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 1.14 2003/11/16 02:41:46 dillon 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. 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) { 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 DELAY(5000000); /* 5 seconds */ 313 } else { 314 printf("done\n"); 315 /* 316 * Unmount filesystems 317 */ 318 if (panicstr == 0) 319 vfs_unmountall(); 320 } 321 DELAY(100000); /* wait for console output to finish */ 322 } 323 324 print_uptime(); 325 326 /* 327 * Ok, now do things that assume all filesystem activity has 328 * been completed. 329 */ 330 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 331 splhigh(); 332 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) 333 dumpsys(); 334 335 /* Now that we're going to really halt the system... */ 336 EVENTHANDLER_INVOKE(shutdown_final, howto); 337 338 for(;;) ; /* safety against shutdown_reset not working */ 339 /* NOTREACHED */ 340 } 341 342 /* 343 * If the shutdown was a clean halt, behave accordingly. 344 */ 345 static void 346 shutdown_halt(void *junk, int howto) 347 { 348 if (howto & RB_HALT) { 349 printf("\n"); 350 printf("The operating system has halted.\n"); 351 printf("Please press any key to reboot.\n\n"); 352 switch (cngetc()) { 353 case -1: /* No console, just die */ 354 cpu_halt(); 355 /* NOTREACHED */ 356 default: 357 howto &= ~RB_HALT; 358 break; 359 } 360 } 361 } 362 363 /* 364 * Check to see if the system paniced, pause and then reboot 365 * according to the specified delay. 366 */ 367 static void 368 shutdown_panic(void *junk, int howto) 369 { 370 int loop; 371 372 if (howto & RB_DUMP) { 373 if (PANIC_REBOOT_WAIT_TIME != 0) { 374 if (PANIC_REBOOT_WAIT_TIME != -1) { 375 printf("Automatic reboot in %d seconds - " 376 "press a key on the console to abort\n", 377 PANIC_REBOOT_WAIT_TIME); 378 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 379 loop > 0; --loop) { 380 DELAY(1000 * 100); /* 1/10th second */ 381 /* Did user type a key? */ 382 if (cncheckc() != -1) 383 break; 384 } 385 if (!loop) 386 return; 387 } 388 } else { /* zero time specified - reboot NOW */ 389 return; 390 } 391 printf("--> Press a key on the console to reboot,\n"); 392 printf("--> or switch off the system now.\n"); 393 cngetc(); 394 } 395 } 396 397 /* 398 * Everything done, now reset 399 */ 400 static void 401 shutdown_reset(void *junk, int howto) 402 { 403 printf("Rebooting...\n"); 404 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 405 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 406 cpu_reset(); 407 /* NOTREACHED */ /* assuming reset worked */ 408 } 409 410 /* 411 * Magic number for savecore 412 * 413 * exported (symorder) and used at least by savecore(8) 414 * 415 */ 416 static u_long const dumpmag = 0x8fca0101UL; 417 418 static int dumpsize = 0; /* also for savecore */ 419 420 static int dodump = 1; 421 422 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 423 "Try to perform coredump on kernel panic"); 424 425 static int 426 setdumpdev(dev) 427 dev_t dev; 428 { 429 int psize; 430 long newdumplo; 431 432 if (dev == NODEV) { 433 dumpdev = dev; 434 return (0); 435 } 436 psize = dev_dpsize(dev); 437 if (psize == -1) 438 return (ENXIO); 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 */ 452 static void dump_conf (void *dummy); 453 static void 454 dump_conf(dummy) 455 void *dummy; 456 { 457 char *path; 458 dev_t dev; 459 460 path = malloc(MNAMELEN, M_TEMP, M_WAITOK); 461 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { 462 dev = getdiskbyname(path); 463 if (dev != NODEV) 464 dumpdev = dev; 465 } 466 free(path, M_TEMP); 467 if (setdumpdev(dumpdev) != 0) 468 dumpdev = NODEV; 469 } 470 471 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) 472 473 static int 474 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 475 { 476 int error; 477 udev_t ndumpdev; 478 479 ndumpdev = dev2udev(dumpdev); 480 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 481 if (error == 0 && req->newptr != NULL) 482 error = setdumpdev(udev2dev(ndumpdev, 0)); 483 return (error); 484 } 485 486 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 487 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,dev_t", ""); 488 489 /* 490 * Doadump comes here after turning off memory management and 491 * getting on the dump stack, either when called above, or by 492 * the auto-restart code. 493 */ 494 static void 495 dumpsys(void) 496 { 497 int error; 498 499 savectx(&dumppcb); 500 if (dumping++) { 501 printf("Dump already in progress, bailing...\n"); 502 return; 503 } 504 if (!dodump) 505 return; 506 if (dumpdev == NODEV) 507 return; 508 dumpsize = Maxmem; 509 printf("\ndumping to dev %s, offset %ld\n", devtoname(dumpdev), dumplo); 510 printf("dump "); 511 error = dev_ddump(dumpdev); 512 if (error == 0) { 513 printf("succeeded\n"); 514 return; 515 } 516 printf("failed, reason: "); 517 switch (error) { 518 case ENOSYS: 519 case ENODEV: 520 printf("device doesn't support a dump routine\n"); 521 break; 522 523 case ENXIO: 524 printf("device bad\n"); 525 break; 526 527 case EFAULT: 528 printf("device not ready\n"); 529 break; 530 531 case EINVAL: 532 printf("area improper\n"); 533 break; 534 535 case EIO: 536 printf("i/o error\n"); 537 break; 538 539 case EINTR: 540 printf("aborted from console\n"); 541 break; 542 543 default: 544 printf("unknown, error = %d\n", error); 545 break; 546 } 547 } 548 549 int 550 dumpstatus(vm_offset_t addr, off_t count) 551 { 552 int c; 553 554 if (addr % (1024 * 1024) == 0) { 555 #ifdef HW_WDOG 556 if (wdog_tickler) 557 (*wdog_tickler)(); 558 #endif 559 printf("%ld ", (long)(count / (1024 * 1024))); 560 } 561 562 if ((c = cncheckc()) == 0x03) 563 return -1; 564 else if (c != -1) 565 printf("[CTRL-C to abort] "); 566 567 return 0; 568 } 569 570 /* 571 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 572 * and then reboots. If we are called twice, then we avoid trying to sync 573 * the disks as this often leads to recursive panics. 574 */ 575 void 576 panic(const char *fmt, ...) 577 { 578 int bootopt, newpanic; 579 __va_list ap; 580 static char buf[256]; 581 582 bootopt = RB_AUTOBOOT | RB_DUMP; 583 if (sync_on_panic == 0) 584 bootopt |= RB_NOSYNC; 585 newpanic = 0; 586 if (panicstr) 587 bootopt |= RB_NOSYNC; 588 else { 589 panicstr = fmt; 590 newpanic = 1; 591 } 592 593 __va_start(ap, fmt); 594 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 595 if (panicstr == fmt) 596 panicstr = buf; 597 __va_end(ap); 598 printf("panic: %s\n", buf); 599 #ifdef SMP 600 /* three seperate prints in case of an unmapped page and trap */ 601 printf("mp_lock = %08x; ", mp_lock); 602 printf("cpuid = %d; ", mycpu->gd_cpuid); 603 printf("lapic.id = %08x\n", lapic.id); 604 #endif 605 606 #if defined(DDB) 607 if (newpanic && trace_on_panic) 608 db_print_backtrace(); 609 if (debugger_on_panic) 610 Debugger ("panic"); 611 #endif 612 boot(bootopt); 613 } 614 615 /* 616 * Support for poweroff delay. 617 */ 618 #ifndef POWEROFF_DELAY 619 # define POWEROFF_DELAY 5000 620 #endif 621 static int poweroff_delay = POWEROFF_DELAY; 622 623 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 624 &poweroff_delay, 0, ""); 625 626 static void 627 poweroff_wait(void *junk, int howto) 628 { 629 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 630 return; 631 DELAY(poweroff_delay * 1000); 632 } 633 634 /* 635 * Some system processes (e.g. syncer) need to be stopped at appropriate 636 * points in their main loops prior to a system shutdown, so that they 637 * won't interfere with the shutdown process (e.g. by holding a disk buf 638 * to cause sync to fail). For each of these system processes, register 639 * shutdown_kproc() as a handler for one of shutdown events. 640 */ 641 static int kproc_shutdown_wait = 60; 642 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 643 &kproc_shutdown_wait, 0, ""); 644 645 void 646 shutdown_kproc(void *arg, int howto) 647 { 648 struct thread *td; 649 struct proc *p; 650 int error; 651 652 if (panicstr) 653 return; 654 655 td = (struct thread *)arg; 656 if ((p = td->td_proc) != NULL) { 657 printf("Waiting (max %d seconds) for system process `%s' to stop...", 658 kproc_shutdown_wait, p->p_comm); 659 } else { 660 printf("Waiting (max %d seconds) for system thread %s to stop...", 661 kproc_shutdown_wait, td->td_comm); 662 } 663 error = suspend_kproc(td, kproc_shutdown_wait * hz); 664 665 if (error == EWOULDBLOCK) 666 printf("timed out\n"); 667 else 668 printf("stopped\n"); 669 } 670