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 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)kern_shutdown.c 8.3 (Berkeley) 1/21/94 35 */ 36 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD$"); 39 40 #include "opt_ddb.h" 41 #include "opt_kdb.h" 42 #include "opt_panic.h" 43 #include "opt_sched.h" 44 #include "opt_watchdog.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/bio.h> 49 #include <sys/buf.h> 50 #include <sys/conf.h> 51 #include <sys/cons.h> 52 #include <sys/eventhandler.h> 53 #include <sys/jail.h> 54 #include <sys/kdb.h> 55 #include <sys/kernel.h> 56 #include <sys/kerneldump.h> 57 #include <sys/kthread.h> 58 #include <sys/ktr.h> 59 #include <sys/malloc.h> 60 #include <sys/mount.h> 61 #include <sys/priv.h> 62 #include <sys/proc.h> 63 #include <sys/reboot.h> 64 #include <sys/resourcevar.h> 65 #include <sys/rwlock.h> 66 #include <sys/sched.h> 67 #include <sys/smp.h> 68 #include <sys/sysctl.h> 69 #include <sys/sysproto.h> 70 #include <sys/vnode.h> 71 #include <sys/watchdog.h> 72 73 #include <ddb/ddb.h> 74 75 #include <machine/cpu.h> 76 #include <machine/pcb.h> 77 #include <machine/smp.h> 78 79 #include <security/mac/mac_framework.h> 80 81 #include <vm/vm.h> 82 #include <vm/vm_object.h> 83 #include <vm/vm_page.h> 84 #include <vm/vm_pager.h> 85 #include <vm/swap_pager.h> 86 87 #include <sys/signalvar.h> 88 89 #ifndef PANIC_REBOOT_WAIT_TIME 90 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 91 #endif 92 static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME; 93 SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RW | CTLFLAG_TUN, 94 &panic_reboot_wait_time, 0, 95 "Seconds to wait before rebooting after a panic"); 96 TUNABLE_INT("kern.panic_reboot_wait_time", &panic_reboot_wait_time); 97 98 /* 99 * Note that stdarg.h and the ANSI style va_start macro is used for both 100 * ANSI and traditional C compilers. 101 */ 102 #include <machine/stdarg.h> 103 104 #ifdef KDB 105 #ifdef KDB_UNATTENDED 106 int debugger_on_panic = 0; 107 #else 108 int debugger_on_panic = 1; 109 #endif 110 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, 111 CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN, 112 &debugger_on_panic, 0, "Run debugger on kernel panic"); 113 TUNABLE_INT("debug.debugger_on_panic", &debugger_on_panic); 114 115 #ifdef KDB_TRACE 116 static int trace_on_panic = 1; 117 #else 118 static int trace_on_panic = 0; 119 #endif 120 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, 121 CTLFLAG_RW | CTLFLAG_SECURE | CTLFLAG_TUN, 122 &trace_on_panic, 0, "Print stack trace on kernel panic"); 123 TUNABLE_INT("debug.trace_on_panic", &trace_on_panic); 124 #endif /* KDB */ 125 126 static int sync_on_panic = 0; 127 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW | CTLFLAG_TUN, 128 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 129 TUNABLE_INT("kern.sync_on_panic", &sync_on_panic); 130 131 static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, 132 "Shutdown environment"); 133 134 #ifndef DIAGNOSTIC 135 static int show_busybufs; 136 #else 137 static int show_busybufs = 1; 138 #endif 139 SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW, 140 &show_busybufs, 0, ""); 141 142 /* 143 * Variable panicstr contains argument to first call to panic; used as flag 144 * to indicate that the kernel has already called panic. 145 */ 146 const char *panicstr; 147 148 int dumping; /* system is dumping */ 149 int rebooting; /* system is rebooting */ 150 static struct dumperinfo dumper; /* our selected dumper */ 151 152 /* Context information for dump-debuggers. */ 153 static struct pcb dumppcb; /* Registers. */ 154 lwpid_t dumptid; /* Thread ID. */ 155 156 static void poweroff_wait(void *, int); 157 static void shutdown_halt(void *junk, int howto); 158 static void shutdown_panic(void *junk, int howto); 159 static void shutdown_reset(void *junk, int howto); 160 static void vpanic(const char *fmt, va_list ap) __dead2; 161 162 /* register various local shutdown events */ 163 static void 164 shutdown_conf(void *unused) 165 { 166 167 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 168 SHUTDOWN_PRI_FIRST); 169 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 170 SHUTDOWN_PRI_LAST + 100); 171 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 172 SHUTDOWN_PRI_LAST + 100); 173 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 174 SHUTDOWN_PRI_LAST + 200); 175 } 176 177 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL); 178 179 /* 180 * The system call that results in a reboot. 181 */ 182 /* ARGSUSED */ 183 int 184 sys_reboot(struct thread *td, struct reboot_args *uap) 185 { 186 int error; 187 188 error = 0; 189 #ifdef MAC 190 error = mac_system_check_reboot(td->td_ucred, uap->opt); 191 #endif 192 if (error == 0) 193 error = priv_check(td, PRIV_REBOOT); 194 if (error == 0) { 195 mtx_lock(&Giant); 196 kern_reboot(uap->opt); 197 mtx_unlock(&Giant); 198 } 199 return (error); 200 } 201 202 /* 203 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 204 */ 205 void 206 shutdown_nice(int howto) 207 { 208 209 if (initproc != NULL) { 210 /* Send a signal to init(8) and have it shutdown the world. */ 211 PROC_LOCK(initproc); 212 if (howto & RB_POWEROFF) 213 kern_psignal(initproc, SIGUSR2); 214 else if (howto & RB_HALT) 215 kern_psignal(initproc, SIGUSR1); 216 else 217 kern_psignal(initproc, SIGINT); 218 PROC_UNLOCK(initproc); 219 } else { 220 /* No init(8) running, so simply reboot. */ 221 kern_reboot(howto | RB_NOSYNC); 222 } 223 } 224 225 static void 226 print_uptime(void) 227 { 228 int f; 229 struct timespec ts; 230 231 getnanouptime(&ts); 232 printf("Uptime: "); 233 f = 0; 234 if (ts.tv_sec >= 86400) { 235 printf("%ldd", (long)ts.tv_sec / 86400); 236 ts.tv_sec %= 86400; 237 f = 1; 238 } 239 if (f || ts.tv_sec >= 3600) { 240 printf("%ldh", (long)ts.tv_sec / 3600); 241 ts.tv_sec %= 3600; 242 f = 1; 243 } 244 if (f || ts.tv_sec >= 60) { 245 printf("%ldm", (long)ts.tv_sec / 60); 246 ts.tv_sec %= 60; 247 f = 1; 248 } 249 printf("%lds\n", (long)ts.tv_sec); 250 } 251 252 int 253 doadump(boolean_t textdump) 254 { 255 boolean_t coredump; 256 257 if (dumping) 258 return (EBUSY); 259 if (dumper.dumper == NULL) 260 return (ENXIO); 261 262 savectx(&dumppcb); 263 dumptid = curthread->td_tid; 264 dumping++; 265 266 coredump = TRUE; 267 #ifdef DDB 268 if (textdump && textdump_pending) { 269 coredump = FALSE; 270 textdump_dumpsys(&dumper); 271 } 272 #endif 273 if (coredump) 274 dumpsys(&dumper); 275 276 dumping--; 277 return (0); 278 } 279 280 static int 281 isbufbusy(struct buf *bp) 282 { 283 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 && 284 BUF_ISLOCKED(bp)) || 285 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)) 286 return (1); 287 return (0); 288 } 289 290 /* 291 * Shutdown the system cleanly to prepare for reboot, halt, or power off. 292 */ 293 void 294 kern_reboot(int howto) 295 { 296 static int first_buf_printf = 1; 297 static int waittime = -1; 298 299 #if defined(SMP) 300 /* 301 * Bind us to CPU 0 so that all shutdown code runs there. Some 302 * systems don't shutdown properly (i.e., ACPI power off) if we 303 * run on another processor. 304 */ 305 if (!SCHEDULER_STOPPED()) { 306 thread_lock(curthread); 307 sched_bind(curthread, 0); 308 thread_unlock(curthread); 309 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0")); 310 } 311 #endif 312 /* We're in the process of rebooting. */ 313 rebooting = 1; 314 315 /* We are out of the debugger now. */ 316 kdb_active = 0; 317 318 /* 319 * Do any callouts that should be done BEFORE syncing the filesystems. 320 */ 321 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 322 323 /* 324 * Now sync filesystems 325 */ 326 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 327 register struct buf *bp; 328 int iter, nbusy, pbusy; 329 #ifndef PREEMPTION 330 int subiter; 331 #endif 332 333 waittime = 0; 334 335 wdog_kern_pat(WD_LASTVAL); 336 sys_sync(curthread, NULL); 337 338 /* 339 * With soft updates, some buffers that are 340 * written will be remarked as dirty until other 341 * buffers are written. 342 */ 343 for (iter = pbusy = 0; iter < 20; iter++) { 344 nbusy = 0; 345 for (bp = &buf[nbuf]; --bp >= buf; ) 346 if (isbufbusy(bp)) 347 nbusy++; 348 if (nbusy == 0) { 349 if (first_buf_printf) 350 printf("All buffers synced."); 351 break; 352 } 353 if (first_buf_printf) { 354 printf("Syncing disks, buffers remaining... "); 355 first_buf_printf = 0; 356 } 357 printf("%d ", nbusy); 358 if (nbusy < pbusy) 359 iter = 0; 360 pbusy = nbusy; 361 362 wdog_kern_pat(WD_LASTVAL); 363 sys_sync(curthread, NULL); 364 365 #ifdef PREEMPTION 366 /* 367 * Drop Giant and spin for a while to allow 368 * interrupt threads to run. 369 */ 370 DROP_GIANT(); 371 DELAY(50000 * iter); 372 PICKUP_GIANT(); 373 #else 374 /* 375 * Drop Giant and context switch several times to 376 * allow interrupt threads to run. 377 */ 378 DROP_GIANT(); 379 for (subiter = 0; subiter < 50 * iter; subiter++) { 380 thread_lock(curthread); 381 mi_switch(SW_VOL, NULL); 382 thread_unlock(curthread); 383 DELAY(1000); 384 } 385 PICKUP_GIANT(); 386 #endif 387 } 388 printf("\n"); 389 /* 390 * Count only busy local buffers to prevent forcing 391 * a fsck if we're just a client of a wedged NFS server 392 */ 393 nbusy = 0; 394 for (bp = &buf[nbuf]; --bp >= buf; ) { 395 if (isbufbusy(bp)) { 396 #if 0 397 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 398 if (bp->b_dev == NULL) { 399 TAILQ_REMOVE(&mountlist, 400 bp->b_vp->v_mount, mnt_list); 401 continue; 402 } 403 #endif 404 nbusy++; 405 if (show_busybufs > 0) { 406 printf( 407 "%d: buf:%p, vnode:%p, flags:%0x, blkno:%jd, lblkno:%jd, buflock:", 408 nbusy, bp, bp->b_vp, bp->b_flags, 409 (intmax_t)bp->b_blkno, 410 (intmax_t)bp->b_lblkno); 411 BUF_LOCKPRINTINFO(bp); 412 if (show_busybufs > 1) 413 vn_printf(bp->b_vp, 414 "vnode content: "); 415 } 416 } 417 } 418 if (nbusy) { 419 /* 420 * Failed to sync all blocks. Indicate this and don't 421 * unmount filesystems (thus forcing an fsck on reboot). 422 */ 423 printf("Giving up on %d buffers\n", nbusy); 424 DELAY(5000000); /* 5 seconds */ 425 } else { 426 if (!first_buf_printf) 427 printf("Final sync complete\n"); 428 /* 429 * Unmount filesystems 430 */ 431 if (panicstr == 0) 432 vfs_unmountall(); 433 } 434 swapoff_all(); 435 DELAY(100000); /* wait for console output to finish */ 436 } 437 438 print_uptime(); 439 440 cngrab(); 441 442 /* 443 * Ok, now do things that assume all filesystem activity has 444 * been completed. 445 */ 446 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 447 448 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 449 doadump(TRUE); 450 451 /* Now that we're going to really halt the system... */ 452 EVENTHANDLER_INVOKE(shutdown_final, howto); 453 454 for(;;) ; /* safety against shutdown_reset not working */ 455 /* NOTREACHED */ 456 } 457 458 /* 459 * If the shutdown was a clean halt, behave accordingly. 460 */ 461 static void 462 shutdown_halt(void *junk, int howto) 463 { 464 465 if (howto & RB_HALT) { 466 printf("\n"); 467 printf("The operating system has halted.\n"); 468 printf("Please press any key to reboot.\n\n"); 469 switch (cngetc()) { 470 case -1: /* No console, just die */ 471 cpu_halt(); 472 /* NOTREACHED */ 473 default: 474 howto &= ~RB_HALT; 475 break; 476 } 477 } 478 } 479 480 /* 481 * Check to see if the system paniced, pause and then reboot 482 * according to the specified delay. 483 */ 484 static void 485 shutdown_panic(void *junk, int howto) 486 { 487 int loop; 488 489 if (howto & RB_DUMP) { 490 if (panic_reboot_wait_time != 0) { 491 if (panic_reboot_wait_time != -1) { 492 printf("Automatic reboot in %d seconds - " 493 "press a key on the console to abort\n", 494 panic_reboot_wait_time); 495 for (loop = panic_reboot_wait_time * 10; 496 loop > 0; --loop) { 497 DELAY(1000 * 100); /* 1/10th second */ 498 /* Did user type a key? */ 499 if (cncheckc() != -1) 500 break; 501 } 502 if (!loop) 503 return; 504 } 505 } else { /* zero time specified - reboot NOW */ 506 return; 507 } 508 printf("--> Press a key on the console to reboot,\n"); 509 printf("--> or switch off the system now.\n"); 510 cngetc(); 511 } 512 } 513 514 /* 515 * Everything done, now reset 516 */ 517 static void 518 shutdown_reset(void *junk, int howto) 519 { 520 521 printf("Rebooting...\n"); 522 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 523 524 /* 525 * Acquiring smp_ipi_mtx here has a double effect: 526 * - it disables interrupts avoiding CPU0 preemption 527 * by fast handlers (thus deadlocking against other CPUs) 528 * - it avoids deadlocks against smp_rendezvous() or, more 529 * generally, threads busy-waiting, with this spinlock held, 530 * and waiting for responses by threads on other CPUs 531 * (ie. smp_tlb_shootdown()). 532 * 533 * For the !SMP case it just needs to handle the former problem. 534 */ 535 #ifdef SMP 536 mtx_lock_spin(&smp_ipi_mtx); 537 #else 538 spinlock_enter(); 539 #endif 540 541 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 542 cpu_reset(); 543 /* NOTREACHED */ /* assuming reset worked */ 544 } 545 546 #if defined(WITNESS) || defined(INVARIANTS) 547 static int kassert_warn_only = 0; 548 #ifdef KDB 549 static int kassert_do_kdb = 0; 550 #endif 551 #ifdef KTR 552 static int kassert_do_ktr = 0; 553 #endif 554 static int kassert_do_log = 1; 555 static int kassert_log_pps_limit = 4; 556 static int kassert_log_mute_at = 0; 557 static int kassert_log_panic_at = 0; 558 static int kassert_warnings = 0; 559 560 SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options"); 561 562 SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RW | CTLFLAG_TUN, 563 &kassert_warn_only, 0, 564 "KASSERT triggers a panic (1) or just a warning (0)"); 565 TUNABLE_INT("debug.kassert.warn_only", &kassert_warn_only); 566 567 #ifdef KDB 568 SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RW | CTLFLAG_TUN, 569 &kassert_do_kdb, 0, "KASSERT will enter the debugger"); 570 TUNABLE_INT("debug.kassert.do_kdb", &kassert_do_kdb); 571 #endif 572 573 #ifdef KTR 574 SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RW | CTLFLAG_TUN, 575 &kassert_do_ktr, 0, 576 "KASSERT does a KTR, set this to the KTRMASK you want"); 577 TUNABLE_INT("debug.kassert.do_ktr", &kassert_do_ktr); 578 #endif 579 580 SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RW | CTLFLAG_TUN, 581 &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)"); 582 TUNABLE_INT("debug.kassert.do_log", &kassert_do_log); 583 584 SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RW | CTLFLAG_TUN, 585 &kassert_warnings, 0, "number of KASSERTs that have been triggered"); 586 TUNABLE_INT("debug.kassert.warnings", &kassert_warnings); 587 588 SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RW | CTLFLAG_TUN, 589 &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic"); 590 TUNABLE_INT("debug.kassert.log_panic_at", &kassert_log_panic_at); 591 592 SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RW | CTLFLAG_TUN, 593 &kassert_log_pps_limit, 0, "limit number of log messages per second"); 594 TUNABLE_INT("debug.kassert.log_pps_limit", &kassert_log_pps_limit); 595 596 SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RW | CTLFLAG_TUN, 597 &kassert_log_mute_at, 0, "max number of KASSERTS to log"); 598 TUNABLE_INT("debug.kassert.log_mute_at", &kassert_log_mute_at); 599 600 static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS); 601 602 SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert, 603 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0, 604 kassert_sysctl_kassert, "I", "set to trigger a test kassert"); 605 606 static int 607 kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS) 608 { 609 int error, i; 610 611 error = sysctl_wire_old_buffer(req, sizeof(int)); 612 if (error == 0) { 613 i = 0; 614 error = sysctl_handle_int(oidp, &i, 0, req); 615 } 616 if (error != 0 || req->newptr == NULL) 617 return (error); 618 KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i)); 619 return (0); 620 } 621 622 /* 623 * Called by KASSERT, this decides if we will panic 624 * or if we will log via printf and/or ktr. 625 */ 626 void 627 kassert_panic(const char *fmt, ...) 628 { 629 static char buf[256]; 630 va_list ap; 631 632 va_start(ap, fmt); 633 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 634 va_end(ap); 635 636 /* 637 * panic if we're not just warning, or if we've exceeded 638 * kassert_log_panic_at warnings. 639 */ 640 if (!kassert_warn_only || 641 (kassert_log_panic_at > 0 && 642 kassert_warnings >= kassert_log_panic_at)) { 643 va_start(ap, fmt); 644 vpanic(fmt, ap); 645 /* NORETURN */ 646 } 647 #ifdef KTR 648 if (kassert_do_ktr) 649 CTR0(ktr_mask, buf); 650 #endif /* KTR */ 651 /* 652 * log if we've not yet met the mute limit. 653 */ 654 if (kassert_do_log && 655 (kassert_log_mute_at == 0 || 656 kassert_warnings < kassert_log_mute_at)) { 657 static struct timeval lasterr; 658 static int curerr; 659 660 if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) { 661 printf("KASSERT failed: %s\n", buf); 662 kdb_backtrace(); 663 } 664 } 665 #ifdef KDB 666 if (kassert_do_kdb) { 667 kdb_enter(KDB_WHY_KASSERT, buf); 668 } 669 #endif 670 atomic_add_int(&kassert_warnings, 1); 671 } 672 #endif 673 674 /* 675 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 676 * and then reboots. If we are called twice, then we avoid trying to sync 677 * the disks as this often leads to recursive panics. 678 */ 679 void 680 panic(const char *fmt, ...) 681 { 682 va_list ap; 683 684 va_start(ap, fmt); 685 vpanic(fmt, ap); 686 } 687 688 static void 689 vpanic(const char *fmt, va_list ap) 690 { 691 #ifdef SMP 692 cpuset_t other_cpus; 693 #endif 694 struct thread *td = curthread; 695 int bootopt, newpanic; 696 static char buf[256]; 697 698 spinlock_enter(); 699 700 #ifdef SMP 701 /* 702 * stop_cpus_hard(other_cpus) should prevent multiple CPUs from 703 * concurrently entering panic. Only the winner will proceed 704 * further. 705 */ 706 if (panicstr == NULL && !kdb_active) { 707 other_cpus = all_cpus; 708 CPU_CLR(PCPU_GET(cpuid), &other_cpus); 709 stop_cpus_hard(other_cpus); 710 } 711 712 /* 713 * We set stop_scheduler here and not in the block above, 714 * because we want to ensure that if panic has been called and 715 * stop_scheduler_on_panic is true, then stop_scheduler will 716 * always be set. Even if panic has been entered from kdb. 717 */ 718 td->td_stopsched = 1; 719 #endif 720 721 bootopt = RB_AUTOBOOT; 722 newpanic = 0; 723 if (panicstr) 724 bootopt |= RB_NOSYNC; 725 else { 726 bootopt |= RB_DUMP; 727 panicstr = fmt; 728 newpanic = 1; 729 } 730 731 if (newpanic) { 732 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 733 panicstr = buf; 734 cngrab(); 735 printf("panic: %s\n", buf); 736 } else { 737 printf("panic: "); 738 vprintf(fmt, ap); 739 printf("\n"); 740 } 741 #ifdef SMP 742 printf("cpuid = %d\n", PCPU_GET(cpuid)); 743 #endif 744 745 #ifdef KDB 746 if (newpanic && trace_on_panic) 747 kdb_backtrace(); 748 if (debugger_on_panic) 749 kdb_enter(KDB_WHY_PANIC, "panic"); 750 #endif 751 /*thread_lock(td); */ 752 td->td_flags |= TDF_INPANIC; 753 /* thread_unlock(td); */ 754 if (!sync_on_panic) 755 bootopt |= RB_NOSYNC; 756 kern_reboot(bootopt); 757 } 758 759 /* 760 * Support for poweroff delay. 761 * 762 * Please note that setting this delay too short might power off your machine 763 * before the write cache on your hard disk has been flushed, leading to 764 * soft-updates inconsistencies. 765 */ 766 #ifndef POWEROFF_DELAY 767 # define POWEROFF_DELAY 5000 768 #endif 769 static int poweroff_delay = POWEROFF_DELAY; 770 771 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 772 &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)"); 773 774 static void 775 poweroff_wait(void *junk, int howto) 776 { 777 778 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 779 return; 780 DELAY(poweroff_delay * 1000); 781 } 782 783 /* 784 * Some system processes (e.g. syncer) need to be stopped at appropriate 785 * points in their main loops prior to a system shutdown, so that they 786 * won't interfere with the shutdown process (e.g. by holding a disk buf 787 * to cause sync to fail). For each of these system processes, register 788 * shutdown_kproc() as a handler for one of shutdown events. 789 */ 790 static int kproc_shutdown_wait = 60; 791 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 792 &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process"); 793 794 void 795 kproc_shutdown(void *arg, int howto) 796 { 797 struct proc *p; 798 int error; 799 800 if (panicstr) 801 return; 802 803 p = (struct proc *)arg; 804 printf("Waiting (max %d seconds) for system process `%s' to stop...", 805 kproc_shutdown_wait, p->p_comm); 806 error = kproc_suspend(p, kproc_shutdown_wait * hz); 807 808 if (error == EWOULDBLOCK) 809 printf("timed out\n"); 810 else 811 printf("done\n"); 812 } 813 814 void 815 kthread_shutdown(void *arg, int howto) 816 { 817 struct thread *td; 818 int error; 819 820 if (panicstr) 821 return; 822 823 td = (struct thread *)arg; 824 printf("Waiting (max %d seconds) for system thread `%s' to stop...", 825 kproc_shutdown_wait, td->td_name); 826 error = kthread_suspend(td, kproc_shutdown_wait * hz); 827 828 if (error == EWOULDBLOCK) 829 printf("timed out\n"); 830 else 831 printf("done\n"); 832 } 833 834 static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)]; 835 SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD, 836 dumpdevname, 0, "Device for kernel dumps"); 837 838 /* Registration of dumpers */ 839 int 840 set_dumper(struct dumperinfo *di, const char *devname) 841 { 842 size_t wantcopy; 843 844 if (di == NULL) { 845 bzero(&dumper, sizeof dumper); 846 dumpdevname[0] = '\0'; 847 return (0); 848 } 849 if (dumper.dumper != NULL) 850 return (EBUSY); 851 dumper = *di; 852 wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname)); 853 if (wantcopy >= sizeof(dumpdevname)) { 854 printf("set_dumper: device name truncated from '%s' -> '%s'\n", 855 devname, dumpdevname); 856 } 857 return (0); 858 } 859 860 /* Call dumper with bounds checking. */ 861 int 862 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical, 863 off_t offset, size_t length) 864 { 865 866 if (length != 0 && (offset < di->mediaoffset || 867 offset - di->mediaoffset + length > di->mediasize)) { 868 printf("Attempt to write outside dump device boundaries.\n" 869 "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n", 870 (intmax_t)offset, (intmax_t)di->mediaoffset, 871 (uintmax_t)length, (intmax_t)di->mediasize); 872 return (ENOSPC); 873 } 874 return (di->dumper(di->priv, virtual, physical, offset, length)); 875 } 876 877 void 878 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver, 879 uint64_t dumplen, uint32_t blksz) 880 { 881 882 bzero(kdh, sizeof(*kdh)); 883 strncpy(kdh->magic, magic, sizeof(kdh->magic)); 884 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture)); 885 kdh->version = htod32(KERNELDUMPVERSION); 886 kdh->architectureversion = htod32(archver); 887 kdh->dumplength = htod64(dumplen); 888 kdh->dumptime = htod64(time_second); 889 kdh->blocksize = htod32(blksz); 890 strncpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname)); 891 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring)); 892 if (panicstr != NULL) 893 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring)); 894 kdh->parity = kerneldump_parity(kdh); 895 } 896