xref: /netbsd-src/sys/kern/kern_subr.c (revision da5f4674a3fc214be3572d358b66af40ab9401e7)
1 /*	$NetBSD: kern_subr.c,v 1.103 2003/08/07 16:31:49 agc 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 1.103 2003/08/07 16:31:49 agc 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 __P((const char *));
118 static struct device *getdisk __P((char *, int, int, dev_t *, int));
119 static struct device *parsedisk __P((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) __P((void *));
127 	void	*hk_arg;
128 };
129 typedef LIST_HEAD(, hook_desc) hook_list_t;
130 
131 static void *hook_establish __P((hook_list_t *, void (*)(void *), void *));
132 static void hook_disestablish __P((hook_list_t *, void *));
133 static void hook_destroy __P((hook_list_t *));
134 static void hook_proc_run __P((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 			uio->uio_iov++;
159 			uio->uio_iovcnt--;
160 			continue;
161 		}
162 		if (cnt > n)
163 			cnt = n;
164 		switch (uio->uio_segflg) {
165 
166 		case UIO_USERSPACE:
167 			if (curlwp->l_cpu->ci_schedstate.spc_flags &
168 			    SPCF_SHOULDYIELD)
169 				preempt(1);
170 			if (__predict_true(p == curproc)) {
171 				if (uio->uio_rw == UIO_READ)
172 					error = copyout(cp, iov->iov_base, cnt);
173 				else
174 					error = copyin(iov->iov_base, cp, cnt);
175 			} else {
176 				if (uio->uio_rw == UIO_READ)
177 					error = copyout_proc(p, cp,
178 					    iov->iov_base, cnt);
179 				else
180 					error = copyin_proc(p, iov->iov_base,
181 					    cp, cnt);
182 			}
183 			if (error)
184 				return (error);
185 			break;
186 
187 		case UIO_SYSSPACE:
188 			if (uio->uio_rw == UIO_READ)
189 				error = kcopy(cp, iov->iov_base, cnt);
190 			else
191 				error = kcopy(iov->iov_base, cp, cnt);
192 			if (error)
193 				return (error);
194 			break;
195 		}
196 		iov->iov_base = (caddr_t)iov->iov_base + cnt;
197 		iov->iov_len -= cnt;
198 		uio->uio_resid -= cnt;
199 		uio->uio_offset += cnt;
200 		cp += cnt;
201 		KDASSERT(cnt <= n);
202 		n -= cnt;
203 	}
204 	return (error);
205 }
206 
207 /*
208  * Give next character to user as result of read.
209  */
210 int
211 ureadc(c, uio)
212 	int c;
213 	struct uio *uio;
214 {
215 	struct iovec *iov;
216 
217 	if (uio->uio_resid <= 0)
218 		panic("ureadc: non-positive resid");
219 again:
220 	if (uio->uio_iovcnt <= 0)
221 		panic("ureadc: non-positive iovcnt");
222 	iov = uio->uio_iov;
223 	if (iov->iov_len <= 0) {
224 		uio->uio_iovcnt--;
225 		uio->uio_iov++;
226 		goto again;
227 	}
228 	switch (uio->uio_segflg) {
229 
230 	case UIO_USERSPACE:
231 		if (subyte(iov->iov_base, c) < 0)
232 			return (EFAULT);
233 		break;
234 
235 	case UIO_SYSSPACE:
236 		*(char *)iov->iov_base = c;
237 		break;
238 	}
239 	iov->iov_base = (caddr_t)iov->iov_base + 1;
240 	iov->iov_len--;
241 	uio->uio_resid--;
242 	uio->uio_offset++;
243 	return (0);
244 }
245 
246 /*
247  * Like copyin(), but operates on an arbitrary process.
248  */
249 int
250 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
251 {
252 	struct iovec iov;
253 	struct uio uio;
254 	int error;
255 
256 	if (len == 0)
257 		return (0);
258 
259 	iov.iov_base = kaddr;
260 	iov.iov_len = len;
261 	uio.uio_iov = &iov;
262 	uio.uio_iovcnt = 1;
263 	uio.uio_offset = (off_t)(intptr_t)uaddr;
264 	uio.uio_resid = len;
265 	uio.uio_segflg = UIO_SYSSPACE;
266 	uio.uio_rw = UIO_READ;
267 	uio.uio_procp = NULL;
268 
269 	/* XXXCDC: how should locking work here? */
270 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
271 		return (EFAULT);
272 	p->p_vmspace->vm_refcnt++;	/* XXX */
273 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
274 	uvmspace_free(p->p_vmspace);
275 
276 	return (error);
277 }
278 
279 /*
280  * Like copyout(), but operates on an arbitrary process.
281  */
282 int
283 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
284 {
285 	struct iovec iov;
286 	struct uio uio;
287 	int error;
288 
289 	if (len == 0)
290 		return (0);
291 
292 	iov.iov_base = (void *) kaddr;	/* XXX cast away const */
293 	iov.iov_len = len;
294 	uio.uio_iov = &iov;
295 	uio.uio_iovcnt = 1;
296 	uio.uio_offset = (off_t)(intptr_t)uaddr;
297 	uio.uio_resid = len;
298 	uio.uio_segflg = UIO_SYSSPACE;
299 	uio.uio_rw = UIO_WRITE;
300 	uio.uio_procp = NULL;
301 
302 	/* XXXCDC: how should locking work here? */
303 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
304 		return (EFAULT);
305 	p->p_vmspace->vm_refcnt++;	/* XXX */
306 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
307 	uvmspace_free(p->p_vmspace);
308 
309 	return (error);
310 }
311 
312 /*
313  * General routine to allocate a hash table.
314  * Allocate enough memory to hold at least `elements' list-head pointers.
315  * Return a pointer to the allocated space and set *hashmask to a pattern
316  * suitable for masking a value to use as an index into the returned array.
317  */
318 void *
319 hashinit(elements, htype, mtype, mflags, hashmask)
320 	u_int elements;
321 	enum hashtype htype;
322 	struct malloc_type *mtype;
323 	int mflags;
324 	u_long *hashmask;
325 {
326 	u_long hashsize, i;
327 	LIST_HEAD(, generic) *hashtbl_list;
328 	TAILQ_HEAD(, generic) *hashtbl_tailq;
329 	size_t esize;
330 	void *p;
331 
332 	if (elements == 0)
333 		panic("hashinit: bad cnt");
334 	for (hashsize = 1; hashsize < elements; hashsize <<= 1)
335 		continue;
336 
337 	switch (htype) {
338 	case HASH_LIST:
339 		esize = sizeof(*hashtbl_list);
340 		break;
341 	case HASH_TAILQ:
342 		esize = sizeof(*hashtbl_tailq);
343 		break;
344 #ifdef DIAGNOSTIC
345 	default:
346 		panic("hashinit: invalid table type");
347 #endif
348 	}
349 
350 	if ((p = malloc(hashsize * esize, mtype, mflags)) == NULL)
351 		return (NULL);
352 
353 	switch (htype) {
354 	case HASH_LIST:
355 		hashtbl_list = p;
356 		for (i = 0; i < hashsize; i++)
357 			LIST_INIT(&hashtbl_list[i]);
358 		break;
359 	case HASH_TAILQ:
360 		hashtbl_tailq = p;
361 		for (i = 0; i < hashsize; i++)
362 			TAILQ_INIT(&hashtbl_tailq[i]);
363 		break;
364 	}
365 	*hashmask = hashsize - 1;
366 	return (p);
367 }
368 
369 /*
370  * Free memory from hash table previosly allocated via hashinit().
371  */
372 void
373 hashdone(hashtbl, mtype)
374 	void *hashtbl;
375 	struct malloc_type *mtype;
376 {
377 
378 	free(hashtbl, mtype);
379 }
380 
381 
382 static void *
383 hook_establish(list, fn, arg)
384 	hook_list_t *list;
385 	void (*fn) __P((void *));
386 	void *arg;
387 {
388 	struct hook_desc *hd;
389 
390 	hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
391 	if (hd == NULL)
392 		return (NULL);
393 
394 	hd->hk_fn = fn;
395 	hd->hk_arg = arg;
396 	LIST_INSERT_HEAD(list, hd, hk_list);
397 
398 	return (hd);
399 }
400 
401 static void
402 hook_disestablish(list, vhook)
403 	hook_list_t *list;
404 	void *vhook;
405 {
406 #ifdef DIAGNOSTIC
407 	struct hook_desc *hd;
408 
409 	LIST_FOREACH(hd, list, hk_list) {
410                 if (hd == vhook)
411 			break;
412 	}
413 
414 	if (hd == NULL)
415 		panic("hook_disestablish: hook %p not established", vhook);
416 #endif
417 	LIST_REMOVE((struct hook_desc *)vhook, hk_list);
418 	free(vhook, M_DEVBUF);
419 }
420 
421 static void
422 hook_destroy(list)
423 	hook_list_t *list;
424 {
425 	struct hook_desc *hd;
426 
427 	while ((hd = LIST_FIRST(list)) != NULL) {
428 		LIST_REMOVE(hd, hk_list);
429 		free(hd, M_DEVBUF);
430 	}
431 }
432 
433 static void
434 hook_proc_run(list, p)
435 	hook_list_t *list;
436 	struct proc *p;
437 {
438 	struct hook_desc *hd;
439 
440 	for (hd = LIST_FIRST(list); hd != NULL; hd = LIST_NEXT(hd, hk_list)) {
441 		((void (*) __P((struct proc *, void *)))*hd->hk_fn)(p,
442 		    hd->hk_arg);
443 	}
444 }
445 
446 /*
447  * "Shutdown hook" types, functions, and variables.
448  *
449  * Should be invoked immediately before the
450  * system is halted or rebooted, i.e. after file systems unmounted,
451  * after crash dump done, etc.
452  *
453  * Each shutdown hook is removed from the list before it's run, so that
454  * it won't be run again.
455  */
456 
457 hook_list_t shutdownhook_list;
458 
459 void *
460 shutdownhook_establish(fn, arg)
461 	void (*fn) __P((void *));
462 	void *arg;
463 {
464 	return hook_establish(&shutdownhook_list, fn, arg);
465 }
466 
467 void
468 shutdownhook_disestablish(vhook)
469 	void *vhook;
470 {
471 	hook_disestablish(&shutdownhook_list, vhook);
472 }
473 
474 /*
475  * Run shutdown hooks.  Should be invoked immediately before the
476  * system is halted or rebooted, i.e. after file systems unmounted,
477  * after crash dump done, etc.
478  *
479  * Each shutdown hook is removed from the list before it's run, so that
480  * it won't be run again.
481  */
482 void
483 doshutdownhooks()
484 {
485 	struct hook_desc *dp;
486 
487 	while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
488 		LIST_REMOVE(dp, hk_list);
489 		(*dp->hk_fn)(dp->hk_arg);
490 #if 0
491 		/*
492 		 * Don't bother freeing the hook structure,, since we may
493 		 * be rebooting because of a memory corruption problem,
494 		 * and this might only make things worse.  It doesn't
495 		 * matter, anyway, since the system is just about to
496 		 * reboot.
497 		 */
498 		free(dp, M_DEVBUF);
499 #endif
500 	}
501 }
502 
503 /*
504  * "Mountroot hook" types, functions, and variables.
505  */
506 
507 hook_list_t mountroothook_list;
508 
509 void *
510 mountroothook_establish(fn, dev)
511 	void (*fn) __P((struct device *));
512 	struct device *dev;
513 {
514 	return hook_establish(&mountroothook_list, (void (*)__P((void *)))fn,
515 	    dev);
516 }
517 
518 void
519 mountroothook_disestablish(vhook)
520 	void *vhook;
521 {
522 	hook_disestablish(&mountroothook_list, vhook);
523 }
524 
525 void
526 mountroothook_destroy()
527 {
528 	hook_destroy(&mountroothook_list);
529 }
530 
531 void
532 domountroothook()
533 {
534 	struct hook_desc *hd;
535 
536 	LIST_FOREACH(hd, &mountroothook_list, hk_list) {
537 		if (hd->hk_arg == (void *)root_device) {
538 			(*hd->hk_fn)(hd->hk_arg);
539 			return;
540 		}
541 	}
542 }
543 
544 hook_list_t exechook_list;
545 
546 void *
547 exechook_establish(fn, arg)
548 	void (*fn) __P((struct proc *, void *));
549 	void *arg;
550 {
551 	return hook_establish(&exechook_list, (void (*) __P((void *)))fn, arg);
552 }
553 
554 void
555 exechook_disestablish(vhook)
556 	void *vhook;
557 {
558 	hook_disestablish(&exechook_list, vhook);
559 }
560 
561 /*
562  * Run exec hooks.
563  */
564 void
565 doexechooks(p)
566 	struct proc *p;
567 {
568 	hook_proc_run(&exechook_list, p);
569 }
570 
571 hook_list_t exithook_list;
572 
573 void *
574 exithook_establish(fn, arg)
575 	void (*fn) __P((struct proc *, void *));
576 	void *arg;
577 {
578 	return hook_establish(&exithook_list, (void (*) __P((void *)))fn, arg);
579 }
580 
581 void
582 exithook_disestablish(vhook)
583 	void *vhook;
584 {
585 	hook_disestablish(&exithook_list, vhook);
586 }
587 
588 /*
589  * Run exit hooks.
590  */
591 void
592 doexithooks(p)
593 	struct proc *p;
594 {
595 	hook_proc_run(&exithook_list, p);
596 }
597 
598 hook_list_t forkhook_list;
599 
600 void *
601 forkhook_establish(fn)
602 	void (*fn) __P((struct proc *, struct proc *));
603 {
604 	return hook_establish(&forkhook_list, (void (*) __P((void *)))fn, NULL);
605 }
606 
607 void
608 forkhook_disestablish(vhook)
609 	void *vhook;
610 {
611 	hook_disestablish(&forkhook_list, vhook);
612 }
613 
614 /*
615  * Run fork hooks.
616  */
617 void
618 doforkhooks(p2, p1)
619 	struct proc *p2, *p1;
620 {
621 	struct hook_desc *hd;
622 
623 	LIST_FOREACH(hd, &forkhook_list, hk_list) {
624 		((void (*) __P((struct proc *, struct proc *)))*hd->hk_fn)
625 		    (p2, p1);
626 	}
627 }
628 
629 /*
630  * "Power hook" types, functions, and variables.
631  * The list of power hooks is kept ordered with the last registered hook
632  * first.
633  * When running the hooks on power down the hooks are called in reverse
634  * registration order, when powering up in registration order.
635  */
636 struct powerhook_desc {
637 	CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
638 	void	(*sfd_fn) __P((int, void *));
639 	void	*sfd_arg;
640 };
641 
642 CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
643 	CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
644 
645 void *
646 powerhook_establish(fn, arg)
647 	void (*fn) __P((int, void *));
648 	void *arg;
649 {
650 	struct powerhook_desc *ndp;
651 
652 	ndp = (struct powerhook_desc *)
653 	    malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
654 	if (ndp == NULL)
655 		return (NULL);
656 
657 	ndp->sfd_fn = fn;
658 	ndp->sfd_arg = arg;
659 	CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
660 
661 	return (ndp);
662 }
663 
664 void
665 powerhook_disestablish(vhook)
666 	void *vhook;
667 {
668 #ifdef DIAGNOSTIC
669 	struct powerhook_desc *dp;
670 
671 	CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
672                 if (dp == vhook)
673 			goto found;
674 	panic("powerhook_disestablish: hook %p not established", vhook);
675  found:
676 #endif
677 
678 	CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
679 	    sfd_list);
680 	free(vhook, M_DEVBUF);
681 }
682 
683 /*
684  * Run power hooks.
685  */
686 void
687 dopowerhooks(why)
688 	int why;
689 {
690 	struct powerhook_desc *dp;
691 
692 	if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
693 		CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
694 			(*dp->sfd_fn)(why, dp->sfd_arg);
695 		}
696 	} else {
697 		CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
698 			(*dp->sfd_fn)(why, dp->sfd_arg);
699 		}
700 	}
701 }
702 
703 /*
704  * Determine the root device and, if instructed to, the root file system.
705  */
706 
707 #include "md.h"
708 #if NMD == 0
709 #undef MEMORY_DISK_HOOKS
710 #endif
711 
712 #ifdef MEMORY_DISK_HOOKS
713 static struct device fakemdrootdev[NMD];
714 #endif
715 
716 #include "raid.h"
717 #if NRAID == 1
718 #define BOOT_FROM_RAID_HOOKS 1
719 #endif
720 
721 #ifdef BOOT_FROM_RAID_HOOKS
722 extern int numraid;
723 extern struct device *raidrootdev;
724 #endif
725 
726 void
727 setroot(bootdv, bootpartition)
728 	struct device *bootdv;
729 	int bootpartition;
730 {
731 	struct device *dv;
732 	int len;
733 #ifdef MEMORY_DISK_HOOKS
734 	int i;
735 #endif
736 	dev_t nrootdev;
737 	dev_t ndumpdev = NODEV;
738 	char buf[128];
739 	const char *rootdevname;
740 	const char *dumpdevname;
741 	struct device *rootdv = NULL;		/* XXX gcc -Wuninitialized */
742 	struct device *dumpdv = NULL;
743 	struct ifnet *ifp;
744 	const char *deffsname;
745 	struct vfsops *vops;
746 
747 #ifdef MEMORY_DISK_HOOKS
748 	for (i = 0; i < NMD; i++) {
749 		fakemdrootdev[i].dv_class  = DV_DISK;
750 		fakemdrootdev[i].dv_cfdata = NULL;
751 		fakemdrootdev[i].dv_unit   = i;
752 		fakemdrootdev[i].dv_parent = NULL;
753 		sprintf(fakemdrootdev[i].dv_xname, "md%d", i);
754 	}
755 #endif /* MEMORY_DISK_HOOKS */
756 
757 #ifdef MEMORY_DISK_IS_ROOT
758 	bootdv = &fakemdrootdev[0];
759 	bootpartition = 0;
760 #endif
761 
762 	/*
763 	 * If NFS is specified as the file system, and we found
764 	 * a DV_DISK boot device (or no boot device at all), then
765 	 * find a reasonable network interface for "rootspec".
766 	 */
767 	vops = vfs_getopsbyname("nfs");
768 	if (vops != NULL && vops->vfs_mountroot == mountroot &&
769 	    rootspec == NULL &&
770 	    (bootdv == NULL || bootdv->dv_class != DV_IFNET)) {
771 		TAILQ_FOREACH(ifp, &ifnet, if_list) {
772 			if ((ifp->if_flags &
773 			     (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
774 				break;
775 		}
776 		if (ifp == NULL) {
777 			/*
778 			 * Can't find a suitable interface; ask the
779 			 * user.
780 			 */
781 			boothowto |= RB_ASKNAME;
782 		} else {
783 			/*
784 			 * Have a suitable interface; behave as if
785 			 * the user specified this interface.
786 			 */
787 			rootspec = (const char *)ifp->if_xname;
788 		}
789 	}
790 
791 	/*
792 	 * If wildcarded root and we the boot device wasn't determined,
793 	 * ask the user.
794 	 */
795 	if (rootspec == NULL && bootdv == NULL)
796 		boothowto |= RB_ASKNAME;
797 
798  top:
799 	if (boothowto & RB_ASKNAME) {
800 		struct device *defdumpdv;
801 
802 		for (;;) {
803 			printf("root device");
804 			if (bootdv != NULL) {
805 				printf(" (default %s", bootdv->dv_xname);
806 				if (bootdv->dv_class == DV_DISK)
807 					printf("%c", bootpartition + 'a');
808 				printf(")");
809 			}
810 			printf(": ");
811 			len = cngetsn(buf, sizeof(buf));
812 			if (len == 0 && bootdv != NULL) {
813 				strlcpy(buf, bootdv->dv_xname, sizeof(buf));
814 				len = strlen(buf);
815 			}
816 			if (len > 0 && buf[len - 1] == '*') {
817 				buf[--len] = '\0';
818 				dv = getdisk(buf, len, 1, &nrootdev, 0);
819 				if (dv != NULL) {
820 					rootdv = dv;
821 					break;
822 				}
823 			}
824 			dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
825 			if (dv != NULL) {
826 				rootdv = dv;
827 				break;
828 			}
829 		}
830 
831 		/*
832 		 * Set up the default dump device.  If root is on
833 		 * a network device, there is no default dump
834 		 * device, since we don't support dumps to the
835 		 * network.
836 		 */
837 		if (rootdv->dv_class == DV_IFNET)
838 			defdumpdv = NULL;
839 		else
840 			defdumpdv = rootdv;
841 
842 		for (;;) {
843 			printf("dump device");
844 			if (defdumpdv != NULL) {
845 				/*
846 				 * Note, we know it's a disk if we get here.
847 				 */
848 				printf(" (default %sb)", defdumpdv->dv_xname);
849 			}
850 			printf(": ");
851 			len = cngetsn(buf, sizeof(buf));
852 			if (len == 0) {
853 				if (defdumpdv != NULL) {
854 					ndumpdev = MAKEDISKDEV(major(nrootdev),
855 					    DISKUNIT(nrootdev), 1);
856 				}
857 				dumpdv = defdumpdv;
858 				break;
859 			}
860 			if (len == 4 && strcmp(buf, "none") == 0) {
861 				dumpdv = NULL;
862 				break;
863 			}
864 			dv = getdisk(buf, len, 1, &ndumpdev, 1);
865 			if (dv != NULL) {
866 				dumpdv = dv;
867 				break;
868 			}
869 		}
870 
871 		rootdev = nrootdev;
872 		dumpdev = ndumpdev;
873 
874 		for (vops = LIST_FIRST(&vfs_list); vops != NULL;
875 		     vops = LIST_NEXT(vops, vfs_list)) {
876 			if (vops->vfs_mountroot != NULL &&
877 			    vops->vfs_mountroot == mountroot)
878 			break;
879 		}
880 
881 		if (vops == NULL) {
882 			mountroot = NULL;
883 			deffsname = "generic";
884 		} else
885 			deffsname = vops->vfs_name;
886 
887 		for (;;) {
888 			printf("file system (default %s): ", deffsname);
889 			len = cngetsn(buf, sizeof(buf));
890 			if (len == 0)
891 				break;
892 			if (len == 4 && strcmp(buf, "halt") == 0)
893 				cpu_reboot(RB_HALT, NULL);
894 			else if (len == 6 && strcmp(buf, "reboot") == 0)
895 				cpu_reboot(0, NULL);
896 #if defined(DDB)
897 			else if (len == 3 && strcmp(buf, "ddb") == 0) {
898 				console_debugger();
899 			}
900 #endif
901 			else if (len == 7 && strcmp(buf, "generic") == 0) {
902 				mountroot = NULL;
903 				break;
904 			}
905 			vops = vfs_getopsbyname(buf);
906 			if (vops == NULL || vops->vfs_mountroot == NULL) {
907 				printf("use one of: generic");
908 				for (vops = LIST_FIRST(&vfs_list);
909 				     vops != NULL;
910 				     vops = LIST_NEXT(vops, vfs_list)) {
911 					if (vops->vfs_mountroot != NULL)
912 						printf(" %s", vops->vfs_name);
913 				}
914 #if defined(DDB)
915 				printf(" ddb");
916 #endif
917 				printf(" halt reboot\n");
918 			} else {
919 				mountroot = vops->vfs_mountroot;
920 				break;
921 			}
922 		}
923 
924 	} else if (rootspec == NULL) {
925 		int majdev;
926 
927 		/*
928 		 * Wildcarded root; use the boot device.
929 		 */
930 		rootdv = bootdv;
931 
932 		majdev = devsw_name2blk(bootdv->dv_xname, NULL, 0);
933 		if (majdev >= 0) {
934 			/*
935 			 * Root is on a disk.  `bootpartition' is root.
936 			 */
937 			rootdev = MAKEDISKDEV(majdev, bootdv->dv_unit,
938 			    bootpartition);
939 		}
940 	} else {
941 
942 		/*
943 		 * `root on <dev> ...'
944 		 */
945 
946 		/*
947 		 * If it's a network interface, we can bail out
948 		 * early.
949 		 */
950 		dv = finddevice(rootspec);
951 		if (dv != NULL && dv->dv_class == DV_IFNET) {
952 			rootdv = dv;
953 			goto haveroot;
954 		}
955 
956 		rootdevname = devsw_blk2name(major(rootdev));
957 		if (rootdevname == NULL) {
958 			printf("unknown device major 0x%x\n", rootdev);
959 			boothowto |= RB_ASKNAME;
960 			goto top;
961 		}
962 		memset(buf, 0, sizeof(buf));
963 		sprintf(buf, "%s%d", rootdevname, DISKUNIT(rootdev));
964 
965 		rootdv = finddevice(buf);
966 		if (rootdv == NULL) {
967 			printf("device %s (0x%x) not configured\n",
968 			    buf, rootdev);
969 			boothowto |= RB_ASKNAME;
970 			goto top;
971 		}
972 	}
973 
974  haveroot:
975 
976 	root_device = rootdv;
977 
978 	switch (rootdv->dv_class) {
979 	case DV_IFNET:
980 		aprint_normal("root on %s", rootdv->dv_xname);
981 		break;
982 
983 	case DV_DISK:
984 		aprint_normal("root on %s%c", rootdv->dv_xname,
985 		    DISKPART(rootdev) + 'a');
986 		break;
987 
988 	default:
989 		printf("can't determine root device\n");
990 		boothowto |= RB_ASKNAME;
991 		goto top;
992 	}
993 
994 	/*
995 	 * Now configure the dump device.
996 	 *
997 	 * If we haven't figured out the dump device, do so, with
998 	 * the following rules:
999 	 *
1000 	 *	(a) We already know dumpdv in the RB_ASKNAME case.
1001 	 *
1002 	 *	(b) If dumpspec is set, try to use it.  If the device
1003 	 *	    is not available, punt.
1004 	 *
1005 	 *	(c) If dumpspec is not set, the dump device is
1006 	 *	    wildcarded or unspecified.  If the root device
1007 	 *	    is DV_IFNET, punt.  Otherwise, use partition b
1008 	 *	    of the root device.
1009 	 */
1010 
1011 	if (boothowto & RB_ASKNAME) {		/* (a) */
1012 		if (dumpdv == NULL)
1013 			goto nodumpdev;
1014 	} else if (dumpspec != NULL) {		/* (b) */
1015 		if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
1016 			/*
1017 			 * Operator doesn't want a dump device.
1018 			 * Or looks like they tried to pick a network
1019 			 * device.  Oops.
1020 			 */
1021 			goto nodumpdev;
1022 		}
1023 
1024 		dumpdevname = devsw_blk2name(major(dumpdev));
1025 		if (dumpdevname == NULL)
1026 			goto nodumpdev;
1027 		memset(buf, 0, sizeof(buf));
1028 		sprintf(buf, "%s%d", dumpdevname, DISKUNIT(dumpdev));
1029 
1030 		dumpdv = finddevice(buf);
1031 		if (dumpdv == NULL) {
1032 			/*
1033 			 * Device not configured.
1034 			 */
1035 			goto nodumpdev;
1036 		}
1037 	} else {				/* (c) */
1038 		if (rootdv->dv_class == DV_IFNET)
1039 			goto nodumpdev;
1040 		else {
1041 			dumpdv = rootdv;
1042 			dumpdev = MAKEDISKDEV(major(rootdev),
1043 			    dumpdv->dv_unit, 1);
1044 		}
1045 	}
1046 
1047 	aprint_normal(" dumps on %s%c\n", dumpdv->dv_xname,
1048 	    DISKPART(dumpdev) + 'a');
1049 	return;
1050 
1051  nodumpdev:
1052 	dumpdev = NODEV;
1053 	aprint_normal("\n");
1054 }
1055 
1056 static struct device *
1057 finddevice(name)
1058 	const char *name;
1059 {
1060 	struct device *dv;
1061 #ifdef BOOT_FROM_RAID_HOOKS
1062 	int j;
1063 
1064 	for (j = 0; j < numraid; j++) {
1065 		if (strcmp(name, raidrootdev[j].dv_xname) == 0) {
1066 			dv = &raidrootdev[j];
1067 			return (dv);
1068 		}
1069 	}
1070 #endif
1071 
1072 	for (dv = TAILQ_FIRST(&alldevs); dv != NULL;
1073 	    dv = TAILQ_NEXT(dv, dv_list))
1074 		if (strcmp(dv->dv_xname, name) == 0)
1075 			break;
1076 	return (dv);
1077 }
1078 
1079 static struct device *
1080 getdisk(str, len, defpart, devp, isdump)
1081 	char *str;
1082 	int len, defpart;
1083 	dev_t *devp;
1084 	int isdump;
1085 {
1086 	struct device	*dv;
1087 #ifdef MEMORY_DISK_HOOKS
1088 	int		i;
1089 #endif
1090 #ifdef BOOT_FROM_RAID_HOOKS
1091 	int 		j;
1092 #endif
1093 
1094 	if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
1095 		printf("use one of:");
1096 #ifdef MEMORY_DISK_HOOKS
1097 		if (isdump == 0)
1098 			for (i = 0; i < NMD; i++)
1099 				printf(" %s[a-%c]", fakemdrootdev[i].dv_xname,
1100 				    'a' + MAXPARTITIONS - 1);
1101 #endif
1102 #ifdef BOOT_FROM_RAID_HOOKS
1103 		if (isdump == 0)
1104 			for (j = 0; j < numraid; j++)
1105 				printf(" %s[a-%c]", raidrootdev[j].dv_xname,
1106 				    'a' + MAXPARTITIONS - 1);
1107 #endif
1108 		TAILQ_FOREACH(dv, &alldevs, dv_list) {
1109 			if (dv->dv_class == DV_DISK)
1110 				printf(" %s[a-%c]", dv->dv_xname,
1111 				    'a' + MAXPARTITIONS - 1);
1112 			if (isdump == 0 && dv->dv_class == DV_IFNET)
1113 				printf(" %s", dv->dv_xname);
1114 		}
1115 		if (isdump)
1116 			printf(" none");
1117 #if defined(DDB)
1118 		printf(" ddb");
1119 #endif
1120 		printf(" halt reboot\n");
1121 	}
1122 	return (dv);
1123 }
1124 
1125 static struct device *
1126 parsedisk(str, len, defpart, devp)
1127 	char *str;
1128 	int len, defpart;
1129 	dev_t *devp;
1130 {
1131 	struct device *dv;
1132 	char *cp, c;
1133 	int majdev, part;
1134 #ifdef MEMORY_DISK_HOOKS
1135 	int i;
1136 #endif
1137 	if (len == 0)
1138 		return (NULL);
1139 
1140 	if (len == 4 && strcmp(str, "halt") == 0)
1141 		cpu_reboot(RB_HALT, NULL);
1142 	else if (len == 6 && strcmp(str, "reboot") == 0)
1143 		cpu_reboot(0, NULL);
1144 #if defined(DDB)
1145 	else if (len == 3 && strcmp(str, "ddb") == 0)
1146 		console_debugger();
1147 #endif
1148 
1149 	cp = str + len - 1;
1150 	c = *cp;
1151 	if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
1152 		part = c - 'a';
1153 		*cp = '\0';
1154 	} else
1155 		part = defpart;
1156 
1157 #ifdef MEMORY_DISK_HOOKS
1158 	for (i = 0; i < NMD; i++)
1159 		if (strcmp(str, fakemdrootdev[i].dv_xname) == 0) {
1160 			dv = &fakemdrootdev[i];
1161 			goto gotdisk;
1162 		}
1163 #endif
1164 
1165 	dv = finddevice(str);
1166 	if (dv != NULL) {
1167 		if (dv->dv_class == DV_DISK) {
1168 #ifdef MEMORY_DISK_HOOKS
1169  gotdisk:
1170 #endif
1171 			majdev = devsw_name2blk(dv->dv_xname, NULL, 0);
1172 			if (majdev < 0)
1173 				panic("parsedisk");
1174 			*devp = MAKEDISKDEV(majdev, dv->dv_unit, part);
1175 		}
1176 
1177 		if (dv->dv_class == DV_IFNET)
1178 			*devp = NODEV;
1179 	}
1180 
1181 	*cp = c;
1182 	return (dv);
1183 }
1184 
1185 /*
1186  * snprintf() `bytes' into `buf', reformatting it so that the number,
1187  * plus a possible `x' + suffix extension) fits into len bytes (including
1188  * the terminating NUL).
1189  * Returns the number of bytes stored in buf, or -1 if there was a problem.
1190  * E.g, given a len of 9 and a suffix of `B':
1191  *	bytes		result
1192  *	-----		------
1193  *	99999		`99999 B'
1194  *	100000		`97 kB'
1195  *	66715648	`65152 kB'
1196  *	252215296	`240 MB'
1197  */
1198 int
1199 humanize_number(buf, len, bytes, suffix, divisor)
1200 	char		*buf;
1201 	size_t		 len;
1202 	u_int64_t	 bytes;
1203 	const char	*suffix;
1204 	int 		divisor;
1205 {
1206        	/* prefixes are: (none), kilo, Mega, Giga, Tera, Peta, Exa */
1207 	const char *prefixes;
1208 	int		r;
1209 	u_int64_t	max;
1210 	size_t		i, suffixlen;
1211 
1212 	if (buf == NULL || suffix == NULL)
1213 		return (-1);
1214 	if (len > 0)
1215 		buf[0] = '\0';
1216 	suffixlen = strlen(suffix);
1217 	/* check if enough room for `x y' + suffix + `\0' */
1218 	if (len < 4 + suffixlen)
1219 		return (-1);
1220 
1221 	if (divisor == 1024) {
1222 		/*
1223 		 * binary multiplies
1224 		 * XXX IEC 60027-2 recommends Ki, Mi, Gi...
1225 		 */
1226 		prefixes = " KMGTPE";
1227 	} else
1228 		prefixes = " kMGTPE"; /* SI for decimal multiplies */
1229 
1230 	max = 1;
1231 	for (i = 0; i < len - suffixlen - 3; i++)
1232 		max *= 10;
1233 	for (i = 0; bytes >= max && prefixes[i + 1]; i++)
1234 		bytes /= divisor;
1235 
1236 	r = snprintf(buf, len, "%qu%s%c%s", (unsigned long long)bytes,
1237 	    i == 0 ? "" : " ", prefixes[i], suffix);
1238 
1239 	return (r);
1240 }
1241 
1242 int
1243 format_bytes(buf, len, bytes)
1244 	char		*buf;
1245 	size_t		 len;
1246 	u_int64_t	 bytes;
1247 {
1248 	int	rv;
1249 	size_t	nlen;
1250 
1251 	rv = humanize_number(buf, len, bytes, "B", 1024);
1252 	if (rv != -1) {
1253 			/* nuke the trailing ` B' if it exists */
1254 		nlen = strlen(buf) - 2;
1255 		if (strcmp(&buf[nlen], " B") == 0)
1256 			buf[nlen] = '\0';
1257 	}
1258 	return (rv);
1259 }
1260 
1261 /*
1262  * Start trace of particular system call. If process is being traced,
1263  * this routine is called by MD syscall dispatch code just before
1264  * a system call is actually executed.
1265  * MD caller guarantees the passed 'code' is within the supported
1266  * system call number range for emulation the process runs under.
1267  */
1268 int
1269 trace_enter(struct lwp *l, register_t code,
1270 	register_t realcode, const struct sysent *callp, void *args,
1271 	register_t rval[])
1272 {
1273 #if defined(KTRACE) || defined(SYSTRACE)
1274 	struct proc *p = l->l_proc;
1275 #endif
1276 
1277 #ifdef SYSCALL_DEBUG
1278 	scdebug_call(l, code, args);
1279 #endif /* SYSCALL_DEBUG */
1280 
1281 #ifdef KTRACE
1282 	if (KTRPOINT(p, KTR_SYSCALL))
1283 		ktrsyscall(p, code, realcode, callp, args);
1284 #endif /* KTRACE */
1285 
1286 #ifdef SYSTRACE
1287 	if (ISSET(p->p_flag, P_SYSTRACE))
1288 		return systrace_enter(p, code, args, rval);
1289 #endif
1290 	return 0;
1291 }
1292 
1293 /*
1294  * End trace of particular system call. If process is being traced,
1295  * this routine is called by MD syscall dispatch code just after
1296  * a system call finishes.
1297  * MD caller guarantees the passed 'code' is within the supported
1298  * system call number range for emulation the process runs under.
1299  */
1300 void
1301 trace_exit(struct lwp *l, register_t code, void *args, register_t rval[],
1302     int error)
1303 {
1304 #if defined(KTRACE) || defined(SYSTRACE)
1305 	struct proc *p = l->l_proc;
1306 #endif
1307 
1308 #ifdef SYSCALL_DEBUG
1309 	scdebug_ret(l, code, error, rval);
1310 #endif /* SYSCALL_DEBUG */
1311 
1312 #ifdef KTRACE
1313 	if (KTRPOINT(p, KTR_SYSRET)) {
1314 		KERNEL_PROC_LOCK(l);
1315 		ktrsysret(p, code, error, rval);
1316 		KERNEL_PROC_UNLOCK(l);
1317 	}
1318 #endif /* KTRACE */
1319 
1320 #ifdef SYSTRACE
1321 	if (ISSET(p->p_flag, P_SYSTRACE))
1322 		systrace_exit(p, code, args, rval, error);
1323 #endif
1324 }
1325