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 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)kern_shutdown.c 8.3 (Berkeley) 1/21/94
39 * $FreeBSD: src/sys/kern/kern_shutdown.c,v 1.72.2.12 2002/02/21 19:15:10 dillon Exp $
40 * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.60 2007/09/21 09:07:03 swildner Exp $
44 #include "opt_ddb_trace.h"
45 #include "opt_hw_wdog.h"
46 #include "opt_panic.h"
47 #include "opt_show_busybufs.h"
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/eventhandler.h>
53 #include <sys/diskslice.h>
54 #include <sys/reboot.h>
56 #include <sys/fcntl.h> /* FREAD */
57 #include <sys/stat.h> /* S_IFCHR */
58 #include <sys/vnode.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/malloc.h>
62 #include <sys/mount.h>
63 #include <sys/queue.h>
64 #include <sys/sysctl.h>
65 #include <sys/vkernel.h>
67 #include <sys/sysproto.h>
68 #include <sys/device.h>
71 #include <sys/kern_syscall.h>
72 #include <vm/vm_map.h>
75 #include <sys/thread2.h>
78 #include <machine/pcb.h>
79 #include <machine/clock.h>
80 #include <machine/md_var.h>
81 #include <machine/smp.h> /* smp_active_mask, cpuid */
82 #include <machine/vmparam.h>
84 #include <sys/signalvar.h>
86 #ifndef PANIC_REBOOT_WAIT_TIME
87 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
91 * Note that stdarg.h and the ANSI style va_start macro is used for both
92 * ANSI and traditional C compilers. We use the machine version to stay
93 * within the confines of the kernel header files.
95 #include <machine/stdarg.h>
99 int debugger_on_panic
= 0;
101 int debugger_on_panic
= 1;
103 SYSCTL_INT(_debug
, OID_AUTO
, debugger_on_panic
, CTLFLAG_RW
,
104 &debugger_on_panic
, 0, "Run debugger on kernel panic");
106 extern void db_print_backtrace(void);
109 int trace_on_panic
= 1;
111 int trace_on_panic
= 0;
113 SYSCTL_INT(_debug
, OID_AUTO
, trace_on_panic
, CTLFLAG_RW
,
114 &trace_on_panic
, 0, "Print stack trace on kernel panic");
117 static int sync_on_panic
= 1;
118 SYSCTL_INT(_kern
, OID_AUTO
, sync_on_panic
, CTLFLAG_RW
,
119 &sync_on_panic
, 0, "Do a sync before rebooting from a panic");
121 SYSCTL_NODE(_kern
, OID_AUTO
, shutdown
, CTLFLAG_RW
, 0, "Shutdown environment");
125 * If there is a hardware watchdog, point this at the function needed to
127 * It's needed when the kernel needs to do some lengthy operations.
128 * e.g. in wd.c when dumping core.. It's most annoying to have
129 * your precious core-dump only half written because the wdog kicked in.
131 watchdog_tickle_fn wdog_tickler
= NULL
;
135 * Variable panicstr contains argument to first call to panic; used as flag
136 * to indicate that the kernel has already called panic.
138 const char *panicstr
;
140 int dumping
; /* system is dumping */
142 u_int panic_cpu_interlock
; /* panic interlock */
143 globaldata_t panic_cpu_gd
; /* which cpu took the panic */
146 int bootverbose
= 0; /* note: assignment to force non-bss */
147 int cold
= 1; /* note: assignment to force non-bss */
148 int dumplo
; /* OBSOLETE - savecore compat */
151 static void boot (int) __dead2
;
152 static void dumpsys (void);
153 static int setdumpdev (cdev_t dev
);
154 static void poweroff_wait (void *, int);
155 static void print_uptime (void);
156 static void shutdown_halt (void *junk
, int howto
);
157 static void shutdown_panic (void *junk
, int howto
);
158 static void shutdown_reset (void *junk
, int howto
);
159 static int shutdown_busycount1(struct buf
*bp
, void *info
);
160 static int shutdown_busycount2(struct buf
*bp
, void *info
);
161 static void shutdown_cleanup_proc(struct proc
*p
);
163 /* register various local shutdown events */
165 shutdown_conf(void *unused
)
167 EVENTHANDLER_REGISTER(shutdown_final
, poweroff_wait
, NULL
, SHUTDOWN_PRI_FIRST
);
168 EVENTHANDLER_REGISTER(shutdown_final
, shutdown_halt
, NULL
, SHUTDOWN_PRI_LAST
+ 100);
169 EVENTHANDLER_REGISTER(shutdown_final
, shutdown_panic
, NULL
, SHUTDOWN_PRI_LAST
+ 100);
170 EVENTHANDLER_REGISTER(shutdown_final
, shutdown_reset
, NULL
, SHUTDOWN_PRI_LAST
+ 200);
173 SYSINIT(shutdown_conf
, SI_BOOT2_MACHDEP
, SI_ORDER_ANY
, shutdown_conf
, NULL
)
178 * The system call that results in a reboot
181 sys_reboot(struct reboot_args
*uap
)
183 struct thread
*td
= curthread
;
186 if ((error
= suser(td
)))
194 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
196 static int shutdown_howto
= 0;
199 shutdown_nice(int howto
)
201 shutdown_howto
= howto
;
203 /* Send a signal to init(8) and have it shutdown the world */
204 if (initproc
!= NULL
) {
205 ksignal(initproc
, SIGINT
);
207 /* No init(8) running, so simply reboot */
212 static int waittime
= -1;
213 static struct thread
*dumpthread
;
214 static struct pcb dumppcb
;
225 if (ts
.tv_sec
>= 86400) {
226 kprintf("%ldd", ts
.tv_sec
/ 86400);
230 if (f
|| ts
.tv_sec
>= 3600) {
231 kprintf("%ldh", ts
.tv_sec
/ 3600);
235 if (f
|| ts
.tv_sec
>= 60) {
236 kprintf("%ldm", ts
.tv_sec
/ 60);
240 kprintf("%lds\n", ts
.tv_sec
);
244 * Go through the rigmarole of shutting down..
245 * this used to be in machdep.c but I'll be dammned if I could see
246 * anything machine dependant in it.
252 * Get rid of any user scheduler baggage and then give
253 * us a high priority.
255 if (curthread
->td_release
)
256 curthread
->td_release(curthread
);
257 lwkt_setpri_self(TDPRI_MAX
);
259 /* collect extra flags that shutdown_nice might have set */
260 howto
|= shutdown_howto
;
264 * We really want to shutdown on the BSP. Subsystems such as ACPI
265 * can't power-down the box otherwise.
267 if (smp_active_mask
> 1) {
268 kprintf("boot() called on cpu#%d\n", mycpu
->gd_cpuid
);
270 if (panicstr
== NULL
&& mycpu
->gd_cpuid
!= 0) {
271 kprintf("Switching to cpu #0 for shutdown\n");
272 lwkt_setcpu_self(globaldata_find(0));
276 * Do any callouts that should be done BEFORE syncing the filesystems.
278 EVENTHANDLER_INVOKE(shutdown_pre_sync
, howto
);
281 * Try to get rid of any remaining FS references. The calling
282 * process, proc0, and init may still hold references. The
283 * VFS cache subsystem may still hold a root reference to root.
285 * XXX this needs work. We really need to SIGSTOP all remaining
286 * processes in order to avoid blowups due to proc0's filesystem
287 * references going away. For now just make sure that the init
288 * process is stopped.
290 if (panicstr
== NULL
) {
291 shutdown_cleanup_proc(curproc
);
292 shutdown_cleanup_proc(&proc0
);
294 if (initproc
!= curproc
) {
295 ksignal(initproc
, SIGSTOP
);
296 tsleep(boot
, 0, "shutdn", hz
/ 20);
298 shutdown_cleanup_proc(initproc
);
300 vfs_cache_setroot(NULL
, NULL
);
304 * Now sync filesystems
306 if (!cold
&& (howto
& RB_NOSYNC
) == 0 && waittime
< 0) {
307 int iter
, nbusy
, pbusy
;
310 kprintf("\nsyncing disks... ");
312 sys_sync(NULL
); /* YYY was sync(&proc0, NULL). why proc0 ? */
315 * With soft updates, some buffers that are
316 * written will be remarked as dirty until other
317 * buffers are written.
319 for (iter
= pbusy
= 0; iter
< 20; iter
++) {
320 nbusy
= scan_all_buffers(shutdown_busycount1
, NULL
);
323 kprintf("%d ", nbusy
);
329 * Process soft update work queue if buffers don't sync
330 * after 6 iterations by permitting the syncer to run.
332 if (iter
> 5 && bioops
.io_sync
)
333 (*bioops
.io_sync
)(NULL
);
334 sys_sync(NULL
); /* YYY was sync(&proc0, NULL). why proc0 ? */
335 tsleep(boot
, 0, "shutdn", hz
* iter
/ 20 + 1);
339 * Count only busy local buffers to prevent forcing
340 * a fsck if we're just a client of a wedged NFS server
342 nbusy
= scan_all_buffers(shutdown_busycount2
, NULL
);
345 * Failed to sync all blocks. Indicate this and don't
346 * unmount filesystems (thus forcing an fsck on reboot).
348 kprintf("giving up on %d buffers\n", nbusy
);
350 Debugger("busy buffer problem");
352 tsleep(boot
, 0, "shutdn", hz
* 5 + 1);
356 * Unmount filesystems
358 if (panicstr
== NULL
)
361 tsleep(boot
, 0, "shutdn", hz
/ 10 + 1);
367 * Dump before doing post_sync shutdown ops
370 if ((howto
& (RB_HALT
|RB_DUMP
)) == RB_DUMP
&& !cold
)
374 * Ok, now do things that assume all filesystem activity has
375 * been completed. This will also call the device shutdown
378 EVENTHANDLER_INVOKE(shutdown_post_sync
, howto
);
380 /* Now that we're going to really halt the system... */
381 EVENTHANDLER_INVOKE(shutdown_final
, howto
);
383 for(;;) ; /* safety against shutdown_reset not working */
388 shutdown_busycount1(struct buf
*bp
, void *info
)
390 if ((bp
->b_flags
& B_INVAL
) == 0 && BUF_REFCNT(bp
) > 0)
392 if ((bp
->b_flags
& (B_DELWRI
| B_INVAL
)) == B_DELWRI
)
398 shutdown_busycount2(struct buf
*bp
, void *info
)
400 if (((bp
->b_flags
& B_INVAL
) == 0 && BUF_REFCNT(bp
)) ||
401 ((bp
->b_flags
& (B_DELWRI
|B_INVAL
)) == B_DELWRI
)) {
403 * Only count buffers undergoing write I/O
404 * on the related vnode.
406 if (bp
->b_vp
== NULL
||
407 bp
->b_vp
->v_track_write
.bk_active
== 0) {
410 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
412 "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
414 bp
->b_flags
, bp
->b_loffset
,
415 bp
->b_bio2
.bio_offset
);
423 * If the shutdown was a clean halt, behave accordingly.
426 shutdown_halt(void *junk
, int howto
)
428 if (howto
& RB_HALT
) {
430 kprintf("The operating system has halted.\n");
431 #ifdef _KERNEL_VIRTUAL
434 kprintf("Please press any key to reboot.\n\n");
436 case -1: /* No console, just die */
448 * Check to see if the system paniced, pause and then reboot
449 * according to the specified delay.
452 shutdown_panic(void *junk
, int howto
)
456 if (howto
& RB_DUMP
) {
457 if (PANIC_REBOOT_WAIT_TIME
!= 0) {
458 if (PANIC_REBOOT_WAIT_TIME
!= -1) {
459 kprintf("Automatic reboot in %d seconds - "
460 "press a key on the console to abort\n",
461 PANIC_REBOOT_WAIT_TIME
);
462 for (loop
= PANIC_REBOOT_WAIT_TIME
* 10;
464 DELAY(1000 * 100); /* 1/10th second */
465 /* Did user type a key? */
466 if (cncheckc() != -1)
472 } else { /* zero time specified - reboot NOW */
475 kprintf("--> Press a key on the console to reboot,\n");
476 kprintf("--> or switch off the system now.\n");
482 * Everything done, now reset
485 shutdown_reset(void *junk
, int howto
)
487 kprintf("Rebooting...\n");
488 DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */
489 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
491 /* NOTREACHED */ /* assuming reset worked */
495 * Try to remove FS references in the specified process. This function
496 * is used during shutdown
500 shutdown_cleanup_proc(struct proc
*p
)
502 struct filedesc
*fdp
;
507 if ((fdp
= p
->p_fd
) != NULL
) {
510 cache_drop(&fdp
->fd_ncdir
);
515 cache_drop(&fdp
->fd_nrdir
);
520 cache_drop(&fdp
->fd_njdir
);
533 pmap_remove_pages(vmspace_pmap(vm
),
535 VM_MAX_USER_ADDRESS
);
536 vm_map_remove(&vm
->vm_map
,
538 VM_MAX_USER_ADDRESS
);
543 * Magic number for savecore
545 * exported (symorder) and used at least by savecore(8)
547 * Mark it as used so that gcc doesn't optimize it away.
549 __attribute__((__used__
))
550 static u_long
const dumpmag
= 0x8fca0101UL
;
552 static int dumpsize
= 0; /* also for savecore */
554 static int dodump
= 1;
556 SYSCTL_INT(_machdep
, OID_AUTO
, do_dump
, CTLFLAG_RW
, &dodump
, 0,
557 "Try to perform coredump on kernel panic");
560 setdumpdev(cdev_t dev
)
562 struct partinfo pinfo
;
571 bzero(&pinfo
, sizeof(pinfo
));
574 * We have to open the device before we can perform ioctls on it,
575 * or the slice/label data may not be present. Device opens are
576 * usually tracked by specfs, but the dump device can be set in
577 * early boot and may not be open so this is somewhat of a hack.
579 doopen
= (dev
->si_sysref
.refcnt
== 1);
581 error
= dev_dopen(dev
, FREAD
, S_IFCHR
, proc0
.p_ucred
);
585 error
= dev_dioctl(dev
, DIOCGPART
, (void *)&pinfo
, 0, proc0
.p_ucred
);
587 dev_dclose(dev
, FREAD
, S_IFCHR
);
588 if (error
|| pinfo
.media_blocks
== 0 || pinfo
.media_blksize
== 0)
591 newdumplo
= pinfo
.media_blocks
-
592 ((u_int64_t
)Maxmem
* PAGE_SIZE
/ DEV_BSIZE
);
593 if ((int64_t)newdumplo
< (int64_t)pinfo
.reserved_blocks
)
596 dumplo64
= newdumplo
;
602 static void dump_conf (void *dummy
);
604 dump_conf(void *dummy
)
609 path
= kmalloc(MNAMELEN
, M_TEMP
, M_WAITOK
);
610 if (TUNABLE_STR_FETCH("dumpdev", path
, MNAMELEN
) != 0) {
611 dev
= kgetdiskbyname(path
);
616 if (setdumpdev(dumpdev
) != 0)
620 SYSINIT(dump_conf
, SI_SUB_DUMP_CONF
, SI_ORDER_FIRST
, dump_conf
, NULL
)
623 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS
)
628 ndumpdev
= dev2udev(dumpdev
);
629 error
= sysctl_handle_opaque(oidp
, &ndumpdev
, sizeof ndumpdev
, req
);
630 if (error
== 0 && req
->newptr
!= NULL
)
631 error
= setdumpdev(udev2dev(ndumpdev
, 0));
635 SYSCTL_PROC(_kern
, KERN_DUMPDEV
, dumpdev
, CTLTYPE_OPAQUE
|CTLFLAG_RW
,
636 0, sizeof dumpdev
, sysctl_kern_dumpdev
, "T,udev_t", "");
639 * Doadump comes here after turning off memory management and
640 * getting on the dump stack, either when called above, or by
641 * the auto-restart code.
649 dumpthread
= curthread
;
651 kprintf("Dump already in progress, bailing...\n");
659 kprintf("\ndumping to dev %s, blockno %lld\n",
660 devtoname(dumpdev
), dumplo64
);
662 error
= dev_ddump(dumpdev
);
664 kprintf("succeeded\n");
667 kprintf("failed, reason: ");
671 kprintf("device doesn't support a dump routine\n");
675 kprintf("device bad\n");
679 kprintf("device not ready\n");
683 kprintf("area improper\n");
687 kprintf("i/o error\n");
691 kprintf("aborted from console\n");
695 kprintf("unknown, error = %d\n", error
);
701 dumpstatus(vm_offset_t addr
, off_t count
)
705 if (addr
% (1024 * 1024) == 0) {
710 kprintf("%ld ", (long)(count
/ (1024 * 1024)));
713 if ((c
= cncheckc()) == 0x03)
716 kprintf("[CTRL-C to abort] ");
722 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
723 * and then reboots. If we are called twice, then we avoid trying to sync
724 * the disks as this often leads to recursive panics.
727 panic(const char *fmt
, ...)
729 int bootopt
, newpanic
;
731 static char buf
[256];
735 * If a panic occurs on multiple cpus before the first is able to
736 * halt the other cpus, only one cpu is allowed to take the panic.
737 * Attempt to be verbose about this situation but if the kprintf()
738 * itself panics don't let us overrun the kernel stack.
740 * Be very nasty about descheduling our thread at the lowest
741 * level possible in an attempt to freeze the thread without
742 * inducing further panics.
744 * Bumping gd_trap_nesting_level will also bypass assertions in
745 * lwkt_switch() and allow us to switch away even if we are a
746 * FAST interrupt or IPI.
748 if (atomic_poll_acquire_int(&panic_cpu_interlock
)) {
749 panic_cpu_gd
= mycpu
;
750 } else if (panic_cpu_gd
!= mycpu
) {
752 ++mycpu
->gd_trap_nesting_level
;
753 if (mycpu
->gd_trap_nesting_level
< 25) {
754 kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
755 mycpu
->gd_cpuid
, curthread
);
757 curthread
->td_release
= NULL
; /* be a grinch */
759 lwkt_deschedule_self(curthread
);
763 /* --mycpu->gd_trap_nesting_level */
767 bootopt
= RB_AUTOBOOT
| RB_DUMP
;
768 if (sync_on_panic
== 0)
769 bootopt
|= RB_NOSYNC
;
772 bootopt
|= RB_NOSYNC
;
779 kvsnprintf(buf
, sizeof(buf
), fmt
, ap
);
783 kprintf("panic: %s\n", buf
);
785 /* two separate prints in case of an unmapped page and trap */
786 kprintf("mp_lock = %08x; ", mp_lock
);
787 kprintf("cpuid = %d\n", mycpu
->gd_cpuid
);
791 if (newpanic
&& trace_on_panic
)
792 db_print_backtrace();
793 if (debugger_on_panic
)
800 * Support for poweroff delay.
802 #ifndef POWEROFF_DELAY
803 # define POWEROFF_DELAY 5000
805 static int poweroff_delay
= POWEROFF_DELAY
;
807 SYSCTL_INT(_kern_shutdown
, OID_AUTO
, poweroff_delay
, CTLFLAG_RW
,
808 &poweroff_delay
, 0, "");
811 poweroff_wait(void *junk
, int howto
)
813 if(!(howto
& RB_POWEROFF
) || poweroff_delay
<= 0)
815 DELAY(poweroff_delay
* 1000);
819 * Some system processes (e.g. syncer) need to be stopped at appropriate
820 * points in their main loops prior to a system shutdown, so that they
821 * won't interfere with the shutdown process (e.g. by holding a disk buf
822 * to cause sync to fail). For each of these system processes, register
823 * shutdown_kproc() as a handler for one of shutdown events.
825 static int kproc_shutdown_wait
= 60;
826 SYSCTL_INT(_kern_shutdown
, OID_AUTO
, kproc_shutdown_wait
, CTLFLAG_RW
,
827 &kproc_shutdown_wait
, 0, "");
830 shutdown_kproc(void *arg
, int howto
)
839 td
= (struct thread
*)arg
;
840 if ((p
= td
->td_proc
) != NULL
) {
841 kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
842 kproc_shutdown_wait
, p
->p_comm
);
844 kprintf("Waiting (max %d seconds) for system thread %s to stop...",
845 kproc_shutdown_wait
, td
->td_comm
);
847 error
= suspend_kproc(td
, kproc_shutdown_wait
* hz
);
849 if (error
== EWOULDBLOCK
)
850 kprintf("timed out\n");
852 kprintf("stopped\n");