Correct bugs introduced in the last syscall separation commit. The
[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 $
e2565a42 40 * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.13 2003/11/09 02:22:36 dillon Exp $
984263bc
MD
41 */
42
43#include "opt_ddb.h"
b1e04573 44#include "opt_ddb_trace.h"
984263bc
MD
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/conf.h>
64#include <sys/sysproto.h>
335dda38 65#include <sys/device.h>
984263bc 66#include <sys/cons.h>
3020e3be 67#include <sys/buf2.h>
984263bc
MD
68
69#include <machine/pcb.h>
70#include <machine/clock.h>
71#include <machine/md_var.h>
72#include <machine/smp.h> /* smp_active, cpuid */
73
74#include <sys/signalvar.h>
75
76#ifndef PANIC_REBOOT_WAIT_TIME
77#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
78#endif
79
80/*
81 * Note that stdarg.h and the ANSI style va_start macro is used for both
e2565a42
MD
82 * ANSI and traditional C compilers. We use the machine version to stay
83 * within the confines of the kernel header files.
984263bc
MD
84 */
85#include <machine/stdarg.h>
86
87#ifdef DDB
88#ifdef DDB_UNATTENDED
89int debugger_on_panic = 0;
90#else
91int debugger_on_panic = 1;
92#endif
93SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
94 &debugger_on_panic, 0, "Run debugger on kernel panic");
b1e04573
JH
95
96extern void db_print_backtrace(void);
97
98#ifdef DDB_TRACE
99int trace_on_panic = 1;
100#else
101int trace_on_panic = 0;
102#endif
103SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
104 &trace_on_panic, 0, "Print stack trace on kernel panic");
984263bc
MD
105#endif
106
107SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
108
109#ifdef HW_WDOG
110/*
111 * If there is a hardware watchdog, point this at the function needed to
112 * hold it off.
113 * It's needed when the kernel needs to do some lengthy operations.
114 * e.g. in wd.c when dumping core.. It's most annoying to have
115 * your precious core-dump only half written because the wdog kicked in.
116 */
117watchdog_tickle_fn wdog_tickler = NULL;
118#endif /* HW_WDOG */
119
120/*
121 * Variable panicstr contains argument to first call to panic; used as flag
122 * to indicate that the kernel has already called panic.
123 */
124const char *panicstr;
125
126int dumping; /* system is dumping */
127
402ed7e1
RG
128static void boot (int) __dead2;
129static void dumpsys (void);
130static int setdumpdev (dev_t dev);
131static void poweroff_wait (void *, int);
132static void print_uptime (void);
133static void shutdown_halt (void *junk, int howto);
134static void shutdown_panic (void *junk, int howto);
135static void shutdown_reset (void *junk, int howto);
984263bc
MD
136
137/* register various local shutdown events */
138static void
139shutdown_conf(void *unused)
140{
141 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
142 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
143 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
144 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
145}
146
147SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
148
149/* ARGSUSED */
150
151/*
152 * The system call that results in a reboot
153 */
154int
41c20dac 155reboot(struct reboot_args *uap)
984263bc 156{
dadab5e9 157 struct thread *td = curthread;
984263bc
MD
158 int error;
159
dadab5e9 160 if ((error = suser(td)))
984263bc
MD
161 return (error);
162
163 boot(uap->opt);
164 return (0);
165}
166
167/*
168 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
169 */
170static int shutdown_howto = 0;
171
172void
173shutdown_nice(int howto)
174{
175 shutdown_howto = howto;
176
177 /* Send a signal to init(8) and have it shutdown the world */
178 if (initproc != NULL) {
179 psignal(initproc, SIGINT);
180 } else {
181 /* No init(8) running, so simply reboot */
182 boot(RB_NOSYNC);
183 }
184 return;
185}
186static int waittime = -1;
187static struct pcb dumppcb;
188
189static void
190print_uptime()
191{
192 int f;
193 struct timespec ts;
194
195 getnanouptime(&ts);
196 printf("Uptime: ");
197 f = 0;
198 if (ts.tv_sec >= 86400) {
199 printf("%ldd", ts.tv_sec / 86400);
200 ts.tv_sec %= 86400;
201 f = 1;
202 }
203 if (f || ts.tv_sec >= 3600) {
204 printf("%ldh", ts.tv_sec / 3600);
205 ts.tv_sec %= 3600;
206 f = 1;
207 }
208 if (f || ts.tv_sec >= 60) {
209 printf("%ldm", ts.tv_sec / 60);
210 ts.tv_sec %= 60;
211 f = 1;
212 }
213 printf("%lds\n", ts.tv_sec);
214}
215
216/*
217 * Go through the rigmarole of shutting down..
218 * this used to be in machdep.c but I'll be dammned if I could see
219 * anything machine dependant in it.
220 */
221static void
41c20dac 222boot(int howto)
984263bc
MD
223{
224
225 /* collect extra flags that shutdown_nice might have set */
226 howto |= shutdown_howto;
227
228#ifdef SMP
229 if (smp_active) {
72740893 230 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
984263bc
MD
231 }
232#endif
233 /*
234 * Do any callouts that should be done BEFORE syncing the filesystems.
235 */
236 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
237
238 /*
239 * Now sync filesystems
240 */
241 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
1fd87d54 242 struct buf *bp;
984263bc
MD
243 int iter, nbusy, pbusy;
244
245 waittime = 0;
246 printf("\nsyncing disks... ");
247
41c20dac 248 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
984263bc
MD
249
250 /*
251 * With soft updates, some buffers that are
252 * written will be remarked as dirty until other
253 * buffers are written.
254 */
255 for (iter = pbusy = 0; iter < 20; iter++) {
256 nbusy = 0;
257 for (bp = &buf[nbuf]; --bp >= buf; ) {
258 if ((bp->b_flags & B_INVAL) == 0 &&
259 BUF_REFCNT(bp) > 0) {
260 nbusy++;
261 } else if ((bp->b_flags & (B_DELWRI | B_INVAL))
262 == B_DELWRI) {
263 /* bawrite(bp);*/
264 nbusy++;
265 }
266 }
267 if (nbusy == 0)
268 break;
269 printf("%d ", nbusy);
270 if (nbusy < pbusy)
271 iter = 0;
272 pbusy = nbusy;
273 if (iter > 5 && bioops.io_sync)
274 (*bioops.io_sync)(NULL);
41c20dac 275 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
984263bc
MD
276 DELAY(50000 * iter);
277 }
278 printf("\n");
279 /*
280 * Count only busy local buffers to prevent forcing
281 * a fsck if we're just a client of a wedged NFS server
282 */
283 nbusy = 0;
284 for (bp = &buf[nbuf]; --bp >= buf; ) {
285 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
286 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
287 if (bp->b_dev == NODEV) {
288 TAILQ_REMOVE(&mountlist,
289 bp->b_vp->v_mount, mnt_list);
290 continue;
291 }
292 nbusy++;
293#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
294 printf(
295 "%p %d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
296 bp, nbusy, devtoname(bp->b_dev),
297 bp->b_flags, (long)bp->b_blkno,
298 (long)bp->b_lblkno);
299#endif
300 }
301 }
302 if (nbusy) {
303 /*
304 * Failed to sync all blocks. Indicate this and don't
305 * unmount filesystems (thus forcing an fsck on reboot).
306 */
307 printf("giving up on %d buffers\n", nbusy);
308 DELAY(5000000); /* 5 seconds */
309 } else {
310 printf("done\n");
311 /*
312 * Unmount filesystems
313 */
314 if (panicstr == 0)
315 vfs_unmountall();
316 }
317 DELAY(100000); /* wait for console output to finish */
318 }
319
320 print_uptime();
321
322 /*
323 * Ok, now do things that assume all filesystem activity has
324 * been completed.
325 */
326 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
327 splhigh();
328 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
329 dumpsys();
330
331 /* Now that we're going to really halt the system... */
332 EVENTHANDLER_INVOKE(shutdown_final, howto);
333
334 for(;;) ; /* safety against shutdown_reset not working */
335 /* NOTREACHED */
336}
337
338/*
339 * If the shutdown was a clean halt, behave accordingly.
340 */
341static void
342shutdown_halt(void *junk, int howto)
343{
344 if (howto & RB_HALT) {
345 printf("\n");
346 printf("The operating system has halted.\n");
347 printf("Please press any key to reboot.\n\n");
348 switch (cngetc()) {
349 case -1: /* No console, just die */
350 cpu_halt();
351 /* NOTREACHED */
352 default:
353 howto &= ~RB_HALT;
354 break;
355 }
356 }
357}
358
359/*
360 * Check to see if the system paniced, pause and then reboot
361 * according to the specified delay.
362 */
363static void
364shutdown_panic(void *junk, int howto)
365{
366 int loop;
367
368 if (howto & RB_DUMP) {
369 if (PANIC_REBOOT_WAIT_TIME != 0) {
370 if (PANIC_REBOOT_WAIT_TIME != -1) {
371 printf("Automatic reboot in %d seconds - "
372 "press a key on the console to abort\n",
373 PANIC_REBOOT_WAIT_TIME);
374 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
375 loop > 0; --loop) {
376 DELAY(1000 * 100); /* 1/10th second */
377 /* Did user type a key? */
378 if (cncheckc() != -1)
379 break;
380 }
381 if (!loop)
382 return;
383 }
384 } else { /* zero time specified - reboot NOW */
385 return;
386 }
387 printf("--> Press a key on the console to reboot,\n");
388 printf("--> or switch off the system now.\n");
389 cngetc();
390 }
391}
392
393/*
394 * Everything done, now reset
395 */
396static void
397shutdown_reset(void *junk, int howto)
398{
399 printf("Rebooting...\n");
400 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
401 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
402 cpu_reset();
403 /* NOTREACHED */ /* assuming reset worked */
404}
405
406/*
407 * Magic number for savecore
408 *
409 * exported (symorder) and used at least by savecore(8)
410 *
411 */
412static u_long const dumpmag = 0x8fca0101UL;
413
414static int dumpsize = 0; /* also for savecore */
415
416static int dodump = 1;
417
418SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
419 "Try to perform coredump on kernel panic");
420
421static int
422setdumpdev(dev)
423 dev_t dev;
424{
425 int psize;
426 long newdumplo;
427
428 if (dev == NODEV) {
429 dumpdev = dev;
430 return (0);
431 }
335dda38 432 psize = dev_dpsize(dev);
984263bc 433 if (psize == -1)
335dda38 434 return (ENXIO);
984263bc
MD
435 /*
436 * XXX should clean up checking in dumpsys() to be more like this.
437 */
438 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
439 if (newdumplo <= LABELSECTOR)
440 return (ENOSPC);
441 dumpdev = dev;
442 dumplo = newdumplo;
443 return (0);
444}
445
446
447/* ARGSUSED */
402ed7e1 448static void dump_conf (void *dummy);
984263bc
MD
449static void
450dump_conf(dummy)
451 void *dummy;
452{
453 char *path;
454 dev_t dev;
455
456 path = malloc(MNAMELEN, M_TEMP, M_WAITOK);
457 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
458 dev = getdiskbyname(path);
459 if (dev != NODEV)
460 dumpdev = dev;
461 }
462 free(path, M_TEMP);
463 if (setdumpdev(dumpdev) != 0)
464 dumpdev = NODEV;
465}
466
467SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
468
469static int
470sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
471{
472 int error;
473 udev_t ndumpdev;
474
475 ndumpdev = dev2udev(dumpdev);
476 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
477 if (error == 0 && req->newptr != NULL)
478 error = setdumpdev(udev2dev(ndumpdev, 0));
479 return (error);
480}
481
482SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
483 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
484
485/*
486 * Doadump comes here after turning off memory management and
487 * getting on the dump stack, either when called above, or by
488 * the auto-restart code.
489 */
490static void
491dumpsys(void)
492{
493 int error;
494
495 savectx(&dumppcb);
496 if (dumping++) {
497 printf("Dump already in progress, bailing...\n");
498 return;
499 }
500 if (!dodump)
501 return;
502 if (dumpdev == NODEV)
503 return;
984263bc
MD
504 dumpsize = Maxmem;
505 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
506 printf("dump ");
335dda38 507 error = dev_ddump(dumpdev);
984263bc
MD
508 if (error == 0) {
509 printf("succeeded\n");
510 return;
511 }
512 printf("failed, reason: ");
513 switch (error) {
335dda38 514 case ENOSYS:
984263bc
MD
515 case ENODEV:
516 printf("device doesn't support a dump routine\n");
517 break;
518
519 case ENXIO:
520 printf("device bad\n");
521 break;
522
523 case EFAULT:
524 printf("device not ready\n");
525 break;
526
527 case EINVAL:
528 printf("area improper\n");
529 break;
530
531 case EIO:
532 printf("i/o error\n");
533 break;
534
535 case EINTR:
536 printf("aborted from console\n");
537 break;
538
539 default:
540 printf("unknown, error = %d\n", error);
541 break;
542 }
543}
544
545int
546dumpstatus(vm_offset_t addr, off_t count)
547{
548 int c;
549
550 if (addr % (1024 * 1024) == 0) {
551#ifdef HW_WDOG
552 if (wdog_tickler)
553 (*wdog_tickler)();
554#endif
555 printf("%ld ", (long)(count / (1024 * 1024)));
556 }
557
558 if ((c = cncheckc()) == 0x03)
559 return -1;
560 else if (c != -1)
561 printf("[CTRL-C to abort] ");
562
563 return 0;
564}
565
566/*
567 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
568 * and then reboots. If we are called twice, then we avoid trying to sync
569 * the disks as this often leads to recursive panics.
570 */
571void
572panic(const char *fmt, ...)
573{
b1e04573 574 int bootopt, newpanic;
e2565a42 575 __va_list ap;
984263bc
MD
576 static char buf[256];
577
578 bootopt = RB_AUTOBOOT | RB_DUMP;
b1e04573 579 newpanic = 0;
984263bc
MD
580 if (panicstr)
581 bootopt |= RB_NOSYNC;
b1e04573 582 else {
984263bc 583 panicstr = fmt;
b1e04573
JH
584 newpanic = 1;
585 }
984263bc 586
e2565a42 587 __va_start(ap, fmt);
984263bc
MD
588 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
589 if (panicstr == fmt)
590 panicstr = buf;
e2565a42 591 __va_end(ap);
984263bc
MD
592 printf("panic: %s\n", buf);
593#ifdef SMP
594 /* three seperate prints in case of an unmapped page and trap */
595 printf("mp_lock = %08x; ", mp_lock);
72740893 596 printf("cpuid = %d; ", mycpu->gd_cpuid);
984263bc
MD
597 printf("lapic.id = %08x\n", lapic.id);
598#endif
599
600#if defined(DDB)
b1e04573
JH
601 if (newpanic && trace_on_panic)
602 db_print_backtrace();
984263bc
MD
603 if (debugger_on_panic)
604 Debugger ("panic");
605#endif
606 boot(bootopt);
607}
608
609/*
610 * Support for poweroff delay.
611 */
612#ifndef POWEROFF_DELAY
613# define POWEROFF_DELAY 5000
614#endif
615static int poweroff_delay = POWEROFF_DELAY;
616
617SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
618 &poweroff_delay, 0, "");
619
620static void
621poweroff_wait(void *junk, int howto)
622{
623 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
624 return;
625 DELAY(poweroff_delay * 1000);
626}
627
628/*
629 * Some system processes (e.g. syncer) need to be stopped at appropriate
630 * points in their main loops prior to a system shutdown, so that they
631 * won't interfere with the shutdown process (e.g. by holding a disk buf
632 * to cause sync to fail). For each of these system processes, register
633 * shutdown_kproc() as a handler for one of shutdown events.
634 */
635static int kproc_shutdown_wait = 60;
636SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
637 &kproc_shutdown_wait, 0, "");
638
639void
640shutdown_kproc(void *arg, int howto)
641{
bc6dffab 642 struct thread *td;
984263bc
MD
643 struct proc *p;
644 int error;
645
646 if (panicstr)
647 return;
648
bc6dffab
MD
649 td = (struct thread *)arg;
650 if ((p = td->td_proc) != NULL) {
651 printf("Waiting (max %d seconds) for system process `%s' to stop...",
652 kproc_shutdown_wait, p->p_comm);
653 } else {
0cfcada1
MD
654 printf("Waiting (max %d seconds) for system thread %s to stop...",
655 kproc_shutdown_wait, td->td_comm);
bc6dffab
MD
656 }
657 error = suspend_kproc(td, kproc_shutdown_wait * hz);
984263bc
MD
658
659 if (error == EWOULDBLOCK)
660 printf("timed out\n");
661 else
662 printf("stopped\n");
663}