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 1.30 2006/06/05 07:26:10 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);
145 static int shutdown_busycount1(struct buf *bp, void *info);
146 static int shutdown_busycount2(struct buf *bp, void *info);
148 /* register various local shutdown events */
150 shutdown_conf(void *unused)
152 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
153 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
154 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
155 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
158 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
163 * The system call that results in a reboot
166 sys_reboot(struct reboot_args *uap)
168 struct thread *td = curthread;
171 if ((error = suser(td)))
179 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
181 static int shutdown_howto = 0;
184 shutdown_nice(int howto)
186 shutdown_howto = howto;
188 /* Send a signal to init(8) and have it shutdown the world */
189 if (initproc != NULL) {
190 psignal(initproc, SIGINT);
192 /* No init(8) running, so simply reboot */
197 static int waittime = -1;
198 static struct thread *dumpthread;
199 static struct pcb dumppcb;
210 if (ts.tv_sec >= 86400) {
211 printf("%ldd", ts.tv_sec / 86400);
215 if (f || ts.tv_sec >= 3600) {
216 printf("%ldh", ts.tv_sec / 3600);
220 if (f || ts.tv_sec >= 60) {
221 printf("%ldm", ts.tv_sec / 60);
225 printf("%lds\n", ts.tv_sec);
229 * Go through the rigmarole of shutting down..
230 * this used to be in machdep.c but I'll be dammned if I could see
231 * anything machine dependant in it.
237 * Get rid of any user scheduler baggage and then give
238 * us a high priority.
240 if (curthread->td_release)
241 curthread->td_release(curthread);
242 lwkt_setpri_self(TDPRI_MAX);
244 /* collect extra flags that shutdown_nice might have set */
245 howto |= shutdown_howto;
249 * We really want to shutdown on the BSP. Subsystems such as ACPI
250 * can't power-down the box otherwise.
252 if (smp_active_mask > 1) {
253 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
255 if (panicstr == NULL && mycpu->gd_cpuid != 0) {
256 printf("Switching to cpu #0 for shutdown\n");
257 lwkt_setcpu_self(globaldata_find(0));
261 * Do any callouts that should be done BEFORE syncing the filesystems.
263 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
266 * Now sync filesystems
268 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
269 int iter, nbusy, pbusy;
272 printf("\nsyncing disks... ");
274 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
277 * With soft updates, some buffers that are
278 * written will be remarked as dirty until other
279 * buffers are written.
281 for (iter = pbusy = 0; iter < 20; iter++) {
282 nbusy = scan_all_buffers(shutdown_busycount1, NULL);
285 printf("%d ", nbusy);
291 * Process soft update work queue if buffers don't sync
292 * after 6 iterations by permitting the syncer to run.
294 if (iter > 5 && bioops.io_sync)
295 (*bioops.io_sync)(NULL);
296 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
297 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
301 * Count only busy local buffers to prevent forcing
302 * a fsck if we're just a client of a wedged NFS server
304 nbusy = scan_all_buffers(shutdown_busycount2, NULL);
307 * Failed to sync all blocks. Indicate this and don't
308 * unmount filesystems (thus forcing an fsck on reboot).
310 printf("giving up on %d buffers\n", nbusy);
312 Debugger("busy buffer problem");
314 tsleep(boot, 0, "shutdn", hz * 5 + 1);
318 * Unmount filesystems
323 tsleep(boot, 0, "shutdn", hz / 10 + 1);
329 * Ok, now do things that assume all filesystem activity has
332 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
334 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
337 /* Now that we're going to really halt the system... */
338 EVENTHANDLER_INVOKE(shutdown_final, howto);
340 for(;;) ; /* safety against shutdown_reset not working */
345 shutdown_busycount1(struct buf *bp, void *info)
347 if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
349 if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
355 shutdown_busycount2(struct buf *bp, void *info)
357 if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
358 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
360 * Only count buffers undergoing write I/O
361 * on the related vnode.
363 if (bp->b_vp == NULL ||
364 bp->b_vp->v_track_write.bk_active == 0) {
367 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
369 "%p dev:?, flags:%08lx, loffset:%lld, doffset:%lld\n",
371 bp->b_flags, bp->b_loffset,
372 bp->b_bio2.bio_offset);
380 * If the shutdown was a clean halt, behave accordingly.
383 shutdown_halt(void *junk, int howto)
385 if (howto & RB_HALT) {
387 printf("The operating system has halted.\n");
388 printf("Please press any key to reboot.\n\n");
390 case -1: /* No console, just die */
401 * Check to see if the system paniced, pause and then reboot
402 * according to the specified delay.
405 shutdown_panic(void *junk, int howto)
409 if (howto & RB_DUMP) {
410 if (PANIC_REBOOT_WAIT_TIME != 0) {
411 if (PANIC_REBOOT_WAIT_TIME != -1) {
412 printf("Automatic reboot in %d seconds - "
413 "press a key on the console to abort\n",
414 PANIC_REBOOT_WAIT_TIME);
415 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
417 DELAY(1000 * 100); /* 1/10th second */
418 /* Did user type a key? */
419 if (cncheckc() != -1)
425 } else { /* zero time specified - reboot NOW */
428 printf("--> Press a key on the console to reboot,\n");
429 printf("--> or switch off the system now.\n");
435 * Everything done, now reset
438 shutdown_reset(void *junk, int howto)
440 printf("Rebooting...\n");
441 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
442 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
444 /* NOTREACHED */ /* assuming reset worked */
448 * Magic number for savecore
450 * exported (symorder) and used at least by savecore(8)
453 static u_long const dumpmag = 0x8fca0101UL;
455 static int dumpsize = 0; /* also for savecore */
457 static int dodump = 1;
459 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
460 "Try to perform coredump on kernel panic");
473 psize = dev_dpsize(dev);
477 * XXX should clean up checking in dumpsys() to be more like this.
479 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
480 if (newdumplo <= LABELSECTOR)
489 static void dump_conf (void *dummy);
497 path = malloc(MNAMELEN, M_TEMP, M_WAITOK);
498 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
499 dev = getdiskbyname(path);
504 if (setdumpdev(dumpdev) != 0)
508 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
511 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
516 ndumpdev = dev2udev(dumpdev);
517 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
518 if (error == 0 && req->newptr != NULL)
519 error = setdumpdev(udev2dev(ndumpdev, 0));
523 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
524 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", "");
527 * Doadump comes here after turning off memory management and
528 * getting on the dump stack, either when called above, or by
529 * the auto-restart code.
537 dumpthread = curthread;
539 printf("Dump already in progress, bailing...\n");
544 if (dumpdev == NODEV)
547 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
549 error = dev_ddump(dumpdev);
551 printf("succeeded\n");
554 printf("failed, reason: ");
558 printf("device doesn't support a dump routine\n");
562 printf("device bad\n");
566 printf("device not ready\n");
570 printf("area improper\n");
574 printf("i/o error\n");
578 printf("aborted from console\n");
582 printf("unknown, error = %d\n", error);
588 dumpstatus(vm_offset_t addr, off_t count)
592 if (addr % (1024 * 1024) == 0) {
597 printf("%ld ", (long)(count / (1024 * 1024)));
600 if ((c = cncheckc()) == 0x03)
603 printf("[CTRL-C to abort] ");
609 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
610 * and then reboots. If we are called twice, then we avoid trying to sync
611 * the disks as this often leads to recursive panics.
614 panic(const char *fmt, ...)
616 int bootopt, newpanic;
618 static char buf[256];
622 * If a panic occurs on multiple cpus before the first is able to
623 * halt the other cpus, only one cpu is allowed to take the panic.
624 * Attempt to be verbose about this situation but if the printf()
625 * itself panics don't let us overrun the kernel stack.
627 * Be very nasty about descheduling our thread at the lowest
628 * level possible in an attempt to freeze the thread without
629 * inducing further panics.
631 * Bumping gd_trap_nesting_level will also bypass assertions in
632 * lwkt_switch() and allow us to switch away even if we are a
633 * FAST interrupt or IPI.
635 if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
636 panic_cpu_gd = mycpu;
637 } else if (panic_cpu_gd != mycpu) {
639 ++mycpu->gd_trap_nesting_level;
640 if (mycpu->gd_trap_nesting_level < 25) {
641 printf("SECONDARY PANIC ON CPU %d THREAD %p\n",
642 mycpu->gd_cpuid, curthread);
644 curthread->td_release = NULL; /* be a grinch */
646 lwkt_deschedule_self(curthread);
650 /* --mycpu->gd_trap_nesting_level */
654 bootopt = RB_AUTOBOOT | RB_DUMP;
655 if (sync_on_panic == 0)
656 bootopt |= RB_NOSYNC;
659 bootopt |= RB_NOSYNC;
666 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
670 printf("panic: %s\n", buf);
672 /* three separate prints in case of an unmapped page and trap */
673 printf("mp_lock = %08x; ", mp_lock);
674 printf("cpuid = %d; ", mycpu->gd_cpuid);
675 printf("lapic.id = %08x\n", lapic.id);
679 if (newpanic && trace_on_panic)
680 db_print_backtrace();
681 if (debugger_on_panic)
688 * Support for poweroff delay.
690 #ifndef POWEROFF_DELAY
691 # define POWEROFF_DELAY 5000
693 static int poweroff_delay = POWEROFF_DELAY;
695 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
696 &poweroff_delay, 0, "");
699 poweroff_wait(void *junk, int howto)
701 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
703 DELAY(poweroff_delay * 1000);
707 * Some system processes (e.g. syncer) need to be stopped at appropriate
708 * points in their main loops prior to a system shutdown, so that they
709 * won't interfere with the shutdown process (e.g. by holding a disk buf
710 * to cause sync to fail). For each of these system processes, register
711 * shutdown_kproc() as a handler for one of shutdown events.
713 static int kproc_shutdown_wait = 60;
714 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
715 &kproc_shutdown_wait, 0, "");
718 shutdown_kproc(void *arg, int howto)
727 td = (struct thread *)arg;
728 if ((p = td->td_proc) != NULL) {
729 printf("Waiting (max %d seconds) for system process `%s' to stop...",
730 kproc_shutdown_wait, p->p_comm);
732 printf("Waiting (max %d seconds) for system thread %s to stop...",
733 kproc_shutdown_wait, td->td_comm);
735 error = suspend_kproc(td, kproc_shutdown_wait * hz);
737 if (error == EWOULDBLOCK)
738 printf("timed out\n");