xref: /dflybsd-src/sys/kern/kern_shutdown.c (revision 744c01d0dc2aa1481a40e5b0988d15691602f5c9)
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 1.47 2007/01/05 22:16:28 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 long dumplo;
147 
148 static void boot (int) __dead2;
149 static void dumpsys (void);
150 static int setdumpdev (cdev_t dev);
151 static void poweroff_wait (void *, int);
152 static void print_uptime (void);
153 static void shutdown_halt (void *junk, int howto);
154 static void shutdown_panic (void *junk, int howto);
155 static void shutdown_reset (void *junk, int howto);
156 static int shutdown_busycount1(struct buf *bp, void *info);
157 static int shutdown_busycount2(struct buf *bp, void *info);
158 static void shutdown_cleanup_proc(struct proc *p);
159 
160 /* register various local shutdown events */
161 static void
162 shutdown_conf(void *unused)
163 {
164 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
165 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
166 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
167 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
168 }
169 
170 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
171 
172 /* ARGSUSED */
173 
174 /*
175  * The system call that results in a reboot
176  */
177 int
178 sys_reboot(struct reboot_args *uap)
179 {
180 	struct thread *td = curthread;
181 	int error;
182 
183 	if ((error = suser(td)))
184 		return (error);
185 
186 	boot(uap->opt);
187 	return (0);
188 }
189 
190 /*
191  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
192  */
193 static int shutdown_howto = 0;
194 
195 void
196 shutdown_nice(int howto)
197 {
198 	shutdown_howto = howto;
199 
200 	/* Send a signal to init(8) and have it shutdown the world */
201 	if (initproc != NULL) {
202 		ksignal(initproc, SIGINT);
203 	} else {
204 		/* No init(8) running, so simply reboot */
205 		boot(RB_NOSYNC);
206 	}
207 	return;
208 }
209 static int	waittime = -1;
210 static struct thread *dumpthread;
211 static struct pcb dumppcb;
212 
213 static void
214 print_uptime(void)
215 {
216 	int f;
217 	struct timespec ts;
218 
219 	getnanouptime(&ts);
220 	kprintf("Uptime: ");
221 	f = 0;
222 	if (ts.tv_sec >= 86400) {
223 		kprintf("%ldd", ts.tv_sec / 86400);
224 		ts.tv_sec %= 86400;
225 		f = 1;
226 	}
227 	if (f || ts.tv_sec >= 3600) {
228 		kprintf("%ldh", ts.tv_sec / 3600);
229 		ts.tv_sec %= 3600;
230 		f = 1;
231 	}
232 	if (f || ts.tv_sec >= 60) {
233 		kprintf("%ldm", ts.tv_sec / 60);
234 		ts.tv_sec %= 60;
235 		f = 1;
236 	}
237 	kprintf("%lds\n", ts.tv_sec);
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  */
245 static void
246 boot(int howto)
247 {
248 	/*
249 	 * Get rid of any user scheduler baggage and then give
250 	 * us a high priority.
251 	 */
252 	if (curthread->td_release)
253 		curthread->td_release(curthread);
254 	lwkt_setpri_self(TDPRI_MAX);
255 
256 	/* collect extra flags that shutdown_nice might have set */
257 	howto |= shutdown_howto;
258 
259 #ifdef SMP
260 	/*
261 	 * We really want to shutdown on the BSP.  Subsystems such as ACPI
262 	 * can't power-down the box otherwise.
263 	 */
264 	if (smp_active_mask > 1) {
265 		kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
266 	}
267 	if (panicstr == NULL && mycpu->gd_cpuid != 0) {
268 		kprintf("Switching to cpu #0 for shutdown\n");
269 		lwkt_setcpu_self(globaldata_find(0));
270 	}
271 #endif
272 	/*
273 	 * Do any callouts that should be done BEFORE syncing the filesystems.
274 	 */
275 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
276 
277 	/*
278 	 * Try to get rid of any remaining FS references.  The calling
279 	 * process, proc0, and init may still hold references.  The
280 	 * VFS cache subsystem may still hold a root reference to root.
281 	 */
282 	if (panicstr == NULL) {
283 		shutdown_cleanup_proc(curproc);
284 		shutdown_cleanup_proc(&proc0);
285 		if (initproc)
286 			shutdown_cleanup_proc(initproc);
287 		vfs_cache_setroot(NULL, NULL);
288 	}
289 
290 	/*
291 	 * Now sync filesystems
292 	 */
293 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
294 		int iter, nbusy, pbusy;
295 
296 		waittime = 0;
297 		kprintf("\nsyncing disks... ");
298 
299 		sys_sync(NULL);	/* YYY was sync(&proc0, NULL). why proc0 ? */
300 
301 		/*
302 		 * With soft updates, some buffers that are
303 		 * written will be remarked as dirty until other
304 		 * buffers are written.
305 		 */
306 		for (iter = pbusy = 0; iter < 20; iter++) {
307 			nbusy = scan_all_buffers(shutdown_busycount1, NULL);
308 			if (nbusy == 0)
309 				break;
310 			kprintf("%d ", nbusy);
311 			if (nbusy < pbusy)
312 				iter = 0;
313 			pbusy = nbusy;
314 			/*
315 			 * XXX:
316 			 * Process soft update work queue if buffers don't sync
317 			 * after 6 iterations by permitting the syncer to run.
318 			 */
319 			if (iter > 5 && bioops.io_sync)
320 				(*bioops.io_sync)(NULL);
321 			sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
322 			tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
323 		}
324 		kprintf("\n");
325 		/*
326 		 * Count only busy local buffers to prevent forcing
327 		 * a fsck if we're just a client of a wedged NFS server
328 		 */
329 		nbusy = scan_all_buffers(shutdown_busycount2, NULL);
330 		if (nbusy) {
331 			/*
332 			 * Failed to sync all blocks. Indicate this and don't
333 			 * unmount filesystems (thus forcing an fsck on reboot).
334 			 */
335 			kprintf("giving up on %d buffers\n", nbusy);
336 #ifdef DDB
337 			Debugger("busy buffer problem");
338 #endif /* DDB */
339 			tsleep(boot, 0, "shutdn", hz * 5 + 1);
340 		} else {
341 			kprintf("done\n");
342 			/*
343 			 * Unmount filesystems
344 			 */
345 			if (panicstr == NULL)
346 				vfs_unmountall();
347 		}
348 		tsleep(boot, 0, "shutdn", hz / 10 + 1);
349 	}
350 
351 	print_uptime();
352 
353 	/*
354 	 * Ok, now do things that assume all filesystem activity has
355 	 * been completed.
356 	 */
357 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
358 	crit_enter();
359 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
360 		dumpsys();
361 
362 	/* Now that we're going to really halt the system... */
363 	EVENTHANDLER_INVOKE(shutdown_final, howto);
364 
365 	for(;;) ;	/* safety against shutdown_reset not working */
366 	/* NOTREACHED */
367 }
368 
369 static int
370 shutdown_busycount1(struct buf *bp, void *info)
371 {
372 	if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
373 		return(1);
374 	if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
375 		return (1);
376 	return (0);
377 }
378 
379 static int
380 shutdown_busycount2(struct buf *bp, void *info)
381 {
382 	if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
383 	    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
384 		/*
385 		 * Only count buffers undergoing write I/O
386 		 * on the related vnode.
387 		 */
388 		if (bp->b_vp == NULL ||
389 		    bp->b_vp->v_track_write.bk_active == 0) {
390 			return (0);
391 		}
392 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
393 		kprintf(
394 	    "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
395 		    bp,
396 		    bp->b_flags, bp->b_loffset,
397 		    bp->b_bio2.bio_offset);
398 #endif
399 		return(1);
400 	}
401 	return(0);
402 }
403 
404 /*
405  * If the shutdown was a clean halt, behave accordingly.
406  */
407 static void
408 shutdown_halt(void *junk, int howto)
409 {
410 	if (howto & RB_HALT) {
411 		kprintf("\n");
412 		kprintf("The operating system has halted.\n");
413 		kprintf("Please press any key to reboot.\n\n");
414 		switch (cngetc()) {
415 		case -1:		/* No console, just die */
416 			cpu_halt();
417 			/* NOTREACHED */
418 		default:
419 			howto &= ~RB_HALT;
420 			break;
421 		}
422 	}
423 }
424 
425 /*
426  * Check to see if the system paniced, pause and then reboot
427  * according to the specified delay.
428  */
429 static void
430 shutdown_panic(void *junk, int howto)
431 {
432 	int loop;
433 
434 	if (howto & RB_DUMP) {
435 		if (PANIC_REBOOT_WAIT_TIME != 0) {
436 			if (PANIC_REBOOT_WAIT_TIME != -1) {
437 				kprintf("Automatic reboot in %d seconds - "
438 				       "press a key on the console to abort\n",
439 					PANIC_REBOOT_WAIT_TIME);
440 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
441 				     loop > 0; --loop) {
442 					DELAY(1000 * 100); /* 1/10th second */
443 					/* Did user type a key? */
444 					if (cncheckc() != -1)
445 						break;
446 				}
447 				if (!loop)
448 					return;
449 			}
450 		} else { /* zero time specified - reboot NOW */
451 			return;
452 		}
453 		kprintf("--> Press a key on the console to reboot,\n");
454 		kprintf("--> or switch off the system now.\n");
455 		cngetc();
456 	}
457 }
458 
459 /*
460  * Everything done, now reset
461  */
462 static void
463 shutdown_reset(void *junk, int howto)
464 {
465 	kprintf("Rebooting...\n");
466 	DELAY(1000000);	/* wait 1 sec for kprintf's to complete and be read */
467 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
468 	cpu_reset();
469 	/* NOTREACHED */ /* assuming reset worked */
470 }
471 
472 /*
473  * Try to remove FS references in the specified process.  This function
474  * is used during shutdown
475  */
476 static
477 void
478 shutdown_cleanup_proc(struct proc *p)
479 {
480 	struct filedesc *fdp;
481 	struct vmspace *vm;
482 
483 	if (p == NULL)
484 		return;
485 	if ((fdp = p->p_fd) != NULL) {
486 		kern_closefrom(0);
487 		if (fdp->fd_cdir) {
488 			cache_drop(&fdp->fd_ncdir);
489 			vrele(fdp->fd_cdir);
490 			fdp->fd_cdir = NULL;
491 		}
492 		if (fdp->fd_rdir) {
493 			cache_drop(&fdp->fd_nrdir);
494 			vrele(fdp->fd_rdir);
495 			fdp->fd_rdir = NULL;
496 		}
497 		if (fdp->fd_jdir) {
498 			cache_drop(&fdp->fd_njdir);
499 			vrele(fdp->fd_jdir);
500 			fdp->fd_jdir = NULL;
501 		}
502 	}
503 	if (p->p_vkernel)
504 		vkernel_exit(p);
505 	if (p->p_textvp) {
506 		vrele(p->p_textvp);
507 		p->p_textvp = NULL;
508 	}
509 	vm = p->p_vmspace;
510 	if (vm != NULL) {
511 		pmap_remove_pages(vmspace_pmap(vm),
512 				  VM_MIN_USER_ADDRESS,
513 				  VM_MAX_USER_ADDRESS);
514 		vm_map_remove(&vm->vm_map,
515 			      VM_MIN_USER_ADDRESS,
516 			      VM_MAX_USER_ADDRESS);
517 	}
518 }
519 
520 /*
521  * Magic number for savecore
522  *
523  * exported (symorder) and used at least by savecore(8)
524  *
525  */
526 static u_long const	dumpmag = 0x8fca0101UL;
527 
528 static int	dumpsize = 0;		/* also for savecore */
529 
530 static int	dodump = 1;
531 
532 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
533     "Try to perform coredump on kernel panic");
534 
535 static int
536 setdumpdev(cdev_t dev)
537 {
538 	int psize;
539 	long newdumplo;
540 
541 	if (dev == NOCDEV) {
542 		dumpdev = dev;
543 		return (0);
544 	}
545 	psize = dev_dpsize(dev);
546 	if (psize == -1)
547 		return (ENXIO);
548 	/*
549 	 * XXX should clean up checking in dumpsys() to be more like this.
550 	 */
551 	newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
552 	if (newdumplo <= LABELSECTOR)
553 		return (ENOSPC);
554 	dumpdev = dev;
555 	dumplo = newdumplo;
556 	return (0);
557 }
558 
559 
560 /* ARGSUSED */
561 static void dump_conf (void *dummy);
562 static void
563 dump_conf(void *dummy)
564 {
565 	char *path;
566 	cdev_t dev;
567 
568 	path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK);
569 	if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
570 		dev = kgetdiskbyname(path);
571 		if (dev != NOCDEV)
572 			dumpdev = dev;
573 	}
574 	kfree(path, M_TEMP);
575 	if (setdumpdev(dumpdev) != 0)
576 		dumpdev = NOCDEV;
577 }
578 
579 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
580 
581 static int
582 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
583 {
584 	int error;
585 	udev_t ndumpdev;
586 
587 	ndumpdev = dev2udev(dumpdev);
588 	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
589 	if (error == 0 && req->newptr != NULL)
590 		error = setdumpdev(udev2dev(ndumpdev, 0));
591 	return (error);
592 }
593 
594 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
595 	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", "");
596 
597 /*
598  * Doadump comes here after turning off memory management and
599  * getting on the dump stack, either when called above, or by
600  * the auto-restart code.
601  */
602 static void
603 dumpsys(void)
604 {
605 	int	error;
606 
607 	savectx(&dumppcb);
608 	dumpthread = curthread;
609 	if (dumping++) {
610 		kprintf("Dump already in progress, bailing...\n");
611 		return;
612 	}
613 	if (!dodump)
614 		return;
615 	if (dumpdev == NOCDEV)
616 		return;
617 	dumpsize = Maxmem;
618 	kprintf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
619 	kprintf("dump ");
620 	error = dev_ddump(dumpdev);
621 	if (error == 0) {
622 		kprintf("succeeded\n");
623 		return;
624 	}
625 	kprintf("failed, reason: ");
626 	switch (error) {
627 	case ENOSYS:
628 	case ENODEV:
629 		kprintf("device doesn't support a dump routine\n");
630 		break;
631 
632 	case ENXIO:
633 		kprintf("device bad\n");
634 		break;
635 
636 	case EFAULT:
637 		kprintf("device not ready\n");
638 		break;
639 
640 	case EINVAL:
641 		kprintf("area improper\n");
642 		break;
643 
644 	case EIO:
645 		kprintf("i/o error\n");
646 		break;
647 
648 	case EINTR:
649 		kprintf("aborted from console\n");
650 		break;
651 
652 	default:
653 		kprintf("unknown, error = %d\n", error);
654 		break;
655 	}
656 }
657 
658 int
659 dumpstatus(vm_offset_t addr, off_t count)
660 {
661 	int c;
662 
663 	if (addr % (1024 * 1024) == 0) {
664 #ifdef HW_WDOG
665 		if (wdog_tickler)
666 			(*wdog_tickler)();
667 #endif
668 		kprintf("%ld ", (long)(count / (1024 * 1024)));
669 	}
670 
671 	if ((c = cncheckc()) == 0x03)
672 		return -1;
673 	else if (c != -1)
674 		kprintf("[CTRL-C to abort] ");
675 
676 	return 0;
677 }
678 
679 /*
680  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
681  * and then reboots.  If we are called twice, then we avoid trying to sync
682  * the disks as this often leads to recursive panics.
683  */
684 void
685 panic(const char *fmt, ...)
686 {
687 	int bootopt, newpanic;
688 	__va_list ap;
689 	static char buf[256];
690 
691 #ifdef SMP
692 	/*
693 	 * If a panic occurs on multiple cpus before the first is able to
694 	 * halt the other cpus, only one cpu is allowed to take the panic.
695 	 * Attempt to be verbose about this situation but if the kprintf()
696 	 * itself panics don't let us overrun the kernel stack.
697 	 *
698 	 * Be very nasty about descheduling our thread at the lowest
699 	 * level possible in an attempt to freeze the thread without
700 	 * inducing further panics.
701 	 *
702 	 * Bumping gd_trap_nesting_level will also bypass assertions in
703 	 * lwkt_switch() and allow us to switch away even if we are a
704 	 * FAST interrupt or IPI.
705 	 */
706 	if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
707 		panic_cpu_gd = mycpu;
708 	} else if (panic_cpu_gd != mycpu) {
709 		crit_enter();
710 		++mycpu->gd_trap_nesting_level;
711 		if (mycpu->gd_trap_nesting_level < 25) {
712 			kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
713 				mycpu->gd_cpuid, curthread);
714 		}
715 		curthread->td_release = NULL;	/* be a grinch */
716 		for (;;) {
717 			lwkt_deschedule_self(curthread);
718 			lwkt_switch();
719 		}
720 		/* NOT REACHED */
721 		/* --mycpu->gd_trap_nesting_level */
722 		/* crit_exit() */
723 	}
724 #endif
725 	bootopt = RB_AUTOBOOT | RB_DUMP;
726 	if (sync_on_panic == 0)
727 		bootopt |= RB_NOSYNC;
728 	newpanic = 0;
729 	if (panicstr)
730 		bootopt |= RB_NOSYNC;
731 	else {
732 		panicstr = fmt;
733 		newpanic = 1;
734 	}
735 
736 	__va_start(ap, fmt);
737 	kvsnprintf(buf, sizeof(buf), fmt, ap);
738 	if (panicstr == fmt)
739 		panicstr = buf;
740 	__va_end(ap);
741 	kprintf("panic: %s\n", buf);
742 #ifdef SMP
743 	/* three separate prints in case of an unmapped page and trap */
744 	kprintf("mp_lock = %08x; ", mp_lock);
745 	kprintf("cpuid = %d; ", mycpu->gd_cpuid);
746 	kprintf("lapic.id = %08x\n", lapic.id);
747 #endif
748 
749 #if defined(DDB)
750 	if (newpanic && trace_on_panic)
751 		db_print_backtrace();
752 	if (debugger_on_panic)
753 		Debugger ("panic");
754 #endif
755 	boot(bootopt);
756 }
757 
758 /*
759  * Support for poweroff delay.
760  */
761 #ifndef POWEROFF_DELAY
762 # define POWEROFF_DELAY 5000
763 #endif
764 static int poweroff_delay = POWEROFF_DELAY;
765 
766 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
767 	&poweroff_delay, 0, "");
768 
769 static void
770 poweroff_wait(void *junk, int howto)
771 {
772 	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
773 		return;
774 	DELAY(poweroff_delay * 1000);
775 }
776 
777 /*
778  * Some system processes (e.g. syncer) need to be stopped at appropriate
779  * points in their main loops prior to a system shutdown, so that they
780  * won't interfere with the shutdown process (e.g. by holding a disk buf
781  * to cause sync to fail).  For each of these system processes, register
782  * shutdown_kproc() as a handler for one of shutdown events.
783  */
784 static int kproc_shutdown_wait = 60;
785 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
786     &kproc_shutdown_wait, 0, "");
787 
788 void
789 shutdown_kproc(void *arg, int howto)
790 {
791 	struct thread *td;
792 	struct proc *p;
793 	int error;
794 
795 	if (panicstr)
796 		return;
797 
798 	td = (struct thread *)arg;
799 	if ((p = td->td_proc) != NULL) {
800 	    kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
801 		kproc_shutdown_wait, p->p_comm);
802 	} else {
803 	    kprintf("Waiting (max %d seconds) for system thread %s to stop...",
804 		kproc_shutdown_wait, td->td_comm);
805 	}
806 	error = suspend_kproc(td, kproc_shutdown_wait * hz);
807 
808 	if (error == EWOULDBLOCK)
809 		kprintf("timed out\n");
810 	else
811 		kprintf("stopped\n");
812 }
813