xref: /netbsd-src/sys/kern/kern_subr.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: kern_subr.c,v 1.117 2005/06/23 23:15:12 thorpej 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 1.117 2005/06/23 23:15:12 thorpej 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 MALLOC_DEFINE(M_IOV, "iov", "large iov's");
132 
133 int
134 uiomove(void *buf, size_t n, struct uio *uio)
135 {
136 	struct iovec *iov;
137 	u_int cnt;
138 	int error = 0;
139 	char *cp = buf;
140 	struct proc *p = uio->uio_procp;
141 	int hold_count;
142 
143 	hold_count = KERNEL_LOCK_RELEASE_ALL();
144 
145 #if defined(LOCKDEBUG) || defined(DIAGNOSTIC)
146 	spinlock_switchcheck();
147 #endif
148 #ifdef LOCKDEBUG
149 	simple_lock_only_held(NULL, "uiomove");
150 #endif
151 
152 #ifdef DIAGNOSTIC
153 	if (uio->uio_rw != UIO_READ && uio->uio_rw != UIO_WRITE)
154 		panic("uiomove: mode");
155 #endif
156 	while (n > 0 && uio->uio_resid) {
157 		iov = uio->uio_iov;
158 		cnt = iov->iov_len;
159 		if (cnt == 0) {
160 			KASSERT(uio->uio_iovcnt > 0);
161 			uio->uio_iov++;
162 			uio->uio_iovcnt--;
163 			continue;
164 		}
165 		if (cnt > n)
166 			cnt = n;
167 		switch (uio->uio_segflg) {
168 
169 		case UIO_USERSPACE:
170 			if (curcpu()->ci_schedstate.spc_flags &
171 			    SPCF_SHOULDYIELD)
172 				preempt(1);
173 			if (__predict_true(p == curproc)) {
174 				if (uio->uio_rw == UIO_READ)
175 					error = copyout(cp, iov->iov_base, cnt);
176 				else
177 					error = copyin(iov->iov_base, cp, cnt);
178 			} else {
179 				if (uio->uio_rw == UIO_READ)
180 					error = copyout_proc(p, cp,
181 					    iov->iov_base, cnt);
182 				else
183 					error = copyin_proc(p, iov->iov_base,
184 					    cp, cnt);
185 			}
186 			if (error)
187 				goto out;
188 			break;
189 
190 		case UIO_SYSSPACE:
191 			if (uio->uio_rw == UIO_READ)
192 				error = kcopy(cp, iov->iov_base, cnt);
193 			else
194 				error = kcopy(iov->iov_base, cp, cnt);
195 			if (error)
196 				goto out;
197 			break;
198 		}
199 		iov->iov_base = (caddr_t)iov->iov_base + cnt;
200 		iov->iov_len -= cnt;
201 		uio->uio_resid -= cnt;
202 		uio->uio_offset += cnt;
203 		cp += cnt;
204 		KDASSERT(cnt <= n);
205 		n -= cnt;
206 	}
207 out:
208 	KERNEL_LOCK_ACQUIRE_COUNT(hold_count);
209 	return (error);
210 }
211 
212 /*
213  * Wrapper for uiomove() that validates the arguments against a known-good
214  * kernel buffer.
215  */
216 int
217 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
218 {
219 	size_t offset;
220 
221 	if (uio->uio_offset < 0 || uio->uio_resid < 0 ||
222 	    (offset = uio->uio_offset) != uio->uio_offset)
223 		return (EINVAL);
224 	if (offset >= buflen)
225 		return (0);
226 	return (uiomove((char *)buf + offset, buflen - offset, uio));
227 }
228 
229 /*
230  * Give next character to user as result of read.
231  */
232 int
233 ureadc(int c, struct uio *uio)
234 {
235 	struct iovec *iov;
236 
237 	if (uio->uio_resid <= 0)
238 		panic("ureadc: non-positive resid");
239 again:
240 	if (uio->uio_iovcnt <= 0)
241 		panic("ureadc: non-positive iovcnt");
242 	iov = uio->uio_iov;
243 	if (iov->iov_len <= 0) {
244 		uio->uio_iovcnt--;
245 		uio->uio_iov++;
246 		goto again;
247 	}
248 	switch (uio->uio_segflg) {
249 
250 	case UIO_USERSPACE:
251 		if (subyte(iov->iov_base, c) < 0)
252 			return (EFAULT);
253 		break;
254 
255 	case UIO_SYSSPACE:
256 		*(char *)iov->iov_base = c;
257 		break;
258 	}
259 	iov->iov_base = (caddr_t)iov->iov_base + 1;
260 	iov->iov_len--;
261 	uio->uio_resid--;
262 	uio->uio_offset++;
263 	return (0);
264 }
265 
266 /*
267  * Like copyin(), but operates on an arbitrary process.
268  */
269 int
270 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
271 {
272 	struct iovec iov;
273 	struct uio uio;
274 	int error;
275 
276 	if (len == 0)
277 		return (0);
278 
279 	iov.iov_base = kaddr;
280 	iov.iov_len = len;
281 	uio.uio_iov = &iov;
282 	uio.uio_iovcnt = 1;
283 	uio.uio_offset = (off_t)(intptr_t)uaddr;
284 	uio.uio_resid = len;
285 	uio.uio_segflg = UIO_SYSSPACE;
286 	uio.uio_rw = UIO_READ;
287 	uio.uio_procp = NULL;
288 
289 	/* XXXCDC: how should locking work here? */
290 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
291 		return (EFAULT);
292 	p->p_vmspace->vm_refcnt++;	/* XXX */
293 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
294 	uvmspace_free(p->p_vmspace);
295 
296 	return (error);
297 }
298 
299 /*
300  * Like copyout(), but operates on an arbitrary process.
301  */
302 int
303 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
304 {
305 	struct iovec iov;
306 	struct uio uio;
307 	int error;
308 
309 	if (len == 0)
310 		return (0);
311 
312 	iov.iov_base = __UNCONST(kaddr); /* XXXUNCONST cast away const */
313 	iov.iov_len = len;
314 	uio.uio_iov = &iov;
315 	uio.uio_iovcnt = 1;
316 	uio.uio_offset = (off_t)(intptr_t)uaddr;
317 	uio.uio_resid = len;
318 	uio.uio_segflg = UIO_SYSSPACE;
319 	uio.uio_rw = UIO_WRITE;
320 	uio.uio_procp = NULL;
321 
322 	/* XXXCDC: how should locking work here? */
323 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
324 		return (EFAULT);
325 	p->p_vmspace->vm_refcnt++;	/* XXX */
326 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
327 	uvmspace_free(p->p_vmspace);
328 
329 	return (error);
330 }
331 
332 /*
333  * General routine to allocate a hash table.
334  * Allocate enough memory to hold at least `elements' list-head pointers.
335  * Return a pointer to the allocated space and set *hashmask to a pattern
336  * suitable for masking a value to use as an index into the returned array.
337  */
338 void *
339 hashinit(u_int elements, enum hashtype htype, struct malloc_type *mtype,
340     int mflags, u_long *hashmask)
341 {
342 	u_long hashsize, i;
343 	LIST_HEAD(, generic) *hashtbl_list;
344 	TAILQ_HEAD(, generic) *hashtbl_tailq;
345 	size_t esize;
346 	void *p;
347 
348 	if (elements == 0)
349 		panic("hashinit: bad cnt");
350 	for (hashsize = 1; hashsize < elements; hashsize <<= 1)
351 		continue;
352 
353 	switch (htype) {
354 	case HASH_LIST:
355 		esize = sizeof(*hashtbl_list);
356 		break;
357 	case HASH_TAILQ:
358 		esize = sizeof(*hashtbl_tailq);
359 		break;
360 	default:
361 #ifdef DIAGNOSTIC
362 		panic("hashinit: invalid table type");
363 #else
364 		return NULL;
365 #endif
366 	}
367 
368 	if ((p = malloc(hashsize * esize, mtype, mflags)) == NULL)
369 		return (NULL);
370 
371 	switch (htype) {
372 	case HASH_LIST:
373 		hashtbl_list = p;
374 		for (i = 0; i < hashsize; i++)
375 			LIST_INIT(&hashtbl_list[i]);
376 		break;
377 	case HASH_TAILQ:
378 		hashtbl_tailq = p;
379 		for (i = 0; i < hashsize; i++)
380 			TAILQ_INIT(&hashtbl_tailq[i]);
381 		break;
382 	}
383 	*hashmask = hashsize - 1;
384 	return (p);
385 }
386 
387 /*
388  * Free memory from hash table previosly allocated via hashinit().
389  */
390 void
391 hashdone(void *hashtbl, struct malloc_type *mtype)
392 {
393 
394 	free(hashtbl, mtype);
395 }
396 
397 
398 static void *
399 hook_establish(hook_list_t *list, void (*fn)(void *), void *arg)
400 {
401 	struct hook_desc *hd;
402 
403 	hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
404 	if (hd == NULL)
405 		return (NULL);
406 
407 	hd->hk_fn = fn;
408 	hd->hk_arg = arg;
409 	LIST_INSERT_HEAD(list, hd, hk_list);
410 
411 	return (hd);
412 }
413 
414 static void
415 hook_disestablish(hook_list_t *list, void *vhook)
416 {
417 #ifdef DIAGNOSTIC
418 	struct hook_desc *hd;
419 
420 	LIST_FOREACH(hd, list, hk_list) {
421                 if (hd == vhook)
422 			break;
423 	}
424 
425 	if (hd == NULL)
426 		panic("hook_disestablish: hook %p not established", vhook);
427 #endif
428 	LIST_REMOVE((struct hook_desc *)vhook, hk_list);
429 	free(vhook, M_DEVBUF);
430 }
431 
432 static void
433 hook_destroy(hook_list_t *list)
434 {
435 	struct hook_desc *hd;
436 
437 	while ((hd = LIST_FIRST(list)) != NULL) {
438 		LIST_REMOVE(hd, hk_list);
439 		free(hd, M_DEVBUF);
440 	}
441 }
442 
443 static void
444 hook_proc_run(hook_list_t *list, struct proc *p)
445 {
446 	struct hook_desc *hd;
447 
448 	for (hd = LIST_FIRST(list); hd != NULL; hd = LIST_NEXT(hd, hk_list)) {
449 		((void (*)(struct proc *, void *))*hd->hk_fn)(p,
450 		    hd->hk_arg);
451 	}
452 }
453 
454 /*
455  * "Shutdown hook" types, functions, and variables.
456  *
457  * Should be invoked immediately before the
458  * system is halted or rebooted, i.e. after file systems unmounted,
459  * after crash dump done, etc.
460  *
461  * Each shutdown hook is removed from the list before it's run, so that
462  * it won't be run again.
463  */
464 
465 static hook_list_t shutdownhook_list;
466 
467 void *
468 shutdownhook_establish(void (*fn)(void *), void *arg)
469 {
470 	return hook_establish(&shutdownhook_list, fn, arg);
471 }
472 
473 void
474 shutdownhook_disestablish(void *vhook)
475 {
476 	hook_disestablish(&shutdownhook_list, vhook);
477 }
478 
479 /*
480  * Run shutdown hooks.  Should be invoked immediately before the
481  * system is halted or rebooted, i.e. after file systems unmounted,
482  * after crash dump done, etc.
483  *
484  * Each shutdown hook is removed from the list before it's run, so that
485  * it won't be run again.
486  */
487 void
488 doshutdownhooks(void)
489 {
490 	struct hook_desc *dp;
491 
492 	while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
493 		LIST_REMOVE(dp, hk_list);
494 		(*dp->hk_fn)(dp->hk_arg);
495 #if 0
496 		/*
497 		 * Don't bother freeing the hook structure,, since we may
498 		 * be rebooting because of a memory corruption problem,
499 		 * and this might only make things worse.  It doesn't
500 		 * matter, anyway, since the system is just about to
501 		 * reboot.
502 		 */
503 		free(dp, M_DEVBUF);
504 #endif
505 	}
506 }
507 
508 /*
509  * "Mountroot hook" types, functions, and variables.
510  */
511 
512 static hook_list_t mountroothook_list;
513 
514 void *
515 mountroothook_establish(void (*fn)(struct device *), struct device *dev)
516 {
517 	return hook_establish(&mountroothook_list, (void (*)(void *))fn, dev);
518 }
519 
520 void
521 mountroothook_disestablish(void *vhook)
522 {
523 	hook_disestablish(&mountroothook_list, vhook);
524 }
525 
526 void
527 mountroothook_destroy(void)
528 {
529 	hook_destroy(&mountroothook_list);
530 }
531 
532 void
533 domountroothook(void)
534 {
535 	struct hook_desc *hd;
536 
537 	LIST_FOREACH(hd, &mountroothook_list, hk_list) {
538 		if (hd->hk_arg == (void *)root_device) {
539 			(*hd->hk_fn)(hd->hk_arg);
540 			return;
541 		}
542 	}
543 }
544 
545 static hook_list_t exechook_list;
546 
547 void *
548 exechook_establish(void (*fn)(struct proc *, void *), void *arg)
549 {
550 	return hook_establish(&exechook_list, (void (*)(void *))fn, arg);
551 }
552 
553 void
554 exechook_disestablish(void *vhook)
555 {
556 	hook_disestablish(&exechook_list, vhook);
557 }
558 
559 /*
560  * Run exec hooks.
561  */
562 void
563 doexechooks(struct proc *p)
564 {
565 	hook_proc_run(&exechook_list, p);
566 }
567 
568 static hook_list_t exithook_list;
569 
570 void *
571 exithook_establish(void (*fn)(struct proc *, void *), void *arg)
572 {
573 	return hook_establish(&exithook_list, (void (*)(void *))fn, arg);
574 }
575 
576 void
577 exithook_disestablish(void *vhook)
578 {
579 	hook_disestablish(&exithook_list, vhook);
580 }
581 
582 /*
583  * Run exit hooks.
584  */
585 void
586 doexithooks(struct proc *p)
587 {
588 	hook_proc_run(&exithook_list, p);
589 }
590 
591 static hook_list_t forkhook_list;
592 
593 void *
594 forkhook_establish(void (*fn)(struct proc *, struct proc *))
595 {
596 	return hook_establish(&forkhook_list, (void (*)(void *))fn, NULL);
597 }
598 
599 void
600 forkhook_disestablish(void *vhook)
601 {
602 	hook_disestablish(&forkhook_list, vhook);
603 }
604 
605 /*
606  * Run fork hooks.
607  */
608 void
609 doforkhooks(struct proc *p2, struct proc *p1)
610 {
611 	struct hook_desc *hd;
612 
613 	LIST_FOREACH(hd, &forkhook_list, hk_list) {
614 		((void (*)(struct proc *, struct proc *))*hd->hk_fn)
615 		    (p2, p1);
616 	}
617 }
618 
619 /*
620  * "Power hook" types, functions, and variables.
621  * The list of power hooks is kept ordered with the last registered hook
622  * first.
623  * When running the hooks on power down the hooks are called in reverse
624  * registration order, when powering up in registration order.
625  */
626 struct powerhook_desc {
627 	CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
628 	void	(*sfd_fn)(int, void *);
629 	void	*sfd_arg;
630 };
631 
632 static CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
633     CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
634 
635 void *
636 powerhook_establish(void (*fn)(int, void *), void *arg)
637 {
638 	struct powerhook_desc *ndp;
639 
640 	ndp = (struct powerhook_desc *)
641 	    malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
642 	if (ndp == NULL)
643 		return (NULL);
644 
645 	ndp->sfd_fn = fn;
646 	ndp->sfd_arg = arg;
647 	CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
648 
649 	return (ndp);
650 }
651 
652 void
653 powerhook_disestablish(void *vhook)
654 {
655 #ifdef DIAGNOSTIC
656 	struct powerhook_desc *dp;
657 
658 	CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
659                 if (dp == vhook)
660 			goto found;
661 	panic("powerhook_disestablish: hook %p not established", vhook);
662  found:
663 #endif
664 
665 	CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
666 	    sfd_list);
667 	free(vhook, M_DEVBUF);
668 }
669 
670 /*
671  * Run power hooks.
672  */
673 void
674 dopowerhooks(int why)
675 {
676 	struct powerhook_desc *dp;
677 
678 	if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
679 		CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
680 			(*dp->sfd_fn)(why, dp->sfd_arg);
681 		}
682 	} else {
683 		CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
684 			(*dp->sfd_fn)(why, dp->sfd_arg);
685 		}
686 	}
687 }
688 
689 /*
690  * Determine the root device and, if instructed to, the root file system.
691  */
692 
693 #include "md.h"
694 #if NMD == 0
695 #undef MEMORY_DISK_HOOKS
696 #endif
697 
698 #ifdef MEMORY_DISK_HOOKS
699 static struct device fakemdrootdev[NMD];
700 #endif
701 
702 #ifdef MEMORY_DISK_IS_ROOT
703 #define BOOT_FROM_MEMORY_HOOKS 1
704 #endif
705 
706 #include "raid.h"
707 #if NRAID == 1
708 #define BOOT_FROM_RAID_HOOKS 1
709 #endif
710 
711 #ifdef BOOT_FROM_RAID_HOOKS
712 extern int numraid;
713 extern struct device *raidrootdev;
714 #endif
715 
716 /*
717  * The device and wedge that we booted from.  If booted_wedge is NULL,
718  * the we might consult booted_partition.
719  */
720 struct device *booted_device;
721 struct device *booted_wedge;
722 int booted_partition;
723 
724 /*
725  * Use partition letters if it's a disk class but not a wedge.
726  * XXX Check for wedge is kinda gross.
727  */
728 #define	DEV_USES_PARTITIONS(dv)						\
729 	((dv)->dv_class == DV_DISK &&					\
730 	((dv)->dv_cfdata == NULL ||					\
731 	 strcmp((dv)->dv_cfdata->cf_name, "dk") != 0))
732 
733 void
734 setroot(struct device *bootdv, int bootpartition)
735 {
736 	struct device *dv;
737 	int len;
738 #ifdef MEMORY_DISK_HOOKS
739 	int i;
740 #endif
741 	dev_t nrootdev;
742 	dev_t ndumpdev = NODEV;
743 	char buf[128];
744 	const char *rootdevname;
745 	const char *dumpdevname;
746 	struct device *rootdv = NULL;		/* XXX gcc -Wuninitialized */
747 	struct device *dumpdv = NULL;
748 	struct ifnet *ifp;
749 	const char *deffsname;
750 	struct vfsops *vops;
751 
752 #ifdef MEMORY_DISK_HOOKS
753 	for (i = 0; i < NMD; i++) {
754 		fakemdrootdev[i].dv_class  = DV_DISK;
755 		fakemdrootdev[i].dv_cfdata = NULL;
756 		fakemdrootdev[i].dv_unit   = i;
757 		fakemdrootdev[i].dv_parent = NULL;
758 		snprintf(fakemdrootdev[i].dv_xname,
759 		    sizeof(fakemdrootdev[i].dv_xname), "md%d", i);
760 	}
761 #endif /* MEMORY_DISK_HOOKS */
762 
763 #ifdef MEMORY_DISK_IS_ROOT
764 	bootdv = &fakemdrootdev[0];
765 	bootpartition = 0;
766 #endif
767 
768 	/*
769 	 * If NFS is specified as the file system, and we found
770 	 * a DV_DISK boot device (or no boot device at all), then
771 	 * find a reasonable network interface for "rootspec".
772 	 */
773 	vops = vfs_getopsbyname("nfs");
774 	if (vops != NULL && vops->vfs_mountroot == mountroot &&
775 	    rootspec == NULL &&
776 	    (bootdv == NULL || bootdv->dv_class != DV_IFNET)) {
777 		IFNET_FOREACH(ifp) {
778 			if ((ifp->if_flags &
779 			     (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
780 				break;
781 		}
782 		if (ifp == NULL) {
783 			/*
784 			 * Can't find a suitable interface; ask the
785 			 * user.
786 			 */
787 			boothowto |= RB_ASKNAME;
788 		} else {
789 			/*
790 			 * Have a suitable interface; behave as if
791 			 * the user specified this interface.
792 			 */
793 			rootspec = (const char *)ifp->if_xname;
794 		}
795 	}
796 
797 	/*
798 	 * If wildcarded root and we the boot device wasn't determined,
799 	 * ask the user.
800 	 */
801 	if (rootspec == NULL && bootdv == NULL)
802 		boothowto |= RB_ASKNAME;
803 
804  top:
805 	if (boothowto & RB_ASKNAME) {
806 		struct device *defdumpdv;
807 
808 		for (;;) {
809 			printf("root device");
810 			if (bootdv != NULL) {
811 				printf(" (default %s", bootdv->dv_xname);
812 				if (DEV_USES_PARTITIONS(bootdv))
813 					printf("%c", bootpartition + 'a');
814 				printf(")");
815 			}
816 			printf(": ");
817 			len = cngetsn(buf, sizeof(buf));
818 			if (len == 0 && bootdv != NULL) {
819 				strlcpy(buf, bootdv->dv_xname, sizeof(buf));
820 				len = strlen(buf);
821 			}
822 			if (len > 0 && buf[len - 1] == '*') {
823 				buf[--len] = '\0';
824 				dv = getdisk(buf, len, 1, &nrootdev, 0);
825 				if (dv != NULL) {
826 					rootdv = dv;
827 					break;
828 				}
829 			}
830 			dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
831 			if (dv != NULL) {
832 				rootdv = dv;
833 				break;
834 			}
835 		}
836 
837 		/*
838 		 * Set up the default dump device.  If root is on
839 		 * a network device, there is no default dump
840 		 * device, since we don't support dumps to the
841 		 * network.
842 		 */
843 		if (DEV_USES_PARTITIONS(rootdv) == 0)
844 			defdumpdv = NULL;
845 		else
846 			defdumpdv = rootdv;
847 
848 		for (;;) {
849 			printf("dump device");
850 			if (defdumpdv != NULL) {
851 				/*
852 				 * Note, we know it's a disk if we get here.
853 				 */
854 				printf(" (default %sb)", defdumpdv->dv_xname);
855 			}
856 			printf(": ");
857 			len = cngetsn(buf, sizeof(buf));
858 			if (len == 0) {
859 				if (defdumpdv != NULL) {
860 					ndumpdev = MAKEDISKDEV(major(nrootdev),
861 					    DISKUNIT(nrootdev), 1);
862 				}
863 				dumpdv = defdumpdv;
864 				break;
865 			}
866 			if (len == 4 && strcmp(buf, "none") == 0) {
867 				dumpdv = NULL;
868 				break;
869 			}
870 			dv = getdisk(buf, len, 1, &ndumpdev, 1);
871 			if (dv != NULL) {
872 				dumpdv = dv;
873 				break;
874 			}
875 		}
876 
877 		rootdev = nrootdev;
878 		dumpdev = ndumpdev;
879 
880 		for (vops = LIST_FIRST(&vfs_list); vops != NULL;
881 		     vops = LIST_NEXT(vops, vfs_list)) {
882 			if (vops->vfs_mountroot != NULL &&
883 			    vops->vfs_mountroot == mountroot)
884 			break;
885 		}
886 
887 		if (vops == NULL) {
888 			mountroot = NULL;
889 			deffsname = "generic";
890 		} else
891 			deffsname = vops->vfs_name;
892 
893 		for (;;) {
894 			printf("file system (default %s): ", deffsname);
895 			len = cngetsn(buf, sizeof(buf));
896 			if (len == 0)
897 				break;
898 			if (len == 4 && strcmp(buf, "halt") == 0)
899 				cpu_reboot(RB_HALT, NULL);
900 			else if (len == 6 && strcmp(buf, "reboot") == 0)
901 				cpu_reboot(0, NULL);
902 #if defined(DDB)
903 			else if (len == 3 && strcmp(buf, "ddb") == 0) {
904 				console_debugger();
905 			}
906 #endif
907 			else if (len == 7 && strcmp(buf, "generic") == 0) {
908 				mountroot = NULL;
909 				break;
910 			}
911 			vops = vfs_getopsbyname(buf);
912 			if (vops == NULL || vops->vfs_mountroot == NULL) {
913 				printf("use one of: generic");
914 				for (vops = LIST_FIRST(&vfs_list);
915 				     vops != NULL;
916 				     vops = LIST_NEXT(vops, vfs_list)) {
917 					if (vops->vfs_mountroot != NULL)
918 						printf(" %s", vops->vfs_name);
919 				}
920 #if defined(DDB)
921 				printf(" ddb");
922 #endif
923 				printf(" halt reboot\n");
924 			} else {
925 				mountroot = vops->vfs_mountroot;
926 				break;
927 			}
928 		}
929 
930 	} else if (rootspec == NULL) {
931 		int majdev;
932 
933 		/*
934 		 * Wildcarded root; use the boot device.
935 		 */
936 		rootdv = bootdv;
937 
938 		majdev = devsw_name2blk(bootdv->dv_xname, NULL, 0);
939 		if (majdev >= 0) {
940 			/*
941 			 * Root is on a disk.  `bootpartition' is root,
942 			 * unless the device does not use partitions.
943 			 */
944 			if (DEV_USES_PARTITIONS(bootdv))
945 				rootdev = MAKEDISKDEV(majdev, bootdv->dv_unit,
946 				    bootpartition);
947 			else
948 				rootdev = makedev(majdev, bootdv->dv_unit);
949 		}
950 	} else {
951 
952 		/*
953 		 * `root on <dev> ...'
954 		 */
955 
956 		/*
957 		 * If it's a network interface, we can bail out
958 		 * early.
959 		 */
960 		dv = finddevice(rootspec);
961 		if (dv != NULL && dv->dv_class == DV_IFNET) {
962 			rootdv = dv;
963 			goto haveroot;
964 		}
965 
966 		rootdevname = devsw_blk2name(major(rootdev));
967 		if (rootdevname == NULL) {
968 			printf("unknown device major 0x%x\n", rootdev);
969 			boothowto |= RB_ASKNAME;
970 			goto top;
971 		}
972 		memset(buf, 0, sizeof(buf));
973 		snprintf(buf, sizeof(buf), "%s%d", rootdevname,
974 		    DISKUNIT(rootdev));
975 
976 		rootdv = finddevice(buf);
977 		if (rootdv == NULL) {
978 			printf("device %s (0x%x) not configured\n",
979 			    buf, rootdev);
980 			boothowto |= RB_ASKNAME;
981 			goto top;
982 		}
983 	}
984 
985  haveroot:
986 
987 	root_device = rootdv;
988 
989 	switch (rootdv->dv_class) {
990 	case DV_IFNET:
991 		aprint_normal("root on %s", rootdv->dv_xname);
992 		break;
993 
994 	case DV_DISK:
995 		aprint_normal("root on %s%c", rootdv->dv_xname,
996 		    DISKPART(rootdev) + 'a');
997 		break;
998 
999 	default:
1000 		printf("can't determine root device\n");
1001 		boothowto |= RB_ASKNAME;
1002 		goto top;
1003 	}
1004 
1005 	/*
1006 	 * Now configure the dump device.
1007 	 *
1008 	 * If we haven't figured out the dump device, do so, with
1009 	 * the following rules:
1010 	 *
1011 	 *	(a) We already know dumpdv in the RB_ASKNAME case.
1012 	 *
1013 	 *	(b) If dumpspec is set, try to use it.  If the device
1014 	 *	    is not available, punt.
1015 	 *
1016 	 *	(c) If dumpspec is not set, the dump device is
1017 	 *	    wildcarded or unspecified.  If the root device
1018 	 *	    is DV_IFNET, punt.  Otherwise, use partition b
1019 	 *	    of the root device.
1020 	 */
1021 
1022 	if (boothowto & RB_ASKNAME) {		/* (a) */
1023 		if (dumpdv == NULL)
1024 			goto nodumpdev;
1025 	} else if (dumpspec != NULL) {		/* (b) */
1026 		if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
1027 			/*
1028 			 * Operator doesn't want a dump device.
1029 			 * Or looks like they tried to pick a network
1030 			 * device.  Oops.
1031 			 */
1032 			goto nodumpdev;
1033 		}
1034 
1035 		dumpdevname = devsw_blk2name(major(dumpdev));
1036 		if (dumpdevname == NULL)
1037 			goto nodumpdev;
1038 		memset(buf, 0, sizeof(buf));
1039 		snprintf(buf, sizeof(buf), "%s%d", dumpdevname,
1040 		    DISKUNIT(dumpdev));
1041 
1042 		dumpdv = finddevice(buf);
1043 		if (dumpdv == NULL) {
1044 			/*
1045 			 * Device not configured.
1046 			 */
1047 			goto nodumpdev;
1048 		}
1049 	} else {				/* (c) */
1050 		if (DEV_USES_PARTITIONS(rootdv) == 0)
1051 			goto nodumpdev;
1052 		else {
1053 			dumpdv = rootdv;
1054 			dumpdev = MAKEDISKDEV(major(rootdev),
1055 			    dumpdv->dv_unit, 1);
1056 		}
1057 	}
1058 
1059 	aprint_normal(" dumps on %s%c\n", dumpdv->dv_xname,
1060 	    DISKPART(dumpdev) + 'a');
1061 	return;
1062 
1063  nodumpdev:
1064 	dumpdev = NODEV;
1065 	aprint_normal("\n");
1066 }
1067 
1068 static struct device *
1069 finddevice(const char *name)
1070 {
1071 	struct device *dv;
1072 #if defined(BOOT_FROM_RAID_HOOKS) || defined(BOOT_FROM_MEMORY_HOOKS)
1073 	int j;
1074 #endif /* BOOT_FROM_RAID_HOOKS || BOOT_FROM_MEMORY_HOOKS */
1075 
1076 #ifdef BOOT_FROM_RAID_HOOKS
1077 	for (j = 0; j < numraid; j++) {
1078 		if (strcmp(name, raidrootdev[j].dv_xname) == 0) {
1079 			dv = &raidrootdev[j];
1080 			return (dv);
1081 		}
1082 	}
1083 #endif /* BOOT_FROM_RAID_HOOKS */
1084 
1085 #ifdef BOOT_FROM_MEMORY_HOOKS
1086 	for (j = 0; j < NMD; j++) {
1087 		if (strcmp(name, fakemdrootdev[j].dv_xname) == 0) {
1088 			dv = &fakemdrootdev[j];
1089 			return (dv);
1090 		}
1091 	}
1092 #endif /* BOOT_FROM_MEMORY_HOOKS */
1093 
1094 	for (dv = TAILQ_FIRST(&alldevs); dv != NULL;
1095 	    dv = TAILQ_NEXT(dv, dv_list))
1096 		if (strcmp(dv->dv_xname, name) == 0)
1097 			break;
1098 	return (dv);
1099 }
1100 
1101 static struct device *
1102 getdisk(char *str, int len, int defpart, dev_t *devp, int isdump)
1103 {
1104 	struct device	*dv;
1105 #ifdef MEMORY_DISK_HOOKS
1106 	int		i;
1107 #endif
1108 #ifdef BOOT_FROM_RAID_HOOKS
1109 	int 		j;
1110 #endif
1111 
1112 	if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
1113 		printf("use one of:");
1114 #ifdef MEMORY_DISK_HOOKS
1115 		if (isdump == 0)
1116 			for (i = 0; i < NMD; i++)
1117 				printf(" %s[a-%c]", fakemdrootdev[i].dv_xname,
1118 				    'a' + MAXPARTITIONS - 1);
1119 #endif
1120 #ifdef BOOT_FROM_RAID_HOOKS
1121 		if (isdump == 0)
1122 			for (j = 0; j < numraid; j++)
1123 				printf(" %s[a-%c]", raidrootdev[j].dv_xname,
1124 				    'a' + MAXPARTITIONS - 1);
1125 #endif
1126 		TAILQ_FOREACH(dv, &alldevs, dv_list) {
1127 			if (DEV_USES_PARTITIONS(dv))
1128 				printf(" %s[a-%c]", dv->dv_xname,
1129 				    'a' + MAXPARTITIONS - 1);
1130 			else if (dv->dv_class == DV_DISK)
1131 				printf(" %s", dv->dv_xname);
1132 			if (isdump == 0 && dv->dv_class == DV_IFNET)
1133 				printf(" %s", dv->dv_xname);
1134 		}
1135 		if (isdump)
1136 			printf(" none");
1137 #if defined(DDB)
1138 		printf(" ddb");
1139 #endif
1140 		printf(" halt reboot\n");
1141 	}
1142 	return (dv);
1143 }
1144 
1145 static struct device *
1146 parsedisk(char *str, int len, int defpart, dev_t *devp)
1147 {
1148 	struct device *dv;
1149 	char *cp, c;
1150 	int majdev, part;
1151 #ifdef MEMORY_DISK_HOOKS
1152 	int i;
1153 #endif
1154 	if (len == 0)
1155 		return (NULL);
1156 
1157 	if (len == 4 && strcmp(str, "halt") == 0)
1158 		cpu_reboot(RB_HALT, NULL);
1159 	else if (len == 6 && strcmp(str, "reboot") == 0)
1160 		cpu_reboot(0, NULL);
1161 #if defined(DDB)
1162 	else if (len == 3 && strcmp(str, "ddb") == 0)
1163 		console_debugger();
1164 #endif
1165 
1166 	cp = str + len - 1;
1167 	c = *cp;
1168 	if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
1169 		part = c - 'a';
1170 		*cp = '\0';
1171 	} else
1172 		part = defpart;
1173 
1174 #ifdef MEMORY_DISK_HOOKS
1175 	for (i = 0; i < NMD; i++)
1176 		if (strcmp(str, fakemdrootdev[i].dv_xname) == 0) {
1177 			dv = &fakemdrootdev[i];
1178 			goto gotdisk;
1179 		}
1180 #endif
1181 
1182 	dv = finddevice(str);
1183 	if (dv != NULL) {
1184 		if (dv->dv_class == DV_DISK) {
1185 #ifdef MEMORY_DISK_HOOKS
1186  gotdisk:
1187 #endif
1188 			majdev = devsw_name2blk(dv->dv_xname, NULL, 0);
1189 			if (majdev < 0)
1190 				panic("parsedisk");
1191 			if (DEV_USES_PARTITIONS(dv))
1192 				*devp = MAKEDISKDEV(majdev, dv->dv_unit, part);
1193 			else
1194 				*devp = makedev(majdev, dv->dv_unit);
1195 		}
1196 
1197 		if (dv->dv_class == DV_IFNET)
1198 			*devp = NODEV;
1199 	}
1200 
1201 	*cp = c;
1202 	return (dv);
1203 }
1204 
1205 /*
1206  * snprintf() `bytes' into `buf', reformatting it so that the number,
1207  * plus a possible `x' + suffix extension) fits into len bytes (including
1208  * the terminating NUL).
1209  * Returns the number of bytes stored in buf, or -1 if there was a problem.
1210  * E.g, given a len of 9 and a suffix of `B':
1211  *	bytes		result
1212  *	-----		------
1213  *	99999		`99999 B'
1214  *	100000		`97 kB'
1215  *	66715648	`65152 kB'
1216  *	252215296	`240 MB'
1217  */
1218 int
1219 humanize_number(char *buf, size_t len, uint64_t bytes, const char *suffix,
1220     int divisor)
1221 {
1222        	/* prefixes are: (none), kilo, Mega, Giga, Tera, Peta, Exa */
1223 	const char *prefixes;
1224 	int		r;
1225 	u_int64_t	umax;
1226 	size_t		i, suffixlen;
1227 
1228 	if (buf == NULL || suffix == NULL)
1229 		return (-1);
1230 	if (len > 0)
1231 		buf[0] = '\0';
1232 	suffixlen = strlen(suffix);
1233 	/* check if enough room for `x y' + suffix + `\0' */
1234 	if (len < 4 + suffixlen)
1235 		return (-1);
1236 
1237 	if (divisor == 1024) {
1238 		/*
1239 		 * binary multiplies
1240 		 * XXX IEC 60027-2 recommends Ki, Mi, Gi...
1241 		 */
1242 		prefixes = " KMGTPE";
1243 	} else
1244 		prefixes = " kMGTPE"; /* SI for decimal multiplies */
1245 
1246 	umax = 1;
1247 	for (i = 0; i < len - suffixlen - 3; i++)
1248 		umax *= 10;
1249 	for (i = 0; bytes >= umax && prefixes[i + 1]; i++)
1250 		bytes /= divisor;
1251 
1252 	r = snprintf(buf, len, "%qu%s%c%s", (unsigned long long)bytes,
1253 	    i == 0 ? "" : " ", prefixes[i], suffix);
1254 
1255 	return (r);
1256 }
1257 
1258 int
1259 format_bytes(char *buf, size_t len, uint64_t bytes)
1260 {
1261 	int	rv;
1262 	size_t	nlen;
1263 
1264 	rv = humanize_number(buf, len, bytes, "B", 1024);
1265 	if (rv != -1) {
1266 			/* nuke the trailing ` B' if it exists */
1267 		nlen = strlen(buf) - 2;
1268 		if (strcmp(&buf[nlen], " B") == 0)
1269 			buf[nlen] = '\0';
1270 	}
1271 	return (rv);
1272 }
1273 
1274 /*
1275  * Start trace of particular system call. If process is being traced,
1276  * this routine is called by MD syscall dispatch code just before
1277  * a system call is actually executed.
1278  * MD caller guarantees the passed 'code' is within the supported
1279  * system call number range for emulation the process runs under.
1280  */
1281 int
1282 trace_enter(struct lwp *l, register_t code,
1283     register_t realcode, const struct sysent *callp, void *args)
1284 {
1285 #if defined(KTRACE) || defined(SYSTRACE)
1286 	struct proc *p = l->l_proc;
1287 #endif
1288 
1289 #ifdef SYSCALL_DEBUG
1290 	scdebug_call(l, code, args);
1291 #endif /* SYSCALL_DEBUG */
1292 
1293 #ifdef KTRACE
1294 	if (KTRPOINT(p, KTR_SYSCALL))
1295 		ktrsyscall(p, code, realcode, callp, args);
1296 #endif /* KTRACE */
1297 
1298 #ifdef SYSTRACE
1299 	if (ISSET(p->p_flag, P_SYSTRACE))
1300 		return systrace_enter(p, code, args);
1301 #endif
1302 	return 0;
1303 }
1304 
1305 /*
1306  * End trace of particular system call. If process is being traced,
1307  * this routine is called by MD syscall dispatch code just after
1308  * a system call finishes.
1309  * MD caller guarantees the passed 'code' is within the supported
1310  * system call number range for emulation the process runs under.
1311  */
1312 void
1313 trace_exit(struct lwp *l, register_t code, void *args, register_t rval[],
1314     int error)
1315 {
1316 #if defined(KTRACE) || defined(SYSTRACE)
1317 	struct proc *p = l->l_proc;
1318 #endif
1319 
1320 #ifdef SYSCALL_DEBUG
1321 	scdebug_ret(l, code, error, rval);
1322 #endif /* SYSCALL_DEBUG */
1323 
1324 #ifdef KTRACE
1325 	if (KTRPOINT(p, KTR_SYSRET)) {
1326 		KERNEL_PROC_LOCK(l);
1327 		ktrsysret(p, code, error, rval);
1328 		KERNEL_PROC_UNLOCK(l);
1329 	}
1330 #endif /* KTRACE */
1331 
1332 #ifdef SYSTRACE
1333 	if (ISSET(p->p_flag, P_SYSTRACE))
1334 		systrace_exit(p, code, args, rval, error);
1335 #endif
1336 }
1337