129 watchdog_tickle_fn wdog_tickler = NULL;
130 #endif	/* HW_WDOG */
131 
132 /*
133  * Variable panicstr contains argument to first call to panic; used as flag
134  * to indicate that the kernel has already called panic.
135  */
136 const char *panicstr;
137 
138 int dumping;				/* system is dumping */
139 #ifdef SMP
140 u_int panic_cpu_interlock;		/* panic interlock */
141 globaldata_t panic_cpu_gd;		/* which cpu took the panic */
142 #endif
143 
144 int bootverbose = 0;			/* note: assignment to force non-bss */
145 int cold = 1;				/* note: assignment to force non-bss */
146 
147 static void boot (int) __dead2;
148 static void dumpsys (void);
149 static int setdumpdev (cdev_t dev);
150 static void poweroff_wait (void *, int);
151 static void print_uptime (void);
152 static void shutdown_halt (void *junk, int howto);
153 static void shutdown_panic (void *junk, int howto);
154 static void shutdown_reset (void *junk, int howto);
155 static int shutdown_busycount1(struct buf *bp, void *info);
156 static int shutdown_busycount2(struct buf *bp, void *info);
157 static void shutdown_cleanup_proc(struct proc *p);
158 
159 /* register various local shutdown events */
160 static void
161 shutdown_conf(void *unused)
162 {
163 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
164 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
165 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
166 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
167 }
168 
169 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
170 
171 /* ARGSUSED */
172 
173 /*
174  * The system call that results in a reboot
175  */
176 int
177 sys_reboot(struct reboot_args *uap)
178 {
179 	struct thread *td = curthread;
180 	int error;
181 
182 	if ((error = suser(td)))
183 		return (error);
184 
185 	boot(uap->opt);
186 	return (0);
187 }
188 
189 /*
190  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
191  */
192 static int shutdown_howto = 0;
193 
194 void
195 shutdown_nice(int howto)
196 {
197 	shutdown_howto = howto;
198 
199 	/* Send a signal to init(8) and have it shutdown the world */
200 	if (initproc != NULL) {
201 		ksignal(initproc, SIGINT);
202 	} else {
203 		/* No init(8) running, so simply reboot */
204 		boot(RB_NOSYNC);
205 	}
206 	return;
207 }
208 static int	waittime = -1;
209 static struct thread *dumpthread;
210 static struct pcb dumppcb;
211 
212 static void
213 print_uptime(void)
214 {
215 	int f;
216 	struct timespec ts;
217 
218 	getnanouptime(&ts);
219 	kprintf("Uptime: ");
220 	f = 0;
221 	if (ts.tv_sec >= 86400) {
222 		kprintf("%ldd", ts.tv_sec / 86400);
223 		ts.tv_sec %= 86400;
224 		f = 1;
225 	}
226 	if (f || ts.tv_sec >= 3600) {
227 		kprintf("%ldh", ts.tv_sec / 3600);
228 		ts.tv_sec %= 3600;
229 		f = 1;
230 	}
231 	if (f || ts.tv_sec >= 60) {
232 		kprintf("%ldm", ts.tv_sec / 60);
233 		ts.tv_sec %= 60;
234 		f = 1;
235 	}
236 	kprintf("%lds\n", ts.tv_sec);
237 }
238 
239 /*
240  *  Go through the rigmarole of shutting down..
241  * this used to be in machdep.c but I'll be dammned if I could see
242  * anything machine dependant in it.
243  */
244 static void
245 boot(int howto)
246 {
247 	/*
248 	 * Get rid of any user scheduler baggage and then give
249 	 * us a high priority.
250 	 */
251 	if (curthread->td_release)
252 		curthread->td_release(curthread);
253 	lwkt_setpri_self(TDPRI_MAX);
254 
255 	/* collect extra flags that shutdown_nice might have set */
256 	howto |= shutdown_howto;
257 
258 #ifdef SMP
259 	/*
260 	 * We really want to shutdown on the BSP.  Subsystems such as ACPI
261 	 * can't power-down the box otherwise.
262 	 */
263 	if (smp_active_mask > 1) {
264 		kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
265 	}
266 	if (panicstr == NULL && mycpu->gd_cpuid != 0) {
267 		kprintf("Switching to cpu #0 for shutdown\n");
268 		lwkt_setcpu_self(globaldata_find(0));
269 	}
270 #endif
271 	/*
272 	 * Do any callouts that should be done BEFORE syncing the filesystems.
273 	 */
274 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
275 
276 	/*
277 	 * Try to get rid of any remaining FS references.  The calling
278 	 * process, proc0, and init may still hold references.  The
279 	 * VFS cache subsystem may still hold a root reference to root.
280 	 */
281 	if (panicstr == NULL) {
282 		shutdown_cleanup_proc(curproc);
283 		shutdown_cleanup_proc(&proc0);
284 		if (initproc)
285 			shutdown_cleanup_proc(initproc);
286 		vfs_cache_setroot(NULL, NULL);
287 	}
288 
289 	/*
290 	 * Now sync filesystems
291 	 */
292 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
293 		int iter, nbusy, pbusy;
294 
295 		waittime = 0;
296 		kprintf("\nsyncing disks... ");
297 
298 		sys_sync(NULL);	/* YYY was sync(&proc0, NULL). why proc0 ? */
299 
300 		/*
301 		 * With soft updates, some buffers that are
302 		 * written will be remarked as dirty until other
303 		 * buffers are written.
304 		 */
305 		for (iter = pbusy = 0; iter < 20; iter++) {
306 			nbusy = scan_all_buffers(shutdown_busycount1, NULL);
307 			if (nbusy == 0)
308 				break;
309 			kprintf("%d ", nbusy);
310 			if (nbusy < pbusy)
311 				iter = 0;
312 			pbusy = nbusy;
313 			/*
314 			 * XXX:
315 			 * Process soft update work queue if buffers don't sync
316 			 * after 6 iterations by permitting the syncer to run.
317 			 */
318 			if (iter > 5 && bioops.io_sync)
319 				(*bioops.io_sync)(NULL);
320 			sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
321 			tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
322 		}
323 		kprintf("\n");
324 		/*
325 		 * Count only busy local buffers to prevent forcing
326 		 * a fsck if we're just a client of a wedged NFS server
327 		 */
328 		nbusy = scan_all_buffers(shutdown_busycount2, NULL);
329 		if (nbusy) {
330 			/*
331 			 * Failed to sync all blocks. Indicate this and don't
332 			 * unmount filesystems (thus forcing an fsck on reboot).
333 			 */
334 			kprintf("giving up on %d buffers\n", nbusy);
335 #ifdef DDB
336 			Debugger("busy buffer problem");
337 #endif /* DDB */
338 			tsleep(boot, 0, "shutdn", hz * 5 + 1);
339 		} else {
340 			kprintf("done\n");
341 			/*
342 			 * Unmount filesystems
343 			 */
344 			if (panicstr == NULL)
345 				vfs_unmountall();
346 		}
347 		tsleep(boot, 0, "shutdn", hz / 10 + 1);
348 	}
349 
350 	print_uptime();
351 
352 	/*
353 	 * Ok, now do things that assume all filesystem activity has
354 	 * been completed.
355 	 */
356 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
357 	crit_enter();
358 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
359 		dumpsys();
360 
361 	/* Now that we're going to really halt the system... */
362 	EVENTHANDLER_INVOKE(shutdown_final, howto);
363 
364 	for(;;) ;	/* safety against shutdown_reset not working */
365 	/* NOTREACHED */
366 }
367 
368 static int
369 shutdown_busycount1(struct buf *bp, void *info)
370 {
371 	if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
372 		return(1);
373 	if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
374 		return (1);
375 	return (0);
376 }
377 
378 static int
379 shutdown_busycount2(struct buf *bp, void *info)
380 {
381 	if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
382 	    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
383 		/*
384 		 * Only count buffers undergoing write I/O
385 		 * on the related vnode.
386 		 */
387 		if (bp->b_vp == NULL ||
388 		    bp->b_vp->v_track_write.bk_active == 0) {
389 			return (0);
390 		}
391 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
392 		kprintf(
393 	    "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
394 		    bp,
395 		    bp->b_flags, bp->b_loffset,
396 		    bp->b_bio2.bio_offset);
397 #endif
398 		return(1);
399 	}
400 	return(0);
401 }
402 
403 /*
404  * If the shutdown was a clean halt, behave accordingly.
405  */
406 static void
407 shutdown_halt(void *junk, int howto)
408 {
409 	if (howto & RB_HALT) {
410 		kprintf("\n");
411 		kprintf("The operating system has halted.\n");
412 		kprintf("Please press any key to reboot.\n\n");
413 		switch (cngetc()) {
414 		case -1:		/* No console, just die */
415 			cpu_halt();
416 			/* NOTREACHED */
417 		default:
418 			howto &= ~RB_HALT;
419 			break;
420 		}
421 	}
422 }
423 
424 /*
425  * Check to see if the system paniced, pause and then reboot
426  * according to the specified delay.
427  */
428 static void
429 shutdown_panic(void *junk, int howto)
430 {
431 	int loop;
432 
433 	if (howto & RB_DUMP) {
434 		if (PANIC_REBOOT_WAIT_TIME != 0) {
435 			if (PANIC_REBOOT_WAIT_TIME != -1) {
436 				kprintf("Automatic reboot in %d seconds - "
437 				       "press a key on the console to abort\n",
438 					PANIC_REBOOT_WAIT_TIME);
439 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
440 				     loop > 0; --loop) {
441 					DELAY(1000 * 100); /* 1/10th second */
442 					/* Did user type a key? */
443 					if (cncheckc() != -1)
444 						break;
445 				}
446 				if (!loop)
447 					return;
448 			}
449 		} else { /* zero time specified - reboot NOW */
450 			return;
451 		}
452 		kprintf("--> Press a key on the console to reboot,\n");
453 		kprintf("--> or switch off the system now.\n");
454 		cngetc();
455 	}
456 }
457 
458 /*
459  * Everything done, now reset
460  */
461 static void
462 shutdown_reset(void *junk, int howto)
463 {
464 	kprintf("Rebooting...\n");
465 	DELAY(1000000);	/* wait 1 sec for kprintf'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 /*
472  * Try to remove FS references in the specified process.  This function
473  * is used during shutdown
474  */
475 static
476 void
477 shutdown_cleanup_proc(struct proc *p)
478 {
479 	struct filedesc *fdp;
480 	struct vmspace *vm;
481 
482 	if (p == NULL)
483 		return;
484 	if ((fdp = p->p_fd) != NULL) {
485 		kern_closefrom(0);
486 		if (fdp->fd_cdir) {
487 			cache_drop(&fdp->fd_ncdir);
488 			vrele(fdp->fd_cdir);
489 			fdp->fd_cdir = NULL;
490 		}
491 		if (fdp->fd_rdir) {
492 			cache_drop(&fdp->fd_nrdir);
493 			vrele(fdp->fd_rdir);
494 			fdp->fd_rdir = NULL;
495 		}
496 		if (fdp->fd_jdir) {
497 			cache_drop(&fdp->fd_njdir);
498 			vrele(fdp->fd_jdir);
499 			fdp->fd_jdir = NULL;
500 		}
501 	}
502 	if (p->p_vkernel)
503 		vkernel_exit(p);
504 	if (p->p_textvp) {
505 		vrele(p->p_textvp);
506 		p->p_textvp = NULL;
507 	}
508 	vm = p->p_vmspace;
509 	if (vm != NULL) {
510 		pmap_remove_pages(vmspace_pmap(vm),
511 				  VM_MIN_USER_ADDRESS,
512 				  VM_MAX_USER_ADDRESS);
513 		vm_map_remove(&vm->vm_map,
514 			      VM_MIN_USER_ADDRESS,
515 			      VM_MAX_USER_ADDRESS);
516 	}
517 }
518 
519 /*
520  * Magic number for savecore
521  *
522  * exported (symorder) and used at least by savecore(8)
523  *
524  */
525 static u_long const	dumpmag = 0x8fca0101UL;
526 
527 static int	dumpsize = 0;		/* also for savecore */
528 
529 static int	dodump = 1;
530 
531 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
532     "Try to perform coredump on kernel panic");
533 
534 static int
535 setdumpdev(cdev_t dev)
536 {
537 	int psize;
538 	long newdumplo;
539 
540 	if (dev == NOCDEV) {
541 		dumpdev = dev;
542 		return (0);
543 	}
544 	psize = dev_dpsize(dev);
545 	if (psize == -1)
546 		return (ENXIO);
547 	/*
548 	 * XXX should clean up checking in dumpsys() to be more like this.
549 	 */
550 	newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
551 	if (newdumplo <= LABELSECTOR)
552 		return (ENOSPC);
553 	dumpdev = dev;
554 	dumplo = newdumplo;
555 	return (0);
556 }
557 
558 
559 /* ARGSUSED */
560 static void dump_conf (void *dummy);
561 static void
562 dump_conf(void *dummy)
563 {
564 	char *path;
565 	cdev_t dev;
566 
567 	path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK);
568 	if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
569 		dev = kgetdiskbyname(path);
570 		if (dev != NOCDEV)
571 			dumpdev = dev;
572 	}
573 	kfree(path, M_TEMP);
574 	if (setdumpdev(dumpdev) != 0)
575 		dumpdev = NOCDEV;
576 }
577 
578 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
579 
580 static int
581 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
582 {
583 	int error;
584 	udev_t ndumpdev;
585 
586 	ndumpdev = dev2udev(dumpdev);
587 	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
588 	if (error == 0 && req->newptr != NULL)
589 		error = setdumpdev(udev2dev(ndumpdev, 0));
590 	return (error);
591 }
592 
593 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
594 	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", "");
595 
596 /*
597  * Doadump comes here after turning off memory management and
598  * getting on the dump stack, either when called above, or by
599  * the auto-restart code.
600  */
601 static void
602 dumpsys(void)
603 {
604 	int	error;
605 
606 	savectx(&dumppcb);
607 	dumpthread = curthread;
608 	if (dumping++) {
609 		kprintf("Dump already in progress, bailing...\n");
610 		return;
611 	}
612 	if (!dodump)
613 		return;
614 	if (dumpdev == NOCDEV)
615 		return;
616 	dumpsize = Maxmem;
617 	kprintf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
618 	kprintf("dump ");
619 	error = dev_ddump(dumpdev);
620 	if (error == 0) {
621 		kprintf("succeeded\n");
622 		return;
623 	}
624 	kprintf("failed, reason: ");
625 	switch (error) {
626 	case ENOSYS:
627 	case ENODEV:
628 		kprintf("device doesn't support a dump routine\n");
629 		break;
630 
631 	case ENXIO:
632 		kprintf("device bad\n");
633 		break;
634 
635 	case EFAULT:
636 		kprintf("device not ready\n");
637 		break;
638 
639 	case EINVAL:
640 		kprintf("area improper\n");
641 		break;
642 
643 	case EIO:
644 		kprintf("i/o error\n");
645 		break;
646 
647 	case EINTR:
648 		kprintf("aborted from console\n");
649 		break;
650 
651 	default:
652 		kprintf("unknown, error = %d\n", error);
653 		break;
654 	}
655 }
656 
657 int
658 dumpstatus(vm_offset_t addr, off_t count)
659 {
660 	int c;
661 
662 	if (addr % (1024 * 1024) == 0) {
663 #ifdef HW_WDOG
664 		if (wdog_tickler)
665 			(*wdog_tickler)();
666 #endif
667 		kprintf("%ld ", (long)(count / (1024 * 1024)));
668 	}
669 
670 	if ((c = cncheckc()) == 0x03)
671 		return -1;
672 	else if (c != -1)
673 		kprintf("[CTRL-C to abort] ");
674 
675 	return 0;
676 }
677 
678 /*
679  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
680  * and then reboots.  If we are called twice, then we avoid trying to sync
681  * the disks as this often leads to recursive panics.
682  */
683 void
684 panic(const char *fmt, ...)
685 {
686 	int bootopt, newpanic;
687 	__va_list ap;
688 	static char buf[256];
689 
690 #ifdef SMP
691 	/*
692 	 * If a panic occurs on multiple cpus before the first is able to
693 	 * halt the other cpus, only one cpu is allowed to take the panic.
694 	 * Attempt to be verbose about this situation but if the kprintf()
695 	 * itself panics don't let us overrun the kernel stack.
696 	 *
697 	 * Be very nasty about descheduling our thread at the lowest
698 	 * level possible in an attempt to freeze the thread without
699 	 * inducing further panics.
700 	 *
701 	 * Bumping gd_trap_nesting_level will also bypass assertions in
702 	 * lwkt_switch() and allow us to switch away even if we are a
703 	 * FAST interrupt or IPI.
704 	 */
705 	if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
706 		panic_cpu_gd = mycpu;
707 	} else if (panic_cpu_gd != mycpu) {
708 		crit_enter();
709 		++mycpu->gd_trap_nesting_level;
710 		if (mycpu->gd_trap_nesting_level < 25) {
711 			kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
712 				mycpu->gd_cpuid, curthread);
713 		}
714 		curthread->td_release = NULL;	/* be a grinch */
715 		for (;;) {
716 			lwkt_deschedule_self(curthread);
717 			lwkt_switch();
718 		}
719 		/* NOT REACHED */
720 		/* --mycpu->gd_trap_nesting_level */
721 		/* crit_exit() */
722 	}
723 #endif
724 	bootopt = RB_AUTOBOOT | RB_DUMP;
725 	if (sync_on_panic == 0)
726 		bootopt |= RB_NOSYNC;
727 	newpanic = 0;
728 	if (panicstr)
729 		bootopt |= RB_NOSYNC;
730 	else {
731 		panicstr = fmt;
732 		newpanic = 1;
733 	}
734 
735 	__va_start(ap, fmt);
736 	kvsnprintf(buf, sizeof(buf), fmt, ap);
737 	if (panicstr == fmt)
738 		panicstr = buf;
739 	__va_end(ap);
740 	kprintf("panic: %s\n", buf);
741 #ifdef SMP
742 	/* three separate prints in case of an unmapped page and trap */
743 	kprintf("mp_lock = %08x; ", mp_lock);
744 	kprintf("cpuid = %d; ", mycpu->gd_cpuid);
745 	kprintf("lapic.id = %08x\n", lapic.id);
746 #endif
747 
748 #if defined(DDB)
749 	if (newpanic && trace_on_panic)
750 		db_print_backtrace();
751 	if (debugger_on_panic)
752 		Debugger ("panic");
753 #endif
754 	boot(bootopt);
755 }
756 
757 /*
758  * Support for poweroff delay.
759  */
760 #ifndef POWEROFF_DELAY
761 # define POWEROFF_DELAY 5000
762 #endif
763 static int poweroff_delay = POWEROFF_DELAY;
764 
765 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
766 	&poweroff_delay, 0, "");
767 
768 static void
769 poweroff_wait(void *junk, int howto)
770 {
771 	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
772 		return;
773 	DELAY(poweroff_delay * 1000);
774 }
775 
776 /*
777  * Some system processes (e.g. syncer) need to be stopped at appropriate
778  * points in their main loops prior to a system shutdown, so that they
779  * won't interfere with the shutdown process (e.g. by holding a disk buf
780  * to cause sync to fail).  For each of these system processes, register
781  * shutdown_kproc() as a handler for one of shutdown events.
782  */
783 static int kproc_shutdown_wait = 60;
784 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
785     &kproc_shutdown_wait, 0, "");
786 
787 void
788 shutdown_kproc(void *arg, int howto)
789 {
790 	struct thread *td;
791 	struct proc *p;
792 	int error;
793 
794 	if (panicstr)
795 		return;
796 
797 	td = (struct thread *)arg;
798 	if ((p = td->td_proc) != NULL) {
799 	    kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
800 		kproc_shutdown_wait, p->p_comm);
801 	} else {
802 	    kprintf("Waiting (max %d seconds) for system thread %s to stop...",
803 		kproc_shutdown_wait, td->td_comm);
804 	}
805 	error = suspend_kproc(td, kproc_shutdown_wait * hz);
806 
807 	if (error == EWOULDBLOCK)
808 		kprintf("timed out\n");
809 	else
810 		kprintf("stopped\n");
811 }
812