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