minor code optimization.
[dragonfly.git] / sys / kern / kern_shutdown.c
CommitLineData
984263bc
MD
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 $
3020e3be 40 * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.3 2003/06/19 01:55:06 dillon Exp $
984263bc
MD
41 */
42
43#include "opt_ddb.h"
44#include "opt_hw_wdog.h"
45#include "opt_panic.h"
46#include "opt_show_busybufs.h"
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/eventhandler.h>
51#include <sys/buf.h>
52#include <sys/disklabel.h>
53#include <sys/reboot.h>
54#include <sys/proc.h>
55#include <sys/vnode.h>
56#include <sys/kernel.h>
57#include <sys/kthread.h>
58#include <sys/malloc.h>
59#include <sys/mount.h>
60#include <sys/queue.h>
61#include <sys/sysctl.h>
62#include <sys/conf.h>
63#include <sys/sysproto.h>
64#include <sys/cons.h>
3020e3be 65#include <sys/buf2.h>
984263bc
MD
66
67#include <machine/pcb.h>
68#include <machine/clock.h>
69#include <machine/md_var.h>
70#include <machine/smp.h> /* smp_active, cpuid */
71
72#include <sys/signalvar.h>
73
74#ifndef PANIC_REBOOT_WAIT_TIME
75#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
76#endif
77
78/*
79 * Note that stdarg.h and the ANSI style va_start macro is used for both
80 * ANSI and traditional C compilers.
81 */
82#include <machine/stdarg.h>
83
84#ifdef DDB
85#ifdef DDB_UNATTENDED
86int debugger_on_panic = 0;
87#else
88int debugger_on_panic = 1;
89#endif
90SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
91 &debugger_on_panic, 0, "Run debugger on kernel panic");
92#endif
93
94SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
95
96#ifdef HW_WDOG
97/*
98 * If there is a hardware watchdog, point this at the function needed to
99 * hold it off.
100 * It's needed when the kernel needs to do some lengthy operations.
101 * e.g. in wd.c when dumping core.. It's most annoying to have
102 * your precious core-dump only half written because the wdog kicked in.
103 */
104watchdog_tickle_fn wdog_tickler = NULL;
105#endif /* HW_WDOG */
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 */
111const char *panicstr;
112
113int dumping; /* system is dumping */
114
115static void boot __P((int)) __dead2;
116static void dumpsys __P((void));
117static int setdumpdev __P((dev_t dev));
118static void poweroff_wait __P((void *, int));
119static void print_uptime __P((void));
120static void shutdown_halt __P((void *junk, int howto));
121static void shutdown_panic __P((void *junk, int howto));
122static void shutdown_reset __P((void *junk, int howto));
123
124/* register various local shutdown events */
125static void
126shutdown_conf(void *unused)
127{
128 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
129 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
130 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
131 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
132}
133
134SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
135
136/* ARGSUSED */
137
138/*
139 * The system call that results in a reboot
140 */
141int
142reboot(p, uap)
143 struct proc *p;
144 struct reboot_args *uap;
145{
146 int error;
147
148 if ((error = suser(p)))
149 return (error);
150
151 boot(uap->opt);
152 return (0);
153}
154
155/*
156 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
157 */
158static int shutdown_howto = 0;
159
160void
161shutdown_nice(int howto)
162{
163 shutdown_howto = howto;
164
165 /* Send a signal to init(8) and have it shutdown the world */
166 if (initproc != NULL) {
167 psignal(initproc, SIGINT);
168 } else {
169 /* No init(8) running, so simply reboot */
170 boot(RB_NOSYNC);
171 }
172 return;
173}
174static int waittime = -1;
175static struct pcb dumppcb;
176
177static void
178print_uptime()
179{
180 int f;
181 struct timespec ts;
182
183 getnanouptime(&ts);
184 printf("Uptime: ");
185 f = 0;
186 if (ts.tv_sec >= 86400) {
187 printf("%ldd", ts.tv_sec / 86400);
188 ts.tv_sec %= 86400;
189 f = 1;
190 }
191 if (f || ts.tv_sec >= 3600) {
192 printf("%ldh", ts.tv_sec / 3600);
193 ts.tv_sec %= 3600;
194 f = 1;
195 }
196 if (f || ts.tv_sec >= 60) {
197 printf("%ldm", ts.tv_sec / 60);
198 ts.tv_sec %= 60;
199 f = 1;
200 }
201 printf("%lds\n", ts.tv_sec);
202}
203
204/*
205 * Go through the rigmarole of shutting down..
206 * this used to be in machdep.c but I'll be dammned if I could see
207 * anything machine dependant in it.
208 */
209static void
210boot(howto)
211 int howto;
212{
213
214 /* collect extra flags that shutdown_nice might have set */
215 howto |= shutdown_howto;
216
217#ifdef SMP
218 if (smp_active) {
219 printf("boot() called on cpu#%d\n", cpuid);
220 }
221#endif
222 /*
223 * Do any callouts that should be done BEFORE syncing the filesystems.
224 */
225 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
226
227 /*
228 * Now sync filesystems
229 */
230 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
231 register struct buf *bp;
232 int iter, nbusy, pbusy;
233
234 waittime = 0;
235 printf("\nsyncing disks... ");
236
237 sync(&proc0, NULL);
238
239 /*
240 * With soft updates, some buffers that are
241 * written will be remarked as dirty until other
242 * buffers are written.
243 */
244 for (iter = pbusy = 0; iter < 20; iter++) {
245 nbusy = 0;
246 for (bp = &buf[nbuf]; --bp >= buf; ) {
247 if ((bp->b_flags & B_INVAL) == 0 &&
248 BUF_REFCNT(bp) > 0) {
249 nbusy++;
250 } else if ((bp->b_flags & (B_DELWRI | B_INVAL))
251 == B_DELWRI) {
252 /* bawrite(bp);*/
253 nbusy++;
254 }
255 }
256 if (nbusy == 0)
257 break;
258 printf("%d ", nbusy);
259 if (nbusy < pbusy)
260 iter = 0;
261 pbusy = nbusy;
262 if (iter > 5 && bioops.io_sync)
263 (*bioops.io_sync)(NULL);
264 sync(&proc0, NULL);
265 DELAY(50000 * iter);
266 }
267 printf("\n");
268 /*
269 * Count only busy local buffers to prevent forcing
270 * a fsck if we're just a client of a wedged NFS server
271 */
272 nbusy = 0;
273 for (bp = &buf[nbuf]; --bp >= buf; ) {
274 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
275 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
276 if (bp->b_dev == NODEV) {
277 TAILQ_REMOVE(&mountlist,
278 bp->b_vp->v_mount, mnt_list);
279 continue;
280 }
281 nbusy++;
282#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
283 printf(
284 "%p %d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
285 bp, nbusy, devtoname(bp->b_dev),
286 bp->b_flags, (long)bp->b_blkno,
287 (long)bp->b_lblkno);
288#endif
289 }
290 }
291 if (nbusy) {
292 /*
293 * Failed to sync all blocks. Indicate this and don't
294 * unmount filesystems (thus forcing an fsck on reboot).
295 */
296 printf("giving up on %d buffers\n", nbusy);
297 DELAY(5000000); /* 5 seconds */
298 } else {
299 printf("done\n");
300 /*
301 * Unmount filesystems
302 */
303 if (panicstr == 0)
304 vfs_unmountall();
305 }
306 DELAY(100000); /* wait for console output to finish */
307 }
308
309 print_uptime();
310
311 /*
312 * Ok, now do things that assume all filesystem activity has
313 * been completed.
314 */
315 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
316 splhigh();
317 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
318 dumpsys();
319
320 /* Now that we're going to really halt the system... */
321 EVENTHANDLER_INVOKE(shutdown_final, howto);
322
323 for(;;) ; /* safety against shutdown_reset not working */
324 /* NOTREACHED */
325}
326
327/*
328 * If the shutdown was a clean halt, behave accordingly.
329 */
330static void
331shutdown_halt(void *junk, int howto)
332{
333 if (howto & RB_HALT) {
334 printf("\n");
335 printf("The operating system has halted.\n");
336 printf("Please press any key to reboot.\n\n");
337 switch (cngetc()) {
338 case -1: /* No console, just die */
339 cpu_halt();
340 /* NOTREACHED */
341 default:
342 howto &= ~RB_HALT;
343 break;
344 }
345 }
346}
347
348/*
349 * Check to see if the system paniced, pause and then reboot
350 * according to the specified delay.
351 */
352static void
353shutdown_panic(void *junk, int howto)
354{
355 int loop;
356
357 if (howto & RB_DUMP) {
358 if (PANIC_REBOOT_WAIT_TIME != 0) {
359 if (PANIC_REBOOT_WAIT_TIME != -1) {
360 printf("Automatic reboot in %d seconds - "
361 "press a key on the console to abort\n",
362 PANIC_REBOOT_WAIT_TIME);
363 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
364 loop > 0; --loop) {
365 DELAY(1000 * 100); /* 1/10th second */
366 /* Did user type a key? */
367 if (cncheckc() != -1)
368 break;
369 }
370 if (!loop)
371 return;
372 }
373 } else { /* zero time specified - reboot NOW */
374 return;
375 }
376 printf("--> Press a key on the console to reboot,\n");
377 printf("--> or switch off the system now.\n");
378 cngetc();
379 }
380}
381
382/*
383 * Everything done, now reset
384 */
385static void
386shutdown_reset(void *junk, int howto)
387{
388 printf("Rebooting...\n");
389 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
390 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
391 cpu_reset();
392 /* NOTREACHED */ /* assuming reset worked */
393}
394
395/*
396 * Magic number for savecore
397 *
398 * exported (symorder) and used at least by savecore(8)
399 *
400 */
401static u_long const dumpmag = 0x8fca0101UL;
402
403static int dumpsize = 0; /* also for savecore */
404
405static int dodump = 1;
406
407SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
408 "Try to perform coredump on kernel panic");
409
410static int
411setdumpdev(dev)
412 dev_t dev;
413{
414 int psize;
415 long newdumplo;
416
417 if (dev == NODEV) {
418 dumpdev = dev;
419 return (0);
420 }
421 if (devsw(dev) == NULL)
422 return (ENXIO); /* XXX is this right? */
423 if (devsw(dev)->d_psize == NULL)
424 return (ENXIO); /* XXX should be ENODEV ? */
425 psize = devsw(dev)->d_psize(dev);
426 if (psize == -1)
427 return (ENXIO); /* XXX should be ENODEV ? */
428 /*
429 * XXX should clean up checking in dumpsys() to be more like this.
430 */
431 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
432 if (newdumplo <= LABELSECTOR)
433 return (ENOSPC);
434 dumpdev = dev;
435 dumplo = newdumplo;
436 return (0);
437}
438
439
440/* ARGSUSED */
441static void dump_conf __P((void *dummy));
442static void
443dump_conf(dummy)
444 void *dummy;
445{
446 char *path;
447 dev_t dev;
448
449 path = malloc(MNAMELEN, M_TEMP, M_WAITOK);
450 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
451 dev = getdiskbyname(path);
452 if (dev != NODEV)
453 dumpdev = dev;
454 }
455 free(path, M_TEMP);
456 if (setdumpdev(dumpdev) != 0)
457 dumpdev = NODEV;
458}
459
460SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
461
462static int
463sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
464{
465 int error;
466 udev_t ndumpdev;
467
468 ndumpdev = dev2udev(dumpdev);
469 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
470 if (error == 0 && req->newptr != NULL)
471 error = setdumpdev(udev2dev(ndumpdev, 0));
472 return (error);
473}
474
475SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
476 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
477
478/*
479 * Doadump comes here after turning off memory management and
480 * getting on the dump stack, either when called above, or by
481 * the auto-restart code.
482 */
483static void
484dumpsys(void)
485{
486 int error;
487
488 savectx(&dumppcb);
489 if (dumping++) {
490 printf("Dump already in progress, bailing...\n");
491 return;
492 }
493 if (!dodump)
494 return;
495 if (dumpdev == NODEV)
496 return;
497 if (!(devsw(dumpdev)))
498 return;
499 if (!(devsw(dumpdev)->d_dump))
500 return;
501 dumpsize = Maxmem;
502 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
503 printf("dump ");
504 error = (*devsw(dumpdev)->d_dump)(dumpdev);
505 if (error == 0) {
506 printf("succeeded\n");
507 return;
508 }
509 printf("failed, reason: ");
510 switch (error) {
511 case ENODEV:
512 printf("device doesn't support a dump routine\n");
513 break;
514
515 case ENXIO:
516 printf("device bad\n");
517 break;
518
519 case EFAULT:
520 printf("device not ready\n");
521 break;
522
523 case EINVAL:
524 printf("area improper\n");
525 break;
526
527 case EIO:
528 printf("i/o error\n");
529 break;
530
531 case EINTR:
532 printf("aborted from console\n");
533 break;
534
535 default:
536 printf("unknown, error = %d\n", error);
537 break;
538 }
539}
540
541int
542dumpstatus(vm_offset_t addr, off_t count)
543{
544 int c;
545
546 if (addr % (1024 * 1024) == 0) {
547#ifdef HW_WDOG
548 if (wdog_tickler)
549 (*wdog_tickler)();
550#endif
551 printf("%ld ", (long)(count / (1024 * 1024)));
552 }
553
554 if ((c = cncheckc()) == 0x03)
555 return -1;
556 else if (c != -1)
557 printf("[CTRL-C to abort] ");
558
559 return 0;
560}
561
562/*
563 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
564 * and then reboots. If we are called twice, then we avoid trying to sync
565 * the disks as this often leads to recursive panics.
566 */
567void
568panic(const char *fmt, ...)
569{
570 int bootopt;
571 va_list ap;
572 static char buf[256];
573
574 bootopt = RB_AUTOBOOT | RB_DUMP;
575 if (panicstr)
576 bootopt |= RB_NOSYNC;
577 else
578 panicstr = fmt;
579
580 va_start(ap, fmt);
581 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
582 if (panicstr == fmt)
583 panicstr = buf;
584 va_end(ap);
585 printf("panic: %s\n", buf);
586#ifdef SMP
587 /* three seperate prints in case of an unmapped page and trap */
588 printf("mp_lock = %08x; ", mp_lock);
589 printf("cpuid = %d; ", cpuid);
590 printf("lapic.id = %08x\n", lapic.id);
591#endif
592
593#if defined(DDB)
594 if (debugger_on_panic)
595 Debugger ("panic");
596#endif
597 boot(bootopt);
598}
599
600/*
601 * Support for poweroff delay.
602 */
603#ifndef POWEROFF_DELAY
604# define POWEROFF_DELAY 5000
605#endif
606static int poweroff_delay = POWEROFF_DELAY;
607
608SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
609 &poweroff_delay, 0, "");
610
611static void
612poweroff_wait(void *junk, int howto)
613{
614 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
615 return;
616 DELAY(poweroff_delay * 1000);
617}
618
619/*
620 * Some system processes (e.g. syncer) need to be stopped at appropriate
621 * points in their main loops prior to a system shutdown, so that they
622 * won't interfere with the shutdown process (e.g. by holding a disk buf
623 * to cause sync to fail). For each of these system processes, register
624 * shutdown_kproc() as a handler for one of shutdown events.
625 */
626static int kproc_shutdown_wait = 60;
627SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
628 &kproc_shutdown_wait, 0, "");
629
630void
631shutdown_kproc(void *arg, int howto)
632{
633 struct proc *p;
634 int error;
635
636 if (panicstr)
637 return;
638
639 p = (struct proc *)arg;
640 printf("Waiting (max %d seconds) for system process `%s' to stop...",
641 kproc_shutdown_wait, p->p_comm);
642 error = suspend_kproc(p, kproc_shutdown_wait * hz);
643
644 if (error == EWOULDBLOCK)
645 printf("timed out\n");
646 else
647 printf("stopped\n");
648}