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/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 */
92 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME;
93 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN,
94 &panic_reboot_wait_time, 0,
95 "Seconds to wait before rebooting after a panic");
98 * Note that stdarg.h and the ANSI style va_start macro is used for both
99 * ANSI and traditional C compilers.
101 #include <machine/stdarg.h>
104 #ifdef KDB_UNATTENDED
105 int debugger_on_panic = 0;
107 int debugger_on_panic = 1;
109 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic,
110 CTLFLAG_RWTUN | CTLFLAG_SECURE,
111 &debugger_on_panic, 0, "Run debugger on kernel 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_RWTUN | CTLFLAG_SECURE,
120 &trace_on_panic, 0, "Print stack trace on kernel panic");
123 static int sync_on_panic = 0;
124 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN,
125 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
127 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0,
128 "Shutdown environment");
131 static int show_busybufs;
133 static int show_busybufs = 1;
135 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW,
136 &show_busybufs, 0, "");
139 * Variable panicstr contains argument to first call to panic; used as flag
140 * to indicate that the kernel has already called panic.
142 const char *panicstr;
144 int dumping; /* system is dumping */
145 int rebooting; /* system is rebooting */
146 static struct dumperinfo dumper; /* our selected dumper */
148 /* Context information for dump-debuggers. */
149 static struct pcb dumppcb; /* Registers. */
150 lwpid_t dumptid; /* Thread ID. */
152 static void poweroff_wait(void *, int);
153 static void shutdown_halt(void *junk, int howto);
154 static void shutdown_panic(void *junk, int howto);
155 static void shutdown_reset(void *junk, int howto);
156 static void vpanic(const char *fmt, va_list ap) __dead2;
158 /* register various local shutdown events */
160 shutdown_conf(void *unused)
163 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
165 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
166 SHUTDOWN_PRI_LAST + 100);
167 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
168 SHUTDOWN_PRI_LAST + 100);
169 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
170 SHUTDOWN_PRI_LAST + 200);
173 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
176 * The system call that results in a reboot.
180 sys_reboot(struct thread *td, struct reboot_args *uap)
186 error = mac_system_check_reboot(td->td_ucred, uap->opt);
189 error = priv_check(td, PRIV_REBOOT);
192 kern_reboot(uap->opt);
199 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
202 shutdown_nice(int howto)
205 if (initproc != NULL) {
206 /* Send a signal to init(8) and have it shutdown the world. */
208 if (howto & RB_POWEROFF)
209 kern_psignal(initproc, SIGUSR2);
210 else if (howto & RB_HALT)
211 kern_psignal(initproc, SIGUSR1);
213 kern_psignal(initproc, SIGINT);
214 PROC_UNLOCK(initproc);
216 /* No init(8) running, so simply reboot. */
217 kern_reboot(howto | RB_NOSYNC);
230 if (ts.tv_sec >= 86400) {
231 printf("%ldd", (long)ts.tv_sec / 86400);
235 if (f || ts.tv_sec >= 3600) {
236 printf("%ldh", (long)ts.tv_sec / 3600);
240 if (f || ts.tv_sec >= 60) {
241 printf("%ldm", (long)ts.tv_sec / 60);
245 printf("%lds\n", (long)ts.tv_sec);
249 doadump(boolean_t textdump)
257 if (dumper.dumper == NULL)
261 dumptid = curthread->td_tid;
266 if (textdump && textdump_pending) {
268 textdump_dumpsys(&dumper);
272 error = 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;
295 static int waittime = -1;
299 * Bind us to CPU 0 so that all shutdown code runs there. Some
300 * systems don't shutdown properly (i.e., ACPI power off) if we
301 * run on another processor.
303 if (!SCHEDULER_STOPPED()) {
304 thread_lock(curthread);
305 sched_bind(curthread, 0);
306 thread_unlock(curthread);
307 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
310 /* We're in the process of rebooting. */
313 /* We are out of the debugger now. */
317 * Do any callouts that should be done BEFORE syncing the filesystems.
319 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
322 * Now sync filesystems
324 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
325 register struct buf *bp;
326 int iter, nbusy, pbusy;
333 wdog_kern_pat(WD_LASTVAL);
334 sys_sync(curthread, NULL);
337 * With soft updates, some buffers that are
338 * written will be remarked as dirty until other
339 * buffers are written.
341 for (iter = pbusy = 0; iter < 20; iter++) {
343 for (bp = &buf[nbuf]; --bp >= buf; )
347 if (first_buf_printf)
348 printf("All buffers synced.");
351 if (first_buf_printf) {
352 printf("Syncing disks, buffers remaining... ");
353 first_buf_printf = 0;
355 printf("%d ", nbusy);
360 wdog_kern_pat(WD_LASTVAL);
361 sys_sync(curthread, NULL);
365 * Drop Giant and spin for a while to allow
366 * interrupt threads to run.
373 * Drop Giant and context switch several times to
374 * allow interrupt threads to run.
377 for (subiter = 0; subiter < 50 * iter; subiter++) {
378 thread_lock(curthread);
379 mi_switch(SW_VOL, NULL);
380 thread_unlock(curthread);
388 * Count only busy local buffers to prevent forcing
389 * a fsck if we're just a client of a wedged NFS server
392 for (bp = &buf[nbuf]; --bp >= buf; ) {
395 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
396 if (bp->b_dev == NULL) {
397 TAILQ_REMOVE(&mountlist,
398 bp->b_vp->v_mount, mnt_list);
403 if (show_busybufs > 0) {
405 "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:",
406 nbusy, bp, bp->b_vp, bp->b_flags,
407 (intmax_t)bp->b_blkno,
408 (intmax_t)bp->b_lblkno);
409 BUF_LOCKPRINTINFO(bp);
410 if (show_busybufs > 1)
418 * Failed to sync all blocks. Indicate this and don't
419 * unmount filesystems (thus forcing an fsck on reboot).
421 printf("Giving up on %d buffers\n", nbusy);
422 DELAY(5000000); /* 5 seconds */
424 if (!first_buf_printf)
425 printf("Final sync complete\n");
427 * Unmount filesystems
433 DELAY(100000); /* wait for console output to finish */
441 * Ok, now do things that assume all filesystem activity has
444 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
446 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
449 /* Now that we're going to really halt the system... */
450 EVENTHANDLER_INVOKE(shutdown_final, howto);
452 for(;;) ; /* safety against shutdown_reset not working */
457 * If the shutdown was a clean halt, behave accordingly.
460 shutdown_halt(void *junk, int howto)
463 if (howto & RB_HALT) {
465 printf("The operating system has halted.\n");
466 printf("Please press any key to reboot.\n\n");
468 case -1: /* No console, just die */
479 * Check to see if the system paniced, pause and then reboot
480 * according to the specified delay.
483 shutdown_panic(void *junk, int howto)
487 if (howto & RB_DUMP) {
488 if (panic_reboot_wait_time != 0) {
489 if (panic_reboot_wait_time != -1) {
490 printf("Automatic reboot in %d seconds - "
491 "press a key on the console to abort\n",
492 panic_reboot_wait_time);
493 for (loop = panic_reboot_wait_time * 10;
495 DELAY(1000 * 100); /* 1/10th second */
496 /* Did user type a key? */
497 if (cncheckc() != -1)
503 } else { /* zero time specified - reboot NOW */
506 printf("--> Press a key on the console to reboot,\n");
507 printf("--> or switch off the system now.\n");
513 * Everything done, now reset
516 shutdown_reset(void *junk, int howto)
519 printf("Rebooting...\n");
520 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
523 * Acquiring smp_ipi_mtx here has a double effect:
524 * - it disables interrupts avoiding CPU0 preemption
525 * by fast handlers (thus deadlocking against other CPUs)
526 * - it avoids deadlocks against smp_rendezvous() or, more
527 * generally, threads busy-waiting, with this spinlock held,
528 * and waiting for responses by threads on other CPUs
529 * (ie. smp_tlb_shootdown()).
531 * For the !SMP case it just needs to handle the former problem.
534 mtx_lock_spin(&smp_ipi_mtx);
539 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
541 /* NOTREACHED */ /* assuming reset worked */
544 #if defined(WITNESS) || defined(INVARIANTS)
545 static int kassert_warn_only = 0;
547 static int kassert_do_kdb = 0;
550 static int kassert_do_ktr = 0;
552 static int kassert_do_log = 1;
553 static int kassert_log_pps_limit = 4;
554 static int kassert_log_mute_at = 0;
555 static int kassert_log_panic_at = 0;
556 static int kassert_warnings = 0;
558 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options");
560 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN,
561 &kassert_warn_only, 0,
562 "KASSERT triggers a panic (1) or just a warning (0)");
565 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN,
566 &kassert_do_kdb, 0, "KASSERT will enter the debugger");
570 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN,
572 "KASSERT does a KTR, set this to the KTRMASK you want");
575 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN,
576 &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)");
578 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN,
579 &kassert_warnings, 0, "number of KASSERTs that have been triggered");
581 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN,
582 &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic");
584 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN,
585 &kassert_log_pps_limit, 0, "limit number of log messages per second");
587 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN,
588 &kassert_log_mute_at, 0, "max number of KASSERTS to log");
590 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS);
592 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert,
593 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0,
594 kassert_sysctl_kassert, "I", "set to trigger a test kassert");
597 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS)
601 error = sysctl_wire_old_buffer(req, sizeof(int));
604 error = sysctl_handle_int(oidp, &i, 0, req);
606 if (error != 0 || req->newptr == NULL)
608 KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i));
613 * Called by KASSERT, this decides if we will panic
614 * or if we will log via printf and/or ktr.
617 kassert_panic(const char *fmt, ...)
619 static char buf[256];
623 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
627 * panic if we're not just warning, or if we've exceeded
628 * kassert_log_panic_at warnings.
630 if (!kassert_warn_only ||
631 (kassert_log_panic_at > 0 &&
632 kassert_warnings >= kassert_log_panic_at)) {
642 * log if we've not yet met the mute limit.
644 if (kassert_do_log &&
645 (kassert_log_mute_at == 0 ||
646 kassert_warnings < kassert_log_mute_at)) {
647 static struct timeval lasterr;
650 if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) {
651 printf("KASSERT failed: %s\n", buf);
656 if (kassert_do_kdb) {
657 kdb_enter(KDB_WHY_KASSERT, buf);
660 atomic_add_int(&kassert_warnings, 1);
665 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
666 * and then reboots. If we are called twice, then we avoid trying to sync
667 * the disks as this often leads to recursive panics.
670 panic(const char *fmt, ...)
679 vpanic(const char *fmt, va_list ap)
684 struct thread *td = curthread;
685 int bootopt, newpanic;
686 static char buf[256];
692 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from
693 * concurrently entering panic. Only the winner will proceed
696 if (panicstr == NULL && !kdb_active) {
697 other_cpus = all_cpus;
698 CPU_CLR(PCPU_GET(cpuid), &other_cpus);
699 stop_cpus_hard(other_cpus);
703 * We set stop_scheduler here and not in the block above,
704 * because we want to ensure that if panic has been called and
705 * stop_scheduler_on_panic is true, then stop_scheduler will
706 * always be set. Even if panic has been entered from kdb.
708 td->td_stopsched = 1;
711 bootopt = RB_AUTOBOOT;
714 bootopt |= RB_NOSYNC;
722 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
725 printf("panic: %s\n", buf);
732 printf("cpuid = %d\n", PCPU_GET(cpuid));
736 if (newpanic && trace_on_panic)
738 if (debugger_on_panic)
739 kdb_enter(KDB_WHY_PANIC, "panic");
741 /*thread_lock(td); */
742 td->td_flags |= TDF_INPANIC;
743 /* thread_unlock(td); */
745 bootopt |= RB_NOSYNC;
746 kern_reboot(bootopt);
750 * Support for poweroff delay.
752 * Please note that setting this delay too short might power off your machine
753 * before the write cache on your hard disk has been flushed, leading to
754 * soft-updates inconsistencies.
756 #ifndef POWEROFF_DELAY
757 # define POWEROFF_DELAY 5000
759 static int poweroff_delay = POWEROFF_DELAY;
761 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
762 &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)");
765 poweroff_wait(void *junk, int howto)
768 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
770 DELAY(poweroff_delay * 1000);
774 * Some system processes (e.g. syncer) need to be stopped at appropriate
775 * points in their main loops prior to a system shutdown, so that they
776 * won't interfere with the shutdown process (e.g. by holding a disk buf
777 * to cause sync to fail). For each of these system processes, register
778 * shutdown_kproc() as a handler for one of shutdown events.
780 static int kproc_shutdown_wait = 60;
781 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
782 &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process");
785 kproc_shutdown(void *arg, int howto)
793 p = (struct proc *)arg;
794 printf("Waiting (max %d seconds) for system process `%s' to stop...",
795 kproc_shutdown_wait, p->p_comm);
796 error = kproc_suspend(p, kproc_shutdown_wait * hz);
798 if (error == EWOULDBLOCK)
799 printf("timed out\n");
805 kthread_shutdown(void *arg, int howto)
813 td = (struct thread *)arg;
814 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
815 kproc_shutdown_wait, td->td_name);
816 error = kthread_suspend(td, kproc_shutdown_wait * hz);
818 if (error == EWOULDBLOCK)
819 printf("timed out\n");
824 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)];
825 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD,
826 dumpdevname, 0, "Device for kernel dumps");
828 /* Registration of dumpers */
830 set_dumper(struct dumperinfo *di, const char *devname)
835 bzero(&dumper, sizeof dumper);
836 dumpdevname[0] = '\0';
839 if (dumper.dumper != NULL)
842 wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname));
843 if (wantcopy >= sizeof(dumpdevname)) {
844 printf("set_dumper: device name truncated from '%s' -> '%s'\n",
845 devname, dumpdevname);
850 /* Call dumper with bounds checking. */
852 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
853 off_t offset, size_t length)
856 if (length != 0 && (offset < di->mediaoffset ||
857 offset - di->mediaoffset + length > di->mediasize)) {
858 printf("Attempt to write outside dump device boundaries.\n"
859 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n",
860 (intmax_t)offset, (intmax_t)di->mediaoffset,
861 (uintmax_t)length, (intmax_t)di->mediasize);
864 return (di->dumper(di->priv, virtual, physical, offset, length));
868 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
869 uint64_t dumplen, uint32_t blksz)
872 bzero(kdh, sizeof(*kdh));
873 strncpy(kdh->magic, magic, sizeof(kdh->magic));
874 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
875 kdh->version = htod32(KERNELDUMPVERSION);
876 kdh->architectureversion = htod32(archver);
877 kdh->dumplength = htod64(dumplen);
878 kdh->dumptime = htod64(time_second);
879 kdh->blocksize = htod32(blksz);
880 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
881 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
882 if (panicstr != NULL)
883 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
884 kdh->parity = kerneldump_parity(kdh);