kern_shutdown.c revision 83703
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 83703 2001-09-20 06:08:53Z peter $ 40 */ 41 42#include "opt_ddb.h" 43#include "opt_hw_wdog.h" 44#include "opt_panic.h" 45#include "opt_show_busybufs.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/bio.h> 50#include <sys/buf.h> 51#include <sys/conf.h> 52#include <sys/cons.h> 53#include <sys/disklabel.h> 54#include <sys/eventhandler.h> 55#include <sys/kernel.h> 56#include <sys/kthread.h> 57#include <sys/mount.h> 58#include <sys/proc.h> 59#include <sys/reboot.h> 60#include <sys/resourcevar.h> 61#include <sys/smp.h> /* smp_active */ 62#include <sys/sysctl.h> 63#include <sys/sysproto.h> 64#include <sys/vnode.h> 65 66#include <machine/pcb.h> 67#include <machine/md_var.h> 68 69#include <sys/signalvar.h> 70#ifdef DDB 71#include <ddb/ddb.h> 72#endif 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 DDB 85#ifdef DDB_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#endif 93 94SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 95 96#ifdef HW_WDOG 97/* 98 * If there is a hardware watchdog, point this at the function needed to 99 * hold it off. 100 * It's needed when the kernel needs to do some lengthy operations. 101 * e.g. in wd.c when dumping core.. It's most annoying to have 102 * your precious core-dump only half written because the wdog kicked in. 103 */ 104watchdog_tickle_fn wdog_tickler = NULL; 105#endif /* HW_WDOG */ 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 */ 114 115static void boot(int) __dead2; 116static void dumpsys(void); 117static void poweroff_wait(void *, int); 118static void shutdown_halt(void *junk, int howto); 119static void shutdown_panic(void *junk, int howto); 120static void shutdown_reset(void *junk, int howto); 121 122/* register various local shutdown events */ 123static void 124shutdown_conf(void *unused) 125{ 126 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 127 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 128 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 129 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 130} 131 132SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 133 134/* 135 * The system call that results in a reboot 136 * 137 * MPSAFE 138 */ 139/* ARGSUSED */ 140int 141reboot(struct thread *td, struct reboot_args *uap) 142{ 143 int error; 144 145 mtx_lock(&Giant); 146 if ((error = suser_td(td)) == 0) 147 boot(uap->opt); 148 mtx_unlock(&Giant); 149 return (error); 150} 151 152/* 153 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 154 */ 155static int shutdown_howto = 0; 156 157void 158shutdown_nice(int howto) 159{ 160 shutdown_howto = howto; 161 162 /* Send a signal to init(8) and have it shutdown the world */ 163 if (initproc != NULL) { 164 PROC_LOCK(initproc); 165 psignal(initproc, SIGINT); 166 PROC_UNLOCK(initproc); 167 } else { 168 /* No init(8) running, so simply reboot */ 169 boot(RB_NOSYNC); 170 } 171 return; 172} 173static int waittime = -1; 174static struct pcb dumppcb; 175 176static void 177print_uptime(void) 178{ 179 int f; 180 struct timespec ts; 181 182 getnanouptime(&ts); 183 printf("Uptime: "); 184 f = 0; 185 if (ts.tv_sec >= 86400) { 186 printf("%ldd", (long)ts.tv_sec / 86400); 187 ts.tv_sec %= 86400; 188 f = 1; 189 } 190 if (f || ts.tv_sec >= 3600) { 191 printf("%ldh", (long)ts.tv_sec / 3600); 192 ts.tv_sec %= 3600; 193 f = 1; 194 } 195 if (f || ts.tv_sec >= 60) { 196 printf("%ldm", (long)ts.tv_sec / 60); 197 ts.tv_sec %= 60; 198 f = 1; 199 } 200 printf("%lds\n", (long)ts.tv_sec); 201} 202 203/* 204 * Go through the rigmarole of shutting down.. 205 * this used to be in machdep.c but I'll be dammned if I could see 206 * anything machine dependant in it. 207 */ 208static void 209boot(int howto) 210{ 211 212 /* collect extra flags that shutdown_nice might have set */ 213 howto |= shutdown_howto; 214 215#ifdef DDB 216 /* We are out of the debugger now. */ 217 db_active = 0; 218#endif 219 220#ifdef SMP 221 if (smp_active) 222 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid)); 223#endif 224 /* 225 * Do any callouts that should be done BEFORE syncing the filesystems. 226 */ 227 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 228 229 /* 230 * Now sync filesystems 231 */ 232 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 233 register struct buf *bp; 234 int iter, nbusy, pbusy; 235 int subiter; 236 237 waittime = 0; 238 printf("\nsyncing disks... "); 239 240 sync(thread0, NULL); 241 242 /* 243 * With soft updates, some buffers that are 244 * written will be remarked as dirty until other 245 * buffers are written. 246 */ 247 for (iter = pbusy = 0; iter < 20; iter++) { 248 nbusy = 0; 249 for (bp = &buf[nbuf]; --bp >= buf; ) { 250 if ((bp->b_flags & B_INVAL) == 0 && 251 BUF_REFCNT(bp) > 0) { 252 nbusy++; 253 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 254 == B_DELWRI) { 255 /* bawrite(bp);*/ 256 nbusy++; 257 } 258 } 259 if (nbusy == 0) 260 break; 261 printf("%d ", nbusy); 262 if (nbusy < pbusy) 263 iter = 0; 264 pbusy = nbusy; 265 sync(thread0, NULL); 266 if (curthread != NULL) { 267 DROP_GIANT_NOSWITCH(); 268 for (subiter = 0; subiter < 50 * iter; subiter++) { 269 mtx_lock_spin(&sched_lock); 270 setrunqueue(curthread); 271 curthread->td_proc->p_stats->p_ru.ru_nvcsw++; 272 mi_switch(); /* Allow interrupt threads to run */ 273 mtx_unlock_spin(&sched_lock); 274 DELAY(1000); 275 } 276 PICKUP_GIANT(); 277 } else 278 DELAY(50000 * iter); 279 } 280 printf("\n"); 281 /* 282 * Count only busy local buffers to prevent forcing 283 * a fsck if we're just a client of a wedged NFS server 284 */ 285 nbusy = 0; 286 for (bp = &buf[nbuf]; --bp >= buf; ) { 287 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 288 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 289 if (bp->b_dev == NODEV) { 290 TAILQ_REMOVE(&mountlist, 291 bp->b_vp->v_mount, mnt_list); 292 continue; 293 } 294 nbusy++; 295#if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 296 printf( 297 "%d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n", 298 nbusy, devtoname(bp->b_dev), 299 bp->b_flags, (long)bp->b_blkno, 300 (long)bp->b_lblkno); 301#endif 302 } 303 } 304 if (nbusy) { 305 /* 306 * Failed to sync all blocks. Indicate this and don't 307 * unmount filesystems (thus forcing an fsck on reboot). 308 */ 309 printf("giving up on %d buffers\n", nbusy); 310 DELAY(5000000); /* 5 seconds */ 311 } else { 312 printf("done\n"); 313 /* 314 * Unmount filesystems 315 */ 316 if (panicstr == 0) 317 vfs_unmountall(); 318 } 319 DELAY(100000); /* wait for console output to finish */ 320 } 321 322 print_uptime(); 323 324 /* 325 * Ok, now do things that assume all filesystem activity has 326 * been completed. 327 */ 328 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 329 splhigh(); 330 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) 331 dumpsys(); 332 333 /* Now that we're going to really halt the system... */ 334 EVENTHANDLER_INVOKE(shutdown_final, howto); 335 336 for(;;) ; /* safety against shutdown_reset not working */ 337 /* NOTREACHED */ 338} 339 340/* 341 * If the shutdown was a clean halt, behave accordingly. 342 */ 343static void 344shutdown_halt(void *junk, int howto) 345{ 346 if (howto & RB_HALT) { 347 printf("\n"); 348 printf("The operating system has halted.\n"); 349 printf("Please press any key to reboot.\n\n"); 350 switch (cngetc()) { 351 case -1: /* No console, just die */ 352 cpu_halt(); 353 /* NOTREACHED */ 354 default: 355 howto &= ~RB_HALT; 356 break; 357 } 358 } 359} 360 361/* 362 * Check to see if the system paniced, pause and then reboot 363 * according to the specified delay. 364 */ 365static void 366shutdown_panic(void *junk, int howto) 367{ 368 int loop; 369 370 if (howto & RB_DUMP) { 371 if (PANIC_REBOOT_WAIT_TIME != 0) { 372 if (PANIC_REBOOT_WAIT_TIME != -1) { 373 printf("Automatic reboot in %d seconds - " 374 "press a key on the console to abort\n", 375 PANIC_REBOOT_WAIT_TIME); 376 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 377 loop > 0; --loop) { 378 DELAY(1000 * 100); /* 1/10th second */ 379 /* Did user type a key? */ 380 if (cncheckc() != -1) 381 break; 382 } 383 if (!loop) 384 return; 385 } 386 } else { /* zero time specified - reboot NOW */ 387 return; 388 } 389 printf("--> Press a key on the console to reboot <--\n"); 390 cngetc(); 391 } 392} 393 394/* 395 * Everything done, now reset 396 */ 397static void 398shutdown_reset(void *junk, int howto) 399{ 400 printf("Rebooting...\n"); 401 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 402 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 403 cpu_reset(); 404 /* NOTREACHED */ /* assuming reset worked */ 405} 406 407/* 408 * Magic number for savecore 409 * 410 * exported (symorder) and used at least by savecore(8) 411 * 412 */ 413static u_long const dumpmag = 0x8fca0101UL; 414 415static int dumpsize = 0; /* also for savecore */ 416 417static int dodump = 1; 418 419SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 420 "Try to perform coredump on kernel panic"); 421 422static int 423setdumpdev(dev_t dev) 424{ 425 int psize; 426 long newdumplo; 427 428 if (dev == NODEV) { 429 dumpdev = dev; 430 return (0); 431 } 432 if (devsw(dev) == NULL) 433 return (ENXIO); /* XXX is this right? */ 434 if (devsw(dev)->d_psize == NULL) 435 return (ENXIO); /* XXX should be ENODEV ? */ 436 psize = devsw(dev)->d_psize(dev); 437 if (psize == -1) 438 return (ENXIO); /* XXX should be ENODEV ? */ 439 /* 440 * XXX should clean up checking in dumpsys() to be more like this. 441 */ 442 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE); 443 if (newdumplo <= LABELSECTOR) 444 return (ENOSPC); 445 dumpdev = dev; 446 dumplo = newdumplo; 447 return (0); 448} 449 450 451/* ARGSUSED */ 452static void 453dump_conf(void *dummy) 454{ 455 if (setdumpdev(dumpdev) != 0) 456 dumpdev = NODEV; 457} 458 459SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) 460 461static int 462sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 463{ 464 int error; 465 udev_t ndumpdev; 466 467 ndumpdev = dev2udev(dumpdev); 468 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 469 if (error == 0 && req->newptr != NULL) 470 error = setdumpdev(udev2dev(ndumpdev, 0)); 471 return (error); 472} 473 474SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 475 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", ""); 476 477/* 478 * Doadump comes here after turning off memory management and 479 * getting on the dump stack, either when called above, or by 480 * the auto-restart code. 481 */ 482static void 483dumpsys(void) 484{ 485 int error; 486 487 savectx(&dumppcb); 488 if (!dodump) 489 return; 490 if (dumpdev == NODEV) 491 return; 492 if (!(devsw(dumpdev))) 493 return; 494 if (!(devsw(dumpdev)->d_dump)) 495 return; 496 if (dumping++) { 497 dumping--; 498 printf("Dump already in progress, bailing...\n"); 499 return; 500 } 501 dumpsize = Maxmem; 502 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo); 503 printf("dump "); 504 error = (*devsw(dumpdev)->d_dump)(dumpdev); 505 dumping--; 506 if (error == 0) { 507 printf("succeeded\n"); 508 return; 509 } 510 printf("failed, reason: "); 511 switch (error) { 512 case ENODEV: 513 printf("device doesn't support a dump routine\n"); 514 break; 515 516 case ENXIO: 517 printf("device bad\n"); 518 break; 519 520 case EFAULT: 521 printf("device not ready\n"); 522 break; 523 524 case EINVAL: 525 printf("area improper\n"); 526 break; 527 528 case EIO: 529 printf("i/o error\n"); 530 break; 531 532 case EINTR: 533 printf("aborted from console\n"); 534 break; 535 536 default: 537 printf("unknown, error = %d\n", error); 538 break; 539 } 540} 541 542int 543dumpstatus(vm_offset_t addr, long count) 544{ 545 int c; 546 547 if (addr % (1024 * 1024) == 0) { 548#ifdef HW_WDOG 549 if (wdog_tickler) 550 (*wdog_tickler)(); 551#endif 552 printf("%ld ", count / (1024 * 1024)); 553 } 554 555 if ((c = cncheckc()) == 0x03) 556 return -1; 557 else if (c != -1) 558 printf("[CTRL-C to abort] "); 559 560 return 0; 561} 562 563#ifdef SMP 564static u_int panic_cpu = NOCPU; 565#endif 566 567/* 568 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 569 * and then reboots. If we are called twice, then we avoid trying to sync 570 * the disks as this often leads to recursive panics. 571 * 572 * MPSAFE 573 */ 574void 575panic(const char *fmt, ...) 576{ 577 int bootopt; 578 va_list ap; 579 static char buf[256]; 580 581#ifdef SMP 582 /* 583 * We don't want multiple CPU's to panic at the same time, so we 584 * use panic_cpu as a simple spinlock. We have to keep checking 585 * panic_cpu if we are spinning in case the panic on the first 586 * CPU is canceled. 587 */ 588 if (panic_cpu != PCPU_GET(cpuid)) 589 while (atomic_cmpset_int(&panic_cpu, NOCPU, 590 PCPU_GET(cpuid)) == 0) 591 while (panic_cpu != NOCPU) 592 ; /* nothing */ 593#endif 594 595 bootopt = RB_AUTOBOOT | RB_DUMP; 596 if (panicstr) 597 bootopt |= RB_NOSYNC; 598 else 599 panicstr = fmt; 600 601 va_start(ap, fmt); 602 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 603 if (panicstr == fmt) 604 panicstr = buf; 605 va_end(ap); 606 printf("panic: %s\n", buf); 607#ifdef SMP 608 /* two separate prints in case of an unmapped page and trap */ 609 printf("cpuid = %d; ", PCPU_GET(cpuid)); 610#ifdef APIC_IO 611 printf("lapic.id = %08x\n", lapic.id); 612#endif 613#endif 614 615#if defined(DDB) 616 if (debugger_on_panic) 617 Debugger ("panic"); 618#ifdef RESTARTABLE_PANICS 619 /* See if the user aborted the panic, in which case we continue. */ 620 if (panicstr == NULL) { 621#ifdef SMP 622 atomic_store_rel_int(&panic_cpu, NOCPU); 623#endif 624 return; 625 } 626#endif 627#endif 628 boot(bootopt); 629} 630 631/* 632 * Support for poweroff delay. 633 */ 634#ifndef POWEROFF_DELAY 635# define POWEROFF_DELAY 5000 636#endif 637static int poweroff_delay = POWEROFF_DELAY; 638 639SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 640 &poweroff_delay, 0, ""); 641 642static void 643poweroff_wait(void *junk, int howto) 644{ 645 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 646 return; 647 DELAY(poweroff_delay * 1000); 648} 649 650/* 651 * Some system processes (e.g. syncer) need to be stopped at appropriate 652 * points in their main loops prior to a system shutdown, so that they 653 * won't interfere with the shutdown process (e.g. by holding a disk buf 654 * to cause sync to fail). For each of these system processes, register 655 * shutdown_kproc() as a handler for one of shutdown events. 656 */ 657static int kproc_shutdown_wait = 60; 658SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 659 &kproc_shutdown_wait, 0, ""); 660 661void 662kproc_shutdown(void *arg, int howto) 663{ 664 struct proc *p; 665 int error; 666 667 if (panicstr) 668 return; 669 670 p = (struct proc *)arg; 671 printf("Waiting (max %d seconds) for system process `%s' to stop...", 672 kproc_shutdown_wait, p->p_comm); 673 error = kthread_suspend(p, kproc_shutdown_wait * hz); 674 675 if (error == EWOULDBLOCK) 676 printf("timed out\n"); 677 else 678 printf("stopped\n"); 679} 680