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_kdb.h" 41 #include "opt_mac.h" 42 #include "opt_panic.h" 43 #include "opt_show_busybufs.h" 44 #include "opt_sched.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/kdb.h> 54 #include <sys/kernel.h> 55 #include <sys/kthread.h> 56 #include <sys/mac.h> 57 #include <sys/malloc.h> 58 #include <sys/mount.h> 59 #include <sys/proc.h> 60 #include <sys/reboot.h> 61 #include <sys/resourcevar.h> 62 #include <sys/sched.h> 63 #include <sys/smp.h> /* smp_active */ 64 #include <sys/sysctl.h> 65 #include <sys/sysproto.h> 66 67 #include <machine/cpu.h> 68 #include <machine/pcb.h> 69 #include <machine/smp.h> 70 71 #include <sys/signalvar.h> 72 73 #ifndef PANIC_REBOOT_WAIT_TIME 74 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 75 #endif 76 77 /* 78 * Note that stdarg.h and the ANSI style va_start macro is used for both 79 * ANSI and traditional C compilers. 80 */ 81 #include <machine/stdarg.h> 82 83 #ifdef KDB 84 #ifdef KDB_UNATTENDED 85 int debugger_on_panic = 0; 86 #else 87 int debugger_on_panic = 1; 88 #endif 89 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 90 &debugger_on_panic, 0, "Run debugger on kernel panic"); 91 92 #ifdef KDB_TRACE 93 int trace_on_panic = 1; 94 #else 95 int trace_on_panic = 0; 96 #endif 97 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 98 &trace_on_panic, 0, "Print stack trace on kernel panic"); 99 #endif /* KDB */ 100 101 int sync_on_panic = 0; 102 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 103 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 104 105 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 106 107 /* 108 * Variable panicstr contains argument to first call to panic; used as flag 109 * to indicate that the kernel has already called panic. 110 */ 111 const char *panicstr; 112 113 int dumping; /* system is dumping */ 114 int rebooting; /* system is rebooting */ 115 static struct dumperinfo dumper; /* our selected dumper */ 116 117 /* Context information for dump-debuggers. */ 118 static struct pcb dumppcb; /* Registers. */ 119 static lwpid_t dumptid; /* Thread ID. */ 120 121 static void boot(int) __dead2; 122 static void poweroff_wait(void *, int); 123 static void shutdown_halt(void *junk, int howto); 124 static void shutdown_panic(void *junk, int howto); 125 static void shutdown_reset(void *junk, int howto); 126 127 /* register various local shutdown events */ 128 static void 129 shutdown_conf(void *unused) 130 { 131 132 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 133 SHUTDOWN_PRI_FIRST); 134 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 135 SHUTDOWN_PRI_LAST + 100); 136 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 137 SHUTDOWN_PRI_LAST + 100); 138 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 139 SHUTDOWN_PRI_LAST + 200); 140 } 141 142 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 143 144 /* 145 * The system call that results in a reboot 146 * 147 * MPSAFE 148 */ 149 /* ARGSUSED */ 150 int 151 reboot(struct thread *td, struct reboot_args *uap) 152 { 153 int error; 154 155 error = 0; 156 #ifdef MAC 157 error = mac_check_system_reboot(td->td_ucred, uap->opt); 158 #endif 159 if (error == 0) 160 error = suser(td); 161 if (error == 0) { 162 mtx_lock(&Giant); 163 boot(uap->opt); 164 mtx_unlock(&Giant); 165 } 166 return (error); 167 } 168 169 /* 170 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 171 */ 172 static int shutdown_howto = 0; 173 174 void 175 shutdown_nice(int howto) 176 { 177 178 shutdown_howto = howto; 179 180 /* Send a signal to init(8) and have it shutdown the world */ 181 if (initproc != NULL) { 182 PROC_LOCK(initproc); 183 psignal(initproc, SIGINT); 184 PROC_UNLOCK(initproc); 185 } else { 186 /* No init(8) running, so simply reboot */ 187 boot(RB_NOSYNC); 188 } 189 return; 190 } 191 static int waittime = -1; 192 193 static void 194 print_uptime(void) 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", (long)ts.tv_sec / 86400); 204 ts.tv_sec %= 86400; 205 f = 1; 206 } 207 if (f || ts.tv_sec >= 3600) { 208 printf("%ldh", (long)ts.tv_sec / 3600); 209 ts.tv_sec %= 3600; 210 f = 1; 211 } 212 if (f || ts.tv_sec >= 60) { 213 printf("%ldm", (long)ts.tv_sec / 60); 214 ts.tv_sec %= 60; 215 f = 1; 216 } 217 printf("%lds\n", (long)ts.tv_sec); 218 } 219 220 static void 221 doadump(void) 222 { 223 224 /* 225 * Sometimes people have to call this from the kernel debugger. 226 * (if 'panic' can not dump) 227 * Give them a clue as to why they can't dump. 228 */ 229 if (dumper.dumper == NULL) { 230 printf("Cannot dump. No dump device defined.\n"); 231 return; 232 } 233 234 savectx(&dumppcb); 235 dumptid = curthread->td_tid; 236 dumping++; 237 dumpsys(&dumper); 238 } 239 240 static int 241 isbufbusy(struct buf *bp) 242 { 243 if (((bp->b_flags & (B_INVAL | B_PERSISTENT)) == 0 && 244 BUF_REFCNT(bp) > 0) || 245 ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)) 246 return (1); 247 return (0); 248 } 249 250 /* 251 * Shutdown the system cleanly to prepare for reboot, halt, or power off. 252 */ 253 static void 254 boot(int howto) 255 { 256 static int first_buf_printf = 1; 257 258 #if defined(SMP) 259 /* 260 * Bind us to CPU 0 so that all shutdown code runs there. Some 261 * systems don't shutdown properly (i.e., ACPI power off) if we 262 * run on another processor. 263 */ 264 mtx_lock_spin(&sched_lock); 265 sched_bind(curthread, 0); 266 mtx_unlock_spin(&sched_lock); 267 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0")); 268 #endif 269 /* We're in the process of rebooting. */ 270 rebooting = 1; 271 272 /* collect extra flags that shutdown_nice might have set */ 273 howto |= shutdown_howto; 274 275 /* We are out of the debugger now. */ 276 kdb_active = 0; 277 278 /* 279 * Do any callouts that should be done BEFORE syncing the filesystems. 280 */ 281 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 282 283 /* 284 * Now sync filesystems 285 */ 286 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 287 register struct buf *bp; 288 int iter, nbusy, pbusy; 289 #ifndef PREEMPTION 290 int subiter; 291 #endif 292 293 waittime = 0; 294 295 sync(curthread, NULL); 296 297 /* 298 * With soft updates, some buffers that are 299 * written will be remarked as dirty until other 300 * buffers are written. 301 */ 302 for (iter = pbusy = 0; iter < 20; iter++) { 303 nbusy = 0; 304 for (bp = &buf[nbuf]; --bp >= buf; ) 305 if (isbufbusy(bp)) 306 nbusy++; 307 if (nbusy == 0) { 308 if (first_buf_printf) 309 printf("All buffers synced."); 310 break; 311 } 312 if (first_buf_printf) { 313 printf("Syncing disks, buffers remaining... "); 314 first_buf_printf = 0; 315 } 316 printf("%d ", nbusy); 317 if (nbusy < pbusy) 318 iter = 0; 319 pbusy = nbusy; 320 sync(curthread, NULL); 321 322 #ifdef PREEMPTION 323 /* 324 * Drop Giant and spin for a while to allow 325 * interrupt threads to run. 326 */ 327 DROP_GIANT(); 328 DELAY(50000 * iter); 329 PICKUP_GIANT(); 330 #else 331 /* 332 * Drop Giant and context switch several times to 333 * allow interrupt threads to run. 334 */ 335 DROP_GIANT(); 336 for (subiter = 0; subiter < 50 * iter; subiter++) { 337 mtx_lock_spin(&sched_lock); 338 mi_switch(SW_VOL, NULL); 339 mtx_unlock_spin(&sched_lock); 340 DELAY(1000); 341 } 342 PICKUP_GIANT(); 343 #endif 344 } 345 printf("\n"); 346 /* 347 * Count only busy local buffers to prevent forcing 348 * a fsck if we're just a client of a wedged NFS server 349 */ 350 nbusy = 0; 351 for (bp = &buf[nbuf]; --bp >= buf; ) { 352 if (isbufbusy(bp)) { 353 #if 0 354 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 355 if (bp->b_dev == NULL) { 356 TAILQ_REMOVE(&mountlist, 357 bp->b_vp->v_mount, mnt_list); 358 continue; 359 } 360 #endif 361 nbusy++; 362 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 363 printf( 364 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n", 365 nbusy, bp->b_bufobj, 366 bp->b_flags, (long)bp->b_blkno, 367 (long)bp->b_lblkno); 368 #endif 369 } 370 } 371 if (nbusy) { 372 /* 373 * Failed to sync all blocks. Indicate this and don't 374 * unmount filesystems (thus forcing an fsck on reboot). 375 */ 376 printf("Giving up on %d buffers\n", nbusy); 377 DELAY(5000000); /* 5 seconds */ 378 } else { 379 if (!first_buf_printf) 380 printf("Final sync complete\n"); 381 /* 382 * Unmount filesystems 383 */ 384 if (panicstr == 0) 385 vfs_unmountall(); 386 } 387 DELAY(100000); /* wait for console output to finish */ 388 } 389 390 print_uptime(); 391 392 /* 393 * Ok, now do things that assume all filesystem activity has 394 * been completed. 395 */ 396 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 397 398 /* XXX This doesn't disable interrupts any more. Reconsider? */ 399 splhigh(); 400 401 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 402 doadump(); 403 404 /* Now that we're going to really halt the system... */ 405 EVENTHANDLER_INVOKE(shutdown_final, howto); 406 407 for(;;) ; /* safety against shutdown_reset not working */ 408 /* NOTREACHED */ 409 } 410 411 /* 412 * If the shutdown was a clean halt, behave accordingly. 413 */ 414 static void 415 shutdown_halt(void *junk, int howto) 416 { 417 418 if (howto & RB_HALT) { 419 printf("\n"); 420 printf("The operating system has halted.\n"); 421 printf("Please press any key to reboot.\n\n"); 422 switch (cngetc()) { 423 case -1: /* No console, just die */ 424 cpu_halt(); 425 /* NOTREACHED */ 426 default: 427 howto &= ~RB_HALT; 428 break; 429 } 430 } 431 } 432 433 /* 434 * Check to see if the system paniced, pause and then reboot 435 * according to the specified delay. 436 */ 437 static void 438 shutdown_panic(void *junk, int howto) 439 { 440 int loop; 441 442 if (howto & RB_DUMP) { 443 if (PANIC_REBOOT_WAIT_TIME != 0) { 444 if (PANIC_REBOOT_WAIT_TIME != -1) { 445 printf("Automatic reboot in %d seconds - " 446 "press a key on the console to abort\n", 447 PANIC_REBOOT_WAIT_TIME); 448 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 449 loop > 0; --loop) { 450 DELAY(1000 * 100); /* 1/10th second */ 451 /* Did user type a key? */ 452 if (cncheckc() != -1) 453 break; 454 } 455 if (!loop) 456 return; 457 } 458 } else { /* zero time specified - reboot NOW */ 459 return; 460 } 461 printf("--> Press a key on the console to reboot,\n"); 462 printf("--> or switch off the system now.\n"); 463 cngetc(); 464 } 465 } 466 467 /* 468 * Everything done, now reset 469 */ 470 static void 471 shutdown_reset(void *junk, int howto) 472 { 473 474 printf("Rebooting...\n"); 475 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 476 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 477 cpu_reset(); 478 /* NOTREACHED */ /* assuming reset worked */ 479 } 480 481 #ifdef SMP 482 static u_int panic_cpu = NOCPU; 483 #endif 484 485 /* 486 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 487 * and then reboots. If we are called twice, then we avoid trying to sync 488 * the disks as this often leads to recursive panics. 489 * 490 * MPSAFE 491 */ 492 void 493 panic(const char *fmt, ...) 494 { 495 struct thread *td = curthread; 496 int bootopt, newpanic; 497 va_list ap; 498 static char buf[256]; 499 500 #ifdef SMP 501 /* 502 * We don't want multiple CPU's to panic at the same time, so we 503 * use panic_cpu as a simple spinlock. We have to keep checking 504 * panic_cpu if we are spinning in case the panic on the first 505 * CPU is canceled. 506 */ 507 if (panic_cpu != PCPU_GET(cpuid)) 508 while (atomic_cmpset_int(&panic_cpu, NOCPU, 509 PCPU_GET(cpuid)) == 0) 510 while (panic_cpu != NOCPU) 511 ; /* nothing */ 512 #endif 513 514 bootopt = RB_AUTOBOOT | RB_DUMP; 515 newpanic = 0; 516 if (panicstr) 517 bootopt |= RB_NOSYNC; 518 else { 519 panicstr = fmt; 520 newpanic = 1; 521 } 522 523 va_start(ap, fmt); 524 if (newpanic) { 525 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 526 panicstr = buf; 527 printf("panic: %s\n", buf); 528 } else { 529 printf("panic: "); 530 vprintf(fmt, ap); 531 printf("\n"); 532 } 533 va_end(ap); 534 #ifdef SMP 535 printf("cpuid = %d\n", PCPU_GET(cpuid)); 536 #endif 537 538 #ifdef KDB 539 if (newpanic && trace_on_panic) 540 kdb_backtrace(); 541 if (debugger_on_panic) 542 kdb_enter("panic"); 543 #ifdef RESTARTABLE_PANICS 544 /* See if the user aborted the panic, in which case we continue. */ 545 if (panicstr == NULL) { 546 #ifdef SMP 547 atomic_store_rel_int(&panic_cpu, NOCPU); 548 #endif 549 return; 550 } 551 #endif 552 #endif 553 mtx_lock_spin(&sched_lock); 554 td->td_flags |= TDF_INPANIC; 555 mtx_unlock_spin(&sched_lock); 556 if (!sync_on_panic) 557 bootopt |= RB_NOSYNC; 558 boot(bootopt); 559 } 560 561 /* 562 * Support for poweroff delay. 563 */ 564 #ifndef POWEROFF_DELAY 565 # define POWEROFF_DELAY 5000 566 #endif 567 static int poweroff_delay = POWEROFF_DELAY; 568 569 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 570 &poweroff_delay, 0, ""); 571 572 static void 573 poweroff_wait(void *junk, int howto) 574 { 575 576 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 577 return; 578 DELAY(poweroff_delay * 1000); 579 } 580 581 /* 582 * Some system processes (e.g. syncer) need to be stopped at appropriate 583 * points in their main loops prior to a system shutdown, so that they 584 * won't interfere with the shutdown process (e.g. by holding a disk buf 585 * to cause sync to fail). For each of these system processes, register 586 * shutdown_kproc() as a handler for one of shutdown events. 587 */ 588 static int kproc_shutdown_wait = 60; 589 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 590 &kproc_shutdown_wait, 0, ""); 591 592 void 593 kproc_shutdown(void *arg, int howto) 594 { 595 struct proc *p; 596 char procname[MAXCOMLEN + 1]; 597 int error; 598 599 if (panicstr) 600 return; 601 602 p = (struct proc *)arg; 603 strlcpy(procname, p->p_comm, sizeof(procname)); 604 printf("Waiting (max %d seconds) for system process `%s' to stop...", 605 kproc_shutdown_wait, procname); 606 error = kthread_suspend(p, kproc_shutdown_wait * hz); 607 608 if (error == EWOULDBLOCK) 609 printf("timed out\n"); 610 else 611 printf("done\n"); 612 } 613 614 /* Registration of dumpers */ 615 int 616 set_dumper(struct dumperinfo *di) 617 { 618 619 if (di == NULL) { 620 bzero(&dumper, sizeof dumper); 621 return (0); 622 } 623 if (dumper.dumper != NULL) 624 return (EBUSY); 625 dumper = *di; 626 return (0); 627 } 628 629 #if defined(__powerpc__) 630 void 631 dumpsys(struct dumperinfo *di __unused) 632 { 633 634 printf("Kernel dumps not implemented on this architecture\n"); 635 } 636 #endif 637