kern_shutdown.c revision 132866
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 132866 2004-07-30 01:30:05Z 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 static int first_buf_printf = 1; 259 260 /* collect extra flags that shutdown_nice might have set */ 261 howto |= shutdown_howto; 262 263 /* We are out of the debugger now. */ 264 kdb_active = 0; 265 266#ifdef SMP 267 if (smp_active) 268 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid)); 269#endif 270 /* 271 * Do any callouts that should be done BEFORE syncing the filesystems. 272 */ 273 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 274 275 /* 276 * Now sync filesystems 277 */ 278 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 279 register struct buf *bp; 280 int iter, nbusy, pbusy; 281#ifndef PREEMPTION 282 int subiter; 283#endif 284 285 waittime = 0; 286 287 sync(&thread0, NULL); 288 289 /* 290 * With soft updates, some buffers that are 291 * written will be remarked as dirty until other 292 * buffers are written. 293 */ 294 for (iter = pbusy = 0; iter < 20; iter++) { 295 nbusy = 0; 296 for (bp = &buf[nbuf]; --bp >= buf; ) { 297 if ((bp->b_flags & B_INVAL) == 0 && 298 BUF_REFCNT(bp) > 0) { 299 nbusy++; 300 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 301 == B_DELWRI) { 302 /* bawrite(bp);*/ 303 nbusy++; 304 } 305 } 306 if (nbusy == 0) 307 break; 308 if (first_buf_printf) { 309 printf("syncing disks, buffers remaining... "); 310 first_buf_printf = 0; 311 } 312 printf("%d ", nbusy); 313 if (nbusy < pbusy) 314 iter = 0; 315 pbusy = nbusy; 316 sync(&thread0, NULL); 317 318#ifdef PREEMPTION 319 /* 320 * Drop Giant and spin for a while to allow 321 * interrupt threads to run. 322 */ 323 DROP_GIANT(); 324 DELAY(50000 * iter); 325 PICKUP_GIANT(); 326#else 327 /* 328 * Drop Giant and context switch several times to 329 * allow interrupt threads to run. 330 */ 331 DROP_GIANT(); 332 for (subiter = 0; subiter < 50 * iter; subiter++) { 333 mtx_lock_spin(&sched_lock); 334 mi_switch(SW_VOL, NULL); 335 mtx_unlock_spin(&sched_lock); 336 DELAY(1000); 337 } 338 PICKUP_GIANT(); 339#endif 340 } 341 342 /* 343 * Count only busy local buffers to prevent forcing 344 * a fsck if we're just a client of a wedged NFS server 345 */ 346 nbusy = 0; 347 for (bp = &buf[nbuf]; --bp >= buf; ) { 348 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 349 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 350 if (bp->b_dev == NULL) { 351 TAILQ_REMOVE(&mountlist, 352 bp->b_vp->v_mount, mnt_list); 353 continue; 354 } 355 nbusy++; 356#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 357 printf( 358 "%d: dev:%s, flags:%0x, blkno:%ld, lblkno:%ld\n", 359 nbusy, devtoname(bp->b_dev), 360 bp->b_flags, (long)bp->b_blkno, 361 (long)bp->b_lblkno); 362#endif 363 } 364 } 365 if (nbusy) { 366 /* 367 * Failed to sync all blocks. Indicate this and don't 368 * unmount filesystems (thus forcing an fsck on reboot). 369 */ 370 printf("giving up on %d buffers\n", nbusy); 371 DELAY(5000000); /* 5 seconds */ 372 } else { 373 /* 374 * Unmount filesystems 375 */ 376 if (panicstr == 0) 377 vfs_unmountall(); 378 } 379 DELAY(100000); /* wait for console output to finish */ 380 } 381 382 print_uptime(); 383 384 /* 385 * Ok, now do things that assume all filesystem activity has 386 * been completed. 387 */ 388 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 389 splhigh(); 390 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 391 doadump(); 392 393 /* Now that we're going to really halt the system... */ 394 EVENTHANDLER_INVOKE(shutdown_final, howto); 395 396 for(;;) ; /* safety against shutdown_reset not working */ 397 /* NOTREACHED */ 398} 399 400/* 401 * If the shutdown was a clean halt, behave accordingly. 402 */ 403static void 404shutdown_halt(void *junk, int howto) 405{ 406 407 if (howto & RB_HALT) { 408 printf("\n"); 409 printf("The operating system has halted.\n"); 410 printf("Please press any key to reboot.\n\n"); 411 switch (cngetc()) { 412 case -1: /* No console, just die */ 413 cpu_halt(); 414 /* NOTREACHED */ 415 default: 416 howto &= ~RB_HALT; 417 break; 418 } 419 } 420} 421 422/* 423 * Check to see if the system paniced, pause and then reboot 424 * according to the specified delay. 425 */ 426static void 427shutdown_panic(void *junk, int howto) 428{ 429 int loop; 430 431 if (howto & RB_DUMP) { 432 if (PANIC_REBOOT_WAIT_TIME != 0) { 433 if (PANIC_REBOOT_WAIT_TIME != -1) { 434 printf("Automatic reboot in %d seconds - " 435 "press a key on the console to abort\n", 436 PANIC_REBOOT_WAIT_TIME); 437 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 438 loop > 0; --loop) { 439 DELAY(1000 * 100); /* 1/10th second */ 440 /* Did user type a key? */ 441 if (cncheckc() != -1) 442 break; 443 } 444 if (!loop) 445 return; 446 } 447 } else { /* zero time specified - reboot NOW */ 448 return; 449 } 450 printf("--> Press a key on the console to reboot,\n"); 451 printf("--> or switch off the system now.\n"); 452 cngetc(); 453 } 454} 455 456/* 457 * Everything done, now reset 458 */ 459static void 460shutdown_reset(void *junk, int howto) 461{ 462 463 printf("Rebooting...\n"); 464 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 465 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 466 cpu_reset(); 467 /* NOTREACHED */ /* assuming reset worked */ 468} 469 470#ifdef SMP 471static u_int panic_cpu = NOCPU; 472#endif 473 474/* 475 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 476 * and then reboots. If we are called twice, then we avoid trying to sync 477 * the disks as this often leads to recursive panics. 478 * 479 * MPSAFE 480 */ 481void 482panic(const char *fmt, ...) 483{ 484 struct thread *td = curthread; 485 int bootopt, newpanic; 486 va_list ap; 487 static char buf[256]; 488 489#ifdef SMP 490 /* 491 * We don't want multiple CPU's to panic at the same time, so we 492 * use panic_cpu as a simple spinlock. We have to keep checking 493 * panic_cpu if we are spinning in case the panic on the first 494 * CPU is canceled. 495 */ 496 if (panic_cpu != PCPU_GET(cpuid)) 497 while (atomic_cmpset_int(&panic_cpu, NOCPU, 498 PCPU_GET(cpuid)) == 0) 499 while (panic_cpu != NOCPU) 500 ; /* nothing */ 501#endif 502 503 bootopt = RB_AUTOBOOT | RB_DUMP; 504 newpanic = 0; 505 if (panicstr) 506 bootopt |= RB_NOSYNC; 507 else { 508 panicstr = fmt; 509 newpanic = 1; 510 } 511 512 va_start(ap, fmt); 513 if (newpanic) { 514 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 515 panicstr = buf; 516 printf("panic: %s\n", buf); 517 } else { 518 printf("panic: "); 519 vprintf(fmt, ap); 520 printf("\n"); 521 } 522 va_end(ap); 523#ifdef SMP 524 /* two separate prints in case of an unmapped page and trap */ 525 printf("cpuid = %d; ", PCPU_GET(cpuid)); 526#ifdef APIC_IO 527 printf("lapic.id = %08x\n", lapic.id); 528#else 529 printf("\n"); 530#endif 531#endif 532 533#ifdef KDB 534 if (newpanic && trace_on_panic) 535 kdb_backtrace(); 536 if (debugger_on_panic) 537 kdb_enter("panic"); 538#ifdef RESTARTABLE_PANICS 539 /* See if the user aborted the panic, in which case we continue. */ 540 if (panicstr == NULL) { 541#ifdef SMP 542 atomic_store_rel_int(&panic_cpu, NOCPU); 543#endif 544 return; 545 } 546#endif 547#endif 548 mtx_lock_spin(&sched_lock); 549 td->td_flags |= TDF_INPANIC; 550 mtx_unlock_spin(&sched_lock); 551 if (!sync_on_panic) 552 bootopt |= RB_NOSYNC; 553 boot(bootopt); 554} 555 556/* 557 * Support for poweroff delay. 558 */ 559#ifndef POWEROFF_DELAY 560# define POWEROFF_DELAY 5000 561#endif 562static int poweroff_delay = POWEROFF_DELAY; 563 564SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 565 &poweroff_delay, 0, ""); 566 567static void 568poweroff_wait(void *junk, int howto) 569{ 570 571 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 572 return; 573 DELAY(poweroff_delay * 1000); 574} 575 576/* 577 * Some system processes (e.g. syncer) need to be stopped at appropriate 578 * points in their main loops prior to a system shutdown, so that they 579 * won't interfere with the shutdown process (e.g. by holding a disk buf 580 * to cause sync to fail). For each of these system processes, register 581 * shutdown_kproc() as a handler for one of shutdown events. 582 */ 583static int kproc_shutdown_wait = 60; 584SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 585 &kproc_shutdown_wait, 0, ""); 586 587void 588kproc_shutdown(void *arg, int howto) 589{ 590 struct proc *p; 591 char procname[MAXCOMLEN + 1]; 592 int error; 593 594 if (panicstr) 595 return; 596 597 p = (struct proc *)arg; 598 strlcpy(procname, p->p_comm, sizeof(procname)); 599 printf("Waiting (max %d seconds) for system process `%s' to stop...", 600 kproc_shutdown_wait, procname); 601 error = kthread_suspend(p, kproc_shutdown_wait * hz); 602 603 if (error == EWOULDBLOCK) 604 printf("timed out\n"); 605 else 606 printf("done\n"); 607} 608 609/* Registration of dumpers */ 610int 611set_dumper(struct dumperinfo *di) 612{ 613 614 if (di == NULL) { 615 bzero(&dumper, sizeof dumper); 616 return (0); 617 } 618 if (dumper.dumper != NULL) 619 return (EBUSY); 620 dumper = *di; 621 return (0); 622} 623 624#if defined(__powerpc__) 625void 626dumpsys(struct dumperinfo *di __unused) 627{ 628 629 printf("Kernel dumps not implemented on this architecture\n"); 630} 631#endif 632