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>
59 #include <sys/malloc.h>
60 #include <sys/mount.h>
63 #include <sys/reboot.h>
64 #include <sys/resourcevar.h>
65 #include <sys/rwlock.h>
66 #include <sys/sched.h>
68 #include <sys/sysctl.h>
69 #include <sys/sysproto.h>
70 #include <sys/vnode.h>
71 #include <sys/watchdog.h>
75 #include <machine/cpu.h>
76 #include <machine/dump.h>
77 #include <machine/pcb.h>
78 #include <machine/smp.h>
80 #include <security/mac/mac_framework.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/vm_pager.h>
86 #include <vm/swap_pager.h>
88 #include <sys/signalvar.h>
90 #ifndef PANIC_REBOOT_WAIT_TIME
91 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
93 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
94 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
95 &panic_reboot_wait_time, 0,
96 "Seconds to wait before rebooting after a panic");
99 * Note that stdarg.h and the ANSI style va_start macro is used for both
100 * ANSI and traditional C compilers.
102 #include <machine/stdarg.h>
105 #ifdef KDB_UNATTENDED
106 int debugger_on_panic = 0;
108 int debugger_on_panic = 1;
110 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
111 CTLFLAG_RWTUN | CTLFLAG_SECURE,
112 &debugger_on_panic, 0, "Run debugger on kernel panic");
115 static int trace_on_panic = 1;
117 static int trace_on_panic = 0;
119 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
120 CTLFLAG_RWTUN | CTLFLAG_SECURE,
121 &trace_on_panic, 0, "Print stack trace on kernel panic");
124 static int sync_on_panic = 0;
125 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
126 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
128 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
129 "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 dumping; /* system is dumping */
146 int rebooting; /* system is rebooting */
147 static struct dumperinfo dumper; /* our selected dumper */
149 /* Context information for dump-debuggers. */
150 static struct pcb dumppcb; /* Registers. */
151 lwpid_t dumptid; /* Thread ID. */
153 static void poweroff_wait(void *, int);
154 static void shutdown_halt(void *junk, int howto);
155 static void shutdown_panic(void *junk, int howto);
156 static void shutdown_reset(void *junk, int howto);
157 static void vpanic(const char *fmt, va_list ap) __dead2;
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
203 shutdown_nice(int howto)
206 if (initproc != NULL) {
207 /* Send a signal to init(8) and have it shutdown the world. */
209 if (howto & RB_POWEROFF)
210 kern_psignal(initproc, SIGUSR2);
211 else if (howto & RB_HALT)
212 kern_psignal(initproc, SIGUSR1);
214 kern_psignal(initproc, SIGINT);
215 PROC_UNLOCK(initproc);
217 /* No init(8) running, so simply reboot. */
218 kern_reboot(howto | RB_NOSYNC);
231 if (ts.tv_sec >= 86400) {
232 printf("%ldd", (long)ts.tv_sec / 86400);
236 if (f || ts.tv_sec >= 3600) {
237 printf("%ldh", (long)ts.tv_sec / 3600);
241 if (f || ts.tv_sec >= 60) {
242 printf("%ldm", (long)ts.tv_sec / 60);
246 printf("%lds\n", (long)ts.tv_sec);
250 doadump(boolean_t textdump)
258 if (dumper.dumper == NULL)
262 dumptid = curthread->td_tid;
267 if (textdump && textdump_pending) {
269 textdump_dumpsys(&dumper);
273 error = dumpsys(&dumper);
280 isbufbusy(struct buf *bp)
282 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
284 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
290 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
293 kern_reboot(int howto)
295 static int first_buf_printf = 1;
296 static int waittime = -1;
300 * Bind us to CPU 0 so that all shutdown code runs there. Some
301 * systems don't shutdown properly (i.e., ACPI power off) if we
302 * run on another processor.
304 if (!SCHEDULER_STOPPED()) {
305 thread_lock(curthread);
306 sched_bind(curthread, 0);
307 thread_unlock(curthread);
308 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
311 /* We're in the process of rebooting. */
314 /* We are out of the debugger now. */
318 * Do any callouts that should be done BEFORE syncing the filesystems.
320 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
323 * Now sync filesystems
325 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
326 register struct buf *bp;
327 int iter, nbusy, pbusy;
334 wdog_kern_pat(WD_LASTVAL);
335 sys_sync(curthread, NULL);
338 * With soft updates, some buffers that are
339 * written will be remarked as dirty until other
340 * buffers are written.
342 for (iter = pbusy = 0; iter < 20; iter++) {
344 for (bp = &buf[nbuf]; --bp >= buf; )
348 if (first_buf_printf)
349 printf("All buffers synced.");
352 if (first_buf_printf) {
353 printf("Syncing disks, buffers remaining... ");
354 first_buf_printf = 0;
356 printf("%d ", nbusy);
361 wdog_kern_pat(WD_LASTVAL);
362 sys_sync(curthread, NULL);
366 * Drop Giant and spin for a while to allow
367 * interrupt threads to run.
374 * Drop Giant and context switch several times to
375 * allow interrupt threads to run.
378 for (subiter = 0; subiter < 50 * iter; subiter++) {
379 thread_lock(curthread);
380 mi_switch(SW_VOL, NULL);
381 thread_unlock(curthread);
389 * Count only busy local buffers to prevent forcing
390 * a fsck if we're just a client of a wedged NFS server
393 for (bp = &buf[nbuf]; --bp >= buf; ) {
396 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
397 if (bp->b_dev == NULL) {
398 TAILQ_REMOVE(&mountlist,
399 bp->b_vp->v_mount, mnt_list);
404 if (show_busybufs > 0) {
406 "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
407 nbusy, bp, bp->b_vp, bp->b_flags,
408 (intmax_t)bp->b_blkno,
409 (intmax_t)bp->b_lblkno);
410 BUF_LOCKPRINTINFO(bp);
411 if (show_busybufs > 1)
419 * Failed to sync all blocks. Indicate this and don't
420 * unmount filesystems (thus forcing an fsck on reboot).
422 printf("Giving up on %d buffers\n", nbusy);
423 DELAY(5000000); /* 5 seconds */
425 if (!first_buf_printf)
426 printf("Final sync complete\n");
428 * Unmount filesystems
434 DELAY(100000); /* wait for console output to finish */
442 * Ok, now do things that assume all filesystem activity has
445 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
447 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
450 /* Now that we're going to really halt the system... */
451 EVENTHANDLER_INVOKE(shutdown_final, howto);
453 for(;;) ; /* safety against shutdown_reset not working */
458 * If the shutdown was a clean halt, behave accordingly.
461 shutdown_halt(void *junk, int howto)
464 if (howto & RB_HALT) {
466 printf("The operating system has halted.\n");
467 printf("Please press any key to reboot.\n\n");
469 case -1: /* No console, just die */
480 * Check to see if the system paniced, pause and then reboot
481 * according to the specified delay.
484 shutdown_panic(void *junk, int howto)
488 if (howto & RB_DUMP) {
489 if (panic_reboot_wait_time != 0) {
490 if (panic_reboot_wait_time != -1) {
491 printf("Automatic reboot in %d seconds - "
492 "press a key on the console to abort\n",
493 panic_reboot_wait_time);
494 for (loop = panic_reboot_wait_time * 10;
496 DELAY(1000 * 100); /* 1/10th second */
497 /* Did user type a key? */
498 if (cncheckc() != -1)
504 } else { /* zero time specified - reboot NOW */
507 printf("--> Press a key on the console to reboot,\n");
508 printf("--> or switch off the system now.\n");
514 * Everything done, now reset
517 shutdown_reset(void *junk, int howto)
520 printf("Rebooting...\n");
521 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
524 * Acquiring smp_ipi_mtx here has a double effect:
525 * - it disables interrupts avoiding CPU0 preemption
526 * by fast handlers (thus deadlocking against other CPUs)
527 * - it avoids deadlocks against smp_rendezvous() or, more
528 * generally, threads busy-waiting, with this spinlock held,
529 * and waiting for responses by threads on other CPUs
530 * (ie. smp_tlb_shootdown()).
532 * For the !SMP case it just needs to handle the former problem.
535 mtx_lock_spin(&smp_ipi_mtx);
540 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
542 /* NOTREACHED */ /* assuming reset worked */
545 #if defined(WITNESS) || defined(INVARIANTS)
546 static int kassert_warn_only = 0;
548 static int kassert_do_kdb = 0;
551 static int kassert_do_ktr = 0;
553 static int kassert_do_log = 1;
554 static int kassert_log_pps_limit = 4;
555 static int kassert_log_mute_at = 0;
556 static int kassert_log_panic_at = 0;
557 static int kassert_warnings = 0;
559 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
561 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
562 &kassert_warn_only, 0,
563 "KASSERT triggers a panic (1) or just a warning (0)");
566 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
567 &kassert_do_kdb, 0, "KASSERT will enter the debugger");
571 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
573 "KASSERT does a KTR, set this to the KTRMASK you want");
576 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
577 &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
579 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
580 &kassert_warnings, 0, "number of KASSERTs that have been triggered");
582 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
583 &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
585 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
586 &kassert_log_pps_limit, 0, "limit number of log messages per second");
588 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
589 &kassert_log_mute_at, 0, "max number of KASSERTS to log");
591 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
593 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
594 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
595 kassert_sysctl_kassert, "I", "set to trigger a test kassert");
598 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
602 error = sysctl_wire_old_buffer(req, sizeof(int));
605 error = sysctl_handle_int(oidp, &i, 0, req);
607 if (error != 0 || req->newptr == NULL)
609 KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
614 * Called by KASSERT, this decides if we will panic
615 * or if we will log via printf and/or ktr.
618 kassert_panic(const char *fmt, ...)
620 static char buf[256];
624 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
628 * panic if we're not just warning, or if we've exceeded
629 * kassert_log_panic_at warnings.
631 if (!kassert_warn_only ||
632 (kassert_log_panic_at > 0 &&
633 kassert_warnings >= kassert_log_panic_at)) {
643 * log if we've not yet met the mute limit.
645 if (kassert_do_log &&
646 (kassert_log_mute_at == 0 ||
647 kassert_warnings < kassert_log_mute_at)) {
648 static struct timeval lasterr;
651 if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
652 printf("KASSERT failed: %s\n", buf);
657 if (kassert_do_kdb) {
658 kdb_enter(KDB_WHY_KASSERT, buf);
661 atomic_add_int(&kassert_warnings, 1);
666 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
667 * and then reboots. If we are called twice, then we avoid trying to sync
668 * the disks as this often leads to recursive panics.
671 panic(const char *fmt, ...)
680 vpanic(const char *fmt, va_list ap)
685 struct thread *td = curthread;
686 int bootopt, newpanic;
687 static char buf[256];
693 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
694 * concurrently entering panic. Only the winner will proceed
697 if (panicstr == NULL && !kdb_active) {
698 other_cpus = all_cpus;
699 CPU_CLR(PCPU_GET(cpuid), &other_cpus);
700 stop_cpus_hard(other_cpus);
704 * We set stop_scheduler here and not in the block above,
705 * because we want to ensure that if panic has been called and
706 * stop_scheduler_on_panic is true, then stop_scheduler will
707 * always be set. Even if panic has been entered from kdb.
709 td->td_stopsched = 1;
712 bootopt = RB_AUTOBOOT;
715 bootopt |= RB_NOSYNC;
723 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
726 printf("panic: %s\n", buf);
733 printf("cpuid = %d\n", PCPU_GET(cpuid));
737 if (newpanic && trace_on_panic)
739 if (debugger_on_panic)
740 kdb_enter(KDB_WHY_PANIC, "panic");
742 /*thread_lock(td); */
743 td->td_flags |= TDF_INPANIC;
744 /* thread_unlock(td); */
746 bootopt |= RB_NOSYNC;
747 kern_reboot(bootopt);
751 * Support for poweroff delay.
753 * Please note that setting this delay too short might power off your machine
754 * before the write cache on your hard disk has been flushed, leading to
755 * soft-updates inconsistencies.
757 #ifndef POWEROFF_DELAY
758 # define POWEROFF_DELAY 5000
760 static int poweroff_delay = POWEROFF_DELAY;
762 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
763 &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
766 poweroff_wait(void *junk, int howto)
769 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
771 DELAY(poweroff_delay * 1000);
775 * Some system processes (e.g. syncer) need to be stopped at appropriate
776 * points in their main loops prior to a system shutdown, so that they
777 * won't interfere with the shutdown process (e.g. by holding a disk buf
778 * to cause sync to fail). For each of these system processes, register
779 * shutdown_kproc() as a handler for one of shutdown events.
781 static int kproc_shutdown_wait = 60;
782 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
783 &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
786 kproc_shutdown(void *arg, int howto)
794 p = (struct proc *)arg;
795 printf("Waiting (max %d seconds) for system process `%s' to stop...",
796 kproc_shutdown_wait, p->p_comm);
797 error = kproc_suspend(p, kproc_shutdown_wait * hz);
799 if (error == EWOULDBLOCK)
800 printf("timed out\n");
806 kthread_shutdown(void *arg, int howto)
814 td = (struct thread *)arg;
815 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
816 kproc_shutdown_wait, td->td_name);
817 error = kthread_suspend(td, kproc_shutdown_wait * hz);
819 if (error == EWOULDBLOCK)
820 printf("timed out\n");
825 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
826 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
827 dumpdevname, 0, "Device for kernel dumps");
829 /* Registration of dumpers */
831 set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
836 error = priv_check(td, PRIV_SETDUMPER);
841 bzero(&dumper, sizeof dumper);
842 dumpdevname[0] = '\0';
845 if (dumper.dumper != NULL)
848 wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
849 if (wantcopy >= sizeof(dumpdevname)) {
850 printf("set_dumper: device name truncated from '%s' -> '%s'\n",
851 devname, dumpdevname);
856 /* Call dumper with bounds checking. */
858 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
859 off_t offset, size_t length)
862 if (length != 0 && (offset < di->mediaoffset ||
863 offset - di->mediaoffset + length > di->mediasize)) {
864 printf("Attempt to write outside dump device boundaries.\n"
865 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
866 (intmax_t)offset, (intmax_t)di->mediaoffset,
867 (uintmax_t)length, (intmax_t)di->mediasize);
870 return (di->dumper(di->priv, virtual, physical, offset, length));
874 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
875 uint64_t dumplen, uint32_t blksz)
878 bzero(kdh, sizeof(*kdh));
879 strncpy(kdh->magic, magic, sizeof(kdh->magic));
880 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
881 kdh->version = htod32(KERNELDUMPVERSION);
882 kdh->architectureversion = htod32(archver);
883 kdh->dumplength = htod64(dumplen);
884 kdh->dumptime = htod64(time_second);
885 kdh->blocksize = htod32(blksz);
886 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
887 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
888 if (panicstr != NULL)
889 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
890 kdh->parity = kerneldump_parity(kdh);