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.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
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.
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
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 1.25 2006/02/21 17:36:38 dillon Exp $
44 #include "opt_ddb_trace.h"
45 #include "opt_hw_wdog.h"
46 #include "opt_panic.h"
47 #include "opt_show_busybufs.h"
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/eventhandler.h>
53 #include <sys/disklabel.h>
54 #include <sys/reboot.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>
64 #include <sys/sysproto.h>
65 #include <sys/device.h>
67 #include <sys/thread2.h>
70 #include <machine/pcb.h>
71 #include <machine/clock.h>
72 #include <machine/md_var.h>
73 #include <machine/smp.h> /* smp_active_mask, cpuid */
75 #include <sys/signalvar.h>
77 #ifndef PANIC_REBOOT_WAIT_TIME
78 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
82 * Note that stdarg.h and the ANSI style va_start macro is used for both
83 * ANSI and traditional C compilers. We use the machine version to stay
84 * within the confines of the kernel header files.
86 #include <machine/stdarg.h>
90 int debugger_on_panic = 0;
92 int debugger_on_panic = 1;
94 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
95 &debugger_on_panic, 0, "Run debugger on kernel panic");
97 extern void db_print_backtrace(void);
100 int trace_on_panic = 1;
102 int trace_on_panic = 0;
104 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
105 &trace_on_panic, 0, "Print stack trace on kernel panic");
108 static int sync_on_panic = 1;
109 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
110 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
112 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
116 * If there is a hardware watchdog, point this at the function needed to
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.
122 watchdog_tickle_fn wdog_tickler = NULL;
126 * Variable panicstr contains argument to first call to panic; used as flag
127 * to indicate that the kernel has already called panic.
129 const char *panicstr;
131 int dumping; /* system is dumping */
133 u_int panic_cpu_interlock; /* panic interlock */
134 globaldata_t panic_cpu_gd; /* which cpu took the panic */
137 static void boot (int) __dead2;
138 static void dumpsys (void);
139 static int setdumpdev (dev_t dev);
140 static void poweroff_wait (void *, int);
141 static void print_uptime (void);
142 static void shutdown_halt (void *junk, int howto);
143 static void shutdown_panic (void *junk, int howto);
144 static void shutdown_reset (void *junk, int howto);
146 /* register various local shutdown events */
148 shutdown_conf(void *unused)
150 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
151 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
152 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
153 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
156 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
161 * The system call that results in a reboot
164 reboot(struct reboot_args *uap)
166 struct thread *td = curthread;
169 if ((error = suser(td)))
177 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
179 static int shutdown_howto = 0;
182 shutdown_nice(int howto)
184 shutdown_howto = howto;
186 /* Send a signal to init(8) and have it shutdown the world */
187 if (initproc != NULL) {
188 psignal(initproc, SIGINT);
190 /* No init(8) running, so simply reboot */
195 static int waittime = -1;
196 static struct pcb dumppcb;
207 if (ts.tv_sec >= 86400) {
208 printf("%ldd", ts.tv_sec / 86400);
212 if (f || ts.tv_sec >= 3600) {
213 printf("%ldh", ts.tv_sec / 3600);
217 if (f || ts.tv_sec >= 60) {
218 printf("%ldm", ts.tv_sec / 60);
222 printf("%lds\n", ts.tv_sec);
226 * Go through the rigmarole of shutting down..
227 * this used to be in machdep.c but I'll be dammned if I could see
228 * anything machine dependant in it.
234 * Get rid of any user scheduler baggage and then give
235 * us a high priority.
237 if (curthread->td_release)
238 curthread->td_release(curthread);
239 lwkt_setpri_self(TDPRI_MAX);
241 /* collect extra flags that shutdown_nice might have set */
242 howto |= shutdown_howto;
246 * We really want to shutdown on the BSP. Subsystems such as ACPI
247 * can't power-down the box otherwise.
249 if (smp_active_mask > 1) {
250 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
252 if (panicstr == NULL && mycpu->gd_cpuid != 0) {
253 printf("Switching to cpu #0 for shutdown\n");
254 lwkt_setcpu_self(globaldata_find(0));
258 * Do any callouts that should be done BEFORE syncing the filesystems.
260 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
263 * Now sync filesystems
265 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
267 int iter, nbusy, pbusy;
270 printf("\nsyncing disks... ");
272 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
275 * With soft updates, some buffers that are
276 * written will be remarked as dirty until other
277 * buffers are written.
279 for (iter = pbusy = 0; iter < 20; iter++) {
281 for (bp = &buf[nbuf]; --bp >= buf; ) {
282 if ((bp->b_flags & B_INVAL) == 0 &&
283 BUF_REFCNT(bp) > 0) {
285 } else if ((bp->b_flags & (B_DELWRI | B_INVAL))
293 printf("%d ", nbusy);
299 * Process soft update work queue if buffers don't sync
300 * after 6 iterations by permitting the syncer to run.
302 if (iter > 5 && bioops.io_sync)
303 (*bioops.io_sync)(NULL);
304 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
305 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
309 * Count only busy local buffers to prevent forcing
310 * a fsck if we're just a client of a wedged NFS server
313 for (bp = &buf[nbuf]; --bp >= buf; ) {
314 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
315 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
317 * Only count buffers undergoing write I/O
318 * on the related vnode.
320 if (bp->b_vp == NULL ||
321 bp->b_vp->v_track_write.bk_active == 0) {
325 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
327 "%p %d: dev:?, flags:%08lx, blkno:%ld, lblkno:%ld\n",
329 bp->b_flags, (long)bp->b_bio1.bio_blkno,
336 * Failed to sync all blocks. Indicate this and don't
337 * unmount filesystems (thus forcing an fsck on reboot).
339 printf("giving up on %d buffers\n", nbusy);
341 Debugger("busy buffer problem");
343 tsleep(boot, 0, "shutdn", hz * 5 + 1);
347 * Unmount filesystems
352 tsleep(boot, 0, "shutdn", hz / 10 + 1);
358 * Ok, now do things that assume all filesystem activity has
361 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
363 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
366 /* Now that we're going to really halt the system... */
367 EVENTHANDLER_INVOKE(shutdown_final, howto);
369 for(;;) ; /* safety against shutdown_reset not working */
374 * If the shutdown was a clean halt, behave accordingly.
377 shutdown_halt(void *junk, int howto)
379 if (howto & RB_HALT) {
381 printf("The operating system has halted.\n");
382 printf("Please press any key to reboot.\n\n");
384 case -1: /* No console, just die */
395 * Check to see if the system paniced, pause and then reboot
396 * according to the specified delay.
399 shutdown_panic(void *junk, int howto)
403 if (howto & RB_DUMP) {
404 if (PANIC_REBOOT_WAIT_TIME != 0) {
405 if (PANIC_REBOOT_WAIT_TIME != -1) {
406 printf("Automatic reboot in %d seconds - "
407 "press a key on the console to abort\n",
408 PANIC_REBOOT_WAIT_TIME);
409 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
411 DELAY(1000 * 100); /* 1/10th second */
412 /* Did user type a key? */
413 if (cncheckc() != -1)
419 } else { /* zero time specified - reboot NOW */
422 printf("--> Press a key on the console to reboot,\n");
423 printf("--> or switch off the system now.\n");
429 * Everything done, now reset
432 shutdown_reset(void *junk, int howto)
434 printf("Rebooting...\n");
435 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
436 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
438 /* NOTREACHED */ /* assuming reset worked */
442 * Magic number for savecore
444 * exported (symorder) and used at least by savecore(8)
447 static u_long const dumpmag = 0x8fca0101UL;
449 static int dumpsize = 0; /* also for savecore */
451 static int dodump = 1;
453 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
454 "Try to perform coredump on kernel panic");
467 psize = dev_dpsize(dev);
471 * XXX should clean up checking in dumpsys() to be more like this.
473 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
474 if (newdumplo <= LABELSECTOR)
483 static void dump_conf (void *dummy);
491 path = malloc(MNAMELEN, M_TEMP, M_WAITOK);
492 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
493 dev = getdiskbyname(path);
498 if (setdumpdev(dumpdev) != 0)
502 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
505 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
510 ndumpdev = dev2udev(dumpdev);
511 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
512 if (error == 0 && req->newptr != NULL)
513 error = setdumpdev(udev2dev(ndumpdev, 0));
517 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
518 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
521 * Doadump comes here after turning off memory management and
522 * getting on the dump stack, either when called above, or by
523 * the auto-restart code.
532 printf("Dump already in progress, bailing...\n");
537 if (dumpdev == NODEV)
540 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
542 error = dev_ddump(dumpdev);
544 printf("succeeded\n");
547 printf("failed, reason: ");
551 printf("device doesn't support a dump routine\n");
555 printf("device bad\n");
559 printf("device not ready\n");
563 printf("area improper\n");
567 printf("i/o error\n");
571 printf("aborted from console\n");
575 printf("unknown, error = %d\n", error);
581 dumpstatus(vm_offset_t addr, off_t count)
585 if (addr % (1024 * 1024) == 0) {
590 printf("%ld ", (long)(count / (1024 * 1024)));
593 if ((c = cncheckc()) == 0x03)
596 printf("[CTRL-C to abort] ");
602 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
603 * and then reboots. If we are called twice, then we avoid trying to sync
604 * the disks as this often leads to recursive panics.
607 panic(const char *fmt, ...)
609 int bootopt, newpanic;
611 static char buf[256];
615 * If a panic occurs on multiple cpus before the first is able to
616 * halt the other cpus, only one cpu is allowed to take the panic.
617 * Attempt to be verbose about this situation but if the printf()
618 * itself panics don't let us overrun the kernel stack.
620 * Be very nasty about descheduling our thread at the lowest
621 * level possible in an attempt to freeze the thread without
622 * inducing further panics.
624 * Bumping gd_trap_nesting_level will also bypass assertions in
625 * lwkt_switch() and allow us to switch away even if we are a
626 * FAST interrupt or IPI.
628 if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
629 panic_cpu_gd = mycpu;
630 } else if (panic_cpu_gd != mycpu) {
632 ++mycpu->gd_trap_nesting_level;
633 if (mycpu->gd_trap_nesting_level < 25) {
634 printf("SECONDARY PANIC ON CPU %d THREAD %p\n",
635 mycpu->gd_cpuid, curthread);
637 curthread->td_release = NULL; /* be a grinch */
639 lwkt_deschedule_self(curthread);
643 /* --mycpu->gd_trap_nesting_level */
647 bootopt = RB_AUTOBOOT | RB_DUMP;
648 if (sync_on_panic == 0)
649 bootopt |= RB_NOSYNC;
652 bootopt |= RB_NOSYNC;
659 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
663 printf("panic: %s\n", buf);
665 /* three separate prints in case of an unmapped page and trap */
666 printf("mp_lock = %08x; ", mp_lock);
667 printf("cpuid = %d; ", mycpu->gd_cpuid);
668 printf("lapic.id = %08x\n", lapic.id);
672 if (newpanic && trace_on_panic)
673 db_print_backtrace();
674 if (debugger_on_panic)
681 * Support for poweroff delay.
683 #ifndef POWEROFF_DELAY
684 # define POWEROFF_DELAY 5000
686 static int poweroff_delay = POWEROFF_DELAY;
688 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
689 &poweroff_delay, 0, "");
692 poweroff_wait(void *junk, int howto)
694 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
696 DELAY(poweroff_delay * 1000);
700 * Some system processes (e.g. syncer) need to be stopped at appropriate
701 * points in their main loops prior to a system shutdown, so that they
702 * won't interfere with the shutdown process (e.g. by holding a disk buf
703 * to cause sync to fail). For each of these system processes, register
704 * shutdown_kproc() as a handler for one of shutdown events.
706 static int kproc_shutdown_wait = 60;
707 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
708 &kproc_shutdown_wait, 0, "");
711 shutdown_kproc(void *arg, int howto)
720 td = (struct thread *)arg;
721 if ((p = td->td_proc) != NULL) {
722 printf("Waiting (max %d seconds) for system process `%s' to stop...",
723 kproc_shutdown_wait, p->p_comm);
725 printf("Waiting (max %d seconds) for system thread %s to stop...",
726 kproc_shutdown_wait, td->td_comm);
728 error = suspend_kproc(td, kproc_shutdown_wait * hz);
730 if (error == EWOULDBLOCK)
731 printf("timed out\n");