xref: /netbsd-src/sys/kern/kern_subr.c (revision 21e37cc72a480a47828990a439cde7ac9ffaf0c6)
1 /*	$NetBSD: kern_subr.c,v 1.111 2004/04/21 20:31:50 matt Exp $	*/
2 
3 /*-
4  * Copyright (c) 1997, 1998, 1999, 2002 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center, and by Luke Mewburn.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1991, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  * (c) UNIX System Laboratories, Inc.
44  * All or some portions of this file are derived from material licensed
45  * to the University of California by American Telephone and Telegraph
46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47  * the permission of UNIX System Laboratories, Inc.
48  *
49  * Copyright (c) 1992, 1993
50  *	The Regents of the University of California.  All rights reserved.
51  *
52  * This software was developed by the Computer Systems Engineering group
53  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
54  * contributed to Berkeley.
55  *
56  * All advertising materials mentioning features or use of this software
57  * must display the following acknowledgement:
58  *	This product includes software developed by the University of
59  *	California, Lawrence Berkeley Laboratory.
60  *
61  * Redistribution and use in source and binary forms, with or without
62  * modification, are permitted provided that the following conditions
63  * are met:
64  * 1. Redistributions of source code must retain the above copyright
65  *    notice, this list of conditions and the following disclaimer.
66  * 2. Redistributions in binary form must reproduce the above copyright
67  *    notice, this list of conditions and the following disclaimer in the
68  *    documentation and/or other materials provided with the distribution.
69  * 3. Neither the name of the University nor the names of its contributors
70  *    may be used to endorse or promote products derived from this software
71  *    without specific prior written permission.
72  *
73  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
74  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
75  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
76  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
77  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
78  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
79  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
80  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
81  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
82  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
83  * SUCH DAMAGE.
84  *
85  *	@(#)kern_subr.c	8.4 (Berkeley) 2/14/95
86  */
87 
88 #include <sys/cdefs.h>
89 __KERNEL_RCSID(0, "$NetBSD: kern_subr.c,v 1.111 2004/04/21 20:31:50 matt Exp $");
90 
91 #include "opt_ddb.h"
92 #include "opt_md.h"
93 #include "opt_syscall_debug.h"
94 #include "opt_ktrace.h"
95 #include "opt_systrace.h"
96 
97 #include <sys/param.h>
98 #include <sys/systm.h>
99 #include <sys/proc.h>
100 #include <sys/malloc.h>
101 #include <sys/mount.h>
102 #include <sys/device.h>
103 #include <sys/reboot.h>
104 #include <sys/conf.h>
105 #include <sys/disklabel.h>
106 #include <sys/queue.h>
107 #include <sys/systrace.h>
108 #include <sys/ktrace.h>
109 
110 #include <uvm/uvm_extern.h>
111 
112 #include <dev/cons.h>
113 
114 #include <net/if.h>
115 
116 /* XXX these should eventually move to subr_autoconf.c */
117 static struct device *finddevice(const char *);
118 static struct device *getdisk(char *, int, int, dev_t *, int);
119 static struct device *parsedisk(char *, int, int, dev_t *);
120 
121 /*
122  * A generic linear hook.
123  */
124 struct hook_desc {
125 	LIST_ENTRY(hook_desc) hk_list;
126 	void	(*hk_fn)(void *);
127 	void	*hk_arg;
128 };
129 typedef LIST_HEAD(, hook_desc) hook_list_t;
130 
131 static void *hook_establish(hook_list_t *, void (*)(void *), void *);
132 static void hook_disestablish(hook_list_t *, void *);
133 static void hook_destroy(hook_list_t *);
134 static void hook_proc_run(hook_list_t *, struct proc *);
135 
136 MALLOC_DEFINE(M_IOV, "iov", "large iov's");
137 
138 int
139 uiomove(buf, n, uio)
140 	void *buf;
141 	size_t n;
142 	struct uio *uio;
143 {
144 	struct iovec *iov;
145 	u_int cnt;
146 	int error = 0;
147 	char *cp = buf;
148 	struct proc *p = uio->uio_procp;
149 
150 #ifdef DIAGNOSTIC
151 	if (uio->uio_rw != UIO_READ && uio->uio_rw != UIO_WRITE)
152 		panic("uiomove: mode");
153 #endif
154 	while (n > 0 && uio->uio_resid) {
155 		iov = uio->uio_iov;
156 		cnt = iov->iov_len;
157 		if (cnt == 0) {
158 			KASSERT(uio->uio_iovcnt > 0);
159 			uio->uio_iov++;
160 			uio->uio_iovcnt--;
161 			continue;
162 		}
163 		if (cnt > n)
164 			cnt = n;
165 		switch (uio->uio_segflg) {
166 
167 		case UIO_USERSPACE:
168 			if (curcpu()->ci_schedstate.spc_flags &
169 			    SPCF_SHOULDYIELD)
170 				preempt(1);
171 			if (__predict_true(p == curproc)) {
172 				if (uio->uio_rw == UIO_READ)
173 					error = copyout(cp, iov->iov_base, cnt);
174 				else
175 					error = copyin(iov->iov_base, cp, cnt);
176 			} else {
177 				if (uio->uio_rw == UIO_READ)
178 					error = copyout_proc(p, cp,
179 					    iov->iov_base, cnt);
180 				else
181 					error = copyin_proc(p, iov->iov_base,
182 					    cp, cnt);
183 			}
184 			if (error)
185 				return (error);
186 			break;
187 
188 		case UIO_SYSSPACE:
189 			if (uio->uio_rw == UIO_READ)
190 				error = kcopy(cp, iov->iov_base, cnt);
191 			else
192 				error = kcopy(iov->iov_base, cp, cnt);
193 			if (error)
194 				return (error);
195 			break;
196 		}
197 		iov->iov_base = (caddr_t)iov->iov_base + cnt;
198 		iov->iov_len -= cnt;
199 		uio->uio_resid -= cnt;
200 		uio->uio_offset += cnt;
201 		cp += cnt;
202 		KDASSERT(cnt <= n);
203 		n -= cnt;
204 	}
205 	return (error);
206 }
207 
208 /*
209  * Wrapper for uiomove() that validates the arguments against a known-good
210  * kernel buffer.
211  */
212 int
213 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
214 {
215 	size_t offset;
216 
217 	if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
218 	    (offset = uio->uio_offset) != uio->uio_offset)
219 		return (EINVAL);
220 	if (offset >= buflen)
221 		return (0);
222 	return (uiomove((char *)buf + offset, buflen - offset, uio));
223 }
224 
225 /*
226  * Give next character to user as result of read.
227  */
228 int
229 ureadc(c, uio)
230 	int c;
231 	struct uio *uio;
232 {
233 	struct iovec *iov;
234 
235 	if (uio->uio_resid <= 0)
236 		panic("ureadc: non-positive resid");
237 again:
238 	if (uio->uio_iovcnt <= 0)
239 		panic("ureadc: non-positive iovcnt");
240 	iov = uio->uio_iov;
241 	if (iov->iov_len <= 0) {
242 		uio->uio_iovcnt--;
243 		uio->uio_iov++;
244 		goto again;
245 	}
246 	switch (uio->uio_segflg) {
247 
248 	case UIO_USERSPACE:
249 		if (subyte(iov->iov_base, c) < 0)
250 			return (EFAULT);
251 		break;
252 
253 	case UIO_SYSSPACE:
254 		*(char *)iov->iov_base = c;
255 		break;
256 	}
257 	iov->iov_base = (caddr_t)iov->iov_base + 1;
258 	iov->iov_len--;
259 	uio->uio_resid--;
260 	uio->uio_offset++;
261 	return (0);
262 }
263 
264 /*
265  * Like copyin(), but operates on an arbitrary process.
266  */
267 int
268 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
269 {
270 	struct iovec iov;
271 	struct uio uio;
272 	int error;
273 
274 	if (len == 0)
275 		return (0);
276 
277 	iov.iov_base = kaddr;
278 	iov.iov_len = len;
279 	uio.uio_iov = &iov;
280 	uio.uio_iovcnt = 1;
281 	uio.uio_offset = (off_t)(intptr_t)uaddr;
282 	uio.uio_resid = len;
283 	uio.uio_segflg = UIO_SYSSPACE;
284 	uio.uio_rw = UIO_READ;
285 	uio.uio_procp = NULL;
286 
287 	/* XXXCDC: how should locking work here? */
288 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
289 		return (EFAULT);
290 	p->p_vmspace->vm_refcnt++;	/* XXX */
291 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
292 	uvmspace_free(p->p_vmspace);
293 
294 	return (error);
295 }
296 
297 /*
298  * Like copyout(), but operates on an arbitrary process.
299  */
300 int
301 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
302 {
303 	struct iovec iov;
304 	struct uio uio;
305 	int error;
306 
307 	if (len == 0)
308 		return (0);
309 
310 	iov.iov_base = (void *) kaddr;	/* XXX cast away const */
311 	iov.iov_len = len;
312 	uio.uio_iov = &iov;
313 	uio.uio_iovcnt = 1;
314 	uio.uio_offset = (off_t)(intptr_t)uaddr;
315 	uio.uio_resid = len;
316 	uio.uio_segflg = UIO_SYSSPACE;
317 	uio.uio_rw = UIO_WRITE;
318 	uio.uio_procp = NULL;
319 
320 	/* XXXCDC: how should locking work here? */
321 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
322 		return (EFAULT);
323 	p->p_vmspace->vm_refcnt++;	/* XXX */
324 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
325 	uvmspace_free(p->p_vmspace);
326 
327 	return (error);
328 }
329 
330 /*
331  * General routine to allocate a hash table.
332  * Allocate enough memory to hold at least `elements' list-head pointers.
333  * Return a pointer to the allocated space and set *hashmask to a pattern
334  * suitable for masking a value to use as an index into the returned array.
335  */
336 void *
337 hashinit(elements, htype, mtype, mflags, hashmask)
338 	u_int elements;
339 	enum hashtype htype;
340 	struct malloc_type *mtype;
341 	int mflags;
342 	u_long *hashmask;
343 {
344 	u_long hashsize, i;
345 	LIST_HEAD(, generic) *hashtbl_list;
346 	TAILQ_HEAD(, generic) *hashtbl_tailq;
347 	size_t esize;
348 	void *p;
349 
350 	if (elements == 0)
351 		panic("hashinit: bad cnt");
352 	for (hashsize = 1; hashsize < elements; hashsize <<= 1)
353 		continue;
354 
355 	switch (htype) {
356 	case HASH_LIST:
357 		esize = sizeof(*hashtbl_list);
358 		break;
359 	case HASH_TAILQ:
360 		esize = sizeof(*hashtbl_tailq);
361 		break;
362 	default:
363 #ifdef DIAGNOSTIC
364 		panic("hashinit: invalid table type");
365 #else
366 		return NULL;
367 #endif
368 	}
369 
370 	if ((p = malloc(hashsize * esize, mtype, mflags)) == NULL)
371 		return (NULL);
372 
373 	switch (htype) {
374 	case HASH_LIST:
375 		hashtbl_list = p;
376 		for (i = 0; i < hashsize; i++)
377 			LIST_INIT(&hashtbl_list[i]);
378 		break;
379 	case HASH_TAILQ:
380 		hashtbl_tailq = p;
381 		for (i = 0; i < hashsize; i++)
382 			TAILQ_INIT(&hashtbl_tailq[i]);
383 		break;
384 	}
385 	*hashmask = hashsize - 1;
386 	return (p);
387 }
388 
389 /*
390  * Free memory from hash table previosly allocated via hashinit().
391  */
392 void
393 hashdone(hashtbl, mtype)
394 	void *hashtbl;
395 	struct malloc_type *mtype;
396 {
397 
398 	free(hashtbl, mtype);
399 }
400 
401 
402 static void *
403 hook_establish(list, fn, arg)
404 	hook_list_t *list;
405 	void (*fn)(void *);
406 	void *arg;
407 {
408 	struct hook_desc *hd;
409 
410 	hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
411 	if (hd == NULL)
412 		return (NULL);
413 
414 	hd->hk_fn = fn;
415 	hd->hk_arg = arg;
416 	LIST_INSERT_HEAD(list, hd, hk_list);
417 
418 	return (hd);
419 }
420 
421 static void
422 hook_disestablish(list, vhook)
423 	hook_list_t *list;
424 	void *vhook;
425 {
426 #ifdef DIAGNOSTIC
427 	struct hook_desc *hd;
428 
429 	LIST_FOREACH(hd, list, hk_list) {
430                 if (hd == vhook)
431 			break;
432 	}
433 
434 	if (hd == NULL)
435 		panic("hook_disestablish: hook %p not established", vhook);
436 #endif
437 	LIST_REMOVE((struct hook_desc *)vhook, hk_list);
438 	free(vhook, M_DEVBUF);
439 }
440 
441 static void
442 hook_destroy(list)
443 	hook_list_t *list;
444 {
445 	struct hook_desc *hd;
446 
447 	while ((hd = LIST_FIRST(list)) != NULL) {
448 		LIST_REMOVE(hd, hk_list);
449 		free(hd, M_DEVBUF);
450 	}
451 }
452 
453 static void
454 hook_proc_run(list, p)
455 	hook_list_t *list;
456 	struct proc *p;
457 {
458 	struct hook_desc *hd;
459 
460 	for (hd = LIST_FIRST(list); hd != NULL; hd = LIST_NEXT(hd, hk_list)) {
461 		((void (*)(struct proc *, void *))*hd->hk_fn)(p,
462 		    hd->hk_arg);
463 	}
464 }
465 
466 /*
467  * "Shutdown hook" types, functions, and variables.
468  *
469  * Should be invoked immediately before the
470  * system is halted or rebooted, i.e. after file systems unmounted,
471  * after crash dump done, etc.
472  *
473  * Each shutdown hook is removed from the list before it's run, so that
474  * it won't be run again.
475  */
476 
477 hook_list_t shutdownhook_list;
478 
479 void *
480 shutdownhook_establish(fn, arg)
481 	void (*fn)(void *);
482 	void *arg;
483 {
484 	return hook_establish(&shutdownhook_list, fn, arg);
485 }
486 
487 void
488 shutdownhook_disestablish(vhook)
489 	void *vhook;
490 {
491 	hook_disestablish(&shutdownhook_list, vhook);
492 }
493 
494 /*
495  * Run shutdown hooks.  Should be invoked immediately before the
496  * system is halted or rebooted, i.e. after file systems unmounted,
497  * after crash dump done, etc.
498  *
499  * Each shutdown hook is removed from the list before it's run, so that
500  * it won't be run again.
501  */
502 void
503 doshutdownhooks()
504 {
505 	struct hook_desc *dp;
506 
507 	while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
508 		LIST_REMOVE(dp, hk_list);
509 		(*dp->hk_fn)(dp->hk_arg);
510 #if 0
511 		/*
512 		 * Don't bother freeing the hook structure,, since we may
513 		 * be rebooting because of a memory corruption problem,
514 		 * and this might only make things worse.  It doesn't
515 		 * matter, anyway, since the system is just about to
516 		 * reboot.
517 		 */
518 		free(dp, M_DEVBUF);
519 #endif
520 	}
521 }
522 
523 /*
524  * "Mountroot hook" types, functions, and variables.
525  */
526 
527 hook_list_t mountroothook_list;
528 
529 void *
530 mountroothook_establish(fn, dev)
531 	void (*fn)(struct device *);
532 	struct device *dev;
533 {
534 	return hook_establish(&mountroothook_list, (void (*)(void *))fn, dev);
535 }
536 
537 void
538 mountroothook_disestablish(vhook)
539 	void *vhook;
540 {
541 	hook_disestablish(&mountroothook_list, vhook);
542 }
543 
544 void
545 mountroothook_destroy()
546 {
547 	hook_destroy(&mountroothook_list);
548 }
549 
550 void
551 domountroothook()
552 {
553 	struct hook_desc *hd;
554 
555 	LIST_FOREACH(hd, &mountroothook_list, hk_list) {
556 		if (hd->hk_arg == (void *)root_device) {
557 			(*hd->hk_fn)(hd->hk_arg);
558 			return;
559 		}
560 	}
561 }
562 
563 hook_list_t exechook_list;
564 
565 void *
566 exechook_establish(fn, arg)
567 	void (*fn)(struct proc *, void *);
568 	void *arg;
569 {
570 	return hook_establish(&exechook_list, (void (*)(void *))fn, arg);
571 }
572 
573 void
574 exechook_disestablish(vhook)
575 	void *vhook;
576 {
577 	hook_disestablish(&exechook_list, vhook);
578 }
579 
580 /*
581  * Run exec hooks.
582  */
583 void
584 doexechooks(p)
585 	struct proc *p;
586 {
587 	hook_proc_run(&exechook_list, p);
588 }
589 
590 hook_list_t exithook_list;
591 
592 void *
593 exithook_establish(fn, arg)
594 	void (*fn)(struct proc *, void *);
595 	void *arg;
596 {
597 	return hook_establish(&exithook_list, (void (*)(void *))fn, arg);
598 }
599 
600 void
601 exithook_disestablish(vhook)
602 	void *vhook;
603 {
604 	hook_disestablish(&exithook_list, vhook);
605 }
606 
607 /*
608  * Run exit hooks.
609  */
610 void
611 doexithooks(p)
612 	struct proc *p;
613 {
614 	hook_proc_run(&exithook_list, p);
615 }
616 
617 hook_list_t forkhook_list;
618 
619 void *
620 forkhook_establish(fn)
621 	void (*fn)(struct proc *, struct proc *);
622 {
623 	return hook_establish(&forkhook_list, (void (*)(void *))fn, NULL);
624 }
625 
626 void
627 forkhook_disestablish(vhook)
628 	void *vhook;
629 {
630 	hook_disestablish(&forkhook_list, vhook);
631 }
632 
633 /*
634  * Run fork hooks.
635  */
636 void
637 doforkhooks(p2, p1)
638 	struct proc *p2, *p1;
639 {
640 	struct hook_desc *hd;
641 
642 	LIST_FOREACH(hd, &forkhook_list, hk_list) {
643 		((void (*)(struct proc *, struct proc *))*hd->hk_fn)
644 		    (p2, p1);
645 	}
646 }
647 
648 /*
649  * "Power hook" types, functions, and variables.
650  * The list of power hooks is kept ordered with the last registered hook
651  * first.
652  * When running the hooks on power down the hooks are called in reverse
653  * registration order, when powering up in registration order.
654  */
655 struct powerhook_desc {
656 	CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
657 	void	(*sfd_fn)(int, void *);
658 	void	*sfd_arg;
659 };
660 
661 CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
662 	CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
663 
664 void *
665 powerhook_establish(fn, arg)
666 	void (*fn)(int, void *);
667 	void *arg;
668 {
669 	struct powerhook_desc *ndp;
670 
671 	ndp = (struct powerhook_desc *)
672 	    malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
673 	if (ndp == NULL)
674 		return (NULL);
675 
676 	ndp->sfd_fn = fn;
677 	ndp->sfd_arg = arg;
678 	CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
679 
680 	return (ndp);
681 }
682 
683 void
684 powerhook_disestablish(vhook)
685 	void *vhook;
686 {
687 #ifdef DIAGNOSTIC
688 	struct powerhook_desc *dp;
689 
690 	CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
691                 if (dp == vhook)
692 			goto found;
693 	panic("powerhook_disestablish: hook %p not established", vhook);
694  found:
695 #endif
696 
697 	CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
698 	    sfd_list);
699 	free(vhook, M_DEVBUF);
700 }
701 
702 /*
703  * Run power hooks.
704  */
705 void
706 dopowerhooks(why)
707 	int why;
708 {
709 	struct powerhook_desc *dp;
710 
711 	if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
712 		CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
713 			(*dp->sfd_fn)(why, dp->sfd_arg);
714 		}
715 	} else {
716 		CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
717 			(*dp->sfd_fn)(why, dp->sfd_arg);
718 		}
719 	}
720 }
721 
722 /*
723  * Determine the root device and, if instructed to, the root file system.
724  */
725 
726 #include "md.h"
727 #if NMD == 0
728 #undef MEMORY_DISK_HOOKS
729 #endif
730 
731 #ifdef MEMORY_DISK_HOOKS
732 static struct device fakemdrootdev[NMD];
733 #endif
734 
735 #ifdef MEMORY_DISK_IS_ROOT
736 #define BOOT_FROM_MEMORY_HOOKS 1
737 #endif
738 
739 #include "raid.h"
740 #if NRAID == 1
741 #define BOOT_FROM_RAID_HOOKS 1
742 #endif
743 
744 #ifdef BOOT_FROM_RAID_HOOKS
745 extern int numraid;
746 extern struct device *raidrootdev;
747 #endif
748 
749 void
750 setroot(bootdv, bootpartition)
751 	struct device *bootdv;
752 	int bootpartition;
753 {
754 	struct device *dv;
755 	int len;
756 #ifdef MEMORY_DISK_HOOKS
757 	int i;
758 #endif
759 	dev_t nrootdev;
760 	dev_t ndumpdev = NODEV;
761 	char buf[128];
762 	const char *rootdevname;
763 	const char *dumpdevname;
764 	struct device *rootdv = NULL;		/* XXX gcc -Wuninitialized */
765 	struct device *dumpdv = NULL;
766 	struct ifnet *ifp;
767 	const char *deffsname;
768 	struct vfsops *vops;
769 
770 #ifdef MEMORY_DISK_HOOKS
771 	for (i = 0; i < NMD; i++) {
772 		fakemdrootdev[i].dv_class  = DV_DISK;
773 		fakemdrootdev[i].dv_cfdata = NULL;
774 		fakemdrootdev[i].dv_unit   = i;
775 		fakemdrootdev[i].dv_parent = NULL;
776 		snprintf(fakemdrootdev[i].dv_xname,
777 		    sizeof(fakemdrootdev[i].dv_xname), "md%d", i);
778 	}
779 #endif /* MEMORY_DISK_HOOKS */
780 
781 #ifdef MEMORY_DISK_IS_ROOT
782 	bootdv = &fakemdrootdev[0];
783 	bootpartition = 0;
784 #endif
785 
786 	/*
787 	 * If NFS is specified as the file system, and we found
788 	 * a DV_DISK boot device (or no boot device at all), then
789 	 * find a reasonable network interface for "rootspec".
790 	 */
791 	vops = vfs_getopsbyname("nfs");
792 	if (vops != NULL && vops->vfs_mountroot == mountroot &&
793 	    rootspec == NULL &&
794 	    (bootdv == NULL || bootdv->dv_class != DV_IFNET)) {
795 		TAILQ_FOREACH(ifp, &ifnet, if_list) {
796 			if ((ifp->if_flags &
797 			     (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
798 				break;
799 		}
800 		if (ifp == NULL) {
801 			/*
802 			 * Can't find a suitable interface; ask the
803 			 * user.
804 			 */
805 			boothowto |= RB_ASKNAME;
806 		} else {
807 			/*
808 			 * Have a suitable interface; behave as if
809 			 * the user specified this interface.
810 			 */
811 			rootspec = (const char *)ifp->if_xname;
812 		}
813 	}
814 
815 	/*
816 	 * If wildcarded root and we the boot device wasn't determined,
817 	 * ask the user.
818 	 */
819 	if (rootspec == NULL && bootdv == NULL)
820 		boothowto |= RB_ASKNAME;
821 
822  top:
823 	if (boothowto & RB_ASKNAME) {
824 		struct device *defdumpdv;
825 
826 		for (;;) {
827 			printf("root device");
828 			if (bootdv != NULL) {
829 				printf(" (default %s", bootdv->dv_xname);
830 				if (bootdv->dv_class == DV_DISK)
831 					printf("%c", bootpartition + 'a');
832 				printf(")");
833 			}
834 			printf(": ");
835 			len = cngetsn(buf, sizeof(buf));
836 			if (len == 0 && bootdv != NULL) {
837 				strlcpy(buf, bootdv->dv_xname, sizeof(buf));
838 				len = strlen(buf);
839 			}
840 			if (len > 0 && buf[len - 1] == '*') {
841 				buf[--len] = '\0';
842 				dv = getdisk(buf, len, 1, &nrootdev, 0);
843 				if (dv != NULL) {
844 					rootdv = dv;
845 					break;
846 				}
847 			}
848 			dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
849 			if (dv != NULL) {
850 				rootdv = dv;
851 				break;
852 			}
853 		}
854 
855 		/*
856 		 * Set up the default dump device.  If root is on
857 		 * a network device, there is no default dump
858 		 * device, since we don't support dumps to the
859 		 * network.
860 		 */
861 		if (rootdv->dv_class == DV_IFNET)
862 			defdumpdv = NULL;
863 		else
864 			defdumpdv = rootdv;
865 
866 		for (;;) {
867 			printf("dump device");
868 			if (defdumpdv != NULL) {
869 				/*
870 				 * Note, we know it's a disk if we get here.
871 				 */
872 				printf(" (default %sb)", defdumpdv->dv_xname);
873 			}
874 			printf(": ");
875 			len = cngetsn(buf, sizeof(buf));
876 			if (len == 0) {
877 				if (defdumpdv != NULL) {
878 					ndumpdev = MAKEDISKDEV(major(nrootdev),
879 					    DISKUNIT(nrootdev), 1);
880 				}
881 				dumpdv = defdumpdv;
882 				break;
883 			}
884 			if (len == 4 && strcmp(buf, "none") == 0) {
885 				dumpdv = NULL;
886 				break;
887 			}
888 			dv = getdisk(buf, len, 1, &ndumpdev, 1);
889 			if (dv != NULL) {
890 				dumpdv = dv;
891 				break;
892 			}
893 		}
894 
895 		rootdev = nrootdev;
896 		dumpdev = ndumpdev;
897 
898 		for (vops = LIST_FIRST(&vfs_list); vops != NULL;
899 		     vops = LIST_NEXT(vops, vfs_list)) {
900 			if (vops->vfs_mountroot != NULL &&
901 			    vops->vfs_mountroot == mountroot)
902 			break;
903 		}
904 
905 		if (vops == NULL) {
906 			mountroot = NULL;
907 			deffsname = "generic";
908 		} else
909 			deffsname = vops->vfs_name;
910 
911 		for (;;) {
912 			printf("file system (default %s): ", deffsname);
913 			len = cngetsn(buf, sizeof(buf));
914 			if (len == 0)
915 				break;
916 			if (len == 4 && strcmp(buf, "halt") == 0)
917 				cpu_reboot(RB_HALT, NULL);
918 			else if (len == 6 && strcmp(buf, "reboot") == 0)
919 				cpu_reboot(0, NULL);
920 #if defined(DDB)
921 			else if (len == 3 && strcmp(buf, "ddb") == 0) {
922 				console_debugger();
923 			}
924 #endif
925 			else if (len == 7 && strcmp(buf, "generic") == 0) {
926 				mountroot = NULL;
927 				break;
928 			}
929 			vops = vfs_getopsbyname(buf);
930 			if (vops == NULL || vops->vfs_mountroot == NULL) {
931 				printf("use one of: generic");
932 				for (vops = LIST_FIRST(&vfs_list);
933 				     vops != NULL;
934 				     vops = LIST_NEXT(vops, vfs_list)) {
935 					if (vops->vfs_mountroot != NULL)
936 						printf(" %s", vops->vfs_name);
937 				}
938 #if defined(DDB)
939 				printf(" ddb");
940 #endif
941 				printf(" halt reboot\n");
942 			} else {
943 				mountroot = vops->vfs_mountroot;
944 				break;
945 			}
946 		}
947 
948 	} else if (rootspec == NULL) {
949 		int majdev;
950 
951 		/*
952 		 * Wildcarded root; use the boot device.
953 		 */
954 		rootdv = bootdv;
955 
956 		majdev = devsw_name2blk(bootdv->dv_xname, NULL, 0);
957 		if (majdev >= 0) {
958 			/*
959 			 * Root is on a disk.  `bootpartition' is root.
960 			 */
961 			rootdev = MAKEDISKDEV(majdev, bootdv->dv_unit,
962 			    bootpartition);
963 		}
964 	} else {
965 
966 		/*
967 		 * `root on <dev> ...'
968 		 */
969 
970 		/*
971 		 * If it's a network interface, we can bail out
972 		 * early.
973 		 */
974 		dv = finddevice(rootspec);
975 		if (dv != NULL && dv->dv_class == DV_IFNET) {
976 			rootdv = dv;
977 			goto haveroot;
978 		}
979 
980 		rootdevname = devsw_blk2name(major(rootdev));
981 		if (rootdevname == NULL) {
982 			printf("unknown device major 0x%x\n", rootdev);
983 			boothowto |= RB_ASKNAME;
984 			goto top;
985 		}
986 		memset(buf, 0, sizeof(buf));
987 		snprintf(buf, sizeof(buf), "%s%d", rootdevname,
988 		    DISKUNIT(rootdev));
989 
990 		rootdv = finddevice(buf);
991 		if (rootdv == NULL) {
992 			printf("device %s (0x%x) not configured\n",
993 			    buf, rootdev);
994 			boothowto |= RB_ASKNAME;
995 			goto top;
996 		}
997 	}
998 
999  haveroot:
1000 
1001 	root_device = rootdv;
1002 
1003 	switch (rootdv->dv_class) {
1004 	case DV_IFNET:
1005 		aprint_normal("root on %s", rootdv->dv_xname);
1006 		break;
1007 
1008 	case DV_DISK:
1009 		aprint_normal("root on %s%c", rootdv->dv_xname,
1010 		    DISKPART(rootdev) + 'a');
1011 		break;
1012 
1013 	default:
1014 		printf("can't determine root device\n");
1015 		boothowto |= RB_ASKNAME;
1016 		goto top;
1017 	}
1018 
1019 	/*
1020 	 * Now configure the dump device.
1021 	 *
1022 	 * If we haven't figured out the dump device, do so, with
1023 	 * the following rules:
1024 	 *
1025 	 *	(a) We already know dumpdv in the RB_ASKNAME case.
1026 	 *
1027 	 *	(b) If dumpspec is set, try to use it.  If the device
1028 	 *	    is not available, punt.
1029 	 *
1030 	 *	(c) If dumpspec is not set, the dump device is
1031 	 *	    wildcarded or unspecified.  If the root device
1032 	 *	    is DV_IFNET, punt.  Otherwise, use partition b
1033 	 *	    of the root device.
1034 	 */
1035 
1036 	if (boothowto & RB_ASKNAME) {		/* (a) */
1037 		if (dumpdv == NULL)
1038 			goto nodumpdev;
1039 	} else if (dumpspec != NULL) {		/* (b) */
1040 		if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
1041 			/*
1042 			 * Operator doesn't want a dump device.
1043 			 * Or looks like they tried to pick a network
1044 			 * device.  Oops.
1045 			 */
1046 			goto nodumpdev;
1047 		}
1048 
1049 		dumpdevname = devsw_blk2name(major(dumpdev));
1050 		if (dumpdevname == NULL)
1051 			goto nodumpdev;
1052 		memset(buf, 0, sizeof(buf));
1053 		snprintf(buf, sizeof(buf), "%s%d", dumpdevname,
1054 		    DISKUNIT(dumpdev));
1055 
1056 		dumpdv = finddevice(buf);
1057 		if (dumpdv == NULL) {
1058 			/*
1059 			 * Device not configured.
1060 			 */
1061 			goto nodumpdev;
1062 		}
1063 	} else {				/* (c) */
1064 		if (rootdv->dv_class == DV_IFNET)
1065 			goto nodumpdev;
1066 		else {
1067 			dumpdv = rootdv;
1068 			dumpdev = MAKEDISKDEV(major(rootdev),
1069 			    dumpdv->dv_unit, 1);
1070 		}
1071 	}
1072 
1073 	aprint_normal(" dumps on %s%c\n", dumpdv->dv_xname,
1074 	    DISKPART(dumpdev) + 'a');
1075 	return;
1076 
1077  nodumpdev:
1078 	dumpdev = NODEV;
1079 	aprint_normal("\n");
1080 }
1081 
1082 static struct device *
1083 finddevice(name)
1084 	const char *name;
1085 {
1086 	struct device *dv;
1087 #if defined(BOOT_FROM_RAID_HOOKS) || defined(BOOT_FROM_MEMORY_HOOKS)
1088 	int j;
1089 #endif /* BOOT_FROM_RAID_HOOKS || BOOT_FROM_MEMORY_HOOKS */
1090 
1091 #ifdef BOOT_FROM_RAID_HOOKS
1092 	for (j = 0; j < numraid; j++) {
1093 		if (strcmp(name, raidrootdev[j].dv_xname) == 0) {
1094 			dv = &raidrootdev[j];
1095 			return (dv);
1096 		}
1097 	}
1098 #endif /* BOOT_FROM_RAID_HOOKS */
1099 
1100 #ifdef BOOT_FROM_MEMORY_HOOKS
1101 	for (j = 0; j < NMD; j++) {
1102 		if (strcmp(name, fakemdrootdev[j].dv_xname) == 0) {
1103 			dv = &fakemdrootdev[j];
1104 			return (dv);
1105 		}
1106 	}
1107 #endif /* BOOT_FROM_MEMORY_HOOKS */
1108 
1109 	for (dv = TAILQ_FIRST(&alldevs); dv != NULL;
1110 	    dv = TAILQ_NEXT(dv, dv_list))
1111 		if (strcmp(dv->dv_xname, name) == 0)
1112 			break;
1113 	return (dv);
1114 }
1115 
1116 static struct device *
1117 getdisk(str, len, defpart, devp, isdump)
1118 	char *str;
1119 	int len, defpart;
1120 	dev_t *devp;
1121 	int isdump;
1122 {
1123 	struct device	*dv;
1124 #ifdef MEMORY_DISK_HOOKS
1125 	int		i;
1126 #endif
1127 #ifdef BOOT_FROM_RAID_HOOKS
1128 	int 		j;
1129 #endif
1130 
1131 	if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
1132 		printf("use one of:");
1133 #ifdef MEMORY_DISK_HOOKS
1134 		if (isdump == 0)
1135 			for (i = 0; i < NMD; i++)
1136 				printf(" %s[a-%c]", fakemdrootdev[i].dv_xname,
1137 				    'a' + MAXPARTITIONS - 1);
1138 #endif
1139 #ifdef BOOT_FROM_RAID_HOOKS
1140 		if (isdump == 0)
1141 			for (j = 0; j < numraid; j++)
1142 				printf(" %s[a-%c]", raidrootdev[j].dv_xname,
1143 				    'a' + MAXPARTITIONS - 1);
1144 #endif
1145 		TAILQ_FOREACH(dv, &alldevs, dv_list) {
1146 			if (dv->dv_class == DV_DISK)
1147 				printf(" %s[a-%c]", dv->dv_xname,
1148 				    'a' + MAXPARTITIONS - 1);
1149 			if (isdump == 0 && dv->dv_class == DV_IFNET)
1150 				printf(" %s", dv->dv_xname);
1151 		}
1152 		if (isdump)
1153 			printf(" none");
1154 #if defined(DDB)
1155 		printf(" ddb");
1156 #endif
1157 		printf(" halt reboot\n");
1158 	}
1159 	return (dv);
1160 }
1161 
1162 static struct device *
1163 parsedisk(str, len, defpart, devp)
1164 	char *str;
1165 	int len, defpart;
1166 	dev_t *devp;
1167 {
1168 	struct device *dv;
1169 	char *cp, c;
1170 	int majdev, part;
1171 #ifdef MEMORY_DISK_HOOKS
1172 	int i;
1173 #endif
1174 	if (len == 0)
1175 		return (NULL);
1176 
1177 	if (len == 4 && strcmp(str, "halt") == 0)
1178 		cpu_reboot(RB_HALT, NULL);
1179 	else if (len == 6 && strcmp(str, "reboot") == 0)
1180 		cpu_reboot(0, NULL);
1181 #if defined(DDB)
1182 	else if (len == 3 && strcmp(str, "ddb") == 0)
1183 		console_debugger();
1184 #endif
1185 
1186 	cp = str + len - 1;
1187 	c = *cp;
1188 	if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
1189 		part = c - 'a';
1190 		*cp = '\0';
1191 	} else
1192 		part = defpart;
1193 
1194 #ifdef MEMORY_DISK_HOOKS
1195 	for (i = 0; i < NMD; i++)
1196 		if (strcmp(str, fakemdrootdev[i].dv_xname) == 0) {
1197 			dv = &fakemdrootdev[i];
1198 			goto gotdisk;
1199 		}
1200 #endif
1201 
1202 	dv = finddevice(str);
1203 	if (dv != NULL) {
1204 		if (dv->dv_class == DV_DISK) {
1205 #ifdef MEMORY_DISK_HOOKS
1206  gotdisk:
1207 #endif
1208 			majdev = devsw_name2blk(dv->dv_xname, NULL, 0);
1209 			if (majdev < 0)
1210 				panic("parsedisk");
1211 			*devp = MAKEDISKDEV(majdev, dv->dv_unit, part);
1212 		}
1213 
1214 		if (dv->dv_class == DV_IFNET)
1215 			*devp = NODEV;
1216 	}
1217 
1218 	*cp = c;
1219 	return (dv);
1220 }
1221 
1222 /*
1223  * snprintf() `bytes' into `buf', reformatting it so that the number,
1224  * plus a possible `x' + suffix extension) fits into len bytes (including
1225  * the terminating NUL).
1226  * Returns the number of bytes stored in buf, or -1 if there was a problem.
1227  * E.g, given a len of 9 and a suffix of `B':
1228  *	bytes		result
1229  *	-----		------
1230  *	99999		`99999 B'
1231  *	100000		`97 kB'
1232  *	66715648	`65152 kB'
1233  *	252215296	`240 MB'
1234  */
1235 int
1236 humanize_number(buf, len, bytes, suffix, divisor)
1237 	char		*buf;
1238 	size_t		 len;
1239 	u_int64_t	 bytes;
1240 	const char	*suffix;
1241 	int 		divisor;
1242 {
1243        	/* prefixes are: (none), kilo, Mega, Giga, Tera, Peta, Exa */
1244 	const char *prefixes;
1245 	int		r;
1246 	u_int64_t	max;
1247 	size_t		i, suffixlen;
1248 
1249 	if (buf == NULL || suffix == NULL)
1250 		return (-1);
1251 	if (len > 0)
1252 		buf[0] = '\0';
1253 	suffixlen = strlen(suffix);
1254 	/* check if enough room for `x y' + suffix + `\0' */
1255 	if (len < 4 + suffixlen)
1256 		return (-1);
1257 
1258 	if (divisor == 1024) {
1259 		/*
1260 		 * binary multiplies
1261 		 * XXX IEC 60027-2 recommends Ki, Mi, Gi...
1262 		 */
1263 		prefixes = " KMGTPE";
1264 	} else
1265 		prefixes = " kMGTPE"; /* SI for decimal multiplies */
1266 
1267 	max = 1;
1268 	for (i = 0; i < len - suffixlen - 3; i++)
1269 		max *= 10;
1270 	for (i = 0; bytes >= max && prefixes[i + 1]; i++)
1271 		bytes /= divisor;
1272 
1273 	r = snprintf(buf, len, "%qu%s%c%s", (unsigned long long)bytes,
1274 	    i == 0 ? "" : " ", prefixes[i], suffix);
1275 
1276 	return (r);
1277 }
1278 
1279 int
1280 format_bytes(buf, len, bytes)
1281 	char		*buf;
1282 	size_t		 len;
1283 	u_int64_t	 bytes;
1284 {
1285 	int	rv;
1286 	size_t	nlen;
1287 
1288 	rv = humanize_number(buf, len, bytes, "B", 1024);
1289 	if (rv != -1) {
1290 			/* nuke the trailing ` B' if it exists */
1291 		nlen = strlen(buf) - 2;
1292 		if (strcmp(&buf[nlen], " B") == 0)
1293 			buf[nlen] = '\0';
1294 	}
1295 	return (rv);
1296 }
1297 
1298 /*
1299  * Start trace of particular system call. If process is being traced,
1300  * this routine is called by MD syscall dispatch code just before
1301  * a system call is actually executed.
1302  * MD caller guarantees the passed 'code' is within the supported
1303  * system call number range for emulation the process runs under.
1304  */
1305 int
1306 trace_enter(struct lwp *l, register_t code,
1307 	register_t realcode, const struct sysent *callp, void *args)
1308 {
1309 #if defined(KTRACE) || defined(SYSTRACE)
1310 	struct proc *p = l->l_proc;
1311 #endif
1312 
1313 #ifdef SYSCALL_DEBUG
1314 	scdebug_call(l, code, args);
1315 #endif /* SYSCALL_DEBUG */
1316 
1317 #ifdef KTRACE
1318 	if (KTRPOINT(p, KTR_SYSCALL))
1319 		ktrsyscall(p, code, realcode, callp, args);
1320 #endif /* KTRACE */
1321 
1322 #ifdef SYSTRACE
1323 	if (ISSET(p->p_flag, P_SYSTRACE))
1324 		return systrace_enter(p, code, args);
1325 #endif
1326 	return 0;
1327 }
1328 
1329 /*
1330  * End trace of particular system call. If process is being traced,
1331  * this routine is called by MD syscall dispatch code just after
1332  * a system call finishes.
1333  * MD caller guarantees the passed 'code' is within the supported
1334  * system call number range for emulation the process runs under.
1335  */
1336 void
1337 trace_exit(struct lwp *l, register_t code, void *args, register_t rval[],
1338     int error)
1339 {
1340 #if defined(KTRACE) || defined(SYSTRACE)
1341 	struct proc *p = l->l_proc;
1342 #endif
1343 
1344 #ifdef SYSCALL_DEBUG
1345 	scdebug_ret(l, code, error, rval);
1346 #endif /* SYSCALL_DEBUG */
1347 
1348 #ifdef KTRACE
1349 	if (KTRPOINT(p, KTR_SYSRET)) {
1350 		KERNEL_PROC_LOCK(l);
1351 		ktrsysret(p, code, error, rval);
1352 		KERNEL_PROC_UNLOCK(l);
1353 	}
1354 #endif /* KTRACE */
1355 
1356 #ifdef SYSTRACE
1357 	if (ISSET(p->p_flag, P_SYSTRACE))
1358 		systrace_exit(p, code, args, rval, error);
1359 #endif
1360 }
1361