ia64/xen-unstable

view freebsd-5.3-xen-sparse/kern/kern_shutdown.c @ 7127:805ee053e61f

VIRQs and IPIs on VCPU#0 are automatically re-bound on save/restore.
Signed-off-by: Keir Fraser <keir@xensource.com>
author kaf24@firebug.cl.cam.ac.uk
date Thu Sep 29 13:05:43 2005 +0100 (2005-09-29)
parents 79b7835ac75d
children
line source
1 /*-
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.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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.
21 *
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
32 * SUCH DAMAGE.
33 *
34 * @(#)kern_shutdown.c 8.3 (Berkeley) 1/21/94
35 */
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD: src/sys/kern/kern_shutdown.c,v 1.163.2.2 2004/09/10 00:04:17 scottl Exp $");
40 #include "opt_kdb.h"
41 #include "opt_hw_wdog.h"
42 #include "opt_mac.h"
43 #include "opt_panic.h"
44 #include "opt_show_busybufs.h"
45 #include "opt_sched.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/bio.h>
50 #include <sys/buf.h>
51 #include <sys/conf.h>
52 #include <sys/cons.h>
53 #include <sys/eventhandler.h>
54 #include <sys/kdb.h>
55 #include <sys/kernel.h>
56 #include <sys/kthread.h>
57 #include <sys/mac.h>
58 #include <sys/malloc.h>
59 #include <sys/mount.h>
60 #include <sys/proc.h>
61 #include <sys/reboot.h>
62 #include <sys/resourcevar.h>
63 #include <sys/smp.h> /* smp_active */
64 #include <sys/sysctl.h>
65 #include <sys/sysproto.h>
66 #include <sys/vnode.h>
68 #include <machine/cpu.h>
69 #include <machine/pcb.h>
70 #include <machine/smp.h>
72 #include <sys/signalvar.h>
74 #ifndef PANIC_REBOOT_WAIT_TIME
75 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
76 #endif
78 /*
79 * Note that stdarg.h and the ANSI style va_start macro is used for both
80 * ANSI and traditional C compilers.
81 */
82 #include <machine/stdarg.h>
84 #ifdef KDB
85 #ifdef KDB_UNATTENDED
86 int debugger_on_panic = 0;
87 #else
88 int debugger_on_panic = 1;
89 #endif
90 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
91 &debugger_on_panic, 0, "Run debugger on kernel panic");
93 #ifdef KDB_TRACE
94 int trace_on_panic = 1;
95 #else
96 int trace_on_panic = 0;
97 #endif
98 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
99 &trace_on_panic, 0, "Print stack trace on kernel panic");
100 #endif /* KDB */
102 int sync_on_panic = 0;
103 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
104 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
106 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
108 #ifdef HW_WDOG
109 /*
110 * If there is a hardware watchdog, point this at the function needed to
111 * hold it off.
112 * It's needed when the kernel needs to do some lengthy operations.
113 * e.g. in wd.c when dumping core.. It's most annoying to have
114 * your precious core-dump only half written because the wdog kicked in.
115 */
116 watchdog_tickle_fn wdog_tickler = NULL;
117 #endif /* HW_WDOG */
119 /*
120 * Variable panicstr contains argument to first call to panic; used as flag
121 * to indicate that the kernel has already called panic.
122 */
123 const char *panicstr;
125 int dumping; /* system is dumping */
126 static struct dumperinfo dumper; /* our selected dumper */
128 /* Context information for dump-debuggers. */
129 static struct pcb dumppcb; /* Registers. */
130 static lwpid_t dumptid; /* Thread ID. */
132 static void boot(int) __dead2;
133 static void poweroff_wait(void *, int);
134 static void shutdown_halt(void *junk, int howto);
135 static void shutdown_panic(void *junk, int howto);
136 static void shutdown_reset(void *junk, int howto);
138 /* register various local shutdown events */
139 static void
140 shutdown_conf(void *unused)
141 {
143 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
144 SHUTDOWN_PRI_FIRST);
145 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
146 SHUTDOWN_PRI_LAST + 100);
147 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
148 SHUTDOWN_PRI_LAST + 100);
149 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
150 SHUTDOWN_PRI_LAST + 200);
151 }
153 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
155 /*
156 * The system call that results in a reboot
157 *
158 * MPSAFE
159 */
160 /* ARGSUSED */
161 int
162 reboot(struct thread *td, struct reboot_args *uap)
163 {
164 int error;
166 error = 0;
167 #ifdef MAC
168 error = mac_check_system_reboot(td->td_ucred, uap->opt);
169 #endif
170 if (error == 0)
171 error = suser(td);
172 if (error == 0) {
173 mtx_lock(&Giant);
174 boot(uap->opt);
175 mtx_unlock(&Giant);
176 }
177 return (error);
178 }
180 /*
181 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
182 */
183 static int shutdown_howto = 0;
185 void
186 shutdown_nice(int howto)
187 {
189 shutdown_howto = howto;
191 /* Send a signal to init(8) and have it shutdown the world */
192 if (initproc != NULL) {
193 PROC_LOCK(initproc);
194 psignal(initproc, SIGINT);
195 PROC_UNLOCK(initproc);
196 } else {
197 /* No init(8) running, so simply reboot */
198 boot(RB_NOSYNC);
199 }
200 return;
201 }
202 static int waittime = -1;
204 static void
205 print_uptime(void)
206 {
207 int f;
208 struct timespec ts;
210 getnanouptime(&ts);
211 printf("Uptime: ");
212 f = 0;
213 if (ts.tv_sec >= 86400) {
214 printf("%ldd", (long)ts.tv_sec / 86400);
215 ts.tv_sec %= 86400;
216 f = 1;
217 }
218 if (f || ts.tv_sec >= 3600) {
219 printf("%ldh", (long)ts.tv_sec / 3600);
220 ts.tv_sec %= 3600;
221 f = 1;
222 }
223 if (f || ts.tv_sec >= 60) {
224 printf("%ldm", (long)ts.tv_sec / 60);
225 ts.tv_sec %= 60;
226 f = 1;
227 }
228 printf("%lds\n", (long)ts.tv_sec);
229 }
231 static void
232 doadump(void)
233 {
235 /*
236 * Sometimes people have to call this from the kernel debugger.
237 * (if 'panic' can not dump)
238 * Give them a clue as to why they can't dump.
239 */
240 if (dumper.dumper == NULL) {
241 printf("Cannot dump. No dump device defined.\n");
242 return;
243 }
245 savectx(&dumppcb);
246 dumptid = curthread->td_tid;
247 dumping++;
248 dumpsys(&dumper);
249 }
251 /*
252 * Go through the rigmarole of shutting down..
253 * this used to be in machdep.c but I'll be dammned if I could see
254 * anything machine dependant in it.
255 */
256 static void
257 boot(int howto)
258 {
259 static int first_buf_printf = 1;
261 /* collect extra flags that shutdown_nice might have set */
262 howto |= shutdown_howto;
264 /* We are out of the debugger now. */
265 kdb_active = 0;
267 #ifdef SMP
268 if (smp_active)
269 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
270 #endif
271 /*
272 * Do any callouts that should be done BEFORE syncing the filesystems.
273 */
274 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
276 /*
277 * Now sync filesystems
278 */
279 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
280 register struct buf *bp;
281 int iter, nbusy, pbusy;
282 #ifndef PREEMPTION
283 int subiter;
284 #endif
286 waittime = 0;
288 sync(&thread0, NULL);
290 /*
291 * With soft updates, some buffers that are
292 * written will be remarked as dirty until other
293 * buffers are written.
294 */
295 for (iter = pbusy = 0; iter < 20; iter++) {
296 nbusy = 0;
297 for (bp = &buf[nbuf]; --bp >= buf; ) {
298 if ((bp->b_flags & B_INVAL) == 0 &&
299 BUF_REFCNT(bp) > 0) {
300 nbusy++;
301 } else if ((bp->b_flags & (B_DELWRI | B_INVAL))
302 == B_DELWRI) {
303 /* bawrite(bp);*/
304 nbusy++;
305 }
306 }
307 if (nbusy == 0) {
308 if (first_buf_printf)
309 printf("No buffers busy after final sync");
310 break;
311 }
312 if (first_buf_printf) {
313 printf("Syncing disks, buffers remaining... ");
314 first_buf_printf = 0;
315 }
316 printf("%d ", nbusy);
317 if (nbusy < pbusy)
318 iter = 0;
319 pbusy = nbusy;
320 sync(&thread0, NULL);
322 #ifdef PREEMPTION
323 /*
324 * Drop Giant and spin for a while to allow
325 * interrupt threads to run.
326 */
327 DROP_GIANT();
328 DELAY(50000 * iter);
329 PICKUP_GIANT();
330 #else
331 /*
332 * Drop Giant and context switch several times to
333 * allow interrupt threads to run.
334 */
335 DROP_GIANT();
336 for (subiter = 0; subiter < 50 * iter; subiter++) {
337 mtx_lock_spin(&sched_lock);
338 mi_switch(SW_VOL, NULL);
339 mtx_unlock_spin(&sched_lock);
340 DELAY(1000);
341 }
342 PICKUP_GIANT();
343 #endif
344 }
345 printf("\n");
346 /*
347 * Count only busy local buffers to prevent forcing
348 * a fsck if we're just a client of a wedged NFS server
349 */
350 nbusy = 0;
351 for (bp = &buf[nbuf]; --bp >= buf; ) {
352 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) ||
353 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
354 if (bp->b_dev == NULL) {
355 TAILQ_REMOVE(&mountlist,
356 bp->b_vp->v_mount, mnt_list);
357 continue;
358 }
359 nbusy++;
360 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
361 printf(
362 "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n",
363 nbusy, devtoname(bp->b_dev),
364 bp->b_flags, (long)bp->b_blkno,
365 (long)bp->b_lblkno);
366 #endif
367 }
368 }
369 if (nbusy) {
370 /*
371 * Failed to sync all blocks. Indicate this and don't
372 * unmount filesystems (thus forcing an fsck on reboot).
373 */
374 printf("Giving up on %d buffers\n", nbusy);
375 DELAY(5000000); /* 5 seconds */
376 } else {
377 if (!first_buf_printf)
378 printf("Final sync complete\n");
379 /*
380 * Unmount filesystems
381 */
382 if (panicstr == 0)
383 vfs_unmountall();
384 }
385 DELAY(100000); /* wait for console output to finish */
386 }
388 print_uptime();
390 /*
391 * Ok, now do things that assume all filesystem activity has
392 * been completed.
393 */
394 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
395 splhigh();
396 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
397 doadump();
399 /* Now that we're going to really halt the system... */
400 EVENTHANDLER_INVOKE(shutdown_final, howto);
402 for(;;) ; /* safety against shutdown_reset not working */
403 /* NOTREACHED */
404 }
406 /*
407 * If the shutdown was a clean halt, behave accordingly.
408 */
409 static void
410 shutdown_halt(void *junk, int howto)
411 {
413 if (howto & RB_HALT) {
414 printf("\n");
415 printf("The operating system has halted.\n");
416 printf("Please press any key to reboot.\n\n");
417 switch (cngetc()) {
418 case -1: /* No console, just die */
419 cpu_halt();
420 /* NOTREACHED */
421 default:
422 howto &= ~RB_HALT;
423 break;
424 }
425 }
426 }
428 /*
429 * Check to see if the system paniced, pause and then reboot
430 * according to the specified delay.
431 */
432 static void
433 shutdown_panic(void *junk, int howto)
434 {
435 int loop;
437 if (howto & RB_DUMP) {
438 if (PANIC_REBOOT_WAIT_TIME != 0) {
439 if (PANIC_REBOOT_WAIT_TIME != -1) {
440 printf("Automatic reboot in %d seconds - "
441 "press a key on the console to abort\n",
442 PANIC_REBOOT_WAIT_TIME);
443 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
444 loop > 0; --loop) {
445 DELAY(1000 * 100); /* 1/10th second */
446 /* Did user type a key? */
447 if (cncheckc() != -1)
448 break;
449 }
450 if (!loop)
451 return;
452 }
453 } else { /* zero time specified - reboot NOW */
454 return;
455 }
456 printf("--> Press a key on the console to reboot,\n");
457 printf("--> or switch off the system now.\n");
458 cngetc();
459 }
460 }
462 /*
463 * Everything done, now reset
464 */
465 static void
466 shutdown_reset(void *junk, int howto)
467 {
469 printf("Rebooting...\n");
470 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
471 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
472 cpu_reset();
473 /* NOTREACHED */ /* assuming reset worked */
474 }
476 #ifdef SMP
477 static u_int panic_cpu = NOCPU;
478 #endif
480 /*
481 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
482 * and then reboots. If we are called twice, then we avoid trying to sync
483 * the disks as this often leads to recursive panics.
484 *
485 * MPSAFE
486 */
487 void
488 panic(const char *fmt, ...)
489 {
490 struct thread *td = curthread;
491 int bootopt, newpanic;
492 va_list ap;
493 static char buf[256];
495 #ifdef SMP
496 /*
497 * We don't want multiple CPU's to panic at the same time, so we
498 * use panic_cpu as a simple spinlock. We have to keep checking
499 * panic_cpu if we are spinning in case the panic on the first
500 * CPU is canceled.
501 */
502 if (panic_cpu != PCPU_GET(cpuid))
503 while (atomic_cmpset_int(&panic_cpu, NOCPU,
504 PCPU_GET(cpuid)) == 0)
505 while (panic_cpu != NOCPU)
506 ; /* nothing */
507 #endif
509 bootopt = RB_AUTOBOOT | RB_DUMP;
510 newpanic = 0;
511 if (panicstr)
512 bootopt |= RB_NOSYNC;
513 else {
514 panicstr = fmt;
515 newpanic = 1;
516 }
518 va_start(ap, fmt);
519 if (newpanic) {
520 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
521 panicstr = buf;
522 printf("panic: %s\n", buf);
523 } else {
524 printf("panic: ");
525 vprintf(fmt, ap);
526 printf("\n");
527 }
528 va_end(ap);
529 #ifdef SMP
530 printf("cpuid = %d\n", PCPU_GET(cpuid));
531 #endif
533 #ifdef KDB
534 if (newpanic && trace_on_panic)
535 kdb_backtrace();
536 if (debugger_on_panic)
537 kdb_enter("panic");
538 #ifdef RESTARTABLE_PANICS
539 /* See if the user aborted the panic, in which case we continue. */
540 if (panicstr == NULL) {
541 #ifdef SMP
542 atomic_store_rel_int(&panic_cpu, NOCPU);
543 #endif
544 return;
545 }
546 #endif
547 #endif
548 mtx_lock_spin(&sched_lock);
549 td->td_flags |= TDF_INPANIC;
550 mtx_unlock_spin(&sched_lock);
551 if (!sync_on_panic)
552 bootopt |= RB_NOSYNC;
553 #ifdef XEN
554 HYPERVISOR_crash();
555 #else
556 boot(bootopt);
557 #endif
558 }
560 /*
561 * Support for poweroff delay.
562 */
563 #ifndef POWEROFF_DELAY
564 # define POWEROFF_DELAY 5000
565 #endif
566 static int poweroff_delay = POWEROFF_DELAY;
568 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
569 &poweroff_delay, 0, "");
571 static void
572 poweroff_wait(void *junk, int howto)
573 {
575 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
576 return;
577 DELAY(poweroff_delay * 1000);
578 }
580 /*
581 * Some system processes (e.g. syncer) need to be stopped at appropriate
582 * points in their main loops prior to a system shutdown, so that they
583 * won't interfere with the shutdown process (e.g. by holding a disk buf
584 * to cause sync to fail). For each of these system processes, register
585 * shutdown_kproc() as a handler for one of shutdown events.
586 */
587 static int kproc_shutdown_wait = 60;
588 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
589 &kproc_shutdown_wait, 0, "");
591 void
592 kproc_shutdown(void *arg, int howto)
593 {
594 struct proc *p;
595 char procname[MAXCOMLEN + 1];
596 int error;
598 if (panicstr)
599 return;
601 p = (struct proc *)arg;
602 strlcpy(procname, p->p_comm, sizeof(procname));
603 printf("Waiting (max %d seconds) for system process `%s' to stop...",
604 kproc_shutdown_wait, procname);
605 error = kthread_suspend(p, kproc_shutdown_wait * hz);
607 if (error == EWOULDBLOCK)
608 printf("timed out\n");
609 else
610 printf("done\n");
611 }
613 /* Registration of dumpers */
614 int
615 set_dumper(struct dumperinfo *di)
616 {
618 if (di == NULL) {
619 bzero(&dumper, sizeof dumper);
620 return (0);
621 }
622 if (dumper.dumper != NULL)
623 return (EBUSY);
624 dumper = *di;
625 return (0);
626 }
628 #if defined(__powerpc__)
629 void
630 dumpsys(struct dumperinfo *di __unused)
631 {
633 printf("Kernel dumps not implemented on this architecture\n");
634 }
635 #endif