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 #include <sys/vnode.h> 67 68 #include <machine/cpu.h> 69 #include <machine/pcb.h> 70 #include <machine/smp.h> 71 72 #include <sys/signalvar.h> 73 74 #ifndef PANIC_REBOOT_WAIT_TIME 75 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 76 #endif 77 78 /* 79 * Note that stdarg.h and the ANSI style va_start macro is used for both 80 * ANSI and traditional C compilers. 81 */ 82 #include <machine/stdarg.h> 83 84 #ifdef KDB 85 #ifdef KDB_UNATTENDED 86 int debugger_on_panic = 0; 87 #else 88 int debugger_on_panic = 1; 89 #endif 90 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 91 &debugger_on_panic, 0, "Run debugger on kernel panic"); 92 93 #ifdef KDB_TRACE 94 int trace_on_panic = 1; 95 #else 96 int trace_on_panic = 0; 97 #endif 98 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 99 &trace_on_panic, 0, "Print stack trace on kernel panic"); 100 #endif /* KDB */ 101 102 int sync_on_panic = 0; 103 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 104 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 105 106 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 107 108 /* 109 * Variable panicstr contains argument to first call to panic; used as flag 110 * to indicate that the kernel has already called panic. 111 */ 112 const char *panicstr; 113 114 int dumping; /* system is dumping */ 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 /* 241 * Shutdown the system cleanly to prepare for reboot, halt, or power off. 242 */ 243 static void 244 boot(int howto) 245 { 246 static int first_buf_printf = 1; 247 248 #if defined(SMP) 249 /* 250 * Bind us to CPU 0 so that all shutdown code runs there. Some 251 * systems don't shutdown properly (i.e., ACPI power off) if we 252 * run on another processor. 253 */ 254 mtx_lock_spin(&sched_lock); 255 sched_bind(curthread, 0); 256 mtx_unlock_spin(&sched_lock); 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 #ifdef SMP 266 if (smp_active) 267 printf("boot() called on cpu#%d\n", PCPU_GET(cpuid)); 268 #endif 269 /* 270 * Do any callouts that should be done BEFORE syncing the filesystems. 271 */ 272 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 273 274 /* 275 * Now sync filesystems 276 */ 277 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 278 register struct buf *bp; 279 int iter, nbusy, pbusy; 280 #ifndef PREEMPTION 281 int subiter; 282 #endif 283 284 waittime = 0; 285 286 sync(&thread0, NULL); 287 288 /* 289 * With soft updates, some buffers that are 290 * written will be remarked as dirty until other 291 * buffers are written. 292 */ 293 for (iter = pbusy = 0; iter < 20; iter++) { 294 nbusy = 0; 295 for (bp = &buf[nbuf]; --bp >= buf; ) { 296 if ((bp->b_flags & B_INVAL) == 0 && 297 BUF_REFCNT(bp) > 0) { 298 nbusy++; 299 } else if ((bp->b_flags & (B_DELWRI | B_INVAL)) 300 == B_DELWRI) { 301 /* bawrite(bp);*/ 302 nbusy++; 303 } 304 } 305 if (nbusy == 0) { 306 if (first_buf_printf) 307 printf("All buffers synced."); 308 break; 309 } 310 if (first_buf_printf) { 311 printf("Syncing disks, buffers remaining... "); 312 first_buf_printf = 0; 313 } 314 printf("%d ", nbusy); 315 if (nbusy < pbusy) 316 iter = 0; 317 pbusy = nbusy; 318 sync(&thread0, NULL); 319 320 #ifdef PREEMPTION 321 /* 322 * Drop Giant and spin for a while to allow 323 * interrupt threads to run. 324 */ 325 DROP_GIANT(); 326 DELAY(50000 * iter); 327 PICKUP_GIANT(); 328 #else 329 /* 330 * Drop Giant and context switch several times to 331 * allow interrupt threads to run. 332 */ 333 DROP_GIANT(); 334 for (subiter = 0; subiter < 50 * iter; subiter++) { 335 mtx_lock_spin(&sched_lock); 336 mi_switch(SW_VOL, NULL); 337 mtx_unlock_spin(&sched_lock); 338 DELAY(1000); 339 } 340 PICKUP_GIANT(); 341 #endif 342 } 343 printf("\n"); 344 /* 345 * Count only busy local buffers to prevent forcing 346 * a fsck if we're just a client of a wedged NFS server 347 */ 348 nbusy = 0; 349 for (bp = &buf[nbuf]; --bp >= buf; ) { 350 if (((bp->b_flags&B_INVAL) == 0 && BUF_REFCNT(bp)) || 351 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 352 #if 0 353 /* XXX: This is bogus. We should probably have a BO_REMOTE flag instead */ 354 if (bp->b_dev == NULL) { 355 TAILQ_REMOVE(&mountlist, 356 bp->b_vp->v_mount, mnt_list); 357 continue; 358 } 359 #endif 360 nbusy++; 361 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 362 printf( 363 "%d: bufobj:%p, flags:%0x, blkno:%ld, lblkno:%ld\n", 364 nbusy, bp->b_bufobj, 365 bp->b_flags, (long)bp->b_blkno, 366 (long)bp->b_lblkno); 367 #endif 368 } 369 } 370 if (nbusy) { 371 /* 372 * Failed to sync all blocks. Indicate this and don't 373 * unmount filesystems (thus forcing an fsck on reboot). 374 */ 375 printf("Giving up on %d buffers\n", nbusy); 376 DELAY(5000000); /* 5 seconds */ 377 } else { 378 if (!first_buf_printf) 379 printf("Final sync complete\n"); 380 /* 381 * Unmount filesystems 382 */ 383 if (panicstr == 0) 384 vfs_unmountall(); 385 } 386 DELAY(100000); /* wait for console output to finish */ 387 } 388 389 print_uptime(); 390 391 /* 392 * Ok, now do things that assume all filesystem activity has 393 * been completed. 394 */ 395 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 396 397 /* XXX This doesn't disable interrupts any more. Reconsider? */ 398 splhigh(); 399 400 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) 401 doadump(); 402 403 /* Now that we're going to really halt the system... */ 404 EVENTHANDLER_INVOKE(shutdown_final, howto); 405 406 for(;;) ; /* safety against shutdown_reset not working */ 407 /* NOTREACHED */ 408 } 409 410 /* 411 * If the shutdown was a clean halt, behave accordingly. 412 */ 413 static void 414 shutdown_halt(void *junk, int howto) 415 { 416 417 if (howto & RB_HALT) { 418 printf("\n"); 419 printf("The operating system has halted.\n"); 420 printf("Please press any key to reboot.\n\n"); 421 switch (cngetc()) { 422 case -1: /* No console, just die */ 423 cpu_halt(); 424 /* NOTREACHED */ 425 default: 426 howto &= ~RB_HALT; 427 break; 428 } 429 } 430 } 431 432 /* 433 * Check to see if the system paniced, pause and then reboot 434 * according to the specified delay. 435 */ 436 static void 437 shutdown_panic(void *junk, int howto) 438 { 439 int loop; 440 441 if (howto & RB_DUMP) { 442 if (PANIC_REBOOT_WAIT_TIME != 0) { 443 if (PANIC_REBOOT_WAIT_TIME != -1) { 444 printf("Automatic reboot in %d seconds - " 445 "press a key on the console to abort\n", 446 PANIC_REBOOT_WAIT_TIME); 447 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 448 loop > 0; --loop) { 449 DELAY(1000 * 100); /* 1/10th second */ 450 /* Did user type a key? */ 451 if (cncheckc() != -1) 452 break; 453 } 454 if (!loop) 455 return; 456 } 457 } else { /* zero time specified - reboot NOW */ 458 return; 459 } 460 printf("--> Press a key on the console to reboot,\n"); 461 printf("--> or switch off the system now.\n"); 462 cngetc(); 463 } 464 } 465 466 /* 467 * Everything done, now reset 468 */ 469 static void 470 shutdown_reset(void *junk, int howto) 471 { 472 473 printf("Rebooting...\n"); 474 DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ 475 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 476 cpu_reset(); 477 /* NOTREACHED */ /* assuming reset worked */ 478 } 479 480 #ifdef SMP 481 static u_int panic_cpu = NOCPU; 482 #endif 483 484 /* 485 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 486 * and then reboots. If we are called twice, then we avoid trying to sync 487 * the disks as this often leads to recursive panics. 488 * 489 * MPSAFE 490 */ 491 void 492 panic(const char *fmt, ...) 493 { 494 struct thread *td = curthread; 495 int bootopt, newpanic; 496 va_list ap; 497 static char buf[256]; 498 499 #ifdef SMP 500 /* 501 * We don't want multiple CPU's to panic at the same time, so we 502 * use panic_cpu as a simple spinlock. We have to keep checking 503 * panic_cpu if we are spinning in case the panic on the first 504 * CPU is canceled. 505 */ 506 if (panic_cpu != PCPU_GET(cpuid)) 507 while (atomic_cmpset_int(&panic_cpu, NOCPU, 508 PCPU_GET(cpuid)) == 0) 509 while (panic_cpu != NOCPU) 510 ; /* nothing */ 511 #endif 512 513 bootopt = RB_AUTOBOOT | RB_DUMP; 514 newpanic = 0; 515 if (panicstr) 516 bootopt |= RB_NOSYNC; 517 else { 518 panicstr = fmt; 519 newpanic = 1; 520 } 521 522 va_start(ap, fmt); 523 if (newpanic) { 524 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 525 panicstr = buf; 526 printf("panic: %s\n", buf); 527 } else { 528 printf("panic: "); 529 vprintf(fmt, ap); 530 printf("\n"); 531 } 532 va_end(ap); 533 #ifdef SMP 534 printf("cpuid = %d\n", PCPU_GET(cpuid)); 535 #endif 536 537 #ifdef KDB 538 if (newpanic && trace_on_panic) 539 kdb_backtrace(); 540 if (debugger_on_panic) 541 kdb_enter("panic"); 542 #ifdef RESTARTABLE_PANICS 543 /* See if the user aborted the panic, in which case we continue. */ 544 if (panicstr == NULL) { 545 #ifdef SMP 546 atomic_store_rel_int(&panic_cpu, NOCPU); 547 #endif 548 return; 549 } 550 #endif 551 #endif 552 mtx_lock_spin(&sched_lock); 553 td->td_flags |= TDF_INPANIC; 554 mtx_unlock_spin(&sched_lock); 555 if (!sync_on_panic) 556 bootopt |= RB_NOSYNC; 557 boot(bootopt); 558 } 559 560 /* 561 * Support for poweroff delay. 562 */ 563 #ifndef POWEROFF_DELAY 564 # define POWEROFF_DELAY 5000 565 #endif 566 static int poweroff_delay = POWEROFF_DELAY; 567 568 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 569 &poweroff_delay, 0, ""); 570 571 static void 572 poweroff_wait(void *junk, int howto) 573 { 574 575 if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) 576 return; 577 DELAY(poweroff_delay * 1000); 578 } 579 580 /* 581 * Some system processes (e.g. syncer) need to be stopped at appropriate 582 * points in their main loops prior to a system shutdown, so that they 583 * won't interfere with the shutdown process (e.g. by holding a disk buf 584 * to cause sync to fail). For each of these system processes, register 585 * shutdown_kproc() as a handler for one of shutdown events. 586 */ 587 static int kproc_shutdown_wait = 60; 588 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 589 &kproc_shutdown_wait, 0, ""); 590 591 void 592 kproc_shutdown(void *arg, int howto) 593 { 594 struct proc *p; 595 char procname[MAXCOMLEN + 1]; 596 int error; 597 598 if (panicstr) 599 return; 600 601 p = (struct proc *)arg; 602 strlcpy(procname, p->p_comm, sizeof(procname)); 603 printf("Waiting (max %d seconds) for system process `%s' to stop...", 604 kproc_shutdown_wait, procname); 605 error = kthread_suspend(p, kproc_shutdown_wait * hz); 606 607 if (error == EWOULDBLOCK) 608 printf("timed out\n"); 609 else 610 printf("done\n"); 611 } 612 613 /* Registration of dumpers */ 614 int 615 set_dumper(struct dumperinfo *di) 616 { 617 618 if (di == NULL) { 619 bzero(&dumper, sizeof dumper); 620 return (0); 621 } 622 if (dumper.dumper != NULL) 623 return (EBUSY); 624 dumper = *di; 625 return (0); 626 } 627 628 #if defined(__powerpc__) 629 void 630 dumpsys(struct dumperinfo *di __unused) 631 { 632 633 printf("Kernel dumps not implemented on this architecture\n"); 634 } 635 #endif 636