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 static struct dumperinfo dumper; /* our selected dumper */ 115 116 /* Context information for dump-debuggers. */ 117 static struct pcb dumppcb; /* Registers. */ 118 static lwpid_t dumptid; /* Thread ID. */ 119 120 static void boot(int) __dead2; 121 static void poweroff_wait(void *, int); 122 static void shutdown_halt(void *junk, int howto); 123 static void shutdown_panic(void *junk, int howto); 124 static void shutdown_reset(void *junk, int howto); 125 126 /* register various local shutdown events */ 127 static void 128 shutdown_conf(void *unused) 129 { 130 131 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, 132 SHUTDOWN_PRI_FIRST); 133 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, 134 SHUTDOWN_PRI_LAST + 100); 135 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, 136 SHUTDOWN_PRI_LAST + 100); 137 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, 138 SHUTDOWN_PRI_LAST + 200); 139 } 140 141 SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL) 142 143 /* 144 * The system call that results in a reboot 145 * 146 * MPSAFE 147 */ 148 /* ARGSUSED */ 149 int 150 reboot(struct thread *td, struct reboot_args *uap) 151 { 152 int error; 153 154 error = 0; 155 #ifdef MAC 156 error = mac_check_system_reboot(td->td_ucred, uap->opt); 157 #endif 158 if (error == 0) 159 error = suser(td); 160 if (error == 0) { 161 mtx_lock(&Giant); 162 boot(uap->opt); 163 mtx_unlock(&Giant); 164 } 165 return (error); 166 } 167 168 /* 169 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 170 */ 171 static int shutdown_howto = 0; 172 173 void 174 shutdown_nice(int howto) 175 { 176 177 shutdown_howto = howto; 178 179 /* Send a signal to init(8) and have it shutdown the world */ 180 if (initproc != NULL) { 181 PROC_LOCK(initproc); 182 psignal(initproc, SIGINT); 183 PROC_UNLOCK(initproc); 184 } else { 185 /* No init(8) running, so simply reboot */ 186 boot(RB_NOSYNC); 187 } 188 return; 189 } 190 static int waittime = -1; 191 192 static void 193 print_uptime(void) 194 { 195 int f; 196 struct timespec ts; 197 198 getnanouptime(&ts); 199 printf("Uptime: "); 200 f = 0; 201 if (ts.tv_sec >= 86400) { 202 printf("%ldd", (long)ts.tv_sec / 86400); 203 ts.tv_sec %= 86400; 204 f = 1; 205 } 206 if (f || ts.tv_sec >= 3600) { 207 printf("%ldh", (long)ts.tv_sec / 3600); 208 ts.tv_sec %= 3600; 209 f = 1; 210 } 211 if (f || ts.tv_sec >= 60) { 212 printf("%ldm", (long)ts.tv_sec / 60); 213 ts.tv_sec %= 60; 214 f = 1; 215 } 216 printf("%lds\n", (long)ts.tv_sec); 217 } 218 219 static void 220 doadump(void) 221 { 222 223 /* 224 * Sometimes people have to call this from the kernel debugger. 225 * (if 'panic' can not dump) 226 * Give them a clue as to why they can't dump. 227 */ 228 if (dumper.dumper == NULL) { 229 printf("Cannot dump. No dump device defined.\n"); 230 return; 231 } 232 233 savectx(&dumppcb); 234 dumptid = curthread->td_tid; 235 dumping++; 236 dumpsys(&dumper); 237 } 238 239 /* 240 * Shutdown the system cleanly to prepare for reboot, halt, or power off. 241 */ 242 static void 243 boot(int howto) 244 { 245 static int first_buf_printf = 1; 246 247 #if defined(SMP) 248 /* 249 * Bind us to CPU 0 so that all shutdown code runs there. Some 250 * systems don't shutdown properly (i.e., ACPI power off) if we 251 * run on another processor. 252 */ 253 mtx_lock_spin(&sched_lock); 254 sched_bind(curthread, 0); 255 mtx_unlock_spin(&sched_lock); 256 KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0")); 257 #endif 258 259 /* collect extra flags that shutdown_nice might have set */ 260 howto |= shutdown_howto; 261 262 /* We are out of the debugger now. */ 263 kdb_active = 0; 264 265 /* 266 * Do any callouts that should be done BEFORE syncing the filesystems. 267 */ 268 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 269 270 /* 271 * Now sync filesystems 272 */ 273 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 274 register struct buf *bp; 275 int iter, nbusy, pbusy; 276 #ifndef PREEMPTION 277 int subiter; 278 #endif 279 280 waittime = 0; 281 282 sync(&thread0, NULL); 283 284 /* 285 * With soft updates, some buffers that are 286 * written will be remarked as dirty until other 287 * buffers are written. 288 */ 289 for (iter = pbusy = 0; iter < 20; iter++) { 290 nbusy = 0; 291 for (bp = &buf[nbuf]; --bp >= buf; ) { 292 if ((bp->b_flags & B_INVAL) == 0 && 293 BUF_REFCNT(bp) > 0) { 294 nbusy++; 295 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 296 == B_DELWRI) { 297 /* bawrite(bp);*/ 298 nbusy++; 299 } 300 } 301 if (nbusy == 0) { 302 if (first_buf_printf) 303 printf("All buffers synced."); 304 break; 305 } 306 if (first_buf_printf) { 307 printf("Syncing disks, buffers remaining... "); 308 first_buf_printf = 0; 309 } 310 printf("%d ", nbusy); 311 if (nbusy < pbusy) 312 iter = 0; 313 pbusy = nbusy; 314 sync(&thread0, NULL); 315 316 #ifdef PREEMPTION 317 /* 318 * Drop Giant and spin for a while to allow 319 * interrupt threads to run. 320 */ 321 DROP_GIANT(); 322 DELAY(50000 * iter); 323 PICKUP_GIANT(); 324 #else 325 /* 326 * Drop Giant and context switch several times to 327 * allow interrupt threads to run. 328 */ 329 DROP_GIANT(); 330 for (subiter = 0; subiter < 50 * iter; subiter++) { 331 mtx_lock_spin(&sched_lock); 332 mi_switch(SW_VOL, NULL); 333 mtx_unlock_spin(&sched_lock); 334 DELAY(1000); 335 } 336 PICKUP_GIANT(); 337 #endif 338 } 339 printf("\n"); 340 /* 341 * Count only busy local buffers to prevent forcing 342 * a fsck if we're just a client of a wedged NFS server 343 */ 344 nbusy = 0; 345 for (bp = &buf[nbuf]; --bp >= buf; ) { 346 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 347 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 348 #if 0 349 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 350 if (bp->b_dev == NULL) { 351 TAILQ_REMOVE(&mountlist, 352 bp->b_vp->v_mount, mnt_list); 353 continue; 354 } 355 #endif 356 nbusy++; 357 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 358 printf( 359 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n", 360 nbusy, bp->b_bufobj, 361 bp->b_flags, (long)bp->b_blkno, 362 (long)bp->b_lblkno); 363 #endif 364 } 365 } 366 if (nbusy) { 367 /* 368 * Failed to sync all blocks. Indicate this and don't 369 * unmount filesystems (thus forcing an fsck on reboot). 370 */ 371 printf("Giving up on %d buffers\n", nbusy); 372 DELAY(5000000); /* 5 seconds */ 373 } else { 374 if (!first_buf_printf) 375 printf("Final sync complete\n"); 376 /* 377 * Unmount filesystems 378 */ 379 if (panicstr == 0) 380 vfs_unmountall(); 381 } 382 DELAY(100000); /* wait for console output to finish */ 383 } 384 385 print_uptime(); 386 387 /* 388 * Ok, now do things that assume all filesystem activity has 389 * been completed. 390 */ 391 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 392 393 /* XXX This doesn't disable interrupts any more. Reconsider? */ 394 splhigh(); 395 396 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 397 doadump(); 398 399 /* Now that we're going to really halt the system... */ 400 EVENTHANDLER_INVOKE(shutdown_final, howto); 401 402 for(;;) ; /* safety against shutdown_reset not working */ 403 /* NOTREACHED */ 404 } 405 406 /* 407 * If the shutdown was a clean halt, behave accordingly. 408 */ 409 static void 410 shutdown_halt(void *junk, int howto) 411 { 412 413 if (howto & RB_HALT) { 414 printf("\n"); 415 printf("The operating system has halted.\n"); 416 printf("Please press any key to reboot.\n\n"); 417 switch (cngetc()) { 418 case -1: /* No console, just die */ 419 cpu_halt(); 420 /* NOTREACHED */ 421 default: 422 howto &= ~RB_HALT; 423 break; 424 } 425 } 426 } 427 428 /* 429 * Check to see if the system paniced, pause and then reboot 430 * according to the specified delay. 431 */ 432 static void 433 shutdown_panic(void *junk, int howto) 434 { 435 int loop; 436 437 if (howto & RB_DUMP) { 438 if (PANIC_REBOOT_WAIT_TIME != 0) { 439 if (PANIC_REBOOT_WAIT_TIME != -1) { 440 printf("Automatic reboot in %d seconds - " 441 "press a key on the console to abort\n", 442 PANIC_REBOOT_WAIT_TIME); 443 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 444 loop > 0; --loop) { 445 DELAY(1000 * 100); /* 1/10th second */ 446 /* Did user type a key? */ 447 if (cncheckc() != -1) 448 break; 449 } 450 if (!loop) 451 return; 452 } 453 } else { /* zero time specified - reboot NOW */ 454 return; 455 } 456 printf("--> Press a key on the console to reboot,\n"); 457 printf("--> or switch off the system now.\n"); 458 cngetc(); 459 } 460 } 461 462 /* 463 * Everything done, now reset 464 */ 465 static void 466 shutdown_reset(void *junk, int howto) 467 { 468 469 printf("Rebooting...\n"); 470 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 471 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 472 cpu_reset(); 473 /* NOTREACHED */ /* assuming reset worked */ 474 } 475 476 #ifdef SMP 477 static u_int panic_cpu = NOCPU; 478 #endif 479 480 /* 481 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 482 * and then reboots. If we are called twice, then we avoid trying to sync 483 * the disks as this often leads to recursive panics. 484 * 485 * MPSAFE 486 */ 487 void 488 panic(const char *fmt, ...) 489 { 490 struct thread *td = curthread; 491 int bootopt, newpanic; 492 va_list ap; 493 static char buf[256]; 494 495 #ifdef SMP 496 /* 497 * We don't want multiple CPU's to panic at the same time, so we 498 * use panic_cpu as a simple spinlock. We have to keep checking 499 * panic_cpu if we are spinning in case the panic on the first 500 * CPU is canceled. 501 */ 502 if (panic_cpu != PCPU_GET(cpuid)) 503 while (atomic_cmpset_int(&panic_cpu, NOCPU, 504 PCPU_GET(cpuid)) == 0) 505 while (panic_cpu != NOCPU) 506 ; /* nothing */ 507 #endif 508 509 bootopt = RB_AUTOBOOT | RB_DUMP; 510 newpanic = 0; 511 if (panicstr) 512 bootopt |= RB_NOSYNC; 513 else { 514 panicstr = fmt; 515 newpanic = 1; 516 } 517 518 va_start(ap, fmt); 519 if (newpanic) { 520 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 521 panicstr = buf; 522 printf("panic: %s\n", buf); 523 } else { 524 printf("panic: "); 525 vprintf(fmt, ap); 526 printf("\n"); 527 } 528 va_end(ap); 529 #ifdef SMP 530 printf("cpuid = %d\n", PCPU_GET(cpuid)); 531 #endif 532 533 #ifdef KDB 534 if (newpanic && trace_on_panic) 535 kdb_backtrace(); 536 if (debugger_on_panic) 537 kdb_enter("panic"); 538 #ifdef RESTARTABLE_PANICS 539 /* See if the user aborted the panic, in which case we continue. */ 540 if (panicstr == NULL) { 541 #ifdef SMP 542 atomic_store_rel_int(&panic_cpu, NOCPU); 543 #endif 544 return; 545 } 546 #endif 547 #endif 548 mtx_lock_spin(&sched_lock); 549 td->td_flags |= TDF_INPANIC; 550 mtx_unlock_spin(&sched_lock); 551 if (!sync_on_panic) 552 bootopt |= RB_NOSYNC; 553 boot(bootopt); 554 } 555 556 /* 557 * Support for poweroff delay. 558 */ 559 #ifndef POWEROFF_DELAY 560 # define POWEROFF_DELAY 5000 561 #endif 562 static int poweroff_delay = POWEROFF_DELAY; 563 564 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 565 &poweroff_delay, 0, ""); 566 567 static void 568 poweroff_wait(void *junk, int howto) 569 { 570 571 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 572 return; 573 DELAY(poweroff_delay * 1000); 574 } 575 576 /* 577 * Some system processes (e.g. syncer) need to be stopped at appropriate 578 * points in their main loops prior to a system shutdown, so that they 579 * won't interfere with the shutdown process (e.g. by holding a disk buf 580 * to cause sync to fail). For each of these system processes, register 581 * shutdown_kproc() as a handler for one of shutdown events. 582 */ 583 static int kproc_shutdown_wait = 60; 584 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 585 &kproc_shutdown_wait, 0, ""); 586 587 void 588 kproc_shutdown(void *arg, int howto) 589 { 590 struct proc *p; 591 char procname[MAXCOMLEN + 1]; 592 int error; 593 594 if (panicstr) 595 return; 596 597 p = (struct proc *)arg; 598 strlcpy(procname, p->p_comm, sizeof(procname)); 599 printf("Waiting (max %d seconds) for system process `%s' to stop...", 600 kproc_shutdown_wait, procname); 601 error = kthread_suspend(p, kproc_shutdown_wait * hz); 602 603 if (error == EWOULDBLOCK) 604 printf("timed out\n"); 605 else 606 printf("done\n"); 607 } 608 609 /* Registration of dumpers */ 610 int 611 set_dumper(struct dumperinfo *di) 612 { 613 614 if (di == NULL) { 615 bzero(&dumper, sizeof dumper); 616 return (0); 617 } 618 if (dumper.dumper != NULL) 619 return (EBUSY); 620 dumper = *di; 621 return (0); 622 } 623 624 #if defined(__powerpc__) 625 void 626 dumpsys(struct dumperinfo *di __unused) 627 { 628 629 printf("Kernel dumps not implemented on this architecture\n"); 630 } 631 #endif 632