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 1.62 2008/01/05 13:23:48 corecode 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>
54 #include <sys/diskslice.h>
55 #include <sys/reboot.h>
58 #include <sys/fcntl.h> /* FREAD */
59 #include <sys/stat.h> /* S_IFCHR */
60 #include <sys/vnode.h>
61 #include <sys/kernel.h>
62 #include <sys/kerneldump.h>
63 #include <sys/kthread.h>
64 #include <sys/malloc.h>
65 #include <sys/mount.h>
66 #include <sys/queue.h>
67 #include <sys/sysctl.h>
68 #include <sys/vkernel.h>
70 #include <sys/sysproto.h>
71 #include <sys/device.h>
74 #include <sys/kerneldump.h>
75 #include <sys/kern_syscall.h>
76 #include <vm/vm_map.h>
79 #include <sys/thread2.h>
81 #include <sys/mplock2.h>
83 #include <machine/clock.h>
84 #include <machine/md_var.h>
85 #include <machine/smp.h> /* smp_active_mask, cpuid */
86 #include <machine/vmparam.h>
87 #include <machine/thread.h>
89 #include <sys/signalvar.h>
92 #include <dev/misc/gpio/gpio.h>
94 #ifndef PANIC_REBOOT_WAIT_TIME
95 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
99 * Note that stdarg.h and the ANSI style va_start macro is used for both
100 * ANSI and traditional C compilers. We use the machine version to stay
101 * within the confines of the kernel header files.
103 #include <machine/stdarg.h>
107 #ifdef DDB_UNATTENDED
108 int debugger_on_panic
= 0;
110 int debugger_on_panic
= 1;
112 SYSCTL_INT(_debug
, OID_AUTO
, debugger_on_panic
, CTLFLAG_RW
,
113 &debugger_on_panic
, 0, "Run debugger on kernel panic");
116 int trace_on_panic
= 1;
118 int trace_on_panic
= 0;
120 SYSCTL_INT(_debug
, OID_AUTO
, trace_on_panic
, CTLFLAG_RW
,
121 &trace_on_panic
, 0, "Print stack trace on kernel panic");
124 static int sync_on_panic
= 0;
125 SYSCTL_INT(_kern
, OID_AUTO
, sync_on_panic
, CTLFLAG_RW
,
126 &sync_on_panic
, 0, "Do a sync before rebooting from a panic");
128 SYSCTL_NODE(_kern
, OID_AUTO
, shutdown
, CTLFLAG_RW
, 0, "Shutdown environment");
132 * If there is a hardware watchdog, point this at the function needed to
134 * It's needed when the kernel needs to do some lengthy operations.
135 * e.g. in wd.c when dumping core.. It's most annoying to have
136 * your precious core-dump only half written because the wdog kicked in.
138 watchdog_tickle_fn wdog_tickler
= NULL
;
142 * Variable panicstr contains argument to first call to panic; used as flag
143 * to indicate that the kernel has already called panic.
145 const char *panicstr
;
147 int dumping
; /* system is dumping */
148 static struct dumperinfo dumper
; /* selected dumper */
151 u_int panic_cpu_interlock
; /* panic interlock */
152 globaldata_t panic_cpu_gd
; /* which cpu took the panic */
155 int bootverbose
= 0; /* note: assignment to force non-bss */
156 SYSCTL_INT(_debug
, OID_AUTO
, bootverbose
, CTLFLAG_RW
,
157 &bootverbose
, 0, "Verbose kernel messages");
159 int cold
= 1; /* note: assignment to force non-bss */
160 int dumplo
; /* OBSOLETE - savecore compat */
163 static void boot (int) __dead2
;
164 static int setdumpdev (cdev_t dev
);
165 static void poweroff_wait (void *, int);
166 static void print_uptime (void);
167 static void shutdown_halt (void *junk
, int howto
);
168 static void shutdown_panic (void *junk
, int howto
);
169 static void shutdown_reset (void *junk
, int howto
);
170 static int shutdown_busycount1(struct buf
*bp
, void *info
);
171 static int shutdown_busycount2(struct buf
*bp
, void *info
);
172 static void shutdown_cleanup_proc(struct proc
*p
);
174 /* register various local shutdown events */
176 shutdown_conf(void *unused
)
178 EVENTHANDLER_REGISTER(shutdown_final
, poweroff_wait
, NULL
, SHUTDOWN_PRI_FIRST
);
179 EVENTHANDLER_REGISTER(shutdown_final
, shutdown_halt
, NULL
, SHUTDOWN_PRI_LAST
+ 100);
180 EVENTHANDLER_REGISTER(shutdown_final
, shutdown_panic
, NULL
, SHUTDOWN_PRI_LAST
+ 100);
181 EVENTHANDLER_REGISTER(shutdown_final
, shutdown_reset
, NULL
, SHUTDOWN_PRI_LAST
+ 200);
184 SYSINIT(shutdown_conf
, SI_BOOT2_MACHDEP
, SI_ORDER_ANY
, shutdown_conf
, NULL
)
189 * The system call that results in a reboot
194 sys_reboot(struct reboot_args
*uap
)
196 struct thread
*td
= curthread
;
199 if ((error
= priv_check(td
, PRIV_REBOOT
)))
209 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
211 static int shutdown_howto
= 0;
214 shutdown_nice(int howto
)
216 shutdown_howto
= howto
;
218 /* Send a signal to init(8) and have it shutdown the world */
219 if (initproc
!= NULL
) {
220 ksignal(initproc
, SIGINT
);
222 /* No init(8) running, so simply reboot */
227 static int waittime
= -1;
229 struct thread
*dumpthread
;
240 if (ts
.tv_sec
>= 86400) {
241 kprintf("%ldd", ts
.tv_sec
/ 86400);
245 if (f
|| ts
.tv_sec
>= 3600) {
246 kprintf("%ldh", ts
.tv_sec
/ 3600);
250 if (f
|| ts
.tv_sec
>= 60) {
251 kprintf("%ldm", ts
.tv_sec
/ 60);
255 kprintf("%lds\n", ts
.tv_sec
);
259 * Go through the rigmarole of shutting down..
260 * this used to be in machdep.c but I'll be dammned if I could see
261 * anything machine dependant in it.
267 * Get rid of any user scheduler baggage and then give
268 * us a high priority.
270 if (curthread
->td_release
)
271 curthread
->td_release(curthread
);
272 lwkt_setpri_self(TDPRI_MAX
);
274 /* collect extra flags that shutdown_nice might have set */
275 howto
|= shutdown_howto
;
279 * We really want to shutdown on the BSP. Subsystems such as ACPI
280 * can't power-down the box otherwise.
282 if (smp_active_mask
> 1) {
283 kprintf("boot() called on cpu#%d\n", mycpu
->gd_cpuid
);
285 if (panicstr
== NULL
&& mycpu
->gd_cpuid
!= 0) {
286 kprintf("Switching to cpu #0 for shutdown\n");
287 lwkt_setcpu_self(globaldata_find(0));
291 * Do any callouts that should be done BEFORE syncing the filesystems.
293 EVENTHANDLER_INVOKE(shutdown_pre_sync
, howto
);
296 * Try to get rid of any remaining FS references. The calling
297 * process, proc0, and init may still hold references. The
298 * VFS cache subsystem may still hold a root reference to root.
300 * XXX this needs work. We really need to SIGSTOP all remaining
301 * processes in order to avoid blowups due to proc0's filesystem
302 * references going away. For now just make sure that the init
303 * process is stopped.
305 if (panicstr
== NULL
) {
306 shutdown_cleanup_proc(curproc
);
307 shutdown_cleanup_proc(&proc0
);
309 if (initproc
!= curproc
) {
310 ksignal(initproc
, SIGSTOP
);
311 tsleep(boot
, 0, "shutdn", hz
/ 20);
313 shutdown_cleanup_proc(initproc
);
315 vfs_cache_setroot(NULL
, NULL
);
319 * Now sync filesystems
321 if (!cold
&& (howto
& RB_NOSYNC
) == 0 && waittime
< 0) {
322 int iter
, nbusy
, pbusy
;
325 kprintf("\nsyncing disks... ");
327 sys_sync(NULL
); /* YYY was sync(&proc0, NULL). why proc0 ? */
330 * With soft updates, some buffers that are
331 * written will be remarked as dirty until other
332 * buffers are written.
334 for (iter
= pbusy
= 0; iter
< 20; iter
++) {
335 nbusy
= scan_all_buffers(shutdown_busycount1
, NULL
);
338 kprintf("%d ", nbusy
);
344 * Process soft update work queue if buffers don't sync
345 * after 6 iterations by permitting the syncer to run.
350 sys_sync(NULL
); /* YYY was sync(&proc0, NULL). why proc0 ? */
351 tsleep(boot
, 0, "shutdn", hz
* iter
/ 20 + 1);
355 * Count only busy local buffers to prevent forcing
356 * a fsck if we're just a client of a wedged NFS server
358 nbusy
= scan_all_buffers(shutdown_busycount2
, NULL
);
361 * Failed to sync all blocks. Indicate this and don't
362 * unmount filesystems (thus forcing an fsck on reboot).
364 kprintf("giving up on %d buffers\n", nbusy
);
366 Debugger("busy buffer problem");
368 tsleep(boot
, 0, "shutdn", hz
* 5 + 1);
372 * Unmount filesystems
374 if (panicstr
== NULL
)
377 tsleep(boot
, 0, "shutdn", hz
/ 10 + 1);
383 * Dump before doing post_sync shutdown ops
386 if ((howto
& (RB_HALT
|RB_DUMP
)) == RB_DUMP
&& !cold
) {
391 * Ok, now do things that assume all filesystem activity has
392 * been completed. This will also call the device shutdown
395 EVENTHANDLER_INVOKE(shutdown_post_sync
, howto
);
397 /* Now that we're going to really halt the system... */
398 EVENTHANDLER_INVOKE(shutdown_final
, howto
);
400 for(;;) ; /* safety against shutdown_reset not working */
405 shutdown_busycount1(struct buf
*bp
, void *info
)
407 if ((bp
->b_flags
& B_INVAL
) == 0 && BUF_REFCNT(bp
) > 0)
409 if ((bp
->b_flags
& (B_DELWRI
| B_INVAL
)) == B_DELWRI
)
415 shutdown_busycount2(struct buf
*bp
, void *info
)
417 if (((bp
->b_flags
& B_INVAL
) == 0 && BUF_REFCNT(bp
)) ||
418 ((bp
->b_flags
& (B_DELWRI
|B_INVAL
)) == B_DELWRI
)) {
420 * Only count buffers undergoing write I/O
421 * on the related vnode.
423 if (bp
->b_vp
== NULL
||
424 bio_track_active(&bp
->b_vp
->v_track_write
) == 0) {
427 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
429 "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
431 bp
->b_flags
, bp
->b_loffset
,
432 bp
->b_bio2
.bio_offset
);
440 * If the shutdown was a clean halt, behave accordingly.
443 shutdown_halt(void *junk
, int howto
)
445 if (howto
& RB_HALT
) {
447 kprintf("The operating system has halted.\n");
448 #ifdef _KERNEL_VIRTUAL
451 kprintf("Please press any key to reboot.\n\n");
453 case -1: /* No console, just die */
465 * Check to see if the system paniced, pause and then reboot
466 * according to the specified delay.
469 shutdown_panic(void *junk
, int howto
)
473 if (howto
& RB_DUMP
) {
474 if (PANIC_REBOOT_WAIT_TIME
!= 0) {
475 if (PANIC_REBOOT_WAIT_TIME
!= -1) {
476 kprintf("Automatic reboot in %d seconds - "
477 "press a key on the console to abort\n",
478 PANIC_REBOOT_WAIT_TIME
);
479 for (loop
= PANIC_REBOOT_WAIT_TIME
* 10;
481 DELAY(1000 * 100); /* 1/10th second */
482 /* Did user type a key? */
483 if (cncheckc() != -1)
489 } else { /* zero time specified - reboot NOW */
492 kprintf("--> Press a key on the console to reboot,\n");
493 kprintf("--> or switch off the system now.\n");
499 * Everything done, now reset
502 shutdown_reset(void *junk
, int howto
)
504 kprintf("Rebooting...\n");
505 DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */
506 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
508 /* NOTREACHED */ /* assuming reset worked */
512 * Try to remove FS references in the specified process. This function
513 * is used during shutdown
517 shutdown_cleanup_proc(struct proc
*p
)
519 struct filedesc
*fdp
;
524 if ((fdp
= p
->p_fd
) != NULL
) {
527 cache_drop(&fdp
->fd_ncdir
);
532 cache_drop(&fdp
->fd_nrdir
);
537 cache_drop(&fdp
->fd_njdir
);
550 pmap_remove_pages(vmspace_pmap(vm
),
552 VM_MAX_USER_ADDRESS
);
553 vm_map_remove(&vm
->vm_map
,
555 VM_MAX_USER_ADDRESS
);
560 * Magic number for savecore
562 * exported (symorder) and used at least by savecore(8)
564 * Mark it as used so that gcc doesn't optimize it away.
566 __attribute__((__used__
))
567 static u_long
const dumpmag
= 0x8fca0101UL
;
569 __attribute__((__used__
))
570 static int dumpsize
= 0; /* also for savecore */
572 static int dodump
= 1;
574 SYSCTL_INT(_machdep
, OID_AUTO
, do_dump
, CTLFLAG_RW
, &dodump
, 0,
575 "Try to perform coredump on kernel panic");
578 mkdumpheader(struct kerneldumpheader
*kdh
, char *magic
, uint32_t archver
,
579 uint64_t dumplen
, uint32_t blksz
)
581 bzero(kdh
, sizeof(*kdh
));
582 strncpy(kdh
->magic
, magic
, sizeof(kdh
->magic
));
583 strncpy(kdh
->architecture
, MACHINE_ARCH
, sizeof(kdh
->architecture
));
584 kdh
->version
= htod32(KERNELDUMPVERSION
);
585 kdh
->architectureversion
= htod32(archver
);
586 kdh
->dumplength
= htod64(dumplen
);
587 kdh
->dumptime
= htod64(time_second
);
588 kdh
->blocksize
= htod32(blksz
);
589 strncpy(kdh
->hostname
, hostname
, sizeof(kdh
->hostname
));
590 strncpy(kdh
->versionstring
, version
, sizeof(kdh
->versionstring
));
591 if (panicstr
!= NULL
)
592 strncpy(kdh
->panicstring
, panicstr
, sizeof(kdh
->panicstring
));
593 kdh
->parity
= kerneldump_parity(kdh
);
597 setdumpdev(cdev_t dev
)
603 disk_dumpconf(NULL
, 0/*off*/);
608 * We have to open the device before we can perform ioctls on it,
609 * or the slice/label data may not be present. Device opens are
610 * usually tracked by specfs, but the dump device can be set in
611 * early boot and may not be open so this is somewhat of a hack.
613 doopen
= (dev
->si_sysref
.refcnt
== 1);
615 error
= dev_dopen(dev
, FREAD
, S_IFCHR
, proc0
.p_ucred
);
619 error
= disk_dumpconf(dev
, 1/*on*/);
625 static void dump_conf (void *dummy
);
627 dump_conf(void *dummy
)
632 path
= kmalloc(MNAMELEN
, M_TEMP
, M_WAITOK
);
633 if (TUNABLE_STR_FETCH("dumpdev", path
, MNAMELEN
) != 0) {
634 dev
= kgetdiskbyname(path
);
639 if (setdumpdev(dumpdev
) != 0)
643 SYSINIT(dump_conf
, SI_SUB_DUMP_CONF
, SI_ORDER_FIRST
, dump_conf
, NULL
)
646 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS
)
651 ndumpdev
= dev2udev(dumpdev
);
652 error
= sysctl_handle_opaque(oidp
, &ndumpdev
, sizeof ndumpdev
, req
);
653 if (error
== 0 && req
->newptr
!= NULL
)
654 error
= setdumpdev(udev2dev(ndumpdev
, 0));
658 SYSCTL_PROC(_kern
, KERN_DUMPDEV
, dumpdev
, CTLTYPE_OPAQUE
|CTLFLAG_RW
,
659 0, sizeof dumpdev
, sysctl_kern_dumpdev
, "T,udev_t", "");
662 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
663 * and then reboots. If we are called twice, then we avoid trying to sync
664 * the disks as this often leads to recursive panics.
667 panic(const char *fmt
, ...)
669 int bootopt
, newpanic
;
671 static char buf
[256];
675 * If a panic occurs on multiple cpus before the first is able to
676 * halt the other cpus, only one cpu is allowed to take the panic.
677 * Attempt to be verbose about this situation but if the kprintf()
678 * itself panics don't let us overrun the kernel stack.
680 * Be very nasty about descheduling our thread at the lowest
681 * level possible in an attempt to freeze the thread without
682 * inducing further panics.
684 * Bumping gd_trap_nesting_level will also bypass assertions in
685 * lwkt_switch() and allow us to switch away even if we are a
686 * FAST interrupt or IPI.
688 if (atomic_poll_acquire_int(&panic_cpu_interlock
)) {
689 panic_cpu_gd
= mycpu
;
690 } else if (panic_cpu_gd
!= mycpu
) {
692 ++mycpu
->gd_trap_nesting_level
;
693 if (mycpu
->gd_trap_nesting_level
< 25) {
694 kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
695 mycpu
->gd_cpuid
, curthread
);
697 curthread
->td_release
= NULL
; /* be a grinch */
699 lwkt_deschedule_self(curthread
);
703 /* --mycpu->gd_trap_nesting_level */
707 bootopt
= RB_AUTOBOOT
| RB_DUMP
;
708 if (sync_on_panic
== 0)
709 bootopt
|= RB_NOSYNC
;
712 bootopt
|= RB_NOSYNC
;
719 kvsnprintf(buf
, sizeof(buf
), fmt
, ap
);
723 kprintf("panic: %s\n", buf
);
725 /* two separate prints in case of an unmapped page and trap */
726 kprintf("mp_lock = %08x; ", mp_lock
);
727 kprintf("cpuid = %d\n", mycpu
->gd_cpuid
);
730 #if (NGPIO > 0) && defined(ERROR_LED_ON_PANIC)
731 led_switch("error", 1);
734 #if defined(WDOG_DISABLE_ON_PANIC) && defined(WATCHDOG_ENABLE)
739 if (newpanic
&& trace_on_panic
)
741 if (debugger_on_panic
)
748 * Support for poweroff delay.
750 #ifndef POWEROFF_DELAY
751 # define POWEROFF_DELAY 5000
753 static int poweroff_delay
= POWEROFF_DELAY
;
755 SYSCTL_INT(_kern_shutdown
, OID_AUTO
, poweroff_delay
, CTLFLAG_RW
,
756 &poweroff_delay
, 0, "");
759 poweroff_wait(void *junk
, int howto
)
761 if(!(howto
& RB_POWEROFF
) || poweroff_delay
<= 0)
763 DELAY(poweroff_delay
* 1000);
767 * Some system processes (e.g. syncer) need to be stopped at appropriate
768 * points in their main loops prior to a system shutdown, so that they
769 * won't interfere with the shutdown process (e.g. by holding a disk buf
770 * to cause sync to fail). For each of these system processes, register
771 * shutdown_kproc() as a handler for one of shutdown events.
773 static int kproc_shutdown_wait
= 60;
774 SYSCTL_INT(_kern_shutdown
, OID_AUTO
, kproc_shutdown_wait
, CTLFLAG_RW
,
775 &kproc_shutdown_wait
, 0, "");
778 shutdown_kproc(void *arg
, int howto
)
787 td
= (struct thread
*)arg
;
788 if ((p
= td
->td_proc
) != NULL
) {
789 kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
790 kproc_shutdown_wait
, p
->p_comm
);
792 kprintf("Waiting (max %d seconds) for system thread %s to stop...",
793 kproc_shutdown_wait
, td
->td_comm
);
795 error
= suspend_kproc(td
, kproc_shutdown_wait
* hz
);
797 if (error
== EWOULDBLOCK
)
798 kprintf("timed out\n");
800 kprintf("stopped\n");
803 /* Registration of dumpers */
805 set_dumper(struct dumperinfo
*di
)
808 bzero(&dumper
, sizeof(dumper
));
812 if (dumper
.dumper
!= NULL
)
822 #if defined (_KERNEL_VIRTUAL)
823 /* VKERNELs don't support dumps */
824 kprintf("VKERNEL doesn't support dumps\n");
828 * If there is a dumper registered and we aren't dumping already, call
829 * the machine dependent dumpsys (md_dumpsys) to do the hard work.
831 * XXX: while right now the md_dumpsys() of x86 and x86_64 could be
832 * factored out completely into here, I rather keep them machine
833 * dependent in case we ever add a platform which does not share
834 * the same dumpsys() code, such as arm.
836 if (dumper
.dumper
!= NULL
&& !dumping
) {