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 1.38 2006/09/24 19:43:55 dillon 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/disklabel.h>
54 #include <sys/reboot.h>
56 #include <sys/vnode.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/malloc.h>
60 #include <sys/mount.h>
61 #include <sys/queue.h>
62 #include <sys/sysctl.h>
63 #include <sys/vkernel.h>
65 #include <sys/sysproto.h>
66 #include <sys/device.h>
69 #include <sys/kern_syscall.h>
70 #include <vm/vm_map.h>
73 #include <sys/thread2.h>
76 #include <machine/pcb.h>
77 #include <machine/clock.h>
78 #include <machine/md_var.h>
79 #include <machine/smp.h> /* smp_active_mask, cpuid */
81 #include <sys/signalvar.h>
83 #ifndef PANIC_REBOOT_WAIT_TIME
84 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
88 * Note that stdarg.h and the ANSI style va_start macro is used for both
89 * ANSI and traditional C compilers. We use the machine version to stay
90 * within the confines of the kernel header files.
92 #include <machine/stdarg.h>
96 int debugger_on_panic = 0;
98 int debugger_on_panic = 1;
100 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
101 &debugger_on_panic, 0, "Run debugger on kernel panic");
103 extern void db_print_backtrace(void);
106 int trace_on_panic = 1;
108 int trace_on_panic = 0;
110 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
111 &trace_on_panic, 0, "Print stack trace on kernel panic");
114 static int sync_on_panic = 1;
115 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
116 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
118 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
122 * If there is a hardware watchdog, point this at the function needed to
124 * It's needed when the kernel needs to do some lengthy operations.
125 * e.g. in wd.c when dumping core.. It's most annoying to have
126 * your precious core-dump only half written because the wdog kicked in.
128 watchdog_tickle_fn wdog_tickler = NULL;
132 * Variable panicstr contains argument to first call to panic; used as flag
133 * to indicate that the kernel has already called panic.
135 const char *panicstr;
137 int dumping; /* system is dumping */
139 u_int panic_cpu_interlock; /* panic interlock */
140 globaldata_t panic_cpu_gd; /* which cpu took the panic */
143 static void boot (int) __dead2;
144 static void dumpsys (void);
145 static int setdumpdev (cdev_t dev);
146 static void poweroff_wait (void *, int);
147 static void print_uptime (void);
148 static void shutdown_halt (void *junk, int howto);
149 static void shutdown_panic (void *junk, int howto);
150 static void shutdown_reset (void *junk, int howto);
151 static int shutdown_busycount1(struct buf *bp, void *info);
152 static int shutdown_busycount2(struct buf *bp, void *info);
153 static void shutdown_cleanup_proc(struct proc *p);
155 /* register various local shutdown events */
157 shutdown_conf(void *unused)
159 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
160 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
161 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
162 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
165 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
170 * The system call that results in a reboot
173 sys_reboot(struct reboot_args *uap)
175 struct thread *td = curthread;
178 if ((error = suser(td)))
186 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
188 static int shutdown_howto = 0;
191 shutdown_nice(int howto)
193 shutdown_howto = howto;
195 /* Send a signal to init(8) and have it shutdown the world */
196 if (initproc != NULL) {
197 ksignal(initproc, SIGINT);
199 /* No init(8) running, so simply reboot */
204 static int waittime = -1;
205 static struct thread *dumpthread;
206 static struct pcb dumppcb;
217 if (ts.tv_sec >= 86400) {
218 printf("%ldd", ts.tv_sec / 86400);
222 if (f || ts.tv_sec >= 3600) {
223 printf("%ldh", ts.tv_sec / 3600);
227 if (f || ts.tv_sec >= 60) {
228 printf("%ldm", ts.tv_sec / 60);
232 printf("%lds\n", ts.tv_sec);
236 * Go through the rigmarole of shutting down..
237 * this used to be in machdep.c but I'll be dammned if I could see
238 * anything machine dependant in it.
244 * Get rid of any user scheduler baggage and then give
245 * us a high priority.
247 if (curthread->td_release)
248 curthread->td_release(curthread);
249 lwkt_setpri_self(TDPRI_MAX);
251 /* collect extra flags that shutdown_nice might have set */
252 howto |= shutdown_howto;
256 * We really want to shutdown on the BSP. Subsystems such as ACPI
257 * can't power-down the box otherwise.
259 if (smp_active_mask > 1) {
260 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
262 if (panicstr == NULL && mycpu->gd_cpuid != 0) {
263 printf("Switching to cpu #0 for shutdown\n");
264 lwkt_setcpu_self(globaldata_find(0));
268 * Do any callouts that should be done BEFORE syncing the filesystems.
270 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
273 * Try to get rid of any remaining FS references. The calling
274 * process, proc0, and init may still hold references. The
275 * VFS cache subsystem may still hold a root reference to root.
277 if (panicstr == NULL) {
278 shutdown_cleanup_proc(curproc);
279 shutdown_cleanup_proc(&proc0);
281 shutdown_cleanup_proc(initproc);
282 vfs_cache_setroot(NULL, NULL);
286 * Now sync filesystems
288 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
289 int iter, nbusy, pbusy;
292 printf("\nsyncing disks... ");
294 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
297 * With soft updates, some buffers that are
298 * written will be remarked as dirty until other
299 * buffers are written.
301 for (iter = pbusy = 0; iter < 20; iter++) {
302 nbusy = scan_all_buffers(shutdown_busycount1, NULL);
305 printf("%d ", nbusy);
311 * Process soft update work queue if buffers don't sync
312 * after 6 iterations by permitting the syncer to run.
314 if (iter > 5 && bioops.io_sync)
315 (*bioops.io_sync)(NULL);
316 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
317 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
321 * Count only busy local buffers to prevent forcing
322 * a fsck if we're just a client of a wedged NFS server
324 nbusy = scan_all_buffers(shutdown_busycount2, NULL);
327 * Failed to sync all blocks. Indicate this and don't
328 * unmount filesystems (thus forcing an fsck on reboot).
330 printf("giving up on %d buffers\n", nbusy);
332 Debugger("busy buffer problem");
334 tsleep(boot, 0, "shutdn", hz * 5 + 1);
338 * Unmount filesystems
340 if (panicstr == NULL)
343 tsleep(boot, 0, "shutdn", hz / 10 + 1);
349 * Ok, now do things that assume all filesystem activity has
352 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
354 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
357 /* Now that we're going to really halt the system... */
358 EVENTHANDLER_INVOKE(shutdown_final, howto);
360 for(;;) ; /* safety against shutdown_reset not working */
365 shutdown_busycount1(struct buf *bp, void *info)
367 if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
369 if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
375 shutdown_busycount2(struct buf *bp, void *info)
377 if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
378 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
380 * Only count buffers undergoing write I/O
381 * on the related vnode.
383 if (bp->b_vp == NULL ||
384 bp->b_vp->v_track_write.bk_active == 0) {
387 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
389 "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
391 bp->b_flags, bp->b_loffset,
392 bp->b_bio2.bio_offset);
400 * If the shutdown was a clean halt, behave accordingly.
403 shutdown_halt(void *junk, int howto)
405 if (howto & RB_HALT) {
407 printf("The operating system has halted.\n");
408 printf("Please press any key to reboot.\n\n");
410 case -1: /* No console, just die */
421 * Check to see if the system paniced, pause and then reboot
422 * according to the specified delay.
425 shutdown_panic(void *junk, int howto)
429 if (howto & RB_DUMP) {
430 if (PANIC_REBOOT_WAIT_TIME != 0) {
431 if (PANIC_REBOOT_WAIT_TIME != -1) {
432 printf("Automatic reboot in %d seconds - "
433 "press a key on the console to abort\n",
434 PANIC_REBOOT_WAIT_TIME);
435 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
437 DELAY(1000 * 100); /* 1/10th second */
438 /* Did user type a key? */
439 if (cncheckc() != -1)
445 } else { /* zero time specified - reboot NOW */
448 printf("--> Press a key on the console to reboot,\n");
449 printf("--> or switch off the system now.\n");
455 * Everything done, now reset
458 shutdown_reset(void *junk, int howto)
460 printf("Rebooting...\n");
461 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
462 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
464 /* NOTREACHED */ /* assuming reset worked */
468 * Try to remove FS references in the specified process. This function
469 * is used during shutdown
473 shutdown_cleanup_proc(struct proc *p)
475 struct filedesc *fdp;
480 if ((fdp = p->p_fd) != NULL) {
483 cache_drop(fdp->fd_ncdir);
486 fdp->fd_ncdir = NULL;
489 cache_drop(fdp->fd_nrdir);
492 fdp->fd_nrdir = NULL;
495 cache_drop(fdp->fd_njdir);
498 fdp->fd_njdir = NULL;
502 vkernel_drop(p->p_vkernel);
511 pmap_remove_pages(vmspace_pmap(vm),
514 vm_map_remove(&vm->vm_map,
521 * Magic number for savecore
523 * exported (symorder) and used at least by savecore(8)
526 static u_long const dumpmag = 0x8fca0101UL;
528 static int dumpsize = 0; /* also for savecore */
530 static int dodump = 1;
532 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
533 "Try to perform coredump on kernel panic");
546 psize = dev_dpsize(dev);
550 * XXX should clean up checking in dumpsys() to be more like this.
552 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE);
553 if (newdumplo <= LABELSECTOR)
562 static void dump_conf (void *dummy);
570 path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK);
571 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
572 dev = kgetdiskbyname(path);
577 if (setdumpdev(dumpdev) != 0)
581 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
584 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
589 ndumpdev = dev2udev(dumpdev);
590 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
591 if (error == 0 && req->newptr != NULL)
592 error = setdumpdev(udev2dev(ndumpdev, 0));
596 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
597 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", "");
600 * Doadump comes here after turning off memory management and
601 * getting on the dump stack, either when called above, or by
602 * the auto-restart code.
610 dumpthread = curthread;
612 printf("Dump already in progress, bailing...\n");
617 if (dumpdev == NOCDEV)
620 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo);
622 error = dev_ddump(dumpdev);
624 printf("succeeded\n");
627 printf("failed, reason: ");
631 printf("device doesn't support a dump routine\n");
635 printf("device bad\n");
639 printf("device not ready\n");
643 printf("area improper\n");
647 printf("i/o error\n");
651 printf("aborted from console\n");
655 printf("unknown, error = %d\n", error);
661 dumpstatus(vm_offset_t addr, off_t count)
665 if (addr % (1024 * 1024) == 0) {
670 printf("%ld ", (long)(count / (1024 * 1024)));
673 if ((c = cncheckc()) == 0x03)
676 printf("[CTRL-C to abort] ");
682 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
683 * and then reboots. If we are called twice, then we avoid trying to sync
684 * the disks as this often leads to recursive panics.
687 panic(const char *fmt, ...)
689 int bootopt, newpanic;
691 static char buf[256];
695 * If a panic occurs on multiple cpus before the first is able to
696 * halt the other cpus, only one cpu is allowed to take the panic.
697 * Attempt to be verbose about this situation but if the printf()
698 * itself panics don't let us overrun the kernel stack.
700 * Be very nasty about descheduling our thread at the lowest
701 * level possible in an attempt to freeze the thread without
702 * inducing further panics.
704 * Bumping gd_trap_nesting_level will also bypass assertions in
705 * lwkt_switch() and allow us to switch away even if we are a
706 * FAST interrupt or IPI.
708 if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
709 panic_cpu_gd = mycpu;
710 } else if (panic_cpu_gd != mycpu) {
712 ++mycpu->gd_trap_nesting_level;
713 if (mycpu->gd_trap_nesting_level < 25) {
714 printf("SECONDARY PANIC ON CPU %d THREAD %p\n",
715 mycpu->gd_cpuid, curthread);
717 curthread->td_release = NULL; /* be a grinch */
719 lwkt_deschedule_self(curthread);
723 /* --mycpu->gd_trap_nesting_level */
727 bootopt = RB_AUTOBOOT | RB_DUMP;
728 if (sync_on_panic == 0)
729 bootopt |= RB_NOSYNC;
732 bootopt |= RB_NOSYNC;
739 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
743 printf("panic: %s\n", buf);
745 /* three separate prints in case of an unmapped page and trap */
746 printf("mp_lock = %08x; ", mp_lock);
747 printf("cpuid = %d; ", mycpu->gd_cpuid);
748 printf("lapic.id = %08x\n", lapic.id);
752 if (newpanic && trace_on_panic)
753 db_print_backtrace();
754 if (debugger_on_panic)
761 * Support for poweroff delay.
763 #ifndef POWEROFF_DELAY
764 # define POWEROFF_DELAY 5000
766 static int poweroff_delay = POWEROFF_DELAY;
768 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
769 &poweroff_delay, 0, "");
772 poweroff_wait(void *junk, int howto)
774 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
776 DELAY(poweroff_delay * 1000);
780 * Some system processes (e.g. syncer) need to be stopped at appropriate
781 * points in their main loops prior to a system shutdown, so that they
782 * won't interfere with the shutdown process (e.g. by holding a disk buf
783 * to cause sync to fail). For each of these system processes, register
784 * shutdown_kproc() as a handler for one of shutdown events.
786 static int kproc_shutdown_wait = 60;
787 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
788 &kproc_shutdown_wait, 0, "");
791 shutdown_kproc(void *arg, int howto)
800 td = (struct thread *)arg;
801 if ((p = td->td_proc) != NULL) {
802 printf("Waiting (max %d seconds) for system process `%s' to stop...",
803 kproc_shutdown_wait, p->p_comm);
805 printf("Waiting (max %d seconds) for system thread %s to stop...",
806 kproc_shutdown_wait, td->td_comm);
808 error = suspend_kproc(td, kproc_shutdown_wait * hz);
810 if (error == EWOULDBLOCK)
811 printf("timed out\n");