xref: /freebsd/sys/kern/kern_shutdown.c (revision cec50dea12481dc578c0805c887ab2097e1c06c5)
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("$FreeBSD$");
39 
40 #include "opt_kdb.h"
41 #include "opt_mac.h"
42 #include "opt_panic.h"
43 #include "opt_show_busybufs.h"
44 #include "opt_sched.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
85 int debugger_on_panic = 0;
86 #else
87 int debugger_on_panic = 1;
88 #endif
89 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
90 	&debugger_on_panic, 0, "Run debugger on kernel panic");
91 
92 #ifdef KDB_TRACE
93 int trace_on_panic = 1;
94 #else
95 int trace_on_panic = 0;
96 #endif
97 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
98 	&trace_on_panic, 0, "Print stack trace on kernel panic");
99 #endif /* KDB */
100 
101 int sync_on_panic = 0;
102 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
103 	&sync_on_panic, 0, "Do a sync before rebooting from a panic");
104 
105 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
106 
107 /*
108  * Variable panicstr contains argument to first call to panic; used as flag
109  * to indicate that the kernel has already called panic.
110  */
111 const char *panicstr;
112 
113 int dumping;				/* system is dumping */
114 static struct dumperinfo dumper;	/* our selected dumper */
115 
116 /* Context information for dump-debuggers. */
117 static struct pcb dumppcb;		/* Registers. */
118 static lwpid_t dumptid;			/* Thread ID. */
119 
120 static void boot(int) __dead2;
121 static void poweroff_wait(void *, int);
122 static void shutdown_halt(void *junk, int howto);
123 static void shutdown_panic(void *junk, int howto);
124 static void shutdown_reset(void *junk, int howto);
125 
126 /* register various local shutdown events */
127 static void
128 shutdown_conf(void *unused)
129 {
130 
131 	EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
132 	    SHUTDOWN_PRI_FIRST);
133 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
134 	    SHUTDOWN_PRI_LAST + 100);
135 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
136 	    SHUTDOWN_PRI_LAST + 100);
137 	EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
138 	    SHUTDOWN_PRI_LAST + 200);
139 }
140 
141 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
142 
143 /*
144  * The system call that results in a reboot
145  *
146  * MPSAFE
147  */
148 /* ARGSUSED */
149 int
150 reboot(struct thread *td, struct reboot_args *uap)
151 {
152 	int error;
153 
154 	error = 0;
155 #ifdef MAC
156 	error = mac_check_system_reboot(td->td_ucred, uap->opt);
157 #endif
158 	if (error == 0)
159 		error = suser(td);
160 	if (error == 0) {
161 		mtx_lock(&Giant);
162 		boot(uap->opt);
163 		mtx_unlock(&Giant);
164 	}
165 	return (error);
166 }
167 
168 /*
169  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
170  */
171 static int shutdown_howto = 0;
172 
173 void
174 shutdown_nice(int howto)
175 {
176 
177 	shutdown_howto = howto;
178 
179 	/* Send a signal to init(8) and have it shutdown the world */
180 	if (initproc != NULL) {
181 		PROC_LOCK(initproc);
182 		psignal(initproc, SIGINT);
183 		PROC_UNLOCK(initproc);
184 	} else {
185 		/* No init(8) running, so simply reboot */
186 		boot(RB_NOSYNC);
187 	}
188 	return;
189 }
190 static int	waittime = -1;
191 
192 static void
193 print_uptime(void)
194 {
195 	int f;
196 	struct timespec ts;
197 
198 	getnanouptime(&ts);
199 	printf("Uptime: ");
200 	f = 0;
201 	if (ts.tv_sec >= 86400) {
202 		printf("%ldd", (long)ts.tv_sec / 86400);
203 		ts.tv_sec %= 86400;
204 		f = 1;
205 	}
206 	if (f || ts.tv_sec >= 3600) {
207 		printf("%ldh", (long)ts.tv_sec / 3600);
208 		ts.tv_sec %= 3600;
209 		f = 1;
210 	}
211 	if (f || ts.tv_sec >= 60) {
212 		printf("%ldm", (long)ts.tv_sec / 60);
213 		ts.tv_sec %= 60;
214 		f = 1;
215 	}
216 	printf("%lds\n", (long)ts.tv_sec);
217 }
218 
219 static void
220 doadump(void)
221 {
222 
223 	/*
224 	 * Sometimes people have to call this from the kernel debugger.
225 	 * (if 'panic' can not dump)
226 	 * Give them a clue as to why they can't dump.
227 	 */
228 	if (dumper.dumper == NULL) {
229 		printf("Cannot dump. No dump device defined.\n");
230 		return;
231 	}
232 
233 	savectx(&dumppcb);
234 	dumptid = curthread->td_tid;
235 	dumping++;
236 	dumpsys(&dumper);
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 	static int first_buf_printf = 1;
248 
249 	/* collect extra flags that shutdown_nice might have set */
250 	howto |= shutdown_howto;
251 
252 	/* We are out of the debugger now. */
253 	kdb_active = 0;
254 
255 #ifdef SMP
256 	if (smp_active)
257 		printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
258 #endif
259 	/*
260 	 * Do any callouts that should be done BEFORE syncing the filesystems.
261 	 */
262 	EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
263 
264 	/*
265 	 * Now sync filesystems
266 	 */
267 	if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
268 		register struct buf *bp;
269 		int iter, nbusy, pbusy;
270 #ifndef PREEMPTION
271 		int subiter;
272 #endif
273 
274 		waittime = 0;
275 
276 		sync(&thread0, NULL);
277 
278 		/*
279 		 * With soft updates, some buffers that are
280 		 * written will be remarked as dirty until other
281 		 * buffers are written.
282 		 */
283 		for (iter = pbusy = 0; iter < 20; iter++) {
284 			nbusy = 0;
285 			for (bp = &buf[nbuf]; --bp >= buf; ) {
286 				if ((bp->b_flags & B_INVAL) == 0 &&
287 				    BUF_REFCNT(bp) > 0) {
288 					nbusy++;
289 				} else if ((bp->b_flags & (B_DELWRI | B_INVAL))
290 						== B_DELWRI) {
291 					/* bawrite(bp);*/
292 					nbusy++;
293 				}
294 			}
295 			if (nbusy == 0) {
296 				if (first_buf_printf)
297 					printf("No buffers busy after final sync");
298 				break;
299 			}
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 		printf("\n");
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 			if (!first_buf_printf)
366 				printf("Final sync complete\n");
367 			/*
368 			 * Unmount filesystems
369 			 */
370 			if (panicstr == 0)
371 				vfs_unmountall();
372 		}
373 		DELAY(100000);		/* wait for console output to finish */
374 	}
375 
376 	print_uptime();
377 
378 	/*
379 	 * Ok, now do things that assume all filesystem activity has
380 	 * been completed.
381 	 */
382 	EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
383 	splhigh();
384 	if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !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  */
397 static void
398 shutdown_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  */
420 static void
421 shutdown_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  */
453 static void
454 shutdown_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
465 static 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  */
475 void
476 panic(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 	printf("cpuid = %d\n", PCPU_GET(cpuid));
519 #endif
520 
521 #ifdef KDB
522 	if (newpanic && trace_on_panic)
523 		kdb_backtrace();
524 	if (debugger_on_panic)
525 		kdb_enter("panic");
526 #ifdef RESTARTABLE_PANICS
527 	/* See if the user aborted the panic, in which case we continue. */
528 	if (panicstr == NULL) {
529 #ifdef SMP
530 		atomic_store_rel_int(&panic_cpu, NOCPU);
531 #endif
532 		return;
533 	}
534 #endif
535 #endif
536 	mtx_lock_spin(&sched_lock);
537 	td->td_flags |= TDF_INPANIC;
538 	mtx_unlock_spin(&sched_lock);
539 	if (!sync_on_panic)
540 		bootopt |= RB_NOSYNC;
541 	boot(bootopt);
542 }
543 
544 /*
545  * Support for poweroff delay.
546  */
547 #ifndef POWEROFF_DELAY
548 # define POWEROFF_DELAY 5000
549 #endif
550 static int poweroff_delay = POWEROFF_DELAY;
551 
552 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
553 	&poweroff_delay, 0, "");
554 
555 static void
556 poweroff_wait(void *junk, int howto)
557 {
558 
559 	if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
560 		return;
561 	DELAY(poweroff_delay * 1000);
562 }
563 
564 /*
565  * Some system processes (e.g. syncer) need to be stopped at appropriate
566  * points in their main loops prior to a system shutdown, so that they
567  * won't interfere with the shutdown process (e.g. by holding a disk buf
568  * to cause sync to fail).  For each of these system processes, register
569  * shutdown_kproc() as a handler for one of shutdown events.
570  */
571 static int kproc_shutdown_wait = 60;
572 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
573     &kproc_shutdown_wait, 0, "");
574 
575 void
576 kproc_shutdown(void *arg, int howto)
577 {
578 	struct proc *p;
579 	char procname[MAXCOMLEN + 1];
580 	int error;
581 
582 	if (panicstr)
583 		return;
584 
585 	p = (struct proc *)arg;
586 	strlcpy(procname, p->p_comm, sizeof(procname));
587 	printf("Waiting (max %d seconds) for system process `%s' to stop...",
588 	    kproc_shutdown_wait, procname);
589 	error = kthread_suspend(p, kproc_shutdown_wait * hz);
590 
591 	if (error == EWOULDBLOCK)
592 		printf("timed out\n");
593 	else
594 		printf("done\n");
595 }
596 
597 /* Registration of dumpers */
598 int
599 set_dumper(struct dumperinfo *di)
600 {
601 
602 	if (di == NULL) {
603 		bzero(&dumper, sizeof dumper);
604 		return (0);
605 	}
606 	if (dumper.dumper != NULL)
607 		return (EBUSY);
608 	dumper = *di;
609 	return (0);
610 }
611 
612 #if defined(__powerpc__)
613 void
614 dumpsys(struct dumperinfo *di __unused)
615 {
616 
617 	printf("Kernel dumps not implemented on this architecture\n");
618 }
619 #endif
620