DELAY() is a spin loop, we can't use it any more because shutdown
[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 $
3ad8cc28 40 * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.21 2005/07/18 02:47:36 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>
9ec81f83 67#include <sys/thread2.h>
3020e3be 68#include <sys/buf2.h>
984263bc
MD
69
70#include <machine/pcb.h>
71#include <machine/clock.h>
72#include <machine/md_var.h>
0f7a3396 73#include <machine/smp.h> /* smp_active_mask, cpuid */
984263bc
MD
74
75#include <sys/signalvar.h>
76
77#ifndef PANIC_REBOOT_WAIT_TIME
78#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
79#endif
80
81/*
82 * Note that stdarg.h and the ANSI style va_start macro is used for both
e2565a42
MD
83 * ANSI and traditional C compilers. We use the machine version to stay
84 * within the confines of the kernel header files.
984263bc
MD
85 */
86#include <machine/stdarg.h>
87
88#ifdef DDB
89#ifdef DDB_UNATTENDED
90int debugger_on_panic = 0;
91#else
92int debugger_on_panic = 1;
93#endif
94SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
95 &debugger_on_panic, 0, "Run debugger on kernel panic");
b1e04573
JH
96
97extern void db_print_backtrace(void);
98
99#ifdef DDB_TRACE
100int trace_on_panic = 1;
101#else
102int trace_on_panic = 0;
103#endif
104SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
105 &trace_on_panic, 0, "Print stack trace on kernel panic");
984263bc
MD
106#endif
107
baf88b6b
MD
108static int sync_on_panic = 1;
109SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
110 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
111
984263bc
MD
112SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
113
114#ifdef HW_WDOG
115/*
116 * If there is a hardware watchdog, point this at the function needed to
117 * hold it off.
118 * It's needed when the kernel needs to do some lengthy operations.
119 * e.g. in wd.c when dumping core.. It's most annoying to have
120 * your precious core-dump only half written because the wdog kicked in.
121 */
122watchdog_tickle_fn wdog_tickler = NULL;
123#endif /* HW_WDOG */
124
125/*
126 * Variable panicstr contains argument to first call to panic; used as flag
127 * to indicate that the kernel has already called panic.
128 */
129const char *panicstr;
130
131int dumping; /* system is dumping */
132
402ed7e1
RG
133static void boot (int) __dead2;
134static void dumpsys (void);
135static int setdumpdev (dev_t dev);
136static void poweroff_wait (void *, int);
137static void print_uptime (void);
138static void shutdown_halt (void *junk, int howto);
139static void shutdown_panic (void *junk, int howto);
140static void shutdown_reset (void *junk, int howto);
984263bc
MD
141
142/* register various local shutdown events */
143static void
144shutdown_conf(void *unused)
145{
146 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
147 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
148 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
149 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
150}
151
152SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
153
154/* ARGSUSED */
155
156/*
157 * The system call that results in a reboot
158 */
159int
41c20dac 160reboot(struct reboot_args *uap)
984263bc 161{
dadab5e9 162 struct thread *td = curthread;
984263bc
MD
163 int error;
164
dadab5e9 165 if ((error = suser(td)))
984263bc
MD
166 return (error);
167
168 boot(uap->opt);
169 return (0);
170}
171
172/*
173 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
174 */
175static int shutdown_howto = 0;
176
177void
178shutdown_nice(int howto)
179{
180 shutdown_howto = howto;
181
182 /* Send a signal to init(8) and have it shutdown the world */
183 if (initproc != NULL) {
184 psignal(initproc, SIGINT);
185 } else {
186 /* No init(8) running, so simply reboot */
187 boot(RB_NOSYNC);
188 }
189 return;
190}
191static int waittime = -1;
192static struct pcb dumppcb;
193
194static void
195print_uptime()
196{
197 int f;
198 struct timespec ts;
199
200 getnanouptime(&ts);
201 printf("Uptime: ");
202 f = 0;
203 if (ts.tv_sec >= 86400) {
204 printf("%ldd", ts.tv_sec / 86400);
205 ts.tv_sec %= 86400;
206 f = 1;
207 }
208 if (f || ts.tv_sec >= 3600) {
209 printf("%ldh", ts.tv_sec / 3600);
210 ts.tv_sec %= 3600;
211 f = 1;
212 }
213 if (f || ts.tv_sec >= 60) {
214 printf("%ldm", ts.tv_sec / 60);
215 ts.tv_sec %= 60;
216 f = 1;
217 }
218 printf("%lds\n", ts.tv_sec);
219}
220
221/*
222 * Go through the rigmarole of shutting down..
223 * this used to be in machdep.c but I'll be dammned if I could see
224 * anything machine dependant in it.
225 */
226static void
41c20dac 227boot(int howto)
984263bc 228{
3ad8cc28
MD
229 /*
230 * Get rid of any user scheduler baggage and then give
231 * us a high priority.
232 */
233 if (curthread->td_release)
234 curthread->td_release(curthread);
235 lwkt_setpri_self(TDPRI_MAX);
984263bc
MD
236
237 /* collect extra flags that shutdown_nice might have set */
238 howto |= shutdown_howto;
239
240#ifdef SMP
0f7a3396 241 if (smp_active_mask > 1) {
72740893 242 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
984263bc
MD
243 }
244#endif
245 /*
246 * Do any callouts that should be done BEFORE syncing the filesystems.
247 */
248 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
249
250 /*
251 * Now sync filesystems
252 */
253 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
1fd87d54 254 struct buf *bp;
984263bc
MD
255 int iter, nbusy, pbusy;
256
257 waittime = 0;
258 printf("\nsyncing disks... ");
259
41c20dac 260 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
984263bc
MD
261
262 /*
263 * With soft updates, some buffers that are
264 * written will be remarked as dirty until other
265 * buffers are written.
266 */
267 for (iter = pbusy = 0; iter < 20; iter++) {
268 nbusy = 0;
269 for (bp = &buf[nbuf]; --bp >= buf; ) {
270 if ((bp->b_flags & B_INVAL) == 0 &&
271 BUF_REFCNT(bp) > 0) {
272 nbusy++;
273 } else if ((bp->b_flags & (B_DELWRI | B_INVAL))
274 == B_DELWRI) {
275 /* bawrite(bp);*/
276 nbusy++;
277 }
278 }
279 if (nbusy == 0)
280 break;
281 printf("%d ", nbusy);
282 if (nbusy < pbusy)
283 iter = 0;
284 pbusy = nbusy;
e200c40d
DR
285 /*
286 * XXX:
287 * Process soft update work queue if buffers don't sync
288 * after 6 iterations by permitting the syncer to run.
289 */
984263bc
MD
290 if (iter > 5 && bioops.io_sync)
291 (*bioops.io_sync)(NULL);
41c20dac 292 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
3ad8cc28 293 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
984263bc
MD
294 }
295 printf("\n");
296 /*
297 * Count only busy local buffers to prevent forcing
298 * a fsck if we're just a client of a wedged NFS server
299 */
300 nbusy = 0;
301 for (bp = &buf[nbuf]; --bp >= buf; ) {
302 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
303 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
304 if (bp->b_dev == NODEV) {
861905fb 305 mountlist_remove(bp->b_vp->v_mount);
984263bc
MD
306 continue;
307 }
308 nbusy++;
309#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
310 printf(
311 "%p %d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n",
312 bp, nbusy, devtoname(bp->b_dev),
313 bp->b_flags, (long)bp->b_blkno,
314 (long)bp->b_lblkno);
315#endif
316 }
317 }
318 if (nbusy) {
319 /*
320 * Failed to sync all blocks. Indicate this and don't
321 * unmount filesystems (thus forcing an fsck on reboot).
322 */
323 printf("giving up on %d buffers\n", nbusy);
a88429e8
DR
324#ifdef DDB
325 Debugger("busy buffer problem");
326#endif /* DDB */
3ad8cc28 327 tsleep(boot, 0, "shutdn", hz * 5 + 1);
984263bc
MD
328 } else {
329 printf("done\n");
330 /*
331 * Unmount filesystems
332 */
333 if (panicstr == 0)
334 vfs_unmountall();
335 }
3ad8cc28 336 tsleep(boot, 0, "shutdn", hz / 10 + 1);
984263bc
MD
337 }
338
339 print_uptime();
340
341 /*
342 * Ok, now do things that assume all filesystem activity has
343 * been completed.
344 */
345 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
9ec81f83 346 crit_enter();
984263bc
MD
347 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
348 dumpsys();
349
350 /* Now that we're going to really halt the system... */
351 EVENTHANDLER_INVOKE(shutdown_final, howto);
352
353 for(;;) ; /* safety against shutdown_reset not working */
354 /* NOTREACHED */
355}
356
357/*
358 * If the shutdown was a clean halt, behave accordingly.
359 */
360static void
361shutdown_halt(void *junk, int howto)
362{
363 if (howto & RB_HALT) {
364 printf("\n");
365 printf("The operating system has halted.\n");
366 printf("Please press any key to reboot.\n\n");
367 switch (cngetc()) {
368 case -1: /* No console, just die */
369 cpu_halt();
370 /* NOTREACHED */
371 default:
372 howto &= ~RB_HALT;
373 break;
374 }
375 }
376}
377
378/*
379 * Check to see if the system paniced, pause and then reboot
380 * according to the specified delay.
381 */
382static void
383shutdown_panic(void *junk, int howto)
384{
385 int loop;
386
387 if (howto & RB_DUMP) {
388 if (PANIC_REBOOT_WAIT_TIME != 0) {
389 if (PANIC_REBOOT_WAIT_TIME != -1) {
390 printf("Automatic reboot in %d seconds - "
391 "press a key on the console to abort\n",
392 PANIC_REBOOT_WAIT_TIME);
393 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
394 loop > 0; --loop) {
395 DELAY(1000 * 100); /* 1/10th second */
396 /* Did user type a key? */
397 if (cncheckc() != -1)
398 break;
399 }
400 if (!loop)
401 return;
402 }
403 } else { /* zero time specified - reboot NOW */
404 return;
405 }
406 printf("--> Press a key on the console to reboot,\n");
407 printf("--> or switch off the system now.\n");
408 cngetc();
409 }
410}
411
412/*
413 * Everything done, now reset
414 */
415static void
416shutdown_reset(void *junk, int howto)
417{
418 printf("Rebooting...\n");
419 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
420 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
421 cpu_reset();
422 /* NOTREACHED */ /* assuming reset worked */
423}
424
425/*
426 * Magic number for savecore
427 *
428 * exported (symorder) and used at least by savecore(8)
429 *
430 */
431static u_long const dumpmag = 0x8fca0101UL;
432
433static int dumpsize = 0; /* also for savecore */
434
435static int dodump = 1;
436
437SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
438 "Try to perform coredump on kernel panic");
439
440static int
441setdumpdev(dev)
442 dev_t dev;
443{
444 int psize;
445 long newdumplo;
446
447 if (dev == NODEV) {
448 dumpdev = dev;
449 return (0);
450 }
335dda38 451 psize = dev_dpsize(dev);
984263bc 452 if (psize == -1)
335dda38 453 return (ENXIO);
984263bc
MD
454 /*
455 * XXX should clean up checking in dumpsys() to be more like this.
456 */
457 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
458 if (newdumplo <= LABELSECTOR)
459 return (ENOSPC);
460 dumpdev = dev;
461 dumplo = newdumplo;
462 return (0);
463}
464
465
466/* ARGSUSED */
402ed7e1 467static void dump_conf (void *dummy);
984263bc
MD
468static void
469dump_conf(dummy)
470 void *dummy;
471{
472 char *path;
473 dev_t dev;
474
475 path = malloc(MNAMELEN, M_TEMP, M_WAITOK);
476 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
477 dev = getdiskbyname(path);
478 if (dev != NODEV)
479 dumpdev = dev;
480 }
481 free(path, M_TEMP);
482 if (setdumpdev(dumpdev) != 0)
483 dumpdev = NODEV;
484}
485
486SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
487
488static int
489sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
490{
491 int error;
492 udev_t ndumpdev;
493
494 ndumpdev = dev2udev(dumpdev);
495 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
496 if (error == 0 && req->newptr != NULL)
497 error = setdumpdev(udev2dev(ndumpdev, 0));
498 return (error);
499}
500
501SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
502 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
503
504/*
505 * Doadump comes here after turning off memory management and
506 * getting on the dump stack, either when called above, or by
507 * the auto-restart code.
508 */
509static void
510dumpsys(void)
511{
512 int error;
513
514 savectx(&dumppcb);
515 if (dumping++) {
516 printf("Dump already in progress, bailing...\n");
517 return;
518 }
519 if (!dodump)
520 return;
521 if (dumpdev == NODEV)
522 return;
984263bc
MD
523 dumpsize = Maxmem;
524 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
525 printf("dump ");
335dda38 526 error = dev_ddump(dumpdev);
984263bc
MD
527 if (error == 0) {
528 printf("succeeded\n");
529 return;
530 }
531 printf("failed, reason: ");
532 switch (error) {
335dda38 533 case ENOSYS:
984263bc
MD
534 case ENODEV:
535 printf("device doesn't support a dump routine\n");
536 break;
537
538 case ENXIO:
539 printf("device bad\n");
540 break;
541
542 case EFAULT:
543 printf("device not ready\n");
544 break;
545
546 case EINVAL:
547 printf("area improper\n");
548 break;
549
550 case EIO:
551 printf("i/o error\n");
552 break;
553
554 case EINTR:
555 printf("aborted from console\n");
556 break;
557
558 default:
559 printf("unknown, error = %d\n", error);
560 break;
561 }
562}
563
564int
565dumpstatus(vm_offset_t addr, off_t count)
566{
567 int c;
568
569 if (addr % (1024 * 1024) == 0) {
570#ifdef HW_WDOG
571 if (wdog_tickler)
572 (*wdog_tickler)();
573#endif
574 printf("%ld ", (long)(count / (1024 * 1024)));
575 }
576
577 if ((c = cncheckc()) == 0x03)
578 return -1;
579 else if (c != -1)
580 printf("[CTRL-C to abort] ");
581
582 return 0;
583}
584
585/*
586 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
587 * and then reboots. If we are called twice, then we avoid trying to sync
588 * the disks as this often leads to recursive panics.
589 */
590void
591panic(const char *fmt, ...)
592{
b1e04573 593 int bootopt, newpanic;
e2565a42 594 __va_list ap;
984263bc
MD
595 static char buf[256];
596
597 bootopt = RB_AUTOBOOT | RB_DUMP;
baf88b6b
MD
598 if (sync_on_panic == 0)
599 bootopt |= RB_NOSYNC;
b1e04573 600 newpanic = 0;
984263bc
MD
601 if (panicstr)
602 bootopt |= RB_NOSYNC;
b1e04573 603 else {
984263bc 604 panicstr = fmt;
b1e04573
JH
605 newpanic = 1;
606 }
984263bc 607
e2565a42 608 __va_start(ap, fmt);
984263bc
MD
609 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
610 if (panicstr == fmt)
611 panicstr = buf;
e2565a42 612 __va_end(ap);
984263bc
MD
613 printf("panic: %s\n", buf);
614#ifdef SMP
7a44d1cb 615 /* three separate prints in case of an unmapped page and trap */
984263bc 616 printf("mp_lock = %08x; ", mp_lock);
72740893 617 printf("cpuid = %d; ", mycpu->gd_cpuid);
984263bc
MD
618 printf("lapic.id = %08x\n", lapic.id);
619#endif
620
621#if defined(DDB)
b1e04573
JH
622 if (newpanic && trace_on_panic)
623 db_print_backtrace();
984263bc
MD
624 if (debugger_on_panic)
625 Debugger ("panic");
626#endif
627 boot(bootopt);
628}
629
630/*
631 * Support for poweroff delay.
632 */
633#ifndef POWEROFF_DELAY
634# define POWEROFF_DELAY 5000
635#endif
636static int poweroff_delay = POWEROFF_DELAY;
637
638SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
639 &poweroff_delay, 0, "");
640
641static void
642poweroff_wait(void *junk, int howto)
643{
644 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
645 return;
646 DELAY(poweroff_delay * 1000);
647}
648
649/*
650 * Some system processes (e.g. syncer) need to be stopped at appropriate
651 * points in their main loops prior to a system shutdown, so that they
652 * won't interfere with the shutdown process (e.g. by holding a disk buf
653 * to cause sync to fail). For each of these system processes, register
654 * shutdown_kproc() as a handler for one of shutdown events.
655 */
656static int kproc_shutdown_wait = 60;
657SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
658 &kproc_shutdown_wait, 0, "");
659
660void
661shutdown_kproc(void *arg, int howto)
662{
bc6dffab 663 struct thread *td;
984263bc
MD
664 struct proc *p;
665 int error;
666
667 if (panicstr)
668 return;
669
bc6dffab
MD
670 td = (struct thread *)arg;
671 if ((p = td->td_proc) != NULL) {
672 printf("Waiting (max %d seconds) for system process `%s' to stop...",
673 kproc_shutdown_wait, p->p_comm);
674 } else {
0cfcada1
MD
675 printf("Waiting (max %d seconds) for system thread %s to stop...",
676 kproc_shutdown_wait, td->td_comm);
bc6dffab
MD
677 }
678 error = suspend_kproc(td, kproc_shutdown_wait * hz);
984263bc
MD
679
680 if (error == EWOULDBLOCK)
681 printf("timed out\n");
682 else
683 printf("stopped\n");
684}