sys/kern/kern_shutdown.c

   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  * 3. All advertising materials mentioning features or use of this software
  19  *    must display the following acknowledgement:
  20  *      This product includes software developed by the University of
  21  *      California, Berkeley and its contributors.
  22  * 4. Neither the name of the University nor the names of its contributors
  23  *    may be used to endorse or promote products derived from this software
  24  *    without specific prior written permission.
  25  *
  26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  36  * SUCH DAMAGE.
  37  *
  38  *      @(#)kern_shutdown.c     8.3 (Berkeley) 1/21/94
  39  * $FreeBSD: src/sys/kern/kern_shutdown.c,v 1.72.2.12 2002/02/21 19:15:10 dillon Exp $
  40  * $DragonFly: src/sys/kern/kern_shutdown.c,v 1.61 2007/11/06 03:49:58 dillon Exp $
  41  */
  42
  43 #include "opt_ddb.h"
  44 #include "opt_ddb_trace.h"
  45 #include "opt_hw_wdog.h"
  46 #include "opt_panic.h"
  47 #include "opt_show_busybufs.h"
  48
  49 #include <sys/param.h>
  50 #include <sys/systm.h>
  51 #include <sys/eventhandler.h>
  52 #include <sys/buf.h>
  53 #include <sys/diskslice.h>
  54 #include <sys/reboot.h>
  55 #include <sys/proc.h>
  56 #include <sys/fcntl.h>          /* FREAD        */
  57 #include <sys/stat.h>           /* S_IFCHR      */
  58 #include <sys/vnode.h>
  59 #include <sys/kernel.h>
  60 #include <sys/kthread.h>
  61 #include <sys/malloc.h>
  62 #include <sys/mount.h>
  63 #include <sys/queue.h>
  64 #include <sys/sysctl.h>
  65 #include <sys/vkernel.h>
  66 #include <sys/conf.h>
  67 #include <sys/sysproto.h>
  68 #include <sys/device.h>
  69 #include <sys/cons.h>
  70 #include <sys/shm.h>
  71 #include <sys/kern_syscall.h>
  72 #include <vm/vm_map.h>
  73 #include <vm/pmap.h>
  74
  75 #include <sys/thread2.h>
  76 #include <sys/buf2.h>
  77
  78 #include <machine/pcb.h>
  79 #include <machine/clock.h>
  80 #include <machine/md_var.h>
  81 #include <machine/smp.h>                /* smp_active_mask, cpuid */
  82 #include <machine/vmparam.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.  We use the machine version to stay
  93  * within the confines of the kernel header files.
  94  */
  95 #include <machine/stdarg.h>
  96
  97 #ifdef DDB
  98 #ifdef DDB_UNATTENDED
  99 int debugger_on_panic = 0;
 100 #else
 101 int debugger_on_panic = 1;
 102 #endif
 103 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW,
 104         &debugger_on_panic, 0, "Run debugger on kernel panic");
 105
 106 extern void db_print_backtrace(void);
 107
 108 #ifdef DDB_TRACE
 109 int trace_on_panic = 1;
 110 #else
 111 int trace_on_panic = 0;
 112 #endif
 113 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW,
 114         &trace_on_panic, 0, "Print stack trace on kernel panic");
 115 #endif
 116
 117 static int sync_on_panic = 1;
 118 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW,
 119         &sync_on_panic, 0, "Do a sync before rebooting from a panic");
 120
 121 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment");
 122
 123 #ifdef  HW_WDOG
 124 /*
 125  * If there is a hardware watchdog, point this at the function needed to
 126  * hold it off.
 127  * It's needed when the kernel needs to do some lengthy operations.
 128  * e.g. in wd.c when dumping core.. It's most annoying to have
 129  * your precious core-dump only half written because the wdog kicked in.
 130  */
 131 watchdog_tickle_fn wdog_tickler = NULL;
 132 #endif  /* HW_WDOG */
 133
 134 /*
 135  * Variable panicstr contains argument to first call to panic; used as flag
 136  * to indicate that the kernel has already called panic.
 137  */
 138 const char *panicstr;
 139
 140 int dumping;                            /* system is dumping */
 141 #ifdef SMP
 142 u_int panic_cpu_interlock;              /* panic interlock */
 143 globaldata_t panic_cpu_gd;              /* which cpu took the panic */
 144 #endif
 145
 146 int bootverbose = 0;                    /* note: assignment to force non-bss */
 147 int cold = 1;                           /* note: assignment to force non-bss */
 148 int dumplo;                             /* OBSOLETE - savecore compat */
 149 u_int64_t dumplo64;
 150
 151 static void boot (int) __dead2;
 152 static void dumpsys (void);
 153 static int setdumpdev (cdev_t dev);
 154 static void poweroff_wait (void *, int);
 155 static void print_uptime (void);
 156 static void shutdown_halt (void *junk, int howto);
 157 static void shutdown_panic (void *junk, int howto);
 158 static void shutdown_reset (void *junk, int howto);
 159 static int shutdown_busycount1(struct buf *bp, void *info);
 160 static int shutdown_busycount2(struct buf *bp, void *info);
 161 static void shutdown_cleanup_proc(struct proc *p);
 162
 163 /* register various local shutdown events */
 164 static void
 165 shutdown_conf(void *unused)
 166 {
 167         EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST);
 168         EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100);
 169         EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100);
 170         EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200);
 171 }
 172
 173 SYSINIT(shutdown_conf, SI_BOOT2_MACHDEP, SI_ORDER_ANY, shutdown_conf, NULL)
 174
 175 /* ARGSUSED */
 176
 177 /*
 178  * The system call that results in a reboot
 179  */
 180 int
 181 sys_reboot(struct reboot_args *uap)
 182 {
 183         struct thread *td = curthread;
 184         int error;
 185
 186         if ((error = suser(td)))
 187                 return (error);
 188
 189         boot(uap->opt);
 190         return (0);
 191 }
 192
 193 /*
 194  * Called by events that want to shut down.. e.g  <CTL><ALT><DEL> on a PC
 195  */
 196 static int shutdown_howto = 0;
 197
 198 void
 199 shutdown_nice(int howto)
 200 {
 201         shutdown_howto = howto;
 202
 203         /* Send a signal to init(8) and have it shutdown the world */
 204         if (initproc != NULL) {
 205                 ksignal(initproc, SIGINT);
 206         } else {
 207                 /* No init(8) running, so simply reboot */
 208                 boot(RB_NOSYNC);
 209         }
 210         return;
 211 }
 212 static int      waittime = -1;
 213 static struct thread *dumpthread;
 214 static struct pcb dumppcb;
 215
 216 static void
 217 print_uptime(void)
 218 {
 219         int f;
 220         struct timespec ts;
 221
 222         getnanouptime(&ts);
 223         kprintf("Uptime: ");
 224         f = 0;
 225         if (ts.tv_sec >= 86400) {
 226                 kprintf("%ldd", ts.tv_sec / 86400);
 227                 ts.tv_sec %= 86400;
 228                 f = 1;
 229         }
 230         if (f || ts.tv_sec >= 3600) {
 231                 kprintf("%ldh", ts.tv_sec / 3600);
 232                 ts.tv_sec %= 3600;
 233                 f = 1;
 234         }
 235         if (f || ts.tv_sec >= 60) {
 236                 kprintf("%ldm", ts.tv_sec / 60);
 237                 ts.tv_sec %= 60;
 238                 f = 1;
 239         }
 240         kprintf("%lds\n", ts.tv_sec);
 241 }
 242
 243 /*
 244  *  Go through the rigmarole of shutting down..
 245  * this used to be in machdep.c but I'll be dammned if I could see
 246  * anything machine dependant in it.
 247  */
 248 static void
 249 boot(int howto)
 250 {
 251         /*
 252          * Get rid of any user scheduler baggage and then give
 253          * us a high priority.
 254          */
 255         if (curthread->td_release)
 256                 curthread->td_release(curthread);
 257         lwkt_setpri_self(TDPRI_MAX);
 258
 259         /* collect extra flags that shutdown_nice might have set */
 260         howto |= shutdown_howto;
 261
 262 #ifdef SMP
 263         /*
 264          * We really want to shutdown on the BSP.  Subsystems such as ACPI
 265          * can't power-down the box otherwise.
 266          */
 267         if (smp_active_mask > 1) {
 268                 kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid);
 269         }
 270         if (panicstr == NULL && mycpu->gd_cpuid != 0) {
 271                 kprintf("Switching to cpu #0 for shutdown\n");
 272                 lwkt_setcpu_self(globaldata_find(0));
 273         }
 274 #endif
 275         /*
 276          * Do any callouts that should be done BEFORE syncing the filesystems.
 277          */
 278         EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
 279
 280         /*
 281          * Try to get rid of any remaining FS references.  The calling
 282          * process, proc0, and init may still hold references.  The
 283          * VFS cache subsystem may still hold a root reference to root.
 284          *
 285          * XXX this needs work.  We really need to SIGSTOP all remaining
 286          * processes in order to avoid blowups due to proc0's filesystem
 287          * references going away.  For now just make sure that the init
 288          * process is stopped.
 289          */
 290         if (panicstr == NULL) {
 291                 shutdown_cleanup_proc(curproc);
 292                 shutdown_cleanup_proc(&proc0);
 293                 if (initproc) {
 294                         if (initproc != curproc) {
 295                                 ksignal(initproc, SIGSTOP);
 296                                 tsleep(boot, 0, "shutdn", hz / 20);
 297                         }
 298                         shutdown_cleanup_proc(initproc);
 299                 }
 300                 vfs_cache_setroot(NULL, NULL);
 301         }
 302
 303         /*
 304          * Now sync filesystems
 305          */
 306         if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) {
 307                 int iter, nbusy, pbusy;
 308
 309                 waittime = 0;
 310                 kprintf("\nsyncing disks... ");
 311
 312                 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
 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 = scan_all_buffers(shutdown_busycount1, NULL);
 321                         if (nbusy == 0)
 322                                 break;
 323                         kprintf("%d ", nbusy);
 324                         if (nbusy < pbusy)
 325                                 iter = 0;
 326                         pbusy = nbusy;
 327                         /*
 328                          * XXX:
 329                          * Process soft update work queue if buffers don't sync
 330                          * after 6 iterations by permitting the syncer to run.
 331                          */
 332                         if (iter > 5)
 333                                 bio_ops_sync(NULL);
 334
 335                         sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */
 336                         tsleep(boot, 0, "shutdn", hz * iter / 20 + 1);
 337                 }
 338                 kprintf("\n");
 339                 /*
 340                  * Count only busy local buffers to prevent forcing
 341                  * a fsck if we're just a client of a wedged NFS server
 342                  */
 343                 nbusy = scan_all_buffers(shutdown_busycount2, NULL);
 344                 if (nbusy) {
 345                         /*
 346                          * Failed to sync all blocks. Indicate this and don't
 347                          * unmount filesystems (thus forcing an fsck on reboot).
 348                          */
 349                         kprintf("giving up on %d buffers\n", nbusy);
 350 #ifdef DDB
 351                         Debugger("busy buffer problem");
 352 #endif /* DDB */
 353                         tsleep(boot, 0, "shutdn", hz * 5 + 1);
 354                 } else {
 355                         kprintf("done\n");
 356                         /*
 357                          * Unmount filesystems
 358                          */
 359                         if (panicstr == NULL)
 360                                 vfs_unmountall();
 361                 }
 362                 tsleep(boot, 0, "shutdn", hz / 10 + 1);
 363         }
 364
 365         print_uptime();
 366
 367         /*
 368          * Dump before doing post_sync shutdown ops
 369          */
 370         crit_enter();
 371         if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold)
 372                 dumpsys();
 373
 374         /*
 375          * Ok, now do things that assume all filesystem activity has
 376          * been completed.  This will also call the device shutdown
 377          * methods.
 378          */
 379         EVENTHANDLER_INVOKE(shutdown_post_sync, howto);
 380
 381         /* Now that we're going to really halt the system... */
 382         EVENTHANDLER_INVOKE(shutdown_final, howto);
 383
 384         for(;;) ;       /* safety against shutdown_reset not working */
 385         /* NOTREACHED */
 386 }
 387
 388 static int
 389 shutdown_busycount1(struct buf *bp, void *info)
 390 {
 391         if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0)
 392                 return(1);
 393         if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI)
 394                 return (1);
 395         return (0);
 396 }
 397
 398 static int
 399 shutdown_busycount2(struct buf *bp, void *info)
 400 {
 401         if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) ||
 402             ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) {
 403                 /*
 404                  * Only count buffers undergoing write I/O
 405                  * on the related vnode.
 406                  */
 407                 if (bp->b_vp == NULL ||
 408                     bp->b_vp->v_track_write.bk_active == 0) {
 409                         return (0);
 410                 }
 411 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC)
 412                 kprintf(
 413             "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n",
 414                     bp,
 415                     bp->b_flags, bp->b_loffset,
 416                     bp->b_bio2.bio_offset);
 417 #endif
 418                 return(1);
 419         }
 420         return(0);
 421 }
 422
 423 /*
 424  * If the shutdown was a clean halt, behave accordingly.
 425  */
 426 static void
 427 shutdown_halt(void *junk, int howto)
 428 {
 429         if (howto & RB_HALT) {
 430                 kprintf("\n");
 431                 kprintf("The operating system has halted.\n");
 432 #ifdef _KERNEL_VIRTUAL
 433                 cpu_halt();
 434 #else
 435                 kprintf("Please press any key to reboot.\n\n");
 436                 switch (cngetc()) {
 437                 case -1:                /* No console, just die */
 438                         cpu_halt();
 439                         /* NOTREACHED */
 440                 default:
 441                         howto &= ~RB_HALT;
 442                         break;
 443                 }
 444 #endif
 445         }
 446 }
 447
 448 /*
 449  * Check to see if the system paniced, pause and then reboot
 450  * according to the specified delay.
 451  */
 452 static void
 453 shutdown_panic(void *junk, int howto)
 454 {
 455         int loop;
 456
 457         if (howto & RB_DUMP) {
 458                 if (PANIC_REBOOT_WAIT_TIME != 0) {
 459                         if (PANIC_REBOOT_WAIT_TIME != -1) {
 460                                 kprintf("Automatic reboot in %d seconds - "
 461                                        "press a key on the console to abort\n",
 462                                         PANIC_REBOOT_WAIT_TIME);
 463                                 for (loop = PANIC_REBOOT_WAIT_TIME * 10;
 464                                      loop > 0; --loop) {
 465                                         DELAY(1000 * 100); /* 1/10th second */
 466                                         /* Did user type a key? */
 467                                         if (cncheckc() != -1)
 468                                                 break;
 469                                 }
 470                                 if (!loop)
 471                                         return;
 472                         }
 473                 } else { /* zero time specified - reboot NOW */
 474                         return;
 475                 }
 476                 kprintf("--> Press a key on the console to reboot,\n");
 477                 kprintf("--> or switch off the system now.\n");
 478                 cngetc();
 479         }
 480 }
 481
 482 /*
 483  * Everything done, now reset
 484  */
 485 static void
 486 shutdown_reset(void *junk, int howto)
 487 {
 488         kprintf("Rebooting...\n");
 489         DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */
 490         /* cpu_boot(howto); */ /* doesn't do anything at the moment */
 491         cpu_reset();
 492         /* NOTREACHED */ /* assuming reset worked */
 493 }
 494
 495 /*
 496  * Try to remove FS references in the specified process.  This function
 497  * is used during shutdown
 498  */
 499 static
 500 void
 501 shutdown_cleanup_proc(struct proc *p)
 502 {
 503         struct filedesc *fdp;
 504         struct vmspace *vm;
 505
 506         if (p == NULL)
 507                 return;
 508         if ((fdp = p->p_fd) != NULL) {
 509                 kern_closefrom(0);
 510                 if (fdp->fd_cdir) {
 511                         cache_drop(&fdp->fd_ncdir);
 512                         vrele(fdp->fd_cdir);
 513                         fdp->fd_cdir = NULL;
 514                 }
 515                 if (fdp->fd_rdir) {
 516                         cache_drop(&fdp->fd_nrdir);
 517                         vrele(fdp->fd_rdir);
 518                         fdp->fd_rdir = NULL;
 519                 }
 520                 if (fdp->fd_jdir) {
 521                         cache_drop(&fdp->fd_njdir);
 522                         vrele(fdp->fd_jdir);
 523                         fdp->fd_jdir = NULL;
 524                 }
 525         }
 526         if (p->p_vkernel)
 527                 vkernel_exit(p);
 528         if (p->p_textvp) {
 529                 vrele(p->p_textvp);
 530                 p->p_textvp = NULL;
 531         }
 532         vm = p->p_vmspace;
 533         if (vm != NULL) {
 534                 pmap_remove_pages(vmspace_pmap(vm),
 535                                   VM_MIN_USER_ADDRESS,
 536                                   VM_MAX_USER_ADDRESS);
 537                 vm_map_remove(&vm->vm_map,
 538                               VM_MIN_USER_ADDRESS,
 539                               VM_MAX_USER_ADDRESS);
 540         }
 541 }
 542
 543 /*
 544  * Magic number for savecore
 545  *
 546  * exported (symorder) and used at least by savecore(8)
 547  *
 548  * Mark it as used so that gcc doesn't optimize it away.
 549  */
 550 __attribute__((__used__))
 551         static u_long const dumpmag = 0x8fca0101UL;
 552
 553 static int      dumpsize = 0;           /* also for savecore */
 554
 555 static int      dodump = 1;
 556
 557 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0,
 558     "Try to perform coredump on kernel panic");
 559
 560 static int
 561 setdumpdev(cdev_t dev)
 562 {
 563         struct partinfo pinfo;
 564         u_int64_t newdumplo;
 565         int error;
 566         int doopen;
 567
 568         if (dev == NULL) {
 569                 dumpdev = dev;
 570                 return (0);
 571         }
 572         bzero(&pinfo, sizeof(pinfo));
 573
 574         /*
 575          * We have to open the device before we can perform ioctls on it,
 576          * or the slice/label data may not be present.  Device opens are
 577          * usually tracked by specfs, but the dump device can be set in
 578          * early boot and may not be open so this is somewhat of a hack.
 579          */
 580         doopen = (dev->si_sysref.refcnt == 1);
 581         if (doopen) {
 582                 error = dev_dopen(dev, FREAD, S_IFCHR, proc0.p_ucred);
 583                 if (error)
 584                         return (error);
 585         }
 586         error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
 587         if (doopen)
 588                 dev_dclose(dev, FREAD, S_IFCHR);
 589         if (error || pinfo.media_blocks == 0 || pinfo.media_blksize == 0)
 590                 return (ENXIO);
 591
 592         newdumplo = pinfo.media_blocks -
 593                     ((u_int64_t)Maxmem * PAGE_SIZE / DEV_BSIZE);
 594         if ((int64_t)newdumplo < (int64_t)pinfo.reserved_blocks)
 595                 return (ENOSPC);
 596         dumpdev = dev;
 597         dumplo64 = newdumplo;
 598         return (0);
 599 }
 600
 601
 602 /* ARGSUSED */
 603 static void dump_conf (void *dummy);
 604 static void
 605 dump_conf(void *dummy)
 606 {
 607         char *path;
 608         cdev_t dev;
 609
 610         path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK);
 611         if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) {
 612                 dev = kgetdiskbyname(path);
 613                 if (dev != NULL)
 614                         dumpdev = dev;
 615         }
 616         kfree(path, M_TEMP);
 617         if (setdumpdev(dumpdev) != 0)
 618                 dumpdev = NULL;
 619 }
 620
 621 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL)
 622
 623 static int
 624 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS)
 625 {
 626         int error;
 627         udev_t ndumpdev;
 628
 629         ndumpdev = dev2udev(dumpdev);
 630         error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req);
 631         if (error == 0 && req->newptr != NULL)
 632                 error = setdumpdev(udev2dev(ndumpdev, 0));
 633         return (error);
 634 }
 635
 636 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW,
 637         0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", "");
 638
 639 /*
 640  * Doadump comes here after turning off memory management and
 641  * getting on the dump stack, either when called above, or by
 642  * the auto-restart code.
 643  */
 644 static void
 645 dumpsys(void)
 646 {
 647         int     error;
 648
 649         savectx(&dumppcb);
 650         dumpthread = curthread;
 651         if (dumping++) {
 652                 kprintf("Dump already in progress, bailing...\n");
 653                 return;
 654         }
 655         if (!dodump)
 656                 return;
 657         if (dumpdev == NULL)
 658                 return;
 659         dumpsize = Maxmem;
 660         kprintf("\ndumping to dev %s, blockno %lld\n",
 661                 devtoname(dumpdev), dumplo64);
 662         kprintf("dump ");
 663         error = dev_ddump(dumpdev);
 664         if (error == 0) {
 665                 kprintf("succeeded\n");
 666                 return;
 667         }
 668         kprintf("failed, reason: ");
 669         switch (error) {
 670         case ENOSYS:
 671         case ENODEV:
 672                 kprintf("device doesn't support a dump routine\n");
 673                 break;
 674
 675         case ENXIO:
 676                 kprintf("device bad\n");
 677                 break;
 678
 679         case EFAULT:
 680                 kprintf("device not ready\n");
 681                 break;
 682
 683         case EINVAL:
 684                 kprintf("area improper\n");
 685                 break;
 686
 687         case EIO:
 688                 kprintf("i/o error\n");
 689                 break;
 690
 691         case EINTR:
 692                 kprintf("aborted from console\n");
 693                 break;
 694
 695         default:
 696                 kprintf("unknown, error = %d\n", error);
 697                 break;
 698         }
 699 }
 700
 701 int
 702 dumpstatus(vm_offset_t addr, off_t count)
 703 {
 704         int c;
 705
 706         if (addr % (1024 * 1024) == 0) {
 707 #ifdef HW_WDOG
 708                 if (wdog_tickler)
 709                         (*wdog_tickler)();
 710 #endif
 711                 kprintf("%ld ", (long)(count / (1024 * 1024)));
 712         }
 713
 714         if ((c = cncheckc()) == 0x03)
 715                 return -1;
 716         else if (c != -1)
 717                 kprintf("[CTRL-C to abort] ");
 718
 719         return 0;
 720 }
 721
 722 /*
 723  * Panic is called on unresolvable fatal errors.  It prints "panic: mesg",
 724  * and then reboots.  If we are called twice, then we avoid trying to sync
 725  * the disks as this often leads to recursive panics.
 726  */
 727 void
 728 panic(const char *fmt, ...)
 729 {
 730         int bootopt, newpanic;
 731         __va_list ap;
 732         static char buf[256];
 733
 734 #ifdef SMP
 735         /*
 736          * If a panic occurs on multiple cpus before the first is able to
 737          * halt the other cpus, only one cpu is allowed to take the panic.
 738          * Attempt to be verbose about this situation but if the kprintf()
 739          * itself panics don't let us overrun the kernel stack.
 740          *
 741          * Be very nasty about descheduling our thread at the lowest
 742          * level possible in an attempt to freeze the thread without
 743          * inducing further panics.
 744          *
 745          * Bumping gd_trap_nesting_level will also bypass assertions in
 746          * lwkt_switch() and allow us to switch away even if we are a
 747          * FAST interrupt or IPI.
 748          */
 749         if (atomic_poll_acquire_int(&panic_cpu_interlock)) {
 750                 panic_cpu_gd = mycpu;
 751         } else if (panic_cpu_gd != mycpu) {
 752                 crit_enter();
 753                 ++mycpu->gd_trap_nesting_level;
 754                 if (mycpu->gd_trap_nesting_level < 25) {
 755                         kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n",
 756                                 mycpu->gd_cpuid, curthread);
 757                 }
 758                 curthread->td_release = NULL;   /* be a grinch */
 759                 for (;;) {
 760                         lwkt_deschedule_self(curthread);
 761                         lwkt_switch();
 762                 }
 763                 /* NOT REACHED */
 764                 /* --mycpu->gd_trap_nesting_level */
 765                 /* crit_exit() */
 766         }
 767 #endif
 768         bootopt = RB_AUTOBOOT | RB_DUMP;
 769         if (sync_on_panic == 0)
 770                 bootopt |= RB_NOSYNC;
 771         newpanic = 0;
 772         if (panicstr)
 773                 bootopt |= RB_NOSYNC;
 774         else {
 775                 panicstr = fmt;
 776                 newpanic = 1;
 777         }
 778
 779         __va_start(ap, fmt);
 780         kvsnprintf(buf, sizeof(buf), fmt, ap);
 781         if (panicstr == fmt)
 782                 panicstr = buf;
 783         __va_end(ap);
 784         kprintf("panic: %s\n", buf);
 785 #ifdef SMP
 786         /* two separate prints in case of an unmapped page and trap */
 787         kprintf("mp_lock = %08x; ", mp_lock);
 788         kprintf("cpuid = %d\n", mycpu->gd_cpuid);
 789 #endif
 790
 791 #if defined(DDB)
 792         if (newpanic && trace_on_panic)
 793                 db_print_backtrace();
 794         if (debugger_on_panic)
 795                 Debugger("panic");
 796 #endif
 797         boot(bootopt);
 798 }
 799
 800 /*
 801  * Support for poweroff delay.
 802  */
 803 #ifndef POWEROFF_DELAY
 804 # define POWEROFF_DELAY 5000
 805 #endif
 806 static int poweroff_delay = POWEROFF_DELAY;
 807
 808 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW,
 809         &poweroff_delay, 0, "");
 810
 811 static void
 812 poweroff_wait(void *junk, int howto)
 813 {
 814         if(!(howto & RB_POWEROFF) || poweroff_delay <= 0)
 815                 return;
 816         DELAY(poweroff_delay * 1000);
 817 }
 818
 819 /*
 820  * Some system processes (e.g. syncer) need to be stopped at appropriate
 821  * points in their main loops prior to a system shutdown, so that they
 822  * won't interfere with the shutdown process (e.g. by holding a disk buf
 823  * to cause sync to fail).  For each of these system processes, register
 824  * shutdown_kproc() as a handler for one of shutdown events.
 825  */
 826 static int kproc_shutdown_wait = 60;
 827 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW,
 828     &kproc_shutdown_wait, 0, "");
 829
 830 void
 831 shutdown_kproc(void *arg, int howto)
 832 {
 833         struct thread *td;
 834         struct proc *p;
 835         int error;
 836
 837         if (panicstr)
 838                 return;
 839
 840         td = (struct thread *)arg;
 841         if ((p = td->td_proc) != NULL) {
 842             kprintf("Waiting (max %d seconds) for system process `%s' to stop...",
 843                 kproc_shutdown_wait, p->p_comm);
 844         } else {
 845             kprintf("Waiting (max %d seconds) for system thread %s to stop...",
 846                 kproc_shutdown_wait, td->td_comm);
 847         }
 848         error = suspend_kproc(td, kproc_shutdown_wait * hz);
 849
 850         if (error == EWOULDBLOCK)
 851                 kprintf("timed out\n");
 852         else
 853                 kprintf("stopped\n");
 854 }