xref: /netbsd-src/sys/kern/kern_subr.c (revision d20841bb642898112fe68f0ad3f7b26dddf56f07)
1 /*	$NetBSD: kern_subr.c,v 1.107 2003/10/31 03:28:14 simonb 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 1.107 2003/10/31 03:28:14 simonb 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 			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  * Give next character to user as result of read.
210  */
211 int
212 ureadc(c, uio)
213 	int c;
214 	struct uio *uio;
215 {
216 	struct iovec *iov;
217 
218 	if (uio->uio_resid <= 0)
219 		panic("ureadc: non-positive resid");
220 again:
221 	if (uio->uio_iovcnt <= 0)
222 		panic("ureadc: non-positive iovcnt");
223 	iov = uio->uio_iov;
224 	if (iov->iov_len <= 0) {
225 		uio->uio_iovcnt--;
226 		uio->uio_iov++;
227 		goto again;
228 	}
229 	switch (uio->uio_segflg) {
230 
231 	case UIO_USERSPACE:
232 		if (subyte(iov->iov_base, c) < 0)
233 			return (EFAULT);
234 		break;
235 
236 	case UIO_SYSSPACE:
237 		*(char *)iov->iov_base = c;
238 		break;
239 	}
240 	iov->iov_base = (caddr_t)iov->iov_base + 1;
241 	iov->iov_len--;
242 	uio->uio_resid--;
243 	uio->uio_offset++;
244 	return (0);
245 }
246 
247 /*
248  * Like copyin(), but operates on an arbitrary process.
249  */
250 int
251 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
252 {
253 	struct iovec iov;
254 	struct uio uio;
255 	int error;
256 
257 	if (len == 0)
258 		return (0);
259 
260 	iov.iov_base = kaddr;
261 	iov.iov_len = len;
262 	uio.uio_iov = &iov;
263 	uio.uio_iovcnt = 1;
264 	uio.uio_offset = (off_t)(intptr_t)uaddr;
265 	uio.uio_resid = len;
266 	uio.uio_segflg = UIO_SYSSPACE;
267 	uio.uio_rw = UIO_READ;
268 	uio.uio_procp = NULL;
269 
270 	/* XXXCDC: how should locking work here? */
271 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
272 		return (EFAULT);
273 	p->p_vmspace->vm_refcnt++;	/* XXX */
274 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
275 	uvmspace_free(p->p_vmspace);
276 
277 	return (error);
278 }
279 
280 /*
281  * Like copyout(), but operates on an arbitrary process.
282  */
283 int
284 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
285 {
286 	struct iovec iov;
287 	struct uio uio;
288 	int error;
289 
290 	if (len == 0)
291 		return (0);
292 
293 	iov.iov_base = (void *) kaddr;	/* XXX cast away const */
294 	iov.iov_len = len;
295 	uio.uio_iov = &iov;
296 	uio.uio_iovcnt = 1;
297 	uio.uio_offset = (off_t)(intptr_t)uaddr;
298 	uio.uio_resid = len;
299 	uio.uio_segflg = UIO_SYSSPACE;
300 	uio.uio_rw = UIO_WRITE;
301 	uio.uio_procp = NULL;
302 
303 	/* XXXCDC: how should locking work here? */
304 	if ((p->p_flag & P_WEXIT) || (p->p_vmspace->vm_refcnt < 1))
305 		return (EFAULT);
306 	p->p_vmspace->vm_refcnt++;	/* XXX */
307 	error = uvm_io(&p->p_vmspace->vm_map, &uio);
308 	uvmspace_free(p->p_vmspace);
309 
310 	return (error);
311 }
312 
313 /*
314  * General routine to allocate a hash table.
315  * Allocate enough memory to hold at least `elements' list-head pointers.
316  * Return a pointer to the allocated space and set *hashmask to a pattern
317  * suitable for masking a value to use as an index into the returned array.
318  */
319 void *
320 hashinit(elements, htype, mtype, mflags, hashmask)
321 	u_int elements;
322 	enum hashtype htype;
323 	struct malloc_type *mtype;
324 	int mflags;
325 	u_long *hashmask;
326 {
327 	u_long hashsize, i;
328 	LIST_HEAD(, generic) *hashtbl_list;
329 	TAILQ_HEAD(, generic) *hashtbl_tailq;
330 	size_t esize;
331 	void *p;
332 
333 	if (elements == 0)
334 		panic("hashinit: bad cnt");
335 	for (hashsize = 1; hashsize < elements; hashsize <<= 1)
336 		continue;
337 
338 	switch (htype) {
339 	case HASH_LIST:
340 		esize = sizeof(*hashtbl_list);
341 		break;
342 	case HASH_TAILQ:
343 		esize = sizeof(*hashtbl_tailq);
344 		break;
345 	default:
346 #ifdef DIAGNOSTIC
347 		panic("hashinit: invalid table type");
348 #else
349 		return NULL;
350 #endif
351 	}
352 
353 	if ((p = malloc(hashsize * esize, mtype, mflags)) == NULL)
354 		return (NULL);
355 
356 	switch (htype) {
357 	case HASH_LIST:
358 		hashtbl_list = p;
359 		for (i = 0; i < hashsize; i++)
360 			LIST_INIT(&hashtbl_list[i]);
361 		break;
362 	case HASH_TAILQ:
363 		hashtbl_tailq = p;
364 		for (i = 0; i < hashsize; i++)
365 			TAILQ_INIT(&hashtbl_tailq[i]);
366 		break;
367 	}
368 	*hashmask = hashsize - 1;
369 	return (p);
370 }
371 
372 /*
373  * Free memory from hash table previosly allocated via hashinit().
374  */
375 void
376 hashdone(hashtbl, mtype)
377 	void *hashtbl;
378 	struct malloc_type *mtype;
379 {
380 
381 	free(hashtbl, mtype);
382 }
383 
384 
385 static void *
386 hook_establish(list, fn, arg)
387 	hook_list_t *list;
388 	void (*fn) __P((void *));
389 	void *arg;
390 {
391 	struct hook_desc *hd;
392 
393 	hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
394 	if (hd == NULL)
395 		return (NULL);
396 
397 	hd->hk_fn = fn;
398 	hd->hk_arg = arg;
399 	LIST_INSERT_HEAD(list, hd, hk_list);
400 
401 	return (hd);
402 }
403 
404 static void
405 hook_disestablish(list, vhook)
406 	hook_list_t *list;
407 	void *vhook;
408 {
409 #ifdef DIAGNOSTIC
410 	struct hook_desc *hd;
411 
412 	LIST_FOREACH(hd, list, hk_list) {
413                 if (hd == vhook)
414 			break;
415 	}
416 
417 	if (hd == NULL)
418 		panic("hook_disestablish: hook %p not established", vhook);
419 #endif
420 	LIST_REMOVE((struct hook_desc *)vhook, hk_list);
421 	free(vhook, M_DEVBUF);
422 }
423 
424 static void
425 hook_destroy(list)
426 	hook_list_t *list;
427 {
428 	struct hook_desc *hd;
429 
430 	while ((hd = LIST_FIRST(list)) != NULL) {
431 		LIST_REMOVE(hd, hk_list);
432 		free(hd, M_DEVBUF);
433 	}
434 }
435 
436 static void
437 hook_proc_run(list, p)
438 	hook_list_t *list;
439 	struct proc *p;
440 {
441 	struct hook_desc *hd;
442 
443 	for (hd = LIST_FIRST(list); hd != NULL; hd = LIST_NEXT(hd, hk_list)) {
444 		((void (*) __P((struct proc *, void *)))*hd->hk_fn)(p,
445 		    hd->hk_arg);
446 	}
447 }
448 
449 /*
450  * "Shutdown hook" types, functions, and variables.
451  *
452  * Should be invoked immediately before the
453  * system is halted or rebooted, i.e. after file systems unmounted,
454  * after crash dump done, etc.
455  *
456  * Each shutdown hook is removed from the list before it's run, so that
457  * it won't be run again.
458  */
459 
460 hook_list_t shutdownhook_list;
461 
462 void *
463 shutdownhook_establish(fn, arg)
464 	void (*fn) __P((void *));
465 	void *arg;
466 {
467 	return hook_establish(&shutdownhook_list, fn, arg);
468 }
469 
470 void
471 shutdownhook_disestablish(vhook)
472 	void *vhook;
473 {
474 	hook_disestablish(&shutdownhook_list, vhook);
475 }
476 
477 /*
478  * Run shutdown hooks.  Should be invoked immediately before the
479  * system is halted or rebooted, i.e. after file systems unmounted,
480  * after crash dump done, etc.
481  *
482  * Each shutdown hook is removed from the list before it's run, so that
483  * it won't be run again.
484  */
485 void
486 doshutdownhooks()
487 {
488 	struct hook_desc *dp;
489 
490 	while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
491 		LIST_REMOVE(dp, hk_list);
492 		(*dp->hk_fn)(dp->hk_arg);
493 #if 0
494 		/*
495 		 * Don't bother freeing the hook structure,, since we may
496 		 * be rebooting because of a memory corruption problem,
497 		 * and this might only make things worse.  It doesn't
498 		 * matter, anyway, since the system is just about to
499 		 * reboot.
500 		 */
501 		free(dp, M_DEVBUF);
502 #endif
503 	}
504 }
505 
506 /*
507  * "Mountroot hook" types, functions, and variables.
508  */
509 
510 hook_list_t mountroothook_list;
511 
512 void *
513 mountroothook_establish(fn, dev)
514 	void (*fn) __P((struct device *));
515 	struct device *dev;
516 {
517 	return hook_establish(&mountroothook_list, (void (*)__P((void *)))fn,
518 	    dev);
519 }
520 
521 void
522 mountroothook_disestablish(vhook)
523 	void *vhook;
524 {
525 	hook_disestablish(&mountroothook_list, vhook);
526 }
527 
528 void
529 mountroothook_destroy()
530 {
531 	hook_destroy(&mountroothook_list);
532 }
533 
534 void
535 domountroothook()
536 {
537 	struct hook_desc *hd;
538 
539 	LIST_FOREACH(hd, &mountroothook_list, hk_list) {
540 		if (hd->hk_arg == (void *)root_device) {
541 			(*hd->hk_fn)(hd->hk_arg);
542 			return;
543 		}
544 	}
545 }
546 
547 hook_list_t exechook_list;
548 
549 void *
550 exechook_establish(fn, arg)
551 	void (*fn) __P((struct proc *, void *));
552 	void *arg;
553 {
554 	return hook_establish(&exechook_list, (void (*) __P((void *)))fn, arg);
555 }
556 
557 void
558 exechook_disestablish(vhook)
559 	void *vhook;
560 {
561 	hook_disestablish(&exechook_list, vhook);
562 }
563 
564 /*
565  * Run exec hooks.
566  */
567 void
568 doexechooks(p)
569 	struct proc *p;
570 {
571 	hook_proc_run(&exechook_list, p);
572 }
573 
574 hook_list_t exithook_list;
575 
576 void *
577 exithook_establish(fn, arg)
578 	void (*fn) __P((struct proc *, void *));
579 	void *arg;
580 {
581 	return hook_establish(&exithook_list, (void (*) __P((void *)))fn, arg);
582 }
583 
584 void
585 exithook_disestablish(vhook)
586 	void *vhook;
587 {
588 	hook_disestablish(&exithook_list, vhook);
589 }
590 
591 /*
592  * Run exit hooks.
593  */
594 void
595 doexithooks(p)
596 	struct proc *p;
597 {
598 	hook_proc_run(&exithook_list, p);
599 }
600 
601 hook_list_t forkhook_list;
602 
603 void *
604 forkhook_establish(fn)
605 	void (*fn) __P((struct proc *, struct proc *));
606 {
607 	return hook_establish(&forkhook_list, (void (*) __P((void *)))fn, NULL);
608 }
609 
610 void
611 forkhook_disestablish(vhook)
612 	void *vhook;
613 {
614 	hook_disestablish(&forkhook_list, vhook);
615 }
616 
617 /*
618  * Run fork hooks.
619  */
620 void
621 doforkhooks(p2, p1)
622 	struct proc *p2, *p1;
623 {
624 	struct hook_desc *hd;
625 
626 	LIST_FOREACH(hd, &forkhook_list, hk_list) {
627 		((void (*) __P((struct proc *, struct proc *)))*hd->hk_fn)
628 		    (p2, p1);
629 	}
630 }
631 
632 /*
633  * "Power hook" types, functions, and variables.
634  * The list of power hooks is kept ordered with the last registered hook
635  * first.
636  * When running the hooks on power down the hooks are called in reverse
637  * registration order, when powering up in registration order.
638  */
639 struct powerhook_desc {
640 	CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
641 	void	(*sfd_fn) __P((int, void *));
642 	void	*sfd_arg;
643 };
644 
645 CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
646 	CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
647 
648 void *
649 powerhook_establish(fn, arg)
650 	void (*fn) __P((int, void *));
651 	void *arg;
652 {
653 	struct powerhook_desc *ndp;
654 
655 	ndp = (struct powerhook_desc *)
656 	    malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
657 	if (ndp == NULL)
658 		return (NULL);
659 
660 	ndp->sfd_fn = fn;
661 	ndp->sfd_arg = arg;
662 	CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
663 
664 	return (ndp);
665 }
666 
667 void
668 powerhook_disestablish(vhook)
669 	void *vhook;
670 {
671 #ifdef DIAGNOSTIC
672 	struct powerhook_desc *dp;
673 
674 	CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
675                 if (dp == vhook)
676 			goto found;
677 	panic("powerhook_disestablish: hook %p not established", vhook);
678  found:
679 #endif
680 
681 	CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
682 	    sfd_list);
683 	free(vhook, M_DEVBUF);
684 }
685 
686 /*
687  * Run power hooks.
688  */
689 void
690 dopowerhooks(why)
691 	int why;
692 {
693 	struct powerhook_desc *dp;
694 
695 	if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
696 		CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
697 			(*dp->sfd_fn)(why, dp->sfd_arg);
698 		}
699 	} else {
700 		CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
701 			(*dp->sfd_fn)(why, dp->sfd_arg);
702 		}
703 	}
704 }
705 
706 /*
707  * Determine the root device and, if instructed to, the root file system.
708  */
709 
710 #include "md.h"
711 #if NMD == 0
712 #undef MEMORY_DISK_HOOKS
713 #endif
714 
715 #ifdef MEMORY_DISK_HOOKS
716 static struct device fakemdrootdev[NMD];
717 #endif
718 
719 #include "raid.h"
720 #if NRAID == 1
721 #define BOOT_FROM_RAID_HOOKS 1
722 #endif
723 
724 #ifdef BOOT_FROM_RAID_HOOKS
725 extern int numraid;
726 extern struct device *raidrootdev;
727 #endif
728 
729 void
730 setroot(bootdv, bootpartition)
731 	struct device *bootdv;
732 	int bootpartition;
733 {
734 	struct device *dv;
735 	int len;
736 #ifdef MEMORY_DISK_HOOKS
737 	int i;
738 #endif
739 	dev_t nrootdev;
740 	dev_t ndumpdev = NODEV;
741 	char buf[128];
742 	const char *rootdevname;
743 	const char *dumpdevname;
744 	struct device *rootdv = NULL;		/* XXX gcc -Wuninitialized */
745 	struct device *dumpdv = NULL;
746 	struct ifnet *ifp;
747 	const char *deffsname;
748 	struct vfsops *vops;
749 
750 #ifdef MEMORY_DISK_HOOKS
751 	for (i = 0; i < NMD; i++) {
752 		fakemdrootdev[i].dv_class  = DV_DISK;
753 		fakemdrootdev[i].dv_cfdata = NULL;
754 		fakemdrootdev[i].dv_unit   = i;
755 		fakemdrootdev[i].dv_parent = NULL;
756 		sprintf(fakemdrootdev[i].dv_xname, "md%d", i);
757 	}
758 #endif /* MEMORY_DISK_HOOKS */
759 
760 #ifdef MEMORY_DISK_IS_ROOT
761 	bootdv = &fakemdrootdev[0];
762 	bootpartition = 0;
763 #endif
764 
765 	/*
766 	 * If NFS is specified as the file system, and we found
767 	 * a DV_DISK boot device (or no boot device at all), then
768 	 * find a reasonable network interface for "rootspec".
769 	 */
770 	vops = vfs_getopsbyname("nfs");
771 	if (vops != NULL && vops->vfs_mountroot == mountroot &&
772 	    rootspec == NULL &&
773 	    (bootdv == NULL || bootdv->dv_class != DV_IFNET)) {
774 		TAILQ_FOREACH(ifp, &ifnet, if_list) {
775 			if ((ifp->if_flags &
776 			     (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
777 				break;
778 		}
779 		if (ifp == NULL) {
780 			/*
781 			 * Can't find a suitable interface; ask the
782 			 * user.
783 			 */
784 			boothowto |= RB_ASKNAME;
785 		} else {
786 			/*
787 			 * Have a suitable interface; behave as if
788 			 * the user specified this interface.
789 			 */
790 			rootspec = (const char *)ifp->if_xname;
791 		}
792 	}
793 
794 	/*
795 	 * If wildcarded root and we the boot device wasn't determined,
796 	 * ask the user.
797 	 */
798 	if (rootspec == NULL && bootdv == NULL)
799 		boothowto |= RB_ASKNAME;
800 
801  top:
802 	if (boothowto & RB_ASKNAME) {
803 		struct device *defdumpdv;
804 
805 		for (;;) {
806 			printf("root device");
807 			if (bootdv != NULL) {
808 				printf(" (default %s", bootdv->dv_xname);
809 				if (bootdv->dv_class == DV_DISK)
810 					printf("%c", bootpartition + 'a');
811 				printf(")");
812 			}
813 			printf(": ");
814 			len = cngetsn(buf, sizeof(buf));
815 			if (len == 0 && bootdv != NULL) {
816 				strlcpy(buf, bootdv->dv_xname, sizeof(buf));
817 				len = strlen(buf);
818 			}
819 			if (len > 0 && buf[len - 1] == '*') {
820 				buf[--len] = '\0';
821 				dv = getdisk(buf, len, 1, &nrootdev, 0);
822 				if (dv != NULL) {
823 					rootdv = dv;
824 					break;
825 				}
826 			}
827 			dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
828 			if (dv != NULL) {
829 				rootdv = dv;
830 				break;
831 			}
832 		}
833 
834 		/*
835 		 * Set up the default dump device.  If root is on
836 		 * a network device, there is no default dump
837 		 * device, since we don't support dumps to the
838 		 * network.
839 		 */
840 		if (rootdv->dv_class == DV_IFNET)
841 			defdumpdv = NULL;
842 		else
843 			defdumpdv = rootdv;
844 
845 		for (;;) {
846 			printf("dump device");
847 			if (defdumpdv != NULL) {
848 				/*
849 				 * Note, we know it's a disk if we get here.
850 				 */
851 				printf(" (default %sb)", defdumpdv->dv_xname);
852 			}
853 			printf(": ");
854 			len = cngetsn(buf, sizeof(buf));
855 			if (len == 0) {
856 				if (defdumpdv != NULL) {
857 					ndumpdev = MAKEDISKDEV(major(nrootdev),
858 					    DISKUNIT(nrootdev), 1);
859 				}
860 				dumpdv = defdumpdv;
861 				break;
862 			}
863 			if (len == 4 && strcmp(buf, "none") == 0) {
864 				dumpdv = NULL;
865 				break;
866 			}
867 			dv = getdisk(buf, len, 1, &ndumpdev, 1);
868 			if (dv != NULL) {
869 				dumpdv = dv;
870 				break;
871 			}
872 		}
873 
874 		rootdev = nrootdev;
875 		dumpdev = ndumpdev;
876 
877 		for (vops = LIST_FIRST(&vfs_list); vops != NULL;
878 		     vops = LIST_NEXT(vops, vfs_list)) {
879 			if (vops->vfs_mountroot != NULL &&
880 			    vops->vfs_mountroot == mountroot)
881 			break;
882 		}
883 
884 		if (vops == NULL) {
885 			mountroot = NULL;
886 			deffsname = "generic";
887 		} else
888 			deffsname = vops->vfs_name;
889 
890 		for (;;) {
891 			printf("file system (default %s): ", deffsname);
892 			len = cngetsn(buf, sizeof(buf));
893 			if (len == 0)
894 				break;
895 			if (len == 4 && strcmp(buf, "halt") == 0)
896 				cpu_reboot(RB_HALT, NULL);
897 			else if (len == 6 && strcmp(buf, "reboot") == 0)
898 				cpu_reboot(0, NULL);
899 #if defined(DDB)
900 			else if (len == 3 && strcmp(buf, "ddb") == 0) {
901 				console_debugger();
902 			}
903 #endif
904 			else if (len == 7 && strcmp(buf, "generic") == 0) {
905 				mountroot = NULL;
906 				break;
907 			}
908 			vops = vfs_getopsbyname(buf);
909 			if (vops == NULL || vops->vfs_mountroot == NULL) {
910 				printf("use one of: generic");
911 				for (vops = LIST_FIRST(&vfs_list);
912 				     vops != NULL;
913 				     vops = LIST_NEXT(vops, vfs_list)) {
914 					if (vops->vfs_mountroot != NULL)
915 						printf(" %s", vops->vfs_name);
916 				}
917 #if defined(DDB)
918 				printf(" ddb");
919 #endif
920 				printf(" halt reboot\n");
921 			} else {
922 				mountroot = vops->vfs_mountroot;
923 				break;
924 			}
925 		}
926 
927 	} else if (rootspec == NULL) {
928 		int majdev;
929 
930 		/*
931 		 * Wildcarded root; use the boot device.
932 		 */
933 		rootdv = bootdv;
934 
935 		majdev = devsw_name2blk(bootdv->dv_xname, NULL, 0);
936 		if (majdev >= 0) {
937 			/*
938 			 * Root is on a disk.  `bootpartition' is root.
939 			 */
940 			rootdev = MAKEDISKDEV(majdev, bootdv->dv_unit,
941 			    bootpartition);
942 		}
943 	} else {
944 
945 		/*
946 		 * `root on <dev> ...'
947 		 */
948 
949 		/*
950 		 * If it's a network interface, we can bail out
951 		 * early.
952 		 */
953 		dv = finddevice(rootspec);
954 		if (dv != NULL && dv->dv_class == DV_IFNET) {
955 			rootdv = dv;
956 			goto haveroot;
957 		}
958 
959 		rootdevname = devsw_blk2name(major(rootdev));
960 		if (rootdevname == NULL) {
961 			printf("unknown device major 0x%x\n", rootdev);
962 			boothowto |= RB_ASKNAME;
963 			goto top;
964 		}
965 		memset(buf, 0, sizeof(buf));
966 		sprintf(buf, "%s%d", rootdevname, DISKUNIT(rootdev));
967 
968 		rootdv = finddevice(buf);
969 		if (rootdv == NULL) {
970 			printf("device %s (0x%x) not configured\n",
971 			    buf, rootdev);
972 			boothowto |= RB_ASKNAME;
973 			goto top;
974 		}
975 	}
976 
977  haveroot:
978 
979 	root_device = rootdv;
980 
981 	switch (rootdv->dv_class) {
982 	case DV_IFNET:
983 		aprint_normal("root on %s", rootdv->dv_xname);
984 		break;
985 
986 	case DV_DISK:
987 		aprint_normal("root on %s%c", rootdv->dv_xname,
988 		    DISKPART(rootdev) + 'a');
989 		break;
990 
991 	default:
992 		printf("can't determine root device\n");
993 		boothowto |= RB_ASKNAME;
994 		goto top;
995 	}
996 
997 	/*
998 	 * Now configure the dump device.
999 	 *
1000 	 * If we haven't figured out the dump device, do so, with
1001 	 * the following rules:
1002 	 *
1003 	 *	(a) We already know dumpdv in the RB_ASKNAME case.
1004 	 *
1005 	 *	(b) If dumpspec is set, try to use it.  If the device
1006 	 *	    is not available, punt.
1007 	 *
1008 	 *	(c) If dumpspec is not set, the dump device is
1009 	 *	    wildcarded or unspecified.  If the root device
1010 	 *	    is DV_IFNET, punt.  Otherwise, use partition b
1011 	 *	    of the root device.
1012 	 */
1013 
1014 	if (boothowto & RB_ASKNAME) {		/* (a) */
1015 		if (dumpdv == NULL)
1016 			goto nodumpdev;
1017 	} else if (dumpspec != NULL) {		/* (b) */
1018 		if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
1019 			/*
1020 			 * Operator doesn't want a dump device.
1021 			 * Or looks like they tried to pick a network
1022 			 * device.  Oops.
1023 			 */
1024 			goto nodumpdev;
1025 		}
1026 
1027 		dumpdevname = devsw_blk2name(major(dumpdev));
1028 		if (dumpdevname == NULL)
1029 			goto nodumpdev;
1030 		memset(buf, 0, sizeof(buf));
1031 		sprintf(buf, "%s%d", dumpdevname, DISKUNIT(dumpdev));
1032 
1033 		dumpdv = finddevice(buf);
1034 		if (dumpdv == NULL) {
1035 			/*
1036 			 * Device not configured.
1037 			 */
1038 			goto nodumpdev;
1039 		}
1040 	} else {				/* (c) */
1041 		if (rootdv->dv_class == DV_IFNET)
1042 			goto nodumpdev;
1043 		else {
1044 			dumpdv = rootdv;
1045 			dumpdev = MAKEDISKDEV(major(rootdev),
1046 			    dumpdv->dv_unit, 1);
1047 		}
1048 	}
1049 
1050 	aprint_normal(" dumps on %s%c\n", dumpdv->dv_xname,
1051 	    DISKPART(dumpdev) + 'a');
1052 	return;
1053 
1054  nodumpdev:
1055 	dumpdev = NODEV;
1056 	aprint_normal("\n");
1057 }
1058 
1059 static struct device *
1060 finddevice(name)
1061 	const char *name;
1062 {
1063 	struct device *dv;
1064 #ifdef BOOT_FROM_RAID_HOOKS
1065 	int j;
1066 
1067 	for (j = 0; j < numraid; j++) {
1068 		if (strcmp(name, raidrootdev[j].dv_xname) == 0) {
1069 			dv = &raidrootdev[j];
1070 			return (dv);
1071 		}
1072 	}
1073 #endif
1074 
1075 	for (dv = TAILQ_FIRST(&alldevs); dv != NULL;
1076 	    dv = TAILQ_NEXT(dv, dv_list))
1077 		if (strcmp(dv->dv_xname, name) == 0)
1078 			break;
1079 	return (dv);
1080 }
1081 
1082 static struct device *
1083 getdisk(str, len, defpart, devp, isdump)
1084 	char *str;
1085 	int len, defpart;
1086 	dev_t *devp;
1087 	int isdump;
1088 {
1089 	struct device	*dv;
1090 #ifdef MEMORY_DISK_HOOKS
1091 	int		i;
1092 #endif
1093 #ifdef BOOT_FROM_RAID_HOOKS
1094 	int 		j;
1095 #endif
1096 
1097 	if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
1098 		printf("use one of:");
1099 #ifdef MEMORY_DISK_HOOKS
1100 		if (isdump == 0)
1101 			for (i = 0; i < NMD; i++)
1102 				printf(" %s[a-%c]", fakemdrootdev[i].dv_xname,
1103 				    'a' + MAXPARTITIONS - 1);
1104 #endif
1105 #ifdef BOOT_FROM_RAID_HOOKS
1106 		if (isdump == 0)
1107 			for (j = 0; j < numraid; j++)
1108 				printf(" %s[a-%c]", raidrootdev[j].dv_xname,
1109 				    'a' + MAXPARTITIONS - 1);
1110 #endif
1111 		TAILQ_FOREACH(dv, &alldevs, dv_list) {
1112 			if (dv->dv_class == DV_DISK)
1113 				printf(" %s[a-%c]", dv->dv_xname,
1114 				    'a' + MAXPARTITIONS - 1);
1115 			if (isdump == 0 && dv->dv_class == DV_IFNET)
1116 				printf(" %s", dv->dv_xname);
1117 		}
1118 		if (isdump)
1119 			printf(" none");
1120 #if defined(DDB)
1121 		printf(" ddb");
1122 #endif
1123 		printf(" halt reboot\n");
1124 	}
1125 	return (dv);
1126 }
1127 
1128 static struct device *
1129 parsedisk(str, len, defpart, devp)
1130 	char *str;
1131 	int len, defpart;
1132 	dev_t *devp;
1133 {
1134 	struct device *dv;
1135 	char *cp, c;
1136 	int majdev, part;
1137 #ifdef MEMORY_DISK_HOOKS
1138 	int i;
1139 #endif
1140 	if (len == 0)
1141 		return (NULL);
1142 
1143 	if (len == 4 && strcmp(str, "halt") == 0)
1144 		cpu_reboot(RB_HALT, NULL);
1145 	else if (len == 6 && strcmp(str, "reboot") == 0)
1146 		cpu_reboot(0, NULL);
1147 #if defined(DDB)
1148 	else if (len == 3 && strcmp(str, "ddb") == 0)
1149 		console_debugger();
1150 #endif
1151 
1152 	cp = str + len - 1;
1153 	c = *cp;
1154 	if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
1155 		part = c - 'a';
1156 		*cp = '\0';
1157 	} else
1158 		part = defpart;
1159 
1160 #ifdef MEMORY_DISK_HOOKS
1161 	for (i = 0; i < NMD; i++)
1162 		if (strcmp(str, fakemdrootdev[i].dv_xname) == 0) {
1163 			dv = &fakemdrootdev[i];
1164 			goto gotdisk;
1165 		}
1166 #endif
1167 
1168 	dv = finddevice(str);
1169 	if (dv != NULL) {
1170 		if (dv->dv_class == DV_DISK) {
1171 #ifdef MEMORY_DISK_HOOKS
1172  gotdisk:
1173 #endif
1174 			majdev = devsw_name2blk(dv->dv_xname, NULL, 0);
1175 			if (majdev < 0)
1176 				panic("parsedisk");
1177 			*devp = MAKEDISKDEV(majdev, dv->dv_unit, part);
1178 		}
1179 
1180 		if (dv->dv_class == DV_IFNET)
1181 			*devp = NODEV;
1182 	}
1183 
1184 	*cp = c;
1185 	return (dv);
1186 }
1187 
1188 /*
1189  * snprintf() `bytes' into `buf', reformatting it so that the number,
1190  * plus a possible `x' + suffix extension) fits into len bytes (including
1191  * the terminating NUL).
1192  * Returns the number of bytes stored in buf, or -1 if there was a problem.
1193  * E.g, given a len of 9 and a suffix of `B':
1194  *	bytes		result
1195  *	-----		------
1196  *	99999		`99999 B'
1197  *	100000		`97 kB'
1198  *	66715648	`65152 kB'
1199  *	252215296	`240 MB'
1200  */
1201 int
1202 humanize_number(buf, len, bytes, suffix, divisor)
1203 	char		*buf;
1204 	size_t		 len;
1205 	u_int64_t	 bytes;
1206 	const char	*suffix;
1207 	int 		divisor;
1208 {
1209        	/* prefixes are: (none), kilo, Mega, Giga, Tera, Peta, Exa */
1210 	const char *prefixes;
1211 	int		r;
1212 	u_int64_t	max;
1213 	size_t		i, suffixlen;
1214 
1215 	if (buf == NULL || suffix == NULL)
1216 		return (-1);
1217 	if (len > 0)
1218 		buf[0] = '\0';
1219 	suffixlen = strlen(suffix);
1220 	/* check if enough room for `x y' + suffix + `\0' */
1221 	if (len < 4 + suffixlen)
1222 		return (-1);
1223 
1224 	if (divisor == 1024) {
1225 		/*
1226 		 * binary multiplies
1227 		 * XXX IEC 60027-2 recommends Ki, Mi, Gi...
1228 		 */
1229 		prefixes = " KMGTPE";
1230 	} else
1231 		prefixes = " kMGTPE"; /* SI for decimal multiplies */
1232 
1233 	max = 1;
1234 	for (i = 0; i < len - suffixlen - 3; i++)
1235 		max *= 10;
1236 	for (i = 0; bytes >= max && prefixes[i + 1]; i++)
1237 		bytes /= divisor;
1238 
1239 	r = snprintf(buf, len, "%qu%s%c%s", (unsigned long long)bytes,
1240 	    i == 0 ? "" : " ", prefixes[i], suffix);
1241 
1242 	return (r);
1243 }
1244 
1245 int
1246 format_bytes(buf, len, bytes)
1247 	char		*buf;
1248 	size_t		 len;
1249 	u_int64_t	 bytes;
1250 {
1251 	int	rv;
1252 	size_t	nlen;
1253 
1254 	rv = humanize_number(buf, len, bytes, "B", 1024);
1255 	if (rv != -1) {
1256 			/* nuke the trailing ` B' if it exists */
1257 		nlen = strlen(buf) - 2;
1258 		if (strcmp(&buf[nlen], " B") == 0)
1259 			buf[nlen] = '\0';
1260 	}
1261 	return (rv);
1262 }
1263 
1264 /*
1265  * Start trace of particular system call. If process is being traced,
1266  * this routine is called by MD syscall dispatch code just before
1267  * a system call is actually executed.
1268  * MD caller guarantees the passed 'code' is within the supported
1269  * system call number range for emulation the process runs under.
1270  */
1271 int
1272 trace_enter(struct lwp *l, register_t code,
1273 	register_t realcode, const struct sysent *callp, void *args)
1274 {
1275 #if defined(KTRACE) || defined(SYSTRACE)
1276 	struct proc *p = l->l_proc;
1277 #endif
1278 
1279 #ifdef SYSCALL_DEBUG
1280 	scdebug_call(l, code, args);
1281 #endif /* SYSCALL_DEBUG */
1282 
1283 #ifdef KTRACE
1284 	if (KTRPOINT(p, KTR_SYSCALL))
1285 		ktrsyscall(p, code, realcode, callp, args);
1286 #endif /* KTRACE */
1287 
1288 #ifdef SYSTRACE
1289 	if (ISSET(p->p_flag, P_SYSTRACE))
1290 		return systrace_enter(p, code, args);
1291 #endif
1292 	return 0;
1293 }
1294 
1295 /*
1296  * End trace of particular system call. If process is being traced,
1297  * this routine is called by MD syscall dispatch code just after
1298  * a system call finishes.
1299  * MD caller guarantees the passed 'code' is within the supported
1300  * system call number range for emulation the process runs under.
1301  */
1302 void
1303 trace_exit(struct lwp *l, register_t code, void *args, register_t rval[],
1304     int error)
1305 {
1306 #if defined(KTRACE) || defined(SYSTRACE)
1307 	struct proc *p = l->l_proc;
1308 #endif
1309 
1310 #ifdef SYSCALL_DEBUG
1311 	scdebug_ret(l, code, error, rval);
1312 #endif /* SYSCALL_DEBUG */
1313 
1314 #ifdef KTRACE
1315 	if (KTRPOINT(p, KTR_SYSRET)) {
1316 		KERNEL_PROC_LOCK(l);
1317 		ktrsysret(p, code, error, rval);
1318 		KERNEL_PROC_UNLOCK(l);
1319 	}
1320 #endif /* KTRACE */
1321 
1322 #ifdef SYSTRACE
1323 	if (ISSET(p->p_flag, P_SYSTRACE))
1324 		systrace_exit(p, code, args, rval, error);
1325 #endif
1326 }
1327