| 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 1.62 2008/01/05 13:23:48 corecode 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/priv.h> |
| 57 | #include <sys/fcntl.h> /* FREAD */ |
| 58 | #include <sys/stat.h> /* S_IFCHR */ |
| 59 | #include <sys/vnode.h> |
| 60 | #include <sys/kernel.h> |
| 61 | #include <sys/kthread.h> |
| 62 | #include <sys/malloc.h> |
| 63 | #include <sys/mount.h> |
| 64 | #include <sys/queue.h> |
| 65 | #include <sys/sysctl.h> |
| 66 | #include <sys/vkernel.h> |
| 67 | #include <sys/conf.h> |
| 68 | #include <sys/sysproto.h> |
| 69 | #include <sys/device.h> |
| 70 | #include <sys/cons.h> |
| 71 | #include <sys/shm.h> |
| 72 | #include <sys/kern_syscall.h> |
| 73 | #include <vm/vm_map.h> |
| 74 | #include <vm/pmap.h> |
| 75 | |
| 76 | #include <sys/thread2.h> |
| 77 | #include <sys/buf2.h> |
| 78 | |
| 79 | #include <machine/pcb.h> |
| 80 | #include <machine/clock.h> |
| 81 | #include <machine/md_var.h> |
| 82 | #include <machine/smp.h> /* smp_active_mask, cpuid */ |
| 83 | #include <machine/vmparam.h> |
| 84 | |
| 85 | #include <sys/signalvar.h> |
| 86 | |
| 87 | #ifndef PANIC_REBOOT_WAIT_TIME |
| 88 | #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ |
| 89 | #endif |
| 90 | |
| 91 | /* |
| 92 | * Note that stdarg.h and the ANSI style va_start macro is used for both |
| 93 | * ANSI and traditional C compilers. We use the machine version to stay |
| 94 | * within the confines of the kernel header files. |
| 95 | */ |
| 96 | #include <machine/stdarg.h> |
| 97 | |
| 98 | #ifdef DDB |
| 99 | #include <ddb/ddb.h> |
| 100 | #ifdef DDB_UNATTENDED |
| 101 | int debugger_on_panic = 0; |
| 102 | #else |
| 103 | int debugger_on_panic = 1; |
| 104 | #endif |
| 105 | SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, |
| 106 | &debugger_on_panic, 0, "Run debugger on kernel panic"); |
| 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 | SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, |
| 148 | &bootverbose, 0, "Verbose kernel messages"); |
| 149 | |
| 150 | int cold = 1; /* note: assignment to force non-bss */ |
| 151 | int dumplo; /* OBSOLETE - savecore compat */ |
| 152 | u_int64_t dumplo64; |
| 153 | |
| 154 | static void boot (int) __dead2; |
| 155 | static void dumpsys (void); |
| 156 | static int setdumpdev (cdev_t dev); |
| 157 | static void poweroff_wait (void *, int); |
| 158 | static void print_uptime (void); |
| 159 | static void shutdown_halt (void *junk, int howto); |
| 160 | static void shutdown_panic (void *junk, int howto); |
| 161 | static void shutdown_reset (void *junk, int howto); |
| 162 | static int shutdown_busycount1(struct buf *bp, void *info); |
| 163 | static int shutdown_busycount2(struct buf *bp, void *info); |
| 164 | static void shutdown_cleanup_proc(struct proc *p); |
| 165 | |
| 166 | /* register various local shutdown events */ |
| 167 | static void |
| 168 | shutdown_conf(void *unused) |
| 169 | { |
| 170 | EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); |
| 171 | EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); |
| 172 | EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); |
| 173 | EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); |
| 174 | } |
| 175 | |
| 176 | SYSINIT(shutdown_conf, SI_BOOT2_MACHDEP, SI_ORDER_ANY, shutdown_conf, NULL) |
| 177 | |
| 178 | /* ARGSUSED */ |
| 179 | |
| 180 | /* |
| 181 | * The system call that results in a reboot |
| 182 | */ |
| 183 | int |
| 184 | sys_reboot(struct reboot_args *uap) |
| 185 | { |
| 186 | struct thread *td = curthread; |
| 187 | int error; |
| 188 | |
| 189 | if ((error = priv_check(td, PRIV_REBOOT))) |
| 190 | return (error); |
| 191 | |
| 192 | boot(uap->opt); |
| 193 | return (0); |
| 194 | } |
| 195 | |
| 196 | /* |
| 197 | * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC |
| 198 | */ |
| 199 | static int shutdown_howto = 0; |
| 200 | |
| 201 | void |
| 202 | shutdown_nice(int howto) |
| 203 | { |
| 204 | shutdown_howto = howto; |
| 205 | |
| 206 | /* Send a signal to init(8) and have it shutdown the world */ |
| 207 | if (initproc != NULL) { |
| 208 | ksignal(initproc, SIGINT); |
| 209 | } else { |
| 210 | /* No init(8) running, so simply reboot */ |
| 211 | boot(RB_NOSYNC); |
| 212 | } |
| 213 | return; |
| 214 | } |
| 215 | static int waittime = -1; |
| 216 | static struct thread *dumpthread; |
| 217 | static struct pcb dumppcb; |
| 218 | |
| 219 | static void |
| 220 | print_uptime(void) |
| 221 | { |
| 222 | int f; |
| 223 | struct timespec ts; |
| 224 | |
| 225 | getnanouptime(&ts); |
| 226 | kprintf("Uptime: "); |
| 227 | f = 0; |
| 228 | if (ts.tv_sec >= 86400) { |
| 229 | kprintf("%ldd", ts.tv_sec / 86400); |
| 230 | ts.tv_sec %= 86400; |
| 231 | f = 1; |
| 232 | } |
| 233 | if (f || ts.tv_sec >= 3600) { |
| 234 | kprintf("%ldh", ts.tv_sec / 3600); |
| 235 | ts.tv_sec %= 3600; |
| 236 | f = 1; |
| 237 | } |
| 238 | if (f || ts.tv_sec >= 60) { |
| 239 | kprintf("%ldm", ts.tv_sec / 60); |
| 240 | ts.tv_sec %= 60; |
| 241 | f = 1; |
| 242 | } |
| 243 | kprintf("%lds\n", ts.tv_sec); |
| 244 | } |
| 245 | |
| 246 | /* |
| 247 | * Go through the rigmarole of shutting down.. |
| 248 | * this used to be in machdep.c but I'll be dammned if I could see |
| 249 | * anything machine dependant in it. |
| 250 | */ |
| 251 | static void |
| 252 | boot(int howto) |
| 253 | { |
| 254 | /* |
| 255 | * Get rid of any user scheduler baggage and then give |
| 256 | * us a high priority. |
| 257 | */ |
| 258 | if (curthread->td_release) |
| 259 | curthread->td_release(curthread); |
| 260 | lwkt_setpri_self(TDPRI_MAX); |
| 261 | |
| 262 | /* collect extra flags that shutdown_nice might have set */ |
| 263 | howto |= shutdown_howto; |
| 264 | |
| 265 | #ifdef SMP |
| 266 | /* |
| 267 | * We really want to shutdown on the BSP. Subsystems such as ACPI |
| 268 | * can't power-down the box otherwise. |
| 269 | */ |
| 270 | if (smp_active_mask > 1) { |
| 271 | kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid); |
| 272 | } |
| 273 | if (panicstr == NULL && mycpu->gd_cpuid != 0) { |
| 274 | kprintf("Switching to cpu #0 for shutdown\n"); |
| 275 | lwkt_setcpu_self(globaldata_find(0)); |
| 276 | } |
| 277 | #endif |
| 278 | /* |
| 279 | * Do any callouts that should be done BEFORE syncing the filesystems. |
| 280 | */ |
| 281 | EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); |
| 282 | |
| 283 | /* |
| 284 | * Try to get rid of any remaining FS references. The calling |
| 285 | * process, proc0, and init may still hold references. The |
| 286 | * VFS cache subsystem may still hold a root reference to root. |
| 287 | * |
| 288 | * XXX this needs work. We really need to SIGSTOP all remaining |
| 289 | * processes in order to avoid blowups due to proc0's filesystem |
| 290 | * references going away. For now just make sure that the init |
| 291 | * process is stopped. |
| 292 | */ |
| 293 | if (panicstr == NULL) { |
| 294 | shutdown_cleanup_proc(curproc); |
| 295 | shutdown_cleanup_proc(&proc0); |
| 296 | if (initproc) { |
| 297 | if (initproc != curproc) { |
| 298 | ksignal(initproc, SIGSTOP); |
| 299 | tsleep(boot, 0, "shutdn", hz / 20); |
| 300 | } |
| 301 | shutdown_cleanup_proc(initproc); |
| 302 | } |
| 303 | vfs_cache_setroot(NULL, NULL); |
| 304 | } |
| 305 | |
| 306 | /* |
| 307 | * Now sync filesystems |
| 308 | */ |
| 309 | if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { |
| 310 | int iter, nbusy, pbusy; |
| 311 | |
| 312 | waittime = 0; |
| 313 | kprintf("\nsyncing disks... "); |
| 314 | |
| 315 | sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ |
| 316 | |
| 317 | /* |
| 318 | * With soft updates, some buffers that are |
| 319 | * written will be remarked as dirty until other |
| 320 | * buffers are written. |
| 321 | */ |
| 322 | for (iter = pbusy = 0; iter < 20; iter++) { |
| 323 | nbusy = scan_all_buffers(shutdown_busycount1, NULL); |
| 324 | if (nbusy == 0) |
| 325 | break; |
| 326 | kprintf("%d ", nbusy); |
| 327 | if (nbusy < pbusy) |
| 328 | iter = 0; |
| 329 | pbusy = nbusy; |
| 330 | /* |
| 331 | * XXX: |
| 332 | * Process soft update work queue if buffers don't sync |
| 333 | * after 6 iterations by permitting the syncer to run. |
| 334 | */ |
| 335 | if (iter > 5) |
| 336 | bio_ops_sync(NULL); |
| 337 | |
| 338 | sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ |
| 339 | tsleep(boot, 0, "shutdn", hz * iter / 20 + 1); |
| 340 | } |
| 341 | kprintf("\n"); |
| 342 | /* |
| 343 | * Count only busy local buffers to prevent forcing |
| 344 | * a fsck if we're just a client of a wedged NFS server |
| 345 | */ |
| 346 | nbusy = scan_all_buffers(shutdown_busycount2, NULL); |
| 347 | if (nbusy) { |
| 348 | /* |
| 349 | * Failed to sync all blocks. Indicate this and don't |
| 350 | * unmount filesystems (thus forcing an fsck on reboot). |
| 351 | */ |
| 352 | kprintf("giving up on %d buffers\n", nbusy); |
| 353 | #ifdef DDB |
| 354 | Debugger("busy buffer problem"); |
| 355 | #endif /* DDB */ |
| 356 | tsleep(boot, 0, "shutdn", hz * 5 + 1); |
| 357 | } else { |
| 358 | kprintf("done\n"); |
| 359 | /* |
| 360 | * Unmount filesystems |
| 361 | */ |
| 362 | if (panicstr == NULL) |
| 363 | vfs_unmountall(); |
| 364 | } |
| 365 | tsleep(boot, 0, "shutdn", hz / 10 + 1); |
| 366 | } |
| 367 | |
| 368 | print_uptime(); |
| 369 | |
| 370 | /* |
| 371 | * Dump before doing post_sync shutdown ops |
| 372 | */ |
| 373 | crit_enter(); |
| 374 | if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) |
| 375 | dumpsys(); |
| 376 | |
| 377 | /* |
| 378 | * Ok, now do things that assume all filesystem activity has |
| 379 | * been completed. This will also call the device shutdown |
| 380 | * methods. |
| 381 | */ |
| 382 | EVENTHANDLER_INVOKE(shutdown_post_sync, howto); |
| 383 | |
| 384 | /* Now that we're going to really halt the system... */ |
| 385 | EVENTHANDLER_INVOKE(shutdown_final, howto); |
| 386 | |
| 387 | for(;;) ; /* safety against shutdown_reset not working */ |
| 388 | /* NOTREACHED */ |
| 389 | } |
| 390 | |
| 391 | static int |
| 392 | shutdown_busycount1(struct buf *bp, void *info) |
| 393 | { |
| 394 | if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0) |
| 395 | return(1); |
| 396 | if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) |
| 397 | return (1); |
| 398 | return (0); |
| 399 | } |
| 400 | |
| 401 | static int |
| 402 | shutdown_busycount2(struct buf *bp, void *info) |
| 403 | { |
| 404 | if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) || |
| 405 | ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { |
| 406 | /* |
| 407 | * Only count buffers undergoing write I/O |
| 408 | * on the related vnode. |
| 409 | */ |
| 410 | if (bp->b_vp == NULL || |
| 411 | bp->b_vp->v_track_write.bk_active == 0) { |
| 412 | return (0); |
| 413 | } |
| 414 | #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) |
| 415 | kprintf( |
| 416 | "%p dev:?, flags:%08x, loffset:%lld, doffset:%lld\n", |
| 417 | bp, |
| 418 | bp->b_flags, bp->b_loffset, |
| 419 | bp->b_bio2.bio_offset); |
| 420 | #endif |
| 421 | return(1); |
| 422 | } |
| 423 | return(0); |
| 424 | } |
| 425 | |
| 426 | /* |
| 427 | * If the shutdown was a clean halt, behave accordingly. |
| 428 | */ |
| 429 | static void |
| 430 | shutdown_halt(void *junk, int howto) |
| 431 | { |
| 432 | if (howto & RB_HALT) { |
| 433 | kprintf("\n"); |
| 434 | kprintf("The operating system has halted.\n"); |
| 435 | #ifdef _KERNEL_VIRTUAL |
| 436 | cpu_halt(); |
| 437 | #else |
| 438 | kprintf("Please press any key to reboot.\n\n"); |
| 439 | switch (cngetc()) { |
| 440 | case -1: /* No console, just die */ |
| 441 | cpu_halt(); |
| 442 | /* NOTREACHED */ |
| 443 | default: |
| 444 | howto &= ~RB_HALT; |
| 445 | break; |
| 446 | } |
| 447 | #endif |
| 448 | } |
| 449 | } |
| 450 | |
| 451 | /* |
| 452 | * Check to see if the system paniced, pause and then reboot |
| 453 | * according to the specified delay. |
| 454 | */ |
| 455 | static void |
| 456 | shutdown_panic(void *junk, int howto) |
| 457 | { |
| 458 | int loop; |
| 459 | |
| 460 | if (howto & RB_DUMP) { |
| 461 | if (PANIC_REBOOT_WAIT_TIME != 0) { |
| 462 | if (PANIC_REBOOT_WAIT_TIME != -1) { |
| 463 | kprintf("Automatic reboot in %d seconds - " |
| 464 | "press a key on the console to abort\n", |
| 465 | PANIC_REBOOT_WAIT_TIME); |
| 466 | for (loop = PANIC_REBOOT_WAIT_TIME * 10; |
| 467 | loop > 0; --loop) { |
| 468 | DELAY(1000 * 100); /* 1/10th second */ |
| 469 | /* Did user type a key? */ |
| 470 | if (cncheckc() != -1) |
| 471 | break; |
| 472 | } |
| 473 | if (!loop) |
| 474 | return; |
| 475 | } |
| 476 | } else { /* zero time specified - reboot NOW */ |
| 477 | return; |
| 478 | } |
| 479 | kprintf("--> Press a key on the console to reboot,\n"); |
| 480 | kprintf("--> or switch off the system now.\n"); |
| 481 | cngetc(); |
| 482 | } |
| 483 | } |
| 484 | |
| 485 | /* |
| 486 | * Everything done, now reset |
| 487 | */ |
| 488 | static void |
| 489 | shutdown_reset(void *junk, int howto) |
| 490 | { |
| 491 | kprintf("Rebooting...\n"); |
| 492 | DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */ |
| 493 | /* cpu_boot(howto); */ /* doesn't do anything at the moment */ |
| 494 | cpu_reset(); |
| 495 | /* NOTREACHED */ /* assuming reset worked */ |
| 496 | } |
| 497 | |
| 498 | /* |
| 499 | * Try to remove FS references in the specified process. This function |
| 500 | * is used during shutdown |
| 501 | */ |
| 502 | static |
| 503 | void |
| 504 | shutdown_cleanup_proc(struct proc *p) |
| 505 | { |
| 506 | struct filedesc *fdp; |
| 507 | struct vmspace *vm; |
| 508 | |
| 509 | if (p == NULL) |
| 510 | return; |
| 511 | if ((fdp = p->p_fd) != NULL) { |
| 512 | kern_closefrom(0); |
| 513 | if (fdp->fd_cdir) { |
| 514 | cache_drop(&fdp->fd_ncdir); |
| 515 | vrele(fdp->fd_cdir); |
| 516 | fdp->fd_cdir = NULL; |
| 517 | } |
| 518 | if (fdp->fd_rdir) { |
| 519 | cache_drop(&fdp->fd_nrdir); |
| 520 | vrele(fdp->fd_rdir); |
| 521 | fdp->fd_rdir = NULL; |
| 522 | } |
| 523 | if (fdp->fd_jdir) { |
| 524 | cache_drop(&fdp->fd_njdir); |
| 525 | vrele(fdp->fd_jdir); |
| 526 | fdp->fd_jdir = NULL; |
| 527 | } |
| 528 | } |
| 529 | if (p->p_vkernel) |
| 530 | vkernel_exit(p); |
| 531 | if (p->p_textvp) { |
| 532 | vrele(p->p_textvp); |
| 533 | p->p_textvp = NULL; |
| 534 | } |
| 535 | vm = p->p_vmspace; |
| 536 | if (vm != NULL) { |
| 537 | pmap_remove_pages(vmspace_pmap(vm), |
| 538 | VM_MIN_USER_ADDRESS, |
| 539 | VM_MAX_USER_ADDRESS); |
| 540 | vm_map_remove(&vm->vm_map, |
| 541 | VM_MIN_USER_ADDRESS, |
| 542 | VM_MAX_USER_ADDRESS); |
| 543 | } |
| 544 | } |
| 545 | |
| 546 | /* |
| 547 | * Magic number for savecore |
| 548 | * |
| 549 | * exported (symorder) and used at least by savecore(8) |
| 550 | * |
| 551 | * Mark it as used so that gcc doesn't optimize it away. |
| 552 | */ |
| 553 | __attribute__((__used__)) |
| 554 | static u_long const dumpmag = 0x8fca0101UL; |
| 555 | |
| 556 | static int dumpsize = 0; /* also for savecore */ |
| 557 | |
| 558 | static int dodump = 1; |
| 559 | |
| 560 | SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, |
| 561 | "Try to perform coredump on kernel panic"); |
| 562 | |
| 563 | static int |
| 564 | setdumpdev(cdev_t dev) |
| 565 | { |
| 566 | struct partinfo pinfo; |
| 567 | u_int64_t newdumplo; |
| 568 | int error; |
| 569 | int doopen; |
| 570 | |
| 571 | if (dev == NULL) { |
| 572 | dumpdev = dev; |
| 573 | return (0); |
| 574 | } |
| 575 | bzero(&pinfo, sizeof(pinfo)); |
| 576 | |
| 577 | /* |
| 578 | * We have to open the device before we can perform ioctls on it, |
| 579 | * or the slice/label data may not be present. Device opens are |
| 580 | * usually tracked by specfs, but the dump device can be set in |
| 581 | * early boot and may not be open so this is somewhat of a hack. |
| 582 | */ |
| 583 | doopen = (dev->si_sysref.refcnt == 1); |
| 584 | if (doopen) { |
| 585 | error = dev_dopen(dev, FREAD, S_IFCHR, proc0.p_ucred); |
| 586 | if (error) |
| 587 | return (error); |
| 588 | } |
| 589 | error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred); |
| 590 | if (doopen) |
| 591 | dev_dclose(dev, FREAD, S_IFCHR); |
| 592 | if (error || pinfo.media_blocks == 0 || pinfo.media_blksize == 0) |
| 593 | return (ENXIO); |
| 594 | |
| 595 | newdumplo = pinfo.media_blocks - |
| 596 | ((u_int64_t)Maxmem * PAGE_SIZE / DEV_BSIZE); |
| 597 | if ((int64_t)newdumplo < (int64_t)pinfo.reserved_blocks) |
| 598 | return (ENOSPC); |
| 599 | dumpdev = dev; |
| 600 | dumplo64 = newdumplo; |
| 601 | return (0); |
| 602 | } |
| 603 | |
| 604 | |
| 605 | /* ARGSUSED */ |
| 606 | static void dump_conf (void *dummy); |
| 607 | static void |
| 608 | dump_conf(void *dummy) |
| 609 | { |
| 610 | char *path; |
| 611 | cdev_t dev; |
| 612 | |
| 613 | path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK); |
| 614 | if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { |
| 615 | dev = kgetdiskbyname(path); |
| 616 | if (dev != NULL) |
| 617 | dumpdev = dev; |
| 618 | } |
| 619 | kfree(path, M_TEMP); |
| 620 | if (setdumpdev(dumpdev) != 0) |
| 621 | dumpdev = NULL; |
| 622 | } |
| 623 | |
| 624 | SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) |
| 625 | |
| 626 | static int |
| 627 | sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) |
| 628 | { |
| 629 | int error; |
| 630 | udev_t ndumpdev; |
| 631 | |
| 632 | ndumpdev = dev2udev(dumpdev); |
| 633 | error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); |
| 634 | if (error == 0 && req->newptr != NULL) |
| 635 | error = setdumpdev(udev2dev(ndumpdev, 0)); |
| 636 | return (error); |
| 637 | } |
| 638 | |
| 639 | SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, |
| 640 | 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", ""); |
| 641 | |
| 642 | /* |
| 643 | * Doadump comes here after turning off memory management and |
| 644 | * getting on the dump stack, either when called above, or by |
| 645 | * the auto-restart code. |
| 646 | */ |
| 647 | static void |
| 648 | dumpsys(void) |
| 649 | { |
| 650 | int error; |
| 651 | |
| 652 | savectx(&dumppcb); |
| 653 | dumpthread = curthread; |
| 654 | if (dumping++) { |
| 655 | kprintf("Dump already in progress, bailing...\n"); |
| 656 | return; |
| 657 | } |
| 658 | if (!dodump) |
| 659 | return; |
| 660 | if (dumpdev == NULL) |
| 661 | return; |
| 662 | dumpsize = Maxmem; |
| 663 | kprintf("\ndumping to dev %s, blockno %lld\n", |
| 664 | devtoname(dumpdev), |
| 665 | (long long)dumplo64); |
| 666 | kprintf("dump "); |
| 667 | error = dev_ddump(dumpdev); |
| 668 | if (error == 0) { |
| 669 | kprintf("succeeded\n"); |
| 670 | return; |
| 671 | } |
| 672 | kprintf("failed, reason: "); |
| 673 | switch (error) { |
| 674 | case ENOSYS: |
| 675 | case ENODEV: |
| 676 | kprintf("device doesn't support a dump routine\n"); |
| 677 | break; |
| 678 | |
| 679 | case ENXIO: |
| 680 | kprintf("device bad\n"); |
| 681 | break; |
| 682 | |
| 683 | case EFAULT: |
| 684 | kprintf("device not ready\n"); |
| 685 | break; |
| 686 | |
| 687 | case EINVAL: |
| 688 | kprintf("area improper\n"); |
| 689 | break; |
| 690 | |
| 691 | case EIO: |
| 692 | kprintf("i/o error\n"); |
| 693 | break; |
| 694 | |
| 695 | case EINTR: |
| 696 | kprintf("aborted from console\n"); |
| 697 | break; |
| 698 | |
| 699 | default: |
| 700 | kprintf("unknown, error = %d\n", error); |
| 701 | break; |
| 702 | } |
| 703 | } |
| 704 | |
| 705 | int |
| 706 | dumpstatus(vm_offset_t addr, off_t count) |
| 707 | { |
| 708 | int c; |
| 709 | |
| 710 | if (addr % (1024 * 1024) == 0) { |
| 711 | #ifdef HW_WDOG |
| 712 | if (wdog_tickler) |
| 713 | (*wdog_tickler)(); |
| 714 | #endif |
| 715 | kprintf("%ld ", (long)(count / (1024 * 1024))); |
| 716 | } |
| 717 | |
| 718 | if ((c = cncheckc()) == 0x03) |
| 719 | return -1; |
| 720 | else if (c != -1) |
| 721 | kprintf("[CTRL-C to abort] "); |
| 722 | |
| 723 | return 0; |
| 724 | } |
| 725 | |
| 726 | /* |
| 727 | * Panic is called on unresolvable fatal errors. It prints "panic: mesg", |
| 728 | * and then reboots. If we are called twice, then we avoid trying to sync |
| 729 | * the disks as this often leads to recursive panics. |
| 730 | */ |
| 731 | void |
| 732 | panic(const char *fmt, ...) |
| 733 | { |
| 734 | int bootopt, newpanic; |
| 735 | __va_list ap; |
| 736 | static char buf[256]; |
| 737 | |
| 738 | #ifdef SMP |
| 739 | /* |
| 740 | * If a panic occurs on multiple cpus before the first is able to |
| 741 | * halt the other cpus, only one cpu is allowed to take the panic. |
| 742 | * Attempt to be verbose about this situation but if the kprintf() |
| 743 | * itself panics don't let us overrun the kernel stack. |
| 744 | * |
| 745 | * Be very nasty about descheduling our thread at the lowest |
| 746 | * level possible in an attempt to freeze the thread without |
| 747 | * inducing further panics. |
| 748 | * |
| 749 | * Bumping gd_trap_nesting_level will also bypass assertions in |
| 750 | * lwkt_switch() and allow us to switch away even if we are a |
| 751 | * FAST interrupt or IPI. |
| 752 | */ |
| 753 | if (atomic_poll_acquire_int(&panic_cpu_interlock)) { |
| 754 | panic_cpu_gd = mycpu; |
| 755 | } else if (panic_cpu_gd != mycpu) { |
| 756 | crit_enter(); |
| 757 | ++mycpu->gd_trap_nesting_level; |
| 758 | if (mycpu->gd_trap_nesting_level < 25) { |
| 759 | kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n", |
| 760 | mycpu->gd_cpuid, curthread); |
| 761 | } |
| 762 | curthread->td_release = NULL; /* be a grinch */ |
| 763 | for (;;) { |
| 764 | lwkt_deschedule_self(curthread); |
| 765 | lwkt_switch(); |
| 766 | } |
| 767 | /* NOT REACHED */ |
| 768 | /* --mycpu->gd_trap_nesting_level */ |
| 769 | /* crit_exit() */ |
| 770 | } |
| 771 | #endif |
| 772 | bootopt = RB_AUTOBOOT | RB_DUMP; |
| 773 | if (sync_on_panic == 0) |
| 774 | bootopt |= RB_NOSYNC; |
| 775 | newpanic = 0; |
| 776 | if (panicstr) |
| 777 | bootopt |= RB_NOSYNC; |
| 778 | else { |
| 779 | panicstr = fmt; |
| 780 | newpanic = 1; |
| 781 | } |
| 782 | |
| 783 | __va_start(ap, fmt); |
| 784 | kvsnprintf(buf, sizeof(buf), fmt, ap); |
| 785 | if (panicstr == fmt) |
| 786 | panicstr = buf; |
| 787 | __va_end(ap); |
| 788 | kprintf("panic: %s\n", buf); |
| 789 | #ifdef SMP |
| 790 | /* two separate prints in case of an unmapped page and trap */ |
| 791 | kprintf("mp_lock = %08x; ", mp_lock); |
| 792 | kprintf("cpuid = %d\n", mycpu->gd_cpuid); |
| 793 | #endif |
| 794 | |
| 795 | #if defined(DDB) |
| 796 | if (newpanic && trace_on_panic) |
| 797 | print_backtrace(); |
| 798 | if (debugger_on_panic) |
| 799 | Debugger("panic"); |
| 800 | #endif |
| 801 | boot(bootopt); |
| 802 | } |
| 803 | |
| 804 | /* |
| 805 | * Support for poweroff delay. |
| 806 | */ |
| 807 | #ifndef POWEROFF_DELAY |
| 808 | # define POWEROFF_DELAY 5000 |
| 809 | #endif |
| 810 | static int poweroff_delay = POWEROFF_DELAY; |
| 811 | |
| 812 | SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, |
| 813 | &poweroff_delay, 0, ""); |
| 814 | |
| 815 | static void |
| 816 | poweroff_wait(void *junk, int howto) |
| 817 | { |
| 818 | if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) |
| 819 | return; |
| 820 | DELAY(poweroff_delay * 1000); |
| 821 | } |
| 822 | |
| 823 | /* |
| 824 | * Some system processes (e.g. syncer) need to be stopped at appropriate |
| 825 | * points in their main loops prior to a system shutdown, so that they |
| 826 | * won't interfere with the shutdown process (e.g. by holding a disk buf |
| 827 | * to cause sync to fail). For each of these system processes, register |
| 828 | * shutdown_kproc() as a handler for one of shutdown events. |
| 829 | */ |
| 830 | static int kproc_shutdown_wait = 60; |
| 831 | SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, |
| 832 | &kproc_shutdown_wait, 0, ""); |
| 833 | |
| 834 | void |
| 835 | shutdown_kproc(void *arg, int howto) |
| 836 | { |
| 837 | struct thread *td; |
| 838 | struct proc *p; |
| 839 | int error; |
| 840 | |
| 841 | if (panicstr) |
| 842 | return; |
| 843 | |
| 844 | td = (struct thread *)arg; |
| 845 | if ((p = td->td_proc) != NULL) { |
| 846 | kprintf("Waiting (max %d seconds) for system process `%s' to stop...", |
| 847 | kproc_shutdown_wait, p->p_comm); |
| 848 | } else { |
| 849 | kprintf("Waiting (max %d seconds) for system thread %s to stop...", |
| 850 | kproc_shutdown_wait, td->td_comm); |
| 851 | } |
| 852 | error = suspend_kproc(td, kproc_shutdown_wait * hz); |
| 853 | |
| 854 | if (error == EWOULDBLOCK) |
| 855 | kprintf("timed out\n"); |
| 856 | else |
| 857 | kprintf("stopped\n"); |
| 858 | } |