kern_shutdown.c revision 138217
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 138217 2004-11-30 06:21:38Z njl $"); 39 40#include "opt_kdb.h" 41#include "opt_mac.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/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/sched.h> 63#include <sys/smp.h> /* smp_active */ 64#include <sys/sysctl.h> 65#include <sys/sysproto.h> 66#include <sys/vnode.h> 67 68#include <machine/cpu.h> 69#include <machine/pcb.h> 70#include <machine/smp.h> 71 72#include <sys/signalvar.h> 73 74#ifndef PANIC_REBOOT_WAIT_TIME 75#define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 76#endif 77 78/* 79 * Note that stdarg.h and the ANSI style va_start macro is used for both 80 * ANSI and traditional C compilers. 81 */ 82#include <machine/stdarg.h> 83 84#ifdef KDB 85#ifdef KDB_UNATTENDED 86int debugger_on_panic = 0; 87#else 88int debugger_on_panic = 1; 89#endif 90SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 91 &debugger_on_panic, 0, "Run debugger on kernel panic"); 92 93#ifdef KDB_TRACE 94int trace_on_panic = 1; 95#else 96int trace_on_panic = 0; 97#endif 98SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 99 &trace_on_panic, 0, "Print stack trace on kernel panic"); 100#endif /* KDB */ 101 102int sync_on_panic = 0; 103SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 104 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 105 106SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 107 108/* 109 * Variable panicstr contains argument to first call to panic; used as flag 110 * to indicate that the kernel has already called panic. 111 */ 112const char *panicstr; 113 114int dumping; /* system is dumping */ 115static struct dumperinfo dumper; /* our selected dumper */ 116 117/* Context information for dump-debuggers. */ 118static struct pcb dumppcb; /* Registers. */ 119static lwpid_t dumptid; /* Thread ID. */ 120 121static void boot(int) __dead2; 122static void poweroff_wait(void *, int); 123static void shutdown_halt(void *junk, int howto); 124static void shutdown_panic(void *junk, int howto); 125static void shutdown_reset(void *junk, int howto); 126 127/* register various local shutdown events */ 128static void 129shutdown_conf(void *unused) 130{ 131 132 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 133 SHUTDOWN_PRI_FIRST); 134 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 135 SHUTDOWN_PRI_LAST + 100); 136 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 137 SHUTDOWN_PRI_LAST + 100); 138 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 139 SHUTDOWN_PRI_LAST + 200); 140} 141 142SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 143 144/* 145 * The system call that results in a reboot 146 * 147 * MPSAFE 148 */ 149/* ARGSUSED */ 150int 151reboot(struct thread *td, struct reboot_args *uap) 152{ 153 int error; 154 155 error = 0; 156#ifdef MAC 157 error = mac_check_system_reboot(td->td_ucred, uap->opt); 158#endif 159 if (error == 0) 160 error = suser(td); 161 if (error == 0) { 162 mtx_lock(&Giant); 163 boot(uap->opt); 164 mtx_unlock(&Giant); 165 } 166 return (error); 167} 168 169/* 170 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 171 */ 172static int shutdown_howto = 0; 173 174void 175shutdown_nice(int howto) 176{ 177 178 shutdown_howto = howto; 179 180 /* Send a signal to init(8) and have it shutdown the world */ 181 if (initproc != NULL) { 182 PROC_LOCK(initproc); 183 psignal(initproc, SIGINT); 184 PROC_UNLOCK(initproc); 185 } else { 186 /* No init(8) running, so simply reboot */ 187 boot(RB_NOSYNC); 188 } 189 return; 190} 191static int waittime = -1; 192 193static void 194print_uptime(void) 195{ 196 int f; 197 struct timespec ts; 198 199 getnanouptime(&ts); 200 printf("Uptime: "); 201 f = 0; 202 if (ts.tv_sec >= 86400) { 203 printf("%ldd", (long)ts.tv_sec / 86400); 204 ts.tv_sec %= 86400; 205 f = 1; 206 } 207 if (f || ts.tv_sec >= 3600) { 208 printf("%ldh", (long)ts.tv_sec / 3600); 209 ts.tv_sec %= 3600; 210 f = 1; 211 } 212 if (f || ts.tv_sec >= 60) { 213 printf("%ldm", (long)ts.tv_sec / 60); 214 ts.tv_sec %= 60; 215 f = 1; 216 } 217 printf("%lds\n", (long)ts.tv_sec); 218} 219 220static void 221doadump(void) 222{ 223 224 /* 225 * Sometimes people have to call this from the kernel debugger. 226 * (if 'panic' can not dump) 227 * Give them a clue as to why they can't dump. 228 */ 229 if (dumper.dumper == NULL) { 230 printf("Cannot dump. No dump device defined.\n"); 231 return; 232 } 233 234 savectx(&dumppcb); 235 dumptid = curthread->td_tid; 236 dumping++; 237 dumpsys(&dumper); 238} 239 240/* 241 * Shutdown the system cleanly to prepare for reboot, halt, or power off. 242 */ 243static void 244boot(int howto) 245{ 246 static int first_buf_printf = 1; 247 248#if defined(SMP) 249 /* 250 * Bind us to CPU 0 so that all shutdown code runs there. Some 251 * systems don't shutdown properly (i.e., ACPI power off) if we 252 * run on another processor. 253 */ 254 mtx_lock_spin(&sched_lock); 255 sched_bind(curthread, 0); 256 mtx_unlock_spin(&sched_lock); 257 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0")); 258#endif 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 /* 267 * Do any callouts that should be done BEFORE syncing the filesystems. 268 */ 269 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 270 271 /* 272 * Now sync filesystems 273 */ 274 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 275 register struct buf *bp; 276 int iter, nbusy, pbusy; 277#ifndef PREEMPTION 278 int subiter; 279#endif 280 281 waittime = 0; 282 283 sync(&thread0, NULL); 284 285 /* 286 * With soft updates, some buffers that are 287 * written will be remarked as dirty until other 288 * buffers are written. 289 */ 290 for (iter = pbusy = 0; iter < 20; iter++) { 291 nbusy = 0; 292 for (bp = &buf[nbuf]; --bp >= buf; ) { 293 if ((bp->b_flags & B_INVAL) == 0 && 294 BUF_REFCNT(bp) > 0) { 295 nbusy++; 296 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 297 == B_DELWRI) { 298 /* bawrite(bp);*/ 299 nbusy++; 300 } 301 } 302 if (nbusy == 0) { 303 if (first_buf_printf) 304 printf("All buffers synced."); 305 break; 306 } 307 if (first_buf_printf) { 308 printf("Syncing disks, buffers remaining... "); 309 first_buf_printf = 0; 310 } 311 printf("%d ", nbusy); 312 if (nbusy < pbusy) 313 iter = 0; 314 pbusy = nbusy; 315 sync(&thread0, NULL); 316 317#ifdef PREEMPTION 318 /* 319 * Drop Giant and spin for a while to allow 320 * interrupt threads to run. 321 */ 322 DROP_GIANT(); 323 DELAY(50000 * iter); 324 PICKUP_GIANT(); 325#else 326 /* 327 * Drop Giant and context switch several times to 328 * allow interrupt threads to run. 329 */ 330 DROP_GIANT(); 331 for (subiter = 0; subiter < 50 * iter; subiter++) { 332 mtx_lock_spin(&sched_lock); 333 mi_switch(SW_VOL, NULL); 334 mtx_unlock_spin(&sched_lock); 335 DELAY(1000); 336 } 337 PICKUP_GIANT(); 338#endif 339 } 340 printf("\n"); 341 /* 342 * Count only busy local buffers to prevent forcing 343 * a fsck if we're just a client of a wedged NFS server 344 */ 345 nbusy = 0; 346 for (bp = &buf[nbuf]; --bp >= buf; ) { 347 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 348 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 349#if 0 350/* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 351 if (bp->b_dev == NULL) { 352 TAILQ_REMOVE(&mountlist, 353 bp->b_vp->v_mount, mnt_list); 354 continue; 355 } 356#endif 357 nbusy++; 358#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 359 printf( 360 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n", 361 nbusy, bp->b_bufobj, 362 bp->b_flags, (long)bp->b_blkno, 363 (long)bp->b_lblkno); 364#endif 365 } 366 } 367 if (nbusy) { 368 /* 369 * Failed to sync all blocks. Indicate this and don't 370 * unmount filesystems (thus forcing an fsck on reboot). 371 */ 372 printf("Giving up on %d buffers\n", nbusy); 373 DELAY(5000000); /* 5 seconds */ 374 } else { 375 if (!first_buf_printf) 376 printf("Final sync complete\n"); 377 /* 378 * Unmount filesystems 379 */ 380 if (panicstr == 0) 381 vfs_unmountall(); 382 } 383 DELAY(100000); /* wait for console output to finish */ 384 } 385 386 print_uptime(); 387 388 /* 389 * Ok, now do things that assume all filesystem activity has 390 * been completed. 391 */ 392 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 393 394 /* XXX This doesn't disable interrupts any more. Reconsider? */ 395 splhigh(); 396 397 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 398 doadump(); 399 400 /* Now that we're going to really halt the system... */ 401 EVENTHANDLER_INVOKE(shutdown_final, howto); 402 403 for(;;) ; /* safety against shutdown_reset not working */ 404 /* NOTREACHED */ 405} 406 407/* 408 * If the shutdown was a clean halt, behave accordingly. 409 */ 410static void 411shutdown_halt(void *junk, int howto) 412{ 413 414 if (howto & RB_HALT) { 415 printf("\n"); 416 printf("The operating system has halted.\n"); 417 printf("Please press any key to reboot.\n\n"); 418 switch (cngetc()) { 419 case -1: /* No console, just die */ 420 cpu_halt(); 421 /* NOTREACHED */ 422 default: 423 howto &= ~RB_HALT; 424 break; 425 } 426 } 427} 428 429/* 430 * Check to see if the system paniced, pause and then reboot 431 * according to the specified delay. 432 */ 433static void 434shutdown_panic(void *junk, int howto) 435{ 436 int loop; 437 438 if (howto & RB_DUMP) { 439 if (PANIC_REBOOT_WAIT_TIME != 0) { 440 if (PANIC_REBOOT_WAIT_TIME != -1) { 441 printf("Automatic reboot in %d seconds - " 442 "press a key on the console to abort\n", 443 PANIC_REBOOT_WAIT_TIME); 444 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 445 loop > 0; --loop) { 446 DELAY(1000 * 100); /* 1/10th second */ 447 /* Did user type a key? */ 448 if (cncheckc() != -1) 449 break; 450 } 451 if (!loop) 452 return; 453 } 454 } else { /* zero time specified - reboot NOW */ 455 return; 456 } 457 printf("--> Press a key on the console to reboot,\n"); 458 printf("--> or switch off the system now.\n"); 459 cngetc(); 460 } 461} 462 463/* 464 * Everything done, now reset 465 */ 466static void 467shutdown_reset(void *junk, int howto) 468{ 469 470 printf("Rebooting...\n"); 471 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 472 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 473 cpu_reset(); 474 /* NOTREACHED */ /* assuming reset worked */ 475} 476 477#ifdef SMP 478static u_int panic_cpu = NOCPU; 479#endif 480 481/* 482 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 483 * and then reboots. If we are called twice, then we avoid trying to sync 484 * the disks as this often leads to recursive panics. 485 * 486 * MPSAFE 487 */ 488void 489panic(const char *fmt, ...) 490{ 491 struct thread *td = curthread; 492 int bootopt, newpanic; 493 va_list ap; 494 static char buf[256]; 495 496#ifdef SMP 497 /* 498 * We don't want multiple CPU's to panic at the same time, so we 499 * use panic_cpu as a simple spinlock. We have to keep checking 500 * panic_cpu if we are spinning in case the panic on the first 501 * CPU is canceled. 502 */ 503 if (panic_cpu != PCPU_GET(cpuid)) 504 while (atomic_cmpset_int(&panic_cpu, NOCPU, 505 PCPU_GET(cpuid)) == 0) 506 while (panic_cpu != NOCPU) 507 ; /* nothing */ 508#endif 509 510 bootopt = RB_AUTOBOOT | RB_DUMP; 511 newpanic = 0; 512 if (panicstr) 513 bootopt |= RB_NOSYNC; 514 else { 515 panicstr = fmt; 516 newpanic = 1; 517 } 518 519 va_start(ap, fmt); 520 if (newpanic) { 521 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 522 panicstr = buf; 523 printf("panic: %s\n", buf); 524 } else { 525 printf("panic: "); 526 vprintf(fmt, ap); 527 printf("\n"); 528 } 529 va_end(ap); 530#ifdef SMP 531 printf("cpuid = %d\n", PCPU_GET(cpuid)); 532#endif 533 534#ifdef KDB 535 if (newpanic && trace_on_panic) 536 kdb_backtrace(); 537 if (debugger_on_panic) 538 kdb_enter("panic"); 539#ifdef RESTARTABLE_PANICS 540 /* See if the user aborted the panic, in which case we continue. */ 541 if (panicstr == NULL) { 542#ifdef SMP 543 atomic_store_rel_int(&panic_cpu, NOCPU); 544#endif 545 return; 546 } 547#endif 548#endif 549 mtx_lock_spin(&sched_lock); 550 td->td_flags |= TDF_INPANIC; 551 mtx_unlock_spin(&sched_lock); 552 if (!sync_on_panic) 553 bootopt |= RB_NOSYNC; 554 boot(bootopt); 555} 556 557/* 558 * Support for poweroff delay. 559 */ 560#ifndef POWEROFF_DELAY 561# define POWEROFF_DELAY 5000 562#endif 563static int poweroff_delay = POWEROFF_DELAY; 564 565SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 566 &poweroff_delay, 0, ""); 567 568static void 569poweroff_wait(void *junk, int howto) 570{ 571 572 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 573 return; 574 DELAY(poweroff_delay * 1000); 575} 576 577/* 578 * Some system processes (e.g. syncer) need to be stopped at appropriate 579 * points in their main loops prior to a system shutdown, so that they 580 * won't interfere with the shutdown process (e.g. by holding a disk buf 581 * to cause sync to fail). For each of these system processes, register 582 * shutdown_kproc() as a handler for one of shutdown events. 583 */ 584static int kproc_shutdown_wait = 60; 585SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 586 &kproc_shutdown_wait, 0, ""); 587 588void 589kproc_shutdown(void *arg, int howto) 590{ 591 struct proc *p; 592 char procname[MAXCOMLEN + 1]; 593 int error; 594 595 if (panicstr) 596 return; 597 598 p = (struct proc *)arg; 599 strlcpy(procname, p->p_comm, sizeof(procname)); 600 printf("Waiting (max %d seconds) for system process `%s' to stop...", 601 kproc_shutdown_wait, procname); 602 error = kthread_suspend(p, kproc_shutdown_wait * hz); 603 604 if (error == EWOULDBLOCK) 605 printf("timed out\n"); 606 else 607 printf("done\n"); 608} 609 610/* Registration of dumpers */ 611int 612set_dumper(struct dumperinfo *di) 613{ 614 615 if (di == NULL) { 616 bzero(&dumper, sizeof dumper); 617 return (0); 618 } 619 if (dumper.dumper != NULL) 620 return (EBUSY); 621 dumper = *di; 622 return (0); 623} 624 625#if defined(__powerpc__) 626void 627dumpsys(struct dumperinfo *di __unused) 628{ 629 630 printf("Kernel dumps not implemented on this architecture\n"); 631} 632#endif 633