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, "");
140 int suspend_blocked = 0;
141 SYSCTL_INT(_kern, OID_AUTO, suspend_blocked, CTLFLAG_RW,
142 &suspend_blocked, 0, "Block suspend due to a pending shutdown");
145 * Variable panicstr contains argument to first call to panic; used as flag
146 * to indicate that the kernel has already called panic.
148 const char *panicstr;
150 int dumping; /* system is dumping */
151 int rebooting; /* system is rebooting */
152 static struct dumperinfo dumper; /* our selected dumper */
154 /* Context information for dump-debuggers. */
155 static struct pcb dumppcb; /* Registers. */
156 lwpid_t dumptid; /* Thread ID. */
158 static struct cdevsw reroot_cdevsw = {
159 .d_version = D_VERSION,
163 static void poweroff_wait(void *, int);
164 static void shutdown_halt(void *junk, int howto);
165 static void shutdown_panic(void *junk, int howto);
166 static void shutdown_reset(void *junk, int howto);
167 static int kern_reroot(void);
169 /* register various local shutdown events */
171 shutdown_conf(void *unused)
174 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
176 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
177 SHUTDOWN_PRI_LAST + 100);
178 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
179 SHUTDOWN_PRI_LAST + 100);
180 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
181 SHUTDOWN_PRI_LAST + 200);
184 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
187 * The only reason this exists is to create the /dev/reroot/ directory,
188 * used by reroot code in init(8) as a mountpoint for tmpfs.
191 reroot_conf(void *unused)
196 error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev,
197 &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot");
199 printf("%s: failed to create device node, error %d",
204 SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL);
207 * The system call that results in a reboot.
211 sys_reboot(struct thread *td, struct reboot_args *uap)
217 error = mac_system_check_reboot(td->td_ucred, uap->opt);
220 error = priv_check(td, PRIV_REBOOT);
222 if (uap->opt & RB_REROOT) {
223 error = kern_reroot();
226 kern_reboot(uap->opt);
234 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
237 shutdown_nice(int howto)
240 if (initproc != NULL) {
241 /* Send a signal to init(8) and have it shutdown the world. */
243 if (howto & RB_POWEROFF)
244 kern_psignal(initproc, SIGUSR2);
245 else if (howto & RB_HALT)
246 kern_psignal(initproc, SIGUSR1);
248 kern_psignal(initproc, SIGINT);
249 PROC_UNLOCK(initproc);
251 /* No init(8) running, so simply reboot. */
252 kern_reboot(howto | RB_NOSYNC);
265 if (ts.tv_sec >= 86400) {
266 printf("%ldd", (long)ts.tv_sec / 86400);
270 if (f || ts.tv_sec >= 3600) {
271 printf("%ldh", (long)ts.tv_sec / 3600);
275 if (f || ts.tv_sec >= 60) {
276 printf("%ldm", (long)ts.tv_sec / 60);
280 printf("%lds\n", (long)ts.tv_sec);
284 doadump(boolean_t textdump)
292 if (dumper.dumper == NULL)
296 dumptid = curthread->td_tid;
301 if (textdump && textdump_pending) {
303 textdump_dumpsys(&dumper);
307 error = dumpsys(&dumper);
314 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
317 kern_reboot(int howto)
323 * Bind us to CPU 0 so that all shutdown code runs there. Some
324 * systems don't shutdown properly (i.e., ACPI power off) if we
325 * run on another processor.
327 if (!SCHEDULER_STOPPED()) {
328 thread_lock(curthread);
329 sched_bind(curthread, 0);
330 thread_unlock(curthread);
331 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
334 /* We're in the process of rebooting. */
337 /* We are out of the debugger now. */
341 * Do any callouts that should be done BEFORE syncing the filesystems.
343 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
346 * Now sync filesystems
348 if (!cold && (howto & RB_NOSYNC) == 0 && once == 0) {
350 bufshutdown(show_busybufs);
358 * Ok, now do things that assume all filesystem activity has
361 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
363 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
366 /* Now that we're going to really halt the system... */
367 EVENTHANDLER_INVOKE(shutdown_final, howto);
369 for(;;) ; /* safety against shutdown_reset not working */
374 * The system call that results in changing the rootfs.
379 struct vnode *oldrootvnode, *vp;
380 struct mount *mp, *devmp;
383 if (curproc != initproc)
387 * Mark the filesystem containing currently-running executable
388 * (the temporary copy of init(8)) busy.
390 vp = curproc->p_textvp;
391 error = vn_lock(vp, LK_SHARED);
395 error = vfs_busy(mp, MBF_NOWAIT);
399 error = vfs_busy(mp, 0);
400 vn_lock(vp, LK_SHARED | LK_RETRY);
406 if (vp->v_iflag & VI_DOOMED) {
415 * Remove the filesystem containing currently-running executable
416 * from the mount list, to prevent it from being unmounted
417 * by vfs_unmountall(), and to avoid confusing vfs_mountroot().
419 * Also preserve /dev - forcibly unmounting it could cause driver
427 mtx_lock(&mountlist_mtx);
428 TAILQ_REMOVE(&mountlist, mp, mnt_list);
429 TAILQ_REMOVE(&mountlist, devmp, mnt_list);
430 mtx_unlock(&mountlist_mtx);
432 oldrootvnode = rootvnode;
435 * Unmount everything except for the two filesystems preserved above.
440 * Add /dev back; vfs_mountroot() will move it into its new place.
442 mtx_lock(&mountlist_mtx);
443 TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list);
444 mtx_unlock(&mountlist_mtx);
449 * Mount the new rootfs.
454 * Update all references to the old rootvnode.
456 mountcheckdirs(oldrootvnode, rootvnode);
459 * Add the temporary filesystem back and unbusy it.
461 mtx_lock(&mountlist_mtx);
462 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
463 mtx_unlock(&mountlist_mtx);
470 * If the shutdown was a clean halt, behave accordingly.
473 shutdown_halt(void *junk, int howto)
476 if (howto & RB_HALT) {
478 printf("The operating system has halted.\n");
479 printf("Please press any key to reboot.\n\n");
481 case -1: /* No console, just die */
492 * Check to see if the system paniced, pause and then reboot
493 * according to the specified delay.
496 shutdown_panic(void *junk, int howto)
500 if (howto & RB_DUMP) {
501 if (panic_reboot_wait_time != 0) {
502 if (panic_reboot_wait_time != -1) {
503 printf("Automatic reboot in %d seconds - "
504 "press a key on the console to abort\n",
505 panic_reboot_wait_time);
506 for (loop = panic_reboot_wait_time * 10;
508 DELAY(1000 * 100); /* 1/10th second */
509 /* Did user type a key? */
510 if (cncheckc() != -1)
516 } else { /* zero time specified - reboot NOW */
519 printf("--> Press a key on the console to reboot,\n");
520 printf("--> or switch off the system now.\n");
526 * Everything done, now reset
529 shutdown_reset(void *junk, int howto)
532 printf("Rebooting...\n");
533 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
536 * Acquiring smp_ipi_mtx here has a double effect:
537 * - it disables interrupts avoiding CPU0 preemption
538 * by fast handlers (thus deadlocking against other CPUs)
539 * - it avoids deadlocks against smp_rendezvous() or, more
540 * generally, threads busy-waiting, with this spinlock held,
541 * and waiting for responses by threads on other CPUs
542 * (ie. smp_tlb_shootdown()).
544 * For the !SMP case it just needs to handle the former problem.
547 mtx_lock_spin(&smp_ipi_mtx);
552 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
554 /* NOTREACHED */ /* assuming reset worked */
557 #if defined(WITNESS) || defined(INVARIANTS)
558 static int kassert_warn_only = 0;
560 static int kassert_do_kdb = 0;
563 static int kassert_do_ktr = 0;
565 static int kassert_do_log = 1;
566 static int kassert_log_pps_limit = 4;
567 static int kassert_log_mute_at = 0;
568 static int kassert_log_panic_at = 0;
569 static int kassert_warnings = 0;
571 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
573 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
574 &kassert_warn_only, 0,
575 "KASSERT triggers a panic (1) or just a warning (0)");
578 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
579 &kassert_do_kdb, 0, "KASSERT will enter the debugger");
583 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
585 "KASSERT does a KTR, set this to the KTRMASK you want");
588 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
589 &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
591 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
592 &kassert_warnings, 0, "number of KASSERTs that have been triggered");
594 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
595 &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
597 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
598 &kassert_log_pps_limit, 0, "limit number of log messages per second");
600 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
601 &kassert_log_mute_at, 0, "max number of KASSERTS to log");
603 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
605 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
606 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
607 kassert_sysctl_kassert, "I", "set to trigger a test kassert");
610 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
614 error = sysctl_wire_old_buffer(req, sizeof(int));
617 error = sysctl_handle_int(oidp, &i, 0, req);
619 if (error != 0 || req->newptr == NULL)
621 KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
626 * Called by KASSERT, this decides if we will panic
627 * or if we will log via printf and/or ktr.
630 kassert_panic(const char *fmt, ...)
632 static char buf[256];
636 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
640 * panic if we're not just warning, or if we've exceeded
641 * kassert_log_panic_at warnings.
643 if (!kassert_warn_only ||
644 (kassert_log_panic_at > 0 &&
645 kassert_warnings >= kassert_log_panic_at)) {
655 * log if we've not yet met the mute limit.
657 if (kassert_do_log &&
658 (kassert_log_mute_at == 0 ||
659 kassert_warnings < kassert_log_mute_at)) {
660 static struct timeval lasterr;
663 if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
664 printf("KASSERT failed: %s\n", buf);
669 if (kassert_do_kdb) {
670 kdb_enter(KDB_WHY_KASSERT, buf);
673 atomic_add_int(&kassert_warnings, 1);
678 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
679 * and then reboots. If we are called twice, then we avoid trying to sync
680 * the disks as this often leads to recursive panics.
683 panic(const char *fmt, ...)
692 vpanic(const char *fmt, va_list ap)
697 struct thread *td = curthread;
698 int bootopt, newpanic;
699 static char buf[256];
705 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
706 * concurrently entering panic. Only the winner will proceed
709 if (panicstr == NULL && !kdb_active) {
710 other_cpus = all_cpus;
711 CPU_CLR(PCPU_GET(cpuid), &other_cpus);
712 stop_cpus_hard(other_cpus);
716 * Ensure that the scheduler is stopped while panicking, even if panic
717 * has been entered from kdb.
719 td->td_stopsched = 1;
722 bootopt = RB_AUTOBOOT;
725 bootopt |= RB_NOSYNC;
733 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
736 printf("panic: %s\n", buf);
743 printf("cpuid = %d\n", PCPU_GET(cpuid));
747 if (newpanic && trace_on_panic)
749 if (debugger_on_panic)
750 kdb_enter(KDB_WHY_PANIC, "panic");
752 /*thread_lock(td); */
753 td->td_flags |= TDF_INPANIC;
754 /* thread_unlock(td); */
756 bootopt |= RB_NOSYNC;
757 kern_reboot(bootopt);
761 * Support for poweroff delay.
763 * Please note that setting this delay too short might power off your machine
764 * before the write cache on your hard disk has been flushed, leading to
765 * soft-updates inconsistencies.
767 #ifndef POWEROFF_DELAY
768 # define POWEROFF_DELAY 5000
770 static int poweroff_delay = POWEROFF_DELAY;
772 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
773 &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
776 poweroff_wait(void *junk, int howto)
779 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
781 DELAY(poweroff_delay * 1000);
785 * Some system processes (e.g. syncer) need to be stopped at appropriate
786 * points in their main loops prior to a system shutdown, so that they
787 * won't interfere with the shutdown process (e.g. by holding a disk buf
788 * to cause sync to fail). For each of these system processes, register
789 * shutdown_kproc() as a handler for one of shutdown events.
791 static int kproc_shutdown_wait = 60;
792 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
793 &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
796 kproc_shutdown(void *arg, int howto)
804 p = (struct proc *)arg;
805 printf("Waiting (max %d seconds) for system process `%s' to stop...",
806 kproc_shutdown_wait, p->p_comm);
807 error = kproc_suspend(p, kproc_shutdown_wait * hz);
809 if (error == EWOULDBLOCK)
810 printf("timed out\n");
816 kthread_shutdown(void *arg, int howto)
824 td = (struct thread *)arg;
825 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
826 kproc_shutdown_wait, td->td_name);
827 error = kthread_suspend(td, kproc_shutdown_wait * hz);
829 if (error == EWOULDBLOCK)
830 printf("timed out\n");
835 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
836 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
837 dumpdevname, 0, "Device for kernel dumps");
839 /* Registration of dumpers */
841 set_dumper(struct dumperinfo *di, const char *devname, struct thread *td)
846 error = priv_check(td, PRIV_SETDUMPER);
851 bzero(&dumper, sizeof dumper);
852 dumpdevname[0] = '\0';
855 if (dumper.dumper != NULL)
858 wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
859 if (wantcopy >= sizeof(dumpdevname)) {
860 printf("set_dumper: device name truncated from '%s' -> '%s'\n",
861 devname, dumpdevname);
866 /* Call dumper with bounds checking. */
868 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
869 off_t offset, size_t length)
872 if (length != 0 && (offset < di->mediaoffset ||
873 offset - di->mediaoffset + length > di->mediasize)) {
874 printf("Attempt to write outside dump device boundaries.\n"
875 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
876 (intmax_t)offset, (intmax_t)di->mediaoffset,
877 (uintmax_t)length, (intmax_t)di->mediasize);
880 return (di->dumper(di->priv, virtual, physical, offset, length));
884 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
885 uint64_t dumplen, uint32_t blksz)
888 bzero(kdh, sizeof(*kdh));
889 strlcpy(kdh->magic, magic, sizeof(kdh->magic));
890 strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
891 kdh->version = htod32(KERNELDUMPVERSION);
892 kdh->architectureversion = htod32(archver);
893 kdh->dumplength = htod64(dumplen);
894 kdh->dumptime = htod64(time_second);
895 kdh->blocksize = htod32(blksz);
896 strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
897 strlcpy(kdh->versionstring, version, sizeof(kdh->versionstring));
898 if (panicstr != NULL)
899 strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
900 kdh->parity = kerneldump_parity(kdh);