xref: /dragonfly/sys/kern/kern_shutdown.c (revision ad7a2457)
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 1.62 2008/01/05 13:23:48 corecode 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/diskslice.h>
54 #include <sys/reboot.h>
55 #include <sys/proc.h>
56 #include <sys/priv.h>
57 #include <sys/fcntl.h>		/* FREAD	*/
58 #include <sys/stat.h>		/* S_IFCHR	*/
59 #include <sys/vnode.h>
60 #include <sys/kernel.h>
61 #include <sys/kthread.h>
62 #include <sys/malloc.h>
63 #include <sys/mount.h>
64 #include <sys/queue.h>
65 #include <sys/sysctl.h>
66 #include <sys/vkernel.h>
67 #include <sys/conf.h>
68 #include <sys/sysproto.h>
69 #include <sys/device.h>
70 #include <sys/cons.h>
71 #include <sys/shm.h>
72 #include <sys/kern_syscall.h>
73 #include <vm/vm_map.h>
74 #include <vm/pmap.h>
75 
76 #include <sys/thread2.h>
77 #include <sys/buf2.h>
78 
79 #include <machine/pcb.h>
80 #include <machine/clock.h>
81 #include <machine/md_var.h>
82 #include <machine/smp.h>		/* smp_active_mask, cpuid */
83 #include <machine/vmparam.h>
84 
85 #include <sys/signalvar.h>
86 
87 #ifndef PANIC_REBOOT_WAIT_TIME
88 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
89 #endif
90 
91 /*
92  * Note that stdarg.h and the ANSI style va_start macro is used for both
93  * ANSI and traditional C compilers.  We use the machine version to stay
94  * within the confines of the kernel header files.
95  */
96 #include <machine/stdarg.h>
97 
98 #ifdef DDB
99 #ifdef DDB_UNATTENDED
100 int debugger_on_panic = 0;
101 #else
102 int debugger_on_panic = 1;
103 #endif
104 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
105 	&debugger_on_panic, 0, "Run debugger on kernel panic");
106 
107 extern void db_print_backtrace(void);
108 
109 #ifdef DDB_TRACE
110 int trace_on_panic = 1;
111 #else
112 int trace_on_panic = 0;
113 #endif
114 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
115 	&trace_on_panic, 0, "Print stack trace on kernel panic");
116 #endif
117 
118 static int sync_on_panic = 1;
119 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
120 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
121 
122 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
123 
124 #ifdef	HW_WDOG
125 /*
126  * If there is a hardware watchdog, point this at the function needed to
127  * hold it off.
128  * It's needed when the kernel needs to do some lengthy operations.
129  * e.g. in wd.c when dumping core.. It's most annoying to have
130  * your precious core-dump only half written because the wdog kicked in.
131  */
132 watchdog_tickle_fn wdog_tickler = NULL;
133 #endif	/* HW_WDOG */
134 
135 /*
136  * Variable panicstr contains argument to first call to panic; used as flag
137  * to indicate that the kernel has already called panic.
138  */
139 const char *panicstr;
140 
141 int dumping;				/* system is dumping */
142 #ifdef SMP
143 u_int panic_cpu_interlock;		/* panic interlock */
144 globaldata_t panic_cpu_gd;		/* which cpu took the panic */
145 #endif
146 
147 int bootverbose = 0;			/* note: assignment to force non-bss */
148 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW,
149 	   &bootverbose, 0, "Verbose kernel messages");
150 
151 int cold = 1;				/* note: assignment to force non-bss */
152 int dumplo;				/* OBSOLETE - savecore compat */
153 u_int64_t dumplo64;
154 
155 static void boot (int) __dead2;
156 static void dumpsys (void);
157 static int setdumpdev (cdev_t dev);
158 static void poweroff_wait (void *, int);
159 static void print_uptime (void);
160 static void shutdown_halt (void *junk, int howto);
161 static void shutdown_panic (void *junk, int howto);
162 static void shutdown_reset (void *junk, int howto);
163 static int shutdown_busycount1(struct buf *bp, void *info);
164 static int shutdown_busycount2(struct buf *bp, void *info);
165 static void shutdown_cleanup_proc(struct proc *p);
166 
167 /* register various local shutdown events */
168 static void
169 shutdown_conf(void *unused)
170 {
171 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
172 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
173 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
174 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
175 }
176 
177 SYSINIT(shutdown_conf, SI_BOOT2_MACHDEP, SI_ORDER_ANY, shutdown_conf, NULL)
178 
179 /* ARGSUSED */
180 
181 /*
182  * The system call that results in a reboot
183  */
184 int
185 sys_reboot(struct reboot_args *uap)
186 {
187 	struct thread *td = curthread;
188 	int error;
189 
190 	if ((error = priv_check(td, PRIV_ROOT)))
191 		return (error);
192 
193 	boot(uap->opt);
194 	return (0);
195 }
196 
197 /*
198  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
199  */
200 static int shutdown_howto = 0;
201 
202 void
203 shutdown_nice(int howto)
204 {
205 	shutdown_howto = howto;
206 
207 	/* Send a signal to init(8) and have it shutdown the world */
208 	if (initproc != NULL) {
209 		ksignal(initproc, SIGINT);
210 	} else {
211 		/* No init(8) running, so simply reboot */
212 		boot(RB_NOSYNC);
213 	}
214 	return;
215 }
216 static int	waittime = -1;
217 static struct thread *dumpthread;
218 static struct pcb dumppcb;
219 
220 static void
221 print_uptime(void)
222 {
223 	int f;
224 	struct timespec ts;
225 
226 	getnanouptime(&ts);
227 	kprintf("Uptime: ");
228 	f = 0;
229 	if (ts.tv_sec >= 86400) {
230 		kprintf("%ldd", ts.tv_sec / 86400);
231 		ts.tv_sec %= 86400;
232 		f = 1;
233 	}
234 	if (f || ts.tv_sec >= 3600) {
235 		kprintf("%ldh", ts.tv_sec / 3600);
236 		ts.tv_sec %= 3600;
237 		f = 1;
238 	}
239 	if (f || ts.tv_sec >= 60) {
240 		kprintf("%ldm", ts.tv_sec / 60);
241 		ts.tv_sec %= 60;
242 		f = 1;
243 	}
244 	kprintf("%lds\n", ts.tv_sec);
245 }
246 
247 /*
248  *  Go through the rigmarole of shutting down..
249  * this used to be in machdep.c but I'll be dammned if I could see
250  * anything machine dependant in it.
251  */
252 static void
253 boot(int howto)
254 {
255 	/*
256 	 * Get rid of any user scheduler baggage and then give
257 	 * us a high priority.
258 	 */
259 	if (curthread->td_release)
260 		curthread->td_release(curthread);
261 	lwkt_setpri_self(TDPRI_MAX);
262 
263 	/* collect extra flags that shutdown_nice might have set */
264 	howto |= shutdown_howto;
265 
266 #ifdef SMP
267 	/*
268 	 * We really want to shutdown on the BSP.  Subsystems such as ACPI
269 	 * can't power-down the box otherwise.
270 	 */
271 	if (smp_active_mask > 1) {
272 		kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
273 	}
274 	if (panicstr == NULL && mycpu->gd_cpuid != 0) {
275 		kprintf("Switching to cpu #0 for shutdown\n");
276 		lwkt_setcpu_self(globaldata_find(0));
277 	}
278 #endif
279 	/*
280 	 * Do any callouts that should be done BEFORE syncing the filesystems.
281 	 */
282 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
283 
284 	/*
285 	 * Try to get rid of any remaining FS references.  The calling
286 	 * process, proc0, and init may still hold references.  The
287 	 * VFS cache subsystem may still hold a root reference to root.
288 	 *
289 	 * XXX this needs work.  We really need to SIGSTOP all remaining
290 	 * processes in order to avoid blowups due to proc0's filesystem
291 	 * references going away.  For now just make sure that the init
292 	 * process is stopped.
293 	 */
294 	if (panicstr == NULL) {
295 		shutdown_cleanup_proc(curproc);
296 		shutdown_cleanup_proc(&proc0);
297 		if (initproc) {
298 			if (initproc != curproc) {
299 				ksignal(initproc, SIGSTOP);
300 				tsleep(boot, 0, "shutdn", hz / 20);
301 			}
302 			shutdown_cleanup_proc(initproc);
303 		}
304 		vfs_cache_setroot(NULL, NULL);
305 	}
306 
307 	/*
308 	 * Now sync filesystems
309 	 */
310 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
311 		int iter, nbusy, pbusy;
312 
313 		waittime = 0;
314 		kprintf("\nsyncing disks... ");
315 
316 		sys_sync(NULL);	/* YYY was sync(&proc0, NULL). why proc0 ? */
317 
318 		/*
319 		 * With soft updates, some buffers that are
320 		 * written will be remarked as dirty until other
321 		 * buffers are written.
322 		 */
323 		for (iter = pbusy = 0; iter < 20; iter++) {
324 			nbusy = scan_all_buffers(shutdown_busycount1, NULL);
325 			if (nbusy == 0)
326 				break;
327 			kprintf("%d ", nbusy);
328 			if (nbusy < pbusy)
329 				iter = 0;
330 			pbusy = nbusy;
331 			/*
332 			 * XXX:
333 			 * Process soft update work queue if buffers don't sync
334 			 * after 6 iterations by permitting the syncer to run.
335 			 */
336 			if (iter > 5)
337 				bio_ops_sync(NULL);
338 
339 			sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
340 			tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
341 		}
342 		kprintf("\n");
343 		/*
344 		 * Count only busy local buffers to prevent forcing
345 		 * a fsck if we're just a client of a wedged NFS server
346 		 */
347 		nbusy = scan_all_buffers(shutdown_busycount2, NULL);
348 		if (nbusy) {
349 			/*
350 			 * Failed to sync all blocks. Indicate this and don't
351 			 * unmount filesystems (thus forcing an fsck on reboot).
352 			 */
353 			kprintf("giving up on %d buffers\n", nbusy);
354 #ifdef DDB
355 			Debugger("busy buffer problem");
356 #endif /* DDB */
357 			tsleep(boot, 0, "shutdn", hz * 5 + 1);
358 		} else {
359 			kprintf("done\n");
360 			/*
361 			 * Unmount filesystems
362 			 */
363 			if (panicstr == NULL)
364 				vfs_unmountall();
365 		}
366 		tsleep(boot, 0, "shutdn", hz / 10 + 1);
367 	}
368 
369 	print_uptime();
370 
371 	/*
372 	 * Dump before doing post_sync shutdown ops
373 	 */
374 	crit_enter();
375 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
376 		dumpsys();
377 
378 	/*
379 	 * Ok, now do things that assume all filesystem activity has
380 	 * been completed.  This will also call the device shutdown
381 	 * methods.
382 	 */
383 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
384 
385 	/* Now that we're going to really halt the system... */
386 	EVENTHANDLER_INVOKE(shutdown_final, howto);
387 
388 	for(;;) ;	/* safety against shutdown_reset not working */
389 	/* NOTREACHED */
390 }
391 
392 static int
393 shutdown_busycount1(struct buf *bp, void *info)
394 {
395 	if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
396 		return(1);
397 	if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
398 		return (1);
399 	return (0);
400 }
401 
402 static int
403 shutdown_busycount2(struct buf *bp, void *info)
404 {
405 	if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
406 	    ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
407 		/*
408 		 * Only count buffers undergoing write I/O
409 		 * on the related vnode.
410 		 */
411 		if (bp->b_vp == NULL ||
412 		    bp->b_vp->v_track_write.bk_active == 0) {
413 			return (0);
414 		}
415 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
416 		kprintf(
417 	    "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
418 		    bp,
419 		    bp->b_flags, bp->b_loffset,
420 		    bp->b_bio2.bio_offset);
421 #endif
422 		return(1);
423 	}
424 	return(0);
425 }
426 
427 /*
428  * If the shutdown was a clean halt, behave accordingly.
429  */
430 static void
431 shutdown_halt(void *junk, int howto)
432 {
433 	if (howto & RB_HALT) {
434 		kprintf("\n");
435 		kprintf("The operating system has halted.\n");
436 #ifdef _KERNEL_VIRTUAL
437 		cpu_halt();
438 #else
439 		kprintf("Please press any key to reboot.\n\n");
440 		switch (cngetc()) {
441 		case -1:		/* No console, just die */
442 			cpu_halt();
443 			/* NOTREACHED */
444 		default:
445 			howto &= ~RB_HALT;
446 			break;
447 		}
448 #endif
449 	}
450 }
451 
452 /*
453  * Check to see if the system paniced, pause and then reboot
454  * according to the specified delay.
455  */
456 static void
457 shutdown_panic(void *junk, int howto)
458 {
459 	int loop;
460 
461 	if (howto & RB_DUMP) {
462 		if (PANIC_REBOOT_WAIT_TIME != 0) {
463 			if (PANIC_REBOOT_WAIT_TIME != -1) {
464 				kprintf("Automatic reboot in %d seconds - "
465 				       "press a key on the console to abort\n",
466 					PANIC_REBOOT_WAIT_TIME);
467 				for (loop = PANIC_REBOOT_WAIT_TIME * 10;
468 				     loop > 0; --loop) {
469 					DELAY(1000 * 100); /* 1/10th second */
470 					/* Did user type a key? */
471 					if (cncheckc() != -1)
472 						break;
473 				}
474 				if (!loop)
475 					return;
476 			}
477 		} else { /* zero time specified - reboot NOW */
478 			return;
479 		}
480 		kprintf("--> Press a key on the console to reboot,\n");
481 		kprintf("--> or switch off the system now.\n");
482 		cngetc();
483 	}
484 }
485 
486 /*
487  * Everything done, now reset
488  */
489 static void
490 shutdown_reset(void *junk, int howto)
491 {
492 	kprintf("Rebooting...\n");
493 	DELAY(1000000);	/* wait 1 sec for kprintf's to complete and be read */
494 	/* cpu_boot(howto); */ /* doesn't do anything at the moment */
495 	cpu_reset();
496 	/* NOTREACHED */ /* assuming reset worked */
497 }
498 
499 /*
500  * Try to remove FS references in the specified process.  This function
501  * is used during shutdown
502  */
503 static
504 void
505 shutdown_cleanup_proc(struct proc *p)
506 {
507 	struct filedesc *fdp;
508 	struct vmspace *vm;
509 
510 	if (p == NULL)
511 		return;
512 	if ((fdp = p->p_fd) != NULL) {
513 		kern_closefrom(0);
514 		if (fdp->fd_cdir) {
515 			cache_drop(&fdp->fd_ncdir);
516 			vrele(fdp->fd_cdir);
517 			fdp->fd_cdir = NULL;
518 		}
519 		if (fdp->fd_rdir) {
520 			cache_drop(&fdp->fd_nrdir);
521 			vrele(fdp->fd_rdir);
522 			fdp->fd_rdir = NULL;
523 		}
524 		if (fdp->fd_jdir) {
525 			cache_drop(&fdp->fd_njdir);
526 			vrele(fdp->fd_jdir);
527 			fdp->fd_jdir = NULL;
528 		}
529 	}
530 	if (p->p_vkernel)
531 		vkernel_exit(p);
532 	if (p->p_textvp) {
533 		vrele(p->p_textvp);
534 		p->p_textvp = NULL;
535 	}
536 	vm = p->p_vmspace;
537 	if (vm != NULL) {
538 		pmap_remove_pages(vmspace_pmap(vm),
539 				  VM_MIN_USER_ADDRESS,
540 				  VM_MAX_USER_ADDRESS);
541 		vm_map_remove(&vm->vm_map,
542 			      VM_MIN_USER_ADDRESS,
543 			      VM_MAX_USER_ADDRESS);
544 	}
545 }
546 
547 /*
548  * Magic number for savecore
549  *
550  * exported (symorder) and used at least by savecore(8)
551  *
552  * Mark it as used so that gcc doesn't optimize it away.
553  */
554 __attribute__((__used__))
555 	static u_long const dumpmag = 0x8fca0101UL;
556 
557 static int	dumpsize = 0;		/* also for savecore */
558 
559 static int	dodump = 1;
560 
561 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
562     "Try to perform coredump on kernel panic");
563 
564 static int
565 setdumpdev(cdev_t dev)
566 {
567 	struct partinfo pinfo;
568 	u_int64_t newdumplo;
569 	int error;
570 	int doopen;
571 
572 	if (dev == NULL) {
573 		dumpdev = dev;
574 		return (0);
575 	}
576 	bzero(&pinfo, sizeof(pinfo));
577 
578 	/*
579 	 * We have to open the device before we can perform ioctls on it,
580 	 * or the slice/label data may not be present.  Device opens are
581 	 * usually tracked by specfs, but the dump device can be set in
582 	 * early boot and may not be open so this is somewhat of a hack.
583 	 */
584 	doopen = (dev->si_sysref.refcnt == 1);
585 	if (doopen) {
586 		error = dev_dopen(dev, FREAD, S_IFCHR, proc0.p_ucred);
587 		if (error)
588 			return (error);
589 	}
590 	error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
591 	if (doopen)
592 		dev_dclose(dev, FREAD, S_IFCHR);
593 	if (error || pinfo.media_blocks == 0 || pinfo.media_blksize == 0)
594 		return (ENXIO);
595 
596 	newdumplo = pinfo.media_blocks -
597 		    ((u_int64_t)Maxmem * PAGE_SIZE / DEV_BSIZE);
598 	if ((int64_t)newdumplo < (int64_t)pinfo.reserved_blocks)
599 		return (ENOSPC);
600 	dumpdev = dev;
601 	dumplo64 = newdumplo;
602 	return (0);
603 }
604 
605 
606 /* ARGSUSED */
607 static void dump_conf (void *dummy);
608 static void
609 dump_conf(void *dummy)
610 {
611 	char *path;
612 	cdev_t dev;
613 
614 	path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK);
615 	if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
616 		dev = kgetdiskbyname(path);
617 		if (dev != NULL)
618 			dumpdev = dev;
619 	}
620 	kfree(path, M_TEMP);
621 	if (setdumpdev(dumpdev) != 0)
622 		dumpdev = NULL;
623 }
624 
625 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
626 
627 static int
628 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
629 {
630 	int error;
631 	udev_t ndumpdev;
632 
633 	ndumpdev = dev2udev(dumpdev);
634 	error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
635 	if (error == 0 && req->newptr != NULL)
636 		error = setdumpdev(udev2dev(ndumpdev, 0));
637 	return (error);
638 }
639 
640 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
641 	0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", "");
642 
643 /*
644  * Doadump comes here after turning off memory management and
645  * getting on the dump stack, either when called above, or by
646  * the auto-restart code.
647  */
648 static void
649 dumpsys(void)
650 {
651 	int	error;
652 
653 	savectx(&dumppcb);
654 	dumpthread = curthread;
655 	if (dumping++) {
656 		kprintf("Dump already in progress, bailing...\n");
657 		return;
658 	}
659 	if (!dodump)
660 		return;
661 	if (dumpdev == NULL)
662 		return;
663 	dumpsize = Maxmem;
664 	kprintf("\ndumping to dev %s, blockno %lld\n",
665 		devtoname(dumpdev), dumplo64);
666 	kprintf("dump ");
667 	error = dev_ddump(dumpdev);
668 	if (error == 0) {
669 		kprintf("succeeded\n");
670 		return;
671 	}
672 	kprintf("failed, reason: ");
673 	switch (error) {
674 	case ENOSYS:
675 	case ENODEV:
676 		kprintf("device doesn't support a dump routine\n");
677 		break;
678 
679 	case ENXIO:
680 		kprintf("device bad\n");
681 		break;
682 
683 	case EFAULT:
684 		kprintf("device not ready\n");
685 		break;
686 
687 	case EINVAL:
688 		kprintf("area improper\n");
689 		break;
690 
691 	case EIO:
692 		kprintf("i/o error\n");
693 		break;
694 
695 	case EINTR:
696 		kprintf("aborted from console\n");
697 		break;
698 
699 	default:
700 		kprintf("unknown, error = %d\n", error);
701 		break;
702 	}
703 }
704 
705 int
706 dumpstatus(vm_offset_t addr, off_t count)
707 {
708 	int c;
709 
710 	if (addr % (1024 * 1024) == 0) {
711 #ifdef HW_WDOG
712 		if (wdog_tickler)
713 			(*wdog_tickler)();
714 #endif
715 		kprintf("%ld ", (long)(count / (1024 * 1024)));
716 	}
717 
718 	if ((c = cncheckc()) == 0x03)
719 		return -1;
720 	else if (c != -1)
721 		kprintf("[CTRL-C to abort] ");
722 
723 	return 0;
724 }
725 
726 /*
727  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
728  * and then reboots.  If we are called twice, then we avoid trying to sync
729  * the disks as this often leads to recursive panics.
730  */
731 void
732 panic(const char *fmt, ...)
733 {
734 	int bootopt, newpanic;
735 	__va_list ap;
736 	static char buf[256];
737 
738 #ifdef SMP
739 	/*
740 	 * If a panic occurs on multiple cpus before the first is able to
741 	 * halt the other cpus, only one cpu is allowed to take the panic.
742 	 * Attempt to be verbose about this situation but if the kprintf()
743 	 * itself panics don't let us overrun the kernel stack.
744 	 *
745 	 * Be very nasty about descheduling our thread at the lowest
746 	 * level possible in an attempt to freeze the thread without
747 	 * inducing further panics.
748 	 *
749 	 * Bumping gd_trap_nesting_level will also bypass assertions in
750 	 * lwkt_switch() and allow us to switch away even if we are a
751 	 * FAST interrupt or IPI.
752 	 */
753 	if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
754 		panic_cpu_gd = mycpu;
755 	} else if (panic_cpu_gd != mycpu) {
756 		crit_enter();
757 		++mycpu->gd_trap_nesting_level;
758 		if (mycpu->gd_trap_nesting_level < 25) {
759 			kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
760 				mycpu->gd_cpuid, curthread);
761 		}
762 		curthread->td_release = NULL;	/* be a grinch */
763 		for (;;) {
764 			lwkt_deschedule_self(curthread);
765 			lwkt_switch();
766 		}
767 		/* NOT REACHED */
768 		/* --mycpu->gd_trap_nesting_level */
769 		/* crit_exit() */
770 	}
771 #endif
772 	bootopt = RB_AUTOBOOT | RB_DUMP;
773 	if (sync_on_panic == 0)
774 		bootopt |= RB_NOSYNC;
775 	newpanic = 0;
776 	if (panicstr)
777 		bootopt |= RB_NOSYNC;
778 	else {
779 		panicstr = fmt;
780 		newpanic = 1;
781 	}
782 
783 	__va_start(ap, fmt);
784 	kvsnprintf(buf, sizeof(buf), fmt, ap);
785 	if (panicstr == fmt)
786 		panicstr = buf;
787 	__va_end(ap);
788 	kprintf("panic: %s\n", buf);
789 #ifdef SMP
790 	/* two separate prints in case of an unmapped page and trap */
791 	kprintf("mp_lock = %08x; ", mp_lock);
792 	kprintf("cpuid = %d\n", mycpu->gd_cpuid);
793 #endif
794 
795 #if defined(DDB)
796 	if (newpanic && trace_on_panic)
797 		db_print_backtrace();
798 	if (debugger_on_panic)
799 		Debugger("panic");
800 #endif
801 	boot(bootopt);
802 }
803 
804 /*
805  * Support for poweroff delay.
806  */
807 #ifndef POWEROFF_DELAY
808 # define POWEROFF_DELAY 5000
809 #endif
810 static int poweroff_delay = POWEROFF_DELAY;
811 
812 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
813 	&poweroff_delay, 0, "");
814 
815 static void
816 poweroff_wait(void *junk, int howto)
817 {
818 	if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
819 		return;
820 	DELAY(poweroff_delay * 1000);
821 }
822 
823 /*
824  * Some system processes (e.g. syncer) need to be stopped at appropriate
825  * points in their main loops prior to a system shutdown, so that they
826  * won't interfere with the shutdown process (e.g. by holding a disk buf
827  * to cause sync to fail).  For each of these system processes, register
828  * shutdown_kproc() as a handler for one of shutdown events.
829  */
830 static int kproc_shutdown_wait = 60;
831 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
832     &kproc_shutdown_wait, 0, "");
833 
834 void
835 shutdown_kproc(void *arg, int howto)
836 {
837 	struct thread *td;
838 	struct proc *p;
839 	int error;
840 
841 	if (panicstr)
842 		return;
843 
844 	td = (struct thread *)arg;
845 	if ((p = td->td_proc) != NULL) {
846 	    kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
847 		kproc_shutdown_wait, p->p_comm);
848 	} else {
849 	    kprintf("Waiting (max %d seconds) for system thread %s to stop...",
850 		kproc_shutdown_wait, td->td_comm);
851 	}
852 	error = suspend_kproc(td, kproc_shutdown_wait * hz);
853 
854 	if (error == EWOULDBLOCK)
855 		kprintf("timed out\n");
856 	else
857 		kprintf("stopped\n");
858 }
859