xref: /netbsd-src/sys/kern/kern_subr.c (revision 1ca06f9c9235889e2ff6dc77279d01d151d70a9a)
1 /*	$NetBSD: kern_subr.c,v 1.203 2009/11/05 18:07:19 dyoung 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 1.203 2009/11/05 18:07:19 dyoung 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/syscallvar.h>
109 #include <sys/xcall.h>
110 #include <sys/module.h>
111 
112 #include <uvm/uvm_extern.h>
113 
114 #include <dev/cons.h>
115 
116 #include <net/if.h>
117 
118 /* XXX these should eventually move to subr_autoconf.c */
119 static device_t finddevice(const char *);
120 static device_t getdisk(char *, int, int, dev_t *, int);
121 static device_t parsedisk(char *, int, int, dev_t *);
122 static const char *getwedgename(const char *, int);
123 
124 /*
125  * A generic linear hook.
126  */
127 struct hook_desc {
128 	LIST_ENTRY(hook_desc) hk_list;
129 	void	(*hk_fn)(void *);
130 	void	*hk_arg;
131 };
132 typedef LIST_HEAD(, hook_desc) hook_list_t;
133 
134 #ifdef TFTPROOT
135 int tftproot_dhcpboot(device_t);
136 #endif
137 
138 dev_t	dumpcdev;	/* for savecore */
139 
140 static void *
141 hook_establish(hook_list_t *list, void (*fn)(void *), void *arg)
142 {
143 	struct hook_desc *hd;
144 
145 	hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
146 	if (hd == NULL)
147 		return (NULL);
148 
149 	hd->hk_fn = fn;
150 	hd->hk_arg = arg;
151 	LIST_INSERT_HEAD(list, hd, hk_list);
152 
153 	return (hd);
154 }
155 
156 static void
157 hook_disestablish(hook_list_t *list, void *vhook)
158 {
159 #ifdef DIAGNOSTIC
160 	struct hook_desc *hd;
161 
162 	LIST_FOREACH(hd, list, hk_list) {
163                 if (hd == vhook)
164 			break;
165 	}
166 
167 	if (hd == NULL)
168 		panic("hook_disestablish: hook %p not established", vhook);
169 #endif
170 	LIST_REMOVE((struct hook_desc *)vhook, hk_list);
171 	free(vhook, M_DEVBUF);
172 }
173 
174 static void
175 hook_destroy(hook_list_t *list)
176 {
177 	struct hook_desc *hd;
178 
179 	while ((hd = LIST_FIRST(list)) != NULL) {
180 		LIST_REMOVE(hd, hk_list);
181 		free(hd, M_DEVBUF);
182 	}
183 }
184 
185 static void
186 hook_proc_run(hook_list_t *list, struct proc *p)
187 {
188 	struct hook_desc *hd;
189 
190 	LIST_FOREACH(hd, list, hk_list)
191 		((void (*)(struct proc *, void *))*hd->hk_fn)(p, hd->hk_arg);
192 }
193 
194 /*
195  * "Shutdown hook" types, functions, and variables.
196  *
197  * Should be invoked immediately before the
198  * system is halted or rebooted, i.e. after file systems unmounted,
199  * after crash dump done, etc.
200  *
201  * Each shutdown hook is removed from the list before it's run, so that
202  * it won't be run again.
203  */
204 
205 static hook_list_t shutdownhook_list;
206 
207 void *
208 shutdownhook_establish(void (*fn)(void *), void *arg)
209 {
210 	return hook_establish(&shutdownhook_list, fn, arg);
211 }
212 
213 void
214 shutdownhook_disestablish(void *vhook)
215 {
216 	hook_disestablish(&shutdownhook_list, vhook);
217 }
218 
219 /*
220  * Run shutdown hooks.  Should be invoked immediately before the
221  * system is halted or rebooted, i.e. after file systems unmounted,
222  * after crash dump done, etc.
223  *
224  * Each shutdown hook is removed from the list before it's run, so that
225  * it won't be run again.
226  */
227 void
228 doshutdownhooks(void)
229 {
230 	struct hook_desc *dp;
231 
232 	while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
233 		LIST_REMOVE(dp, hk_list);
234 		(*dp->hk_fn)(dp->hk_arg);
235 #if 0
236 		/*
237 		 * Don't bother freeing the hook structure,, since we may
238 		 * be rebooting because of a memory corruption problem,
239 		 * and this might only make things worse.  It doesn't
240 		 * matter, anyway, since the system is just about to
241 		 * reboot.
242 		 */
243 		free(dp, M_DEVBUF);
244 #endif
245 	}
246 }
247 
248 /*
249  * "Mountroot hook" types, functions, and variables.
250  */
251 
252 static hook_list_t mountroothook_list;
253 
254 void *
255 mountroothook_establish(void (*fn)(device_t), device_t dev)
256 {
257 	return hook_establish(&mountroothook_list, (void (*)(void *))fn, dev);
258 }
259 
260 void
261 mountroothook_disestablish(void *vhook)
262 {
263 	hook_disestablish(&mountroothook_list, vhook);
264 }
265 
266 void
267 mountroothook_destroy(void)
268 {
269 	hook_destroy(&mountroothook_list);
270 }
271 
272 void
273 domountroothook(void)
274 {
275 	struct hook_desc *hd;
276 
277 	LIST_FOREACH(hd, &mountroothook_list, hk_list) {
278 		if (hd->hk_arg == (void *)root_device) {
279 			(*hd->hk_fn)(hd->hk_arg);
280 			return;
281 		}
282 	}
283 }
284 
285 static hook_list_t exechook_list;
286 
287 void *
288 exechook_establish(void (*fn)(struct proc *, void *), void *arg)
289 {
290 	return hook_establish(&exechook_list, (void (*)(void *))fn, arg);
291 }
292 
293 void
294 exechook_disestablish(void *vhook)
295 {
296 	hook_disestablish(&exechook_list, vhook);
297 }
298 
299 /*
300  * Run exec hooks.
301  */
302 void
303 doexechooks(struct proc *p)
304 {
305 	hook_proc_run(&exechook_list, p);
306 }
307 
308 static hook_list_t exithook_list;
309 extern krwlock_t exec_lock;
310 
311 void *
312 exithook_establish(void (*fn)(struct proc *, void *), void *arg)
313 {
314 	void *rv;
315 
316 	rw_enter(&exec_lock, RW_WRITER);
317 	rv = hook_establish(&exithook_list, (void (*)(void *))fn, arg);
318 	rw_exit(&exec_lock);
319 	return rv;
320 }
321 
322 void
323 exithook_disestablish(void *vhook)
324 {
325 
326 	rw_enter(&exec_lock, RW_WRITER);
327 	hook_disestablish(&exithook_list, vhook);
328 	rw_exit(&exec_lock);
329 }
330 
331 /*
332  * Run exit hooks.
333  */
334 void
335 doexithooks(struct proc *p)
336 {
337 	hook_proc_run(&exithook_list, p);
338 }
339 
340 static hook_list_t forkhook_list;
341 
342 void *
343 forkhook_establish(void (*fn)(struct proc *, struct proc *))
344 {
345 	return hook_establish(&forkhook_list, (void (*)(void *))fn, NULL);
346 }
347 
348 void
349 forkhook_disestablish(void *vhook)
350 {
351 	hook_disestablish(&forkhook_list, vhook);
352 }
353 
354 /*
355  * Run fork hooks.
356  */
357 void
358 doforkhooks(struct proc *p2, struct proc *p1)
359 {
360 	struct hook_desc *hd;
361 
362 	LIST_FOREACH(hd, &forkhook_list, hk_list) {
363 		((void (*)(struct proc *, struct proc *))*hd->hk_fn)
364 		    (p2, p1);
365 	}
366 }
367 
368 /*
369  * "Power hook" types, functions, and variables.
370  * The list of power hooks is kept ordered with the last registered hook
371  * first.
372  * When running the hooks on power down the hooks are called in reverse
373  * registration order, when powering up in registration order.
374  */
375 struct powerhook_desc {
376 	CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
377 	void	(*sfd_fn)(int, void *);
378 	void	*sfd_arg;
379 	char	sfd_name[16];
380 };
381 
382 static CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
383     CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
384 
385 void *
386 powerhook_establish(const char *name, void (*fn)(int, void *), void *arg)
387 {
388 	struct powerhook_desc *ndp;
389 
390 	ndp = (struct powerhook_desc *)
391 	    malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
392 	if (ndp == NULL)
393 		return (NULL);
394 
395 	ndp->sfd_fn = fn;
396 	ndp->sfd_arg = arg;
397 	strlcpy(ndp->sfd_name, name, sizeof(ndp->sfd_name));
398 	CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
399 
400 	aprint_error("%s: WARNING: powerhook_establish is deprecated\n", name);
401 	return (ndp);
402 }
403 
404 void
405 powerhook_disestablish(void *vhook)
406 {
407 #ifdef DIAGNOSTIC
408 	struct powerhook_desc *dp;
409 
410 	CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
411                 if (dp == vhook)
412 			goto found;
413 	panic("powerhook_disestablish: hook %p not established", vhook);
414  found:
415 #endif
416 
417 	CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
418 	    sfd_list);
419 	free(vhook, M_DEVBUF);
420 }
421 
422 /*
423  * Run power hooks.
424  */
425 void
426 dopowerhooks(int why)
427 {
428 	struct powerhook_desc *dp;
429 
430 #ifdef POWERHOOK_DEBUG
431 	const char *why_name;
432 	static const char * pwr_names[] = {PWR_NAMES};
433 	why_name = why < __arraycount(pwr_names) ? pwr_names[why] : "???";
434 #endif
435 
436 	if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
437 		CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
438 #ifdef POWERHOOK_DEBUG
439 			printf("dopowerhooks %s: %s (%p)\n", why_name, dp->sfd_name, dp);
440 #endif
441 			(*dp->sfd_fn)(why, dp->sfd_arg);
442 		}
443 	} else {
444 		CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
445 #ifdef POWERHOOK_DEBUG
446 			printf("dopowerhooks %s: %s (%p)\n", why_name, dp->sfd_name, dp);
447 #endif
448 			(*dp->sfd_fn)(why, dp->sfd_arg);
449 		}
450 	}
451 
452 #ifdef POWERHOOK_DEBUG
453 	printf("dopowerhooks: %s done\n", why_name);
454 #endif
455 }
456 
457 static int
458 isswap(device_t dv)
459 {
460 	struct dkwedge_info wi;
461 	struct vnode *vn;
462 	int error;
463 
464 	if (device_class(dv) != DV_DISK || !device_is_a(dv, "dk"))
465 		return 0;
466 
467 	if ((vn = opendisk(dv)) == NULL)
468 		return 0;
469 
470 	error = VOP_IOCTL(vn, DIOCGWEDGEINFO, &wi, FREAD, NOCRED);
471 	VOP_CLOSE(vn, FREAD, NOCRED);
472 	vput(vn);
473 	if (error) {
474 #ifdef DEBUG_WEDGE
475 		printf("%s: Get wedge info returned %d\n", device_xname(dv), error);
476 #endif
477 		return 0;
478 	}
479 	return strcmp(wi.dkw_ptype, DKW_PTYPE_SWAP) == 0;
480 }
481 
482 /*
483  * Determine the root device and, if instructed to, the root file system.
484  */
485 
486 #include "md.h"
487 
488 #if NMD > 0
489 extern struct cfdriver md_cd;
490 #ifdef MEMORY_DISK_IS_ROOT
491 int md_is_root = 1;
492 #else
493 int md_is_root = 0;
494 #endif
495 #endif
496 
497 /*
498  * The device and wedge that we booted from.  If booted_wedge is NULL,
499  * the we might consult booted_partition.
500  */
501 device_t booted_device;
502 device_t booted_wedge;
503 int booted_partition;
504 
505 /*
506  * Use partition letters if it's a disk class but not a wedge.
507  * XXX Check for wedge is kinda gross.
508  */
509 #define	DEV_USES_PARTITIONS(dv)						\
510 	(device_class((dv)) == DV_DISK &&				\
511 	 !device_is_a((dv), "dk"))
512 
513 void
514 setroot(device_t bootdv, int bootpartition)
515 {
516 	device_t dv;
517 	deviter_t di;
518 	int len, majdev;
519 	dev_t nrootdev;
520 	dev_t ndumpdev = NODEV;
521 	char buf[128];
522 	const char *rootdevname;
523 	const char *dumpdevname;
524 	device_t rootdv = NULL;		/* XXX gcc -Wuninitialized */
525 	device_t dumpdv = NULL;
526 	struct ifnet *ifp;
527 	const char *deffsname;
528 	struct vfsops *vops;
529 
530 #ifdef TFTPROOT
531 	if (tftproot_dhcpboot(bootdv) != 0)
532 		boothowto |= RB_ASKNAME;
533 #endif
534 
535 #if NMD > 0
536 	if (md_is_root) {
537 		/*
538 		 * XXX there should be "root on md0" in the config file,
539 		 * but it isn't always
540 		 */
541 		bootdv = md_cd.cd_devs[0];
542 		bootpartition = 0;
543 	}
544 #endif
545 
546 	/*
547 	 * If NFS is specified as the file system, and we found
548 	 * a DV_DISK boot device (or no boot device at all), then
549 	 * find a reasonable network interface for "rootspec".
550 	 */
551 	vops = vfs_getopsbyname(MOUNT_NFS);
552 	if (vops != NULL && strcmp(rootfstype, MOUNT_NFS) == 0 &&
553 	    rootspec == NULL &&
554 	    (bootdv == NULL || device_class(bootdv) != DV_IFNET)) {
555 		IFNET_FOREACH(ifp) {
556 			if ((ifp->if_flags &
557 			     (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
558 				break;
559 		}
560 		if (ifp == NULL) {
561 			/*
562 			 * Can't find a suitable interface; ask the
563 			 * user.
564 			 */
565 			boothowto |= RB_ASKNAME;
566 		} else {
567 			/*
568 			 * Have a suitable interface; behave as if
569 			 * the user specified this interface.
570 			 */
571 			rootspec = (const char *)ifp->if_xname;
572 		}
573 	}
574 	if (vops != NULL)
575 		vfs_delref(vops);
576 
577 	/*
578 	 * If wildcarded root and we the boot device wasn't determined,
579 	 * ask the user.
580 	 */
581 	if (rootspec == NULL && bootdv == NULL)
582 		boothowto |= RB_ASKNAME;
583 
584  top:
585 	if (boothowto & RB_ASKNAME) {
586 		device_t defdumpdv;
587 
588 		for (;;) {
589 			printf("root device");
590 			if (bootdv != NULL) {
591 				printf(" (default %s", device_xname(bootdv));
592 				if (DEV_USES_PARTITIONS(bootdv))
593 					printf("%c", bootpartition + 'a');
594 				printf(")");
595 			}
596 			printf(": ");
597 			len = cngetsn(buf, sizeof(buf));
598 			if (len == 0 && bootdv != NULL) {
599 				strlcpy(buf, device_xname(bootdv), sizeof(buf));
600 				len = strlen(buf);
601 			}
602 			if (len > 0 && buf[len - 1] == '*') {
603 				buf[--len] = '\0';
604 				dv = getdisk(buf, len, 1, &nrootdev, 0);
605 				if (dv != NULL) {
606 					rootdv = dv;
607 					break;
608 				}
609 			}
610 			dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
611 			if (dv != NULL) {
612 				rootdv = dv;
613 				break;
614 			}
615 		}
616 
617 		/*
618 		 * Set up the default dump device.  If root is on
619 		 * a network device, there is no default dump
620 		 * device, since we don't support dumps to the
621 		 * network.
622 		 */
623 		if (DEV_USES_PARTITIONS(rootdv) == 0)
624 			defdumpdv = NULL;
625 		else
626 			defdumpdv = rootdv;
627 
628 		for (;;) {
629 			printf("dump device");
630 			if (defdumpdv != NULL) {
631 				/*
632 				 * Note, we know it's a disk if we get here.
633 				 */
634 				printf(" (default %sb)", device_xname(defdumpdv));
635 			}
636 			printf(": ");
637 			len = cngetsn(buf, sizeof(buf));
638 			if (len == 0) {
639 				if (defdumpdv != NULL) {
640 					ndumpdev = MAKEDISKDEV(major(nrootdev),
641 					    DISKUNIT(nrootdev), 1);
642 				}
643 				dumpdv = defdumpdv;
644 				break;
645 			}
646 			if (len == 4 && strcmp(buf, "none") == 0) {
647 				dumpdv = NULL;
648 				break;
649 			}
650 			dv = getdisk(buf, len, 1, &ndumpdev, 1);
651 			if (dv != NULL) {
652 				dumpdv = dv;
653 				break;
654 			}
655 		}
656 
657 		rootdev = nrootdev;
658 		dumpdev = ndumpdev;
659 
660 		for (vops = LIST_FIRST(&vfs_list); vops != NULL;
661 		     vops = LIST_NEXT(vops, vfs_list)) {
662 			if (vops->vfs_mountroot != NULL &&
663 			    strcmp(rootfstype, vops->vfs_name) == 0)
664 			break;
665 		}
666 
667 		if (vops == NULL) {
668 			deffsname = "generic";
669 		} else
670 			deffsname = vops->vfs_name;
671 
672 		for (;;) {
673 			printf("file system (default %s): ", deffsname);
674 			len = cngetsn(buf, sizeof(buf));
675 			if (len == 0) {
676 				if (strcmp(deffsname, "generic") == 0)
677 					rootfstype = ROOT_FSTYPE_ANY;
678 				break;
679 			}
680 			if (len == 4 && strcmp(buf, "halt") == 0)
681 				cpu_reboot(RB_HALT, NULL);
682 			else if (len == 6 && strcmp(buf, "reboot") == 0)
683 				cpu_reboot(0, NULL);
684 #if defined(DDB)
685 			else if (len == 3 && strcmp(buf, "ddb") == 0) {
686 				console_debugger();
687 			}
688 #endif
689 			else if (len == 7 && strcmp(buf, "generic") == 0) {
690 				rootfstype = ROOT_FSTYPE_ANY;
691 				break;
692 			}
693 			vops = vfs_getopsbyname(buf);
694 			if (vops == NULL || vops->vfs_mountroot == NULL) {
695 				printf("use one of: generic");
696 				for (vops = LIST_FIRST(&vfs_list);
697 				     vops != NULL;
698 				     vops = LIST_NEXT(vops, vfs_list)) {
699 					if (vops->vfs_mountroot != NULL)
700 						printf(" %s", vops->vfs_name);
701 				}
702 				if (vops != NULL)
703 					vfs_delref(vops);
704 #if defined(DDB)
705 				printf(" ddb");
706 #endif
707 				printf(" halt reboot\n");
708 			} else {
709 				/*
710 				 * XXX If *vops gets freed between here and
711 				 * the call to mountroot(), rootfstype will
712 				 * point to something unexpected.  But in
713 				 * this case the system will fail anyway.
714 				 */
715 				rootfstype = vops->vfs_name;
716 				vfs_delref(vops);
717 				break;
718 			}
719 		}
720 
721 	} else if (rootspec == NULL) {
722 		/*
723 		 * Wildcarded root; use the boot device.
724 		 */
725 		rootdv = bootdv;
726 
727 		if (bootdv)
728 			majdev = devsw_name2blk(device_xname(bootdv), NULL, 0);
729 		else
730 			majdev = -1;
731 		if (majdev >= 0) {
732 			/*
733 			 * Root is on a disk.  `bootpartition' is root,
734 			 * unless the device does not use partitions.
735 			 */
736 			if (DEV_USES_PARTITIONS(bootdv))
737 				rootdev = MAKEDISKDEV(majdev,
738 						      device_unit(bootdv),
739 						      bootpartition);
740 			else
741 				rootdev = makedev(majdev, device_unit(bootdv));
742 		}
743 	} else {
744 
745 		/*
746 		 * `root on <dev> ...'
747 		 */
748 
749 		/*
750 		 * If it's a network interface, we can bail out
751 		 * early.
752 		 */
753 		dv = finddevice(rootspec);
754 		if (dv != NULL && device_class(dv) == DV_IFNET) {
755 			rootdv = dv;
756 			goto haveroot;
757 		}
758 
759 		if (rootdev == NODEV &&
760 		    device_class(dv) == DV_DISK && device_is_a(dv, "dk") &&
761 		    (majdev = devsw_name2blk(device_xname(dv), NULL, 0)) >= 0)
762 			rootdev = makedev(majdev, device_unit(dv));
763 
764 		rootdevname = devsw_blk2name(major(rootdev));
765 		if (rootdevname == NULL) {
766 			printf("unknown device major 0x%llx\n",
767 			    (unsigned long long)rootdev);
768 			boothowto |= RB_ASKNAME;
769 			goto top;
770 		}
771 		memset(buf, 0, sizeof(buf));
772 		snprintf(buf, sizeof(buf), "%s%llu", rootdevname,
773 		    (unsigned long long)DISKUNIT(rootdev));
774 
775 		rootdv = finddevice(buf);
776 		if (rootdv == NULL) {
777 			printf("device %s (0x%llx) not configured\n",
778 			    buf, (unsigned long long)rootdev);
779 			boothowto |= RB_ASKNAME;
780 			goto top;
781 		}
782 	}
783 
784  haveroot:
785 
786 	root_device = rootdv;
787 
788 	switch (device_class(rootdv)) {
789 	case DV_IFNET:
790 	case DV_DISK:
791 		aprint_normal("root on %s", device_xname(rootdv));
792 		if (DEV_USES_PARTITIONS(rootdv))
793 			aprint_normal("%c", (int)DISKPART(rootdev) + 'a');
794 		break;
795 
796 	default:
797 		printf("can't determine root device\n");
798 		boothowto |= RB_ASKNAME;
799 		goto top;
800 	}
801 
802 	/*
803 	 * Now configure the dump device.
804 	 *
805 	 * If we haven't figured out the dump device, do so, with
806 	 * the following rules:
807 	 *
808 	 *	(a) We already know dumpdv in the RB_ASKNAME case.
809 	 *
810 	 *	(b) If dumpspec is set, try to use it.  If the device
811 	 *	    is not available, punt.
812 	 *
813 	 *	(c) If dumpspec is not set, the dump device is
814 	 *	    wildcarded or unspecified.  If the root device
815 	 *	    is DV_IFNET, punt.  Otherwise, use partition b
816 	 *	    of the root device.
817 	 */
818 
819 	if (boothowto & RB_ASKNAME) {		/* (a) */
820 		if (dumpdv == NULL)
821 			goto nodumpdev;
822 	} else if (dumpspec != NULL) {		/* (b) */
823 		if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
824 			/*
825 			 * Operator doesn't want a dump device.
826 			 * Or looks like they tried to pick a network
827 			 * device.  Oops.
828 			 */
829 			goto nodumpdev;
830 		}
831 
832 		dumpdevname = devsw_blk2name(major(dumpdev));
833 		if (dumpdevname == NULL)
834 			goto nodumpdev;
835 		memset(buf, 0, sizeof(buf));
836 		snprintf(buf, sizeof(buf), "%s%llu", dumpdevname,
837 		    (unsigned long long)DISKUNIT(dumpdev));
838 
839 		dumpdv = finddevice(buf);
840 		if (dumpdv == NULL) {
841 			/*
842 			 * Device not configured.
843 			 */
844 			goto nodumpdev;
845 		}
846 	} else {				/* (c) */
847 		if (DEV_USES_PARTITIONS(rootdv) == 0) {
848 			for (dv = deviter_first(&di, DEVITER_F_ROOT_FIRST);
849 			     dv != NULL;
850 			     dv = deviter_next(&di))
851 				if (isswap(dv))
852 					break;
853 			deviter_release(&di);
854 			if (dv == NULL)
855 				goto nodumpdev;
856 
857 			majdev = devsw_name2blk(device_xname(dv), NULL, 0);
858 			if (majdev < 0)
859 				goto nodumpdev;
860 			dumpdv = dv;
861 			dumpdev = makedev(majdev, device_unit(dumpdv));
862 		} else {
863 			dumpdv = rootdv;
864 			dumpdev = MAKEDISKDEV(major(rootdev),
865 			    device_unit(dumpdv), 1);
866 		}
867 	}
868 
869 	dumpcdev = devsw_blk2chr(dumpdev);
870 	aprint_normal(" dumps on %s", device_xname(dumpdv));
871 	if (DEV_USES_PARTITIONS(dumpdv))
872 		aprint_normal("%c", (int)DISKPART(dumpdev) + 'a');
873 	aprint_normal("\n");
874 	return;
875 
876  nodumpdev:
877 	dumpdev = NODEV;
878 	dumpcdev = NODEV;
879 	aprint_normal("\n");
880 }
881 
882 static device_t
883 finddevice(const char *name)
884 {
885 	const char *wname;
886 
887 	if ((wname = getwedgename(name, strlen(name))) != NULL)
888 		return dkwedge_find_by_wname(wname);
889 
890 	return device_find_by_xname(name);
891 }
892 
893 static device_t
894 getdisk(char *str, int len, int defpart, dev_t *devp, int isdump)
895 {
896 	device_t dv;
897 	deviter_t di;
898 
899 	if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
900 		printf("use one of:");
901 		for (dv = deviter_first(&di, DEVITER_F_ROOT_FIRST); dv != NULL;
902 		     dv = deviter_next(&di)) {
903 			if (DEV_USES_PARTITIONS(dv))
904 				printf(" %s[a-%c]", device_xname(dv),
905 				    'a' + MAXPARTITIONS - 1);
906 			else if (device_class(dv) == DV_DISK)
907 				printf(" %s", device_xname(dv));
908 			if (isdump == 0 && device_class(dv) == DV_IFNET)
909 				printf(" %s", device_xname(dv));
910 		}
911 		deviter_release(&di);
912 		dkwedge_print_wnames();
913 		if (isdump)
914 			printf(" none");
915 #if defined(DDB)
916 		printf(" ddb");
917 #endif
918 		printf(" halt reboot\n");
919 	}
920 	return dv;
921 }
922 
923 static const char *
924 getwedgename(const char *name, int namelen)
925 {
926 	const char *wpfx = "wedge:";
927 	const int wpfxlen = strlen(wpfx);
928 
929 	if (namelen < wpfxlen || strncmp(name, wpfx, wpfxlen) != 0)
930 		return NULL;
931 
932 	return name + wpfxlen;
933 }
934 
935 static device_t
936 parsedisk(char *str, int len, int defpart, dev_t *devp)
937 {
938 	device_t dv;
939 	const char *wname;
940 	char *cp, c;
941 	int majdev, part;
942 	if (len == 0)
943 		return (NULL);
944 
945 	if (len == 4 && strcmp(str, "halt") == 0)
946 		cpu_reboot(RB_HALT, NULL);
947 	else if (len == 6 && strcmp(str, "reboot") == 0)
948 		cpu_reboot(0, NULL);
949 #if defined(DDB)
950 	else if (len == 3 && strcmp(str, "ddb") == 0)
951 		console_debugger();
952 #endif
953 
954 	cp = str + len - 1;
955 	c = *cp;
956 
957 	if ((wname = getwedgename(str, len)) != NULL) {
958 		if ((dv = dkwedge_find_by_wname(wname)) == NULL)
959 			return NULL;
960 		part = defpart;
961 		goto gotdisk;
962 	} else if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
963 		part = c - 'a';
964 		*cp = '\0';
965 	} else
966 		part = defpart;
967 
968 	dv = finddevice(str);
969 	if (dv != NULL) {
970 		if (device_class(dv) == DV_DISK) {
971  gotdisk:
972 			majdev = devsw_name2blk(device_xname(dv), NULL, 0);
973 			if (majdev < 0)
974 				panic("parsedisk");
975 			if (DEV_USES_PARTITIONS(dv))
976 				*devp = MAKEDISKDEV(majdev, device_unit(dv),
977 						    part);
978 			else
979 				*devp = makedev(majdev, device_unit(dv));
980 		}
981 
982 		if (device_class(dv) == DV_IFNET)
983 			*devp = NODEV;
984 	}
985 
986 	*cp = c;
987 	return (dv);
988 }
989 
990 /*
991  * Return true if system call tracing is enabled for the specified process.
992  */
993 bool
994 trace_is_enabled(struct proc *p)
995 {
996 #ifdef SYSCALL_DEBUG
997 	return (true);
998 #endif
999 #ifdef KTRACE
1000 	if (ISSET(p->p_traceflag, (KTRFAC_SYSCALL | KTRFAC_SYSRET)))
1001 		return (true);
1002 #endif
1003 #ifdef PTRACE
1004 	if (ISSET(p->p_slflag, PSL_SYSCALL))
1005 		return (true);
1006 #endif
1007 
1008 	return (false);
1009 }
1010 
1011 /*
1012  * Start trace of particular system call. If process is being traced,
1013  * this routine is called by MD syscall dispatch code just before
1014  * a system call is actually executed.
1015  */
1016 int
1017 trace_enter(register_t code, const register_t *args, int narg)
1018 {
1019 #ifdef SYSCALL_DEBUG
1020 	scdebug_call(code, args);
1021 #endif /* SYSCALL_DEBUG */
1022 
1023 	ktrsyscall(code, args, narg);
1024 
1025 #ifdef PTRACE
1026 	if ((curlwp->l_proc->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
1027 	    (PSL_SYSCALL|PSL_TRACED))
1028 		process_stoptrace();
1029 #endif
1030 	return 0;
1031 }
1032 
1033 /*
1034  * End trace of particular system call. If process is being traced,
1035  * this routine is called by MD syscall dispatch code just after
1036  * a system call finishes.
1037  * MD caller guarantees the passed 'code' is within the supported
1038  * system call number range for emulation the process runs under.
1039  */
1040 void
1041 trace_exit(register_t code, register_t rval[], int error)
1042 {
1043 #ifdef SYSCALL_DEBUG
1044 	scdebug_ret(code, error, rval);
1045 #endif /* SYSCALL_DEBUG */
1046 
1047 	ktrsysret(code, error, rval);
1048 
1049 #ifdef PTRACE
1050 	if ((curlwp->l_proc->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
1051 	    (PSL_SYSCALL|PSL_TRACED))
1052 		process_stoptrace();
1053 #endif
1054 }
1055 
1056 int
1057 syscall_establish(const struct emul *em, const struct syscall_package *sp)
1058 {
1059 	struct sysent *sy;
1060 	int i;
1061 
1062 	KASSERT(mutex_owned(&module_lock));
1063 
1064 	if (em == NULL) {
1065 		em = &emul_netbsd;
1066 	}
1067 	sy = em->e_sysent;
1068 
1069 	/*
1070 	 * Ensure that all preconditions are valid, since this is
1071 	 * an all or nothing deal.  Once a system call is entered,
1072 	 * it can become busy and we could be unable to remove it
1073 	 * on error.
1074 	 */
1075 	for (i = 0; sp[i].sp_call != NULL; i++) {
1076 		if (sy[sp[i].sp_code].sy_call != sys_nomodule) {
1077 #ifdef DIAGNOSTIC
1078 			printf("syscall %d is busy\n", sp[i].sp_code);
1079 #endif
1080 			return EBUSY;
1081 		}
1082 	}
1083 	/* Everything looks good, patch them in. */
1084 	for (i = 0; sp[i].sp_call != NULL; i++) {
1085 		sy[sp[i].sp_code].sy_call = sp[i].sp_call;
1086 	}
1087 
1088 	return 0;
1089 }
1090 
1091 int
1092 syscall_disestablish(const struct emul *em, const struct syscall_package *sp)
1093 {
1094 	struct sysent *sy;
1095 	uint64_t where;
1096 	lwp_t *l;
1097 	int i;
1098 
1099 	KASSERT(mutex_owned(&module_lock));
1100 
1101 	if (em == NULL) {
1102 		em = &emul_netbsd;
1103 	}
1104 	sy = em->e_sysent;
1105 
1106 	/*
1107 	 * First, patch the system calls to sys_nomodule to gate further
1108 	 * activity.
1109 	 */
1110 	for (i = 0; sp[i].sp_call != NULL; i++) {
1111 		KASSERT(sy[sp[i].sp_code].sy_call == sp[i].sp_call);
1112 		sy[sp[i].sp_code].sy_call = sys_nomodule;
1113 	}
1114 
1115 	/*
1116 	 * Run a cross call to cycle through all CPUs.  This does two
1117 	 * things: lock activity provides a barrier and makes our update
1118 	 * of sy_call visible to all CPUs, and upon return we can be sure
1119 	 * that we see pertinent values of l_sysent posted by remote CPUs.
1120 	 */
1121 	where = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
1122 	xc_wait(where);
1123 
1124 	/*
1125 	 * Now it's safe to check l_sysent.  Run through all LWPs and see
1126 	 * if anyone is still using the system call.
1127 	 */
1128 	for (i = 0; sp[i].sp_call != NULL; i++) {
1129 		mutex_enter(proc_lock);
1130 		LIST_FOREACH(l, &alllwp, l_list) {
1131 			if (l->l_sysent == &sy[sp[i].sp_code]) {
1132 				break;
1133 			}
1134 		}
1135 		mutex_exit(proc_lock);
1136 		if (l == NULL) {
1137 			continue;
1138 		}
1139 		/*
1140 		 * We lose: one or more calls are still in use.  Put back
1141 		 * the old entrypoints and act like nothing happened.
1142 		 * When we drop module_lock, any system calls held in
1143 		 * sys_nomodule() will be restarted.
1144 		 */
1145 		for (i = 0; sp[i].sp_call != NULL; i++) {
1146 			sy[sp[i].sp_code].sy_call = sp[i].sp_call;
1147 		}
1148 		return EBUSY;
1149 	}
1150 
1151 	return 0;
1152 }
1153