xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision d9ac053afa92f3d7a493f791e10558f738234217)
1 /*	$NetBSD: kernfs_vnops.c,v 1.121 2006/06/07 22:33:41 kardel Exp $	*/
2 
3 /*
4  * Copyright (c) 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software donated to Berkeley by
8  * Jan-Simon Pendry.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)kernfs_vnops.c	8.15 (Berkeley) 5/21/95
35  */
36 
37 /*
38  * Kernel parameter filesystem (/kern)
39  */
40 
41 #include <sys/cdefs.h>
42 __KERNEL_RCSID(0, "$NetBSD: kernfs_vnops.c,v 1.121 2006/06/07 22:33:41 kardel Exp $");
43 
44 #ifdef _KERNEL_OPT
45 #include "opt_ipsec.h"
46 #endif
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/kernel.h>
51 #include <sys/vmmeter.h>
52 #include <sys/time.h>
53 #include <sys/proc.h>
54 #include <sys/vnode.h>
55 #include <sys/malloc.h>
56 #include <sys/file.h>
57 #include <sys/stat.h>
58 #include <sys/mount.h>
59 #include <sys/namei.h>
60 #include <sys/buf.h>
61 #include <sys/dirent.h>
62 #include <sys/msgbuf.h>
63 
64 #include <miscfs/genfs/genfs.h>
65 #include <miscfs/kernfs/kernfs.h>
66 
67 #ifdef IPSEC
68 #include <sys/mbuf.h>
69 #include <net/route.h>
70 #include <netinet/in.h>
71 #include <netinet6/ipsec.h>
72 #include <netkey/key.h>
73 #endif
74 
75 #include <uvm/uvm_extern.h>
76 
77 #define KSTRING	256		/* Largest I/O available via this filesystem */
78 #define	UIO_MX 32
79 
80 #define	READ_MODE	(S_IRUSR|S_IRGRP|S_IROTH)
81 #define	WRITE_MODE	(S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH)
82 #define	UREAD_MODE	(S_IRUSR)
83 #define	DIR_MODE	(S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
84 #define	UDIR_MODE	(S_IRUSR|S_IXUSR)
85 
86 #define N(s) sizeof(s)-1, s
87 const struct kern_target kern_targets[] = {
88 /* NOTE: The name must be less than UIO_MX-16 chars in length */
89      /*        name            data          tag           type  ro/rw */
90      { DT_DIR, N("."),         0,            KFSkern,        VDIR, DIR_MODE   },
91      { DT_DIR, N(".."),        0,            KFSroot,        VDIR, DIR_MODE   },
92      { DT_REG, N("boottime"),  &boottime.tv_sec, KFSint,     VREG, READ_MODE  },
93 			/* XXXUNCONST */
94      { DT_REG, N("copyright"), __UNCONST(copyright),
95      					     KFSstring,      VREG, READ_MODE  },
96      { DT_REG, N("hostname"),  0,            KFShostname,    VREG, WRITE_MODE },
97      { DT_REG, N("hz"),        &hz,          KFSint,         VREG, READ_MODE  },
98 #ifdef IPSEC
99      { DT_DIR, N("ipsecsa"),   0,	     KFSipsecsadir,  VDIR, UDIR_MODE  },
100      { DT_DIR, N("ipsecsp"),   0,	     KFSipsecspdir,  VDIR, UDIR_MODE  },
101 #endif
102      { DT_REG, N("loadavg"),   0,            KFSavenrun,     VREG, READ_MODE  },
103      { DT_REG, N("msgbuf"),    0,	     KFSmsgbuf,      VREG, READ_MODE  },
104      { DT_REG, N("pagesize"),  &uvmexp.pagesize, KFSint,     VREG, READ_MODE  },
105      { DT_REG, N("physmem"),   &physmem,     KFSint,         VREG, READ_MODE  },
106 #if 0
107      { DT_DIR, N("root"),      0,            KFSnull,        VDIR, DIR_MODE   },
108 #endif
109      { DT_BLK, N("rootdev"),   &rootdev,     KFSdevice,      VBLK, READ_MODE  },
110      { DT_CHR, N("rrootdev"),  &rrootdev,    KFSdevice,      VCHR, READ_MODE  },
111      { DT_REG, N("time"),      0,            KFStime,        VREG, READ_MODE  },
112 			/* XXXUNCONST */
113      { DT_REG, N("version"),   __UNCONST(version),
114      					     KFSstring,      VREG, READ_MODE  },
115 };
116 const struct kern_target subdir_targets[] = {
117 /* NOTE: The name must be less than UIO_MX-16 chars in length */
118      /*        name            data          tag           type  ro/rw */
119      { DT_DIR, N("."),         0,            KFSsubdir,      VDIR, DIR_MODE   },
120      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
121 };
122 #ifdef IPSEC
123 const struct kern_target ipsecsa_targets[] = {
124 /* NOTE: The name must be less than UIO_MX-16 chars in length */
125      /*        name            data          tag           type  ro/rw */
126      { DT_DIR, N("."),         0,            KFSipsecsadir,  VDIR, DIR_MODE   },
127      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
128 };
129 const struct kern_target ipsecsp_targets[] = {
130 /* NOTE: The name must be less than UIO_MX-16 chars in length */
131      /*        name            data          tag           type  ro/rw */
132      { DT_DIR, N("."),         0,            KFSipsecspdir,  VDIR, DIR_MODE   },
133      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
134 };
135 const struct kern_target ipsecsa_kt =
136      { DT_DIR, N(""),          0,            KFSipsecsa,     VREG, UREAD_MODE };
137 const struct kern_target ipsecsp_kt =
138      { DT_DIR, N(""),          0,            KFSipsecsp,     VREG, UREAD_MODE };
139 #endif
140 #undef N
141 SIMPLEQ_HEAD(,dyn_kern_target) dyn_kern_targets =
142 	SIMPLEQ_HEAD_INITIALIZER(dyn_kern_targets);
143 int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
144 const int static_nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
145 #ifdef IPSEC
146 int nipsecsa_targets = sizeof(ipsecsa_targets) / sizeof(ipsecsa_targets[0]);
147 int nipsecsp_targets = sizeof(ipsecsp_targets) / sizeof(ipsecsp_targets[0]);
148 int nkern_dirs = 4; /* 2 extra subdirs */
149 #else
150 int nkern_dirs = 2;
151 #endif
152 
153 int kernfs_try_fileop(kfstype, kfsfileop, void *, int);
154 int kernfs_try_xread(kfstype, const struct kernfs_node *, char *,
155     size_t, int);
156 int kernfs_try_xwrite(kfstype, const struct kernfs_node *, char *,
157     size_t, int);
158 
159 static int kernfs_default_xread(void *v);
160 static int kernfs_default_xwrite(void *v);
161 static int kernfs_default_fileop_getattr(void *);
162 
163 /* must include all fileop's */
164 const struct kernfs_fileop kernfs_default_fileops[] = {
165   { .kf_fileop = KERNFS_XREAD },
166   { .kf_fileop = KERNFS_XWRITE },
167   { .kf_fileop = KERNFS_FILEOP_OPEN },
168   { .kf_fileop = KERNFS_FILEOP_GETATTR,
169     .kf_genop = {kernfs_default_fileop_getattr} },
170   { .kf_fileop = KERNFS_FILEOP_IOCTL },
171   { .kf_fileop = KERNFS_FILEOP_CLOSE },
172   { .kf_fileop = KERNFS_FILEOP_READ, .kf_genop = {kernfs_default_xread} },
173   { .kf_fileop = KERNFS_FILEOP_WRITE, .kf_genop = {kernfs_default_xwrite} },
174 };
175 
176 int	kernfs_lookup(void *);
177 #define	kernfs_create	genfs_eopnotsupp
178 #define	kernfs_mknod	genfs_eopnotsupp
179 int	kernfs_open(void *);
180 int	kernfs_close(void *);
181 int	kernfs_access(void *);
182 int	kernfs_getattr(void *);
183 int	kernfs_setattr(void *);
184 int	kernfs_read(void *);
185 int	kernfs_write(void *);
186 #define	kernfs_fcntl	genfs_fcntl
187 int	kernfs_ioctl(void *);
188 #define	kernfs_poll	genfs_poll
189 #define kernfs_revoke	genfs_revoke
190 #define	kernfs_fsync	genfs_nullop
191 #define	kernfs_seek	genfs_nullop
192 #define	kernfs_remove	genfs_eopnotsupp
193 int	kernfs_link(void *);
194 #define	kernfs_rename	genfs_eopnotsupp
195 #define	kernfs_mkdir	genfs_eopnotsupp
196 #define	kernfs_rmdir	genfs_eopnotsupp
197 int	kernfs_symlink(void *);
198 int	kernfs_readdir(void *);
199 #define	kernfs_readlink	genfs_eopnotsupp
200 #define	kernfs_abortop	genfs_abortop
201 int	kernfs_inactive(void *);
202 int	kernfs_reclaim(void *);
203 #define	kernfs_lock	genfs_lock
204 #define	kernfs_unlock	genfs_unlock
205 #define	kernfs_bmap	genfs_badop
206 #define	kernfs_strategy	genfs_badop
207 int	kernfs_print(void *);
208 #define	kernfs_islocked	genfs_islocked
209 int	kernfs_pathconf(void *);
210 #define	kernfs_advlock	genfs_einval
211 #define	kernfs_bwrite	genfs_eopnotsupp
212 #define	kernfs_putpages	genfs_putpages
213 
214 static int	kernfs_xread(struct kernfs_node *, int, char **,
215 				size_t, size_t *);
216 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
217 
218 int (**kernfs_vnodeop_p)(void *);
219 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
220 	{ &vop_default_desc, vn_default_error },
221 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
222 	{ &vop_create_desc, kernfs_create },		/* create */
223 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
224 	{ &vop_open_desc, kernfs_open },		/* open */
225 	{ &vop_close_desc, kernfs_close },		/* close */
226 	{ &vop_access_desc, kernfs_access },		/* access */
227 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
228 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
229 	{ &vop_read_desc, kernfs_read },		/* read */
230 	{ &vop_write_desc, kernfs_write },		/* write */
231 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
232 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
233 	{ &vop_poll_desc, kernfs_poll },		/* poll */
234 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
235 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
236 	{ &vop_seek_desc, kernfs_seek },		/* seek */
237 	{ &vop_remove_desc, kernfs_remove },		/* remove */
238 	{ &vop_link_desc, kernfs_link },		/* link */
239 	{ &vop_rename_desc, kernfs_rename },		/* rename */
240 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
241 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
242 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
243 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
244 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
245 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
246 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
247 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
248 	{ &vop_lock_desc, kernfs_lock },		/* lock */
249 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
250 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
251 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
252 	{ &vop_print_desc, kernfs_print },		/* print */
253 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
254 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
255 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
256 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
257 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
258 	{ NULL, NULL }
259 };
260 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
261 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
262 
263 static inline int
264 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
265 {
266 	if (a->kf_type < b->kf_type)
267 		return -1;
268 	if (a->kf_type > b->kf_type)
269 		return 1;
270 	if (a->kf_fileop < b->kf_fileop)
271 		return -1;
272 	if (a->kf_fileop > b->kf_fileop)
273 		return 1;
274 	return (0);
275 }
276 
277 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
278 	SPLAY_INITIALIZER(kfsfileoptree);
279 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
280 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
281 
282 kfstype
283 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
284 {
285 	static u_char nextfreetype = KFSlasttype;
286 	struct kernfs_fileop *dkf, *fkf, skf;
287 	int i;
288 
289 	/* XXX need to keep track of dkf's memory if we support
290            deallocating types */
291 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
292 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
293 
294 	for (i = 0; i < sizeof(kernfs_default_fileops) /
295 		     sizeof(kernfs_default_fileops[0]); i++) {
296 		dkf[i].kf_type = nextfreetype;
297 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
298 	}
299 
300 	for (i = 0; i < nkf; i++) {
301 		skf.kf_type = nextfreetype;
302 		skf.kf_fileop = kf[i].kf_fileop;
303 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
304 			fkf->kf_genop = kf[i].kf_genop;
305 	}
306 
307 	return nextfreetype++;
308 }
309 
310 int
311 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
312 {
313 	struct kernfs_fileop *kf, skf;
314 
315 	skf.kf_type = type;
316 	skf.kf_fileop = fileop;
317 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
318 		if (kf->kf_vop)
319 			return kf->kf_vop(v);
320 	return error;
321 }
322 
323 int
324 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
325     size_t len, int error)
326 {
327 	struct kernfs_fileop *kf, skf;
328 
329 	skf.kf_type = type;
330 	skf.kf_fileop = KERNFS_XWRITE;
331 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
332 		if (kf->kf_xwrite)
333 			return kf->kf_xwrite(kfs, bf, len);
334 	return error;
335 }
336 
337 int
338 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
339 {
340 	struct kernfs_subdir *ks, *parent;
341 
342 	if (pkt == NULL) {
343 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
344 		nkern_targets++;
345 		if (dkt->dkt_kt.kt_vtype == VDIR)
346 			nkern_dirs++;
347 	} else {
348 		parent = (struct kernfs_subdir *)pkt->kt_data;
349 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
350 		parent->ks_nentries++;
351 		if (dkt->dkt_kt.kt_vtype == VDIR)
352 			parent->ks_dirs++;
353 	}
354 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
355 		ks = malloc(sizeof(struct kernfs_subdir),
356 		    M_TEMP, M_WAITOK);
357 		SIMPLEQ_INIT(&ks->ks_entries);
358 		ks->ks_nentries = 2; /* . and .. */
359 		ks->ks_dirs = 2;
360 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
361 		dkt->dkt_kt.kt_data = ks;
362 	}
363 	return 0;
364 }
365 
366 static int
367 kernfs_xread(kfs, off, bufp, len, wrlen)
368 	struct kernfs_node *kfs;
369 	int off;
370 	char **bufp;
371 	size_t len;
372 	size_t *wrlen;
373 {
374 	const struct kern_target *kt;
375 #ifdef IPSEC
376 	struct mbuf *m;
377 #endif
378 
379 	kt = kfs->kfs_kt;
380 
381 	switch (kfs->kfs_type) {
382 	case KFStime: {
383 		struct timeval tv;
384 
385 		microtime(&tv);
386 		snprintf(*bufp, len, "%ld %ld\n", tv.tv_sec, tv.tv_usec);
387 		break;
388 	}
389 
390 	case KFSint: {
391 		int *ip = kt->kt_data;
392 
393 		snprintf(*bufp, len, "%d\n", *ip);
394 		break;
395 	}
396 
397 	case KFSstring: {
398 		char *cp = kt->kt_data;
399 
400 		*bufp = cp;
401 		break;
402 	}
403 
404 	case KFSmsgbuf: {
405 		long n;
406 
407 		/*
408 		 * deal with cases where the message buffer has
409 		 * become corrupted.
410 		 */
411 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
412 			msgbufenabled = 0;
413 			return (ENXIO);
414 		}
415 
416 		/*
417 		 * Note that reads of /kern/msgbuf won't necessarily yield
418 		 * consistent results, if the message buffer is modified
419 		 * while the read is in progress.  The worst that can happen
420 		 * is that incorrect data will be read.  There's no way
421 		 * that this can crash the system unless the values in the
422 		 * message buffer header are corrupted, but that'll cause
423 		 * the system to die anyway.
424 		 */
425 		if (off >= msgbufp->msg_bufs) {
426 			*wrlen = 0;
427 			return (0);
428 		}
429 		n = msgbufp->msg_bufx + off;
430 		if (n >= msgbufp->msg_bufs)
431 			n -= msgbufp->msg_bufs;
432 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
433 		*bufp = msgbufp->msg_bufc + n;
434 		*wrlen = len;
435 		return (0);
436 	}
437 
438 	case KFShostname: {
439 		char *cp = hostname;
440 		int xlen = hostnamelen;
441 
442 		if (xlen >= (len - 2))
443 			return (EINVAL);
444 
445 		memcpy(*bufp, cp, xlen);
446 		(*bufp)[xlen] = '\n';
447 		(*bufp)[xlen+1] = '\0';
448 		len = strlen(*bufp);
449 		break;
450 	}
451 
452 	case KFSavenrun:
453 		averunnable.fscale = FSCALE;
454 		snprintf(*bufp, len, "%d %d %d %ld\n",
455 		    averunnable.ldavg[0], averunnable.ldavg[1],
456 		    averunnable.ldavg[2], averunnable.fscale);
457 		break;
458 
459 #ifdef IPSEC
460 	case KFSipsecsa:
461 		/*
462 		 * Note that SA configuration could be changed during the
463 		 * read operation, resulting in garbled output.
464 		 */
465 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
466 		if (!m)
467 			return (ENOBUFS);
468 		if (off >= m->m_pkthdr.len) {
469 			*wrlen = 0;
470 			m_freem(m);
471 			return (0);
472 		}
473 		if (len > m->m_pkthdr.len - off)
474 			len = m->m_pkthdr.len - off;
475 		m_copydata(m, off, len, *bufp);
476 		*wrlen = len;
477 		m_freem(m);
478 		return (0);
479 
480 	case KFSipsecsp:
481 		/*
482 		 * Note that SP configuration could be changed during the
483 		 * read operation, resulting in garbled output.
484 		 */
485 		if (!kfs->kfs_v) {
486 			struct secpolicy *sp;
487 
488 			sp = key_getspbyid(kfs->kfs_value);
489 			if (sp)
490 				kfs->kfs_v = sp;
491 			else
492 				return (ENOENT);
493 		}
494 		m = key_setdumpsp((struct secpolicy *)kfs->kfs_v,
495 		    SADB_X_SPDGET, 0, 0);
496 		if (!m)
497 			return (ENOBUFS);
498 		if (off >= m->m_pkthdr.len) {
499 			*wrlen = 0;
500 			m_freem(m);
501 			return (0);
502 		}
503 		if (len > m->m_pkthdr.len - off)
504 			len = m->m_pkthdr.len - off;
505 		m_copydata(m, off, len, *bufp);
506 		*wrlen = len;
507 		m_freem(m);
508 		return (0);
509 #endif
510 
511 	default:
512 		*wrlen = 0;
513 		return (0);
514 	}
515 
516 	len = strlen(*bufp);
517 	if (len <= off)
518 		*wrlen = 0;
519 	else {
520 		*bufp += off;
521 		*wrlen = len - off;
522 	}
523 	return (0);
524 }
525 
526 static int
527 kernfs_xwrite(kfs, bf, len)
528 	const struct kernfs_node *kfs;
529 	char *bf;
530 	size_t len;
531 {
532 
533 	switch (kfs->kfs_type) {
534 	case KFShostname:
535 		if (bf[len-1] == '\n')
536 			--len;
537 		memcpy(hostname, bf, len);
538 		hostname[len] = '\0';
539 		hostnamelen = (size_t) len;
540 		return (0);
541 
542 	default:
543 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
544 	}
545 }
546 
547 
548 /*
549  * vp is the current namei directory
550  * ndp is the name to locate in that directory...
551  */
552 int
553 kernfs_lookup(v)
554 	void *v;
555 {
556 	struct vop_lookup_args /* {
557 		struct vnode * a_dvp;
558 		struct vnode ** a_vpp;
559 		struct componentname * a_cnp;
560 	} */ *ap = v;
561 	struct componentname *cnp = ap->a_cnp;
562 	struct vnode **vpp = ap->a_vpp;
563 	struct vnode *dvp = ap->a_dvp;
564 	const char *pname = cnp->cn_nameptr;
565 	const struct kernfs_node *kfs;
566 	const struct kern_target *kt;
567 	const struct dyn_kern_target *dkt;
568 	const struct kernfs_subdir *ks;
569 	int error, i, wantpunlock;
570 #ifdef IPSEC
571 	char *ep;
572 	u_int32_t id;
573 #endif
574 
575 	*vpp = NULLVP;
576 	cnp->cn_flags &= ~PDIRUNLOCK;
577 
578 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
579 		return (EROFS);
580 
581 	if (cnp->cn_namelen == 1 && *pname == '.') {
582 		*vpp = dvp;
583 		VREF(dvp);
584 		return (0);
585 	}
586 
587 	wantpunlock = (~cnp->cn_flags & (LOCKPARENT | ISLASTCN));
588 	kfs = VTOKERN(dvp);
589 	switch (kfs->kfs_type) {
590 	case KFSkern:
591 		/*
592 		 * Shouldn't get here with .. in the root node.
593 		 */
594 		if (cnp->cn_flags & ISDOTDOT)
595 			return (EIO);
596 
597 		for (i = 0; i < static_nkern_targets; i++) {
598 			kt = &kern_targets[i];
599 			if (cnp->cn_namelen == kt->kt_namlen &&
600 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
601 				goto found;
602 		}
603 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
604 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
605 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
606 				kt = &dkt->dkt_kt;
607 				goto found;
608 			}
609 		}
610 		break;
611 
612 	found:
613 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
614 		if ((error == 0) && wantpunlock) {
615 			VOP_UNLOCK(dvp, 0);
616 			cnp->cn_flags |= PDIRUNLOCK;
617 		}
618 		return (error);
619 
620 	case KFSsubdir:
621 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
622 		if (cnp->cn_flags & ISDOTDOT) {
623 			kt = ks->ks_parent;
624 			goto found;
625 		}
626 
627 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
628 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
629 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
630 				kt = &dkt->dkt_kt;
631 				goto found;
632 			}
633 		}
634 		break;
635 
636 #ifdef IPSEC
637 	case KFSipsecsadir:
638 		if (cnp->cn_flags & ISDOTDOT) {
639 			kt = &kern_targets[0];
640 			goto found;
641 		}
642 
643 		for (i = 2; i < nipsecsa_targets; i++) {
644 			kt = &ipsecsa_targets[i];
645 			if (cnp->cn_namelen == kt->kt_namlen &&
646 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
647 				goto found;
648 		}
649 
650 		ep = NULL;
651 		id = strtoul(pname, &ep, 10);
652 		if (!ep || *ep || ep == pname)
653 			break;
654 
655 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id);
656 		if ((error == 0) && wantpunlock) {
657 			VOP_UNLOCK(dvp, 0);
658 			cnp->cn_flags |= PDIRUNLOCK;
659 		}
660 		return (error);
661 
662 	case KFSipsecspdir:
663 		if (cnp->cn_flags & ISDOTDOT) {
664 			kt = &kern_targets[0];
665 			goto found;
666 		}
667 
668 		for (i = 2; i < nipsecsp_targets; i++) {
669 			kt = &ipsecsp_targets[i];
670 			if (cnp->cn_namelen == kt->kt_namlen &&
671 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
672 				goto found;
673 		}
674 
675 		ep = NULL;
676 		id = strtoul(pname, &ep, 10);
677 		if (!ep || *ep || ep == pname)
678 			break;
679 
680 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id);
681 		if ((error == 0) && wantpunlock) {
682 			VOP_UNLOCK(dvp, 0);
683 			cnp->cn_flags |= PDIRUNLOCK;
684 		}
685 		return (error);
686 #endif
687 
688 	default:
689 		return (ENOTDIR);
690 	}
691 
692 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
693 }
694 
695 int
696 kernfs_open(v)
697 	void *v;
698 {
699 	struct vop_open_args /* {
700 		struct vnode *a_vp;
701 		int a_mode;
702 		kauth_cred_t a_cred;
703 		struct lwp *a_l;
704 	} */ *ap = v;
705 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
706 #ifdef IPSEC
707 	struct mbuf *m;
708 	struct secpolicy *sp;
709 #endif
710 
711 	switch (kfs->kfs_type) {
712 #ifdef IPSEC
713 	case KFSipsecsa:
714 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
715 		if (m) {
716 			m_freem(m);
717 			return (0);
718 		} else
719 			return (ENOENT);
720 
721 	case KFSipsecsp:
722 		sp = key_getspbyid(kfs->kfs_value);
723 		if (sp) {
724 			kfs->kfs_v = sp;
725 			return (0);
726 		} else
727 			return (ENOENT);
728 #endif
729 
730 	default:
731 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN,
732 		    v, 0);
733 	}
734 }
735 
736 int
737 kernfs_close(v)
738 	void *v;
739 {
740 	struct vop_close_args /* {
741 		struct vnode *a_vp;
742 		int a_fflag;
743 		kauth_cred_t a_cred;
744 		struct lwp *a_l;
745 	} */ *ap = v;
746 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
747 
748 	switch (kfs->kfs_type) {
749 #ifdef IPSEC
750 	case KFSipsecsp:
751 		key_freesp((struct secpolicy *)kfs->kfs_v);
752 		break;
753 #endif
754 
755 	default:
756 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE,
757 		    v, 0);
758 	}
759 
760 	return (0);
761 }
762 
763 int
764 kernfs_access(v)
765 	void *v;
766 {
767 	struct vop_access_args /* {
768 		struct vnode *a_vp;
769 		int a_mode;
770 		kauth_cred_t a_cred;
771 		struct lwp *a_l;
772 	} */ *ap = v;
773 	struct vattr va;
774 	int error;
775 
776 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_l)) != 0)
777 		return (error);
778 
779 	return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid,
780 	    ap->a_mode, ap->a_cred));
781 }
782 
783 static int
784 kernfs_default_fileop_getattr(v)
785 	void *v;
786 {
787 	struct vop_getattr_args /* {
788 		struct vnode *a_vp;
789 		struct vattr *a_vap;
790 		kauth_cred_t a_cred;
791 		struct lwp *a_l;
792 	} */ *ap = v;
793 	struct vattr *vap = ap->a_vap;
794 
795 	vap->va_nlink = 1;
796 	vap->va_bytes = vap->va_size = 0;
797 
798 	return 0;
799 }
800 
801 int
802 kernfs_getattr(v)
803 	void *v;
804 {
805 	struct vop_getattr_args /* {
806 		struct vnode *a_vp;
807 		struct vattr *a_vap;
808 		kauth_cred_t a_cred;
809 		struct lwp *a_l;
810 	} */ *ap = v;
811 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
812 	struct kernfs_subdir *ks;
813 	struct vattr *vap = ap->a_vap;
814 	int error = 0;
815 	char strbuf[KSTRING], *bf;
816 	size_t nread, total;
817 
818 	VATTR_NULL(vap);
819 	vap->va_type = ap->a_vp->v_type;
820 	vap->va_uid = 0;
821 	vap->va_gid = 0;
822 	vap->va_mode = kfs->kfs_mode;
823 	vap->va_fileid = kfs->kfs_fileno;
824 	vap->va_flags = 0;
825 	vap->va_size = 0;
826 	vap->va_blocksize = DEV_BSIZE;
827 	/* Make all times be current TOD, except for the "boottime" node. */
828 	if (kfs->kfs_kt && kfs->kfs_kt->kt_namlen == 8 &&
829 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
830 		TIMEVAL_TO_TIMESPEC(&boottime, &vap->va_ctime);
831 	} else {
832 		getnanotime(&vap->va_ctime);
833 	}
834 	vap->va_atime = vap->va_mtime = vap->va_ctime;
835 	vap->va_gen = 0;
836 	vap->va_flags = 0;
837 	vap->va_rdev = 0;
838 	vap->va_bytes = 0;
839 
840 	switch (kfs->kfs_type) {
841 	case KFSkern:
842 		vap->va_nlink = nkern_dirs;
843 		vap->va_bytes = vap->va_size = DEV_BSIZE;
844 		break;
845 
846 	case KFSroot:
847 		vap->va_nlink = 1;
848 		vap->va_bytes = vap->va_size = DEV_BSIZE;
849 		break;
850 
851 	case KFSsubdir:
852 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
853 		vap->va_nlink = ks->ks_dirs;
854 		vap->va_bytes = vap->va_size = DEV_BSIZE;
855 		break;
856 
857 	case KFSnull:
858 	case KFStime:
859 	case KFSint:
860 	case KFSstring:
861 	case KFShostname:
862 	case KFSavenrun:
863 	case KFSdevice:
864 	case KFSmsgbuf:
865 #ifdef IPSEC
866 	case KFSipsecsa:
867 	case KFSipsecsp:
868 #endif
869 		vap->va_nlink = 1;
870 		total = 0;
871 		do {
872 			bf = strbuf;
873 			error = kernfs_xread(kfs, total, &bf,
874 			    sizeof(strbuf), &nread);
875 			total += nread;
876 		} while (error == 0 && nread != 0);
877 		vap->va_bytes = vap->va_size = total;
878 		break;
879 
880 #ifdef IPSEC
881 	case KFSipsecsadir:
882 	case KFSipsecspdir:
883 		vap->va_nlink = 2;
884 		vap->va_bytes = vap->va_size = DEV_BSIZE;
885 		break;
886 #endif
887 
888 	default:
889 		error = kernfs_try_fileop(kfs->kfs_type,
890 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
891 		break;
892 	}
893 
894 	return (error);
895 }
896 
897 /*ARGSUSED*/
898 int
899 kernfs_setattr(v)
900 	void *v;
901 {
902 
903 	/*
904 	 * Silently ignore attribute changes.
905 	 * This allows for open with truncate to have no
906 	 * effect until some data is written.  I want to
907 	 * do it this way because all writes are atomic.
908 	 */
909 	return (0);
910 }
911 
912 int
913 kernfs_default_xread(v)
914 	void *v;
915 {
916 	struct vop_read_args /* {
917 		struct vnode *a_vp;
918 		struct uio *a_uio;
919 		int  a_ioflag;
920 		kauth_cred_t a_cred;
921 	} */ *ap = v;
922 	struct uio *uio = ap->a_uio;
923 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
924 	char strbuf[KSTRING], *bf;
925 	int off;
926 	size_t len;
927 	int error;
928 
929 	if (ap->a_vp->v_type == VDIR)
930 		return (EOPNOTSUPP);
931 
932 	off = (int)uio->uio_offset;
933 	/* Don't allow negative offsets */
934 	if (off < 0)
935 		return EINVAL;
936 
937 	bf = strbuf;
938 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
939 		error = uiomove(bf, len, uio);
940 	return (error);
941 }
942 
943 int
944 kernfs_read(v)
945 	void *v;
946 {
947 	struct vop_read_args /* {
948 		struct vnode *a_vp;
949 		struct uio *a_uio;
950 		int  a_ioflag;
951 		struct ucred *a_cred;
952 	} */ *ap = v;
953 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
954 
955 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v, 0);
956 }
957 
958 static int
959 kernfs_default_xwrite(v)
960 	void *v;
961 {
962 	struct vop_write_args /* {
963 		struct vnode *a_vp;
964 		struct uio *a_uio;
965 		int  a_ioflag;
966 		kauth_cred_t a_cred;
967 	} */ *ap = v;
968 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
969 	struct uio *uio = ap->a_uio;
970 	int error, xlen;
971 	char strbuf[KSTRING];
972 
973 	if (uio->uio_offset != 0)
974 		return (EINVAL);
975 
976 	xlen = min(uio->uio_resid, KSTRING-1);
977 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
978 		return (error);
979 
980 	if (uio->uio_resid != 0)
981 		return (EIO);
982 
983 	strbuf[xlen] = '\0';
984 	xlen = strlen(strbuf);
985 	return (kernfs_xwrite(kfs, strbuf, xlen));
986 }
987 
988 int
989 kernfs_write(v)
990 	void *v;
991 {
992 	struct vop_write_args /* {
993 		struct vnode *a_vp;
994 		struct uio *a_uio;
995 		int  a_ioflag;
996 		kauth_cred_t a_cred;
997 	} */ *ap = v;
998 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
999 
1000 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v, 0);
1001 }
1002 
1003 int
1004 kernfs_ioctl(v)
1005 	void *v;
1006 {
1007 	struct vop_ioctl_args /* {
1008 		const struct vnodeop_desc *a_desc;
1009 		struct vnode *a_vp;
1010 		u_long a_command;
1011 		void *a_data;
1012 		int a_fflag;
1013 		kauth_cred_t a_cred;
1014 		struct lwp *a_l;
1015 	} */ *ap = v;
1016 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1017 
1018 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1019 	    EPASSTHROUGH);
1020 }
1021 
1022 static int
1023 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1024     u_int32_t value, struct vop_readdir_args *ap)
1025 {
1026 	struct kernfs_node *kfs;
1027 	struct vnode *vp;
1028 	int error;
1029 
1030 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1031 	    value)) != 0)
1032 		return error;
1033 	if (kt->kt_tag == KFSdevice) {
1034 		struct vattr va;
1035 
1036 		error = VOP_GETATTR(vp, &va, ap->a_cred, curlwp);
1037 		if (error != 0) {
1038 			return error;
1039 		}
1040 		d->d_fileno = va.va_fileid;
1041 	} else {
1042 		kfs = VTOKERN(vp);
1043 		d->d_fileno = kfs->kfs_fileno;
1044 	}
1045 	vput(vp);
1046 	return 0;
1047 }
1048 
1049 static int
1050 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1051     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1052     const struct kern_target *kt, struct vop_readdir_args *ap)
1053 {
1054 	const struct kern_target *ikt;
1055 	int error;
1056 
1057 	switch (entry) {
1058 	case 0:
1059 		d->d_fileno = thisdir_kfs->kfs_fileno;
1060 		return 0;
1061 	case 1:
1062 		ikt = parent_kt;
1063 		break;
1064 	default:
1065 		ikt = kt;
1066 		break;
1067 	}
1068 	if (ikt != thisdir_kfs->kfs_kt) {
1069 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1070 			return error;
1071 	} else
1072 		d->d_fileno = thisdir_kfs->kfs_fileno;
1073 	return 0;
1074 }
1075 
1076 int
1077 kernfs_readdir(v)
1078 	void *v;
1079 {
1080 	struct vop_readdir_args /* {
1081 		struct vnode *a_vp;
1082 		struct uio *a_uio;
1083 		kauth_cred_t a_cred;
1084 		int *a_eofflag;
1085 		off_t **a_cookies;
1086 		int a_*ncookies;
1087 	} */ *ap = v;
1088 	struct uio *uio = ap->a_uio;
1089 	struct dirent d;
1090 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1091 	const struct kern_target *kt;
1092 	const struct dyn_kern_target *dkt = NULL;
1093 	const struct kernfs_subdir *ks;
1094 	off_t i, j;
1095 	int error;
1096 	off_t *cookies = NULL;
1097 	int ncookies = 0, n;
1098 #ifdef IPSEC
1099 	struct secasvar *sav, *sav2;
1100 	struct secpolicy *sp;
1101 #endif
1102 
1103 	if (uio->uio_resid < UIO_MX)
1104 		return (EINVAL);
1105 	if (uio->uio_offset < 0)
1106 		return (EINVAL);
1107 
1108 	error = 0;
1109 	i = uio->uio_offset;
1110 	memset(&d, 0, sizeof(d));
1111 	d.d_reclen = UIO_MX;
1112 	ncookies = uio->uio_resid / UIO_MX;
1113 
1114 	switch (kfs->kfs_type) {
1115 	case KFSkern:
1116 		if (i >= nkern_targets)
1117 			return (0);
1118 
1119 		if (ap->a_ncookies) {
1120 			ncookies = min(ncookies, (nkern_targets - i));
1121 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1122 			    M_WAITOK);
1123 			*ap->a_cookies = cookies;
1124 		}
1125 
1126 		n = 0;
1127 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1128 			if (i < static_nkern_targets)
1129 				kt = &kern_targets[i];
1130 			else {
1131 				if (dkt == NULL) {
1132 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1133 					for (j = static_nkern_targets; j < i &&
1134 						     dkt != NULL; j++)
1135 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1136 					if (j != i)
1137 						break;
1138 				} else {
1139 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1140 				}
1141 				if (dkt == NULL)
1142 					break;
1143 				kt = &dkt->dkt_kt;
1144 			}
1145 			if (kt->kt_tag == KFSdevice) {
1146 				dev_t *dp = kt->kt_data;
1147 				struct vnode *fvp;
1148 
1149 				if (*dp == NODEV ||
1150 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1151 					continue;
1152 			}
1153 			d.d_namlen = kt->kt_namlen;
1154 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1155 			    &kern_targets[0], kt, ap)) != 0)
1156 				break;
1157 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1158 			d.d_type = kt->kt_type;
1159 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1160 				break;
1161 			if (cookies)
1162 				*cookies++ = i + 1;
1163 			n++;
1164 		}
1165 		ncookies = n;
1166 		break;
1167 
1168 	case KFSroot:
1169 		if (i >= 2)
1170 			return 0;
1171 
1172 		if (ap->a_ncookies) {
1173 			ncookies = min(ncookies, (2 - i));
1174 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1175 			    M_WAITOK);
1176 			*ap->a_cookies = cookies;
1177 		}
1178 
1179 		n = 0;
1180 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1181 			kt = &kern_targets[i];
1182 			d.d_namlen = kt->kt_namlen;
1183 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
1184 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1185 			d.d_type = kt->kt_type;
1186 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1187 				break;
1188 			if (cookies)
1189 				*cookies++ = i + 1;
1190 			n++;
1191 		}
1192 		ncookies = n;
1193 		break;
1194 
1195 	case KFSsubdir:
1196 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
1197 		if (i >= ks->ks_nentries)
1198 			return (0);
1199 
1200 		if (ap->a_ncookies) {
1201 			ncookies = min(ncookies, (ks->ks_nentries - i));
1202 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1203 			    M_WAITOK);
1204 			*ap->a_cookies = cookies;
1205 		}
1206 
1207 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1208 		for (j = 0; j < i && dkt != NULL; j++)
1209 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1210 		n = 0;
1211 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1212 			if (i < 2)
1213 				kt = &subdir_targets[i];
1214 			else {
1215 				/* check if ks_nentries lied to us */
1216 				if (dkt == NULL)
1217 					break;
1218 				kt = &dkt->dkt_kt;
1219 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1220 			}
1221 			if (kt->kt_tag == KFSdevice) {
1222 				dev_t *dp = kt->kt_data;
1223 				struct vnode *fvp;
1224 
1225 				if (*dp == NODEV ||
1226 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1227 					continue;
1228 			}
1229 			d.d_namlen = kt->kt_namlen;
1230 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1231 			    ks->ks_parent, kt, ap)) != 0)
1232 				break;
1233 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1234 			d.d_type = kt->kt_type;
1235 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1236 				break;
1237 			if (cookies)
1238 				*cookies++ = i + 1;
1239 			n++;
1240 		}
1241 		ncookies = n;
1242 		break;
1243 
1244 #ifdef IPSEC
1245 	case KFSipsecsadir:
1246 		/* count SA in the system */
1247 		n = 0;
1248 		TAILQ_FOREACH(sav, &satailq, tailq) {
1249 			for (sav2 = TAILQ_FIRST(&satailq);
1250 			    sav2 != sav;
1251 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1252 				if (sav->spi == sav2->spi) {
1253 					/* multiple SA with same SPI */
1254 					break;
1255 				}
1256 			}
1257 			if (sav == sav2 || sav->spi != sav2->spi)
1258 				n++;
1259 		}
1260 
1261 		if (i >= nipsecsa_targets + n)
1262 			return (0);
1263 
1264 		if (ap->a_ncookies) {
1265 			ncookies = min(ncookies, (n - i));
1266 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1267 			    M_WAITOK);
1268 			*ap->a_cookies = cookies;
1269 		}
1270 
1271 		n = 0;
1272 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1273 			kt = &ipsecsa_targets[i];
1274 			d.d_namlen = kt->kt_namlen;
1275 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1276 			    &kern_targets[0], kt, ap)) != 0)
1277 				break;
1278 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1279 			d.d_type = kt->kt_type;
1280 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1281 				break;
1282 			if (cookies)
1283 				*cookies++ = i + 1;
1284 			n++;
1285 		}
1286 		if (error) {
1287 			ncookies = n;
1288 			break;
1289 		}
1290 
1291 		TAILQ_FOREACH(sav, &satailq, tailq) {
1292 			for (sav2 = TAILQ_FIRST(&satailq);
1293 			    sav2 != sav;
1294 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1295 				if (sav->spi == sav2->spi) {
1296 					/* multiple SA with same SPI */
1297 					break;
1298 				}
1299 			}
1300 			if (sav != sav2 && sav->spi == sav2->spi)
1301 				continue;
1302 			if (uio->uio_resid < UIO_MX)
1303 				break;
1304 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1305 			    sav->spi, ap)) != 0)
1306 				break;
1307 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1308 			    "%u", ntohl(sav->spi));
1309 			d.d_type = DT_REG;
1310 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1311 				break;
1312 			if (cookies)
1313 				*cookies++ = i + 1;
1314 			n++;
1315 			i++;
1316 		}
1317 		ncookies = n;
1318 		break;
1319 
1320 	case KFSipsecspdir:
1321 		/* count SP in the system */
1322 		n = 0;
1323 		TAILQ_FOREACH(sp, &sptailq, tailq)
1324 			n++;
1325 
1326 		if (i >= nipsecsp_targets + n)
1327 			return (0);
1328 
1329 		if (ap->a_ncookies) {
1330 			ncookies = min(ncookies, (n - i));
1331 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1332 			    M_WAITOK);
1333 			*ap->a_cookies = cookies;
1334 		}
1335 
1336 		n = 0;
1337 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1338 			kt = &ipsecsp_targets[i];
1339 			d.d_namlen = kt->kt_namlen;
1340 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1341 			    &kern_targets[0], kt, ap)) != 0)
1342 				break;
1343 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1344 			d.d_type = kt->kt_type;
1345 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1346 				break;
1347 			if (cookies)
1348 				*cookies++ = i + 1;
1349 			n++;
1350 		}
1351 		if (error) {
1352 			ncookies = n;
1353 			break;
1354 		}
1355 
1356 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1357 			if (uio->uio_resid < UIO_MX)
1358 				break;
1359 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1360 			    sp->id, ap)) != 0)
1361 				break;
1362 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1363 			    "%u", sp->id);
1364 			d.d_type = DT_REG;
1365 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1366 				break;
1367 			if (cookies)
1368 				*cookies++ = i + 1;
1369 			n++;
1370 			i++;
1371 		}
1372 		ncookies = n;
1373 		break;
1374 #endif
1375 
1376 	default:
1377 		error = ENOTDIR;
1378 		break;
1379 	}
1380 
1381 	if (ap->a_ncookies) {
1382 		if (error) {
1383 			if (cookies)
1384 				free(*ap->a_cookies, M_TEMP);
1385 			*ap->a_ncookies = 0;
1386 			*ap->a_cookies = NULL;
1387 		} else
1388 			*ap->a_ncookies = ncookies;
1389 	}
1390 
1391 	uio->uio_offset = i;
1392 	return (error);
1393 }
1394 
1395 int
1396 kernfs_inactive(v)
1397 	void *v;
1398 {
1399 	struct vop_inactive_args /* {
1400 		struct vnode *a_vp;
1401 		struct lwp *a_l;
1402 	} */ *ap = v;
1403 	struct vnode *vp = ap->a_vp;
1404 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1405 #ifdef IPSEC
1406 	struct mbuf *m;
1407 	struct secpolicy *sp;
1408 #endif
1409 
1410 	VOP_UNLOCK(vp, 0);
1411 	switch (kfs->kfs_type) {
1412 #ifdef IPSEC
1413 	case KFSipsecsa:
1414 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1415 		if (m)
1416 			m_freem(m);
1417 		else
1418 			vgone(vp);
1419 		break;
1420 	case KFSipsecsp:
1421 		sp = key_getspbyid(kfs->kfs_value);
1422 		if (sp)
1423 			key_freesp(sp);
1424 		else {
1425 			/* should never happen as we hold a refcnt */
1426 			vgone(vp);
1427 		}
1428 		break;
1429 #endif
1430 	default:
1431 		break;
1432 	}
1433 	return (0);
1434 }
1435 
1436 int
1437 kernfs_reclaim(v)
1438 	void *v;
1439 {
1440 	struct vop_reclaim_args /* {
1441 		struct vnode *a_vp;
1442 	} */ *ap = v;
1443 
1444 	return (kernfs_freevp(ap->a_vp));
1445 }
1446 
1447 /*
1448  * Return POSIX pathconf information applicable to special devices.
1449  */
1450 int
1451 kernfs_pathconf(v)
1452 	void *v;
1453 {
1454 	struct vop_pathconf_args /* {
1455 		struct vnode *a_vp;
1456 		int a_name;
1457 		register_t *a_retval;
1458 	} */ *ap = v;
1459 
1460 	switch (ap->a_name) {
1461 	case _PC_LINK_MAX:
1462 		*ap->a_retval = LINK_MAX;
1463 		return (0);
1464 	case _PC_MAX_CANON:
1465 		*ap->a_retval = MAX_CANON;
1466 		return (0);
1467 	case _PC_MAX_INPUT:
1468 		*ap->a_retval = MAX_INPUT;
1469 		return (0);
1470 	case _PC_PIPE_BUF:
1471 		*ap->a_retval = PIPE_BUF;
1472 		return (0);
1473 	case _PC_CHOWN_RESTRICTED:
1474 		*ap->a_retval = 1;
1475 		return (0);
1476 	case _PC_VDISABLE:
1477 		*ap->a_retval = _POSIX_VDISABLE;
1478 		return (0);
1479 	case _PC_SYNC_IO:
1480 		*ap->a_retval = 1;
1481 		return (0);
1482 	default:
1483 		return (EINVAL);
1484 	}
1485 	/* NOTREACHED */
1486 }
1487 
1488 /*
1489  * Print out the contents of a /dev/fd vnode.
1490  */
1491 /* ARGSUSED */
1492 int
1493 kernfs_print(v)
1494 	void *v;
1495 {
1496 
1497 	printf("tag VT_KERNFS, kernfs vnode\n");
1498 	return (0);
1499 }
1500 
1501 int
1502 kernfs_link(v)
1503 	void *v;
1504 {
1505 	struct vop_link_args /* {
1506 		struct vnode *a_dvp;
1507 		struct vnode *a_vp;
1508 		struct componentname *a_cnp;
1509 	} */ *ap = v;
1510 
1511 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1512 	vput(ap->a_dvp);
1513 	return (EROFS);
1514 }
1515 
1516 int
1517 kernfs_symlink(v)
1518 	void *v;
1519 {
1520 	struct vop_symlink_args /* {
1521 		struct vnode *a_dvp;
1522 		struct vnode **a_vpp;
1523 		struct componentname *a_cnp;
1524 		struct vattr *a_vap;
1525 		char *a_target;
1526 	} */ *ap = v;
1527 
1528 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1529 	vput(ap->a_dvp);
1530 	return (EROFS);
1531 }
1532