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("$FreeBSD$");
39
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"
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/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/sched.h>
64 #include <sys/smp.h> /* smp_active */
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/vnode.h>
68
69 #include <machine/cpu.h>
70 #include <machine/pcb.h>
71 #include <machine/smp.h>
72
73 #include <sys/signalvar.h>
74
75 #ifndef PANIC_REBOOT_WAIT_TIME
76 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */
77 #endif
78
79 /*
80 * Note that stdarg.h and the ANSI style va_start macro is used for both
81 * ANSI and traditional C compilers.
82 */
83 #include <machine/stdarg.h>
84
85 #ifdef KDB
86 #ifdef KDB_UNATTENDED
87 int debugger_on_panic = 0;
88 #else
89 int debugger_on_panic = 1;
90 #endif
91 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
92 &debugger_on_panic, 0, "Run debugger on kernel panic");
93
94 #ifdef KDB_TRACE
95 int trace_on_panic = 1;
96 #else
97 int trace_on_panic = 0;
98 #endif
99 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
100 &trace_on_panic, 0, "Print stack trace on kernel panic");
101 #endif /* KDB */
102
103 int sync_on_panic = 0;
104 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
105 &sync_on_panic, 0, "Do a sync before rebooting from a panic");
106
107 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
108
109 #ifdef HW_WDOG
110 /*
111 * If there is a hardware watchdog, point this at the function needed to
112 * hold it off.
113 * It's needed when the kernel needs to do some lengthy operations.
114 * e.g. in wd.c when dumping core.. It's most annoying to have
115 * your precious core-dump only half written because the wdog kicked in.
116 */
117 watchdog_tickle_fn wdog_tickler = NULL;
118 #endif /* HW_WDOG */
119
120 /*
121 * Variable panicstr contains argument to first call to panic; used as flag
122 * to indicate that the kernel has already called panic.
123 */
124 const char *panicstr;
125
126 int dumping; /* system is dumping */
127 static struct dumperinfo dumper; /* our selected dumper */
128
129 /* Context information for dump-debuggers. */
130 static struct pcb dumppcb; /* Registers. */
131 static lwpid_t dumptid; /* Thread ID. */
132
133 static void boot(int) __dead2;
134 static void poweroff_wait(void *, int);
135 static void shutdown_halt(void *junk, int howto);
136 static void shutdown_panic(void *junk, int howto);
137 static void shutdown_reset(void *junk, int howto);
138
139 /* register various local shutdown events */
140 static void
141 shutdown_conf(void *unused)
142 {
143
144 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL,
145 SHUTDOWN_PRI_FIRST);
146 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL,
147 SHUTDOWN_PRI_LAST + 100);
148 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL,
149 SHUTDOWN_PRI_LAST + 100);
150 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL,
151 SHUTDOWN_PRI_LAST + 200);
152 }
153
154 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL)
155
156 /*
157 * The system call that results in a reboot
158 *
159 * MPSAFE
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 = suser(td);
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 dumpsys(&dumper);
250 }
251
252 static int
253 isbufbusy(struct buf *bp)
254 {
255 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 &&
256 BUF_REFCNT(bp) > 0) ||
257 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI))
258 return (1);
259 return (0);
260 }
261
262 /*
263 * Shutdown the system cleanly to prepare for reboot, halt, or power off.
264 */
265 static void
266 boot(int howto)
267 {
268 static int first_buf_printf = 1;
269
270 #if defined(SMP)
271 /*
272 * Bind us to CPU 0 so that all shutdown code runs there. Some
273 * systems don't shutdown properly (i.e., ACPI power off) if we
274 * run on another processor.
275 */
276 mtx_lock_spin(&sched_lock);
277 sched_bind(curthread, 0);
278 mtx_unlock_spin(&sched_lock);
279 #endif
280
281 /* collect extra flags that shutdown_nice might have set */
282 howto |= shutdown_howto;
283
284 /* We are out of the debugger now. */
285 kdb_active = 0;
286
287 #ifdef SMP
288 if (smp_active)
289 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
290 #endif
291 /*
292 * Do any callouts that should be done BEFORE syncing the filesystems.
293 */
294 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
295
296 /*
297 * Now sync filesystems
298 */
299 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
300 register struct buf *bp;
301 int iter, nbusy, pbusy;
302 #ifndef PREEMPTION
303 int subiter;
304 #endif
305
306 waittime = 0;
307
308 sync(&thread0, NULL);
309
310 /*
311 * With soft updates, some buffers that are
312 * written will be remarked as dirty until other
313 * buffers are written.
314 */
315 for (iter = pbusy = 0; iter < 20; iter++) {
316 nbusy = 0;
317 for (bp = &buf[nbuf]; --bp >= buf; )
318 if (isbufbusy(bp))
319 nbusy++;
320 if (nbusy == 0) {
321 if (first_buf_printf)
322 printf("No buffers busy after final sync");
323 break;
324 }
325 if (first_buf_printf) {
326 printf("Syncing disks, buffers remaining... ");
327 first_buf_printf = 0;
328 }
329 printf("%d ", nbusy);
330 if (nbusy < pbusy)
331 iter = 0;
332 pbusy = nbusy;
333 sync(&thread0, NULL);
334
335 #ifdef PREEMPTION
336 /*
337 * Drop Giant and spin for a while to allow
338 * interrupt threads to run.
339 */
340 DROP_GIANT();
341 DELAY(50000 * iter);
342 PICKUP_GIANT();
343 #else
344 /*
345 * Drop Giant and context switch several times to
346 * allow interrupt threads to run.
347 */
348 DROP_GIANT();
349 for (subiter = 0; subiter < 50 * iter; subiter++) {
350 mtx_lock_spin(&sched_lock);
351 mi_switch(SW_VOL, NULL);
352 mtx_unlock_spin(&sched_lock);
353 DELAY(1000);
354 }
355 PICKUP_GIANT();
356 #endif
357 }
358 printf("\n");
359 /*
360 * Count only busy local buffers to prevent forcing
361 * a fsck if we're just a client of a wedged NFS server
362 */
363 nbusy = 0;
364 for (bp = &buf[nbuf]; --bp >= buf; ) {
365 if (isbufbusy(bp)) {
366 if (bp->b_dev == NULL) {
367 TAILQ_REMOVE(&mountlist,
368 bp->b_vp->v_mount, mnt_list);
369 continue;
370 }
371 nbusy++;
372 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
373 printf(
374 "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n",
375 nbusy, devtoname(bp->b_dev),
376 bp->b_flags, (long)bp->b_blkno,
377 (long)bp->b_lblkno);
378 #endif
379 }
380 }
381 if (nbusy) {
382 /*
383 * Failed to sync all blocks. Indicate this and don't
384 * unmount filesystems (thus forcing an fsck on reboot).
385 */
386 printf("Giving up on %d buffers\n", nbusy);
387 DELAY(5000000); /* 5 seconds */
388 } else {
389 if (!first_buf_printf)
390 printf("Final sync complete\n");
391 /*
392 * Unmount filesystems
393 */
394 if (panicstr == 0)
395 vfs_unmountall();
396 }
397 DELAY(100000); /* wait for console output to finish */
398 }
399
400 print_uptime();
401
402 /*
403 * Ok, now do things that assume all filesystem activity has
404 * been completed.
405 */
406 EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
407
408 /* XXX This doesn't disable interrupts any more. Reconsider? */
409 splhigh();
410
411 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping)
412 doadump();
413
414 /* Now that we're going to really halt the system... */
415 EVENTHANDLER_INVOKE(shutdown_final, howto);
416
417 for(;;) ; /* safety against shutdown_reset not working */
418 /* NOTREACHED */
419 }
420
421 /*
422 * If the shutdown was a clean halt, behave accordingly.
423 */
424 static void
425 shutdown_halt(void *junk, int howto)
426 {
427
428 if (howto & RB_HALT) {
429 printf("\n");
430 printf("The operating system has halted.\n");
431 printf("Please press any key to reboot.\n\n");
432 switch (cngetc()) {
433 case -1: /* No console, just die */
434 cpu_halt();
435 /* NOTREACHED */
436 default:
437 howto &= ~RB_HALT;
438 break;
439 }
440 }
441 }
442
443 /*
444 * Check to see if the system paniced, pause and then reboot
445 * according to the specified delay.
446 */
447 static void
448 shutdown_panic(void *junk, int howto)
449 {
450 int loop;
451
452 if (howto & RB_DUMP) {
453 if (PANIC_REBOOT_WAIT_TIME != 0) {
454 if (PANIC_REBOOT_WAIT_TIME != -1) {
455 printf("Automatic reboot in %d seconds - "
456 "press a key on the console to abort\n",
457 PANIC_REBOOT_WAIT_TIME);
458 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
459 loop > 0; --loop) {
460 DELAY(1000 * 100); /* 1/10th second */
461 /* Did user type a key? */
462 if (cncheckc() != -1)
463 break;
464 }
465 if (!loop)
466 return;
467 }
468 } else { /* zero time specified - reboot NOW */
469 return;
470 }
471 printf("--> Press a key on the console to reboot,\n");
472 printf("--> or switch off the system now.\n");
473 cngetc();
474 }
475 }
476
477 /*
478 * Everything done, now reset
479 */
480 static void
481 shutdown_reset(void *junk, int howto)
482 {
483
484 printf("Rebooting...\n");
485 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */
486 /* cpu_boot(howto); */ /* doesn't do anything at the moment */
487 cpu_reset();
488 /* NOTREACHED */ /* assuming reset worked */
489 }
490
491 #ifdef SMP
492 static u_int panic_cpu = NOCPU;
493 #endif
494
495 /*
496 * Panic is called on unresolvable fatal errors. It prints "panic: mesg",
497 * and then reboots. If we are called twice, then we avoid trying to sync
498 * the disks as this often leads to recursive panics.
499 *
500 * MPSAFE
501 */
502 void
503 panic(const char *fmt, ...)
504 {
505 struct thread *td = curthread;
506 int bootopt, newpanic;
507 va_list ap;
508 static char buf[256];
509
510 #ifdef SMP
511 /*
512 * We don't want multiple CPU's to panic at the same time, so we
513 * use panic_cpu as a simple spinlock. We have to keep checking
514 * panic_cpu if we are spinning in case the panic on the first
515 * CPU is canceled.
516 */
517 if (panic_cpu != PCPU_GET(cpuid))
518 while (atomic_cmpset_int(&panic_cpu, NOCPU,
519 PCPU_GET(cpuid)) == 0)
520 while (panic_cpu != NOCPU)
521 ; /* nothing */
522 #endif
523
524 bootopt = RB_AUTOBOOT | RB_DUMP;
525 newpanic = 0;
526 if (panicstr)
527 bootopt |= RB_NOSYNC;
528 else {
529 panicstr = fmt;
530 newpanic = 1;
531 }
532
533 va_start(ap, fmt);
534 if (newpanic) {
535 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
536 panicstr = buf;
537 printf("panic: %s\n", buf);
538 } else {
539 printf("panic: ");
540 vprintf(fmt, ap);
541 printf("\n");
542 }
543 va_end(ap);
544 #ifdef SMP
545 printf("cpuid = %d\n", PCPU_GET(cpuid));
546 #endif
547
548 #ifdef KDB
549 if (newpanic && trace_on_panic)
550 kdb_backtrace();
551 if (debugger_on_panic)
552 kdb_enter("panic");
553 #ifdef RESTARTABLE_PANICS
554 /* See if the user aborted the panic, in which case we continue. */
555 if (panicstr == NULL) {
556 #ifdef SMP
557 atomic_store_rel_int(&panic_cpu, NOCPU);
558 #endif
559 return;
560 }
561 #endif
562 #endif
563 mtx_lock_spin(&sched_lock);
564 td->td_flags |= TDF_INPANIC;
565 mtx_unlock_spin(&sched_lock);
566 if (!sync_on_panic)
567 bootopt |= RB_NOSYNC;
568 boot(bootopt);
569 }
570
571 /*
572 * Support for poweroff delay.
573 */
574 #ifndef POWEROFF_DELAY
575 # define POWEROFF_DELAY 5000
576 #endif
577 static int poweroff_delay = POWEROFF_DELAY;
578
579 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
580 &poweroff_delay, 0, "");
581
582 static void
583 poweroff_wait(void *junk, int howto)
584 {
585
586 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0)
587 return;
588 DELAY(poweroff_delay * 1000);
589 }
590
591 /*
592 * Some system processes (e.g. syncer) need to be stopped at appropriate
593 * points in their main loops prior to a system shutdown, so that they
594 * won't interfere with the shutdown process (e.g. by holding a disk buf
595 * to cause sync to fail). For each of these system processes, register
596 * shutdown_kproc() as a handler for one of shutdown events.
597 */
598 static int kproc_shutdown_wait = 60;
599 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
600 &kproc_shutdown_wait, 0, "");
601
602 void
603 kproc_shutdown(void *arg, int howto)
604 {
605 struct proc *p;
606 char procname[MAXCOMLEN + 1];
607 int error;
608
609 if (panicstr)
610 return;
611
612 p = (struct proc *)arg;
613 strlcpy(procname, p->p_comm, sizeof(procname));
614 printf("Waiting (max %d seconds) for system process `%s' to stop...",
615 kproc_shutdown_wait, procname);
616 error = kthread_suspend(p, kproc_shutdown_wait * hz);
617
618 if (error == EWOULDBLOCK)
619 printf("timed out\n");
620 else
621 printf("done\n");
622 }
623
624 /* Registration of dumpers */
625 int
626 set_dumper(struct dumperinfo *di)
627 {
628
629 if (di == NULL) {
630 bzero(&dumper, sizeof dumper);
631 return (0);
632 }
633 if (dumper.dumper != NULL)
634 return (EBUSY);
635 dumper = *di;
636 return (0);
637 }
638
639 #if defined(__powerpc__)
640 void
641 dumpsys(struct dumperinfo *di __unused)
642 {
643
644 printf("Kernel dumps not implemented on this architecture\n");
645 }
646 #endif
Cache object: 5190532703d404a7a7a7b7742d8f1877
|