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>
69 #include <sys/watchdog.h>
73 #include <machine/cpu.h>
74 #include <machine/pcb.h>
75 #include <machine/smp.h>
77 #include <security/mac/mac_framework.h>
80 #include <vm/vm_object.h>
81 #include <vm/vm_page.h>
82 #include <vm/vm_pager.h>
83 #include <vm/swap_pager.h>
85 #include <sys/signalvar.h>
87 #ifndef PANIC_REBOOT_WAIT_TIME
88 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
90 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
91 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RW | CTLFLAG_TUN,
92 &panic_reboot_wait_time, 0,
93 "Seconds to wait before rebooting after a panic");
94 TUNABLE_INT("kern.panic_reboot_wait_time", &panic_reboot_wait_time);
97 * Note that stdarg.h and the ANSI style va_start macro is used for both
98 * ANSI and traditional C compilers.
100 #include <machine/stdarg.h>
103 #ifdef KDB_UNATTENDED
104 int debugger_on_panic = 0;
106 int debugger_on_panic = 1;
108 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
109 CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN,
110 &debugger_on_panic, 0, "Run debugger on kernel panic");
111 TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
114 static int trace_on_panic = 1;
116 static int trace_on_panic = 0;
118 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
119 CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN,
120 &trace_on_panic, 0, "Print stack trace on kernel panic");
121 TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
124 static int sync_on_panic = 0;
125 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
126 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
127 TUNABLE_INT("kern.sync_on_panic", &sync_on_panic);
129 static int stop_scheduler_on_panic = 1;
130 SYSCTL_INT(_kern, OID_AUTO, stop_scheduler_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
131 &stop_scheduler_on_panic, 0, "stop scheduler upon entering panic");
132 TUNABLE_INT("kern.stop_scheduler_on_panic", &stop_scheduler_on_panic);
134 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
135 "Shutdown environment");
138 static int show_busybufs;
140 static int show_busybufs = 1;
142 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
143 &show_busybufs, 0, "");
146 * Variable panicstr contains argument to first call to panic; used as flag
147 * to indicate that the kernel has already called panic.
149 const char *panicstr;
151 int dumping; /* system is dumping */
152 int rebooting; /* system is rebooting */
153 static struct dumperinfo dumper; /* our selected dumper */
155 /* Context information for dump-debuggers. */
156 static struct pcb dumppcb; /* Registers. */
157 lwpid_t dumptid; /* Thread ID. */
159 static void poweroff_wait(void *, int);
160 static void shutdown_halt(void *junk, int howto);
161 static void shutdown_panic(void *junk, int howto);
162 static void shutdown_reset(void *junk, int howto);
164 /* register various local shutdown events */
166 shutdown_conf(void *unused)
169 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
171 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
172 SHUTDOWN_PRI_LAST + 100);
173 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
174 SHUTDOWN_PRI_LAST + 100);
175 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
176 SHUTDOWN_PRI_LAST + 200);
179 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
182 * The system call that results in a reboot.
186 sys_reboot(struct thread *td, struct reboot_args *uap)
192 error = mac_system_check_reboot(td->td_ucred, uap->opt);
195 error = priv_check(td, PRIV_REBOOT);
198 kern_reboot(uap->opt);
205 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
207 static int shutdown_howto = 0;
210 shutdown_nice(int howto)
213 shutdown_howto = howto;
215 /* Send a signal to init(8) and have it shutdown the world */
216 if (initproc != NULL) {
218 kern_psignal(initproc, SIGINT);
219 PROC_UNLOCK(initproc);
221 /* No init(8) running, so simply reboot */
222 kern_reboot(RB_NOSYNC);
226 static int waittime = -1;
237 if (ts.tv_sec >= 86400) {
238 printf("%ldd", (long)ts.tv_sec / 86400);
242 if (f || ts.tv_sec >= 3600) {
243 printf("%ldh", (long)ts.tv_sec / 3600);
247 if (f || ts.tv_sec >= 60) {
248 printf("%ldm", (long)ts.tv_sec / 60);
252 printf("%lds\n", (long)ts.tv_sec);
256 doadump(boolean_t textdump)
262 if (dumper.dumper == NULL)
266 dumptid = curthread->td_tid;
271 if (textdump && textdump_pending) {
273 textdump_dumpsys(&dumper);
284 isbufbusy(struct buf *bp)
286 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
288 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
294 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
297 kern_reboot(int howto)
299 static int first_buf_printf = 1;
303 * Bind us to CPU 0 so that all shutdown code runs there. Some
304 * systems don't shutdown properly (i.e., ACPI power off) if we
305 * run on another processor.
307 if (!SCHEDULER_STOPPED()) {
308 thread_lock(curthread);
309 sched_bind(curthread, 0);
310 thread_unlock(curthread);
311 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
314 /* We're in the process of rebooting. */
317 /* collect extra flags that shutdown_nice might have set */
318 howto |= shutdown_howto;
320 /* We are out of the debugger now. */
324 * Do any callouts that should be done BEFORE syncing the filesystems.
326 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
329 * Now sync filesystems
331 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
332 register struct buf *bp;
333 int iter, nbusy, pbusy;
340 wdog_kern_pat(WD_LASTVAL);
341 sys_sync(curthread, NULL);
344 * With soft updates, some buffers that are
345 * written will be remarked as dirty until other
346 * buffers are written.
348 for (iter = pbusy = 0; iter < 20; iter++) {
350 for (bp = &buf[nbuf]; --bp >= buf; )
354 if (first_buf_printf)
355 printf("All buffers synced.");
358 if (first_buf_printf) {
359 printf("Syncing disks, buffers remaining... ");
360 first_buf_printf = 0;
362 printf("%d ", nbusy);
367 wdog_kern_pat(WD_LASTVAL);
368 sys_sync(curthread, NULL);
372 * Drop Giant and spin for a while to allow
373 * interrupt threads to run.
380 * Drop Giant and context switch several times to
381 * allow interrupt threads to run.
384 for (subiter = 0; subiter < 50 * iter; subiter++) {
385 thread_lock(curthread);
386 mi_switch(SW_VOL, NULL);
387 thread_unlock(curthread);
395 * Count only busy local buffers to prevent forcing
396 * a fsck if we're just a client of a wedged NFS server
399 for (bp = &buf[nbuf]; --bp >= buf; ) {
402 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
403 if (bp->b_dev == NULL) {
404 TAILQ_REMOVE(&mountlist,
405 bp->b_vp->v_mount, mnt_list);
410 if (show_busybufs > 0) {
412 "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
413 nbusy, bp, bp->b_vp, bp->b_flags,
414 (intmax_t)bp->b_blkno,
415 (intmax_t)bp->b_lblkno);
416 BUF_LOCKPRINTINFO(bp);
417 if (show_busybufs > 1)
425 * Failed to sync all blocks. Indicate this and don't
426 * unmount filesystems (thus forcing an fsck on reboot).
428 printf("Giving up on %d buffers\n", nbusy);
429 DELAY(5000000); /* 5 seconds */
431 if (!first_buf_printf)
432 printf("Final sync complete\n");
434 * Unmount filesystems
440 DELAY(100000); /* wait for console output to finish */
448 * Ok, now do things that assume all filesystem activity has
451 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
453 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
456 /* Now that we're going to really halt the system... */
457 EVENTHANDLER_INVOKE(shutdown_final, howto);
459 for(;;) ; /* safety against shutdown_reset not working */
464 * If the shutdown was a clean halt, behave accordingly.
467 shutdown_halt(void *junk, int howto)
470 if (howto & RB_HALT) {
472 printf("The operating system has halted.\n");
473 printf("Please press any key to reboot.\n\n");
475 case -1: /* No console, just die */
486 * Check to see if the system paniced, pause and then reboot
487 * according to the specified delay.
490 shutdown_panic(void *junk, int howto)
494 if (howto & RB_DUMP) {
495 if (panic_reboot_wait_time != 0) {
496 if (panic_reboot_wait_time != -1) {
497 printf("Automatic reboot in %d seconds - "
498 "press a key on the console to abort\n",
499 panic_reboot_wait_time);
500 for (loop = panic_reboot_wait_time * 10;
502 DELAY(1000 * 100); /* 1/10th second */
503 /* Did user type a key? */
504 if (cncheckc() != -1)
510 } else { /* zero time specified - reboot NOW */
513 printf("--> Press a key on the console to reboot,\n");
514 printf("--> or switch off the system now.\n");
520 * Everything done, now reset
523 shutdown_reset(void *junk, int howto)
526 printf("Rebooting...\n");
527 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
530 * Acquiring smp_ipi_mtx here has a double effect:
531 * - it disables interrupts avoiding CPU0 preemption
532 * by fast handlers (thus deadlocking against other CPUs)
533 * - it avoids deadlocks against smp_rendezvous() or, more
534 * generally, threads busy-waiting, with this spinlock held,
535 * and waiting for responses by threads on other CPUs
536 * (ie. smp_tlb_shootdown()).
538 * For the !SMP case it just needs to handle the former problem.
541 mtx_lock_spin(&smp_ipi_mtx);
546 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
548 /* NOTREACHED */ /* assuming reset worked */
552 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
553 * and then reboots. If we are called twice, then we avoid trying to sync
554 * the disks as this often leads to recursive panics.
557 panic(const char *fmt, ...)
560 static volatile u_int panic_cpu = NOCPU;
563 struct thread *td = curthread;
564 int bootopt, newpanic;
566 static char buf[256];
568 if (stop_scheduler_on_panic)
575 * We don't want multiple CPU's to panic at the same time, so we
576 * use panic_cpu as a simple spinlock. We have to keep checking
577 * panic_cpu if we are spinning in case the panic on the first
580 if (panic_cpu != PCPU_GET(cpuid))
581 while (atomic_cmpset_int(&panic_cpu, NOCPU,
582 PCPU_GET(cpuid)) == 0)
583 while (panic_cpu != NOCPU)
586 if (stop_scheduler_on_panic) {
587 if (panicstr == NULL && !kdb_active) {
588 other_cpus = all_cpus;
589 CPU_CLR(PCPU_GET(cpuid), &other_cpus);
590 stop_cpus_hard(other_cpus);
594 * We set stop_scheduler here and not in the block above,
595 * because we want to ensure that if panic has been called and
596 * stop_scheduler_on_panic is true, then stop_scheduler will
597 * always be set. Even if panic has been entered from kdb.
599 td->td_stopsched = 1;
603 bootopt = RB_AUTOBOOT;
606 bootopt |= RB_NOSYNC;
615 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
618 printf("panic: %s\n", buf);
626 printf("cpuid = %d\n", PCPU_GET(cpuid));
630 if (newpanic && trace_on_panic)
632 if (debugger_on_panic)
633 kdb_enter(KDB_WHY_PANIC, "panic");
635 /*thread_lock(td); */
636 td->td_flags |= TDF_INPANIC;
637 /* thread_unlock(td); */
639 bootopt |= RB_NOSYNC;
640 if (!stop_scheduler_on_panic)
642 kern_reboot(bootopt);
646 * Support for poweroff delay.
648 * Please note that setting this delay too short might power off your machine
649 * before the write cache on your hard disk has been flushed, leading to
650 * soft-updates inconsistencies.
652 #ifndef POWEROFF_DELAY
653 # define POWEROFF_DELAY 5000
655 static int poweroff_delay = POWEROFF_DELAY;
657 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
658 &poweroff_delay, 0, "");
661 poweroff_wait(void *junk, int howto)
664 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
666 DELAY(poweroff_delay * 1000);
670 * Some system processes (e.g. syncer) need to be stopped at appropriate
671 * points in their main loops prior to a system shutdown, so that they
672 * won't interfere with the shutdown process (e.g. by holding a disk buf
673 * to cause sync to fail). For each of these system processes, register
674 * shutdown_kproc() as a handler for one of shutdown events.
676 static int kproc_shutdown_wait = 60;
677 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
678 &kproc_shutdown_wait, 0, "");
681 kproc_shutdown(void *arg, int howto)
689 p = (struct proc *)arg;
690 printf("Waiting (max %d seconds) for system process `%s' to stop...",
691 kproc_shutdown_wait, p->p_comm);
692 error = kproc_suspend(p, kproc_shutdown_wait * hz);
694 if (error == EWOULDBLOCK)
695 printf("timed out\n");
701 kthread_shutdown(void *arg, int howto)
709 td = (struct thread *)arg;
710 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
711 kproc_shutdown_wait, td->td_name);
712 error = kthread_suspend(td, kproc_shutdown_wait * hz);
714 if (error == EWOULDBLOCK)
715 printf("timed out\n");
720 /* Registration of dumpers */
722 set_dumper(struct dumperinfo *di)
726 bzero(&dumper, sizeof dumper);
729 if (dumper.dumper != NULL)
735 /* Call dumper with bounds checking. */
737 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
738 off_t offset, size_t length)
741 if (length != 0 && (offset < di->mediaoffset ||
742 offset - di->mediaoffset + length > di->mediasize)) {
743 printf("Attempt to write outside dump device boundaries.\n"
744 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
745 (intmax_t)offset, (intmax_t)di->mediaoffset,
746 (uintmax_t)length, (intmax_t)di->mediasize);
749 return (di->dumper(di->priv, virtual, physical, offset, length));
753 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
754 uint64_t dumplen, uint32_t blksz)
757 bzero(kdh, sizeof(*kdh));
758 strncpy(kdh->magic, magic, sizeof(kdh->magic));
759 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
760 kdh->version = htod32(KERNELDUMPVERSION);
761 kdh->architectureversion = htod32(archver);
762 kdh->dumplength = htod64(dumplen);
763 kdh->dumptime = htod64(time_second);
764 kdh->blocksize = htod32(blksz);
765 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
766 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
767 if (panicstr != NULL)
768 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
769 kdh->parity = kerneldump_parity(kdh);