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 1.12 2003/09/05 17:46:54 hsu Exp $ 41 */ 42 43 #include "opt_ddb.h" 44 #include "opt_ddb_trace.h" 45 #include "opt_hw_wdog.h" 46 #include "opt_panic.h" 47 #include "opt_show_busybufs.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/eventhandler.h> 52 #include <sys/buf.h> 53 #include <sys/disklabel.h> 54 #include <sys/reboot.h> 55 #include <sys/proc.h> 56 #include <sys/vnode.h> 57 #include <sys/kernel.h> 58 #include <sys/kthread.h> 59 #include <sys/malloc.h> 60 #include <sys/mount.h> 61 #include <sys/queue.h> 62 #include <sys/sysctl.h> 63 #include <sys/conf.h> 64 #include <sys/sysproto.h> 65 #include <sys/device.h> 66 #include <sys/cons.h> 67 #include <sys/buf2.h> 68 69 #include <machine/pcb.h> 70 #include <machine/clock.h> 71 #include <machine/md_var.h> 72 #include <machine/smp.h> /* smp_active, cpuid */ 73 74 #include <sys/signalvar.h> 75 76 #ifndef PANIC_REBOOT_WAIT_TIME 77 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 78 #endif 79 80 /* 81 * Note that stdarg.h and the ANSI style va_start macro is used for both 82 * ANSI and traditional C compilers. 83 */ 84 #include <machine/stdarg.h> 85 86 #ifdef DDB 87 #ifdef DDB_UNATTENDED 88 int debugger_on_panic = 0; 89 #else 90 int debugger_on_panic = 1; 91 #endif 92 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 93 &debugger_on_panic, 0, "Run debugger on kernel panic"); 94 95 extern void db_print_backtrace(void); 96 97 #ifdef DDB_TRACE 98 int trace_on_panic = 1; 99 #else 100 int trace_on_panic = 0; 101 #endif 102 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 103 &trace_on_panic, 0, "Print stack trace on kernel panic"); 104 #endif 105 106 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 107 108 #ifdef HW_WDOG 109 /* 110 * If there is a hardware watchdog, point this at the function needed to 111 * hold it off. 112 * It's needed when the kernel needs to do some lengthy operations. 113 * e.g. in wd.c when dumping core.. It's most annoying to have 114 * your precious core-dump only half written because the wdog kicked in. 115 */ 116 watchdog_tickle_fn wdog_tickler = NULL; 117 #endif /* HW_WDOG */ 118 119 /* 120 * Variable panicstr contains argument to first call to panic; used as flag 121 * to indicate that the kernel has already called panic. 122 */ 123 const char *panicstr; 124 125 int dumping; /* system is dumping */ 126 127 static void boot (int) __dead2; 128 static void dumpsys (void); 129 static int setdumpdev (dev_t dev); 130 static void poweroff_wait (void *, int); 131 static void print_uptime (void); 132 static void shutdown_halt (void *junk, int howto); 133 static void shutdown_panic (void *junk, int howto); 134 static void shutdown_reset (void *junk, int howto); 135 136 /* register various local shutdown events */ 137 static void 138 shutdown_conf(void *unused) 139 { 140 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 141 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 142 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 143 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 144 } 145 146 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 147 148 /* ARGSUSED */ 149 150 /* 151 * The system call that results in a reboot 152 */ 153 int 154 reboot(struct reboot_args *uap) 155 { 156 struct thread *td = curthread; 157 int error; 158 159 if ((error = suser(td))) 160 return (error); 161 162 boot(uap->opt); 163 return (0); 164 } 165 166 /* 167 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 168 */ 169 static int shutdown_howto = 0; 170 171 void 172 shutdown_nice(int howto) 173 { 174 shutdown_howto = howto; 175 176 /* Send a signal to init(8) and have it shutdown the world */ 177 if (initproc != NULL) { 178 psignal(initproc, SIGINT); 179 } else { 180 /* No init(8) running, so simply reboot */ 181 boot(RB_NOSYNC); 182 } 183 return; 184 } 185 static int waittime = -1; 186 static struct pcb dumppcb; 187 188 static void 189 print_uptime() 190 { 191 int f; 192 struct timespec ts; 193 194 getnanouptime(&ts); 195 printf("Uptime: "); 196 f = 0; 197 if (ts.tv_sec >= 86400) { 198 printf("%ldd", ts.tv_sec / 86400); 199 ts.tv_sec %= 86400; 200 f = 1; 201 } 202 if (f || ts.tv_sec >= 3600) { 203 printf("%ldh", ts.tv_sec / 3600); 204 ts.tv_sec %= 3600; 205 f = 1; 206 } 207 if (f || ts.tv_sec >= 60) { 208 printf("%ldm", ts.tv_sec / 60); 209 ts.tv_sec %= 60; 210 f = 1; 211 } 212 printf("%lds\n", ts.tv_sec); 213 } 214 215 /* 216 * Go through the rigmarole of shutting down.. 217 * this used to be in machdep.c but I'll be dammned if I could see 218 * anything machine dependant in it. 219 */ 220 static void 221 boot(int howto) 222 { 223 224 /* collect extra flags that shutdown_nice might have set */ 225 howto |= shutdown_howto; 226 227 #ifdef SMP 228 if (smp_active) { 229 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid); 230 } 231 #endif 232 /* 233 * Do any callouts that should be done BEFORE syncing the filesystems. 234 */ 235 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 236 237 /* 238 * Now sync filesystems 239 */ 240 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 241 struct buf *bp; 242 int iter, nbusy, pbusy; 243 244 waittime = 0; 245 printf("\nsyncing disks... "); 246 247 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 248 249 /* 250 * With soft updates, some buffers that are 251 * written will be remarked as dirty until other 252 * buffers are written. 253 */ 254 for (iter = pbusy = 0; iter < 20; iter++) { 255 nbusy = 0; 256 for (bp = &buf[nbuf]; --bp >= buf; ) { 257 if ((bp->b_flags & B_INVAL) == 0 && 258 BUF_REFCNT(bp) > 0) { 259 nbusy++; 260 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 261 == B_DELWRI) { 262 /* bawrite(bp);*/ 263 nbusy++; 264 } 265 } 266 if (nbusy == 0) 267 break; 268 printf("%d ", nbusy); 269 if (nbusy < pbusy) 270 iter = 0; 271 pbusy = nbusy; 272 if (iter > 5 && bioops.io_sync) 273 (*bioops.io_sync)(NULL); 274 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 275 DELAY(50000 * iter); 276 } 277 printf("\n"); 278 /* 279 * Count only busy local buffers to prevent forcing 280 * a fsck if we're just a client of a wedged NFS server 281 */ 282 nbusy = 0; 283 for (bp = &buf[nbuf]; --bp >= buf; ) { 284 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 285 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 286 if (bp->b_dev == NODEV) { 287 TAILQ_REMOVE(&mountlist, 288 bp->b_vp->v_mount, mnt_list); 289 continue; 290 } 291 nbusy++; 292 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 293 printf( 294 "%p %d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n", 295 bp, nbusy, devtoname(bp->b_dev), 296 bp->b_flags, (long)bp->b_blkno, 297 (long)bp->b_lblkno); 298 #endif 299 } 300 } 301 if (nbusy) { 302 /* 303 * Failed to sync all blocks. Indicate this and don't 304 * unmount filesystems (thus forcing an fsck on reboot). 305 */ 306 printf("giving up on %d buffers\n", nbusy); 307 DELAY(5000000); /* 5 seconds */ 308 } else { 309 printf("done\n"); 310 /* 311 * Unmount filesystems 312 */ 313 if (panicstr == 0) 314 vfs_unmountall(); 315 } 316 DELAY(100000); /* wait for console output to finish */ 317 } 318 319 print_uptime(); 320 321 /* 322 * Ok, now do things that assume all filesystem activity has 323 * been completed. 324 */ 325 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 326 splhigh(); 327 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) 328 dumpsys(); 329 330 /* Now that we're going to really halt the system... */ 331 EVENTHANDLER_INVOKE(shutdown_final, howto); 332 333 for(;;) ; /* safety against shutdown_reset not working */ 334 /* NOTREACHED */ 335 } 336 337 /* 338 * If the shutdown was a clean halt, behave accordingly. 339 */ 340 static void 341 shutdown_halt(void *junk, int howto) 342 { 343 if (howto & RB_HALT) { 344 printf("\n"); 345 printf("The operating system has halted.\n"); 346 printf("Please press any key to reboot.\n\n"); 347 switch (cngetc()) { 348 case -1: /* No console, just die */ 349 cpu_halt(); 350 /* NOTREACHED */ 351 default: 352 howto &= ~RB_HALT; 353 break; 354 } 355 } 356 } 357 358 /* 359 * Check to see if the system paniced, pause and then reboot 360 * according to the specified delay. 361 */ 362 static void 363 shutdown_panic(void *junk, int howto) 364 { 365 int loop; 366 367 if (howto & RB_DUMP) { 368 if (PANIC_REBOOT_WAIT_TIME != 0) { 369 if (PANIC_REBOOT_WAIT_TIME != -1) { 370 printf("Automatic reboot in %d seconds - " 371 "press a key on the console to abort\n", 372 PANIC_REBOOT_WAIT_TIME); 373 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 374 loop > 0; --loop) { 375 DELAY(1000 * 100); /* 1/10th second */ 376 /* Did user type a key? */ 377 if (cncheckc() != -1) 378 break; 379 } 380 if (!loop) 381 return; 382 } 383 } else { /* zero time specified - reboot NOW */ 384 return; 385 } 386 printf("--> Press a key on the console to reboot,\n"); 387 printf("--> or switch off the system now.\n"); 388 cngetc(); 389 } 390 } 391 392 /* 393 * Everything done, now reset 394 */ 395 static void 396 shutdown_reset(void *junk, int howto) 397 { 398 printf("Rebooting...\n"); 399 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 400 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 401 cpu_reset(); 402 /* NOTREACHED */ /* assuming reset worked */ 403 } 404 405 /* 406 * Magic number for savecore 407 * 408 * exported (symorder) and used at least by savecore(8) 409 * 410 */ 411 static u_long const dumpmag = 0x8fca0101UL; 412 413 static int dumpsize = 0; /* also for savecore */ 414 415 static int dodump = 1; 416 417 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 418 "Try to perform coredump on kernel panic"); 419 420 static int 421 setdumpdev(dev) 422 dev_t dev; 423 { 424 int psize; 425 long newdumplo; 426 427 if (dev == NODEV) { 428 dumpdev = dev; 429 return (0); 430 } 431 psize = dev_dpsize(dev); 432 if (psize == -1) 433 return (ENXIO); 434 /* 435 * XXX should clean up checking in dumpsys() to be more like this. 436 */ 437 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE); 438 if (newdumplo <= LABELSECTOR) 439 return (ENOSPC); 440 dumpdev = dev; 441 dumplo = newdumplo; 442 return (0); 443 } 444 445 446 /* ARGSUSED */ 447 static void dump_conf (void *dummy); 448 static void 449 dump_conf(dummy) 450 void *dummy; 451 { 452 char *path; 453 dev_t dev; 454 455 path = malloc(MNAMELEN, M_TEMP, M_WAITOK); 456 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { 457 dev = getdiskbyname(path); 458 if (dev != NODEV) 459 dumpdev = dev; 460 } 461 free(path, M_TEMP); 462 if (setdumpdev(dumpdev) != 0) 463 dumpdev = NODEV; 464 } 465 466 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) 467 468 static int 469 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 470 { 471 int error; 472 udev_t ndumpdev; 473 474 ndumpdev = dev2udev(dumpdev); 475 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 476 if (error == 0 && req->newptr != NULL) 477 error = setdumpdev(udev2dev(ndumpdev, 0)); 478 return (error); 479 } 480 481 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 482 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", ""); 483 484 /* 485 * Doadump comes here after turning off memory management and 486 * getting on the dump stack, either when called above, or by 487 * the auto-restart code. 488 */ 489 static void 490 dumpsys(void) 491 { 492 int error; 493 494 savectx(&dumppcb); 495 if (dumping++) { 496 printf("Dump already in progress, bailing...\n"); 497 return; 498 } 499 if (!dodump) 500 return; 501 if (dumpdev == NODEV) 502 return; 503 dumpsize = Maxmem; 504 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo); 505 printf("dump "); 506 error = dev_ddump(dumpdev); 507 if (error == 0) { 508 printf("succeeded\n"); 509 return; 510 } 511 printf("failed, reason: "); 512 switch (error) { 513 case ENOSYS: 514 case ENODEV: 515 printf("device doesn't support a dump routine\n"); 516 break; 517 518 case ENXIO: 519 printf("device bad\n"); 520 break; 521 522 case EFAULT: 523 printf("device not ready\n"); 524 break; 525 526 case EINVAL: 527 printf("area improper\n"); 528 break; 529 530 case EIO: 531 printf("i/o error\n"); 532 break; 533 534 case EINTR: 535 printf("aborted from console\n"); 536 break; 537 538 default: 539 printf("unknown, error = %d\n", error); 540 break; 541 } 542 } 543 544 int 545 dumpstatus(vm_offset_t addr, off_t count) 546 { 547 int c; 548 549 if (addr % (1024 * 1024) == 0) { 550 #ifdef HW_WDOG 551 if (wdog_tickler) 552 (*wdog_tickler)(); 553 #endif 554 printf("%ld ", (long)(count / (1024 * 1024))); 555 } 556 557 if ((c = cncheckc()) == 0x03) 558 return -1; 559 else if (c != -1) 560 printf("[CTRL-C to abort] "); 561 562 return 0; 563 } 564 565 /* 566 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 567 * and then reboots. If we are called twice, then we avoid trying to sync 568 * the disks as this often leads to recursive panics. 569 */ 570 void 571 panic(const char *fmt, ...) 572 { 573 int bootopt, newpanic; 574 va_list ap; 575 static char buf[256]; 576 577 bootopt = RB_AUTOBOOT | RB_DUMP; 578 newpanic = 0; 579 if (panicstr) 580 bootopt |= RB_NOSYNC; 581 else { 582 panicstr = fmt; 583 newpanic = 1; 584 } 585 586 va_start(ap, fmt); 587 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 588 if (panicstr == fmt) 589 panicstr = buf; 590 va_end(ap); 591 printf("panic: %s\n", buf); 592 #ifdef SMP 593 /* three seperate prints in case of an unmapped page and trap */ 594 printf("mp_lock = %08x; ", mp_lock); 595 printf("cpuid = %d; ", mycpu->gd_cpuid); 596 printf("lapic.id = %08x\n", lapic.id); 597 #endif 598 599 #if defined(DDB) 600 if (newpanic && trace_on_panic) 601 db_print_backtrace(); 602 if (debugger_on_panic) 603 Debugger ("panic"); 604 #endif 605 boot(bootopt); 606 } 607 608 /* 609 * Support for poweroff delay. 610 */ 611 #ifndef POWEROFF_DELAY 612 # define POWEROFF_DELAY 5000 613 #endif 614 static int poweroff_delay = POWEROFF_DELAY; 615 616 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 617 &poweroff_delay, 0, ""); 618 619 static void 620 poweroff_wait(void *junk, int howto) 621 { 622 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 623 return; 624 DELAY(poweroff_delay * 1000); 625 } 626 627 /* 628 * Some system processes (e.g. syncer) need to be stopped at appropriate 629 * points in their main loops prior to a system shutdown, so that they 630 * won't interfere with the shutdown process (e.g. by holding a disk buf 631 * to cause sync to fail). For each of these system processes, register 632 * shutdown_kproc() as a handler for one of shutdown events. 633 */ 634 static int kproc_shutdown_wait = 60; 635 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 636 &kproc_shutdown_wait, 0, ""); 637 638 void 639 shutdown_kproc(void *arg, int howto) 640 { 641 struct thread *td; 642 struct proc *p; 643 int error; 644 645 if (panicstr) 646 return; 647 648 td = (struct thread *)arg; 649 if ((p = td->td_proc) != NULL) { 650 printf("Waiting (max %d seconds) for system process `%s' to stop...", 651 kproc_shutdown_wait, p->p_comm); 652 } else { 653 printf("Waiting (max %d seconds) for system thread %s to stop...", 654 kproc_shutdown_wait, td->td_comm); 655 } 656 error = suspend_kproc(td, kproc_shutdown_wait * hz); 657 658 if (error == EWOULDBLOCK) 659 printf("timed out\n"); 660 else 661 printf("stopped\n"); 662 } 663