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>
53 #include <sys/filedesc.h>
56 #include <sys/kernel.h>
57 #include <sys/kerneldump.h>
58 #include <sys/kthread.h>
60 #include <sys/malloc.h>
61 #include <sys/mount.h>
64 #include <sys/reboot.h>
65 #include <sys/resourcevar.h>
66 #include <sys/rwlock.h>
67 #include <sys/sched.h>
69 #include <sys/sysctl.h>
70 #include <sys/sysproto.h>
71 #include <sys/vnode.h>
72 #include <sys/watchdog.h>
76 #include <machine/cpu.h>
77 #include <machine/dump.h>
78 #include <machine/pcb.h>
79 #include <machine/smp.h>
81 #include <security/mac/mac_framework.h>
84 #include <vm/vm_object.h>
85 #include <vm/vm_page.h>
86 #include <vm/vm_pager.h>
87 #include <vm/swap_pager.h>
89 #include <sys/signalvar.h>
91 #ifndef PANIC_REBOOT_WAIT_TIME
92 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
94 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
95 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
96 &panic_reboot_wait_time, 0,
97 "Seconds to wait before rebooting after a panic");
100 * Note that stdarg.h and the ANSI style va_start macro is used for both
101 * ANSI and traditional C compilers.
103 #include <machine/stdarg.h>
106 #ifdef KDB_UNATTENDED
107 int debugger_on_panic = 0;
109 int debugger_on_panic = 1;
111 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
112 CTLFLAG_RWTUN | CTLFLAG_SECURE,
113 &debugger_on_panic, 0, "Run debugger on kernel panic");
116 static int trace_on_panic = 1;
118 static int trace_on_panic = 0;
120 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic,
121 CTLFLAG_RWTUN | CTLFLAG_SECURE,
122 &trace_on_panic, 0, "Print stack trace on kernel panic");
125 static int sync_on_panic = 0;
126 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
127 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
129 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
130 "Shutdown environment");
133 static int show_busybufs;
135 static int show_busybufs = 1;
137 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
138 &show_busybufs, 0, "");
141 * Variable panicstr contains argument to first call to panic; used as flag
142 * to indicate that the kernel has already called panic.
144 const char *panicstr;
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 lwpid_t dumptid; /* Thread ID. */
154 static struct cdevsw reroot_cdevsw = {
155 .d_version = D_VERSION,
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);
163 static int kern_reroot(void);
165 /* register various local shutdown events */
167 shutdown_conf(void *unused)
170 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
172 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
173 SHUTDOWN_PRI_LAST + 100);
174 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
175 SHUTDOWN_PRI_LAST + 100);
176 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
177 SHUTDOWN_PRI_LAST + 200);
180 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
183 * The only reason this exists is to create the /dev/reroot/ directory,
184 * used by reroot code in init(8) as a mountpoint for tmpfs.
187 reroot_conf(void *unused)
192 error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev,
193 &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot");
195 printf("%s: failed to create device node, error %d",
200 SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL);
203 * The system call that results in a reboot.
207 sys_reboot(struct thread *td, struct reboot_args *uap)
213 error = mac_system_check_reboot(td->td_ucred, uap->opt);
216 error = priv_check(td, PRIV_REBOOT);
218 if (uap->opt & RB_REROOT) {
219 error = kern_reroot();
222 kern_reboot(uap->opt);
230 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
233 shutdown_nice(int howto)
236 if (initproc != NULL) {
237 /* Send a signal to init(8) and have it shutdown the world. */
239 if (howto & RB_POWEROFF)
240 kern_psignal(initproc, SIGUSR2);
241 else if (howto & RB_HALT)
242 kern_psignal(initproc, SIGUSR1);
244 kern_psignal(initproc, SIGINT);
245 PROC_UNLOCK(initproc);
247 /* No init(8) running, so simply reboot. */
248 kern_reboot(howto | RB_NOSYNC);
261 if (ts.tv_sec >= 86400) {
262 printf("%ldd", (long)ts.tv_sec / 86400);
266 if (f || ts.tv_sec >= 3600) {
267 printf("%ldh", (long)ts.tv_sec / 3600);
271 if (f || ts.tv_sec >= 60) {
272 printf("%ldm", (long)ts.tv_sec / 60);
276 printf("%lds\n", (long)ts.tv_sec);
280 doadump(boolean_t textdump)
288 if (dumper.dumper == NULL)
292 dumptid = curthread->td_tid;
297 if (textdump && textdump_pending) {
299 textdump_dumpsys(&dumper);
303 error = dumpsys(&dumper);
310 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
313 kern_reboot(int howto)
319 * Bind us to CPU 0 so that all shutdown code runs there. Some
320 * systems don't shutdown properly (i.e., ACPI power off) if we
321 * run on another processor.
323 if (!SCHEDULER_STOPPED()) {
324 thread_lock(curthread);
325 sched_bind(curthread, 0);
326 thread_unlock(curthread);
327 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
330 /* We're in the process of rebooting. */
333 /* We are out of the debugger now. */
337 * Do any callouts that should be done BEFORE syncing the filesystems.
339 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
342 * Now sync filesystems
344 if (!cold && (howto & RB_NOSYNC) == 0 && once == 0) {
346 bufshutdown(show_busybufs);
354 * Ok, now do things that assume all filesystem activity has
357 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
359 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
362 /* Now that we're going to really halt the system... */
363 EVENTHANDLER_INVOKE(shutdown_final, howto);
365 for(;;) ; /* safety against shutdown_reset not working */
370 * The system call that results in changing the rootfs.
375 struct vnode *oldrootvnode, *vp;
376 struct mount *mp, *devmp;
379 if (curproc != initproc)
383 * Mark the filesystem containing currently-running executable
384 * (the temporary copy of init(8)) busy.
386 vp = curproc->p_textvp;
387 error = vn_lock(vp, LK_SHARED);
391 error = vfs_busy(mp, MBF_NOWAIT);
395 error = vfs_busy(mp, 0);
396 vn_lock(vp, LK_SHARED | LK_RETRY);
402 if (vp->v_iflag & VI_DOOMED) {
411 * Remove the filesystem containing currently-running executable
412 * from the mount list, to prevent it from being unmounted
413 * by vfs_unmountall(), and to avoid confusing vfs_mountroot().
415 * Also preserve /dev - forcibly unmounting it could cause driver
423 mtx_lock(&mountlist_mtx);
424 TAILQ_REMOVE(&mountlist, mp, mnt_list);
425 TAILQ_REMOVE(&mountlist, devmp, mnt_list);
426 mtx_unlock(&mountlist_mtx);
428 oldrootvnode = rootvnode;
431 * Unmount everything except for the two filesystems preserved above.
436 * Add /dev back; vfs_mountroot() will move it into its new place.
438 mtx_lock(&mountlist_mtx);
439 TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list);
440 mtx_unlock(&mountlist_mtx);
445 * Mount the new rootfs.
450 * Update all references to the old rootvnode.
452 mountcheckdirs(oldrootvnode, rootvnode);
455 * Add the temporary filesystem back and unbusy it.
457 mtx_lock(&mountlist_mtx);
458 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
459 mtx_unlock(&mountlist_mtx);
466 * If the shutdown was a clean halt, behave accordingly.
469 shutdown_halt(void *junk, int howto)
472 if (howto & RB_HALT) {
474 printf("The operating system has halted.\n");
475 printf("Please press any key to reboot.\n\n");
477 case -1: /* No console, just die */
488 * Check to see if the system paniced, pause and then reboot
489 * according to the specified delay.
492 shutdown_panic(void *junk, int howto)
496 if (howto & RB_DUMP) {
497 if (panic_reboot_wait_time != 0) {
498 if (panic_reboot_wait_time != -1) {
499 printf("Automatic reboot in %d seconds - "
500 "press a key on the console to abort\n",
501 panic_reboot_wait_time);
502 for (loop = panic_reboot_wait_time * 10;
504 DELAY(1000 * 100); /* 1/10th second */
505 /* Did user type a key? */
506 if (cncheckc() != -1)
512 } else { /* zero time specified - reboot NOW */
515 printf("--> Press a key on the console to reboot,\n");
516 printf("--> or switch off the system now.\n");
522 * Everything done, now reset
525 shutdown_reset(void *junk, int howto)
528 printf("Rebooting...\n");
529 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
532 * Acquiring smp_ipi_mtx here has a double effect:
533 * - it disables interrupts avoiding CPU0 preemption
534 * by fast handlers (thus deadlocking against other CPUs)
535 * - it avoids deadlocks against smp_rendezvous() or, more
536 * generally, threads busy-waiting, with this spinlock held,
537 * and waiting for responses by threads on other CPUs
538 * (ie. smp_tlb_shootdown()).
540 * For the !SMP case it just needs to handle the former problem.
543 mtx_lock_spin(&smp_ipi_mtx);
548 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
550 /* NOTREACHED */ /* assuming reset worked */
553 #if defined(WITNESS) || defined(INVARIANTS)
554 static int kassert_warn_only = 0;
556 static int kassert_do_kdb = 0;
559 static int kassert_do_ktr = 0;
561 static int kassert_do_log = 1;
562 static int kassert_log_pps_limit = 4;
563 static int kassert_log_mute_at = 0;
564 static int kassert_log_panic_at = 0;
565 static int kassert_warnings = 0;
567 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
569 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
570 &kassert_warn_only, 0,
571 "KASSERT triggers a panic (1) or just a warning (0)");
574 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
575 &kassert_do_kdb, 0, "KASSERT will enter the debugger");
579 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
581 "KASSERT does a KTR, set this to the KTRMASK you want");
584 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
585 &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
587 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
588 &kassert_warnings, 0, "number of KASSERTs that have been triggered");
590 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
591 &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
593 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
594 &kassert_log_pps_limit, 0, "limit number of log messages per second");
596 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
597 &kassert_log_mute_at, 0, "max number of KASSERTS to log");
599 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
601 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
602 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
603 kassert_sysctl_kassert, "I", "set to trigger a test kassert");
606 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
610 error = sysctl_wire_old_buffer(req, sizeof(int));
613 error = sysctl_handle_int(oidp, &i, 0, req);
615 if (error != 0 || req->newptr == NULL)
617 KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
622 * Called by KASSERT, this decides if we will panic
623 * or if we will log via printf and/or ktr.
626 kassert_panic(const char *fmt, ...)
628 static char buf[256];
632 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
636 * panic if we're not just warning, or if we've exceeded
637 * kassert_log_panic_at warnings.
639 if (!kassert_warn_only ||
640 (kassert_log_panic_at > 0 &&
641 kassert_warnings >= kassert_log_panic_at)) {
651 * log if we've not yet met the mute limit.
653 if (kassert_do_log &&
654 (kassert_log_mute_at == 0 ||
655 kassert_warnings < kassert_log_mute_at)) {
656 static struct timeval lasterr;
659 if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
660 printf("KASSERT failed: %s\n", buf);
665 if (kassert_do_kdb) {
666 kdb_enter(KDB_WHY_KASSERT, buf);
669 atomic_add_int(&kassert_warnings, 1);
674 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
675 * and then reboots. If we are called twice, then we avoid trying to sync
676 * the disks as this often leads to recursive panics.
679 panic(const char *fmt, ...)
688 vpanic(const char *fmt, va_list ap)
693 struct thread *td = curthread;
694 int bootopt, newpanic;
695 static char buf[256];
701 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
702 * concurrently entering panic. Only the winner will proceed
705 if (panicstr == NULL && !kdb_active) {
706 other_cpus = all_cpus;
707 CPU_CLR(PCPU_GET(cpuid), &other_cpus);
708 stop_cpus_hard(other_cpus);
712 * Ensure that the scheduler is stopped while panicking, even if panic
713 * has been entered from kdb.
715 td->td_stopsched = 1;
718 bootopt = RB_AUTOBOOT;
721 bootopt |= RB_NOSYNC;
729 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
732 printf("panic: %s\n", buf);
739 printf("cpuid = %d\n", PCPU_GET(cpuid));
743 if (newpanic && trace_on_panic)
745 if (debugger_on_panic)
746 kdb_enter(KDB_WHY_PANIC, "panic");
748 /*thread_lock(td); */
749 td->td_flags |= TDF_INPANIC;
750 /* thread_unlock(td); */
752 bootopt |= RB_NOSYNC;
753 kern_reboot(bootopt);
757 * Support for poweroff delay.
759 * Please note that setting this delay too short might power off your machine
760 * before the write cache on your hard disk has been flushed, leading to
761 * soft-updates inconsistencies.
763 #ifndef POWEROFF_DELAY
764 # define POWEROFF_DELAY 5000
766 static int poweroff_delay = POWEROFF_DELAY;
768 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
769 &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
772 poweroff_wait(void *junk, int howto)
775 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
777 DELAY(poweroff_delay * 1000);
781 * Some system processes (e.g. syncer) need to be stopped at appropriate
782 * points in their main loops prior to a system shutdown, so that they
783 * won't interfere with the shutdown process (e.g. by holding a disk buf
784 * to cause sync to fail). For each of these system processes, register
785 * shutdown_kproc() as a handler for one of shutdown events.
787 static int kproc_shutdown_wait = 60;
788 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
789 &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
792 kproc_shutdown(void *arg, int howto)
800 p = (struct proc *)arg;
801 printf("Waiting (max %d seconds) for system process `%s' to stop...",
802 kproc_shutdown_wait, p->p_comm);
803 error = kproc_suspend(p, kproc_shutdown_wait * hz);
805 if (error == EWOULDBLOCK)
806 printf("timed out\n");
812 kthread_shutdown(void *arg, int howto)
820 td = (struct thread *)arg;
821 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
822 kproc_shutdown_wait, td->td_name);
823 error = kthread_suspend(td, kproc_shutdown_wait * hz);
825 if (error == EWOULDBLOCK)
826 printf("timed out\n");
831 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
832 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
833 dumpdevname, 0, "Device for kernel dumps");
835 /* Registration of dumpers */
837 set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
842 error = priv_check(td, PRIV_SETDUMPER);
847 bzero(&dumper, sizeof dumper);
848 dumpdevname[0] = '\0';
851 if (dumper.dumper != NULL)
854 wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
855 if (wantcopy >= sizeof(dumpdevname)) {
856 printf("set_dumper: device name truncated from '%s' -> '%s'\n",
857 devname, dumpdevname);
862 /* Call dumper with bounds checking. */
864 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
865 off_t offset, size_t length)
868 if (length != 0 && (offset < di->mediaoffset ||
869 offset - di->mediaoffset + length > di->mediasize)) {
870 printf("Attempt to write outside dump device boundaries.\n"
871 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
872 (intmax_t)offset, (intmax_t)di->mediaoffset,
873 (uintmax_t)length, (intmax_t)di->mediasize);
876 return (di->dumper(di->priv, virtual, physical, offset, length));
880 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
881 uint64_t dumplen, uint32_t blksz)
884 bzero(kdh, sizeof(*kdh));
885 strlcpy(kdh->magic, magic, sizeof(kdh->magic));
886 strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
887 kdh->version = htod32(KERNELDUMPVERSION);
888 kdh->architectureversion = htod32(archver);
889 kdh->dumplength = htod64(dumplen);
890 kdh->dumptime = htod64(time_second);
891 kdh->blocksize = htod32(blksz);
892 strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
893 strlcpy(kdh->versionstring, version, sizeof(kdh->versionstring));
894 if (panicstr != NULL)
895 strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
896 kdh->parity = kerneldump_parity(kdh);