kern_shutdown.c revision 137186
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 137186 2004-11-04 07:59:57Z phk $"); 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/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/* 108 * Variable panicstr contains argument to first call to panic; used as flag 109 * to indicate that the kernel has already called panic. 110 */ 111const char *panicstr; 112 113int dumping; /* system is dumping */ 114static struct dumperinfo dumper; /* our selected dumper */ 115 116/* Context information for dump-debuggers. */ 117static struct pcb dumppcb; /* Registers. */ 118static lwpid_t dumptid; /* Thread ID. */ 119 120static void boot(int) __dead2; 121static void poweroff_wait(void *, int); 122static void shutdown_halt(void *junk, int howto); 123static void shutdown_panic(void *junk, int howto); 124static void shutdown_reset(void *junk, int howto); 125 126/* register various local shutdown events */ 127static void 128shutdown_conf(void *unused) 129{ 130 131 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 132 SHUTDOWN_PRI_FIRST); 133 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 134 SHUTDOWN_PRI_LAST + 100); 135 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 136 SHUTDOWN_PRI_LAST + 100); 137 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 138 SHUTDOWN_PRI_LAST + 200); 139} 140 141SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 142 143/* 144 * The system call that results in a reboot 145 * 146 * MPSAFE 147 */ 148/* ARGSUSED */ 149int 150reboot(struct thread *td, struct reboot_args *uap) 151{ 152 int error; 153 154 error = 0; 155#ifdef MAC 156 error = mac_check_system_reboot(td->td_ucred, uap->opt); 157#endif 158 if (error == 0) 159 error = suser(td); 160 if (error == 0) { 161 mtx_lock(&Giant); 162 boot(uap->opt); 163 mtx_unlock(&Giant); 164 } 165 return (error); 166} 167 168/* 169 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 170 */ 171static int shutdown_howto = 0; 172 173void 174shutdown_nice(int howto) 175{ 176 177 shutdown_howto = howto; 178 179 /* Send a signal to init(8) and have it shutdown the world */ 180 if (initproc != NULL) { 181 PROC_LOCK(initproc); 182 psignal(initproc, SIGINT); 183 PROC_UNLOCK(initproc); 184 } else { 185 /* No init(8) running, so simply reboot */ 186 boot(RB_NOSYNC); 187 } 188 return; 189} 190static int waittime = -1; 191 192static void 193print_uptime(void) 194{ 195 int f; 196 struct timespec ts; 197 198 getnanouptime(&ts); 199 printf("Uptime: "); 200 f = 0; 201 if (ts.tv_sec >= 86400) { 202 printf("%ldd", (long)ts.tv_sec / 86400); 203 ts.tv_sec %= 86400; 204 f = 1; 205 } 206 if (f || ts.tv_sec >= 3600) { 207 printf("%ldh", (long)ts.tv_sec / 3600); 208 ts.tv_sec %= 3600; 209 f = 1; 210 } 211 if (f || ts.tv_sec >= 60) { 212 printf("%ldm", (long)ts.tv_sec / 60); 213 ts.tv_sec %= 60; 214 f = 1; 215 } 216 printf("%lds\n", (long)ts.tv_sec); 217} 218 219static void 220doadump(void) 221{ 222 223 /* 224 * Sometimes people have to call this from the kernel debugger. 225 * (if 'panic' can not dump) 226 * Give them a clue as to why they can't dump. 227 */ 228 if (dumper.dumper == NULL) { 229 printf("Cannot dump. No dump device defined.\n"); 230 return; 231 } 232 233 savectx(&dumppcb); 234 dumptid = curthread->td_tid; 235 dumping++; 236 dumpsys(&dumper); 237} 238 239/* 240 * Go through the rigmarole of shutting down.. 241 * this used to be in machdep.c but I'll be dammned if I could see 242 * anything machine dependant in it. 243 */ 244static void 245boot(int howto) 246{ 247 static int first_buf_printf = 1; 248 249 /* collect extra flags that shutdown_nice might have set */ 250 howto |= shutdown_howto; 251 252 /* We are out of the debugger now. */ 253 kdb_active = 0; 254 255#ifdef SMP 256 if (smp_active) 257 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid)); 258#endif 259 /* 260 * Do any callouts that should be done BEFORE syncing the filesystems. 261 */ 262 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 263 264 /* 265 * Now sync filesystems 266 */ 267 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 268 register struct buf *bp; 269 int iter, nbusy, pbusy; 270#ifndef PREEMPTION 271 int subiter; 272#endif 273 274 waittime = 0; 275 276 sync(&thread0, NULL); 277 278 /* 279 * With soft updates, some buffers that are 280 * written will be remarked as dirty until other 281 * buffers are written. 282 */ 283 for (iter = pbusy = 0; iter < 20; iter++) { 284 nbusy = 0; 285 for (bp = &buf[nbuf]; --bp >= buf; ) { 286 if ((bp->b_flags & B_INVAL) == 0 && 287 BUF_REFCNT(bp) > 0) { 288 nbusy++; 289 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 290 == B_DELWRI) { 291 /* bawrite(bp);*/ 292 nbusy++; 293 } 294 } 295 if (nbusy == 0) { 296 if (first_buf_printf) 297 printf("All buffers synced."); 298 break; 299 } 300 if (first_buf_printf) { 301 printf("Syncing disks, buffers remaining... "); 302 first_buf_printf = 0; 303 } 304 printf("%d ", nbusy); 305 if (nbusy < pbusy) 306 iter = 0; 307 pbusy = nbusy; 308 sync(&thread0, NULL); 309 310#ifdef PREEMPTION 311 /* 312 * Drop Giant and spin for a while to allow 313 * interrupt threads to run. 314 */ 315 DROP_GIANT(); 316 DELAY(50000 * iter); 317 PICKUP_GIANT(); 318#else 319 /* 320 * Drop Giant and context switch several times to 321 * allow interrupt threads to run. 322 */ 323 DROP_GIANT(); 324 for (subiter = 0; subiter < 50 * iter; subiter++) { 325 mtx_lock_spin(&sched_lock); 326 mi_switch(SW_VOL, NULL); 327 mtx_unlock_spin(&sched_lock); 328 DELAY(1000); 329 } 330 PICKUP_GIANT(); 331#endif 332 } 333 printf("\n"); 334 /* 335 * Count only busy local buffers to prevent forcing 336 * a fsck if we're just a client of a wedged NFS server 337 */ 338 nbusy = 0; 339 for (bp = &buf[nbuf]; --bp >= buf; ) { 340 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 341 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 342#if 0 343/* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 344 if (bp->b_dev == NULL) { 345 TAILQ_REMOVE(&mountlist, 346 bp->b_vp->v_mount, mnt_list); 347 continue; 348 } 349#endif 350 nbusy++; 351#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 352 printf( 353 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n", 354 nbusy, bp->b_bufobj, 355 bp->b_flags, (long)bp->b_blkno, 356 (long)bp->b_lblkno); 357#endif 358 } 359 } 360 if (nbusy) { 361 /* 362 * Failed to sync all blocks. Indicate this and don't 363 * unmount filesystems (thus forcing an fsck on reboot). 364 */ 365 printf("Giving up on %d buffers\n", nbusy); 366 DELAY(5000000); /* 5 seconds */ 367 } else { 368 if (!first_buf_printf) 369 printf("Final sync complete\n"); 370 /* 371 * Unmount filesystems 372 */ 373 if (panicstr == 0) 374 vfs_unmountall(); 375 } 376 DELAY(100000); /* wait for console output to finish */ 377 } 378 379 print_uptime(); 380 381 /* 382 * Ok, now do things that assume all filesystem activity has 383 * been completed. 384 */ 385 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 386 splhigh(); 387 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 388 doadump(); 389 390 /* Now that we're going to really halt the system... */ 391 EVENTHANDLER_INVOKE(shutdown_final, howto); 392 393 for(;;) ; /* safety against shutdown_reset not working */ 394 /* NOTREACHED */ 395} 396 397/* 398 * If the shutdown was a clean halt, behave accordingly. 399 */ 400static void 401shutdown_halt(void *junk, int howto) 402{ 403 404 if (howto & RB_HALT) { 405 printf("\n"); 406 printf("The operating system has halted.\n"); 407 printf("Please press any key to reboot.\n\n"); 408 switch (cngetc()) { 409 case -1: /* No console, just die */ 410 cpu_halt(); 411 /* NOTREACHED */ 412 default: 413 howto &= ~RB_HALT; 414 break; 415 } 416 } 417} 418 419/* 420 * Check to see if the system paniced, pause and then reboot 421 * according to the specified delay. 422 */ 423static void 424shutdown_panic(void *junk, int howto) 425{ 426 int loop; 427 428 if (howto & RB_DUMP) { 429 if (PANIC_REBOOT_WAIT_TIME != 0) { 430 if (PANIC_REBOOT_WAIT_TIME != -1) { 431 printf("Automatic reboot in %d seconds - " 432 "press a key on the console to abort\n", 433 PANIC_REBOOT_WAIT_TIME); 434 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 435 loop > 0; --loop) { 436 DELAY(1000 * 100); /* 1/10th second */ 437 /* Did user type a key? */ 438 if (cncheckc() != -1) 439 break; 440 } 441 if (!loop) 442 return; 443 } 444 } else { /* zero time specified - reboot NOW */ 445 return; 446 } 447 printf("--> Press a key on the console to reboot,\n"); 448 printf("--> or switch off the system now.\n"); 449 cngetc(); 450 } 451} 452 453/* 454 * Everything done, now reset 455 */ 456static void 457shutdown_reset(void *junk, int howto) 458{ 459 460 printf("Rebooting...\n"); 461 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 462 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 463 cpu_reset(); 464 /* NOTREACHED */ /* assuming reset worked */ 465} 466 467#ifdef SMP 468static u_int panic_cpu = NOCPU; 469#endif 470 471/* 472 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 473 * and then reboots. If we are called twice, then we avoid trying to sync 474 * the disks as this often leads to recursive panics. 475 * 476 * MPSAFE 477 */ 478void 479panic(const char *fmt, ...) 480{ 481 struct thread *td = curthread; 482 int bootopt, newpanic; 483 va_list ap; 484 static char buf[256]; 485 486#ifdef SMP 487 /* 488 * We don't want multiple CPU's to panic at the same time, so we 489 * use panic_cpu as a simple spinlock. We have to keep checking 490 * panic_cpu if we are spinning in case the panic on the first 491 * CPU is canceled. 492 */ 493 if (panic_cpu != PCPU_GET(cpuid)) 494 while (atomic_cmpset_int(&panic_cpu, NOCPU, 495 PCPU_GET(cpuid)) == 0) 496 while (panic_cpu != NOCPU) 497 ; /* nothing */ 498#endif 499 500 bootopt = RB_AUTOBOOT | RB_DUMP; 501 newpanic = 0; 502 if (panicstr) 503 bootopt |= RB_NOSYNC; 504 else { 505 panicstr = fmt; 506 newpanic = 1; 507 } 508 509 va_start(ap, fmt); 510 if (newpanic) { 511 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 512 panicstr = buf; 513 printf("panic: %s\n", buf); 514 } else { 515 printf("panic: "); 516 vprintf(fmt, ap); 517 printf("\n"); 518 } 519 va_end(ap); 520#ifdef SMP 521 printf("cpuid = %d\n", PCPU_GET(cpuid)); 522#endif 523 524#ifdef KDB 525 if (newpanic && trace_on_panic) 526 kdb_backtrace(); 527 if (debugger_on_panic) 528 kdb_enter("panic"); 529#ifdef RESTARTABLE_PANICS 530 /* See if the user aborted the panic, in which case we continue. */ 531 if (panicstr == NULL) { 532#ifdef SMP 533 atomic_store_rel_int(&panic_cpu, NOCPU); 534#endif 535 return; 536 } 537#endif 538#endif 539 mtx_lock_spin(&sched_lock); 540 td->td_flags |= TDF_INPANIC; 541 mtx_unlock_spin(&sched_lock); 542 if (!sync_on_panic) 543 bootopt |= RB_NOSYNC; 544 boot(bootopt); 545} 546 547/* 548 * Support for poweroff delay. 549 */ 550#ifndef POWEROFF_DELAY 551# define POWEROFF_DELAY 5000 552#endif 553static int poweroff_delay = POWEROFF_DELAY; 554 555SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 556 &poweroff_delay, 0, ""); 557 558static void 559poweroff_wait(void *junk, int howto) 560{ 561 562 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 563 return; 564 DELAY(poweroff_delay * 1000); 565} 566 567/* 568 * Some system processes (e.g. syncer) need to be stopped at appropriate 569 * points in their main loops prior to a system shutdown, so that they 570 * won't interfere with the shutdown process (e.g. by holding a disk buf 571 * to cause sync to fail). For each of these system processes, register 572 * shutdown_kproc() as a handler for one of shutdown events. 573 */ 574static int kproc_shutdown_wait = 60; 575SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 576 &kproc_shutdown_wait, 0, ""); 577 578void 579kproc_shutdown(void *arg, int howto) 580{ 581 struct proc *p; 582 char procname[MAXCOMLEN + 1]; 583 int error; 584 585 if (panicstr) 586 return; 587 588 p = (struct proc *)arg; 589 strlcpy(procname, p->p_comm, sizeof(procname)); 590 printf("Waiting (max %d seconds) for system process `%s' to stop...", 591 kproc_shutdown_wait, procname); 592 error = kthread_suspend(p, kproc_shutdown_wait * hz); 593 594 if (error == EWOULDBLOCK) 595 printf("timed out\n"); 596 else 597 printf("done\n"); 598} 599 600/* Registration of dumpers */ 601int 602set_dumper(struct dumperinfo *di) 603{ 604 605 if (di == NULL) { 606 bzero(&dumper, sizeof dumper); 607 return (0); 608 } 609 if (dumper.dumper != NULL) 610 return (EBUSY); 611 dumper = *di; 612 return (0); 613} 614 615#if defined(__powerpc__) 616void 617dumpsys(struct dumperinfo *di __unused) 618{ 619 620 printf("Kernel dumps not implemented on this architecture\n"); 621} 622#endif 623