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/8.2/sys/kern/kern_shutdown.c 213934 2010-10-16 19:58:49Z avg $");
39
40 #include "opt_ddb.h"
41 #include "opt_kdb.h"
42 #include "opt_panic.h"
43 #include "opt_show_busybufs.h"
44 #include "opt_sched.h"
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/bio.h>
49 #include <sys/buf.h>
50 #include <sys/conf.h>
51 #include <sys/cons.h>
52 #include <sys/eventhandler.h>
53 #include <sys/jail.h>
54 #include <sys/kdb.h>
55 #include <sys/kernel.h>
56 #include <sys/kerneldump.h>
57 #include <sys/kthread.h>
58 #include <sys/malloc.h>
59 #include <sys/mount.h>
60 #include <sys/priv.h>
61 #include <sys/proc.h>
62 #include <sys/reboot.h>
63 #include <sys/resourcevar.h>
64 #include <sys/sched.h>
65 #include <sys/smp.h>
66 #include <sys/sysctl.h>
67 #include <sys/sysproto.h>
68
69 #include <ddb/ddb.h>
70
71 #include <machine/cpu.h>
72 #include <machine/pcb.h>
73 #include <machine/smp.h>
74
75 #include <security/mac/mac_framework.h>
76
77 #include <vm/vm.h>
78 #include <vm/vm_object.h>
79 #include <vm/vm_page.h>
80 #include <vm/vm_pager.h>
81 #include <vm/swap_pager.h>
82
83 #include <sys/signalvar.h>
84
85 #ifndef PANIC_REBOOT_WAIT_TIME
86 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
87 #endif
88
89 /*
90 * Note that stdarg.h and the ANSI style va_start macro is used for both
91 * ANSI and traditional C compilers.
92 */
93 #include <machine/stdarg.h>
94
95 #ifdef KDB
96 #ifdef KDB_UNATTENDED
97 int debugger_on_panic = 0;
98 #else
99 int debugger_on_panic = 1;
100 #endif
101 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
102 &debugger_on_panic, 0, "Run debugger on kernel panic");
103 TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic);
104
105 #ifdef KDB_TRACE
106 static int trace_on_panic = 1;
107 #else
108 static int trace_on_panic = 0;
109 #endif
110 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
111 &trace_on_panic, 0, "Print stack trace on kernel panic");
112 TUNABLE_INT("debug.trace_on_panic", &trace_on_panic);
113 #endif /* KDB */
114
115 static int sync_on_panic = 0;
116 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN,
117 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
118 TUNABLE_INT("kern.sync_on_panic", &sync_on_panic);
119
120 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
121
122 /*
123 * Variable panicstr contains argument to first call to panic; used as flag
124 * to indicate that the kernel has already called panic.
125 */
126 const char *panicstr;
127
128 int dumping; /* system is dumping */
129 int rebooting; /* system is rebooting */
130 static struct dumperinfo dumper; /* our selected dumper */
131
132 /* Context information for dump-debuggers. */
133 static struct pcb dumppcb; /* Registers. */
134 static lwpid_t dumptid; /* Thread ID. */
135
136 static void boot(int) __dead2;
137 static void poweroff_wait(void *, int);
138 static void shutdown_halt(void *junk, int howto);
139 static void shutdown_panic(void *junk, int howto);
140 static void shutdown_reset(void *junk, int howto);
141
142 /* register various local shutdown events */
143 static void
144 shutdown_conf(void *unused)
145 {
146
147 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
148 SHUTDOWN_PRI_FIRST);
149 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
150 SHUTDOWN_PRI_LAST + 100);
151 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
152 SHUTDOWN_PRI_LAST + 100);
153 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
154 SHUTDOWN_PRI_LAST + 200);
155 }
156
157 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL);
158
159 /*
160 * The system call that results in a reboot.
161 */
162 /* ARGSUSED */
163 int
164 reboot(struct thread *td, struct reboot_args *uap)
165 {
166 int error;
167
168 error = 0;
169 #ifdef MAC
170 error = mac_system_check_reboot(td->td_ucred, uap->opt);
171 #endif
172 if (error == 0)
173 error = priv_check(td, PRIV_REBOOT);
174 if (error == 0) {
175 mtx_lock(&Giant);
176 boot(uap->opt);
177 mtx_unlock(&Giant);
178 }
179 return (error);
180 }
181
182 /*
183 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
184 */
185 static int shutdown_howto = 0;
186
187 void
188 shutdown_nice(int howto)
189 {
190
191 shutdown_howto = howto;
192
193 /* Send a signal to init(8) and have it shutdown the world */
194 if (initproc != NULL) {
195 PROC_LOCK(initproc);
196 psignal(initproc, SIGINT);
197 PROC_UNLOCK(initproc);
198 } else {
199 /* No init(8) running, so simply reboot */
200 boot(RB_NOSYNC);
201 }
202 return;
203 }
204 static int waittime = -1;
205
206 static void
207 print_uptime(void)
208 {
209 int f;
210 struct timespec ts;
211
212 getnanouptime(&ts);
213 printf("Uptime: ");
214 f = 0;
215 if (ts.tv_sec >= 86400) {
216 printf("%ldd", (long)ts.tv_sec / 86400);
217 ts.tv_sec %= 86400;
218 f = 1;
219 }
220 if (f || ts.tv_sec >= 3600) {
221 printf("%ldh", (long)ts.tv_sec / 3600);
222 ts.tv_sec %= 3600;
223 f = 1;
224 }
225 if (f || ts.tv_sec >= 60) {
226 printf("%ldm", (long)ts.tv_sec / 60);
227 ts.tv_sec %= 60;
228 f = 1;
229 }
230 printf("%lds\n", (long)ts.tv_sec);
231 }
232
233 static void
234 doadump(void)
235 {
236
237 /*
238 * Sometimes people have to call this from the kernel debugger.
239 * (if 'panic' can not dump)
240 * Give them a clue as to why they can't dump.
241 */
242 if (dumper.dumper == NULL) {
243 printf("Cannot dump. Device not defined or unavailable.\n");
244 return;
245 }
246
247 savectx(&dumppcb);
248 dumptid = curthread->td_tid;
249 dumping++;
250 #ifdef DDB
251 if (textdump_pending)
252 textdump_dumpsys(&dumper);
253 else
254 #endif
255 dumpsys(&dumper);
256 dumping--;
257 }
258
259 static int
260 isbufbusy(struct buf *bp)
261 {
262 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
263 BUF_ISLOCKED(bp)) ||
264 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
265 return (1);
266 return (0);
267 }
268
269 /*
270 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
271 */
272 static void
273 boot(int howto)
274 {
275 static int first_buf_printf = 1;
276
277 #if defined(SMP)
278 /*
279 * Bind us to CPU 0 so that all shutdown code runs there. Some
280 * systems don't shutdown properly (i.e., ACPI power off) if we
281 * run on another processor.
282 */
283 thread_lock(curthread);
284 sched_bind(curthread, 0);
285 thread_unlock(curthread);
286 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0"));
287 #endif
288 /* We're in the process of rebooting. */
289 rebooting = 1;
290
291 /* collect extra flags that shutdown_nice might have set */
292 howto |= shutdown_howto;
293
294 /* We are out of the debugger now. */
295 kdb_active = 0;
296
297 /*
298 * Do any callouts that should be done BEFORE syncing the filesystems.
299 */
300 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
301
302 /*
303 * Now sync filesystems
304 */
305 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
306 register struct buf *bp;
307 int iter, nbusy, pbusy;
308 #ifndef PREEMPTION
309 int subiter;
310 #endif
311
312 waittime = 0;
313
314 sync(curthread, NULL);
315
316 /*
317 * With soft updates, some buffers that are
318 * written will be remarked as dirty until other
319 * buffers are written.
320 */
321 for (iter = pbusy = 0; iter < 20; iter++) {
322 nbusy = 0;
323 for (bp = &buf[nbuf]; --bp >= buf; )
324 if (isbufbusy(bp))
325 nbusy++;
326 if (nbusy == 0) {
327 if (first_buf_printf)
328 printf("All buffers synced.");
329 break;
330 }
331 if (first_buf_printf) {
332 printf("Syncing disks, buffers remaining... ");
333 first_buf_printf = 0;
334 }
335 printf("%d ", nbusy);
336 if (nbusy < pbusy)
337 iter = 0;
338 pbusy = nbusy;
339 sync(curthread, NULL);
340
341 #ifdef PREEMPTION
342 /*
343 * Drop Giant and spin for a while to allow
344 * interrupt threads to run.
345 */
346 DROP_GIANT();
347 DELAY(50000 * iter);
348 PICKUP_GIANT();
349 #else
350 /*
351 * Drop Giant and context switch several times to
352 * allow interrupt threads to run.
353 */
354 DROP_GIANT();
355 for (subiter = 0; subiter < 50 * iter; subiter++) {
356 thread_lock(curthread);
357 mi_switch(SW_VOL, NULL);
358 thread_unlock(curthread);
359 DELAY(1000);
360 }
361 PICKUP_GIANT();
362 #endif
363 }
364 printf("\n");
365 /*
366 * Count only busy local buffers to prevent forcing
367 * a fsck if we're just a client of a wedged NFS server
368 */
369 nbusy = 0;
370 for (bp = &buf[nbuf]; --bp >= buf; ) {
371 if (isbufbusy(bp)) {
372 #if 0
373 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */
374 if (bp->b_dev == NULL) {
375 TAILQ_REMOVE(&mountlist,
376 bp->b_vp->v_mount, mnt_list);
377 continue;
378 }
379 #endif
380 nbusy++;
381 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
382 printf(
383 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n",
384 nbusy, bp->b_bufobj,
385 bp->b_flags, (long)bp->b_blkno,
386 (long)bp->b_lblkno);
387 #endif
388 }
389 }
390 if (nbusy) {
391 /*
392 * Failed to sync all blocks. Indicate this and don't
393 * unmount filesystems (thus forcing an fsck on reboot).
394 */
395 printf("Giving up on %d buffers\n", nbusy);
396 DELAY(5000000); /* 5 seconds */
397 } else {
398 if (!first_buf_printf)
399 printf("Final sync complete\n");
400 /*
401 * Unmount filesystems
402 */
403 if (panicstr == 0)
404 vfs_unmountall();
405 }
406 swapoff_all();
407 DELAY(100000); /* wait for console output to finish */
408 }
409
410 print_uptime();
411
412 /*
413 * Ok, now do things that assume all filesystem activity has
414 * been completed.
415 */
416 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
417
418 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
419 doadump();
420
421 /* Now that we're going to really halt the system... */
422 EVENTHANDLER_INVOKE(shutdown_final, howto);
423
424 for(;;) ; /* safety against shutdown_reset not working */
425 /* NOTREACHED */
426 }
427
428 /*
429 * If the shutdown was a clean halt, behave accordingly.
430 */
431 static void
432 shutdown_halt(void *junk, int howto)
433 {
434
435 if (howto & RB_HALT) {
436 printf("\n");
437 printf("The operating system has halted.\n");
438 printf("Please press any key to reboot.\n\n");
439 switch (cngetc()) {
440 case -1: /* No console, just die */
441 cpu_halt();
442 /* NOTREACHED */
443 default:
444 howto &= ~RB_HALT;
445 break;
446 }
447 }
448 }
449
450 /*
451 * Check to see if the system paniced, pause and then reboot
452 * according to the specified delay.
453 */
454 static void
455 shutdown_panic(void *junk, int howto)
456 {
457 int loop;
458
459 if (howto & RB_DUMP) {
460 if (PANIC_REBOOT_WAIT_TIME != 0) {
461 if (PANIC_REBOOT_WAIT_TIME != -1) {
462 printf("Automatic reboot in %d seconds - "
463 "press a key on the console to abort\n",
464 PANIC_REBOOT_WAIT_TIME);
465 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
466 loop > 0; --loop) {
467 DELAY(1000 * 100); /* 1/10th second */
468 /* Did user type a key? */
469 if (cncheckc() != -1)
470 break;
471 }
472 if (!loop)
473 return;
474 }
475 } else { /* zero time specified - reboot NOW */
476 return;
477 }
478 printf("--> Press a key on the console to reboot,\n");
479 printf("--> or switch off the system now.\n");
480 cngetc();
481 }
482 }
483
484 /*
485 * Everything done, now reset
486 */
487 static void
488 shutdown_reset(void *junk, int howto)
489 {
490
491 printf("Rebooting...\n");
492 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
493
494 /*
495 * Acquiring smp_ipi_mtx here has a double effect:
496 * - it disables interrupts avoiding CPU0 preemption
497 * by fast handlers (thus deadlocking against other CPUs)
498 * - it avoids deadlocks against smp_rendezvous() or, more
499 * generally, threads busy-waiting, with this spinlock held,
500 * and waiting for responses by threads on other CPUs
501 * (ie. smp_tlb_shootdown()).
502 *
503 * For the !SMP case it just needs to handle the former problem.
504 */
505 #ifdef SMP
506 mtx_lock_spin(&smp_ipi_mtx);
507 #else
508 spinlock_enter();
509 #endif
510
511 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
512 cpu_reset();
513 /* NOTREACHED */ /* assuming reset worked */
514 }
515
516 /*
517 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
518 * and then reboots. If we are called twice, then we avoid trying to sync
519 * the disks as this often leads to recursive panics.
520 */
521 void
522 panic(const char *fmt, ...)
523 {
524 #ifdef SMP
525 static volatile u_int panic_cpu = NOCPU;
526 #endif
527 struct thread *td = curthread;
528 int bootopt, newpanic;
529 va_list ap;
530 static char buf[256];
531
532 critical_enter();
533 #ifdef SMP
534 /*
535 * We don't want multiple CPU's to panic at the same time, so we
536 * use panic_cpu as a simple spinlock. We have to keep checking
537 * panic_cpu if we are spinning in case the panic on the first
538 * CPU is canceled.
539 */
540 if (panic_cpu != PCPU_GET(cpuid))
541 while (atomic_cmpset_int(&panic_cpu, NOCPU,
542 PCPU_GET(cpuid)) == 0)
543 while (panic_cpu != NOCPU)
544 ; /* nothing */
545 #endif
546
547 bootopt = RB_AUTOBOOT | RB_DUMP;
548 newpanic = 0;
549 if (panicstr)
550 bootopt |= RB_NOSYNC;
551 else {
552 panicstr = fmt;
553 newpanic = 1;
554 }
555
556 va_start(ap, fmt);
557 if (newpanic) {
558 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
559 panicstr = buf;
560 printf("panic: %s\n", buf);
561 } else {
562 printf("panic: ");
563 vprintf(fmt, ap);
564 printf("\n");
565 }
566 va_end(ap);
567 #ifdef SMP
568 printf("cpuid = %d\n", PCPU_GET(cpuid));
569 #endif
570
571 #ifdef KDB
572 if (newpanic && trace_on_panic)
573 kdb_backtrace();
574 if (debugger_on_panic)
575 kdb_enter(KDB_WHY_PANIC, "panic");
576 #ifdef RESTARTABLE_PANICS
577 /* See if the user aborted the panic, in which case we continue. */
578 if (panicstr == NULL) {
579 #ifdef SMP
580 atomic_store_rel_int(&panic_cpu, NOCPU);
581 #endif
582 return;
583 }
584 #endif
585 #endif
586 /*thread_lock(td); */
587 td->td_flags |= TDF_INPANIC;
588 /* thread_unlock(td); */
589 if (!sync_on_panic)
590 bootopt |= RB_NOSYNC;
591 critical_exit();
592 boot(bootopt);
593 }
594
595 /*
596 * Support for poweroff delay.
597 */
598 #ifndef POWEROFF_DELAY
599 # define POWEROFF_DELAY 5000
600 #endif
601 static int poweroff_delay = POWEROFF_DELAY;
602
603 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
604 &poweroff_delay, 0, "");
605
606 static void
607 poweroff_wait(void *junk, int howto)
608 {
609
610 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
611 return;
612 DELAY(poweroff_delay * 1000);
613 }
614
615 /*
616 * Some system processes (e.g. syncer) need to be stopped at appropriate
617 * points in their main loops prior to a system shutdown, so that they
618 * won't interfere with the shutdown process (e.g. by holding a disk buf
619 * to cause sync to fail). For each of these system processes, register
620 * shutdown_kproc() as a handler for one of shutdown events.
621 */
622 static int kproc_shutdown_wait = 60;
623 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
624 &kproc_shutdown_wait, 0, "");
625
626 void
627 kproc_shutdown(void *arg, int howto)
628 {
629 struct proc *p;
630 char procname[MAXCOMLEN + 1];
631 int error;
632
633 if (panicstr)
634 return;
635
636 p = (struct proc *)arg;
637 strlcpy(procname, p->p_comm, sizeof(procname));
638 printf("Waiting (max %d seconds) for system process `%s' to stop...",
639 kproc_shutdown_wait, procname);
640 error = kproc_suspend(p, kproc_shutdown_wait * hz);
641
642 if (error == EWOULDBLOCK)
643 printf("timed out\n");
644 else
645 printf("done\n");
646 }
647
648 void
649 kthread_shutdown(void *arg, int howto)
650 {
651 struct thread *td;
652 char procname[MAXCOMLEN + 1];
653 int error;
654
655 if (panicstr)
656 return;
657
658 td = (struct thread *)arg;
659 strlcpy(procname, td->td_name, sizeof(procname));
660 printf("Waiting (max %d seconds) for system thread `%s' to stop...",
661 kproc_shutdown_wait, procname);
662 error = kthread_suspend(td, kproc_shutdown_wait * hz);
663
664 if (error == EWOULDBLOCK)
665 printf("timed out\n");
666 else
667 printf("done\n");
668 }
669
670 /* Registration of dumpers */
671 int
672 set_dumper(struct dumperinfo *di)
673 {
674
675 if (di == NULL) {
676 bzero(&dumper, sizeof dumper);
677 return (0);
678 }
679 if (dumper.dumper != NULL)
680 return (EBUSY);
681 dumper = *di;
682 return (0);
683 }
684
685 /* Call dumper with bounds checking. */
686 int
687 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical,
688 off_t offset, size_t length)
689 {
690
691 if (length != 0 && (offset < di->mediaoffset ||
692 offset - di->mediaoffset + length > di->mediasize)) {
693 printf("Attempt to write outside dump device boundaries.\n");
694 return (ENXIO);
695 }
696 return (di->dumper(di->priv, virtual, physical, offset, length));
697 }
698
699 void
700 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver,
701 uint64_t dumplen, uint32_t blksz)
702 {
703
704 bzero(kdh, sizeof(*kdh));
705 strncpy(kdh->magic, magic, sizeof(kdh->magic));
706 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
707 kdh->version = htod32(KERNELDUMPVERSION);
708 kdh->architectureversion = htod32(archver);
709 kdh->dumplength = htod64(dumplen);
710 kdh->dumptime = htod64(time_second);
711 kdh->blocksize = htod32(blksz);
712 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname));
713 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
714 if (panicstr != NULL)
715 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
716 kdh->parity = kerneldump_parity(kdh);
717 }
Cache object: 8781150b48ca41b04944373011f619d3
|