kern_shutdown.c revision 108682
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 * 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. 25 * 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 36 * SUCH DAMAGE. 37 * 38 * @(#)kern_shutdown.c 8.3 (Berkeley) 1/21/94 39 * $FreeBSD: head/sys/kern/kern_shutdown.c 108682 2003-01-04 20:54:58Z phk $ 40 */ 41 42#include "opt_ddb.h" 43#include "opt_ddb_trace.h" 44#include "opt_ddb_unattended.h" 45#include "opt_hw_wdog.h" 46#include "opt_mac.h" 47#include "opt_panic.h" 48#include "opt_show_busybufs.h" 49 50#include <sys/param.h> 51#include <sys/systm.h> 52#include <sys/bio.h> 53#include <sys/buf.h> 54#include <sys/conf.h> 55#include <sys/cons.h> 56#include <sys/disklabel.h> 57#include <sys/eventhandler.h> 58#include <sys/kernel.h> 59#include <sys/kthread.h> 60#include <sys/mac.h> 61#include <sys/malloc.h> 62#include <sys/mount.h> 63#include <sys/proc.h> 64#include <sys/reboot.h> 65#include <sys/resourcevar.h> 66#include <sys/smp.h> /* smp_active */ 67#include <sys/sysctl.h> 68#include <sys/sysproto.h> 69#include <sys/vnode.h> 70 71#include <machine/pcb.h> 72#include <machine/md_var.h> 73#include <machine/smp.h> 74 75#include <sys/signalvar.h> 76#ifdef DDB 77#include <ddb/ddb.h> 78#endif 79 80#ifndef PANIC_REBOOT_WAIT_TIME 81#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 82#endif 83 84/* 85 * Note that stdarg.h and the ANSI style va_start macro is used for both 86 * ANSI and traditional C compilers. 87 */ 88#include <machine/stdarg.h> 89 90#ifdef DDB 91#ifdef DDB_UNATTENDED 92int debugger_on_panic = 0; 93#else 94int debugger_on_panic = 1; 95#endif 96SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 97 &debugger_on_panic, 0, "Run debugger on kernel panic"); 98 99#ifdef DDB_TRACE 100int trace_on_panic = 1; 101#else 102int trace_on_panic = 0; 103#endif 104SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 105 &trace_on_panic, 0, "Print stack trace on kernel panic"); 106#endif 107 108int sync_on_panic = 1; 109SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 110 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 111 112SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 113 114#ifdef HW_WDOG 115/* 116 * If there is a hardware watchdog, point this at the function needed to 117 * hold it off. 118 * It's needed when the kernel needs to do some lengthy operations. 119 * e.g. in wd.c when dumping core.. It's most annoying to have 120 * your precious core-dump only half written because the wdog kicked in. 121 */ 122watchdog_tickle_fn wdog_tickler = NULL; 123#endif /* HW_WDOG */ 124 125/* 126 * Variable panicstr contains argument to first call to panic; used as flag 127 * to indicate that the kernel has already called panic. 128 */ 129const char *panicstr; 130 131int dumping; /* system is dumping */ 132static struct dumperinfo dumper; /* our selected dumper */ 133static struct pcb dumppcb; /* "You Are Here" sign for dump-debuggers */ 134 135static void boot(int) __dead2; 136static void poweroff_wait(void *, int); 137static void shutdown_halt(void *junk, int howto); 138static void shutdown_panic(void *junk, int howto); 139static void shutdown_reset(void *junk, int howto); 140 141/* register various local shutdown events */ 142static void 143shutdown_conf(void *unused) 144{ 145 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 146 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 147 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 148 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 149} 150 151SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 152 153/* 154 * The system call that results in a reboot 155 * 156 * MPSAFE 157 */ 158/* ARGSUSED */ 159int 160reboot(struct thread *td, struct reboot_args *uap) 161{ 162 int error; 163 164 error = 0; 165#ifdef MAC 166 error = mac_check_system_reboot(td->td_ucred, uap->opt); 167#endif 168 if (error == 0) 169 error = suser(td); 170 if (error == 0) { 171 mtx_lock(&Giant); 172 boot(uap->opt); 173 mtx_unlock(&Giant); 174 } 175 return (error); 176} 177 178/* 179 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 180 */ 181static int shutdown_howto = 0; 182 183void 184shutdown_nice(int howto) 185{ 186 shutdown_howto = howto; 187 188 /* Send a signal to init(8) and have it shutdown the world */ 189 if (initproc != NULL) { 190 PROC_LOCK(initproc); 191 psignal(initproc, SIGINT); 192 PROC_UNLOCK(initproc); 193 } else { 194 /* No init(8) running, so simply reboot */ 195 boot(RB_NOSYNC); 196 } 197 return; 198} 199static int waittime = -1; 200 201static void 202print_uptime(void) 203{ 204 int f; 205 struct timespec ts; 206 207 getnanouptime(&ts); 208 printf("Uptime: "); 209 f = 0; 210 if (ts.tv_sec >= 86400) { 211 printf("%ldd", (long)ts.tv_sec / 86400); 212 ts.tv_sec %= 86400; 213 f = 1; 214 } 215 if (f || ts.tv_sec >= 3600) { 216 printf("%ldh", (long)ts.tv_sec / 3600); 217 ts.tv_sec %= 3600; 218 f = 1; 219 } 220 if (f || ts.tv_sec >= 60) { 221 printf("%ldm", (long)ts.tv_sec / 60); 222 ts.tv_sec %= 60; 223 f = 1; 224 } 225 printf("%lds\n", (long)ts.tv_sec); 226} 227 228static void 229doadump(void) 230{ 231 savectx(&dumppcb); 232 dumping++; 233 dumpsys(&dumper); 234} 235 236/* 237 * Go through the rigmarole of shutting down.. 238 * this used to be in machdep.c but I'll be dammned if I could see 239 * anything machine dependant in it. 240 */ 241static void 242boot(int howto) 243{ 244 245 /* collect extra flags that shutdown_nice might have set */ 246 howto |= shutdown_howto; 247 248#ifdef DDB 249 /* We are out of the debugger now. */ 250 db_active = 0; 251#endif 252 253#ifdef SMP 254 if (smp_active) 255 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid)); 256#endif 257 /* 258 * Do any callouts that should be done BEFORE syncing the filesystems. 259 */ 260 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 261 262 /* 263 * Now sync filesystems 264 */ 265 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 266 register struct buf *bp; 267 int iter, nbusy, pbusy; 268 int subiter; 269 270 waittime = 0; 271 printf("\nsyncing disks, buffers remaining... "); 272 273 sync(&thread0, NULL); 274 275 /* 276 * With soft updates, some buffers that are 277 * written will be remarked as dirty until other 278 * buffers are written. 279 */ 280 for (iter = pbusy = 0; iter < 20; iter++) { 281 nbusy = 0; 282 for (bp = &buf[nbuf]; --bp >= buf; ) { 283 if ((bp->b_flags & B_INVAL) == 0 && 284 BUF_REFCNT(bp) > 0) { 285 nbusy++; 286 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 287 == B_DELWRI) { 288 /* bawrite(bp);*/ 289 nbusy++; 290 } 291 } 292 if (nbusy == 0) 293 break; 294 printf("%d ", nbusy); 295 if (nbusy < pbusy) 296 iter = 0; 297 pbusy = nbusy; 298 sync(&thread0, NULL); 299 if (curthread != NULL) { 300 DROP_GIANT(); 301 for (subiter = 0; subiter < 50 * iter; subiter++) { 302 mtx_lock_spin(&sched_lock); 303 curthread->td_proc->p_stats->p_ru.ru_nvcsw++; 304 mi_switch(); /* Allow interrupt threads to run */ 305 mtx_unlock_spin(&sched_lock); 306 DELAY(1000); 307 } 308 PICKUP_GIANT(); 309 } else 310 DELAY(50000 * iter); 311 } 312 printf("\n"); 313 /* 314 * Count only busy local buffers to prevent forcing 315 * a fsck if we're just a client of a wedged NFS server 316 */ 317 nbusy = 0; 318 for (bp = &buf[nbuf]; --bp >= buf; ) { 319 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 320 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 321 if (bp->b_dev == NODEV) { 322 TAILQ_REMOVE(&mountlist, 323 bp->b_vp->v_mount, mnt_list); 324 continue; 325 } 326 nbusy++; 327#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 328 printf( 329 "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n", 330 nbusy, devtoname(bp->b_dev), 331 bp->b_flags, (long)bp->b_blkno, 332 (long)bp->b_lblkno); 333#endif 334 } 335 } 336 if (nbusy) { 337 /* 338 * Failed to sync all blocks. Indicate this and don't 339 * unmount filesystems (thus forcing an fsck on reboot). 340 */ 341 printf("giving up on %d buffers\n", nbusy); 342 DELAY(5000000); /* 5 seconds */ 343 } else { 344 printf("done\n"); 345 /* 346 * Unmount filesystems 347 */ 348 if (panicstr == 0) 349 vfs_unmountall(); 350 } 351 DELAY(100000); /* wait for console output to finish */ 352 } 353 354 print_uptime(); 355 356 /* 357 * Ok, now do things that assume all filesystem activity has 358 * been completed. 359 */ 360 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 361 splhigh(); 362 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && 363 !cold && dumper.dumper != NULL && !dumping) 364 doadump(); 365 366 /* Now that we're going to really halt the system... */ 367 EVENTHANDLER_INVOKE(shutdown_final, howto); 368 369 for(;;) ; /* safety against shutdown_reset not working */ 370 /* NOTREACHED */ 371} 372 373/* 374 * If the shutdown was a clean halt, behave accordingly. 375 */ 376static void 377shutdown_halt(void *junk, int howto) 378{ 379 if (howto & RB_HALT) { 380 printf("\n"); 381 printf("The operating system has halted.\n"); 382 printf("Please press any key to reboot.\n\n"); 383 switch (cngetc()) { 384 case -1: /* No console, just die */ 385 cpu_halt(); 386 /* NOTREACHED */ 387 default: 388 howto &= ~RB_HALT; 389 break; 390 } 391 } 392} 393 394/* 395 * Check to see if the system paniced, pause and then reboot 396 * according to the specified delay. 397 */ 398static void 399shutdown_panic(void *junk, int howto) 400{ 401 int loop; 402 403 if (howto & RB_DUMP) { 404 if (PANIC_REBOOT_WAIT_TIME != 0) { 405 if (PANIC_REBOOT_WAIT_TIME != -1) { 406 printf("Automatic reboot in %d seconds - " 407 "press a key on the console to abort\n", 408 PANIC_REBOOT_WAIT_TIME); 409 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 410 loop > 0; --loop) { 411 DELAY(1000 * 100); /* 1/10th second */ 412 /* Did user type a key? */ 413 if (cncheckc() != -1) 414 break; 415 } 416 if (!loop) 417 return; 418 } 419 } else { /* zero time specified - reboot NOW */ 420 return; 421 } 422 printf("--> Press a key on the console to reboot,\n"); 423 printf("--> or switch off the system now.\n"); 424 cngetc(); 425 } 426} 427 428/* 429 * Everything done, now reset 430 */ 431static void 432shutdown_reset(void *junk, int howto) 433{ 434 printf("Rebooting...\n"); 435 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 436 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 437 cpu_reset(); 438 /* NOTREACHED */ /* assuming reset worked */ 439} 440 441/* 442 * Print a backtrace if we can. 443 */ 444 445void 446backtrace(void) 447{ 448#ifdef DDB 449 db_print_backtrace(); 450#else 451 printf("Sorry, need DDB option to print backtrace"); 452#endif 453} 454 455#ifdef SMP 456static u_int panic_cpu = NOCPU; 457#endif 458 459/* 460 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 461 * and then reboots. If we are called twice, then we avoid trying to sync 462 * the disks as this often leads to recursive panics. 463 * 464 * MPSAFE 465 */ 466void 467panic(const char *fmt, ...) 468{ 469 struct thread *td = curthread; 470 int bootopt, newpanic; 471 va_list ap; 472 static char buf[256]; 473 474#ifdef SMP 475 /* 476 * We don't want multiple CPU's to panic at the same time, so we 477 * use panic_cpu as a simple spinlock. We have to keep checking 478 * panic_cpu if we are spinning in case the panic on the first 479 * CPU is canceled. 480 */ 481 if (panic_cpu != PCPU_GET(cpuid)) 482 while (atomic_cmpset_int(&panic_cpu, NOCPU, 483 PCPU_GET(cpuid)) == 0) 484 while (panic_cpu != NOCPU) 485 ; /* nothing */ 486#endif 487 488 bootopt = RB_AUTOBOOT | RB_DUMP; 489 newpanic = 0; 490 if (panicstr) 491 bootopt |= RB_NOSYNC; 492 else { 493 panicstr = fmt; 494 newpanic = 1; 495 } 496 497 va_start(ap, fmt); 498 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 499 if (panicstr == fmt) 500 panicstr = buf; 501 va_end(ap); 502 printf("panic: %s\n", buf); 503#ifdef SMP 504 /* two separate prints in case of an unmapped page and trap */ 505 printf("cpuid = %d; ", PCPU_GET(cpuid)); 506#ifdef APIC_IO 507 printf("lapic.id = %08x\n", lapic.id); 508#else 509 printf("\n"); 510#endif 511#endif 512 513#if defined(DDB) 514 if (newpanic && trace_on_panic) 515 db_print_backtrace(); 516 if (debugger_on_panic) 517 Debugger ("panic"); 518#ifdef RESTARTABLE_PANICS 519 /* See if the user aborted the panic, in which case we continue. */ 520 if (panicstr == NULL) { 521#ifdef SMP 522 atomic_store_rel_int(&panic_cpu, NOCPU); 523#endif 524 return; 525 } 526#endif 527#endif 528 td->td_flags |= TDF_INPANIC; 529 if (!sync_on_panic) 530 bootopt |= RB_NOSYNC; 531 boot(bootopt); 532} 533 534/* 535 * Support for poweroff delay. 536 */ 537#ifndef POWEROFF_DELAY 538# define POWEROFF_DELAY 5000 539#endif 540static int poweroff_delay = POWEROFF_DELAY; 541 542SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 543 &poweroff_delay, 0, ""); 544 545static void 546poweroff_wait(void *junk, int howto) 547{ 548 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 549 return; 550 DELAY(poweroff_delay * 1000); 551} 552 553/* 554 * Some system processes (e.g. syncer) need to be stopped at appropriate 555 * points in their main loops prior to a system shutdown, so that they 556 * won't interfere with the shutdown process (e.g. by holding a disk buf 557 * to cause sync to fail). For each of these system processes, register 558 * shutdown_kproc() as a handler for one of shutdown events. 559 */ 560static int kproc_shutdown_wait = 60; 561SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 562 &kproc_shutdown_wait, 0, ""); 563 564void 565kproc_shutdown(void *arg, int howto) 566{ 567 struct proc *p; 568 int error; 569 570 if (panicstr) 571 return; 572 573 p = (struct proc *)arg; 574 printf("Waiting (max %d seconds) for system process `%s' to stop...", 575 kproc_shutdown_wait, p->p_comm); 576 error = kthread_suspend(p, kproc_shutdown_wait * hz); 577 578 if (error == EWOULDBLOCK) 579 printf("timed out\n"); 580 else 581 printf("stopped\n"); 582} 583 584/* Registration of dumpers */ 585int 586set_dumper(struct dumperinfo *di) 587{ 588 if (di == NULL) { 589 bzero(&dumper, sizeof dumper); 590 return (0); 591 } 592 if (dumper.dumper != NULL) 593 return (EBUSY); 594 dumper = *di; 595 return (0); 596} 597 598#if defined(__powerpc__) 599void 600dumpsys(struct dumperinfo *di __unused) 601{ 602 603 printf("Kernel dumps not implemented on this architecture\n"); 604} 605#endif 606