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 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)kern_shutdown.c 8.3 (Berkeley) 1/21/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
42 #include "opt_panic.h"
43 #include "opt_sched.h"
44 #include "opt_watchdog.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
52 #include <sys/eventhandler.h>
55 #include <sys/kernel.h>
56 #include <sys/kerneldump.h>
57 #include <sys/kthread.h>
58 #include <sys/malloc.h>
59 #include <sys/mount.h>
62 #include <sys/reboot.h>
63 #include <sys/resourcevar.h>
64 #include <sys/sched.h>
66 #include <sys/sysctl.h>
67 #include <sys/sysproto.h>
68 #include <sys/vnode.h>
70 #include <sys/watchdog.h>
75 #include <machine/cpu.h>
76 #include <machine/pcb.h>
77 #include <machine/smp.h>
79 #include <security/mac/mac_framework.h>
82 #include <vm/vm_object.h>
83 #include <vm/vm_page.h>
84 #include <vm/vm_pager.h>
85 #include <vm/swap_pager.h>
87 #include <sys/signalvar.h>
89 #ifndef PANIC_REBOOT_WAIT_TIME
90 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
94 * Note that stdarg.h and the ANSI style va_start macro is used for both
95 * ANSI and traditional C compilers.
97 #include <machine/stdarg.h>
100 #ifdef KDB_UNATTENDED
101 int debugger_on_panic = 0;
103 int debugger_on_panic = 1;
105 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
106 &debugger_on_panic, 0, "Run debugger on kernel panic");
107 TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
110 static int trace_on_panic = 1;
112 static int trace_on_panic = 0;
114 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
115 &trace_on_panic, 0, "Print stack trace on kernel panic");
116 TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
119 static int sync_on_panic = 0;
120 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
121 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
122 TUNABLE_INT("kern.sync_on_panic", &sync_on_panic);
124 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
127 static int show_busybufs;
129 static int show_busybufs = 1;
131 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
132 &show_busybufs, 0, "");
135 * Variable panicstr contains argument to first call to panic; used as flag
136 * to indicate that the kernel has already called panic.
138 const char *panicstr;
140 int dumping; /* system is dumping */
141 int rebooting; /* system is rebooting */
142 static struct dumperinfo dumper; /* our selected dumper */
144 /* Context information for dump-debuggers. */
145 static struct pcb dumppcb; /* Registers. */
146 static lwpid_t dumptid; /* Thread ID. */
148 static void poweroff_wait(void *, int);
149 static void shutdown_halt(void *junk, int howto);
150 static void shutdown_panic(void *junk, int howto);
151 static void shutdown_reset(void *junk, int howto);
153 /* register various local shutdown events */
155 shutdown_conf(void *unused)
158 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
160 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
161 SHUTDOWN_PRI_LAST + 100);
162 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
163 SHUTDOWN_PRI_LAST + 100);
164 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
165 SHUTDOWN_PRI_LAST + 200);
168 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
171 * The system call that results in a reboot.
175 sys_reboot(struct thread *td, struct reboot_args *uap)
181 error = mac_system_check_reboot(td->td_ucred, uap->opt);
184 error = priv_check(td, PRIV_REBOOT);
187 kern_reboot(uap->opt);
194 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
196 static int shutdown_howto = 0;
199 shutdown_nice(int howto)
202 shutdown_howto = howto;
204 /* Send a signal to init(8) and have it shutdown the world */
205 if (initproc != NULL) {
207 kern_psignal(initproc, SIGINT);
208 PROC_UNLOCK(initproc);
210 /* No init(8) running, so simply reboot */
211 kern_reboot(RB_NOSYNC);
215 static int waittime = -1;
226 if (ts.tv_sec >= 86400) {
227 printf("%ldd", (long)ts.tv_sec / 86400);
231 if (f || ts.tv_sec >= 3600) {
232 printf("%ldh", (long)ts.tv_sec / 3600);
236 if (f || ts.tv_sec >= 60) {
237 printf("%ldm", (long)ts.tv_sec / 60);
241 printf("%lds\n", (long)ts.tv_sec);
245 doadump(boolean_t textdump)
251 if (dumper.dumper == NULL)
255 dumptid = curthread->td_tid;
260 if (textdump && textdump_pending) {
262 textdump_dumpsys(&dumper);
273 isbufbusy(struct buf *bp)
275 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
277 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
283 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
286 kern_reboot(int howto)
288 static int first_buf_printf = 1;
292 * Bind us to CPU 0 so that all shutdown code runs there. Some
293 * systems don't shutdown properly (i.e., ACPI power off) if we
294 * run on another processor.
296 thread_lock(curthread);
297 sched_bind(curthread, 0);
298 thread_unlock(curthread);
299 KASSERT(PCPU_GET(cpuid) == 0, ("%s: not running on cpu 0", __func__));
301 /* We're in the process of rebooting. */
304 /* collect extra flags that shutdown_nice might have set */
305 howto |= shutdown_howto;
307 /* We are out of the debugger now. */
311 * Do any callouts that should be done BEFORE syncing the filesystems.
313 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
316 * Now sync filesystems
318 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
319 register struct buf *bp;
320 int iter, nbusy, pbusy;
328 wdog_kern_pat(WD_LASTVAL);
330 sys_sync(curthread, NULL);
333 * With soft updates, some buffers that are
334 * written will be remarked as dirty until other
335 * buffers are written.
337 for (iter = pbusy = 0; iter < 20; iter++) {
339 for (bp = &buf[nbuf]; --bp >= buf; )
343 if (first_buf_printf)
344 printf("All buffers synced.");
347 if (first_buf_printf) {
348 printf("Syncing disks, buffers remaining... ");
349 first_buf_printf = 0;
351 printf("%d ", nbusy);
356 wdog_kern_pat(WD_LASTVAL);
358 sys_sync(curthread, NULL);
362 * Drop Giant and spin for a while to allow
363 * interrupt threads to run.
370 * Drop Giant and context switch several times to
371 * allow interrupt threads to run.
374 for (subiter = 0; subiter < 50 * iter; subiter++) {
375 thread_lock(curthread);
376 mi_switch(SW_VOL, NULL);
377 thread_unlock(curthread);
385 * Count only busy local buffers to prevent forcing
386 * a fsck if we're just a client of a wedged NFS server
389 for (bp = &buf[nbuf]; --bp >= buf; ) {
392 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
393 if (bp->b_dev == NULL) {
394 TAILQ_REMOVE(&mountlist,
395 bp->b_vp->v_mount, mnt_list);
400 if (show_busybufs > 0) {
402 "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
403 nbusy, bp, bp->b_vp, bp->b_flags,
404 (intmax_t)bp->b_blkno,
405 (intmax_t)bp->b_lblkno);
406 BUF_LOCKPRINTINFO(bp);
407 if (show_busybufs > 1)
415 * Failed to sync all blocks. Indicate this and don't
416 * unmount filesystems (thus forcing an fsck on reboot).
418 printf("Giving up on %d buffers\n", nbusy);
419 DELAY(5000000); /* 5 seconds */
421 if (!first_buf_printf)
422 printf("Final sync complete\n");
424 * Unmount filesystems
430 DELAY(100000); /* wait for console output to finish */
436 * Ok, now do things that assume all filesystem activity has
439 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
441 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
444 /* Now that we're going to really halt the system... */
445 EVENTHANDLER_INVOKE(shutdown_final, howto);
447 for(;;) ; /* safety against shutdown_reset not working */
452 * If the shutdown was a clean halt, behave accordingly.
455 shutdown_halt(void *junk, int howto)
458 if (howto & RB_HALT) {
460 printf("The operating system has halted.\n");
461 printf("Please press any key to reboot.\n\n");
463 case -1: /* No console, just die */
474 * Check to see if the system paniced, pause and then reboot
475 * according to the specified delay.
478 shutdown_panic(void *junk, int howto)
482 if (howto & RB_DUMP) {
483 if (PANIC_REBOOT_WAIT_TIME != 0) {
484 if (PANIC_REBOOT_WAIT_TIME != -1) {
485 printf("Automatic reboot in %d seconds - "
486 "press a key on the console to abort\n",
487 PANIC_REBOOT_WAIT_TIME);
488 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
490 DELAY(1000 * 100); /* 1/10th second */
491 /* Did user type a key? */
492 if (cncheckc() != -1)
498 } else { /* zero time specified - reboot NOW */
501 printf("--> Press a key on the console to reboot,\n");
502 printf("--> or switch off the system now.\n");
508 * Everything done, now reset
511 shutdown_reset(void *junk, int howto)
514 printf("Rebooting...\n");
515 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
518 * Acquiring smp_ipi_mtx here has a double effect:
519 * - it disables interrupts avoiding CPU0 preemption
520 * by fast handlers (thus deadlocking against other CPUs)
521 * - it avoids deadlocks against smp_rendezvous() or, more
522 * generally, threads busy-waiting, with this spinlock held,
523 * and waiting for responses by threads on other CPUs
524 * (ie. smp_tlb_shootdown()).
526 * For the !SMP case it just needs to handle the former problem.
529 mtx_lock_spin(&smp_ipi_mtx);
534 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
536 /* NOTREACHED */ /* assuming reset worked */
540 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
541 * and then reboots. If we are called twice, then we avoid trying to sync
542 * the disks as this often leads to recursive panics.
545 panic(const char *fmt, ...)
548 static volatile u_int panic_cpu = NOCPU;
550 struct thread *td = curthread;
551 int bootopt, newpanic;
553 static char buf[256];
558 * We don't want multiple CPU's to panic at the same time, so we
559 * use panic_cpu as a simple spinlock. We have to keep checking
560 * panic_cpu if we are spinning in case the panic on the first
563 if (panic_cpu != PCPU_GET(cpuid))
564 while (atomic_cmpset_int(&panic_cpu, NOCPU,
565 PCPU_GET(cpuid)) == 0)
566 while (panic_cpu != NOCPU)
570 bootopt = RB_AUTOBOOT;
573 bootopt |= RB_NOSYNC;
582 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
584 printf("panic: %s\n", buf);
592 printf("cpuid = %d\n", PCPU_GET(cpuid));
596 if (newpanic && trace_on_panic)
598 if (debugger_on_panic)
599 kdb_enter(KDB_WHY_PANIC, "panic");
601 /*thread_lock(td); */
602 td->td_flags |= TDF_INPANIC;
603 /* thread_unlock(td); */
605 bootopt |= RB_NOSYNC;
607 kern_reboot(bootopt);
611 * Support for poweroff delay.
613 * Please note that setting this delay too short might power off your machine
614 * before the write cache on your hard disk has been flushed, leading to
615 * soft-updates inconsistencies.
617 #ifndef POWEROFF_DELAY
618 # define POWEROFF_DELAY 5000
620 static int poweroff_delay = POWEROFF_DELAY;
622 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
623 &poweroff_delay, 0, "");
626 poweroff_wait(void *junk, int howto)
629 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
631 DELAY(poweroff_delay * 1000);
635 * Some system processes (e.g. syncer) need to be stopped at appropriate
636 * points in their main loops prior to a system shutdown, so that they
637 * won't interfere with the shutdown process (e.g. by holding a disk buf
638 * to cause sync to fail). For each of these system processes, register
639 * shutdown_kproc() as a handler for one of shutdown events.
641 static int kproc_shutdown_wait = 60;
642 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
643 &kproc_shutdown_wait, 0, "");
646 kproc_shutdown(void *arg, int howto)
654 p = (struct proc *)arg;
655 printf("Waiting (max %d seconds) for system process `%s' to stop...",
656 kproc_shutdown_wait, p->p_comm);
657 error = kproc_suspend(p, kproc_shutdown_wait * hz);
659 if (error == EWOULDBLOCK)
660 printf("timed out\n");
666 kthread_shutdown(void *arg, int howto)
674 td = (struct thread *)arg;
675 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
676 kproc_shutdown_wait, td->td_name);
677 error = kthread_suspend(td, kproc_shutdown_wait * hz);
679 if (error == EWOULDBLOCK)
680 printf("timed out\n");
685 /* Registration of dumpers */
687 set_dumper(struct dumperinfo *di)
691 bzero(&dumper, sizeof dumper);
694 if (dumper.dumper != NULL)
700 /* Call dumper with bounds checking. */
702 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
703 off_t offset, size_t length)
706 if (length != 0 && (offset < di->mediaoffset ||
707 offset - di->mediaoffset + length > di->mediasize)) {
708 printf("Attempt to write outside dump device boundaries.\n"
709 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
710 (intmax_t)offset, (intmax_t)di->mediaoffset,
711 (uintmax_t)length, (intmax_t)di->mediasize);
714 return (di->dumper(di->priv, virtual, physical, offset, length));
718 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
719 uint64_t dumplen, uint32_t blksz)
722 bzero(kdh, sizeof(*kdh));
723 strncpy(kdh->magic, magic, sizeof(kdh->magic));
724 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
725 kdh->version = htod32(KERNELDUMPVERSION);
726 kdh->architectureversion = htod32(archver);
727 kdh->dumplength = htod64(dumplen);
728 kdh->dumptime = htod64(time_second);
729 kdh->blocksize = htod32(blksz);
730 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
731 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
732 if (panicstr != NULL)
733 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
734 kdh->parity = kerneldump_parity(kdh);