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 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD: releng/7.4/sys/kern/kern_shutdown.c 213935 2010-10-16 20:01:41Z avg $");
39
40 #include "opt_ddb.h"
41 #include "opt_kdb.h"
42 #include "opt_mac.h"
43 #include "opt_panic.h"
44 #include "opt_show_busybufs.h"
45 #include "opt_sched.h"
46
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/malloc.h>
58 #include <sys/mount.h>
59 #include <sys/priv.h>
60 #include <sys/proc.h>
61 #include <sys/reboot.h>
62 #include <sys/resourcevar.h>
63 #include <sys/sched.h>
64 #include <sys/smp.h> /* smp_active */
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67
68 #include <ddb/ddb.h>
69
70 #include <machine/cpu.h>
71 #include <machine/pcb.h>
72 #include <machine/smp.h>
73
74 #include <security/mac/mac_framework.h>
75
76 #include <vm/vm.h>
77 #include <vm/vm_object.h>
78 #include <vm/vm_page.h>
79 #include <vm/vm_pager.h>
80 #include <vm/swap_pager.h>
81
82 #include <sys/signalvar.h>
83
84 #ifndef PANIC_REBOOT_WAIT_TIME
85 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
86 #endif
87
88 /*
89 * Note that stdarg.h and the ANSI style va_start macro is used for both
90 * ANSI and traditional C compilers.
91 */
92 #include <machine/stdarg.h>
93
94 #ifdef KDB
95 #ifdef KDB_UNATTENDED
96 int debugger_on_panic = 0;
97 #else
98 int debugger_on_panic = 1;
99 #endif
100 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
101 &debugger_on_panic, 0, "Run debugger on kernel panic");
102 TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
103
104 #ifdef KDB_TRACE
105 static int trace_on_panic = 1;
106 #else
107 static int trace_on_panic = 0;
108 #endif
109 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
110 &trace_on_panic, 0, "Print stack trace on kernel panic");
111 TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
112 #endif /* KDB */
113
114 static int sync_on_panic = 0;
115 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
116 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
117 TUNABLE_INT("kern.sync_on_panic", &sync_on_panic);
118
119 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
120
121 /*
122 * Variable panicstr contains argument to first call to panic; used as flag
123 * to indicate that the kernel has already called panic.
124 */
125 const char *panicstr;
126
127 int dumping; /* system is dumping */
128 int rebooting; /* system is rebooting */
129 static struct dumperinfo dumper; /* our selected dumper */
130
131 /* Context information for dump-debuggers. */
132 static struct pcb dumppcb; /* Registers. */
133 static lwpid_t dumptid; /* Thread ID. */
134
135 static void boot(int) __dead2;
136 static void poweroff_wait(void *, int);
137 static void shutdown_halt(void *junk, int howto);
138 static void shutdown_panic(void *junk, int howto);
139 static void shutdown_reset(void *junk, int howto);
140
141 /* register various local shutdown events */
142 static void
143 shutdown_conf(void *unused)
144 {
145
146 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
147 SHUTDOWN_PRI_FIRST);
148 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
149 SHUTDOWN_PRI_LAST + 100);
150 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
151 SHUTDOWN_PRI_LAST + 100);
152 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
153 SHUTDOWN_PRI_LAST + 200);
154 }
155
156 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
157
158 /*
159 * The system call that results in a reboot.
160 */
161 /* ARGSUSED */
162 int
163 reboot(struct thread *td, struct reboot_args *uap)
164 {
165 int error;
166
167 error = 0;
168 #ifdef MAC
169 error = mac_check_system_reboot(td->td_ucred, uap->opt);
170 #endif
171 if (error == 0)
172 error = priv_check(td, PRIV_REBOOT);
173 if (error == 0) {
174 mtx_lock(&Giant);
175 boot(uap->opt);
176 mtx_unlock(&Giant);
177 }
178 return (error);
179 }
180
181 /*
182 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
183 */
184 static int shutdown_howto = 0;
185
186 void
187 shutdown_nice(int howto)
188 {
189
190 shutdown_howto = howto;
191
192 /* Send a signal to init(8) and have it shutdown the world */
193 if (initproc != NULL) {
194 PROC_LOCK(initproc);
195 psignal(initproc, SIGINT);
196 PROC_UNLOCK(initproc);
197 } else {
198 /* No init(8) running, so simply reboot */
199 boot(RB_NOSYNC);
200 }
201 return;
202 }
203 static int waittime = -1;
204
205 static void
206 print_uptime(void)
207 {
208 int f;
209 struct timespec ts;
210
211 getnanouptime(&ts);
212 printf("Uptime: ");
213 f = 0;
214 if (ts.tv_sec >= 86400) {
215 printf("%ldd", (long)ts.tv_sec / 86400);
216 ts.tv_sec %= 86400;
217 f = 1;
218 }
219 if (f || ts.tv_sec >= 3600) {
220 printf("%ldh", (long)ts.tv_sec / 3600);
221 ts.tv_sec %= 3600;
222 f = 1;
223 }
224 if (f || ts.tv_sec >= 60) {
225 printf("%ldm", (long)ts.tv_sec / 60);
226 ts.tv_sec %= 60;
227 f = 1;
228 }
229 printf("%lds\n", (long)ts.tv_sec);
230 }
231
232 static void
233 doadump(void)
234 {
235
236 /*
237 * Sometimes people have to call this from the kernel debugger.
238 * (if 'panic' can not dump)
239 * Give them a clue as to why they can't dump.
240 */
241 if (dumper.dumper == NULL) {
242 printf("Cannot dump. No dump device defined.\n");
243 return;
244 }
245
246 savectx(&dumppcb);
247 dumptid = curthread->td_tid;
248 dumping++;
249 #ifdef DDB
250 if (textdump_pending)
251 textdump_dumpsys(&dumper);
252 else
253 #endif
254 dumpsys(&dumper);
255 dumping--;
256 }
257
258 static int
259 isbufbusy(struct buf *bp)
260 {
261 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
262 BUF_REFCNT(bp) > 0) ||
263 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
264 return (1);
265 return (0);
266 }
267
268 /*
269 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
270 */
271 static void
272 boot(int howto)
273 {
274 static int first_buf_printf = 1;
275
276 #if defined(SMP)
277 /*
278 * Bind us to CPU 0 so that all shutdown code runs there. Some
279 * systems don't shutdown properly (i.e., ACPI power off) if we
280 * run on another processor.
281 */
282 thread_lock(curthread);
283 sched_bind(curthread, 0);
284 thread_unlock(curthread);
285 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
286 #endif
287 /* We're in the process of rebooting. */
288 rebooting = 1;
289
290 /* collect extra flags that shutdown_nice might have set */
291 howto |= shutdown_howto;
292
293 /* We are out of the debugger now. */
294 kdb_active = 0;
295
296 /*
297 * Do any callouts that should be done BEFORE syncing the filesystems.
298 */
299 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
300
301 /*
302 * Now sync filesystems
303 */
304 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
305 register struct buf *bp;
306 int iter, nbusy, pbusy;
307 #ifndef PREEMPTION
308 int subiter;
309 #endif
310
311 waittime = 0;
312
313 sync(curthread, NULL);
314
315 /*
316 * With soft updates, some buffers that are
317 * written will be remarked as dirty until other
318 * buffers are written.
319 */
320 for (iter = pbusy = 0; iter < 20; iter++) {
321 nbusy = 0;
322 for (bp = &buf[nbuf]; --bp >= buf; )
323 if (isbufbusy(bp))
324 nbusy++;
325 if (nbusy == 0) {
326 if (first_buf_printf)
327 printf("All buffers synced.");
328 break;
329 }
330 if (first_buf_printf) {
331 printf("Syncing disks, buffers remaining... ");
332 first_buf_printf = 0;
333 }
334 printf("%d ", nbusy);
335 if (nbusy < pbusy)
336 iter = 0;
337 pbusy = nbusy;
338 sync(curthread, NULL);
339
340 #ifdef PREEMPTION
341 /*
342 * Drop Giant and spin for a while to allow
343 * interrupt threads to run.
344 */
345 DROP_GIANT();
346 DELAY(50000 * iter);
347 PICKUP_GIANT();
348 #else
349 /*
350 * Drop Giant and context switch several times to
351 * allow interrupt threads to run.
352 */
353 DROP_GIANT();
354 for (subiter = 0; subiter < 50 * iter; subiter++) {
355 thread_lock(curthread);
356 mi_switch(SW_VOL, NULL);
357 thread_unlock(curthread);
358 DELAY(1000);
359 }
360 PICKUP_GIANT();
361 #endif
362 }
363 printf("\n");
364 /*
365 * Count only busy local buffers to prevent forcing
366 * a fsck if we're just a client of a wedged NFS server
367 */
368 nbusy = 0;
369 for (bp = &buf[nbuf]; --bp >= buf; ) {
370 if (isbufbusy(bp)) {
371 #if 0
372 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
373 if (bp->b_dev == NULL) {
374 TAILQ_REMOVE(&mountlist,
375 bp->b_vp->v_mount, mnt_list);
376 continue;
377 }
378 #endif
379 nbusy++;
380 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
381 printf(
382 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n",
383 nbusy, bp->b_bufobj,
384 bp->b_flags, (long)bp->b_blkno,
385 (long)bp->b_lblkno);
386 #endif
387 }
388 }
389 if (nbusy) {
390 /*
391 * Failed to sync all blocks. Indicate this and don't
392 * unmount filesystems (thus forcing an fsck on reboot).
393 */
394 printf("Giving up on %d buffers\n", nbusy);
395 DELAY(5000000); /* 5 seconds */
396 } else {
397 if (!first_buf_printf)
398 printf("Final sync complete\n");
399 /*
400 * Unmount filesystems
401 */
402 if (panicstr == 0)
403 vfs_unmountall();
404 }
405 swapoff_all();
406 DELAY(100000); /* wait for console output to finish */
407 }
408
409 print_uptime();
410
411 /*
412 * Ok, now do things that assume all filesystem activity has
413 * been completed.
414 */
415 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
416
417 /* XXX This doesn't disable interrupts any more. Reconsider? */
418 splhigh();
419
420 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
421 doadump();
422
423 /* Now that we're going to really halt the system... */
424 EVENTHANDLER_INVOKE(shutdown_final, howto);
425
426 for(;;) ; /* safety against shutdown_reset not working */
427 /* NOTREACHED */
428 }
429
430 /*
431 * If the shutdown was a clean halt, behave accordingly.
432 */
433 static void
434 shutdown_halt(void *junk, int howto)
435 {
436
437 if (howto & RB_HALT) {
438 printf("\n");
439 printf("The operating system has halted.\n");
440 printf("Please press any key to reboot.\n\n");
441 switch (cngetc()) {
442 case -1: /* No console, just die */
443 cpu_halt();
444 /* NOTREACHED */
445 default:
446 howto &= ~RB_HALT;
447 break;
448 }
449 }
450 }
451
452 /*
453 * Check to see if the system paniced, pause and then reboot
454 * according to the specified delay.
455 */
456 static void
457 shutdown_panic(void *junk, int howto)
458 {
459 int loop;
460
461 if (howto & RB_DUMP) {
462 if (PANIC_REBOOT_WAIT_TIME != 0) {
463 if (PANIC_REBOOT_WAIT_TIME != -1) {
464 printf("Automatic reboot in %d seconds - "
465 "press a key on the console to abort\n",
466 PANIC_REBOOT_WAIT_TIME);
467 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
468 loop > 0; --loop) {
469 DELAY(1000 * 100); /* 1/10th second */
470 /* Did user type a key? */
471 if (cncheckc() != -1)
472 break;
473 }
474 if (!loop)
475 return;
476 }
477 } else { /* zero time specified - reboot NOW */
478 return;
479 }
480 printf("--> Press a key on the console to reboot,\n");
481 printf("--> or switch off the system now.\n");
482 cngetc();
483 }
484 }
485
486 /*
487 * Everything done, now reset
488 */
489 static void
490 shutdown_reset(void *junk, int howto)
491 {
492
493 printf("Rebooting...\n");
494 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
495 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
496 cpu_reset();
497 /* NOTREACHED */ /* assuming reset worked */
498 }
499
500 /*
501 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
502 * and then reboots. If we are called twice, then we avoid trying to sync
503 * the disks as this often leads to recursive panics.
504 */
505 void
506 panic(const char *fmt, ...)
507 {
508 #ifdef SMP
509 static volatile u_int panic_cpu = NOCPU;
510 #endif
511 struct thread *td = curthread;
512 int bootopt, newpanic;
513 va_list ap;
514 static char buf[256];
515
516 critical_enter();
517 #ifdef SMP
518 /*
519 * We don't want multiple CPU's to panic at the same time, so we
520 * use panic_cpu as a simple spinlock. We have to keep checking
521 * panic_cpu if we are spinning in case the panic on the first
522 * CPU is canceled.
523 */
524 if (panic_cpu != PCPU_GET(cpuid))
525 while (atomic_cmpset_int(&panic_cpu, NOCPU,
526 PCPU_GET(cpuid)) == 0)
527 while (panic_cpu != NOCPU)
528 ; /* nothing */
529 #endif
530
531 bootopt = RB_AUTOBOOT | RB_DUMP;
532 newpanic = 0;
533 if (panicstr)
534 bootopt |= RB_NOSYNC;
535 else {
536 panicstr = fmt;
537 newpanic = 1;
538 }
539
540 va_start(ap, fmt);
541 if (newpanic) {
542 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
543 panicstr = buf;
544 printf("panic: %s\n", buf);
545 } else {
546 printf("panic: ");
547 vprintf(fmt, ap);
548 printf("\n");
549 }
550 va_end(ap);
551 #ifdef SMP
552 printf("cpuid = %d\n", PCPU_GET(cpuid));
553 #endif
554
555 #ifdef KDB
556 if (newpanic && trace_on_panic)
557 kdb_backtrace();
558 if (debugger_on_panic)
559 kdb_enter_why(KDB_WHY_PANIC, "panic");
560 #ifdef RESTARTABLE_PANICS
561 /* See if the user aborted the panic, in which case we continue. */
562 if (panicstr == NULL) {
563 #ifdef SMP
564 atomic_store_rel_int(&panic_cpu, NOCPU);
565 #endif
566 return;
567 }
568 #endif
569 #endif
570 /*thread_lock(td); */
571 td->td_flags |= TDF_INPANIC;
572 /* thread_unlock(td); */
573 if (!sync_on_panic)
574 bootopt |= RB_NOSYNC;
575 critical_exit();
576 boot(bootopt);
577 }
578
579 /*
580 * Support for poweroff delay.
581 */
582 #ifndef POWEROFF_DELAY
583 # define POWEROFF_DELAY 5000
584 #endif
585 static int poweroff_delay = POWEROFF_DELAY;
586
587 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
588 &poweroff_delay, 0, "");
589
590 static void
591 poweroff_wait(void *junk, int howto)
592 {
593
594 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
595 return;
596 DELAY(poweroff_delay * 1000);
597 }
598
599 /*
600 * Some system processes (e.g. syncer) need to be stopped at appropriate
601 * points in their main loops prior to a system shutdown, so that they
602 * won't interfere with the shutdown process (e.g. by holding a disk buf
603 * to cause sync to fail). For each of these system processes, register
604 * shutdown_kproc() as a handler for one of shutdown events.
605 */
606 static int kproc_shutdown_wait = 60;
607 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
608 &kproc_shutdown_wait, 0, "");
609
610 void
611 kproc_shutdown(void *arg, int howto)
612 {
613 struct proc *p;
614 char procname[MAXCOMLEN + 1];
615 int error;
616
617 if (panicstr)
618 return;
619
620 p = (struct proc *)arg;
621 strlcpy(procname, p->p_comm, sizeof(procname));
622 printf("Waiting (max %d seconds) for system process `%s' to stop...",
623 kproc_shutdown_wait, procname);
624 error = kthread_suspend(p, kproc_shutdown_wait * hz);
625
626 if (error == EWOULDBLOCK)
627 printf("timed out\n");
628 else
629 printf("done\n");
630 }
631
632 /* Registration of dumpers */
633 int
634 set_dumper(struct dumperinfo *di)
635 {
636
637 if (di == NULL) {
638 bzero(&dumper, sizeof dumper);
639 return (0);
640 }
641 if (dumper.dumper != NULL)
642 return (EBUSY);
643 dumper = *di;
644 return (0);
645 }
646
647 /* Call dumper with bounds checking. */
648 int
649 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
650 off_t offset, size_t length)
651 {
652
653 if (length != 0 && (offset < di->mediaoffset ||
654 offset - di->mediaoffset + length > di->mediasize)) {
655 printf("Attempt to write outside dump device boundaries.\n");
656 return (ENXIO);
657 }
658 return (di->dumper(di->priv, virtual, physical, offset, length));
659 }
660
661 #if defined(__powerpc__)
662 void
663 dumpsys(struct dumperinfo *di __unused)
664 {
665
666 printf("Kernel dumps not implemented on this architecture\n");
667 }
668 #endif
Cache object: af259f3e49568eba9b3a54f4b4a2f321
|