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_show_busybufs.h"
44 #include "opt_sched.h"
45 #include "opt_watchdog.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
53 #include <sys/eventhandler.h>
56 #include <sys/kernel.h>
57 #include <sys/kerneldump.h>
58 #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/sched.h>
67 #include <sys/sysctl.h>
68 #include <sys/sysproto.h>
70 #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 */
94 * Note that stdarg.h and the ANSI style va_start macro is used for both
95 * ANSI and traditional C compilers.
97 #include <machine/stdarg.h>
100 #ifdef KDB_UNATTENDED
101 int debugger_on_panic = 0;
103 int debugger_on_panic = 1;
105 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
106 &debugger_on_panic, 0, "Run debugger on kernel panic");
107 TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
110 static int trace_on_panic = 1;
112 static int trace_on_panic = 0;
114 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
115 &trace_on_panic, 0, "Print stack trace on kernel panic");
116 TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
119 static int sync_on_panic = 0;
120 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
121 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
122 TUNABLE_INT("kern.sync_on_panic", &sync_on_panic);
124 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
127 * Variable panicstr contains argument to first call to panic; used as flag
128 * to indicate that the kernel has already called panic.
130 const char *panicstr;
132 int dumping; /* system is dumping */
133 int rebooting; /* system is rebooting */
134 static struct dumperinfo dumper; /* our selected dumper */
136 /* Context information for dump-debuggers. */
137 static struct pcb dumppcb; /* Registers. */
138 static lwpid_t dumptid; /* Thread ID. */
140 static void boot(int) __dead2;
141 static void poweroff_wait(void *, int);
142 static void shutdown_halt(void *junk, int howto);
143 static void shutdown_panic(void *junk, int howto);
144 static void shutdown_reset(void *junk, int howto);
146 /* register various local shutdown events */
148 shutdown_conf(void *unused)
151 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
153 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
154 SHUTDOWN_PRI_LAST + 100);
155 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
156 SHUTDOWN_PRI_LAST + 100);
157 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
158 SHUTDOWN_PRI_LAST + 200);
161 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
164 * The system call that results in a reboot.
168 reboot(struct thread *td, struct reboot_args *uap)
174 error = mac_system_check_reboot(td->td_ucred, uap->opt);
177 error = priv_check(td, PRIV_REBOOT);
187 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
189 static int shutdown_howto = 0;
192 shutdown_nice(int howto)
195 shutdown_howto = howto;
197 /* Send a signal to init(8) and have it shutdown the world */
198 if (initproc != NULL) {
200 psignal(initproc, SIGINT);
201 PROC_UNLOCK(initproc);
203 /* No init(8) running, so simply reboot */
208 static int waittime = -1;
219 if (ts.tv_sec >= 86400) {
220 printf("%ldd", (long)ts.tv_sec / 86400);
224 if (f || ts.tv_sec >= 3600) {
225 printf("%ldh", (long)ts.tv_sec / 3600);
229 if (f || ts.tv_sec >= 60) {
230 printf("%ldm", (long)ts.tv_sec / 60);
234 printf("%lds\n", (long)ts.tv_sec);
242 * Sometimes people have to call this from the kernel debugger.
243 * (if 'panic' can not dump)
244 * Give them a clue as to why they can't dump.
246 if (dumper.dumper == NULL) {
247 printf("Cannot dump. Device not defined or unavailable.\n");
252 dumptid = curthread->td_tid;
255 if (textdump_pending)
256 textdump_dumpsys(&dumper);
264 isbufbusy(struct buf *bp)
266 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
268 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
274 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
279 static int first_buf_printf = 1;
283 * Bind us to CPU 0 so that all shutdown code runs there. Some
284 * systems don't shutdown properly (i.e., ACPI power off) if we
285 * run on another processor.
287 thread_lock(curthread);
288 sched_bind(curthread, 0);
289 thread_unlock(curthread);
290 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
292 /* We're in the process of rebooting. */
295 /* collect extra flags that shutdown_nice might have set */
296 howto |= shutdown_howto;
298 /* We are out of the debugger now. */
302 * Do any callouts that should be done BEFORE syncing the filesystems.
304 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
307 * Now sync filesystems
309 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
310 register struct buf *bp;
311 int iter, nbusy, pbusy;
319 wdog_kern_pat(WD_LASTVAL);
321 sync(curthread, NULL);
324 * With soft updates, some buffers that are
325 * written will be remarked as dirty until other
326 * buffers are written.
328 for (iter = pbusy = 0; iter < 20; iter++) {
330 for (bp = &buf[nbuf]; --bp >= buf; )
334 if (first_buf_printf)
335 printf("All buffers synced.");
338 if (first_buf_printf) {
339 printf("Syncing disks, buffers remaining... ");
340 first_buf_printf = 0;
342 printf("%d ", nbusy);
347 wdog_kern_pat(WD_LASTVAL);
349 sync(curthread, NULL);
353 * Drop Giant and spin for a while to allow
354 * interrupt threads to run.
361 * Drop Giant and context switch several times to
362 * allow interrupt threads to run.
365 for (subiter = 0; subiter < 50 * iter; subiter++) {
366 thread_lock(curthread);
367 mi_switch(SW_VOL, NULL);
368 thread_unlock(curthread);
376 * Count only busy local buffers to prevent forcing
377 * a fsck if we're just a client of a wedged NFS server
380 for (bp = &buf[nbuf]; --bp >= buf; ) {
383 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
384 if (bp->b_dev == NULL) {
385 TAILQ_REMOVE(&mountlist,
386 bp->b_vp->v_mount, mnt_list);
391 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
393 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n",
395 bp->b_flags, (long)bp->b_blkno,
402 * Failed to sync all blocks. Indicate this and don't
403 * unmount filesystems (thus forcing an fsck on reboot).
405 printf("Giving up on %d buffers\n", nbusy);
406 DELAY(5000000); /* 5 seconds */
408 if (!first_buf_printf)
409 printf("Final sync complete\n");
411 * Unmount filesystems
417 DELAY(100000); /* wait for console output to finish */
423 * Ok, now do things that assume all filesystem activity has
426 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
428 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
431 /* Now that we're going to really halt the system... */
432 EVENTHANDLER_INVOKE(shutdown_final, howto);
434 for(;;) ; /* safety against shutdown_reset not working */
439 * If the shutdown was a clean halt, behave accordingly.
442 shutdown_halt(void *junk, int howto)
445 if (howto & RB_HALT) {
447 printf("The operating system has halted.\n");
448 printf("Please press any key to reboot.\n\n");
450 case -1: /* No console, just die */
461 * Check to see if the system paniced, pause and then reboot
462 * according to the specified delay.
465 shutdown_panic(void *junk, int howto)
469 if (howto & RB_DUMP) {
470 if (PANIC_REBOOT_WAIT_TIME != 0) {
471 if (PANIC_REBOOT_WAIT_TIME != -1) {
472 printf("Automatic reboot in %d seconds - "
473 "press a key on the console to abort\n",
474 PANIC_REBOOT_WAIT_TIME);
475 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
477 DELAY(1000 * 100); /* 1/10th second */
478 /* Did user type a key? */
479 if (cncheckc() != -1)
485 } else { /* zero time specified - reboot NOW */
488 printf("--> Press a key on the console to reboot,\n");
489 printf("--> or switch off the system now.\n");
495 * Everything done, now reset
498 shutdown_reset(void *junk, int howto)
501 printf("Rebooting...\n");
502 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
505 * Acquiring smp_ipi_mtx here has a double effect:
506 * - it disables interrupts avoiding CPU0 preemption
507 * by fast handlers (thus deadlocking against other CPUs)
508 * - it avoids deadlocks against smp_rendezvous() or, more
509 * generally, threads busy-waiting, with this spinlock held,
510 * and waiting for responses by threads on other CPUs
511 * (ie. smp_tlb_shootdown()).
513 * For the !SMP case it just needs to handle the former problem.
516 mtx_lock_spin(&smp_ipi_mtx);
521 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
523 /* NOTREACHED */ /* assuming reset worked */
527 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
528 * and then reboots. If we are called twice, then we avoid trying to sync
529 * the disks as this often leads to recursive panics.
532 panic(const char *fmt, ...)
535 static volatile u_int panic_cpu = NOCPU;
537 struct thread *td = curthread;
538 int bootopt, newpanic;
540 static char buf[256];
545 * We don't want multiple CPU's to panic at the same time, so we
546 * use panic_cpu as a simple spinlock. We have to keep checking
547 * panic_cpu if we are spinning in case the panic on the first
550 if (panic_cpu != PCPU_GET(cpuid))
551 while (atomic_cmpset_int(&panic_cpu, NOCPU,
552 PCPU_GET(cpuid)) == 0)
553 while (panic_cpu != NOCPU)
557 bootopt = RB_AUTOBOOT | RB_DUMP;
560 bootopt |= RB_NOSYNC;
568 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
570 printf("panic: %s\n", buf);
578 printf("cpuid = %d\n", PCPU_GET(cpuid));
582 if (newpanic && trace_on_panic)
584 if (debugger_on_panic)
585 kdb_enter(KDB_WHY_PANIC, "panic");
586 #ifdef RESTARTABLE_PANICS
587 /* See if the user aborted the panic, in which case we continue. */
588 if (panicstr == NULL) {
590 atomic_store_rel_int(&panic_cpu, NOCPU);
596 /*thread_lock(td); */
597 td->td_flags |= TDF_INPANIC;
598 /* thread_unlock(td); */
600 bootopt |= RB_NOSYNC;
606 * Support for poweroff delay.
608 #ifndef POWEROFF_DELAY
609 # define POWEROFF_DELAY 5000
611 static int poweroff_delay = POWEROFF_DELAY;
613 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
614 &poweroff_delay, 0, "");
617 poweroff_wait(void *junk, int howto)
620 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
622 DELAY(poweroff_delay * 1000);
626 * Some system processes (e.g. syncer) need to be stopped at appropriate
627 * points in their main loops prior to a system shutdown, so that they
628 * won't interfere with the shutdown process (e.g. by holding a disk buf
629 * to cause sync to fail). For each of these system processes, register
630 * shutdown_kproc() as a handler for one of shutdown events.
632 static int kproc_shutdown_wait = 60;
633 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
634 &kproc_shutdown_wait, 0, "");
637 kproc_shutdown(void *arg, int howto)
640 char procname[MAXCOMLEN + 1];
646 p = (struct proc *)arg;
647 strlcpy(procname, p->p_comm, sizeof(procname));
648 printf("Waiting (max %d seconds) for system process `%s' to stop...",
649 kproc_shutdown_wait, procname);
650 error = kproc_suspend(p, kproc_shutdown_wait * hz);
652 if (error == EWOULDBLOCK)
653 printf("timed out\n");
659 kthread_shutdown(void *arg, int howto)
662 char procname[MAXCOMLEN + 1];
668 td = (struct thread *)arg;
669 strlcpy(procname, td->td_name, sizeof(procname));
670 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
671 kproc_shutdown_wait, procname);
672 error = kthread_suspend(td, kproc_shutdown_wait * hz);
674 if (error == EWOULDBLOCK)
675 printf("timed out\n");
680 /* Registration of dumpers */
682 set_dumper(struct dumperinfo *di)
686 bzero(&dumper, sizeof dumper);
689 if (dumper.dumper != NULL)
695 /* Call dumper with bounds checking. */
697 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
698 off_t offset, size_t length)
701 if (length != 0 && (offset < di->mediaoffset ||
702 offset - di->mediaoffset + length > di->mediasize)) {
703 printf("Attempt to write outside dump device boundaries.\n");
706 return (di->dumper(di->priv, virtual, physical, offset, length));
710 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
711 uint64_t dumplen, uint32_t blksz)
714 bzero(kdh, sizeof(*kdh));
715 strncpy(kdh->magic, magic, sizeof(kdh->magic));
716 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
717 kdh->version = htod32(KERNELDUMPVERSION);
718 kdh->architectureversion = htod32(archver);
719 kdh->dumplength = htod64(dumplen);
720 kdh->dumptime = htod64(time_second);
721 kdh->blocksize = htod32(blksz);
722 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
723 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
724 if (panicstr != NULL)
725 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
726 kdh->parity = kerneldump_parity(kdh);