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. 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 * $FreeBSD: src/sys/kern/kern_shutdown.c,v 1.72.2.12 2002/02/21 19:15:10 dillon Exp $ 36 */ 37 38 #include "opt_ddb.h" 39 #include "opt_ddb_trace.h" 40 #include "opt_panic.h" 41 #include "opt_show_busybufs.h" 42 #include "use_gpio.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/eventhandler.h> 47 #include <sys/buf.h> 48 #include <sys/disk.h> 49 #include <sys/diskslice.h> 50 #include <sys/reboot.h> 51 #include <sys/proc.h> 52 #include <sys/priv.h> 53 #include <sys/fcntl.h> /* FREAD */ 54 #include <sys/stat.h> /* S_IFCHR */ 55 #include <sys/vnode.h> 56 #include <sys/kernel.h> 57 #include <sys/kerneldump.h> 58 #include <sys/kthread.h> 59 #include <sys/malloc.h> 60 #include <sys/mount.h> 61 #include <sys/queue.h> 62 #include <sys/sysctl.h> 63 #include <sys/vkernel.h> 64 #include <sys/conf.h> 65 #include <sys/sysproto.h> 66 #include <sys/device.h> 67 #include <sys/cons.h> 68 #include <sys/kbio.h> 69 #include <sys/shm.h> 70 #include <sys/kern_syscall.h> 71 #include <vm/vm_map.h> 72 #include <vm/pmap.h> 73 74 #include <sys/thread2.h> 75 #include <sys/buf2.h> 76 #include <sys/mplock2.h> 77 78 #include <machine/cpu.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 #include <machine/thread.h> 84 85 #include <sys/signalvar.h> 86 87 #include <sys/wdog.h> 88 #include <dev/acpica/acpi_pvpanic/panic_notifier.h> 89 #include <dev/misc/gpio/gpio.h> 90 91 #ifndef PANIC_REBOOT_WAIT_TIME 92 #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ 93 #endif 94 95 /* 96 * Note that stdarg.h and the ANSI style va_start macro is used for both 97 * ANSI and traditional C compilers. We use the machine version to stay 98 * within the confines of the kernel header files. 99 */ 100 #include <machine/stdarg.h> 101 102 #ifdef DDB 103 #include <ddb/ddb.h> 104 #ifdef DDB_UNATTENDED 105 int debugger_on_panic = 0; 106 #else 107 int debugger_on_panic = 1; 108 #endif 109 SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RW, 110 &debugger_on_panic, 0, "Run debugger on kernel panic"); 111 112 #ifdef DDB_TRACE 113 int trace_on_panic = 1; 114 #else 115 int trace_on_panic = 0; 116 #endif 117 SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RW, 118 &trace_on_panic, 0, "Print stack trace on kernel panic"); 119 #endif 120 121 static int sync_on_panic = 0; 122 SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RW, 123 &sync_on_panic, 0, "Do a sync before rebooting from a panic"); 124 125 SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); 126 127 /* 128 * Variable panicstr contains argument to first call to panic; used as flag 129 * to indicate that the kernel has already called panic. 130 */ 131 const char *panicstr; 132 133 int dumping; /* system is dumping */ 134 static struct dumperinfo dumper; /* selected dumper */ 135 136 globaldata_t panic_cpu_gd; /* which cpu took the panic */ 137 struct lwkt_tokref panic_tokens[LWKT_MAXTOKENS]; 138 int panic_tokens_count; 139 140 int bootverbose = 0; /* note: assignment to force non-bss */ 141 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, 142 &bootverbose, 0, "Verbose kernel messages"); 143 144 int cold = 1; /* note: assignment to force non-bss */ 145 int dumplo; /* OBSOLETE - savecore compat */ 146 u_int64_t dumplo64; 147 148 static void boot (int) __dead2; 149 static int setdumpdev (cdev_t dev); 150 static void poweroff_wait (void *, int); 151 static void print_uptime (void); 152 static void shutdown_halt (void *junk, int howto); 153 static void shutdown_panic (void *junk, int howto); 154 static void shutdown_reset (void *junk, int howto); 155 static int shutdown_busycount1(struct buf *bp, void *info); 156 static int shutdown_busycount2(struct buf *bp, void *info); 157 static void shutdown_cleanup_proc(struct proc *p); 158 159 /* register various local shutdown events */ 160 static void 161 shutdown_conf(void *unused) 162 { 163 EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); 164 EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); 165 EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); 166 EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); 167 } 168 169 SYSINIT(shutdown_conf, SI_BOOT2_MACHDEP, SI_ORDER_ANY, shutdown_conf, NULL); 170 171 /* ARGSUSED */ 172 173 /* 174 * The system call that results in a reboot 175 * 176 * MPALMOSTSAFE 177 */ 178 int 179 sys_reboot(struct reboot_args *uap) 180 { 181 struct thread *td = curthread; 182 int error; 183 184 if ((error = priv_check(td, PRIV_REBOOT))) 185 return (error); 186 187 get_mplock(); 188 boot(uap->opt); 189 rel_mplock(); 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 struct pcb dumppcb; 214 struct thread *dumpthread; 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 /* 263 * We really want to shutdown on the BSP. Subsystems such as ACPI 264 * can't power-down the box otherwise. 265 */ 266 if (!CPUMASK_ISUP(smp_active_mask)) { 267 kprintf("boot() called on cpu#%d\n", mycpu->gd_cpuid); 268 } 269 if (panicstr == NULL && mycpu->gd_cpuid != 0) { 270 kprintf("Switching to cpu #0 for shutdown\n"); 271 lwkt_setcpu_self(globaldata_find(0)); 272 } 273 /* 274 * Do any callouts that should be done BEFORE syncing the filesystems. 275 */ 276 EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); 277 278 /* 279 * Try to get rid of any remaining FS references. The calling 280 * process, proc0, and init may still hold references. The 281 * VFS cache subsystem may still hold a root reference to root. 282 * 283 * XXX this needs work. We really need to SIGSTOP all remaining 284 * processes in order to avoid blowups due to proc0's filesystem 285 * references going away. For now just make sure that the init 286 * process is stopped. 287 */ 288 if (panicstr == NULL) { 289 shutdown_cleanup_proc(curproc); 290 shutdown_cleanup_proc(&proc0); 291 if (initproc) { 292 if (initproc != curproc) { 293 ksignal(initproc, SIGSTOP); 294 tsleep(boot, 0, "shutdn", hz / 20); 295 } 296 shutdown_cleanup_proc(initproc); 297 } 298 vfs_cache_setroot(NULL, NULL); 299 } 300 301 /* 302 * Now sync filesystems 303 */ 304 if (!cold && (howto & RB_NOSYNC) == 0 && waittime < 0) { 305 int iter, nbusy, pbusy; 306 307 waittime = 0; 308 kprintf("\nsyncing disks... "); 309 310 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 311 312 /* 313 * With soft updates, some buffers that are 314 * written will be remarked as dirty until other 315 * buffers are written. 316 */ 317 for (iter = pbusy = 0; iter < 20; iter++) { 318 nbusy = scan_all_buffers(shutdown_busycount1, NULL); 319 if (nbusy == 0) 320 break; 321 kprintf("%d ", nbusy); 322 if (nbusy < pbusy) 323 iter = 0; 324 pbusy = nbusy; 325 /* 326 * XXX: 327 * Process soft update work queue if buffers don't sync 328 * after 6 iterations by permitting the syncer to run. 329 */ 330 if (iter > 5) 331 bio_ops_sync(NULL); 332 333 sys_sync(NULL); /* YYY was sync(&proc0, NULL). why proc0 ? */ 334 tsleep(boot, 0, "shutdn", hz * iter / 20 + 1); 335 } 336 kprintf("\n"); 337 /* 338 * Count only busy local buffers to prevent forcing 339 * a fsck if we're just a client of a wedged NFS server 340 */ 341 nbusy = scan_all_buffers(shutdown_busycount2, NULL); 342 if (nbusy) { 343 /* 344 * Failed to sync all blocks. Indicate this and don't 345 * unmount filesystems (thus forcing an fsck on reboot). 346 */ 347 kprintf("giving up on %d buffers\n", nbusy); 348 #ifdef DDB 349 if (debugger_on_panic) 350 Debugger("busy buffer problem"); 351 #endif /* DDB */ 352 tsleep(boot, 0, "shutdn", hz * 5 + 1); 353 } else { 354 kprintf("done\n"); 355 /* 356 * Unmount filesystems 357 */ 358 if (panicstr == NULL) 359 vfs_unmountall(); 360 } 361 tsleep(boot, 0, "shutdn", hz / 10 + 1); 362 } 363 364 print_uptime(); 365 366 /* 367 * Dump before doing post_sync shutdown ops 368 */ 369 crit_enter(); 370 if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold) { 371 dumpsys(); 372 } 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 /* 389 * Pass 1 - Figure out if there are any busy or dirty buffers still present. 390 * 391 * We ignore TMPFS mounts in this pass. 392 */ 393 static int 394 shutdown_busycount1(struct buf *bp, void *info) 395 { 396 struct vnode *vp; 397 398 if ((vp = bp->b_vp) != NULL && vp->v_tag == VT_TMPFS) 399 return (0); 400 if ((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp) > 0) 401 return(1); 402 if ((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) 403 return (1); 404 return (0); 405 } 406 407 /* 408 * Pass 2 - only run after pass 1 has completed or has given up 409 * 410 * We ignore TMPFS, NFS, MFS, and SMBFS mounts in this pass. 411 */ 412 static int 413 shutdown_busycount2(struct buf *bp, void *info) 414 { 415 struct vnode *vp; 416 417 /* 418 * Ignore tmpfs and nfs mounts 419 */ 420 if ((vp = bp->b_vp) != NULL) { 421 if (vp->v_tag == VT_TMPFS) 422 return (0); 423 if (vp->v_tag == VT_NFS) 424 return (0); 425 if (vp->v_tag == VT_MFS) 426 return (0); 427 if (vp->v_tag == VT_SMBFS) 428 return (0); 429 } 430 431 /* 432 * Only count buffers stuck on I/O, ignore everything else 433 */ 434 if (((bp->b_flags & B_INVAL) == 0 && BUF_REFCNT(bp)) || 435 ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI)) { 436 /* 437 * Only count buffers undergoing write I/O 438 * on the related vnode. 439 */ 440 if (bp->b_vp == NULL || 441 bio_track_active(&bp->b_vp->v_track_write) == 0) { 442 return (0); 443 } 444 #if defined(SHOW_BUSYBUFS) || defined(DIAGNOSTIC) 445 kprintf( 446 "%p dev:?, flags:%08x, loffset:%jd, doffset:%jd\n", 447 bp, 448 bp->b_flags, (intmax_t)bp->b_loffset, 449 (intmax_t)bp->b_bio2.bio_offset); 450 #endif 451 return(1); 452 } 453 return(0); 454 } 455 456 /* 457 * If the shutdown was a clean halt, behave accordingly. 458 */ 459 static void 460 shutdown_halt(void *junk, int howto) 461 { 462 if (howto & RB_HALT) { 463 kprintf("\n"); 464 kprintf("The operating system has halted.\n"); 465 #ifdef _KERNEL_VIRTUAL 466 cpu_halt(); 467 #else 468 kprintf("Please press any key to reboot.\n\n"); 469 switch (cngetc()) { 470 case -1: /* No console, just die */ 471 cpu_halt(); 472 /* NOTREACHED */ 473 default: 474 howto &= ~RB_HALT; 475 break; 476 } 477 #endif 478 } 479 } 480 481 /* 482 * Check to see if the system paniced, pause and then reboot 483 * according to the specified delay. 484 */ 485 static void 486 shutdown_panic(void *junk, int howto) 487 { 488 int loop; 489 int c; 490 491 if (howto & RB_DUMP) { 492 if (PANIC_REBOOT_WAIT_TIME != 0) { 493 if (PANIC_REBOOT_WAIT_TIME != -1) { 494 kprintf("Automatic reboot in %d seconds - " 495 "press a key on the console to abort\n", 496 PANIC_REBOOT_WAIT_TIME); 497 for (loop = PANIC_REBOOT_WAIT_TIME * 10; 498 loop > 0; --loop) { 499 DELAY(1000 * 100); /* 1/10th second */ 500 /* Did user type a key? */ 501 c = cncheckc(); 502 if (c != -1 && c != NOKEY) 503 break; 504 } 505 if (!loop) 506 return; 507 } 508 } else { /* zero time specified - reboot NOW */ 509 return; 510 } 511 kprintf("--> Press a key on the console to reboot,\n"); 512 kprintf("--> or switch off the system now.\n"); 513 cngetc(); 514 } 515 } 516 517 /* 518 * Everything done, now reset 519 */ 520 static void 521 shutdown_reset(void *junk, int howto) 522 { 523 kprintf("Rebooting...\n"); 524 DELAY(1000000); /* wait 1 sec for kprintf's to complete and be read */ 525 /* cpu_boot(howto); */ /* doesn't do anything at the moment */ 526 cpu_reset(); 527 /* NOTREACHED */ /* assuming reset worked */ 528 } 529 530 /* 531 * Try to remove FS references in the specified process. This function 532 * is used during shutdown 533 */ 534 static 535 void 536 shutdown_cleanup_proc(struct proc *p) 537 { 538 struct filedesc *fdp; 539 struct vmspace *vm; 540 541 if (p == NULL) 542 return; 543 if ((fdp = p->p_fd) != NULL) { 544 kern_closefrom(0); 545 if (fdp->fd_cdir) { 546 cache_drop(&fdp->fd_ncdir); 547 vrele(fdp->fd_cdir); 548 fdp->fd_cdir = NULL; 549 } 550 if (fdp->fd_rdir) { 551 cache_drop(&fdp->fd_nrdir); 552 vrele(fdp->fd_rdir); 553 fdp->fd_rdir = NULL; 554 } 555 if (fdp->fd_jdir) { 556 cache_drop(&fdp->fd_njdir); 557 vrele(fdp->fd_jdir); 558 fdp->fd_jdir = NULL; 559 } 560 } 561 if (p->p_vkernel) 562 vkernel_exit(p); 563 if (p->p_textvp) { 564 vrele(p->p_textvp); 565 p->p_textvp = NULL; 566 } 567 vm = p->p_vmspace; 568 if (vm != NULL) { 569 pmap_remove_pages(vmspace_pmap(vm), 570 VM_MIN_USER_ADDRESS, 571 VM_MAX_USER_ADDRESS); 572 vm_map_remove(&vm->vm_map, 573 VM_MIN_USER_ADDRESS, 574 VM_MAX_USER_ADDRESS); 575 } 576 } 577 578 /* 579 * Magic number for savecore 580 * 581 * exported (symorder) and used at least by savecore(8) 582 * 583 * Mark it as used so that gcc doesn't optimize it away. 584 */ 585 __attribute__((__used__)) 586 static u_long const dumpmag = 0x8fca0101UL; 587 588 __attribute__((__used__)) 589 static int dumpsize = 0; /* also for savecore */ 590 591 static int dodump = 1; 592 593 SYSCTL_INT(_machdep, OID_AUTO, do_dump, CTLFLAG_RW, &dodump, 0, 594 "Try to perform coredump on kernel panic"); 595 596 void 597 mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver, 598 uint64_t dumplen, uint32_t blksz) 599 { 600 bzero(kdh, sizeof(*kdh)); 601 strncpy(kdh->magic, magic, sizeof(kdh->magic)); 602 strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture)); 603 kdh->version = htod32(KERNELDUMPVERSION); 604 kdh->architectureversion = htod32(archver); 605 kdh->dumplength = htod64(dumplen); 606 kdh->dumptime = htod64(time_second); 607 kdh->blocksize = htod32(blksz); 608 strncpy(kdh->hostname, hostname, sizeof(kdh->hostname)); 609 strncpy(kdh->versionstring, version, sizeof(kdh->versionstring)); 610 if (panicstr != NULL) 611 strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring)); 612 kdh->parity = kerneldump_parity(kdh); 613 } 614 615 static int 616 setdumpdev(cdev_t dev) 617 { 618 int error; 619 int doopen; 620 621 if (dev == NULL) { 622 disk_dumpconf(NULL, 0/*off*/); 623 return (0); 624 } 625 626 /* 627 * We have to open the device before we can perform ioctls on it, 628 * or the slice/label data may not be present. Device opens are 629 * usually tracked by specfs, but the dump device can be set in 630 * early boot and may not be open so this is somewhat of a hack. 631 */ 632 doopen = (dev->si_sysref.refcnt == 1); 633 if (doopen) { 634 error = dev_dopen(dev, FREAD, S_IFCHR, proc0.p_ucred, NULL); 635 if (error) 636 return (error); 637 } 638 error = disk_dumpconf(dev, 1/*on*/); 639 640 return error; 641 } 642 643 /* ARGSUSED */ 644 static void dump_conf (void *dummy); 645 static void 646 dump_conf(void *dummy) 647 { 648 char *path; 649 cdev_t dev; 650 int _dummy; 651 652 path = kmalloc(MNAMELEN, M_TEMP, M_WAITOK); 653 if (TUNABLE_STR_FETCH("dumpdev", path, MNAMELEN) != 0) { 654 /* 655 * Make sure all disk devices created so far have also been 656 * probed, and also make sure that the newly created device 657 * nodes for probed disks are ready, too. 658 * 659 * XXX - Delay an additional 2 seconds to help drivers which 660 * pickup devices asynchronously and are not caught by 661 * CAM's initial probe. 662 */ 663 sync_devs(); 664 tsleep(&_dummy, 0, "syncer", hz*2); 665 666 dev = kgetdiskbyname(path); 667 if (dev != NULL) 668 dumpdev = dev; 669 } 670 kfree(path, M_TEMP); 671 if (setdumpdev(dumpdev) != 0) 672 dumpdev = NULL; 673 } 674 675 SYSINIT(dump_conf, SI_SUB_DUMP_CONF, SI_ORDER_FIRST, dump_conf, NULL); 676 677 static int 678 sysctl_kern_dumpdev(SYSCTL_HANDLER_ARGS) 679 { 680 int error; 681 udev_t ndumpdev; 682 683 ndumpdev = dev2udev(dumpdev); 684 error = sysctl_handle_opaque(oidp, &ndumpdev, sizeof ndumpdev, req); 685 if (error == 0 && req->newptr != NULL) 686 error = setdumpdev(udev2dev(ndumpdev, 0)); 687 return (error); 688 } 689 690 SYSCTL_PROC(_kern, KERN_DUMPDEV, dumpdev, CTLTYPE_OPAQUE|CTLFLAG_RW, 691 0, sizeof dumpdev, sysctl_kern_dumpdev, "T,udev_t", ""); 692 693 static struct panicerinfo *panic_notifier; 694 695 int 696 set_panic_notifier(struct panicerinfo *info) 697 { 698 if (info == NULL) 699 panic_notifier = NULL; 700 else if (panic_notifier != NULL) 701 return 1; 702 else 703 panic_notifier = info; 704 705 return 0; 706 } 707 708 /* 709 * Panic is called on unresolvable fatal errors. It prints "panic: mesg", 710 * and then reboots. If we are called twice, then we avoid trying to sync 711 * the disks as this often leads to recursive panics. 712 */ 713 void 714 panic(const char *fmt, ...) 715 { 716 int bootopt, newpanic; 717 globaldata_t gd = mycpu; 718 thread_t td = gd->gd_curthread; 719 __va_list ap; 720 static char buf[256]; 721 722 /* 723 * If a panic occurs on multiple cpus before the first is able to 724 * halt the other cpus, only one cpu is allowed to take the panic. 725 * Attempt to be verbose about this situation but if the kprintf() 726 * itself panics don't let us overrun the kernel stack. 727 * 728 * Be very nasty about descheduling our thread at the lowest 729 * level possible in an attempt to freeze the thread without 730 * inducing further panics. 731 * 732 * Bumping gd_trap_nesting_level will also bypass assertions in 733 * lwkt_switch() and allow us to switch away even if we are a 734 * FAST interrupt or IPI. 735 * 736 * The setting of panic_cpu_gd also determines how kprintf() 737 * spin-locks itself. DDB can set panic_cpu_gd as well. 738 */ 739 for (;;) { 740 globaldata_t xgd = panic_cpu_gd; 741 742 /* 743 * Someone else got the panic cpu 744 */ 745 if (xgd && xgd != gd) { 746 crit_enter(); 747 ++mycpu->gd_trap_nesting_level; 748 if (mycpu->gd_trap_nesting_level < 25) { 749 kprintf("SECONDARY PANIC ON CPU %d THREAD %p\n", 750 mycpu->gd_cpuid, td); 751 } 752 td->td_release = NULL; /* be a grinch */ 753 for (;;) { 754 lwkt_deschedule_self(td); 755 lwkt_switch(); 756 } 757 /* NOT REACHED */ 758 /* --mycpu->gd_trap_nesting_level */ 759 /* crit_exit() */ 760 } 761 762 /* 763 * Reentrant panic 764 */ 765 if (xgd && xgd == gd) 766 break; 767 768 /* 769 * We got it 770 */ 771 if (atomic_cmpset_ptr(&panic_cpu_gd, NULL, gd)) 772 break; 773 } 774 /* 775 * Try to get the system into a working state. Save information 776 * we are about to destroy. 777 */ 778 kvcreinitspin(); 779 if (panicstr == NULL) { 780 bcopy(td->td_toks_array, panic_tokens, sizeof(panic_tokens)); 781 panic_tokens_count = td->td_toks_stop - &td->td_toks_base; 782 } 783 lwkt_relalltokens(td); 784 td->td_toks_stop = &td->td_toks_base; 785 if (gd->gd_spinlocks) 786 kprintf("panic with %d spinlocks held\n", gd->gd_spinlocks); 787 gd->gd_spinlocks = 0; 788 789 /* 790 * Setup 791 */ 792 bootopt = RB_AUTOBOOT | RB_DUMP; 793 if (sync_on_panic == 0) 794 bootopt |= RB_NOSYNC; 795 newpanic = 0; 796 if (panicstr) { 797 bootopt |= RB_NOSYNC; 798 } else { 799 panicstr = fmt; 800 newpanic = 1; 801 } 802 803 /* 804 * Format the panic string. 805 */ 806 __va_start(ap, fmt); 807 kvsnprintf(buf, sizeof(buf), fmt, ap); 808 if (panicstr == fmt) 809 panicstr = buf; 810 __va_end(ap); 811 if (panic_notifier != NULL) 812 panic_notifier->notifier(panic_notifier->arg); 813 kprintf("panic: %s\n", buf); 814 /* two separate prints in case of an unmapped page and trap */ 815 kprintf("cpuid = %d\n", mycpu->gd_cpuid); 816 817 #if (NGPIO > 0) && defined(ERROR_LED_ON_PANIC) 818 led_switch("error", 1); 819 #endif 820 821 #if defined(WDOG_DISABLE_ON_PANIC) 822 wdog_disable(); 823 #endif 824 825 /* 826 * Make sure kgdb knows who we are, there won't be a stoppcbs[] 827 * entry since our cpu wasn't stopped. 828 */ 829 savectx(&dumppcb); 830 dumpthread = curthread; 831 832 /* 833 * Enter the debugger or fall through & dump. Entering the 834 * debugger will stop cpus. If not entering the debugger stop 835 * cpus here. 836 * 837 * Limit the trace history to leave more panic data on a 838 * potentially row-limited console. 839 */ 840 841 #if defined(DDB) 842 if (newpanic && trace_on_panic) 843 print_backtrace(6); 844 if (debugger_on_panic) 845 Debugger("panic"); 846 else 847 #endif 848 if (newpanic) 849 stop_cpus(mycpu->gd_other_cpus); 850 boot(bootopt); 851 } 852 853 /* 854 * Support for poweroff delay. 855 */ 856 #ifndef POWEROFF_DELAY 857 # define POWEROFF_DELAY 5000 858 #endif 859 static int poweroff_delay = POWEROFF_DELAY; 860 861 SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, 862 &poweroff_delay, 0, ""); 863 864 static void 865 poweroff_wait(void *junk, int howto) 866 { 867 if(!(howto & RB_POWEROFF) || poweroff_delay <= 0) 868 return; 869 DELAY(poweroff_delay * 1000); 870 } 871 872 /* 873 * Some system processes (e.g. syncer) need to be stopped at appropriate 874 * points in their main loops prior to a system shutdown, so that they 875 * won't interfere with the shutdown process (e.g. by holding a disk buf 876 * to cause sync to fail). For each of these system processes, register 877 * shutdown_kproc() as a handler for one of shutdown events. 878 */ 879 static int kproc_shutdown_wait = 60; 880 SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, 881 &kproc_shutdown_wait, 0, ""); 882 883 void 884 shutdown_kproc(void *arg, int howto) 885 { 886 struct thread *td; 887 struct proc *p; 888 int error; 889 890 if (panicstr) 891 return; 892 893 td = (struct thread *)arg; 894 if ((p = td->td_proc) != NULL) { 895 kprintf("Waiting (max %d seconds) for system process `%s' to stop...", 896 kproc_shutdown_wait, p->p_comm); 897 } else { 898 kprintf("Waiting (max %d seconds) for system thread %s to stop...", 899 kproc_shutdown_wait, td->td_comm); 900 } 901 error = suspend_kproc(td, kproc_shutdown_wait * hz); 902 903 if (error == EWOULDBLOCK) 904 kprintf("timed out\n"); 905 else 906 kprintf("stopped\n"); 907 } 908 909 /* Registration of dumpers */ 910 int 911 set_dumper(struct dumperinfo *di) 912 { 913 if (di == NULL) { 914 bzero(&dumper, sizeof(dumper)); 915 return 0; 916 } 917 918 if (dumper.dumper != NULL) 919 return (EBUSY); 920 921 dumper = *di; 922 return 0; 923 } 924 925 void 926 dumpsys(void) 927 { 928 #if defined (_KERNEL_VIRTUAL) 929 /* vkernels don't support dumps */ 930 kprintf("vkernels don't support dumps\n"); 931 return; 932 #endif 933 /* 934 * If there is a dumper registered and we aren't dumping already, call 935 * the machine dependent dumpsys (md_dumpsys) to do the hard work. 936 * 937 * XXX: while right now the md_dumpsys() of x86 and x86_64 could be 938 * factored out completely into here, I rather keep them machine 939 * dependent in case we ever add a platform which does not share 940 * the same dumpsys() code, such as arm. 941 */ 942 if (dumper.dumper != NULL && !dumping) { 943 dumping++; 944 md_dumpsys(&dumper); 945 } 946 } 947 948 int dump_stop_usertds = 0; 949 950 static 951 void 952 need_user_resched_remote(void *dummy) 953 { 954 need_user_resched(); 955 } 956 957 void 958 dump_reactivate_cpus(void) 959 { 960 globaldata_t gd; 961 int cpu, seq; 962 963 dump_stop_usertds = 1; 964 965 need_user_resched(); 966 967 for (cpu = 0; cpu < ncpus; cpu++) { 968 gd = globaldata_find(cpu); 969 seq = lwkt_send_ipiq(gd, need_user_resched_remote, NULL); 970 lwkt_wait_ipiq(gd, seq); 971 } 972 973 restart_cpus(stopped_cpus); 974 } 975