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 #include "use_gpio.h" 49 50 #include <sys/param.h> 51 #include <sys/systm.h> 52 #include <sys/eventhandler.h> 53 #include <sys/buf.h> 54 #include <sys/disk.h> 55 #include <sys/diskslice.h> 56 #include <sys/reboot.h> 57 #include <sys/proc.h> 58 #include <sys/priv.h> 59 #include <sys/fcntl.h> /* FREAD */ 60 #include <sys/stat.h> /* S_IFCHR */ 61 #include <sys/vnode.h> 62 #include <sys/kernel.h> 63 #include <sys/kerneldump.h> 64 #include <sys/kthread.h> 65 #include <sys/malloc.h> 66 #include <sys/mount.h> 67 #include <sys/queue.h> 68 #include <sys/sysctl.h> 69 #include <sys/vkernel.h> 70 #include <sys/conf.h> 71 #include <sys/sysproto.h> 72 #include <sys/device.h> 73 #include <sys/cons.h> 74 #include <sys/shm.h> 75 #include <sys/kerneldump.h> 76 #include <sys/kern_syscall.h> 77 #include <vm/vm_map.h> 78 #include <vm/pmap.h> 79 80 #include <sys/thread2.h> 81 #include <sys/buf2.h> 82 #include <sys/mplock2.h> 83 84 #include <machine/cpu.h> 85 #include <machine/clock.h> 86 #include <machine/md_var.h> 87 #include <machine/smp.h> /* smp_active_mask, cpuid */ 88 #include <machine/vmparam.h> 89 #include <machine/thread.h> 90 91 #include <sys/signalvar.h> 92 93 #include <sys/wdog.h> 94 #include <dev/misc/gpio/gpio.h> 95 96 #ifndef PANIC_REBOOT_WAIT_TIME 97 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 98 #endif 99 100 /* 101 * Note that stdarg.h and the ANSI style va_start macro is used for both 102 * ANSI and traditional C compilers. We use the machine version to stay 103 * within the confines of the kernel header files. 104 */ 105 #include <machine/stdarg.h> 106 107 #ifdef DDB 108 #include <ddb/ddb.h> 109 #ifdef DDB_UNATTENDED 110 int debugger_on_panic = 0; 111 #else 112 int debugger_on_panic = 1; 113 #endif 114 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 115 &debugger_on_panic, 0, "Run debugger on kernel panic"); 116 117 #ifdef DDB_TRACE 118 int trace_on_panic = 1; 119 #else 120 int trace_on_panic = 0; 121 #endif 122 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 123 &trace_on_panic, 0, "Print stack trace on kernel panic"); 124 #endif 125 126 static int sync_on_panic = 0; 127 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 128 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 129 130 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 131 132 #ifdef HW_WDOG 133 /* 134 * If there is a hardware watchdog, point this at the function needed to 135 * hold it off. 136 * It's needed when the kernel needs to do some lengthy operations. 137 * e.g. in wd.c when dumping core.. It's most annoying to have 138 * your precious core-dump only half written because the wdog kicked in. 139 */ 140 watchdog_tickle_fn wdog_tickler = NULL; 141 #endif /* HW_WDOG */ 142 143 /* 144 * Variable panicstr contains argument to first call to panic; used as flag 145 * to indicate that the kernel has already called panic. 146 */ 147 const char *panicstr; 148 149 int dumping; /* system is dumping */ 150 static struct dumperinfo dumper; /* selected dumper */ 151 152 globaldata_t panic_cpu_gd; /* which cpu took the panic */ 153 struct lwkt_tokref panic_tokens[LWKT_MAXTOKENS]; 154 int panic_tokens_count; 155 156 int bootverbose = 0; /* note: assignment to force non-bss */ 157 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, 158 &bootverbose, 0, "Verbose kernel messages"); 159 160 int cold = 1; /* note: assignment to force non-bss */ 161 int dumplo; /* OBSOLETE - savecore compat */ 162 u_int64_t dumplo64; 163 164 static void boot (int) __dead2; 165 static int setdumpdev (cdev_t dev); 166 static void poweroff_wait (void *, int); 167 static void print_uptime (void); 168 static void shutdown_halt (void *junk, int howto); 169 static void shutdown_panic (void *junk, int howto); 170 static void shutdown_reset (void *junk, int howto); 171 static int shutdown_busycount1(struct buf *bp, void *info); 172 static int shutdown_busycount2(struct buf *bp, void *info); 173 static void shutdown_cleanup_proc(struct proc *p); 174 175 /* register various local shutdown events */ 176 static void 177 shutdown_conf(void *unused) 178 { 179 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 180 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 181 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 182 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 183 } 184 185 SYSINIT(shutdown_conf, SI_BOOT2_MACHDEP, SI_ORDER_ANY, shutdown_conf, NULL) 186 187 /* ARGSUSED */ 188 189 /* 190 * The system call that results in a reboot 191 * 192 * MPALMOSTSAFE 193 */ 194 int 195 sys_reboot(struct reboot_args *uap) 196 { 197 struct thread *td = curthread; 198 int error; 199 200 if ((error = priv_check(td, PRIV_REBOOT))) 201 return (error); 202 203 get_mplock(); 204 boot(uap->opt); 205 rel_mplock(); 206 return (0); 207 } 208 209 /* 210 * Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC 211 */ 212 static int shutdown_howto = 0; 213 214 void 215 shutdown_nice(int howto) 216 { 217 shutdown_howto = howto; 218 219 /* Send a signal to init(8) and have it shutdown the world */ 220 if (initproc != NULL) { 221 ksignal(initproc, SIGINT); 222 } else { 223 /* No init(8) running, so simply reboot */ 224 boot(RB_NOSYNC); 225 } 226 return; 227 } 228 static int waittime = -1; 229 struct pcb dumppcb; 230 struct thread *dumpthread; 231 232 static void 233 print_uptime(void) 234 { 235 int f; 236 struct timespec ts; 237 238 getnanouptime(&ts); 239 kprintf("Uptime: "); 240 f = 0; 241 if (ts.tv_sec >= 86400) { 242 kprintf("%ldd", ts.tv_sec / 86400); 243 ts.tv_sec %= 86400; 244 f = 1; 245 } 246 if (f || ts.tv_sec >= 3600) { 247 kprintf("%ldh", ts.tv_sec / 3600); 248 ts.tv_sec %= 3600; 249 f = 1; 250 } 251 if (f || ts.tv_sec >= 60) { 252 kprintf("%ldm", ts.tv_sec / 60); 253 ts.tv_sec %= 60; 254 f = 1; 255 } 256 kprintf("%lds\n", ts.tv_sec); 257 } 258 259 /* 260 * Go through the rigmarole of shutting down.. 261 * this used to be in machdep.c but I'll be dammned if I could see 262 * anything machine dependant in it. 263 */ 264 static void 265 boot(int howto) 266 { 267 /* 268 * Get rid of any user scheduler baggage and then give 269 * us a high priority. 270 */ 271 if (curthread->td_release) 272 curthread->td_release(curthread); 273 lwkt_setpri_self(TDPRI_MAX); 274 275 /* collect extra flags that shutdown_nice might have set */ 276 howto |= shutdown_howto; 277 278 #ifdef SMP 279 /* 280 * We really want to shutdown on the BSP. Subsystems such as ACPI 281 * can't power-down the box otherwise. 282 */ 283 if (smp_active_mask > 1) { 284 kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid); 285 } 286 if (panicstr == NULL && mycpu->gd_cpuid != 0) { 287 kprintf("Switching to cpu #0 for shutdown\n"); 288 lwkt_setcpu_self(globaldata_find(0)); 289 } 290 #endif 291 /* 292 * Do any callouts that should be done BEFORE syncing the filesystems. 293 */ 294 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 295 296 /* 297 * Try to get rid of any remaining FS references. The calling 298 * process, proc0, and init may still hold references. The 299 * VFS cache subsystem may still hold a root reference to root. 300 * 301 * XXX this needs work. We really need to SIGSTOP all remaining 302 * processes in order to avoid blowups due to proc0's filesystem 303 * references going away. For now just make sure that the init 304 * process is stopped. 305 */ 306 if (panicstr == NULL) { 307 shutdown_cleanup_proc(curproc); 308 shutdown_cleanup_proc(&proc0); 309 if (initproc) { 310 if (initproc != curproc) { 311 ksignal(initproc, SIGSTOP); 312 tsleep(boot, 0, "shutdn", hz / 20); 313 } 314 shutdown_cleanup_proc(initproc); 315 } 316 vfs_cache_setroot(NULL, NULL); 317 } 318 319 /* 320 * Now sync filesystems 321 */ 322 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 323 int iter, nbusy, pbusy; 324 325 waittime = 0; 326 kprintf("\nsyncing disks... "); 327 328 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 329 330 /* 331 * With soft updates, some buffers that are 332 * written will be remarked as dirty until other 333 * buffers are written. 334 */ 335 for (iter = pbusy = 0; iter < 20; iter++) { 336 nbusy = scan_all_buffers(shutdown_busycount1, NULL); 337 if (nbusy == 0) 338 break; 339 kprintf("%d ", nbusy); 340 if (nbusy < pbusy) 341 iter = 0; 342 pbusy = nbusy; 343 /* 344 * XXX: 345 * Process soft update work queue if buffers don't sync 346 * after 6 iterations by permitting the syncer to run. 347 */ 348 if (iter > 5) 349 bio_ops_sync(NULL); 350 351 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 352 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1); 353 } 354 kprintf("\n"); 355 /* 356 * Count only busy local buffers to prevent forcing 357 * a fsck if we're just a client of a wedged NFS server 358 */ 359 nbusy = scan_all_buffers(shutdown_busycount2, NULL); 360 if (nbusy) { 361 /* 362 * Failed to sync all blocks. Indicate this and don't 363 * unmount filesystems (thus forcing an fsck on reboot). 364 */ 365 kprintf("giving up on %d buffers\n", nbusy); 366 #ifdef DDB 367 if (debugger_on_panic) 368 Debugger("busy buffer problem"); 369 #endif /* DDB */ 370 tsleep(boot, 0, "shutdn", hz * 5 + 1); 371 } else { 372 kprintf("done\n"); 373 /* 374 * Unmount filesystems 375 */ 376 if (panicstr == NULL) 377 vfs_unmountall(); 378 } 379 tsleep(boot, 0, "shutdn", hz / 10 + 1); 380 } 381 382 print_uptime(); 383 384 /* 385 * Dump before doing post_sync shutdown ops 386 */ 387 crit_enter(); 388 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) { 389 dumpsys(); 390 } 391 392 /* 393 * Ok, now do things that assume all filesystem activity has 394 * been completed. This will also call the device shutdown 395 * methods. 396 */ 397 EVENTHANDLER_INVOKE(shutdown_post_sync, howto); 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 * Pass 1 - Figure out if there are any busy or dirty buffers still present. 408 * 409 * We ignore TMPFS mounts in this pass. 410 */ 411 static int 412 shutdown_busycount1(struct buf *bp, void *info) 413 { 414 struct vnode *vp; 415 416 if ((vp = bp->b_vp) != NULL && vp->v_tag == VT_TMPFS) 417 return (0); 418 if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0) 419 return(1); 420 if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) 421 return (1); 422 return (0); 423 } 424 425 /* 426 * Pass 2 - only run after pass 1 has completed or has given up 427 * 428 * We ignore TMPFS, NFS, MFS, and SMBFS mounts in this pass. 429 */ 430 static int 431 shutdown_busycount2(struct buf *bp, void *info) 432 { 433 struct vnode *vp; 434 435 /* 436 * Ignore tmpfs and nfs mounts 437 */ 438 if ((vp = bp->b_vp) != NULL) { 439 if (vp->v_tag == VT_TMPFS) 440 return (0); 441 if (vp->v_tag == VT_NFS) 442 return (0); 443 if (vp->v_tag == VT_MFS) 444 return (0); 445 if (vp->v_tag == VT_SMBFS) 446 return (0); 447 } 448 449 /* 450 * Only count buffers stuck on I/O, ignore everything else 451 */ 452 if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) || 453 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 454 /* 455 * Only count buffers undergoing write I/O 456 * on the related vnode. 457 */ 458 if (bp->b_vp == NULL || 459 bio_track_active(&bp->b_vp->v_track_write) == 0) { 460 return (0); 461 } 462 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 463 kprintf( 464 "%p dev:?, flags:%08x, loffset:%jd, doffset:%jd\n", 465 bp, 466 bp->b_flags, (intmax_t)bp->b_loffset, 467 (intmax_t)bp->b_bio2.bio_offset); 468 #endif 469 return(1); 470 } 471 return(0); 472 } 473 474 /* 475 * If the shutdown was a clean halt, behave accordingly. 476 */ 477 static void 478 shutdown_halt(void *junk, int howto) 479 { 480 if (howto & RB_HALT) { 481 kprintf("\n"); 482 kprintf("The operating system has halted.\n"); 483 #ifdef _KERNEL_VIRTUAL 484 cpu_halt(); 485 #else 486 kprintf("Please press any key to reboot.\n\n"); 487 switch (cngetc()) { 488 case -1: /* No console, just die */ 489 cpu_halt(); 490 /* NOTREACHED */ 491 default: 492 howto &= ~RB_HALT; 493 break; 494 } 495 #endif 496 } 497 } 498 499 /* 500 * Check to see if the system paniced, pause and then reboot 501 * according to the specified delay. 502 */ 503 static void 504 shutdown_panic(void *junk, int howto) 505 { 506 int loop; 507 508 if (howto & RB_DUMP) { 509 if (PANIC_REBOOT_WAIT_TIME != 0) { 510 if (PANIC_REBOOT_WAIT_TIME != -1) { 511 kprintf("Automatic reboot in %d seconds - " 512 "press a key on the console to abort\n", 513 PANIC_REBOOT_WAIT_TIME); 514 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 515 loop > 0; --loop) { 516 DELAY(1000 * 100); /* 1/10th second */ 517 /* Did user type a key? */ 518 if (cncheckc() != -1) 519 break; 520 } 521 if (!loop) 522 return; 523 } 524 } else { /* zero time specified - reboot NOW */ 525 return; 526 } 527 kprintf("--> Press a key on the console to reboot,\n"); 528 kprintf("--> or switch off the system now.\n"); 529 cngetc(); 530 } 531 } 532 533 /* 534 * Everything done, now reset 535 */ 536 static void 537 shutdown_reset(void *junk, int howto) 538 { 539 kprintf("Rebooting...\n"); 540 DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */ 541 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 542 cpu_reset(); 543 /* NOTREACHED */ /* assuming reset worked */ 544 } 545 546 /* 547 * Try to remove FS references in the specified process. This function 548 * is used during shutdown 549 */ 550 static 551 void 552 shutdown_cleanup_proc(struct proc *p) 553 { 554 struct filedesc *fdp; 555 struct vmspace *vm; 556 557 if (p == NULL) 558 return; 559 if ((fdp = p->p_fd) != NULL) { 560 kern_closefrom(0); 561 if (fdp->fd_cdir) { 562 cache_drop(&fdp->fd_ncdir); 563 vrele(fdp->fd_cdir); 564 fdp->fd_cdir = NULL; 565 } 566 if (fdp->fd_rdir) { 567 cache_drop(&fdp->fd_nrdir); 568 vrele(fdp->fd_rdir); 569 fdp->fd_rdir = NULL; 570 } 571 if (fdp->fd_jdir) { 572 cache_drop(&fdp->fd_njdir); 573 vrele(fdp->fd_jdir); 574 fdp->fd_jdir = NULL; 575 } 576 } 577 if (p->p_vkernel) 578 vkernel_exit(p); 579 if (p->p_textvp) { 580 vrele(p->p_textvp); 581 p->p_textvp = NULL; 582 } 583 vm = p->p_vmspace; 584 if (vm != NULL) { 585 pmap_remove_pages(vmspace_pmap(vm), 586 VM_MIN_USER_ADDRESS, 587 VM_MAX_USER_ADDRESS); 588 vm_map_remove(&vm->vm_map, 589 VM_MIN_USER_ADDRESS, 590 VM_MAX_USER_ADDRESS); 591 } 592 } 593 594 /* 595 * Magic number for savecore 596 * 597 * exported (symorder) and used at least by savecore(8) 598 * 599 * Mark it as used so that gcc doesn't optimize it away. 600 */ 601 __attribute__((__used__)) 602 static u_long const dumpmag = 0x8fca0101UL; 603 604 __attribute__((__used__)) 605 static int dumpsize = 0; /* also for savecore */ 606 607 static int dodump = 1; 608 609 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 610 "Try to perform coredump on kernel panic"); 611 612 void 613 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver, 614 uint64_t dumplen, uint32_t blksz) 615 { 616 bzero(kdh, sizeof(*kdh)); 617 strncpy(kdh->magic, magic, sizeof(kdh->magic)); 618 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture)); 619 kdh->version = htod32(KERNELDUMPVERSION); 620 kdh->architectureversion = htod32(archver); 621 kdh->dumplength = htod64(dumplen); 622 kdh->dumptime = htod64(time_second); 623 kdh->blocksize = htod32(blksz); 624 strncpy(kdh->hostname, hostname, sizeof(kdh->hostname)); 625 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring)); 626 if (panicstr != NULL) 627 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring)); 628 kdh->parity = kerneldump_parity(kdh); 629 } 630 631 static int 632 setdumpdev(cdev_t dev) 633 { 634 int error; 635 int doopen; 636 637 if (dev == NULL) { 638 disk_dumpconf(NULL, 0/*off*/); 639 return (0); 640 } 641 642 /* 643 * We have to open the device before we can perform ioctls on it, 644 * or the slice/label data may not be present. Device opens are 645 * usually tracked by specfs, but the dump device can be set in 646 * early boot and may not be open so this is somewhat of a hack. 647 */ 648 doopen = (dev->si_sysref.refcnt == 1); 649 if (doopen) { 650 error = dev_dopen(dev, FREAD, S_IFCHR, proc0.p_ucred); 651 if (error) 652 return (error); 653 } 654 error = disk_dumpconf(dev, 1/*on*/); 655 656 return error; 657 } 658 659 /* ARGSUSED */ 660 static void dump_conf (void *dummy); 661 static void 662 dump_conf(void *dummy) 663 { 664 char *path; 665 cdev_t dev; 666 667 path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK); 668 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { 669 sync_devs(); 670 dev = kgetdiskbyname(path); 671 if (dev != NULL) 672 dumpdev = dev; 673 } 674 kfree(path, M_TEMP); 675 if (setdumpdev(dumpdev) != 0) 676 dumpdev = NULL; 677 } 678 679 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL) 680 681 static int 682 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 683 { 684 int error; 685 udev_t ndumpdev; 686 687 ndumpdev = dev2udev(dumpdev); 688 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 689 if (error == 0 && req->newptr != NULL) 690 error = setdumpdev(udev2dev(ndumpdev, 0)); 691 return (error); 692 } 693 694 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 695 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", ""); 696 697 /* 698 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 699 * and then reboots. If we are called twice, then we avoid trying to sync 700 * the disks as this often leads to recursive panics. 701 */ 702 void 703 panic(const char *fmt, ...) 704 { 705 int bootopt, newpanic; 706 globaldata_t gd = mycpu; 707 thread_t td = gd->gd_curthread; 708 __va_list ap; 709 static char buf[256]; 710 711 #ifdef SMP 712 /* 713 * If a panic occurs on multiple cpus before the first is able to 714 * halt the other cpus, only one cpu is allowed to take the panic. 715 * Attempt to be verbose about this situation but if the kprintf() 716 * itself panics don't let us overrun the kernel stack. 717 * 718 * Be very nasty about descheduling our thread at the lowest 719 * level possible in an attempt to freeze the thread without 720 * inducing further panics. 721 * 722 * Bumping gd_trap_nesting_level will also bypass assertions in 723 * lwkt_switch() and allow us to switch away even if we are a 724 * FAST interrupt or IPI. 725 * 726 * The setting of panic_cpu_gd also determines how kprintf() 727 * spin-locks itself. DDB can set panic_cpu_gd as well. 728 */ 729 for (;;) { 730 globaldata_t xgd = panic_cpu_gd; 731 732 /* 733 * Someone else got the panic cpu 734 */ 735 if (xgd && xgd != gd) { 736 crit_enter(); 737 ++mycpu->gd_trap_nesting_level; 738 if (mycpu->gd_trap_nesting_level < 25) { 739 kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n", 740 mycpu->gd_cpuid, td); 741 } 742 td->td_release = NULL; /* be a grinch */ 743 for (;;) { 744 lwkt_deschedule_self(td); 745 lwkt_switch(); 746 } 747 /* NOT REACHED */ 748 /* --mycpu->gd_trap_nesting_level */ 749 /* crit_exit() */ 750 } 751 752 /* 753 * Reentrant panic 754 */ 755 if (xgd && xgd == gd) 756 break; 757 758 /* 759 * We got it 760 */ 761 if (atomic_cmpset_ptr(&panic_cpu_gd, NULL, gd)) 762 break; 763 } 764 #else 765 panic_cpu_gd = gd; 766 #endif 767 /* 768 * Try to get the system into a working state. Save information 769 * we are about to destroy. 770 */ 771 kvcreinitspin(); 772 if (panicstr == NULL) { 773 bcopy(td->td_toks_array, panic_tokens, sizeof(panic_tokens)); 774 panic_tokens_count = td->td_toks_stop - &td->td_toks_base; 775 } 776 lwkt_relalltokens(td); 777 td->td_toks_stop = &td->td_toks_base; 778 779 /* 780 * Setup 781 */ 782 bootopt = RB_AUTOBOOT | RB_DUMP; 783 if (sync_on_panic == 0) 784 bootopt |= RB_NOSYNC; 785 newpanic = 0; 786 if (panicstr) { 787 bootopt |= RB_NOSYNC; 788 } else { 789 panicstr = fmt; 790 newpanic = 1; 791 } 792 793 /* 794 * Format the panic string. 795 */ 796 __va_start(ap, fmt); 797 kvsnprintf(buf, sizeof(buf), fmt, ap); 798 if (panicstr == fmt) 799 panicstr = buf; 800 __va_end(ap); 801 kprintf("panic: %s\n", buf); 802 #ifdef SMP 803 /* two separate prints in case of an unmapped page and trap */ 804 kprintf("cpuid = %d\n", mycpu->gd_cpuid); 805 #endif 806 807 #if (NGPIO > 0) && defined(ERROR_LED_ON_PANIC) 808 led_switch("error", 1); 809 #endif 810 811 #if defined(WDOG_DISABLE_ON_PANIC) && defined(WATCHDOG_ENABLE) 812 wdog_disable(); 813 #endif 814 815 /* 816 * Enter the debugger or fall through & dump. Entering the 817 * debugger will stop cpus. If not entering the debugger stop 818 * cpus here. 819 */ 820 #if defined(DDB) 821 if (newpanic && trace_on_panic) 822 print_backtrace(-1); 823 if (debugger_on_panic) 824 Debugger("panic"); 825 else 826 #endif 827 #ifdef SMP 828 if (newpanic) 829 stop_cpus(mycpu->gd_other_cpus); 830 #else 831 ; 832 #endif 833 boot(bootopt); 834 } 835 836 /* 837 * Support for poweroff delay. 838 */ 839 #ifndef POWEROFF_DELAY 840 # define POWEROFF_DELAY 5000 841 #endif 842 static int poweroff_delay = POWEROFF_DELAY; 843 844 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 845 &poweroff_delay, 0, ""); 846 847 static void 848 poweroff_wait(void *junk, int howto) 849 { 850 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 851 return; 852 DELAY(poweroff_delay * 1000); 853 } 854 855 /* 856 * Some system processes (e.g. syncer) need to be stopped at appropriate 857 * points in their main loops prior to a system shutdown, so that they 858 * won't interfere with the shutdown process (e.g. by holding a disk buf 859 * to cause sync to fail). For each of these system processes, register 860 * shutdown_kproc() as a handler for one of shutdown events. 861 */ 862 static int kproc_shutdown_wait = 60; 863 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 864 &kproc_shutdown_wait, 0, ""); 865 866 void 867 shutdown_kproc(void *arg, int howto) 868 { 869 struct thread *td; 870 struct proc *p; 871 int error; 872 873 if (panicstr) 874 return; 875 876 td = (struct thread *)arg; 877 if ((p = td->td_proc) != NULL) { 878 kprintf("Waiting (max %d seconds) for system process `%s' to stop...", 879 kproc_shutdown_wait, p->p_comm); 880 } else { 881 kprintf("Waiting (max %d seconds) for system thread %s to stop...", 882 kproc_shutdown_wait, td->td_comm); 883 } 884 error = suspend_kproc(td, kproc_shutdown_wait * hz); 885 886 if (error == EWOULDBLOCK) 887 kprintf("timed out\n"); 888 else 889 kprintf("stopped\n"); 890 } 891 892 /* Registration of dumpers */ 893 int 894 set_dumper(struct dumperinfo *di) 895 { 896 if (di == NULL) { 897 bzero(&dumper, sizeof(dumper)); 898 return 0; 899 } 900 901 if (dumper.dumper != NULL) 902 return (EBUSY); 903 904 dumper = *di; 905 return 0; 906 } 907 908 void 909 dumpsys(void) 910 { 911 #if defined (_KERNEL_VIRTUAL) 912 /* VKERNELs don't support dumps */ 913 kprintf("VKERNEL doesn't support dumps\n"); 914 return; 915 #endif 916 /* 917 * If there is a dumper registered and we aren't dumping already, call 918 * the machine dependent dumpsys (md_dumpsys) to do the hard work. 919 * 920 * XXX: while right now the md_dumpsys() of x86 and x86_64 could be 921 * factored out completely into here, I rather keep them machine 922 * dependent in case we ever add a platform which does not share 923 * the same dumpsys() code, such as arm. 924 */ 925 if (dumper.dumper != NULL && !dumping) { 926 dumping++; 927 md_dumpsys(&dumper); 928 } 929 } 930 931 int dump_stop_usertds = 0; 932 933 #ifdef SMP 934 static 935 void 936 need_user_resched_remote(void *dummy) 937 { 938 need_user_resched(); 939 } 940 #endif 941 942 void 943 dump_reactivate_cpus(void) 944 { 945 #ifdef SMP 946 globaldata_t gd; 947 int cpu, seq; 948 #endif 949 950 dump_stop_usertds = 1; 951 952 need_user_resched(); 953 954 #ifdef SMP 955 for (cpu = 0; cpu < ncpus; cpu++) { 956 gd = globaldata_find(cpu); 957 seq = lwkt_send_ipiq(gd, need_user_resched_remote, NULL); 958 lwkt_wait_ipiq(gd, seq); 959 } 960 961 restart_cpus(stopped_cpus); 962 #endif 963 } 964