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: src/sys/kern/kern_shutdown.c,v 1.72.2.12 2002/02/21 19:15:10 dillon Exp $ 40 * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.10 2003/07/26 19:42:11 rob Exp $ 41 */ 42 43 #include "opt_ddb.h" 44 #include "opt_hw_wdog.h" 45 #include "opt_panic.h" 46 #include "opt_show_busybufs.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/eventhandler.h> 51 #include <sys/buf.h> 52 #include <sys/disklabel.h> 53 #include <sys/reboot.h> 54 #include <sys/proc.h> 55 #include <sys/vnode.h> 56 #include <sys/kernel.h> 57 #include <sys/kthread.h> 58 #include <sys/malloc.h> 59 #include <sys/mount.h> 60 #include <sys/queue.h> 61 #include <sys/sysctl.h> 62 #include <sys/conf.h> 63 #include <sys/sysproto.h> 64 #include <sys/device.h> 65 #include <sys/cons.h> 66 #include <sys/buf2.h> 67 68 #include <machine/pcb.h> 69 #include <machine/clock.h> 70 #include <machine/md_var.h> 71 #include <machine/smp.h> /* smp_active, cpuid */ 72 73 #include <sys/signalvar.h> 74 75 #ifndef PANIC_REBOOT_WAIT_TIME 76 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 77 #endif 78 79 /* 80 * Note that stdarg.h and the ANSI style va_start macro is used for both 81 * ANSI and traditional C compilers. 82 */ 83 #include <machine/stdarg.h> 84 85 #ifdef DDB 86 #ifdef DDB_UNATTENDED 87 int debugger_on_panic = 0; 88 #else 89 int debugger_on_panic = 1; 90 #endif 91 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 92 &debugger_on_panic, 0, "Run debugger on kernel panic"); 93 #endif 94 95 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 96 97 #ifdef HW_WDOG 98 /* 99 * If there is a hardware watchdog, point this at the function needed to 100 * hold it off. 101 * It's needed when the kernel needs to do some lengthy operations. 102 * e.g. in wd.c when dumping core.. It's most annoying to have 103 * your precious core-dump only half written because the wdog kicked in. 104 */ 105 watchdog_tickle_fn wdog_tickler = NULL; 106 #endif /* HW_WDOG */ 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 */ 112 const char *panicstr; 113 114 int dumping; /* system is dumping */ 115 116 static void boot __P((int)) __dead2; 117 static void dumpsys __P((void)); 118 static int setdumpdev __P((dev_t dev)); 119 static void poweroff_wait __P((void *, int)); 120 static void print_uptime __P((void)); 121 static void shutdown_halt __P((void *junk, int howto)); 122 static void shutdown_panic __P((void *junk, int howto)); 123 static void shutdown_reset __P((void *junk, int howto)); 124 125 /* register various local shutdown events */ 126 static void 127 shutdown_conf(void *unused) 128 { 129 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 130 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 131 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 132 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 133 } 134 135 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 136 137 /* ARGSUSED */ 138 139 /* 140 * The system call that results in a reboot 141 */ 142 int 143 reboot(struct reboot_args *uap) 144 { 145 struct thread *td = curthread; 146 int error; 147 148 if ((error = suser(td))) 149 return (error); 150 151 boot(uap->opt); 152 return (0); 153 } 154 155 /* 156 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 157 */ 158 static int shutdown_howto = 0; 159 160 void 161 shutdown_nice(int howto) 162 { 163 shutdown_howto = howto; 164 165 /* Send a signal to init(8) and have it shutdown the world */ 166 if (initproc != NULL) { 167 psignal(initproc, SIGINT); 168 } else { 169 /* No init(8) running, so simply reboot */ 170 boot(RB_NOSYNC); 171 } 172 return; 173 } 174 static int waittime = -1; 175 static struct pcb dumppcb; 176 177 static void 178 print_uptime() 179 { 180 int f; 181 struct timespec ts; 182 183 getnanouptime(&ts); 184 printf("Uptime: "); 185 f = 0; 186 if (ts.tv_sec >= 86400) { 187 printf("%ldd", ts.tv_sec / 86400); 188 ts.tv_sec %= 86400; 189 f = 1; 190 } 191 if (f || ts.tv_sec >= 3600) { 192 printf("%ldh", ts.tv_sec / 3600); 193 ts.tv_sec %= 3600; 194 f = 1; 195 } 196 if (f || ts.tv_sec >= 60) { 197 printf("%ldm", ts.tv_sec / 60); 198 ts.tv_sec %= 60; 199 f = 1; 200 } 201 printf("%lds\n", ts.tv_sec); 202 } 203 204 /* 205 * Go through the rigmarole of shutting down.. 206 * this used to be in machdep.c but I'll be dammned if I could see 207 * anything machine dependant in it. 208 */ 209 static void 210 boot(int howto) 211 { 212 213 /* collect extra flags that shutdown_nice might have set */ 214 howto |= shutdown_howto; 215 216 #ifdef SMP 217 if (smp_active) { 218 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid); 219 } 220 #endif 221 /* 222 * Do any callouts that should be done BEFORE syncing the filesystems. 223 */ 224 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 225 226 /* 227 * Now sync filesystems 228 */ 229 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 230 struct buf *bp; 231 int iter, nbusy, pbusy; 232 233 waittime = 0; 234 printf("\nsyncing disks... "); 235 236 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 237 238 /* 239 * With soft updates, some buffers that are 240 * written will be remarked as dirty until other 241 * buffers are written. 242 */ 243 for (iter = pbusy = 0; iter < 20; iter++) { 244 nbusy = 0; 245 for (bp = &buf[nbuf]; --bp >= buf; ) { 246 if ((bp->b_flags & B_INVAL) == 0 && 247 BUF_REFCNT(bp) > 0) { 248 nbusy++; 249 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 250 == B_DELWRI) { 251 /* bawrite(bp);*/ 252 nbusy++; 253 } 254 } 255 if (nbusy == 0) 256 break; 257 printf("%d ", nbusy); 258 if (nbusy < pbusy) 259 iter = 0; 260 pbusy = nbusy; 261 if (iter > 5 && bioops.io_sync) 262 (*bioops.io_sync)(NULL); 263 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 264 DELAY(50000 * iter); 265 } 266 printf("\n"); 267 /* 268 * Count only busy local buffers to prevent forcing 269 * a fsck if we're just a client of a wedged NFS server 270 */ 271 nbusy = 0; 272 for (bp = &buf[nbuf]; --bp >= buf; ) { 273 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 274 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 275 if (bp->b_dev == NODEV) { 276 TAILQ_REMOVE(&mountlist, 277 bp->b_vp->v_mount, mnt_list); 278 continue; 279 } 280 nbusy++; 281 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 282 printf( 283 "%p %d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n", 284 bp, nbusy, devtoname(bp->b_dev), 285 bp->b_flags, (long)bp->b_blkno, 286 (long)bp->b_lblkno); 287 #endif 288 } 289 } 290 if (nbusy) { 291 /* 292 * Failed to sync all blocks. Indicate this and don't 293 * unmount filesystems (thus forcing an fsck on reboot). 294 */ 295 printf("giving up on %d buffers\n", nbusy); 296 DELAY(5000000); /* 5 seconds */ 297 } else { 298 printf("done\n"); 299 /* 300 * Unmount filesystems 301 */ 302 if (panicstr == 0) 303 vfs_unmountall(); 304 } 305 DELAY(100000); /* wait for console output to finish */ 306 } 307 308 print_uptime(); 309 310 /* 311 * Ok, now do things that assume all filesystem activity has 312 * been completed. 313 */ 314 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 315 splhigh(); 316 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) 317 dumpsys(); 318 319 /* Now that we're going to really halt the system... */ 320 EVENTHANDLER_INVOKE(shutdown_final, howto); 321 322 for(;;) ; /* safety against shutdown_reset not working */ 323 /* NOTREACHED */ 324 } 325 326 /* 327 * If the shutdown was a clean halt, behave accordingly. 328 */ 329 static void 330 shutdown_halt(void *junk, int howto) 331 { 332 if (howto & RB_HALT) { 333 printf("\n"); 334 printf("The operating system has halted.\n"); 335 printf("Please press any key to reboot.\n\n"); 336 switch (cngetc()) { 337 case -1: /* No console, just die */ 338 cpu_halt(); 339 /* NOTREACHED */ 340 default: 341 howto &= ~RB_HALT; 342 break; 343 } 344 } 345 } 346 347 /* 348 * Check to see if the system paniced, pause and then reboot 349 * according to the specified delay. 350 */ 351 static void 352 shutdown_panic(void *junk, int howto) 353 { 354 int loop; 355 356 if (howto & RB_DUMP) { 357 if (PANIC_REBOOT_WAIT_TIME != 0) { 358 if (PANIC_REBOOT_WAIT_TIME != -1) { 359 printf("Automatic reboot in %d seconds - " 360 "press a key on the console to abort\n", 361 PANIC_REBOOT_WAIT_TIME); 362 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 363 loop > 0; --loop) { 364 DELAY(1000 * 100); /* 1/10th second */ 365 /* Did user type a key? */ 366 if (cncheckc() != -1) 367 break; 368 } 369 if (!loop) 370 return; 371 } 372 } else { /* zero time specified - reboot NOW */ 373 return; 374 } 375 printf("--> Press a key on the console to reboot,\n"); 376 printf("--> or switch off the system now.\n"); 377 cngetc(); 378 } 379 } 380 381 /* 382 * Everything done, now reset 383 */ 384 static void 385 shutdown_reset(void *junk, int howto) 386 { 387 printf("Rebooting...\n"); 388 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 389 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 390 cpu_reset(); 391 /* NOTREACHED */ /* assuming reset worked */ 392 } 393 394 /* 395 * Magic number for savecore 396 * 397 * exported (symorder) and used at least by savecore(8) 398 * 399 */ 400 static u_long const dumpmag = 0x8fca0101UL; 401 402 static int dumpsize = 0; /* also for savecore */ 403 404 static int dodump = 1; 405 406 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 407 "Try to perform coredump on kernel panic"); 408 409 static int 410 setdumpdev(dev) 411 dev_t dev; 412 { 413 int psize; 414 long newdumplo; 415 416 if (dev == NODEV) { 417 dumpdev = dev; 418 return (0); 419 } 420 psize = dev_dpsize(dev); 421 if (psize == -1) 422 return (ENXIO); 423 /* 424 * XXX should clean up checking in dumpsys() to be more like this. 425 */ 426 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE); 427 if (newdumplo <= LABELSECTOR) 428 return (ENOSPC); 429 dumpdev = dev; 430 dumplo = newdumplo; 431 return (0); 432 } 433 434 435 /* ARGSUSED */ 436 static void dump_conf __P((void *dummy)); 437 static void 438 dump_conf(dummy) 439 void *dummy; 440 { 441 char *path; 442 dev_t dev; 443 444 path = malloc(MNAMELEN, M_TEMP, M_WAITOK); 445 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { 446 dev = getdiskbyname(path); 447 if (dev != NODEV) 448 dumpdev = dev; 449 } 450 free(path, M_TEMP); 451 if (setdumpdev(dumpdev) != 0) 452 dumpdev = NODEV; 453 } 454 455 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) 456 457 static int 458 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 459 { 460 int error; 461 udev_t ndumpdev; 462 463 ndumpdev = dev2udev(dumpdev); 464 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 465 if (error == 0 && req->newptr != NULL) 466 error = setdumpdev(udev2dev(ndumpdev, 0)); 467 return (error); 468 } 469 470 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 471 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", ""); 472 473 /* 474 * Doadump comes here after turning off memory management and 475 * getting on the dump stack, either when called above, or by 476 * the auto-restart code. 477 */ 478 static void 479 dumpsys(void) 480 { 481 int error; 482 483 savectx(&dumppcb); 484 if (dumping++) { 485 printf("Dump already in progress, bailing...\n"); 486 return; 487 } 488 if (!dodump) 489 return; 490 if (dumpdev == NODEV) 491 return; 492 dumpsize = Maxmem; 493 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo); 494 printf("dump "); 495 error = dev_ddump(dumpdev); 496 if (error == 0) { 497 printf("succeeded\n"); 498 return; 499 } 500 printf("failed, reason: "); 501 switch (error) { 502 case ENOSYS: 503 case ENODEV: 504 printf("device doesn't support a dump routine\n"); 505 break; 506 507 case ENXIO: 508 printf("device bad\n"); 509 break; 510 511 case EFAULT: 512 printf("device not ready\n"); 513 break; 514 515 case EINVAL: 516 printf("area improper\n"); 517 break; 518 519 case EIO: 520 printf("i/o error\n"); 521 break; 522 523 case EINTR: 524 printf("aborted from console\n"); 525 break; 526 527 default: 528 printf("unknown, error = %d\n", error); 529 break; 530 } 531 } 532 533 int 534 dumpstatus(vm_offset_t addr, off_t count) 535 { 536 int c; 537 538 if (addr % (1024 * 1024) == 0) { 539 #ifdef HW_WDOG 540 if (wdog_tickler) 541 (*wdog_tickler)(); 542 #endif 543 printf("%ld ", (long)(count / (1024 * 1024))); 544 } 545 546 if ((c = cncheckc()) == 0x03) 547 return -1; 548 else if (c != -1) 549 printf("[CTRL-C to abort] "); 550 551 return 0; 552 } 553 554 /* 555 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 556 * and then reboots. If we are called twice, then we avoid trying to sync 557 * the disks as this often leads to recursive panics. 558 */ 559 void 560 panic(const char *fmt, ...) 561 { 562 int bootopt; 563 va_list ap; 564 static char buf[256]; 565 566 bootopt = RB_AUTOBOOT | RB_DUMP; 567 if (panicstr) 568 bootopt |= RB_NOSYNC; 569 else 570 panicstr = fmt; 571 572 va_start(ap, fmt); 573 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 574 if (panicstr == fmt) 575 panicstr = buf; 576 va_end(ap); 577 printf("panic: %s\n", buf); 578 #ifdef SMP 579 /* three seperate prints in case of an unmapped page and trap */ 580 printf("mp_lock = %08x; ", mp_lock); 581 printf("cpuid = %d; ", mycpu->gd_cpuid); 582 printf("lapic.id = %08x\n", lapic.id); 583 #endif 584 585 #if defined(DDB) 586 if (debugger_on_panic) 587 Debugger ("panic"); 588 #endif 589 boot(bootopt); 590 } 591 592 /* 593 * Support for poweroff delay. 594 */ 595 #ifndef POWEROFF_DELAY 596 # define POWEROFF_DELAY 5000 597 #endif 598 static int poweroff_delay = POWEROFF_DELAY; 599 600 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 601 &poweroff_delay, 0, ""); 602 603 static void 604 poweroff_wait(void *junk, int howto) 605 { 606 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 607 return; 608 DELAY(poweroff_delay * 1000); 609 } 610 611 /* 612 * Some system processes (e.g. syncer) need to be stopped at appropriate 613 * points in their main loops prior to a system shutdown, so that they 614 * won't interfere with the shutdown process (e.g. by holding a disk buf 615 * to cause sync to fail). For each of these system processes, register 616 * shutdown_kproc() as a handler for one of shutdown events. 617 */ 618 static int kproc_shutdown_wait = 60; 619 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 620 &kproc_shutdown_wait, 0, ""); 621 622 void 623 shutdown_kproc(void *arg, int howto) 624 { 625 struct thread *td; 626 struct proc *p; 627 int error; 628 629 if (panicstr) 630 return; 631 632 td = (struct thread *)arg; 633 if ((p = td->td_proc) != NULL) { 634 printf("Waiting (max %d seconds) for system process `%s' to stop...", 635 kproc_shutdown_wait, p->p_comm); 636 } else { 637 printf("Waiting (max %d seconds) for system thread %s to stop...", 638 kproc_shutdown_wait, td->td_comm); 639 } 640 error = suspend_kproc(td, kproc_shutdown_wait * hz); 641 642 if (error == EWOULDBLOCK) 643 printf("timed out\n"); 644 else 645 printf("stopped\n"); 646 } 647