xref: /netbsd-src/sys/kern/kern_subr.c (revision 2de962bd804263c16657f586aa00f1704045df8e)
1 /*	$NetBSD: kern_subr.c,v 1.187 2008/05/02 13:02:31 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 1997, 1998, 1999, 2002, 2007, 2008 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  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  * (c) UNIX System Laboratories, Inc.
37  * All or some portions of this file are derived from material licensed
38  * to the University of California by American Telephone and Telegraph
39  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
40  * the permission of UNIX System Laboratories, Inc.
41  *
42  * Copyright (c) 1992, 1993
43  *	The Regents of the University of California.  All rights reserved.
44  *
45  * This software was developed by the Computer Systems Engineering group
46  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
47  * contributed to Berkeley.
48  *
49  * All advertising materials mentioning features or use of this software
50  * must display the following acknowledgement:
51  *	This product includes software developed by the University of
52  *	California, Lawrence Berkeley Laboratory.
53  *
54  * Redistribution and use in source and binary forms, with or without
55  * modification, are permitted provided that the following conditions
56  * are met:
57  * 1. Redistributions of source code must retain the above copyright
58  *    notice, this list of conditions and the following disclaimer.
59  * 2. Redistributions in binary form must reproduce the above copyright
60  *    notice, this list of conditions and the following disclaimer in the
61  *    documentation and/or other materials provided with the distribution.
62  * 3. Neither the name of the University nor the names of its contributors
63  *    may be used to endorse or promote products derived from this software
64  *    without specific prior written permission.
65  *
66  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
67  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
68  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
69  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
70  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
71  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
72  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
73  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
74  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
75  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
76  * SUCH DAMAGE.
77  *
78  *	@(#)kern_subr.c	8.4 (Berkeley) 2/14/95
79  */
80 
81 #include <sys/cdefs.h>
82 __KERNEL_RCSID(0, "$NetBSD: kern_subr.c,v 1.187 2008/05/02 13:02:31 ad Exp $");
83 
84 #include "opt_ddb.h"
85 #include "opt_md.h"
86 #include "opt_syscall_debug.h"
87 #include "opt_ktrace.h"
88 #include "opt_ptrace.h"
89 #include "opt_powerhook.h"
90 #include "opt_tftproot.h"
91 
92 #include <sys/param.h>
93 #include <sys/systm.h>
94 #include <sys/proc.h>
95 #include <sys/malloc.h>
96 #include <sys/mount.h>
97 #include <sys/device.h>
98 #include <sys/reboot.h>
99 #include <sys/conf.h>
100 #include <sys/disk.h>
101 #include <sys/disklabel.h>
102 #include <sys/queue.h>
103 #include <sys/ktrace.h>
104 #include <sys/ptrace.h>
105 #include <sys/fcntl.h>
106 #include <sys/kauth.h>
107 #include <sys/vnode.h>
108 #include <sys/pmf.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 static const char *getwedgename(const char *, int);
121 
122 /*
123  * A generic linear hook.
124  */
125 struct hook_desc {
126 	LIST_ENTRY(hook_desc) hk_list;
127 	void	(*hk_fn)(void *);
128 	void	*hk_arg;
129 };
130 typedef LIST_HEAD(, hook_desc) hook_list_t;
131 
132 MALLOC_DEFINE(M_IOV, "iov", "large iov's");
133 
134 #ifdef TFTPROOT
135 int tftproot_dhcpboot(struct device *);
136 #endif
137 
138 dev_t	dumpcdev;	/* for savecore */
139 
140 void
141 uio_setup_sysspace(struct uio *uio)
142 {
143 
144 	uio->uio_vmspace = vmspace_kernel();
145 }
146 
147 int
148 uiomove(void *buf, size_t n, struct uio *uio)
149 {
150 	struct vmspace *vm = uio->uio_vmspace;
151 	struct iovec *iov;
152 	size_t cnt;
153 	int error = 0;
154 	char *cp = buf;
155 
156 	ASSERT_SLEEPABLE();
157 
158 #ifdef DIAGNOSTIC
159 	if (uio->uio_rw != UIO_READ && uio->uio_rw != UIO_WRITE)
160 		panic("uiomove: mode");
161 #endif
162 	while (n > 0 && uio->uio_resid) {
163 		iov = uio->uio_iov;
164 		cnt = iov->iov_len;
165 		if (cnt == 0) {
166 			KASSERT(uio->uio_iovcnt > 0);
167 			uio->uio_iov++;
168 			uio->uio_iovcnt--;
169 			continue;
170 		}
171 		if (cnt > n)
172 			cnt = n;
173 		if (!VMSPACE_IS_KERNEL_P(vm)) {
174 			if (curcpu()->ci_schedstate.spc_flags &
175 			    SPCF_SHOULDYIELD)
176 				preempt();
177 		}
178 
179 		if (uio->uio_rw == UIO_READ) {
180 			error = copyout_vmspace(vm, cp, iov->iov_base,
181 			    cnt);
182 		} else {
183 			error = copyin_vmspace(vm, iov->iov_base, cp,
184 			    cnt);
185 		}
186 		if (error) {
187 			break;
188 		}
189 		iov->iov_base = (char *)iov->iov_base + cnt;
190 		iov->iov_len -= cnt;
191 		uio->uio_resid -= cnt;
192 		uio->uio_offset += cnt;
193 		cp += cnt;
194 		KDASSERT(cnt <= n);
195 		n -= cnt;
196 	}
197 
198 	return (error);
199 }
200 
201 /*
202  * Wrapper for uiomove() that validates the arguments against a known-good
203  * kernel buffer.
204  */
205 int
206 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
207 {
208 	size_t offset;
209 
210 	if (uio->uio_offset < 0 || /* uio->uio_resid < 0 || */
211 	    (offset = uio->uio_offset) != uio->uio_offset)
212 		return (EINVAL);
213 	if (offset >= buflen)
214 		return (0);
215 	return (uiomove((char *)buf + offset, buflen - offset, uio));
216 }
217 
218 /*
219  * Give next character to user as result of read.
220  */
221 int
222 ureadc(int c, struct uio *uio)
223 {
224 	struct iovec *iov;
225 
226 	if (uio->uio_resid <= 0)
227 		panic("ureadc: non-positive resid");
228 again:
229 	if (uio->uio_iovcnt <= 0)
230 		panic("ureadc: non-positive iovcnt");
231 	iov = uio->uio_iov;
232 	if (iov->iov_len <= 0) {
233 		uio->uio_iovcnt--;
234 		uio->uio_iov++;
235 		goto again;
236 	}
237 	if (!VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) {
238 		if (subyte(iov->iov_base, c) < 0)
239 			return (EFAULT);
240 	} else {
241 		*(char *)iov->iov_base = c;
242 	}
243 	iov->iov_base = (char *)iov->iov_base + 1;
244 	iov->iov_len--;
245 	uio->uio_resid--;
246 	uio->uio_offset++;
247 	return (0);
248 }
249 
250 /*
251  * Like copyin(), but operates on an arbitrary vmspace.
252  */
253 int
254 copyin_vmspace(struct vmspace *vm, const void *uaddr, void *kaddr, size_t len)
255 {
256 	struct iovec iov;
257 	struct uio uio;
258 	int error;
259 
260 	if (len == 0)
261 		return (0);
262 
263 	if (VMSPACE_IS_KERNEL_P(vm)) {
264 		return kcopy(uaddr, kaddr, len);
265 	}
266 	if (__predict_true(vm == curproc->p_vmspace)) {
267 		return copyin(uaddr, kaddr, len);
268 	}
269 
270 	iov.iov_base = kaddr;
271 	iov.iov_len = len;
272 	uio.uio_iov = &iov;
273 	uio.uio_iovcnt = 1;
274 	uio.uio_offset = (off_t)(intptr_t)uaddr;
275 	uio.uio_resid = len;
276 	uio.uio_rw = UIO_READ;
277 	UIO_SETUP_SYSSPACE(&uio);
278 	error = uvm_io(&vm->vm_map, &uio);
279 
280 	return (error);
281 }
282 
283 /*
284  * Like copyout(), but operates on an arbitrary vmspace.
285  */
286 int
287 copyout_vmspace(struct vmspace *vm, const void *kaddr, void *uaddr, size_t len)
288 {
289 	struct iovec iov;
290 	struct uio uio;
291 	int error;
292 
293 	if (len == 0)
294 		return (0);
295 
296 	if (VMSPACE_IS_KERNEL_P(vm)) {
297 		return kcopy(kaddr, uaddr, len);
298 	}
299 	if (__predict_true(vm == curproc->p_vmspace)) {
300 		return copyout(kaddr, uaddr, len);
301 	}
302 
303 	iov.iov_base = __UNCONST(kaddr); /* XXXUNCONST cast away const */
304 	iov.iov_len = len;
305 	uio.uio_iov = &iov;
306 	uio.uio_iovcnt = 1;
307 	uio.uio_offset = (off_t)(intptr_t)uaddr;
308 	uio.uio_resid = len;
309 	uio.uio_rw = UIO_WRITE;
310 	UIO_SETUP_SYSSPACE(&uio);
311 	error = uvm_io(&vm->vm_map, &uio);
312 
313 	return (error);
314 }
315 
316 /*
317  * Like copyin(), but operates on an arbitrary process.
318  */
319 int
320 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
321 {
322 	struct vmspace *vm;
323 	int error;
324 
325 	error = proc_vmspace_getref(p, &vm);
326 	if (error) {
327 		return error;
328 	}
329 	error = copyin_vmspace(vm, uaddr, kaddr, len);
330 	uvmspace_free(vm);
331 
332 	return error;
333 }
334 
335 /*
336  * Like copyout(), but operates on an arbitrary process.
337  */
338 int
339 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
340 {
341 	struct vmspace *vm;
342 	int error;
343 
344 	error = proc_vmspace_getref(p, &vm);
345 	if (error) {
346 		return error;
347 	}
348 	error = copyout_vmspace(vm, kaddr, uaddr, len);
349 	uvmspace_free(vm);
350 
351 	return error;
352 }
353 
354 /*
355  * Like copyin(), except it operates on kernel addresses when the FKIOCTL
356  * flag is passed in `ioctlflags' from the ioctl call.
357  */
358 int
359 ioctl_copyin(int ioctlflags, const void *src, void *dst, size_t len)
360 {
361 	if (ioctlflags & FKIOCTL)
362 		return kcopy(src, dst, len);
363 	return copyin(src, dst, len);
364 }
365 
366 /*
367  * Like copyout(), except it operates on kernel addresses when the FKIOCTL
368  * flag is passed in `ioctlflags' from the ioctl call.
369  */
370 int
371 ioctl_copyout(int ioctlflags, const void *src, void *dst, size_t len)
372 {
373 	if (ioctlflags & FKIOCTL)
374 		return kcopy(src, dst, len);
375 	return copyout(src, dst, len);
376 }
377 
378 static void *
379 hook_establish(hook_list_t *list, void (*fn)(void *), void *arg)
380 {
381 	struct hook_desc *hd;
382 
383 	hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
384 	if (hd == NULL)
385 		return (NULL);
386 
387 	hd->hk_fn = fn;
388 	hd->hk_arg = arg;
389 	LIST_INSERT_HEAD(list, hd, hk_list);
390 
391 	return (hd);
392 }
393 
394 static void
395 hook_disestablish(hook_list_t *list, void *vhook)
396 {
397 #ifdef DIAGNOSTIC
398 	struct hook_desc *hd;
399 
400 	LIST_FOREACH(hd, list, hk_list) {
401                 if (hd == vhook)
402 			break;
403 	}
404 
405 	if (hd == NULL)
406 		panic("hook_disestablish: hook %p not established", vhook);
407 #endif
408 	LIST_REMOVE((struct hook_desc *)vhook, hk_list);
409 	free(vhook, M_DEVBUF);
410 }
411 
412 static void
413 hook_destroy(hook_list_t *list)
414 {
415 	struct hook_desc *hd;
416 
417 	while ((hd = LIST_FIRST(list)) != NULL) {
418 		LIST_REMOVE(hd, hk_list);
419 		free(hd, M_DEVBUF);
420 	}
421 }
422 
423 static void
424 hook_proc_run(hook_list_t *list, struct proc *p)
425 {
426 	struct hook_desc *hd;
427 
428 	LIST_FOREACH(hd, list, hk_list)
429 		((void (*)(struct proc *, void *))*hd->hk_fn)(p, hd->hk_arg);
430 }
431 
432 /*
433  * "Shutdown hook" types, functions, and variables.
434  *
435  * Should be invoked immediately before the
436  * system is halted or rebooted, i.e. after file systems unmounted,
437  * after crash dump done, etc.
438  *
439  * Each shutdown hook is removed from the list before it's run, so that
440  * it won't be run again.
441  */
442 
443 static hook_list_t shutdownhook_list;
444 
445 void *
446 shutdownhook_establish(void (*fn)(void *), void *arg)
447 {
448 	return hook_establish(&shutdownhook_list, fn, arg);
449 }
450 
451 void
452 shutdownhook_disestablish(void *vhook)
453 {
454 	hook_disestablish(&shutdownhook_list, vhook);
455 }
456 
457 /*
458  * Run shutdown hooks.  Should be invoked immediately before the
459  * system is halted or rebooted, i.e. after file systems unmounted,
460  * after crash dump done, etc.
461  *
462  * Each shutdown hook is removed from the list before it's run, so that
463  * it won't be run again.
464  */
465 void
466 doshutdownhooks(void)
467 {
468 	struct hook_desc *dp;
469 
470 	if (panicstr != NULL) {
471 		/*
472 		 * Do as few things as possible after a panic.
473 		 * We don't know the state the system is in.
474 		 */
475 		return;
476 	}
477 
478 	while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
479 		LIST_REMOVE(dp, hk_list);
480 		(*dp->hk_fn)(dp->hk_arg);
481 #if 0
482 		/*
483 		 * Don't bother freeing the hook structure,, since we may
484 		 * be rebooting because of a memory corruption problem,
485 		 * and this might only make things worse.  It doesn't
486 		 * matter, anyway, since the system is just about to
487 		 * reboot.
488 		 */
489 		free(dp, M_DEVBUF);
490 #endif
491 	}
492 
493 	pmf_system_shutdown(boothowto);
494 }
495 
496 /*
497  * "Mountroot hook" types, functions, and variables.
498  */
499 
500 static hook_list_t mountroothook_list;
501 
502 void *
503 mountroothook_establish(void (*fn)(struct device *), struct device *dev)
504 {
505 	return hook_establish(&mountroothook_list, (void (*)(void *))fn, dev);
506 }
507 
508 void
509 mountroothook_disestablish(void *vhook)
510 {
511 	hook_disestablish(&mountroothook_list, vhook);
512 }
513 
514 void
515 mountroothook_destroy(void)
516 {
517 	hook_destroy(&mountroothook_list);
518 }
519 
520 void
521 domountroothook(void)
522 {
523 	struct hook_desc *hd;
524 
525 	LIST_FOREACH(hd, &mountroothook_list, hk_list) {
526 		if (hd->hk_arg == (void *)root_device) {
527 			(*hd->hk_fn)(hd->hk_arg);
528 			return;
529 		}
530 	}
531 }
532 
533 static hook_list_t exechook_list;
534 
535 void *
536 exechook_establish(void (*fn)(struct proc *, void *), void *arg)
537 {
538 	return hook_establish(&exechook_list, (void (*)(void *))fn, arg);
539 }
540 
541 void
542 exechook_disestablish(void *vhook)
543 {
544 	hook_disestablish(&exechook_list, vhook);
545 }
546 
547 /*
548  * Run exec hooks.
549  */
550 void
551 doexechooks(struct proc *p)
552 {
553 	hook_proc_run(&exechook_list, p);
554 }
555 
556 static hook_list_t exithook_list;
557 
558 void *
559 exithook_establish(void (*fn)(struct proc *, void *), void *arg)
560 {
561 	return hook_establish(&exithook_list, (void (*)(void *))fn, arg);
562 }
563 
564 void
565 exithook_disestablish(void *vhook)
566 {
567 	hook_disestablish(&exithook_list, vhook);
568 }
569 
570 /*
571  * Run exit hooks.
572  */
573 void
574 doexithooks(struct proc *p)
575 {
576 	hook_proc_run(&exithook_list, p);
577 }
578 
579 static hook_list_t forkhook_list;
580 
581 void *
582 forkhook_establish(void (*fn)(struct proc *, struct proc *))
583 {
584 	return hook_establish(&forkhook_list, (void (*)(void *))fn, NULL);
585 }
586 
587 void
588 forkhook_disestablish(void *vhook)
589 {
590 	hook_disestablish(&forkhook_list, vhook);
591 }
592 
593 /*
594  * Run fork hooks.
595  */
596 void
597 doforkhooks(struct proc *p2, struct proc *p1)
598 {
599 	struct hook_desc *hd;
600 
601 	LIST_FOREACH(hd, &forkhook_list, hk_list) {
602 		((void (*)(struct proc *, struct proc *))*hd->hk_fn)
603 		    (p2, p1);
604 	}
605 }
606 
607 /*
608  * "Power hook" types, functions, and variables.
609  * The list of power hooks is kept ordered with the last registered hook
610  * first.
611  * When running the hooks on power down the hooks are called in reverse
612  * registration order, when powering up in registration order.
613  */
614 struct powerhook_desc {
615 	CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
616 	void	(*sfd_fn)(int, void *);
617 	void	*sfd_arg;
618 	char	sfd_name[16];
619 };
620 
621 static CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
622     CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
623 
624 void *
625 powerhook_establish(const char *name, void (*fn)(int, void *), void *arg)
626 {
627 	struct powerhook_desc *ndp;
628 
629 	ndp = (struct powerhook_desc *)
630 	    malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
631 	if (ndp == NULL)
632 		return (NULL);
633 
634 	ndp->sfd_fn = fn;
635 	ndp->sfd_arg = arg;
636 	strlcpy(ndp->sfd_name, name, sizeof(ndp->sfd_name));
637 	CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
638 
639 	aprint_error("%s: WARNING: powerhook_establish is deprecated\n", name);
640 	return (ndp);
641 }
642 
643 void
644 powerhook_disestablish(void *vhook)
645 {
646 #ifdef DIAGNOSTIC
647 	struct powerhook_desc *dp;
648 
649 	CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
650                 if (dp == vhook)
651 			goto found;
652 	panic("powerhook_disestablish: hook %p not established", vhook);
653  found:
654 #endif
655 
656 	CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
657 	    sfd_list);
658 	free(vhook, M_DEVBUF);
659 }
660 
661 /*
662  * Run power hooks.
663  */
664 void
665 dopowerhooks(int why)
666 {
667 	struct powerhook_desc *dp;
668 
669 #ifdef POWERHOOK_DEBUG
670 	const char *why_name;
671 	static const char * pwr_names[] = {PWR_NAMES};
672 	why_name = why < __arraycount(pwr_names) ? pwr_names[why] : "???";
673 #endif
674 
675 	if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
676 		CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
677 #ifdef POWERHOOK_DEBUG
678 			printf("dopowerhooks %s: %s (%p)\n", why_name, dp->sfd_name, dp);
679 #endif
680 			(*dp->sfd_fn)(why, dp->sfd_arg);
681 		}
682 	} else {
683 		CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
684 #ifdef POWERHOOK_DEBUG
685 			printf("dopowerhooks %s: %s (%p)\n", why_name, dp->sfd_name, dp);
686 #endif
687 			(*dp->sfd_fn)(why, dp->sfd_arg);
688 		}
689 	}
690 
691 #ifdef POWERHOOK_DEBUG
692 	printf("dopowerhooks: %s done\n", why_name);
693 #endif
694 }
695 
696 static int
697 isswap(struct device *dv)
698 {
699 	struct dkwedge_info wi;
700 	struct vnode *vn;
701 	int error;
702 
703 	if (device_class(dv) != DV_DISK || !device_is_a(dv, "dk"))
704 		return 0;
705 
706 	if ((vn = opendisk(dv)) == NULL)
707 		return 0;
708 
709 	error = VOP_IOCTL(vn, DIOCGWEDGEINFO, &wi, FREAD, NOCRED);
710 	VOP_CLOSE(vn, FREAD, NOCRED);
711 	vput(vn);
712 	if (error) {
713 #ifdef DEBUG_WEDGE
714 		printf("%s: Get wedge info returned %d\n", device_xname(dv), error);
715 #endif
716 		return 0;
717 	}
718 	return strcmp(wi.dkw_ptype, DKW_PTYPE_SWAP) == 0;
719 }
720 
721 /*
722  * Determine the root device and, if instructed to, the root file system.
723  */
724 
725 #include "md.h"
726 
727 #if NMD > 0
728 static struct device fakemdrootdev[NMD];
729 extern struct cfdriver md_cd;
730 #endif
731 
732 #ifdef MEMORY_DISK_IS_ROOT
733 int md_is_root = 1;
734 #else
735 int md_is_root = 0;
736 #endif
737 
738 /*
739  * The device and wedge that we booted from.  If booted_wedge is NULL,
740  * the we might consult booted_partition.
741  */
742 struct device *booted_device;
743 struct device *booted_wedge;
744 int booted_partition;
745 
746 /*
747  * Use partition letters if it's a disk class but not a wedge.
748  * XXX Check for wedge is kinda gross.
749  */
750 #define	DEV_USES_PARTITIONS(dv)						\
751 	(device_class((dv)) == DV_DISK &&				\
752 	 !device_is_a((dv), "dk"))
753 
754 void
755 setroot(struct device *bootdv, int bootpartition)
756 {
757 	struct device *dv;
758 	int len, majdev;
759 	dev_t nrootdev;
760 	dev_t ndumpdev = NODEV;
761 	char buf[128];
762 	const char *rootdevname;
763 	const char *dumpdevname;
764 	struct device *rootdv = NULL;		/* XXX gcc -Wuninitialized */
765 	struct device *dumpdv = NULL;
766 	struct ifnet *ifp;
767 	const char *deffsname;
768 	struct vfsops *vops;
769 
770 #ifdef TFTPROOT
771 	if (tftproot_dhcpboot(bootdv) != 0)
772 		boothowto |= RB_ASKNAME;
773 #endif
774 
775 #if NMD > 0
776 	if (md_is_root) {
777 		int i;
778 		for (i = 0; i < NMD; i++) {
779 			fakemdrootdev[i].dv_class  = DV_DISK;
780 			fakemdrootdev[i].dv_cfdata = NULL;
781 			fakemdrootdev[i].dv_cfdriver = &md_cd;
782 			fakemdrootdev[i].dv_unit   = i;
783 			fakemdrootdev[i].dv_parent = NULL;
784 			snprintf(fakemdrootdev[i].dv_xname,
785 			    sizeof(fakemdrootdev[i].dv_xname), "md%d", i);
786 		}
787 		bootdv = &fakemdrootdev[0];
788 		bootpartition = 0;
789 	}
790 #endif
791 
792 	/*
793 	 * If NFS is specified as the file system, and we found
794 	 * a DV_DISK boot device (or no boot device at all), then
795 	 * find a reasonable network interface for "rootspec".
796 	 */
797 	vops = vfs_getopsbyname("nfs");
798 	if (vops != NULL && vops->vfs_mountroot == mountroot &&
799 	    rootspec == NULL &&
800 	    (bootdv == NULL || device_class(bootdv) != DV_IFNET)) {
801 		IFNET_FOREACH(ifp) {
802 			if ((ifp->if_flags &
803 			     (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
804 				break;
805 		}
806 		if (ifp == NULL) {
807 			/*
808 			 * Can't find a suitable interface; ask the
809 			 * user.
810 			 */
811 			boothowto |= RB_ASKNAME;
812 		} else {
813 			/*
814 			 * Have a suitable interface; behave as if
815 			 * the user specified this interface.
816 			 */
817 			rootspec = (const char *)ifp->if_xname;
818 		}
819 	}
820 	if (vops != NULL)
821 		vfs_delref(vops);
822 
823 	/*
824 	 * If wildcarded root and we the boot device wasn't determined,
825 	 * ask the user.
826 	 */
827 	if (rootspec == NULL && bootdv == NULL)
828 		boothowto |= RB_ASKNAME;
829 
830  top:
831 	if (boothowto & RB_ASKNAME) {
832 		struct device *defdumpdv;
833 
834 		for (;;) {
835 			printf("root device");
836 			if (bootdv != NULL) {
837 				printf(" (default %s", device_xname(bootdv));
838 				if (DEV_USES_PARTITIONS(bootdv))
839 					printf("%c", bootpartition + 'a');
840 				printf(")");
841 			}
842 			printf(": ");
843 			len = cngetsn(buf, sizeof(buf));
844 			if (len == 0 && bootdv != NULL) {
845 				strlcpy(buf, device_xname(bootdv), sizeof(buf));
846 				len = strlen(buf);
847 			}
848 			if (len > 0 && buf[len - 1] == '*') {
849 				buf[--len] = '\0';
850 				dv = getdisk(buf, len, 1, &nrootdev, 0);
851 				if (dv != NULL) {
852 					rootdv = dv;
853 					break;
854 				}
855 			}
856 			dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
857 			if (dv != NULL) {
858 				rootdv = dv;
859 				break;
860 			}
861 		}
862 
863 		/*
864 		 * Set up the default dump device.  If root is on
865 		 * a network device, there is no default dump
866 		 * device, since we don't support dumps to the
867 		 * network.
868 		 */
869 		if (DEV_USES_PARTITIONS(rootdv) == 0)
870 			defdumpdv = NULL;
871 		else
872 			defdumpdv = rootdv;
873 
874 		for (;;) {
875 			printf("dump device");
876 			if (defdumpdv != NULL) {
877 				/*
878 				 * Note, we know it's a disk if we get here.
879 				 */
880 				printf(" (default %sb)", device_xname(defdumpdv));
881 			}
882 			printf(": ");
883 			len = cngetsn(buf, sizeof(buf));
884 			if (len == 0) {
885 				if (defdumpdv != NULL) {
886 					ndumpdev = MAKEDISKDEV(major(nrootdev),
887 					    DISKUNIT(nrootdev), 1);
888 				}
889 				dumpdv = defdumpdv;
890 				break;
891 			}
892 			if (len == 4 && strcmp(buf, "none") == 0) {
893 				dumpdv = NULL;
894 				break;
895 			}
896 			dv = getdisk(buf, len, 1, &ndumpdev, 1);
897 			if (dv != NULL) {
898 				dumpdv = dv;
899 				break;
900 			}
901 		}
902 
903 		rootdev = nrootdev;
904 		dumpdev = ndumpdev;
905 
906 		for (vops = LIST_FIRST(&vfs_list); vops != NULL;
907 		     vops = LIST_NEXT(vops, vfs_list)) {
908 			if (vops->vfs_mountroot != NULL &&
909 			    vops->vfs_mountroot == mountroot)
910 			break;
911 		}
912 
913 		if (vops == NULL) {
914 			mountroot = NULL;
915 			deffsname = "generic";
916 		} else
917 			deffsname = vops->vfs_name;
918 
919 		for (;;) {
920 			printf("file system (default %s): ", deffsname);
921 			len = cngetsn(buf, sizeof(buf));
922 			if (len == 0)
923 				break;
924 			if (len == 4 && strcmp(buf, "halt") == 0)
925 				cpu_reboot(RB_HALT, NULL);
926 			else if (len == 6 && strcmp(buf, "reboot") == 0)
927 				cpu_reboot(0, NULL);
928 #if defined(DDB)
929 			else if (len == 3 && strcmp(buf, "ddb") == 0) {
930 				console_debugger();
931 			}
932 #endif
933 			else if (len == 7 && strcmp(buf, "generic") == 0) {
934 				mountroot = NULL;
935 				break;
936 			}
937 			vops = vfs_getopsbyname(buf);
938 			if (vops == NULL || vops->vfs_mountroot == NULL) {
939 				printf("use one of: generic");
940 				for (vops = LIST_FIRST(&vfs_list);
941 				     vops != NULL;
942 				     vops = LIST_NEXT(vops, vfs_list)) {
943 					if (vops->vfs_mountroot != NULL)
944 						printf(" %s", vops->vfs_name);
945 				}
946 #if defined(DDB)
947 				printf(" ddb");
948 #endif
949 				printf(" halt reboot\n");
950 			} else {
951 				mountroot = vops->vfs_mountroot;
952 				vfs_delref(vops);
953 				break;
954 			}
955 		}
956 
957 	} else if (rootspec == NULL) {
958 		/*
959 		 * Wildcarded root; use the boot device.
960 		 */
961 		rootdv = bootdv;
962 
963 		majdev = devsw_name2blk(device_xname(bootdv), NULL, 0);
964 		if (majdev >= 0) {
965 			/*
966 			 * Root is on a disk.  `bootpartition' is root,
967 			 * unless the device does not use partitions.
968 			 */
969 			if (DEV_USES_PARTITIONS(bootdv))
970 				rootdev = MAKEDISKDEV(majdev,
971 						      device_unit(bootdv),
972 						      bootpartition);
973 			else
974 				rootdev = makedev(majdev, device_unit(bootdv));
975 		}
976 	} else {
977 
978 		/*
979 		 * `root on <dev> ...'
980 		 */
981 
982 		/*
983 		 * If it's a network interface, we can bail out
984 		 * early.
985 		 */
986 		dv = finddevice(rootspec);
987 		if (dv != NULL && device_class(dv) == DV_IFNET) {
988 			rootdv = dv;
989 			goto haveroot;
990 		}
991 
992 		if (rootdev == NODEV &&
993 		    device_class(dv) == DV_DISK && device_is_a(dv, "dk") &&
994 		    (majdev = devsw_name2blk(device_xname(dv), NULL, 0)) >= 0)
995 			rootdev = makedev(majdev, device_unit(dv));
996 
997 		rootdevname = devsw_blk2name(major(rootdev));
998 		if (rootdevname == NULL) {
999 			printf("unknown device major 0x%x\n", rootdev);
1000 			boothowto |= RB_ASKNAME;
1001 			goto top;
1002 		}
1003 		memset(buf, 0, sizeof(buf));
1004 		snprintf(buf, sizeof(buf), "%s%d", rootdevname,
1005 		    DISKUNIT(rootdev));
1006 
1007 		rootdv = finddevice(buf);
1008 		if (rootdv == NULL) {
1009 			printf("device %s (0x%x) not configured\n",
1010 			    buf, rootdev);
1011 			boothowto |= RB_ASKNAME;
1012 			goto top;
1013 		}
1014 	}
1015 
1016  haveroot:
1017 
1018 	root_device = rootdv;
1019 
1020 	switch (device_class(rootdv)) {
1021 	case DV_IFNET:
1022 	case DV_DISK:
1023 		aprint_normal("root on %s", device_xname(rootdv));
1024 		if (DEV_USES_PARTITIONS(rootdv))
1025 			aprint_normal("%c", DISKPART(rootdev) + 'a');
1026 		break;
1027 
1028 	default:
1029 		printf("can't determine root device\n");
1030 		boothowto |= RB_ASKNAME;
1031 		goto top;
1032 	}
1033 
1034 	/*
1035 	 * Now configure the dump device.
1036 	 *
1037 	 * If we haven't figured out the dump device, do so, with
1038 	 * the following rules:
1039 	 *
1040 	 *	(a) We already know dumpdv in the RB_ASKNAME case.
1041 	 *
1042 	 *	(b) If dumpspec is set, try to use it.  If the device
1043 	 *	    is not available, punt.
1044 	 *
1045 	 *	(c) If dumpspec is not set, the dump device is
1046 	 *	    wildcarded or unspecified.  If the root device
1047 	 *	    is DV_IFNET, punt.  Otherwise, use partition b
1048 	 *	    of the root device.
1049 	 */
1050 
1051 	if (boothowto & RB_ASKNAME) {		/* (a) */
1052 		if (dumpdv == NULL)
1053 			goto nodumpdev;
1054 	} else if (dumpspec != NULL) {		/* (b) */
1055 		if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
1056 			/*
1057 			 * Operator doesn't want a dump device.
1058 			 * Or looks like they tried to pick a network
1059 			 * device.  Oops.
1060 			 */
1061 			goto nodumpdev;
1062 		}
1063 
1064 		dumpdevname = devsw_blk2name(major(dumpdev));
1065 		if (dumpdevname == NULL)
1066 			goto nodumpdev;
1067 		memset(buf, 0, sizeof(buf));
1068 		snprintf(buf, sizeof(buf), "%s%d", dumpdevname,
1069 		    DISKUNIT(dumpdev));
1070 
1071 		dumpdv = finddevice(buf);
1072 		if (dumpdv == NULL) {
1073 			/*
1074 			 * Device not configured.
1075 			 */
1076 			goto nodumpdev;
1077 		}
1078 	} else {				/* (c) */
1079 		if (DEV_USES_PARTITIONS(rootdv) == 0) {
1080 			for (dv = TAILQ_FIRST(&alldevs); dv != NULL;
1081 			    dv = TAILQ_NEXT(dv, dv_list))
1082 				if (isswap(dv))
1083 					break;
1084 			if (dv == NULL)
1085 				goto nodumpdev;
1086 
1087 			majdev = devsw_name2blk(device_xname(dv), NULL, 0);
1088 			if (majdev < 0)
1089 				goto nodumpdev;
1090 			dumpdv = dv;
1091 			dumpdev = makedev(majdev, device_unit(dumpdv));
1092 		} else {
1093 			dumpdv = rootdv;
1094 			dumpdev = MAKEDISKDEV(major(rootdev),
1095 			    device_unit(dumpdv), 1);
1096 		}
1097 	}
1098 
1099 	dumpcdev = devsw_blk2chr(dumpdev);
1100 	aprint_normal(" dumps on %s", device_xname(dumpdv));
1101 	if (DEV_USES_PARTITIONS(dumpdv))
1102 		aprint_normal("%c", DISKPART(dumpdev) + 'a');
1103 	aprint_normal("\n");
1104 	return;
1105 
1106  nodumpdev:
1107 	dumpdev = NODEV;
1108 	dumpcdev = NODEV;
1109 	aprint_normal("\n");
1110 }
1111 
1112 static struct device *
1113 finddevice(const char *name)
1114 {
1115 	const char *wname;
1116 
1117 	if ((wname = getwedgename(name, strlen(name))) != NULL)
1118 		return dkwedge_find_by_wname(wname);
1119 
1120 #if NMD > 0
1121 	if (md_is_root) {
1122 		int j;
1123 		for (j = 0; j < NMD; j++) {
1124 			if (strcmp(name, fakemdrootdev[j].dv_xname) == 0)
1125 				return &fakemdrootdev[j];
1126 		}
1127 	}
1128 #endif
1129 
1130 	return device_find_by_xname(name);
1131 }
1132 
1133 static struct device *
1134 getdisk(char *str, int len, int defpart, dev_t *devp, int isdump)
1135 {
1136 	struct device	*dv;
1137 
1138 	if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
1139 		printf("use one of:");
1140 #if NMD > 0
1141 		if (isdump == 0 && md_is_root) {
1142 			int i;
1143 			for (i = 0; i < NMD; i++)
1144 				printf(" %s[a-%c]", fakemdrootdev[i].dv_xname,
1145 				    'a' + MAXPARTITIONS - 1);
1146 		}
1147 #endif
1148 		TAILQ_FOREACH(dv, &alldevs, dv_list) {
1149 			if (DEV_USES_PARTITIONS(dv))
1150 				printf(" %s[a-%c]", device_xname(dv),
1151 				    'a' + MAXPARTITIONS - 1);
1152 			else if (device_class(dv) == DV_DISK)
1153 				printf(" %s", device_xname(dv));
1154 			if (isdump == 0 && device_class(dv) == DV_IFNET)
1155 				printf(" %s", device_xname(dv));
1156 		}
1157 		dkwedge_print_wnames();
1158 		if (isdump)
1159 			printf(" none");
1160 #if defined(DDB)
1161 		printf(" ddb");
1162 #endif
1163 		printf(" halt reboot\n");
1164 	}
1165 	return dv;
1166 }
1167 
1168 static const char *
1169 getwedgename(const char *name, int namelen)
1170 {
1171 	const char *wpfx = "wedge:";
1172 	const int wpfxlen = strlen(wpfx);
1173 
1174 	if (namelen < wpfxlen || strncmp(name, wpfx, wpfxlen) != 0)
1175 		return NULL;
1176 
1177 	return name + wpfxlen;
1178 }
1179 
1180 static struct device *
1181 parsedisk(char *str, int len, int defpart, dev_t *devp)
1182 {
1183 	struct device *dv;
1184 	const char *wname;
1185 	char *cp, c;
1186 	int majdev, part;
1187 	if (len == 0)
1188 		return (NULL);
1189 
1190 	if (len == 4 && strcmp(str, "halt") == 0)
1191 		cpu_reboot(RB_HALT, NULL);
1192 	else if (len == 6 && strcmp(str, "reboot") == 0)
1193 		cpu_reboot(0, NULL);
1194 #if defined(DDB)
1195 	else if (len == 3 && strcmp(str, "ddb") == 0)
1196 		console_debugger();
1197 #endif
1198 
1199 	cp = str + len - 1;
1200 	c = *cp;
1201 
1202 	if ((wname = getwedgename(str, len)) != NULL) {
1203 		if ((dv = dkwedge_find_by_wname(wname)) == NULL)
1204 			return NULL;
1205 		part = defpart;
1206 		goto gotdisk;
1207 	} else if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
1208 		part = c - 'a';
1209 		*cp = '\0';
1210 	} else
1211 		part = defpart;
1212 
1213 #if NMD > 0
1214 	if (md_is_root) {
1215 		int i;
1216 		for (i = 0; i < NMD; i++) {
1217 			if (strcmp(str, fakemdrootdev[i].dv_xname) == 0) {
1218 				dv = &fakemdrootdev[i];
1219 				goto gotdisk;
1220 			}
1221 		}
1222 	}
1223 #endif
1224 
1225 	dv = finddevice(str);
1226 	if (dv != NULL) {
1227 		if (device_class(dv) == DV_DISK) {
1228  gotdisk:
1229 			majdev = devsw_name2blk(device_xname(dv), NULL, 0);
1230 			if (majdev < 0)
1231 				panic("parsedisk");
1232 			if (DEV_USES_PARTITIONS(dv))
1233 				*devp = MAKEDISKDEV(majdev, device_unit(dv),
1234 						    part);
1235 			else
1236 				*devp = makedev(majdev, device_unit(dv));
1237 		}
1238 
1239 		if (device_class(dv) == DV_IFNET)
1240 			*devp = NODEV;
1241 	}
1242 
1243 	*cp = c;
1244 	return (dv);
1245 }
1246 
1247 /*
1248  * snprintf() `bytes' into `buf', reformatting it so that the number,
1249  * plus a possible `x' + suffix extension) fits into len bytes (including
1250  * the terminating NUL).
1251  * Returns the number of bytes stored in buf, or -1 if there was a problem.
1252  * E.g, given a len of 9 and a suffix of `B':
1253  *	bytes		result
1254  *	-----		------
1255  *	99999		`99999 B'
1256  *	100000		`97 kB'
1257  *	66715648	`65152 kB'
1258  *	252215296	`240 MB'
1259  */
1260 int
1261 humanize_number(char *buf, size_t len, uint64_t bytes, const char *suffix,
1262     int divisor)
1263 {
1264        	/* prefixes are: (none), kilo, Mega, Giga, Tera, Peta, Exa */
1265 	const char *prefixes;
1266 	int		r;
1267 	uint64_t	umax;
1268 	size_t		i, suffixlen;
1269 
1270 	if (buf == NULL || suffix == NULL)
1271 		return (-1);
1272 	if (len > 0)
1273 		buf[0] = '\0';
1274 	suffixlen = strlen(suffix);
1275 	/* check if enough room for `x y' + suffix + `\0' */
1276 	if (len < 4 + suffixlen)
1277 		return (-1);
1278 
1279 	if (divisor == 1024) {
1280 		/*
1281 		 * binary multiplies
1282 		 * XXX IEC 60027-2 recommends Ki, Mi, Gi...
1283 		 */
1284 		prefixes = " KMGTPE";
1285 	} else
1286 		prefixes = " kMGTPE"; /* SI for decimal multiplies */
1287 
1288 	umax = 1;
1289 	for (i = 0; i < len - suffixlen - 3; i++)
1290 		umax *= 10;
1291 	for (i = 0; bytes >= umax && prefixes[i + 1]; i++)
1292 		bytes /= divisor;
1293 
1294 	r = snprintf(buf, len, "%qu%s%c%s", (unsigned long long)bytes,
1295 	    i == 0 ? "" : " ", prefixes[i], suffix);
1296 
1297 	return (r);
1298 }
1299 
1300 int
1301 format_bytes(char *buf, size_t len, uint64_t bytes)
1302 {
1303 	int	rv;
1304 	size_t	nlen;
1305 
1306 	rv = humanize_number(buf, len, bytes, "B", 1024);
1307 	if (rv != -1) {
1308 			/* nuke the trailing ` B' if it exists */
1309 		nlen = strlen(buf) - 2;
1310 		if (strcmp(&buf[nlen], " B") == 0)
1311 			buf[nlen] = '\0';
1312 	}
1313 	return (rv);
1314 }
1315 
1316 /*
1317  * Return true if system call tracing is enabled for the specified process.
1318  */
1319 bool
1320 trace_is_enabled(struct proc *p)
1321 {
1322 #ifdef SYSCALL_DEBUG
1323 	return (true);
1324 #endif
1325 #ifdef KTRACE
1326 	if (ISSET(p->p_traceflag, (KTRFAC_SYSCALL | KTRFAC_SYSRET)))
1327 		return (true);
1328 #endif
1329 #ifdef PTRACE
1330 	if (ISSET(p->p_slflag, PSL_SYSCALL))
1331 		return (true);
1332 #endif
1333 
1334 	return (false);
1335 }
1336 
1337 /*
1338  * Start trace of particular system call. If process is being traced,
1339  * this routine is called by MD syscall dispatch code just before
1340  * a system call is actually executed.
1341  */
1342 int
1343 trace_enter(register_t code, const register_t *args, int narg)
1344 {
1345 #ifdef SYSCALL_DEBUG
1346 	scdebug_call(code, args);
1347 #endif /* SYSCALL_DEBUG */
1348 
1349 	ktrsyscall(code, args, narg);
1350 
1351 #ifdef PTRACE
1352 	if ((curlwp->l_proc->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
1353 	    (PSL_SYSCALL|PSL_TRACED))
1354 		process_stoptrace();
1355 #endif
1356 	return 0;
1357 }
1358 
1359 /*
1360  * End trace of particular system call. If process is being traced,
1361  * this routine is called by MD syscall dispatch code just after
1362  * a system call finishes.
1363  * MD caller guarantees the passed 'code' is within the supported
1364  * system call number range for emulation the process runs under.
1365  */
1366 void
1367 trace_exit(register_t code, register_t rval[], int error)
1368 {
1369 #ifdef SYSCALL_DEBUG
1370 	scdebug_ret(code, error, rval);
1371 #endif /* SYSCALL_DEBUG */
1372 
1373 	ktrsysret(code, error, rval);
1374 
1375 #ifdef PTRACE
1376 	if ((curlwp->l_proc->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
1377 	    (PSL_SYSCALL|PSL_TRACED))
1378 		process_stoptrace();
1379 #endif
1380 }
1381