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_mac.h" 43 #include "opt_panic.h" 44 #include "opt_show_busybufs.h" 45 #include "opt_sched.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/bio.h> 50 #include <sys/buf.h> 51 #include <sys/conf.h> 52 #include <sys/cons.h> 53 #include <sys/eventhandler.h> 54 #include <sys/kdb.h> 55 #include <sys/kernel.h> 56 #include <sys/kerneldump.h> 57 #include <sys/kthread.h> 58 #include <sys/malloc.h> 59 #include <sys/mount.h> 60 #include <sys/priv.h> 61 #include <sys/proc.h> 62 #include <sys/reboot.h> 63 #include <sys/resourcevar.h> 64 #include <sys/sched.h> 65 #include <sys/smp.h> /* smp_active */ 66 #include <sys/sysctl.h> 67 #include <sys/sysproto.h> 68 #include <sys/vimage.h> 69 70 #include <ddb/ddb.h> 71 72 #include <machine/cpu.h> 73 #include <machine/pcb.h> 74 #include <machine/smp.h> 75 76 #include <security/mac/mac_framework.h> 77 78 #include <vm/vm.h> 79 #include <vm/vm_object.h> 80 #include <vm/vm_page.h> 81 #include <vm/vm_pager.h> 82 #include <vm/swap_pager.h> 83 84 #include <sys/signalvar.h> 85 86 #ifndef PANIC_REBOOT_WAIT_TIME 87 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 88 #endif 89 90 /* 91 * Note that stdarg.h and the ANSI style va_start macro is used for both 92 * ANSI and traditional C compilers. 93 */ 94 #include <machine/stdarg.h> 95 96 #ifdef KDB 97 #ifdef KDB_UNATTENDED 98 int debugger_on_panic = 0; 99 #else 100 int debugger_on_panic = 1; 101 #endif 102 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 103 &debugger_on_panic, 0, "Run debugger on kernel panic"); 104 105 #ifdef KDB_TRACE 106 int trace_on_panic = 1; 107 #else 108 int trace_on_panic = 0; 109 #endif 110 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 111 &trace_on_panic, 0, "Print stack trace on kernel panic"); 112 #endif /* KDB */ 113 114 int sync_on_panic = 0; 115 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 116 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 117 118 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 119 120 /* 121 * Variable panicstr contains argument to first call to panic; used as flag 122 * to indicate that the kernel has already called panic. 123 */ 124 const char *panicstr; 125 126 int dumping; /* system is dumping */ 127 int rebooting; /* system is rebooting */ 128 static struct dumperinfo dumper; /* our selected dumper */ 129 130 /* Context information for dump-debuggers. */ 131 static struct pcb dumppcb; /* Registers. */ 132 static lwpid_t dumptid; /* Thread ID. */ 133 134 static void boot(int) __dead2; 135 static void poweroff_wait(void *, int); 136 static void shutdown_halt(void *junk, int howto); 137 static void shutdown_panic(void *junk, int howto); 138 static void shutdown_reset(void *junk, int howto); 139 140 /* register various local shutdown events */ 141 static void 142 shutdown_conf(void *unused) 143 { 144 145 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 146 SHUTDOWN_PRI_FIRST); 147 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 148 SHUTDOWN_PRI_LAST + 100); 149 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 150 SHUTDOWN_PRI_LAST + 100); 151 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 152 SHUTDOWN_PRI_LAST + 200); 153 } 154 155 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL); 156 157 /* 158 * The system call that results in a reboot. 159 */ 160 /* ARGSUSED */ 161 int 162 reboot(struct thread *td, struct reboot_args *uap) 163 { 164 int error; 165 166 error = 0; 167 #ifdef MAC 168 error = mac_system_check_reboot(td->td_ucred, uap->opt); 169 #endif 170 if (error == 0) 171 error = priv_check(td, PRIV_REBOOT); 172 if (error == 0) { 173 mtx_lock(&Giant); 174 boot(uap->opt); 175 mtx_unlock(&Giant); 176 } 177 return (error); 178 } 179 180 /* 181 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 182 */ 183 static int shutdown_howto = 0; 184 185 void 186 shutdown_nice(int howto) 187 { 188 189 shutdown_howto = howto; 190 191 /* Send a signal to init(8) and have it shutdown the world */ 192 if (initproc != NULL) { 193 PROC_LOCK(initproc); 194 psignal(initproc, SIGINT); 195 PROC_UNLOCK(initproc); 196 } else { 197 /* No init(8) running, so simply reboot */ 198 boot(RB_NOSYNC); 199 } 200 return; 201 } 202 static int waittime = -1; 203 204 static void 205 print_uptime(void) 206 { 207 int f; 208 struct timespec ts; 209 210 getnanouptime(&ts); 211 printf("Uptime: "); 212 f = 0; 213 if (ts.tv_sec >= 86400) { 214 printf("%ldd", (long)ts.tv_sec / 86400); 215 ts.tv_sec %= 86400; 216 f = 1; 217 } 218 if (f || ts.tv_sec >= 3600) { 219 printf("%ldh", (long)ts.tv_sec / 3600); 220 ts.tv_sec %= 3600; 221 f = 1; 222 } 223 if (f || ts.tv_sec >= 60) { 224 printf("%ldm", (long)ts.tv_sec / 60); 225 ts.tv_sec %= 60; 226 f = 1; 227 } 228 printf("%lds\n", (long)ts.tv_sec); 229 } 230 231 static void 232 doadump(void) 233 { 234 235 /* 236 * Sometimes people have to call this from the kernel debugger. 237 * (if 'panic' can not dump) 238 * Give them a clue as to why they can't dump. 239 */ 240 if (dumper.dumper == NULL) { 241 printf("Cannot dump. No dump device defined.\n"); 242 return; 243 } 244 245 savectx(&dumppcb); 246 dumptid = curthread->td_tid; 247 dumping++; 248 #ifdef DDB 249 if (textdump_pending) 250 textdump_dumpsys(&dumper); 251 else 252 #endif 253 dumpsys(&dumper); 254 dumping--; 255 } 256 257 static int 258 isbufbusy(struct buf *bp) 259 { 260 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 && 261 BUF_ISLOCKED(bp)) || 262 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)) 263 return (1); 264 return (0); 265 } 266 267 /* 268 * Shutdown the system cleanly to prepare for reboot, halt, or power off. 269 */ 270 static void 271 boot(int howto) 272 { 273 static int first_buf_printf = 1; 274 275 #if defined(SMP) 276 /* 277 * Bind us to CPU 0 so that all shutdown code runs there. Some 278 * systems don't shutdown properly (i.e., ACPI power off) if we 279 * run on another processor. 280 */ 281 thread_lock(curthread); 282 sched_bind(curthread, 0); 283 thread_unlock(curthread); 284 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0")); 285 #endif 286 /* We're in the process of rebooting. */ 287 rebooting = 1; 288 289 /* collect extra flags that shutdown_nice might have set */ 290 howto |= shutdown_howto; 291 292 /* We are out of the debugger now. */ 293 kdb_active = 0; 294 295 /* 296 * Do any callouts that should be done BEFORE syncing the filesystems. 297 */ 298 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 299 300 /* 301 * Now sync filesystems 302 */ 303 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 304 register struct buf *bp; 305 int iter, nbusy, pbusy; 306 #ifndef PREEMPTION 307 int subiter; 308 #endif 309 310 waittime = 0; 311 312 sync(curthread, NULL); 313 314 /* 315 * With soft updates, some buffers that are 316 * written will be remarked as dirty until other 317 * buffers are written. 318 */ 319 for (iter = pbusy = 0; iter < 20; iter++) { 320 nbusy = 0; 321 for (bp = &buf[nbuf]; --bp >= buf; ) 322 if (isbufbusy(bp)) 323 nbusy++; 324 if (nbusy == 0) { 325 if (first_buf_printf) 326 printf("All buffers synced."); 327 break; 328 } 329 if (first_buf_printf) { 330 printf("Syncing disks, buffers remaining... "); 331 first_buf_printf = 0; 332 } 333 printf("%d ", nbusy); 334 if (nbusy < pbusy) 335 iter = 0; 336 pbusy = nbusy; 337 sync(curthread, NULL); 338 339 #ifdef PREEMPTION 340 /* 341 * Drop Giant and spin for a while to allow 342 * interrupt threads to run. 343 */ 344 DROP_GIANT(); 345 DELAY(50000 * iter); 346 PICKUP_GIANT(); 347 #else 348 /* 349 * Drop Giant and context switch several times to 350 * allow interrupt threads to run. 351 */ 352 DROP_GIANT(); 353 for (subiter = 0; subiter < 50 * iter; subiter++) { 354 thread_lock(curthread); 355 mi_switch(SW_VOL, NULL); 356 thread_unlock(curthread); 357 DELAY(1000); 358 } 359 PICKUP_GIANT(); 360 #endif 361 } 362 printf("\n"); 363 /* 364 * Count only busy local buffers to prevent forcing 365 * a fsck if we're just a client of a wedged NFS server 366 */ 367 nbusy = 0; 368 for (bp = &buf[nbuf]; --bp >= buf; ) { 369 if (isbufbusy(bp)) { 370 #if 0 371 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 372 if (bp->b_dev == NULL) { 373 TAILQ_REMOVE(&mountlist, 374 bp->b_vp->v_mount, mnt_list); 375 continue; 376 } 377 #endif 378 nbusy++; 379 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 380 printf( 381 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n", 382 nbusy, bp->b_bufobj, 383 bp->b_flags, (long)bp->b_blkno, 384 (long)bp->b_lblkno); 385 #endif 386 } 387 } 388 if (nbusy) { 389 /* 390 * Failed to sync all blocks. Indicate this and don't 391 * unmount filesystems (thus forcing an fsck on reboot). 392 */ 393 printf("Giving up on %d buffers\n", nbusy); 394 DELAY(5000000); /* 5 seconds */ 395 } else { 396 if (!first_buf_printf) 397 printf("Final sync complete\n"); 398 /* 399 * Unmount filesystems 400 */ 401 if (panicstr == 0) 402 vfs_unmountall(); 403 } 404 swapoff_all(); 405 DELAY(100000); /* wait for console output to finish */ 406 } 407 408 print_uptime(); 409 410 /* 411 * Ok, now do things that assume all filesystem activity has 412 * been completed. 413 */ 414 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 415 416 /* XXX This doesn't disable interrupts any more. Reconsider? */ 417 splhigh(); 418 419 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 420 doadump(); 421 422 /* Now that we're going to really halt the system... */ 423 EVENTHANDLER_INVOKE(shutdown_final, howto); 424 425 for(;;) ; /* safety against shutdown_reset not working */ 426 /* NOTREACHED */ 427 } 428 429 /* 430 * If the shutdown was a clean halt, behave accordingly. 431 */ 432 static void 433 shutdown_halt(void *junk, int howto) 434 { 435 436 if (howto & RB_HALT) { 437 printf("\n"); 438 printf("The operating system has halted.\n"); 439 printf("Please press any key to reboot.\n\n"); 440 switch (cngetc()) { 441 case -1: /* No console, just die */ 442 cpu_halt(); 443 /* NOTREACHED */ 444 default: 445 howto &= ~RB_HALT; 446 break; 447 } 448 } 449 } 450 451 /* 452 * Check to see if the system paniced, pause and then reboot 453 * according to the specified delay. 454 */ 455 static void 456 shutdown_panic(void *junk, int howto) 457 { 458 int loop; 459 460 if (howto & RB_DUMP) { 461 if (PANIC_REBOOT_WAIT_TIME != 0) { 462 if (PANIC_REBOOT_WAIT_TIME != -1) { 463 printf("Automatic reboot in %d seconds - " 464 "press a key on the console to abort\n", 465 PANIC_REBOOT_WAIT_TIME); 466 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 467 loop > 0; --loop) { 468 DELAY(1000 * 100); /* 1/10th second */ 469 /* Did user type a key? */ 470 if (cncheckc() != -1) 471 break; 472 } 473 if (!loop) 474 return; 475 } 476 } else { /* zero time specified - reboot NOW */ 477 return; 478 } 479 printf("--> Press a key on the console to reboot,\n"); 480 printf("--> or switch off the system now.\n"); 481 cngetc(); 482 } 483 } 484 485 /* 486 * Everything done, now reset 487 */ 488 static void 489 shutdown_reset(void *junk, int howto) 490 { 491 492 printf("Rebooting...\n"); 493 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 494 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 495 cpu_reset(); 496 /* NOTREACHED */ /* assuming reset worked */ 497 } 498 499 #ifdef SMP 500 static u_int panic_cpu = NOCPU; 501 #endif 502 503 /* 504 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 505 * and then reboots. If we are called twice, then we avoid trying to sync 506 * the disks as this often leads to recursive panics. 507 */ 508 void 509 panic(const char *fmt, ...) 510 { 511 struct thread *td = curthread; 512 int bootopt, newpanic; 513 va_list ap; 514 static char buf[256]; 515 516 critical_enter(); 517 #ifdef SMP 518 /* 519 * We don't want multiple CPU's to panic at the same time, so we 520 * use panic_cpu as a simple spinlock. We have to keep checking 521 * panic_cpu if we are spinning in case the panic on the first 522 * CPU is canceled. 523 */ 524 if (panic_cpu != PCPU_GET(cpuid)) 525 while (atomic_cmpset_int(&panic_cpu, NOCPU, 526 PCPU_GET(cpuid)) == 0) 527 while (panic_cpu != NOCPU) 528 ; /* nothing */ 529 #endif 530 531 bootopt = RB_AUTOBOOT | RB_DUMP; 532 newpanic = 0; 533 if (panicstr) 534 bootopt |= RB_NOSYNC; 535 else { 536 panicstr = fmt; 537 newpanic = 1; 538 } 539 540 va_start(ap, fmt); 541 if (newpanic) { 542 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 543 panicstr = buf; 544 printf("panic: %s\n", buf); 545 } else { 546 printf("panic: "); 547 vprintf(fmt, ap); 548 printf("\n"); 549 } 550 va_end(ap); 551 #ifdef SMP 552 printf("cpuid = %d\n", PCPU_GET(cpuid)); 553 #endif 554 555 #ifdef KDB 556 if (newpanic && trace_on_panic) 557 kdb_backtrace(); 558 if (debugger_on_panic) 559 kdb_enter(KDB_WHY_PANIC, "panic"); 560 #ifdef RESTARTABLE_PANICS 561 /* See if the user aborted the panic, in which case we continue. */ 562 if (panicstr == NULL) { 563 #ifdef SMP 564 atomic_store_rel_int(&panic_cpu, NOCPU); 565 #endif 566 return; 567 } 568 #endif 569 #endif 570 /*thread_lock(td); */ 571 td->td_flags |= TDF_INPANIC; 572 /* thread_unlock(td); */ 573 if (!sync_on_panic) 574 bootopt |= RB_NOSYNC; 575 critical_exit(); 576 boot(bootopt); 577 } 578 579 /* 580 * Support for poweroff delay. 581 */ 582 #ifndef POWEROFF_DELAY 583 # define POWEROFF_DELAY 5000 584 #endif 585 static int poweroff_delay = POWEROFF_DELAY; 586 587 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 588 &poweroff_delay, 0, ""); 589 590 static void 591 poweroff_wait(void *junk, int howto) 592 { 593 594 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 595 return; 596 DELAY(poweroff_delay * 1000); 597 } 598 599 /* 600 * Some system processes (e.g. syncer) need to be stopped at appropriate 601 * points in their main loops prior to a system shutdown, so that they 602 * won't interfere with the shutdown process (e.g. by holding a disk buf 603 * to cause sync to fail). For each of these system processes, register 604 * shutdown_kproc() as a handler for one of shutdown events. 605 */ 606 static int kproc_shutdown_wait = 60; 607 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 608 &kproc_shutdown_wait, 0, ""); 609 610 void 611 kproc_shutdown(void *arg, int howto) 612 { 613 struct proc *p; 614 char procname[MAXCOMLEN + 1]; 615 int error; 616 617 if (panicstr) 618 return; 619 620 p = (struct proc *)arg; 621 strlcpy(procname, p->p_comm, sizeof(procname)); 622 printf("Waiting (max %d seconds) for system process `%s' to stop...", 623 kproc_shutdown_wait, procname); 624 error = kproc_suspend(p, kproc_shutdown_wait * hz); 625 626 if (error == EWOULDBLOCK) 627 printf("timed out\n"); 628 else 629 printf("done\n"); 630 } 631 632 void 633 kthread_shutdown(void *arg, int howto) 634 { 635 struct thread *td; 636 char procname[MAXCOMLEN + 1]; 637 int error; 638 639 if (panicstr) 640 return; 641 642 td = (struct thread *)arg; 643 strlcpy(procname, td->td_name, sizeof(procname)); 644 printf("Waiting (max %d seconds) for system thread `%s' to stop...", 645 kproc_shutdown_wait, procname); 646 error = kthread_suspend(td, kproc_shutdown_wait * hz); 647 648 if (error == EWOULDBLOCK) 649 printf("timed out\n"); 650 else 651 printf("done\n"); 652 } 653 654 /* Registration of dumpers */ 655 int 656 set_dumper(struct dumperinfo *di) 657 { 658 659 if (di == NULL) { 660 bzero(&dumper, sizeof dumper); 661 return (0); 662 } 663 if (dumper.dumper != NULL) 664 return (EBUSY); 665 dumper = *di; 666 return (0); 667 } 668 669 /* Call dumper with bounds checking. */ 670 int 671 dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical, 672 off_t offset, size_t length) 673 { 674 675 if (length != 0 && (offset < di->mediaoffset || 676 offset - di->mediaoffset + length > di->mediasize)) { 677 printf("Attempt to write outside dump device boundaries.\n"); 678 return (ENXIO); 679 } 680 return (di->dumper(di->priv, virtual, physical, offset, length)); 681 } 682 683 #if defined(__powerpc__) 684 void 685 dumpsys(struct dumperinfo *di __unused) 686 { 687 688 printf("Kernel dumps not implemented on this architecture\n"); 689 } 690 #endif 691 692 void 693 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver, 694 uint64_t dumplen, uint32_t blksz) 695 { 696 697 bzero(kdh, sizeof(*kdh)); 698 strncpy(kdh->magic, magic, sizeof(kdh->magic)); 699 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture)); 700 kdh->version = htod32(KERNELDUMPVERSION); 701 kdh->architectureversion = htod32(archver); 702 kdh->dumplength = htod64(dumplen); 703 kdh->dumptime = htod64(time_second); 704 kdh->blocksize = htod32(blksz); 705 strncpy(kdh->hostname, G_hostname, sizeof(kdh->hostname)); 706 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring)); 707 if (panicstr != NULL) 708 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring)); 709 kdh->parity = kerneldump_parity(kdh); 710 } 711