kern_shutdown.c revision 133418
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: head/sys/kern/kern_shutdown.c 133418 2004-08-10 01:32:05Z njl $"); 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 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/kdb.h> 54#include <sys/kernel.h> 55#include <sys/kthread.h> 56#include <sys/mac.h> 57#include <sys/malloc.h> 58#include <sys/mount.h> 59#include <sys/proc.h> 60#include <sys/reboot.h> 61#include <sys/resourcevar.h> 62#include <sys/smp.h> /* smp_active */ 63#include <sys/sysctl.h> 64#include <sys/sysproto.h> 65#include <sys/vnode.h> 66 67#include <machine/cpu.h> 68#include <machine/pcb.h> 69#include <machine/smp.h> 70 71#include <sys/signalvar.h> 72 73#ifndef PANIC_REBOOT_WAIT_TIME 74#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 75#endif 76 77/* 78 * Note that stdarg.h and the ANSI style va_start macro is used for both 79 * ANSI and traditional C compilers. 80 */ 81#include <machine/stdarg.h> 82 83#ifdef KDB 84#ifdef KDB_UNATTENDED 85int debugger_on_panic = 0; 86#else 87int debugger_on_panic = 1; 88#endif 89SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 90 &debugger_on_panic, 0, "Run debugger on kernel panic"); 91 92#ifdef KDB_TRACE 93int trace_on_panic = 1; 94#else 95int trace_on_panic = 0; 96#endif 97SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 98 &trace_on_panic, 0, "Print stack trace on kernel panic"); 99#endif /* KDB */ 100 101int sync_on_panic = 0; 102SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 103 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 104 105SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 106 107#ifdef HW_WDOG 108/* 109 * If there is a hardware watchdog, point this at the function needed to 110 * hold it off. 111 * It's needed when the kernel needs to do some lengthy operations. 112 * e.g. in wd.c when dumping core.. It's most annoying to have 113 * your precious core-dump only half written because the wdog kicked in. 114 */ 115watchdog_tickle_fn wdog_tickler = NULL; 116#endif /* HW_WDOG */ 117 118/* 119 * Variable panicstr contains argument to first call to panic; used as flag 120 * to indicate that the kernel has already called panic. 121 */ 122const char *panicstr; 123 124int dumping; /* system is dumping */ 125static struct dumperinfo dumper; /* our selected dumper */ 126 127/* Context information for dump-debuggers. */ 128static struct pcb dumppcb; /* Registers. */ 129static lwpid_t dumptid; /* Thread ID. */ 130 131static void boot(int) __dead2; 132static void poweroff_wait(void *, int); 133static void shutdown_halt(void *junk, int howto); 134static void shutdown_panic(void *junk, int howto); 135static void shutdown_reset(void *junk, int howto); 136 137/* register various local shutdown events */ 138static void 139shutdown_conf(void *unused) 140{ 141 142 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 143 SHUTDOWN_PRI_FIRST); 144 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 145 SHUTDOWN_PRI_LAST + 100); 146 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 147 SHUTDOWN_PRI_LAST + 100); 148 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 149 SHUTDOWN_PRI_LAST + 200); 150} 151 152SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 153 154/* 155 * The system call that results in a reboot 156 * 157 * MPSAFE 158 */ 159/* ARGSUSED */ 160int 161reboot(struct thread *td, struct reboot_args *uap) 162{ 163 int error; 164 165 error = 0; 166#ifdef MAC 167 error = mac_check_system_reboot(td->td_ucred, uap->opt); 168#endif 169 if (error == 0) 170 error = suser(td); 171 if (error == 0) { 172 mtx_lock(&Giant); 173 boot(uap->opt); 174 mtx_unlock(&Giant); 175 } 176 return (error); 177} 178 179/* 180 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 181 */ 182static int shutdown_howto = 0; 183 184void 185shutdown_nice(int howto) 186{ 187 188 shutdown_howto = howto; 189 190 /* Send a signal to init(8) and have it shutdown the world */ 191 if (initproc != NULL) { 192 PROC_LOCK(initproc); 193 psignal(initproc, SIGINT); 194 PROC_UNLOCK(initproc); 195 } else { 196 /* No init(8) running, so simply reboot */ 197 boot(RB_NOSYNC); 198 } 199 return; 200} 201static int waittime = -1; 202 203static void 204print_uptime(void) 205{ 206 int f; 207 struct timespec ts; 208 209 getnanouptime(&ts); 210 printf("Uptime: "); 211 f = 0; 212 if (ts.tv_sec >= 86400) { 213 printf("%ldd", (long)ts.tv_sec / 86400); 214 ts.tv_sec %= 86400; 215 f = 1; 216 } 217 if (f || ts.tv_sec >= 3600) { 218 printf("%ldh", (long)ts.tv_sec / 3600); 219 ts.tv_sec %= 3600; 220 f = 1; 221 } 222 if (f || ts.tv_sec >= 60) { 223 printf("%ldm", (long)ts.tv_sec / 60); 224 ts.tv_sec %= 60; 225 f = 1; 226 } 227 printf("%lds\n", (long)ts.tv_sec); 228} 229 230static void 231doadump(void) 232{ 233 234 /* 235 * Sometimes people have to call this from the kernel debugger. 236 * (if 'panic' can not dump) 237 * Give them a clue as to why they can't dump. 238 */ 239 if (dumper.dumper == NULL) { 240 printf("Cannot dump. No dump device defined.\n"); 241 return; 242 } 243 244 savectx(&dumppcb); 245 dumptid = curthread->td_tid; 246 dumping++; 247 dumpsys(&dumper); 248} 249 250/* 251 * Go through the rigmarole of shutting down.. 252 * this used to be in machdep.c but I'll be dammned if I could see 253 * anything machine dependant in it. 254 */ 255static void 256boot(int howto) 257{ 258 259 /* collect extra flags that shutdown_nice might have set */ 260 howto |= shutdown_howto; 261 262 /* We are out of the debugger now. */ 263 kdb_active = 0; 264 265#ifdef SMP 266 if (smp_active) 267 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid)); 268#endif 269 /* 270 * Do any callouts that should be done BEFORE syncing the filesystems. 271 */ 272 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 273 274 /* 275 * Now sync filesystems 276 */ 277 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 278 register struct buf *bp; 279 int iter, nbusy, pbusy; 280#ifndef PREEMPTION 281 int subiter; 282#endif 283 284 for (nbusy = 0, bp = &buf[nbuf]; --bp >= buf; ) 285 if (((bp->b_flags & B_INVAL) == 0 && 286 BUF_REFCNT(bp) > 0) || 287 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) 288 nbusy++; 289 if (nbusy == 0) { 290 printf("Skipping final sync, no buffers remaining\n"); 291 goto unmountall; 292 } 293 294 waittime = 0; 295 printf("Syncing disks, buffers remaining... "); 296 297 sync(&thread0, NULL); 298 299 /* 300 * With soft updates, some buffers that are 301 * written will be remarked as dirty until other 302 * buffers are written. 303 */ 304 for (iter = pbusy = 0; iter < 20; iter++) { 305 nbusy = 0; 306 for (bp = &buf[nbuf]; --bp >= buf; ) { 307 if ((bp->b_flags & B_INVAL) == 0 && 308 BUF_REFCNT(bp) > 0) { 309 nbusy++; 310 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 311 == B_DELWRI) { 312 /* bawrite(bp);*/ 313 nbusy++; 314 } 315 } 316 if (nbusy == 0) 317 break; 318 printf("%d ", nbusy); 319 if (nbusy < pbusy) 320 iter = 0; 321 pbusy = nbusy; 322 sync(&thread0, NULL); 323 324#ifdef PREEMPTION 325 /* 326 * Drop Giant and spin for a while to allow 327 * interrupt threads to run. 328 */ 329 DROP_GIANT(); 330 DELAY(50000 * iter); 331 PICKUP_GIANT(); 332#else 333 /* 334 * Drop Giant and context switch several times to 335 * allow interrupt threads to run. 336 */ 337 DROP_GIANT(); 338 for (subiter = 0; subiter < 50 * iter; subiter++) { 339 mtx_lock_spin(&sched_lock); 340 mi_switch(SW_VOL, NULL); 341 mtx_unlock_spin(&sched_lock); 342 DELAY(1000); 343 } 344 PICKUP_GIANT(); 345#endif 346 } 347 printf("\n"); 348 349 /* 350 * Count only busy local buffers to prevent forcing 351 * a fsck if we're just a client of a wedged NFS server 352 */ 353 nbusy = 0; 354 for (bp = &buf[nbuf]; --bp >= buf; ) { 355 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 356 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 357 if (bp->b_dev == NULL) { 358 TAILQ_REMOVE(&mountlist, 359 bp->b_vp->v_mount, mnt_list); 360 continue; 361 } 362 nbusy++; 363#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 364 printf( 365 "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n", 366 nbusy, devtoname(bp->b_dev), 367 bp->b_flags, (long)bp->b_blkno, 368 (long)bp->b_lblkno); 369#endif 370 } 371 } 372 if (nbusy) { 373 /* 374 * Failed to sync all blocks. Indicate this and don't 375 * unmount filesystems (thus forcing an fsck on reboot). 376 */ 377 printf("giving up on %d buffers\n", nbusy); 378 DELAY(5000000); /* 5 seconds */ 379 } else { 380 /* 381 * Unmount filesystems 382 */ 383unmountall: 384 if (panicstr == 0) 385 vfs_unmountall(); 386 } 387 DELAY(100000); /* wait for console output to finish */ 388 } 389 390 print_uptime(); 391 392 /* 393 * Ok, now do things that assume all filesystem activity has 394 * been completed. 395 */ 396 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 397 splhigh(); 398 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 399 doadump(); 400 401 /* Now that we're going to really halt the system... */ 402 EVENTHANDLER_INVOKE(shutdown_final, howto); 403 404 for(;;) ; /* safety against shutdown_reset not working */ 405 /* NOTREACHED */ 406} 407 408/* 409 * If the shutdown was a clean halt, behave accordingly. 410 */ 411static void 412shutdown_halt(void *junk, int howto) 413{ 414 415 if (howto & RB_HALT) { 416 printf("\n"); 417 printf("The operating system has halted.\n"); 418 printf("Please press any key to reboot.\n\n"); 419 switch (cngetc()) { 420 case -1: /* No console, just die */ 421 cpu_halt(); 422 /* NOTREACHED */ 423 default: 424 howto &= ~RB_HALT; 425 break; 426 } 427 } 428} 429 430/* 431 * Check to see if the system paniced, pause and then reboot 432 * according to the specified delay. 433 */ 434static void 435shutdown_panic(void *junk, int howto) 436{ 437 int loop; 438 439 if (howto & RB_DUMP) { 440 if (PANIC_REBOOT_WAIT_TIME != 0) { 441 if (PANIC_REBOOT_WAIT_TIME != -1) { 442 printf("Automatic reboot in %d seconds - " 443 "press a key on the console to abort\n", 444 PANIC_REBOOT_WAIT_TIME); 445 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 446 loop > 0; --loop) { 447 DELAY(1000 * 100); /* 1/10th second */ 448 /* Did user type a key? */ 449 if (cncheckc() != -1) 450 break; 451 } 452 if (!loop) 453 return; 454 } 455 } else { /* zero time specified - reboot NOW */ 456 return; 457 } 458 printf("--> Press a key on the console to reboot,\n"); 459 printf("--> or switch off the system now.\n"); 460 cngetc(); 461 } 462} 463 464/* 465 * Everything done, now reset 466 */ 467static void 468shutdown_reset(void *junk, int howto) 469{ 470 471 printf("Rebooting...\n"); 472 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 473 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 474 cpu_reset(); 475 /* NOTREACHED */ /* assuming reset worked */ 476} 477 478#ifdef SMP 479static u_int panic_cpu = NOCPU; 480#endif 481 482/* 483 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 484 * and then reboots. If we are called twice, then we avoid trying to sync 485 * the disks as this often leads to recursive panics. 486 * 487 * MPSAFE 488 */ 489void 490panic(const char *fmt, ...) 491{ 492 struct thread *td = curthread; 493 int bootopt, newpanic; 494 va_list ap; 495 static char buf[256]; 496 497#ifdef SMP 498 /* 499 * We don't want multiple CPU's to panic at the same time, so we 500 * use panic_cpu as a simple spinlock. We have to keep checking 501 * panic_cpu if we are spinning in case the panic on the first 502 * CPU is canceled. 503 */ 504 if (panic_cpu != PCPU_GET(cpuid)) 505 while (atomic_cmpset_int(&panic_cpu, NOCPU, 506 PCPU_GET(cpuid)) == 0) 507 while (panic_cpu != NOCPU) 508 ; /* nothing */ 509#endif 510 511 bootopt = RB_AUTOBOOT | RB_DUMP; 512 newpanic = 0; 513 if (panicstr) 514 bootopt |= RB_NOSYNC; 515 else { 516 panicstr = fmt; 517 newpanic = 1; 518 } 519 520 va_start(ap, fmt); 521 if (newpanic) { 522 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 523 panicstr = buf; 524 printf("panic: %s\n", buf); 525 } else { 526 printf("panic: "); 527 vprintf(fmt, ap); 528 printf("\n"); 529 } 530 va_end(ap); 531#ifdef SMP 532 /* two separate prints in case of an unmapped page and trap */ 533 printf("cpuid = %d; ", PCPU_GET(cpuid)); 534#ifdef APIC_IO 535 printf("lapic.id = %08x\n", lapic.id); 536#else 537 printf("\n"); 538#endif 539#endif 540 541#ifdef KDB 542 if (newpanic && trace_on_panic) 543 kdb_backtrace(); 544 if (debugger_on_panic) 545 kdb_enter("panic"); 546#ifdef RESTARTABLE_PANICS 547 /* See if the user aborted the panic, in which case we continue. */ 548 if (panicstr == NULL) { 549#ifdef SMP 550 atomic_store_rel_int(&panic_cpu, NOCPU); 551#endif 552 return; 553 } 554#endif 555#endif 556 mtx_lock_spin(&sched_lock); 557 td->td_flags |= TDF_INPANIC; 558 mtx_unlock_spin(&sched_lock); 559 if (!sync_on_panic) 560 bootopt |= RB_NOSYNC; 561 boot(bootopt); 562} 563 564/* 565 * Support for poweroff delay. 566 */ 567#ifndef POWEROFF_DELAY 568# define POWEROFF_DELAY 5000 569#endif 570static int poweroff_delay = POWEROFF_DELAY; 571 572SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 573 &poweroff_delay, 0, ""); 574 575static void 576poweroff_wait(void *junk, int howto) 577{ 578 579 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 580 return; 581 DELAY(poweroff_delay * 1000); 582} 583 584/* 585 * Some system processes (e.g. syncer) need to be stopped at appropriate 586 * points in their main loops prior to a system shutdown, so that they 587 * won't interfere with the shutdown process (e.g. by holding a disk buf 588 * to cause sync to fail). For each of these system processes, register 589 * shutdown_kproc() as a handler for one of shutdown events. 590 */ 591static int kproc_shutdown_wait = 60; 592SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 593 &kproc_shutdown_wait, 0, ""); 594 595void 596kproc_shutdown(void *arg, int howto) 597{ 598 struct proc *p; 599 char procname[MAXCOMLEN + 1]; 600 int error; 601 602 if (panicstr) 603 return; 604 605 p = (struct proc *)arg; 606 strlcpy(procname, p->p_comm, sizeof(procname)); 607 printf("Waiting (max %d seconds) for system process `%s' to stop...", 608 kproc_shutdown_wait, procname); 609 error = kthread_suspend(p, kproc_shutdown_wait * hz); 610 611 if (error == EWOULDBLOCK) 612 printf("timed out\n"); 613 else 614 printf("done\n"); 615} 616 617/* Registration of dumpers */ 618int 619set_dumper(struct dumperinfo *di) 620{ 621 622 if (di == NULL) { 623 bzero(&dumper, sizeof dumper); 624 return (0); 625 } 626 if (dumper.dumper != NULL) 627 return (EBUSY); 628 dumper = *di; 629 return (0); 630} 631 632#if defined(__powerpc__) 633void 634dumpsys(struct dumperinfo *di __unused) 635{ 636 637 printf("Kernel dumps not implemented on this architecture\n"); 638} 639#endif 640