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>
82 #include <machine/clock.h>
83 #include <machine/md_var.h>
84 #include <machine/smp.h> /* smp_active_mask, cpuid */
85 #include <machine/vmparam.h>
86 #include <machine/thread.h>
88 #include <sys/signalvar.h>
91 #include <dev/misc/gpio/gpio.h>
93 #ifndef PANIC_REBOOT_WAIT_TIME
94 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
98 * Note that stdarg.h and the ANSI style va_start macro is used for both
99 * ANSI and traditional C compilers. We use the machine version to stay
100 * within the confines of the kernel header files.
102 #include <machine/stdarg.h>
106 #ifdef DDB_UNATTENDED
107 int debugger_on_panic = 0;
109 int debugger_on_panic = 1;
111 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
112 &debugger_on_panic, 0, "Run debugger on kernel panic");
115 int trace_on_panic = 1;
117 int trace_on_panic = 0;
119 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
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_RW,
125 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
127 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
131 * If there is a hardware watchdog, point this at the function needed to
133 * It's needed when the kernel needs to do some lengthy operations.
134 * e.g. in wd.c when dumping core.. It's most annoying to have
135 * your precious core-dump only half written because the wdog kicked in.
137 watchdog_tickle_fn wdog_tickler = NULL;
141 * Variable panicstr contains argument to first call to panic; used as flag
142 * to indicate that the kernel has already called panic.
144 const char *panicstr;
146 int dumping; /* system is dumping */
147 static struct dumperinfo dumper; /* selected dumper */
150 u_int panic_cpu_interlock; /* panic interlock */
151 globaldata_t panic_cpu_gd; /* which cpu took the panic */
154 int bootverbose = 0; /* note: assignment to force non-bss */
155 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW,
156 &bootverbose, 0, "Verbose kernel messages");
158 int cold = 1; /* note: assignment to force non-bss */
159 int dumplo; /* OBSOLETE - savecore compat */
162 static void boot (int) __dead2;
163 static int setdumpdev (cdev_t dev);
164 static void poweroff_wait (void *, int);
165 static void print_uptime (void);
166 static void shutdown_halt (void *junk, int howto);
167 static void shutdown_panic (void *junk, int howto);
168 static void shutdown_reset (void *junk, int howto);
169 static int shutdown_busycount1(struct buf *bp, void *info);
170 static int shutdown_busycount2(struct buf *bp, void *info);
171 static void shutdown_cleanup_proc(struct proc *p);
173 /* register various local shutdown events */
175 shutdown_conf(void *unused)
177 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
178 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
179 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
180 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
183 SYSINIT(shutdown_conf, SI_BOOT2_MACHDEP, SI_ORDER_ANY, shutdown_conf, NULL)
188 * The system call that results in a reboot
193 sys_reboot(struct reboot_args *uap)
195 struct thread *td = curthread;
198 if ((error = priv_check(td, PRIV_REBOOT)))
208 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
210 static int shutdown_howto = 0;
213 shutdown_nice(int howto)
215 shutdown_howto = howto;
217 /* Send a signal to init(8) and have it shutdown the world */
218 if (initproc != NULL) {
219 ksignal(initproc, SIGINT);
221 /* No init(8) running, so simply reboot */
226 static int waittime = -1;
228 struct thread *dumpthread;
239 if (ts.tv_sec >= 86400) {
240 kprintf("%ldd", ts.tv_sec / 86400);
244 if (f || ts.tv_sec >= 3600) {
245 kprintf("%ldh", ts.tv_sec / 3600);
249 if (f || ts.tv_sec >= 60) {
250 kprintf("%ldm", ts.tv_sec / 60);
254 kprintf("%lds\n", ts.tv_sec);
258 * Go through the rigmarole of shutting down..
259 * this used to be in machdep.c but I'll be dammned if I could see
260 * anything machine dependant in it.
266 * Get rid of any user scheduler baggage and then give
267 * us a high priority.
269 if (curthread->td_release)
270 curthread->td_release(curthread);
271 lwkt_setpri_self(TDPRI_MAX);
273 /* collect extra flags that shutdown_nice might have set */
274 howto |= shutdown_howto;
278 * We really want to shutdown on the BSP. Subsystems such as ACPI
279 * can't power-down the box otherwise.
281 if (smp_active_mask > 1) {
282 kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
284 if (panicstr == NULL && mycpu->gd_cpuid != 0) {
285 kprintf("Switching to cpu #0 for shutdown\n");
286 lwkt_setcpu_self(globaldata_find(0));
290 * Do any callouts that should be done BEFORE syncing the filesystems.
292 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
295 * Try to get rid of any remaining FS references. The calling
296 * process, proc0, and init may still hold references. The
297 * VFS cache subsystem may still hold a root reference to root.
299 * XXX this needs work. We really need to SIGSTOP all remaining
300 * processes in order to avoid blowups due to proc0's filesystem
301 * references going away. For now just make sure that the init
302 * process is stopped.
304 if (panicstr == NULL) {
305 shutdown_cleanup_proc(curproc);
306 shutdown_cleanup_proc(&proc0);
308 if (initproc != curproc) {
309 ksignal(initproc, SIGSTOP);
310 tsleep(boot, 0, "shutdn", hz / 20);
312 shutdown_cleanup_proc(initproc);
314 vfs_cache_setroot(NULL, NULL);
318 * Now sync filesystems
320 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
321 int iter, nbusy, pbusy;
324 kprintf("\nsyncing disks... ");
326 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
329 * With soft updates, some buffers that are
330 * written will be remarked as dirty until other
331 * buffers are written.
333 for (iter = pbusy = 0; iter < 20; iter++) {
334 nbusy = scan_all_buffers(shutdown_busycount1, NULL);
337 kprintf("%d ", nbusy);
343 * Process soft update work queue if buffers don't sync
344 * after 6 iterations by permitting the syncer to run.
349 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
350 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
354 * Count only busy local buffers to prevent forcing
355 * a fsck if we're just a client of a wedged NFS server
357 nbusy = scan_all_buffers(shutdown_busycount2, NULL);
360 * Failed to sync all blocks. Indicate this and don't
361 * unmount filesystems (thus forcing an fsck on reboot).
363 kprintf("giving up on %d buffers\n", nbusy);
365 Debugger("busy buffer problem");
367 tsleep(boot, 0, "shutdn", hz * 5 + 1);
371 * Unmount filesystems
373 if (panicstr == NULL)
376 tsleep(boot, 0, "shutdn", hz / 10 + 1);
382 * Dump before doing post_sync shutdown ops
385 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) {
390 * Ok, now do things that assume all filesystem activity has
391 * been completed. This will also call the device shutdown
394 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
396 /* Now that we're going to really halt the system... */
397 EVENTHANDLER_INVOKE(shutdown_final, howto);
399 for(;;) ; /* safety against shutdown_reset not working */
404 shutdown_busycount1(struct buf *bp, void *info)
406 if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
408 if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
414 shutdown_busycount2(struct buf *bp, void *info)
416 if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
417 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
419 * Only count buffers undergoing write I/O
420 * on the related vnode.
422 if (bp->b_vp == NULL ||
423 bio_track_active(&bp->b_vp->v_track_write) == 0) {
426 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
428 "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
430 bp->b_flags, bp->b_loffset,
431 bp->b_bio2.bio_offset);
439 * If the shutdown was a clean halt, behave accordingly.
442 shutdown_halt(void *junk, int howto)
444 if (howto & RB_HALT) {
446 kprintf("The operating system has halted.\n");
447 #ifdef _KERNEL_VIRTUAL
450 kprintf("Please press any key to reboot.\n\n");
452 case -1: /* No console, just die */
464 * Check to see if the system paniced, pause and then reboot
465 * according to the specified delay.
468 shutdown_panic(void *junk, int howto)
472 if (howto & RB_DUMP) {
473 if (PANIC_REBOOT_WAIT_TIME != 0) {
474 if (PANIC_REBOOT_WAIT_TIME != -1) {
475 kprintf("Automatic reboot in %d seconds - "
476 "press a key on the console to abort\n",
477 PANIC_REBOOT_WAIT_TIME);
478 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
480 DELAY(1000 * 100); /* 1/10th second */
481 /* Did user type a key? */
482 if (cncheckc() != -1)
488 } else { /* zero time specified - reboot NOW */
491 kprintf("--> Press a key on the console to reboot,\n");
492 kprintf("--> or switch off the system now.\n");
498 * Everything done, now reset
501 shutdown_reset(void *junk, int howto)
503 kprintf("Rebooting...\n");
504 DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */
505 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
507 /* NOTREACHED */ /* assuming reset worked */
511 * Try to remove FS references in the specified process. This function
512 * is used during shutdown
516 shutdown_cleanup_proc(struct proc *p)
518 struct filedesc *fdp;
523 if ((fdp = p->p_fd) != NULL) {
526 cache_drop(&fdp->fd_ncdir);
531 cache_drop(&fdp->fd_nrdir);
536 cache_drop(&fdp->fd_njdir);
549 pmap_remove_pages(vmspace_pmap(vm),
551 VM_MAX_USER_ADDRESS);
552 vm_map_remove(&vm->vm_map,
554 VM_MAX_USER_ADDRESS);
559 * Magic number for savecore
561 * exported (symorder) and used at least by savecore(8)
563 * Mark it as used so that gcc doesn't optimize it away.
565 __attribute__((__used__))
566 static u_long const dumpmag = 0x8fca0101UL;
568 __attribute__((__used__))
569 static int dumpsize = 0; /* also for savecore */
571 static int dodump = 1;
573 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
574 "Try to perform coredump on kernel panic");
577 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
578 uint64_t dumplen, uint32_t blksz)
580 bzero(kdh, sizeof(*kdh));
581 strncpy(kdh->magic, magic, sizeof(kdh->magic));
582 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
583 kdh->version = htod32(KERNELDUMPVERSION);
584 kdh->architectureversion = htod32(archver);
585 kdh->dumplength = htod64(dumplen);
586 kdh->dumptime = htod64(time_second);
587 kdh->blocksize = htod32(blksz);
588 strncpy(kdh->hostname, hostname, sizeof(kdh->hostname));
589 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
590 if (panicstr != NULL)
591 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
592 kdh->parity = kerneldump_parity(kdh);
596 setdumpdev(cdev_t dev)
602 disk_dumpconf(NULL, 0/*off*/);
607 * We have to open the device before we can perform ioctls on it,
608 * or the slice/label data may not be present. Device opens are
609 * usually tracked by specfs, but the dump device can be set in
610 * early boot and may not be open so this is somewhat of a hack.
612 doopen = (dev->si_sysref.refcnt == 1);
614 error = dev_dopen(dev, FREAD, S_IFCHR, proc0.p_ucred);
618 error = disk_dumpconf(dev, 1/*on*/);
624 static void dump_conf (void *dummy);
626 dump_conf(void *dummy)
631 path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK);
632 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
633 dev = kgetdiskbyname(path);
638 if (setdumpdev(dumpdev) != 0)
642 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
645 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
650 ndumpdev = dev2udev(dumpdev);
651 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
652 if (error == 0 && req->newptr != NULL)
653 error = setdumpdev(udev2dev(ndumpdev, 0));
657 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
658 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", "");
661 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
662 * and then reboots. If we are called twice, then we avoid trying to sync
663 * the disks as this often leads to recursive panics.
666 panic(const char *fmt, ...)
668 int bootopt, newpanic;
670 static char buf[256];
674 * If a panic occurs on multiple cpus before the first is able to
675 * halt the other cpus, only one cpu is allowed to take the panic.
676 * Attempt to be verbose about this situation but if the kprintf()
677 * itself panics don't let us overrun the kernel stack.
679 * Be very nasty about descheduling our thread at the lowest
680 * level possible in an attempt to freeze the thread without
681 * inducing further panics.
683 * Bumping gd_trap_nesting_level will also bypass assertions in
684 * lwkt_switch() and allow us to switch away even if we are a
685 * FAST interrupt or IPI.
687 if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
688 panic_cpu_gd = mycpu;
689 } else if (panic_cpu_gd != mycpu) {
691 ++mycpu->gd_trap_nesting_level;
692 if (mycpu->gd_trap_nesting_level < 25) {
693 kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
694 mycpu->gd_cpuid, curthread);
696 curthread->td_release = NULL; /* be a grinch */
698 lwkt_deschedule_self(curthread);
702 /* --mycpu->gd_trap_nesting_level */
706 bootopt = RB_AUTOBOOT | RB_DUMP;
707 if (sync_on_panic == 0)
708 bootopt |= RB_NOSYNC;
711 bootopt |= RB_NOSYNC;
718 kvsnprintf(buf, sizeof(buf), fmt, ap);
722 kprintf("panic: %s\n", buf);
724 /* two separate prints in case of an unmapped page and trap */
725 kprintf("mp_lock = %08x; ", mp_lock);
726 kprintf("cpuid = %d\n", mycpu->gd_cpuid);
729 #if (NGPIO > 0) && defined(ERROR_LED_ON_PANIC)
730 led_switch("error", 1);
733 #if defined(WDOG_DISABLE_ON_PANIC) && defined(WATCHDOG_ENABLE)
738 if (newpanic && trace_on_panic)
740 if (debugger_on_panic)
747 * Support for poweroff delay.
749 #ifndef POWEROFF_DELAY
750 # define POWEROFF_DELAY 5000
752 static int poweroff_delay = POWEROFF_DELAY;
754 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
755 &poweroff_delay, 0, "");
758 poweroff_wait(void *junk, int howto)
760 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
762 DELAY(poweroff_delay * 1000);
766 * Some system processes (e.g. syncer) need to be stopped at appropriate
767 * points in their main loops prior to a system shutdown, so that they
768 * won't interfere with the shutdown process (e.g. by holding a disk buf
769 * to cause sync to fail). For each of these system processes, register
770 * shutdown_kproc() as a handler for one of shutdown events.
772 static int kproc_shutdown_wait = 60;
773 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
774 &kproc_shutdown_wait, 0, "");
777 shutdown_kproc(void *arg, int howto)
786 td = (struct thread *)arg;
787 if ((p = td->td_proc) != NULL) {
788 kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
789 kproc_shutdown_wait, p->p_comm);
791 kprintf("Waiting (max %d seconds) for system thread %s to stop...",
792 kproc_shutdown_wait, td->td_comm);
794 error = suspend_kproc(td, kproc_shutdown_wait * hz);
796 if (error == EWOULDBLOCK)
797 kprintf("timed out\n");
799 kprintf("stopped\n");
802 /* Registration of dumpers */
804 set_dumper(struct dumperinfo *di)
807 bzero(&dumper, sizeof(dumper));
811 if (dumper.dumper != NULL)
821 #if defined (_KERNEL_VIRTUAL)
822 /* VKERNELs don't support dumps */
823 kprintf("VKERNEL doesn't support dumps\n");
827 * If there is a dumper registered and we aren't dumping already, call
828 * the machine dependent dumpsys (md_dumpsys) to do the hard work.
830 * XXX: while right now the md_dumpsys() of x86 and x86_64 could be
831 * factored out completely into here, I rather keep them machine
832 * dependent in case we ever add a platform which does not share
833 * the same dumpsys() code, such as arm.
835 if (dumper.dumper != NULL && !dumping) {