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 static int stop_scheduler_on_panic = 1;
125 SYSCTL_INT(_kern, OID_AUTO, stop_scheduler_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
126 &stop_scheduler_on_panic, 0, "stop scheduler upon entering panic");
127 TUNABLE_INT("kern.stop_scheduler_on_panic", &stop_scheduler_on_panic);
129 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
132 static int show_busybufs;
134 static int show_busybufs = 1;
136 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
137 &show_busybufs, 0, "");
140 * Variable panicstr contains argument to first call to panic; used as flag
141 * to indicate that the kernel has already called panic.
143 const char *panicstr;
145 int stop_scheduler; /* system stopped CPUs for panic */
146 int dumping; /* system is dumping */
147 int rebooting; /* system is rebooting */
148 static struct dumperinfo dumper; /* our selected dumper */
150 /* Context information for dump-debuggers. */
151 static struct pcb dumppcb; /* Registers. */
152 static lwpid_t dumptid; /* Thread ID. */
154 static void poweroff_wait(void *, int);
155 static void shutdown_halt(void *junk, int howto);
156 static void shutdown_panic(void *junk, int howto);
157 static void shutdown_reset(void *junk, int howto);
159 /* register various local shutdown events */
161 shutdown_conf(void *unused)
164 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
166 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
167 SHUTDOWN_PRI_LAST + 100);
168 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
169 SHUTDOWN_PRI_LAST + 100);
170 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
171 SHUTDOWN_PRI_LAST + 200);
174 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
177 * The system call that results in a reboot.
181 sys_reboot(struct thread *td, struct reboot_args *uap)
187 error = mac_system_check_reboot(td->td_ucred, uap->opt);
190 error = priv_check(td, PRIV_REBOOT);
193 kern_reboot(uap->opt);
200 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
202 static int shutdown_howto = 0;
205 shutdown_nice(int howto)
208 shutdown_howto = howto;
210 /* Send a signal to init(8) and have it shutdown the world */
211 if (initproc != NULL) {
213 kern_psignal(initproc, SIGINT);
214 PROC_UNLOCK(initproc);
216 /* No init(8) running, so simply reboot */
217 kern_reboot(RB_NOSYNC);
221 static int waittime = -1;
232 if (ts.tv_sec >= 86400) {
233 printf("%ldd", (long)ts.tv_sec / 86400);
237 if (f || ts.tv_sec >= 3600) {
238 printf("%ldh", (long)ts.tv_sec / 3600);
242 if (f || ts.tv_sec >= 60) {
243 printf("%ldm", (long)ts.tv_sec / 60);
247 printf("%lds\n", (long)ts.tv_sec);
251 doadump(boolean_t textdump)
257 if (dumper.dumper == NULL)
261 dumptid = curthread->td_tid;
266 if (textdump && textdump_pending) {
268 textdump_dumpsys(&dumper);
279 isbufbusy(struct buf *bp)
281 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
283 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
289 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
292 kern_reboot(int howto)
294 static int first_buf_printf = 1;
298 * Bind us to CPU 0 so that all shutdown code runs there. Some
299 * systems don't shutdown properly (i.e., ACPI power off) if we
300 * run on another processor.
302 if (!SCHEDULER_STOPPED()) {
303 thread_lock(curthread);
304 sched_bind(curthread, 0);
305 thread_unlock(curthread);
306 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
309 /* We're in the process of rebooting. */
312 /* collect extra flags that shutdown_nice might have set */
313 howto |= shutdown_howto;
315 /* We are out of the debugger now. */
319 * Do any callouts that should be done BEFORE syncing the filesystems.
321 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
324 * Now sync filesystems
326 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
327 register struct buf *bp;
328 int iter, nbusy, pbusy;
336 wdog_kern_pat(WD_LASTVAL);
338 sys_sync(curthread, NULL);
341 * With soft updates, some buffers that are
342 * written will be remarked as dirty until other
343 * buffers are written.
345 for (iter = pbusy = 0; iter < 20; iter++) {
347 for (bp = &buf[nbuf]; --bp >= buf; )
351 if (first_buf_printf)
352 printf("All buffers synced.");
355 if (first_buf_printf) {
356 printf("Syncing disks, buffers remaining... ");
357 first_buf_printf = 0;
359 printf("%d ", nbusy);
364 wdog_kern_pat(WD_LASTVAL);
366 sys_sync(curthread, NULL);
370 * Drop Giant and spin for a while to allow
371 * interrupt threads to run.
378 * Drop Giant and context switch several times to
379 * allow interrupt threads to run.
382 for (subiter = 0; subiter < 50 * iter; subiter++) {
383 thread_lock(curthread);
384 mi_switch(SW_VOL, NULL);
385 thread_unlock(curthread);
393 * Count only busy local buffers to prevent forcing
394 * a fsck if we're just a client of a wedged NFS server
397 for (bp = &buf[nbuf]; --bp >= buf; ) {
400 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
401 if (bp->b_dev == NULL) {
402 TAILQ_REMOVE(&mountlist,
403 bp->b_vp->v_mount, mnt_list);
408 if (show_busybufs > 0) {
410 "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
411 nbusy, bp, bp->b_vp, bp->b_flags,
412 (intmax_t)bp->b_blkno,
413 (intmax_t)bp->b_lblkno);
414 BUF_LOCKPRINTINFO(bp);
415 if (show_busybufs > 1)
423 * Failed to sync all blocks. Indicate this and don't
424 * unmount filesystems (thus forcing an fsck on reboot).
426 printf("Giving up on %d buffers\n", nbusy);
427 DELAY(5000000); /* 5 seconds */
429 if (!first_buf_printf)
430 printf("Final sync complete\n");
432 * Unmount filesystems
438 DELAY(100000); /* wait for console output to finish */
446 * Ok, now do things that assume all filesystem activity has
449 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
451 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
454 /* Now that we're going to really halt the system... */
455 EVENTHANDLER_INVOKE(shutdown_final, howto);
457 for(;;) ; /* safety against shutdown_reset not working */
462 * If the shutdown was a clean halt, behave accordingly.
465 shutdown_halt(void *junk, int howto)
468 if (howto & RB_HALT) {
470 printf("The operating system has halted.\n");
471 printf("Please press any key to reboot.\n\n");
473 case -1: /* No console, just die */
484 * Check to see if the system paniced, pause and then reboot
485 * according to the specified delay.
488 shutdown_panic(void *junk, int howto)
492 if (howto & RB_DUMP) {
493 if (PANIC_REBOOT_WAIT_TIME != 0) {
494 if (PANIC_REBOOT_WAIT_TIME != -1) {
495 printf("Automatic reboot in %d seconds - "
496 "press a key on the console to abort\n",
497 PANIC_REBOOT_WAIT_TIME);
498 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
500 DELAY(1000 * 100); /* 1/10th second */
501 /* Did user type a key? */
502 if (cncheckc() != -1)
508 } else { /* zero time specified - reboot NOW */
511 printf("--> Press a key on the console to reboot,\n");
512 printf("--> or switch off the system now.\n");
518 * Everything done, now reset
521 shutdown_reset(void *junk, int howto)
524 printf("Rebooting...\n");
525 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
528 * Acquiring smp_ipi_mtx here has a double effect:
529 * - it disables interrupts avoiding CPU0 preemption
530 * by fast handlers (thus deadlocking against other CPUs)
531 * - it avoids deadlocks against smp_rendezvous() or, more
532 * generally, threads busy-waiting, with this spinlock held,
533 * and waiting for responses by threads on other CPUs
534 * (ie. smp_tlb_shootdown()).
536 * For the !SMP case it just needs to handle the former problem.
539 mtx_lock_spin(&smp_ipi_mtx);
544 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
546 /* NOTREACHED */ /* assuming reset worked */
550 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
551 * and then reboots. If we are called twice, then we avoid trying to sync
552 * the disks as this often leads to recursive panics.
555 panic(const char *fmt, ...)
558 static volatile u_int panic_cpu = NOCPU;
561 struct thread *td = curthread;
562 int bootopt, newpanic;
564 static char buf[256];
566 if (stop_scheduler_on_panic)
573 * We don't want multiple CPU's to panic at the same time, so we
574 * use panic_cpu as a simple spinlock. We have to keep checking
575 * panic_cpu if we are spinning in case the panic on the first
578 if (panic_cpu != PCPU_GET(cpuid))
579 while (atomic_cmpset_int(&panic_cpu, NOCPU,
580 PCPU_GET(cpuid)) == 0)
581 while (panic_cpu != NOCPU)
584 if (stop_scheduler_on_panic) {
585 if (panicstr == NULL && !kdb_active) {
586 other_cpus = all_cpus;
587 CPU_CLR(PCPU_GET(cpuid), &other_cpus);
588 stop_cpus_hard(other_cpus);
592 * We set stop_scheduler here and not in the block above,
593 * because we want to ensure that if panic has been called and
594 * stop_scheduler_on_panic is true, then stop_scheduler will
595 * always be set. Even if panic has been entered from kdb.
601 bootopt = RB_AUTOBOOT;
604 bootopt |= RB_NOSYNC;
613 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
616 printf("panic: %s\n", buf);
624 printf("cpuid = %d\n", PCPU_GET(cpuid));
628 if (newpanic && trace_on_panic)
630 if (debugger_on_panic)
631 kdb_enter(KDB_WHY_PANIC, "panic");
633 /*thread_lock(td); */
634 td->td_flags |= TDF_INPANIC;
635 /* thread_unlock(td); */
637 bootopt |= RB_NOSYNC;
638 if (!stop_scheduler_on_panic)
640 kern_reboot(bootopt);
644 * Support for poweroff delay.
646 * Please note that setting this delay too short might power off your machine
647 * before the write cache on your hard disk has been flushed, leading to
648 * soft-updates inconsistencies.
650 #ifndef POWEROFF_DELAY
651 # define POWEROFF_DELAY 5000
653 static int poweroff_delay = POWEROFF_DELAY;
655 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
656 &poweroff_delay, 0, "");
659 poweroff_wait(void *junk, int howto)
662 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
664 DELAY(poweroff_delay * 1000);
668 * Some system processes (e.g. syncer) need to be stopped at appropriate
669 * points in their main loops prior to a system shutdown, so that they
670 * won't interfere with the shutdown process (e.g. by holding a disk buf
671 * to cause sync to fail). For each of these system processes, register
672 * shutdown_kproc() as a handler for one of shutdown events.
674 static int kproc_shutdown_wait = 60;
675 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
676 &kproc_shutdown_wait, 0, "");
679 kproc_shutdown(void *arg, int howto)
687 p = (struct proc *)arg;
688 printf("Waiting (max %d seconds) for system process `%s' to stop...",
689 kproc_shutdown_wait, p->p_comm);
690 error = kproc_suspend(p, kproc_shutdown_wait * hz);
692 if (error == EWOULDBLOCK)
693 printf("timed out\n");
699 kthread_shutdown(void *arg, int howto)
707 td = (struct thread *)arg;
708 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
709 kproc_shutdown_wait, td->td_name);
710 error = kthread_suspend(td, kproc_shutdown_wait * hz);
712 if (error == EWOULDBLOCK)
713 printf("timed out\n");
718 /* Registration of dumpers */
720 set_dumper(struct dumperinfo *di)
724 bzero(&dumper, sizeof dumper);
727 if (dumper.dumper != NULL)
733 /* Call dumper with bounds checking. */
735 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
736 off_t offset, size_t length)
739 if (length != 0 && (offset < di->mediaoffset ||
740 offset - di->mediaoffset + length > di->mediasize)) {
741 printf("Attempt to write outside dump device boundaries.\n"
742 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
743 (intmax_t)offset, (intmax_t)di->mediaoffset,
744 (uintmax_t)length, (intmax_t)di->mediasize);
747 return (di->dumper(di->priv, virtual, physical, offset, length));
751 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
752 uint64_t dumplen, uint32_t blksz)
755 bzero(kdh, sizeof(*kdh));
756 strncpy(kdh->magic, magic, sizeof(kdh->magic));
757 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
758 kdh->version = htod32(KERNELDUMPVERSION);
759 kdh->architectureversion = htod32(archver);
760 kdh->dumplength = htod64(dumplen);
761 kdh->dumptime = htod64(time_second);
762 kdh->blocksize = htod32(blksz);
763 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
764 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
765 if (panicstr != NULL)
766 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
767 kdh->parity = kerneldump_parity(kdh);