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 1.17 2004/06/24 16:49:13 drhodus 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_mask, 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. We use the machine version to stay 83 * within the confines of the kernel header files. 84 */ 85 #include <machine/stdarg.h> 86 87 #ifdef DDB 88 #ifdef DDB_UNATTENDED 89 int debugger_on_panic = 0; 90 #else 91 int debugger_on_panic = 1; 92 #endif 93 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 94 &debugger_on_panic, 0, "Run debugger on kernel panic"); 95 96 extern void db_print_backtrace(void); 97 98 #ifdef DDB_TRACE 99 int trace_on_panic = 1; 100 #else 101 int trace_on_panic = 0; 102 #endif 103 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 104 &trace_on_panic, 0, "Print stack trace on kernel panic"); 105 #endif 106 107 static int sync_on_panic = 1; 108 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 109 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 110 111 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 112 113 #ifdef HW_WDOG 114 /* 115 * If there is a hardware watchdog, point this at the function needed to 116 * hold it off. 117 * It's needed when the kernel needs to do some lengthy operations. 118 * e.g. in wd.c when dumping core.. It's most annoying to have 119 * your precious core-dump only half written because the wdog kicked in. 120 */ 121 watchdog_tickle_fn wdog_tickler = NULL; 122 #endif /* HW_WDOG */ 123 124 /* 125 * Variable panicstr contains argument to first call to panic; used as flag 126 * to indicate that the kernel has already called panic. 127 */ 128 const char *panicstr; 129 130 int dumping; /* system is dumping */ 131 132 static void boot (int) __dead2; 133 static void dumpsys (void); 134 static int setdumpdev (dev_t dev); 135 static void poweroff_wait (void *, int); 136 static void print_uptime (void); 137 static void shutdown_halt (void *junk, int howto); 138 static void shutdown_panic (void *junk, int howto); 139 static void shutdown_reset (void *junk, int howto); 140 141 /* register various local shutdown events */ 142 static void 143 shutdown_conf(void *unused) 144 { 145 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 146 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 147 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 148 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 149 } 150 151 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 152 153 /* ARGSUSED */ 154 155 /* 156 * The system call that results in a reboot 157 */ 158 int 159 reboot(struct reboot_args *uap) 160 { 161 struct thread *td = curthread; 162 int error; 163 164 if ((error = suser(td))) 165 return (error); 166 167 boot(uap->opt); 168 return (0); 169 } 170 171 /* 172 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 173 */ 174 static int shutdown_howto = 0; 175 176 void 177 shutdown_nice(int howto) 178 { 179 shutdown_howto = howto; 180 181 /* Send a signal to init(8) and have it shutdown the world */ 182 if (initproc != NULL) { 183 psignal(initproc, SIGINT); 184 } else { 185 /* No init(8) running, so simply reboot */ 186 boot(RB_NOSYNC); 187 } 188 return; 189 } 190 static int waittime = -1; 191 static struct pcb dumppcb; 192 193 static void 194 print_uptime() 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", ts.tv_sec / 86400); 204 ts.tv_sec %= 86400; 205 f = 1; 206 } 207 if (f || ts.tv_sec >= 3600) { 208 printf("%ldh", ts.tv_sec / 3600); 209 ts.tv_sec %= 3600; 210 f = 1; 211 } 212 if (f || ts.tv_sec >= 60) { 213 printf("%ldm", ts.tv_sec / 60); 214 ts.tv_sec %= 60; 215 f = 1; 216 } 217 printf("%lds\n", ts.tv_sec); 218 } 219 220 /* 221 * Go through the rigmarole of shutting down.. 222 * this used to be in machdep.c but I'll be dammned if I could see 223 * anything machine dependant in it. 224 */ 225 static void 226 boot(int howto) 227 { 228 229 /* collect extra flags that shutdown_nice might have set */ 230 howto |= shutdown_howto; 231 232 #ifdef SMP 233 if (smp_active_mask > 1) { 234 printf("boot() called on cpu#%d\n", mycpu->gd_cpuid); 235 } 236 #endif 237 /* 238 * Do any callouts that should be done BEFORE syncing the filesystems. 239 */ 240 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 241 242 /* 243 * Now sync filesystems 244 */ 245 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 246 struct buf *bp; 247 int iter, nbusy, pbusy; 248 249 waittime = 0; 250 printf("\nsyncing disks... "); 251 252 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 253 254 /* 255 * With soft updates, some buffers that are 256 * written will be remarked as dirty until other 257 * buffers are written. 258 */ 259 for (iter = pbusy = 0; iter < 20; iter++) { 260 nbusy = 0; 261 for (bp = &buf[nbuf]; --bp >= buf; ) { 262 if ((bp->b_flags & B_INVAL) == 0 && 263 BUF_REFCNT(bp) > 0) { 264 nbusy++; 265 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 266 == B_DELWRI) { 267 /* bawrite(bp);*/ 268 nbusy++; 269 } 270 } 271 if (nbusy == 0) 272 break; 273 printf("%d ", nbusy); 274 if (nbusy < pbusy) 275 iter = 0; 276 pbusy = nbusy; 277 if (iter > 5 && bioops.io_sync) 278 (*bioops.io_sync)(NULL); 279 sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 280 DELAY(50000 * iter); 281 } 282 printf("\n"); 283 /* 284 * Count only busy local buffers to prevent forcing 285 * a fsck if we're just a client of a wedged NFS server 286 */ 287 nbusy = 0; 288 for (bp = &buf[nbuf]; --bp >= buf; ) { 289 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 290 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 291 if (bp->b_dev == NODEV) { 292 TAILQ_REMOVE(&mountlist, 293 bp->b_vp->v_mount, mnt_list); 294 continue; 295 } 296 nbusy++; 297 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 298 printf( 299 "%p %d: dev:%s, flags:%08lx, blkno:%ld, lblkno:%ld\n", 300 bp, nbusy, devtoname(bp->b_dev), 301 bp->b_flags, (long)bp->b_blkno, 302 (long)bp->b_lblkno); 303 #endif 304 } 305 } 306 if (nbusy) { 307 /* 308 * Failed to sync all blocks. Indicate this and don't 309 * unmount filesystems (thus forcing an fsck on reboot). 310 */ 311 printf("giving up on %d buffers\n", nbusy); 312 #ifdef DDB 313 Debugger("busy buffer problem"); 314 #endif /* DDB */ 315 DELAY(5000000); /* 5 seconds */ 316 } else { 317 printf("done\n"); 318 /* 319 * Unmount filesystems 320 */ 321 if (panicstr == 0) 322 vfs_unmountall(); 323 } 324 DELAY(100000); /* wait for console output to finish */ 325 } 326 327 print_uptime(); 328 329 /* 330 * Ok, now do things that assume all filesystem activity has 331 * been completed. 332 */ 333 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 334 splhigh(); 335 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) 336 dumpsys(); 337 338 /* Now that we're going to really halt the system... */ 339 EVENTHANDLER_INVOKE(shutdown_final, howto); 340 341 for(;;) ; /* safety against shutdown_reset not working */ 342 /* NOTREACHED */ 343 } 344 345 /* 346 * If the shutdown was a clean halt, behave accordingly. 347 */ 348 static void 349 shutdown_halt(void *junk, int howto) 350 { 351 if (howto & RB_HALT) { 352 printf("\n"); 353 printf("The operating system has halted.\n"); 354 printf("Please press any key to reboot.\n\n"); 355 switch (cngetc()) { 356 case -1: /* No console, just die */ 357 cpu_halt(); 358 /* NOTREACHED */ 359 default: 360 howto &= ~RB_HALT; 361 break; 362 } 363 } 364 } 365 366 /* 367 * Check to see if the system paniced, pause and then reboot 368 * according to the specified delay. 369 */ 370 static void 371 shutdown_panic(void *junk, int howto) 372 { 373 int loop; 374 375 if (howto & RB_DUMP) { 376 if (PANIC_REBOOT_WAIT_TIME != 0) { 377 if (PANIC_REBOOT_WAIT_TIME != -1) { 378 printf("Automatic reboot in %d seconds - " 379 "press a key on the console to abort\n", 380 PANIC_REBOOT_WAIT_TIME); 381 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 382 loop > 0; --loop) { 383 DELAY(1000 * 100); /* 1/10th second */ 384 /* Did user type a key? */ 385 if (cncheckc() != -1) 386 break; 387 } 388 if (!loop) 389 return; 390 } 391 } else { /* zero time specified - reboot NOW */ 392 return; 393 } 394 printf("--> Press a key on the console to reboot,\n"); 395 printf("--> or switch off the system now.\n"); 396 cngetc(); 397 } 398 } 399 400 /* 401 * Everything done, now reset 402 */ 403 static void 404 shutdown_reset(void *junk, int howto) 405 { 406 printf("Rebooting...\n"); 407 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 408 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 409 cpu_reset(); 410 /* NOTREACHED */ /* assuming reset worked */ 411 } 412 413 /* 414 * Magic number for savecore 415 * 416 * exported (symorder) and used at least by savecore(8) 417 * 418 */ 419 static u_long const dumpmag = 0x8fca0101UL; 420 421 static int dumpsize = 0; /* also for savecore */ 422 423 static int dodump = 1; 424 425 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 426 "Try to perform coredump on kernel panic"); 427 428 static int 429 setdumpdev(dev) 430 dev_t dev; 431 { 432 int psize; 433 long newdumplo; 434 435 if (dev == NODEV) { 436 dumpdev = dev; 437 return (0); 438 } 439 psize = dev_dpsize(dev); 440 if (psize == -1) 441 return (ENXIO); 442 /* 443 * XXX should clean up checking in dumpsys() to be more like this. 444 */ 445 newdumplo = psize - Maxmem * (PAGE_SIZE / DEV_BSIZE); 446 if (newdumplo <= LABELSECTOR) 447 return (ENOSPC); 448 dumpdev = dev; 449 dumplo = newdumplo; 450 return (0); 451 } 452 453 454 /* ARGSUSED */ 455 static void dump_conf (void *dummy); 456 static void 457 dump_conf(dummy) 458 void *dummy; 459 { 460 char *path; 461 dev_t dev; 462 463 path = malloc(MNAMELEN, M_TEMP, M_WAITOK); 464 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { 465 dev = getdiskbyname(path); 466 if (dev != NODEV) 467 dumpdev = dev; 468 } 469 free(path, M_TEMP); 470 if (setdumpdev(dumpdev) != 0) 471 dumpdev = NODEV; 472 } 473 474 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) 475 476 static int 477 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 478 { 479 int error; 480 udev_t ndumpdev; 481 482 ndumpdev = dev2udev(dumpdev); 483 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 484 if (error == 0 && req->newptr != NULL) 485 error = setdumpdev(udev2dev(ndumpdev, 0)); 486 return (error); 487 } 488 489 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 490 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", ""); 491 492 /* 493 * Doadump comes here after turning off memory management and 494 * getting on the dump stack, either when called above, or by 495 * the auto-restart code. 496 */ 497 static void 498 dumpsys(void) 499 { 500 int error; 501 502 savectx(&dumppcb); 503 if (dumping++) { 504 printf("Dump already in progress, bailing...\n"); 505 return; 506 } 507 if (!dodump) 508 return; 509 if (dumpdev == NODEV) 510 return; 511 dumpsize = Maxmem; 512 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo); 513 printf("dump "); 514 error = dev_ddump(dumpdev); 515 if (error == 0) { 516 printf("succeeded\n"); 517 return; 518 } 519 printf("failed, reason: "); 520 switch (error) { 521 case ENOSYS: 522 case ENODEV: 523 printf("device doesn't support a dump routine\n"); 524 break; 525 526 case ENXIO: 527 printf("device bad\n"); 528 break; 529 530 case EFAULT: 531 printf("device not ready\n"); 532 break; 533 534 case EINVAL: 535 printf("area improper\n"); 536 break; 537 538 case EIO: 539 printf("i/o error\n"); 540 break; 541 542 case EINTR: 543 printf("aborted from console\n"); 544 break; 545 546 default: 547 printf("unknown, error = %d\n", error); 548 break; 549 } 550 } 551 552 int 553 dumpstatus(vm_offset_t addr, off_t count) 554 { 555 int c; 556 557 if (addr % (1024 * 1024) == 0) { 558 #ifdef HW_WDOG 559 if (wdog_tickler) 560 (*wdog_tickler)(); 561 #endif 562 printf("%ld ", (long)(count / (1024 * 1024))); 563 } 564 565 if ((c = cncheckc()) == 0x03) 566 return -1; 567 else if (c != -1) 568 printf("[CTRL-C to abort] "); 569 570 return 0; 571 } 572 573 /* 574 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 575 * and then reboots. If we are called twice, then we avoid trying to sync 576 * the disks as this often leads to recursive panics. 577 */ 578 void 579 panic(const char *fmt, ...) 580 { 581 int bootopt, newpanic; 582 __va_list ap; 583 static char buf[256]; 584 585 bootopt = RB_AUTOBOOT | RB_DUMP; 586 if (sync_on_panic == 0) 587 bootopt |= RB_NOSYNC; 588 newpanic = 0; 589 if (panicstr) 590 bootopt |= RB_NOSYNC; 591 else { 592 panicstr = fmt; 593 newpanic = 1; 594 } 595 596 __va_start(ap, fmt); 597 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 598 if (panicstr == fmt) 599 panicstr = buf; 600 __va_end(ap); 601 printf("panic: %s\n", buf); 602 #ifdef SMP 603 /* three separate prints in case of an unmapped page and trap */ 604 printf("mp_lock = %08x; ", mp_lock); 605 printf("cpuid = %d; ", mycpu->gd_cpuid); 606 printf("lapic.id = %08x\n", lapic.id); 607 #endif 608 609 #if defined(DDB) 610 if (newpanic && trace_on_panic) 611 db_print_backtrace(); 612 if (debugger_on_panic) 613 Debugger ("panic"); 614 #endif 615 boot(bootopt); 616 } 617 618 /* 619 * Support for poweroff delay. 620 */ 621 #ifndef POWEROFF_DELAY 622 # define POWEROFF_DELAY 5000 623 #endif 624 static int poweroff_delay = POWEROFF_DELAY; 625 626 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 627 &poweroff_delay, 0, ""); 628 629 static void 630 poweroff_wait(void *junk, int howto) 631 { 632 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 633 return; 634 DELAY(poweroff_delay * 1000); 635 } 636 637 /* 638 * Some system processes (e.g. syncer) need to be stopped at appropriate 639 * points in their main loops prior to a system shutdown, so that they 640 * won't interfere with the shutdown process (e.g. by holding a disk buf 641 * to cause sync to fail). For each of these system processes, register 642 * shutdown_kproc() as a handler for one of shutdown events. 643 */ 644 static int kproc_shutdown_wait = 60; 645 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 646 &kproc_shutdown_wait, 0, ""); 647 648 void 649 shutdown_kproc(void *arg, int howto) 650 { 651 struct thread *td; 652 struct proc *p; 653 int error; 654 655 if (panicstr) 656 return; 657 658 td = (struct thread *)arg; 659 if ((p = td->td_proc) != NULL) { 660 printf("Waiting (max %d seconds) for system process `%s' to stop...", 661 kproc_shutdown_wait, p->p_comm); 662 } else { 663 printf("Waiting (max %d seconds) for system thread %s to stop...", 664 kproc_shutdown_wait, td->td_comm); 665 } 666 error = suspend_kproc(td, kproc_shutdown_wait * hz); 667 668 if (error == EWOULDBLOCK) 669 printf("timed out\n"); 670 else 671 printf("stopped\n"); 672 } 673