xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: kernfs_vnops.c,v 1.126 2006/10/12 01:32:27 christos 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 1.126 2006/10/12 01:32:27 christos 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_vop = kernfs_default_fileop_getattr },
170   { .kf_fileop = KERNFS_FILEOP_IOCTL },
171   { .kf_fileop = KERNFS_FILEOP_CLOSE },
172   { .kf_fileop = KERNFS_FILEOP_READ,
173     .kf_vop = kernfs_default_xread },
174   { .kf_fileop = KERNFS_FILEOP_WRITE,
175     .kf_vop = kernfs_default_xwrite },
176 };
177 
178 int	kernfs_lookup(void *);
179 #define	kernfs_create	genfs_eopnotsupp
180 #define	kernfs_mknod	genfs_eopnotsupp
181 int	kernfs_open(void *);
182 int	kernfs_close(void *);
183 int	kernfs_access(void *);
184 int	kernfs_getattr(void *);
185 int	kernfs_setattr(void *);
186 int	kernfs_read(void *);
187 int	kernfs_write(void *);
188 #define	kernfs_fcntl	genfs_fcntl
189 int	kernfs_ioctl(void *);
190 #define	kernfs_poll	genfs_poll
191 #define kernfs_revoke	genfs_revoke
192 #define	kernfs_fsync	genfs_nullop
193 #define	kernfs_seek	genfs_nullop
194 #define	kernfs_remove	genfs_eopnotsupp
195 int	kernfs_link(void *);
196 #define	kernfs_rename	genfs_eopnotsupp
197 #define	kernfs_mkdir	genfs_eopnotsupp
198 #define	kernfs_rmdir	genfs_eopnotsupp
199 int	kernfs_symlink(void *);
200 int	kernfs_readdir(void *);
201 #define	kernfs_readlink	genfs_eopnotsupp
202 #define	kernfs_abortop	genfs_abortop
203 int	kernfs_inactive(void *);
204 int	kernfs_reclaim(void *);
205 #define	kernfs_lock	genfs_lock
206 #define	kernfs_unlock	genfs_unlock
207 #define	kernfs_bmap	genfs_badop
208 #define	kernfs_strategy	genfs_badop
209 int	kernfs_print(void *);
210 #define	kernfs_islocked	genfs_islocked
211 int	kernfs_pathconf(void *);
212 #define	kernfs_advlock	genfs_einval
213 #define	kernfs_bwrite	genfs_eopnotsupp
214 #define	kernfs_putpages	genfs_putpages
215 
216 static int	kernfs_xread(struct kernfs_node *, int, char **,
217 				size_t, size_t *);
218 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
219 
220 int (**kernfs_vnodeop_p)(void *);
221 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
222 	{ &vop_default_desc, vn_default_error },
223 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
224 	{ &vop_create_desc, kernfs_create },		/* create */
225 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
226 	{ &vop_open_desc, kernfs_open },		/* open */
227 	{ &vop_close_desc, kernfs_close },		/* close */
228 	{ &vop_access_desc, kernfs_access },		/* access */
229 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
230 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
231 	{ &vop_read_desc, kernfs_read },		/* read */
232 	{ &vop_write_desc, kernfs_write },		/* write */
233 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
234 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
235 	{ &vop_poll_desc, kernfs_poll },		/* poll */
236 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
237 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
238 	{ &vop_seek_desc, kernfs_seek },		/* seek */
239 	{ &vop_remove_desc, kernfs_remove },		/* remove */
240 	{ &vop_link_desc, kernfs_link },		/* link */
241 	{ &vop_rename_desc, kernfs_rename },		/* rename */
242 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
243 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
244 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
245 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
246 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
247 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
248 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
249 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
250 	{ &vop_lock_desc, kernfs_lock },		/* lock */
251 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
252 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
253 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
254 	{ &vop_print_desc, kernfs_print },		/* print */
255 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
256 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
257 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
258 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
259 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
260 	{ NULL, NULL }
261 };
262 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
263 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
264 
265 static inline int
266 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
267 {
268 	if (a->kf_type < b->kf_type)
269 		return -1;
270 	if (a->kf_type > b->kf_type)
271 		return 1;
272 	if (a->kf_fileop < b->kf_fileop)
273 		return -1;
274 	if (a->kf_fileop > b->kf_fileop)
275 		return 1;
276 	return (0);
277 }
278 
279 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
280 	SPLAY_INITIALIZER(kfsfileoptree);
281 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
282 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
283 
284 kfstype
285 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
286 {
287 	static u_char nextfreetype = KFSlasttype;
288 	struct kernfs_fileop *dkf, *fkf, skf;
289 	int i;
290 
291 	/* XXX need to keep track of dkf's memory if we support
292            deallocating types */
293 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
294 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
295 
296 	for (i = 0; i < sizeof(kernfs_default_fileops) /
297 		     sizeof(kernfs_default_fileops[0]); i++) {
298 		dkf[i].kf_type = nextfreetype;
299 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
300 	}
301 
302 	for (i = 0; i < nkf; i++) {
303 		skf.kf_type = nextfreetype;
304 		skf.kf_fileop = kf[i].kf_fileop;
305 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
306 			fkf->kf_vop = kf[i].kf_vop;
307 	}
308 
309 	return nextfreetype++;
310 }
311 
312 int
313 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
314 {
315 	struct kernfs_fileop *kf, skf;
316 
317 	skf.kf_type = type;
318 	skf.kf_fileop = fileop;
319 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
320 		if (kf->kf_vop)
321 			return kf->kf_vop(v);
322 	return error;
323 }
324 
325 int
326 kernfs_try_xread(kfstype type, const struct kernfs_node *kfs, char **bfp,
327     size_t len, int error)
328 {
329 	struct kernfs_fileop *kf, skf;
330 
331 	skf.kf_type = type;
332 	skf.kf_fileop = KERNFS_XREAD;
333 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
334 		if (kf->kf_xread)
335 			return kf->kf_xread(kfs, bfp, len);
336 	return error;
337 }
338 
339 int
340 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
341     size_t len, int error)
342 {
343 	struct kernfs_fileop *kf, skf;
344 
345 	skf.kf_type = type;
346 	skf.kf_fileop = KERNFS_XWRITE;
347 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
348 		if (kf->kf_xwrite)
349 			return kf->kf_xwrite(kfs, bf, len);
350 	return error;
351 }
352 
353 int
354 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
355 {
356 	struct kernfs_subdir *ks, *parent;
357 
358 	if (pkt == NULL) {
359 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
360 		nkern_targets++;
361 		if (dkt->dkt_kt.kt_vtype == VDIR)
362 			nkern_dirs++;
363 	} else {
364 		parent = (struct kernfs_subdir *)pkt->kt_data;
365 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
366 		parent->ks_nentries++;
367 		if (dkt->dkt_kt.kt_vtype == VDIR)
368 			parent->ks_dirs++;
369 	}
370 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
371 		ks = malloc(sizeof(struct kernfs_subdir),
372 		    M_TEMP, M_WAITOK);
373 		SIMPLEQ_INIT(&ks->ks_entries);
374 		ks->ks_nentries = 2; /* . and .. */
375 		ks->ks_dirs = 2;
376 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
377 		dkt->dkt_kt.kt_data = ks;
378 	}
379 	return 0;
380 }
381 
382 static int
383 kernfs_xread(kfs, off, bufp, len, wrlen)
384 	struct kernfs_node *kfs;
385 	int off;
386 	char **bufp;
387 	size_t len;
388 	size_t *wrlen;
389 {
390 	const struct kern_target *kt;
391 #ifdef IPSEC
392 	struct mbuf *m;
393 #endif
394 	int err;
395 
396 	kt = kfs->kfs_kt;
397 
398 	switch (kfs->kfs_type) {
399 	case KFStime: {
400 		struct timeval tv;
401 
402 		microtime(&tv);
403 		snprintf(*bufp, len, "%ld %ld\n", tv.tv_sec, tv.tv_usec);
404 		break;
405 	}
406 
407 	case KFSint: {
408 		int *ip = kt->kt_data;
409 
410 		snprintf(*bufp, len, "%d\n", *ip);
411 		break;
412 	}
413 
414 	case KFSstring: {
415 		char *cp = kt->kt_data;
416 
417 		*bufp = cp;
418 		break;
419 	}
420 
421 	case KFSmsgbuf: {
422 		long n;
423 
424 		/*
425 		 * deal with cases where the message buffer has
426 		 * become corrupted.
427 		 */
428 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
429 			msgbufenabled = 0;
430 			return (ENXIO);
431 		}
432 
433 		/*
434 		 * Note that reads of /kern/msgbuf won't necessarily yield
435 		 * consistent results, if the message buffer is modified
436 		 * while the read is in progress.  The worst that can happen
437 		 * is that incorrect data will be read.  There's no way
438 		 * that this can crash the system unless the values in the
439 		 * message buffer header are corrupted, but that'll cause
440 		 * the system to die anyway.
441 		 */
442 		if (off >= msgbufp->msg_bufs) {
443 			*wrlen = 0;
444 			return (0);
445 		}
446 		n = msgbufp->msg_bufx + off;
447 		if (n >= msgbufp->msg_bufs)
448 			n -= msgbufp->msg_bufs;
449 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
450 		*bufp = msgbufp->msg_bufc + n;
451 		*wrlen = len;
452 		return (0);
453 	}
454 
455 	case KFShostname: {
456 		char *cp = hostname;
457 		int xlen = hostnamelen;
458 
459 		if (xlen >= (len - 2))
460 			return (EINVAL);
461 
462 		memcpy(*bufp, cp, xlen);
463 		(*bufp)[xlen] = '\n';
464 		(*bufp)[xlen+1] = '\0';
465 		len = strlen(*bufp);
466 		break;
467 	}
468 
469 	case KFSavenrun:
470 		averunnable.fscale = FSCALE;
471 		snprintf(*bufp, len, "%d %d %d %ld\n",
472 		    averunnable.ldavg[0], averunnable.ldavg[1],
473 		    averunnable.ldavg[2], averunnable.fscale);
474 		break;
475 
476 #ifdef IPSEC
477 	case KFSipsecsa:
478 		/*
479 		 * Note that SA configuration could be changed during the
480 		 * read operation, resulting in garbled output.
481 		 */
482 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
483 		if (!m)
484 			return (ENOBUFS);
485 		if (off >= m->m_pkthdr.len) {
486 			*wrlen = 0;
487 			m_freem(m);
488 			return (0);
489 		}
490 		if (len > m->m_pkthdr.len - off)
491 			len = m->m_pkthdr.len - off;
492 		m_copydata(m, off, len, *bufp);
493 		*wrlen = len;
494 		m_freem(m);
495 		return (0);
496 
497 	case KFSipsecsp:
498 		/*
499 		 * Note that SP configuration could be changed during the
500 		 * read operation, resulting in garbled output.
501 		 */
502 		if (!kfs->kfs_v) {
503 			struct secpolicy *sp;
504 
505 			sp = key_getspbyid(kfs->kfs_value);
506 			if (sp)
507 				kfs->kfs_v = sp;
508 			else
509 				return (ENOENT);
510 		}
511 		m = key_setdumpsp((struct secpolicy *)kfs->kfs_v,
512 		    SADB_X_SPDGET, 0, 0);
513 		if (!m)
514 			return (ENOBUFS);
515 		if (off >= m->m_pkthdr.len) {
516 			*wrlen = 0;
517 			m_freem(m);
518 			return (0);
519 		}
520 		if (len > m->m_pkthdr.len - off)
521 			len = m->m_pkthdr.len - off;
522 		m_copydata(m, off, len, *bufp);
523 		*wrlen = len;
524 		m_freem(m);
525 		return (0);
526 #endif
527 
528 	default:
529 		err = kernfs_try_xread(kfs->kfs_type, kfs, bufp, len,
530 		    EOPNOTSUPP);
531 		if (err)
532 			return err;
533 	}
534 
535 	len = strlen(*bufp);
536 	if (len <= off)
537 		*wrlen = 0;
538 	else {
539 		*bufp += off;
540 		*wrlen = len - off;
541 	}
542 	return (0);
543 }
544 
545 static int
546 kernfs_xwrite(kfs, bf, len)
547 	const struct kernfs_node *kfs;
548 	char *bf;
549 	size_t len;
550 {
551 
552 	switch (kfs->kfs_type) {
553 	case KFShostname:
554 		if (bf[len-1] == '\n')
555 			--len;
556 		memcpy(hostname, bf, len);
557 		hostname[len] = '\0';
558 		hostnamelen = (size_t) len;
559 		return (0);
560 
561 	default:
562 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
563 	}
564 }
565 
566 
567 /*
568  * vp is the current namei directory
569  * ndp is the name to locate in that directory...
570  */
571 int
572 kernfs_lookup(v)
573 	void *v;
574 {
575 	struct vop_lookup_args /* {
576 		struct vnode * a_dvp;
577 		struct vnode ** a_vpp;
578 		struct componentname * a_cnp;
579 	} */ *ap = v;
580 	struct componentname *cnp = ap->a_cnp;
581 	struct vnode **vpp = ap->a_vpp;
582 	struct vnode *dvp = ap->a_dvp;
583 	const char *pname = cnp->cn_nameptr;
584 	const struct kernfs_node *kfs;
585 	const struct kern_target *kt;
586 	const struct dyn_kern_target *dkt;
587 	const struct kernfs_subdir *ks;
588 	int error, i, wantpunlock;
589 #ifdef IPSEC
590 	char *ep;
591 	u_int32_t id;
592 #endif
593 
594 	*vpp = NULLVP;
595 	cnp->cn_flags &= ~PDIRUNLOCK;
596 
597 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
598 		return (EROFS);
599 
600 	if (cnp->cn_namelen == 1 && *pname == '.') {
601 		*vpp = dvp;
602 		VREF(dvp);
603 		return (0);
604 	}
605 
606 	wantpunlock = (~cnp->cn_flags & (LOCKPARENT | ISLASTCN));
607 	kfs = VTOKERN(dvp);
608 	switch (kfs->kfs_type) {
609 	case KFSkern:
610 		/*
611 		 * Shouldn't get here with .. in the root node.
612 		 */
613 		if (cnp->cn_flags & ISDOTDOT)
614 			return (EIO);
615 
616 		for (i = 0; i < static_nkern_targets; i++) {
617 			kt = &kern_targets[i];
618 			if (cnp->cn_namelen == kt->kt_namlen &&
619 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
620 				goto found;
621 		}
622 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
623 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
624 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
625 				kt = &dkt->dkt_kt;
626 				goto found;
627 			}
628 		}
629 		break;
630 
631 	found:
632 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
633 		if ((error == 0) && wantpunlock) {
634 			VOP_UNLOCK(dvp, 0);
635 			cnp->cn_flags |= PDIRUNLOCK;
636 		}
637 		return (error);
638 
639 	case KFSsubdir:
640 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
641 		if (cnp->cn_flags & ISDOTDOT) {
642 			kt = ks->ks_parent;
643 			goto found;
644 		}
645 
646 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
647 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
648 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
649 				kt = &dkt->dkt_kt;
650 				goto found;
651 			}
652 		}
653 		break;
654 
655 #ifdef IPSEC
656 	case KFSipsecsadir:
657 		if (cnp->cn_flags & ISDOTDOT) {
658 			kt = &kern_targets[0];
659 			goto found;
660 		}
661 
662 		for (i = 2; i < nipsecsa_targets; i++) {
663 			kt = &ipsecsa_targets[i];
664 			if (cnp->cn_namelen == kt->kt_namlen &&
665 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
666 				goto found;
667 		}
668 
669 		ep = NULL;
670 		id = strtoul(pname, &ep, 10);
671 		if (!ep || *ep || ep == pname)
672 			break;
673 
674 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id);
675 		if ((error == 0) && wantpunlock) {
676 			VOP_UNLOCK(dvp, 0);
677 			cnp->cn_flags |= PDIRUNLOCK;
678 		}
679 		return (error);
680 
681 	case KFSipsecspdir:
682 		if (cnp->cn_flags & ISDOTDOT) {
683 			kt = &kern_targets[0];
684 			goto found;
685 		}
686 
687 		for (i = 2; i < nipsecsp_targets; i++) {
688 			kt = &ipsecsp_targets[i];
689 			if (cnp->cn_namelen == kt->kt_namlen &&
690 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
691 				goto found;
692 		}
693 
694 		ep = NULL;
695 		id = strtoul(pname, &ep, 10);
696 		if (!ep || *ep || ep == pname)
697 			break;
698 
699 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id);
700 		if ((error == 0) && wantpunlock) {
701 			VOP_UNLOCK(dvp, 0);
702 			cnp->cn_flags |= PDIRUNLOCK;
703 		}
704 		return (error);
705 #endif
706 
707 	default:
708 		return (ENOTDIR);
709 	}
710 
711 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
712 }
713 
714 int
715 kernfs_open(v)
716 	void *v;
717 {
718 	struct vop_open_args /* {
719 		struct vnode *a_vp;
720 		int a_mode;
721 		kauth_cred_t a_cred;
722 		struct lwp *a_l;
723 	} */ *ap = v;
724 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
725 #ifdef IPSEC
726 	struct mbuf *m;
727 	struct secpolicy *sp;
728 #endif
729 
730 	switch (kfs->kfs_type) {
731 #ifdef IPSEC
732 	case KFSipsecsa:
733 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
734 		if (m) {
735 			m_freem(m);
736 			return (0);
737 		} else
738 			return (ENOENT);
739 
740 	case KFSipsecsp:
741 		sp = key_getspbyid(kfs->kfs_value);
742 		if (sp) {
743 			kfs->kfs_v = sp;
744 			return (0);
745 		} else
746 			return (ENOENT);
747 #endif
748 
749 	default:
750 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN,
751 		    v, 0);
752 	}
753 }
754 
755 int
756 kernfs_close(v)
757 	void *v;
758 {
759 	struct vop_close_args /* {
760 		struct vnode *a_vp;
761 		int a_fflag;
762 		kauth_cred_t a_cred;
763 		struct lwp *a_l;
764 	} */ *ap = v;
765 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
766 
767 	switch (kfs->kfs_type) {
768 #ifdef IPSEC
769 	case KFSipsecsp:
770 		key_freesp((struct secpolicy *)kfs->kfs_v);
771 		break;
772 #endif
773 
774 	default:
775 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE,
776 		    v, 0);
777 	}
778 
779 	return (0);
780 }
781 
782 int
783 kernfs_access(v)
784 	void *v;
785 {
786 	struct vop_access_args /* {
787 		struct vnode *a_vp;
788 		int a_mode;
789 		kauth_cred_t a_cred;
790 		struct lwp *a_l;
791 	} */ *ap = v;
792 	struct vattr va;
793 	int error;
794 
795 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_l)) != 0)
796 		return (error);
797 
798 	return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid,
799 	    ap->a_mode, ap->a_cred));
800 }
801 
802 static int
803 kernfs_default_fileop_getattr(v)
804 	void *v;
805 {
806 	struct vop_getattr_args /* {
807 		struct vnode *a_vp;
808 		struct vattr *a_vap;
809 		kauth_cred_t a_cred;
810 		struct lwp *a_l;
811 	} */ *ap = v;
812 	struct vattr *vap = ap->a_vap;
813 
814 	vap->va_nlink = 1;
815 	vap->va_bytes = vap->va_size = 0;
816 
817 	return 0;
818 }
819 
820 int
821 kernfs_getattr(v)
822 	void *v;
823 {
824 	struct vop_getattr_args /* {
825 		struct vnode *a_vp;
826 		struct vattr *a_vap;
827 		kauth_cred_t a_cred;
828 		struct lwp *a_l;
829 	} */ *ap = v;
830 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
831 	struct kernfs_subdir *ks;
832 	struct vattr *vap = ap->a_vap;
833 	int error = 0;
834 	char strbuf[KSTRING], *bf;
835 	size_t nread, total;
836 
837 	VATTR_NULL(vap);
838 	vap->va_type = ap->a_vp->v_type;
839 	vap->va_uid = 0;
840 	vap->va_gid = 0;
841 	vap->va_mode = kfs->kfs_mode;
842 	vap->va_fileid = kfs->kfs_fileno;
843 	vap->va_flags = 0;
844 	vap->va_size = 0;
845 	vap->va_blocksize = DEV_BSIZE;
846 	/* Make all times be current TOD, except for the "boottime" node. */
847 	if (kfs->kfs_kt && kfs->kfs_kt->kt_namlen == 8 &&
848 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
849 		TIMEVAL_TO_TIMESPEC(&boottime, &vap->va_ctime);
850 	} else {
851 		getnanotime(&vap->va_ctime);
852 	}
853 	vap->va_atime = vap->va_mtime = vap->va_ctime;
854 	vap->va_gen = 0;
855 	vap->va_flags = 0;
856 	vap->va_rdev = 0;
857 	vap->va_bytes = 0;
858 
859 	switch (kfs->kfs_type) {
860 	case KFSkern:
861 		vap->va_nlink = nkern_dirs;
862 		vap->va_bytes = vap->va_size = DEV_BSIZE;
863 		break;
864 
865 	case KFSroot:
866 		vap->va_nlink = 1;
867 		vap->va_bytes = vap->va_size = DEV_BSIZE;
868 		break;
869 
870 	case KFSsubdir:
871 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
872 		vap->va_nlink = ks->ks_dirs;
873 		vap->va_bytes = vap->va_size = DEV_BSIZE;
874 		break;
875 
876 	case KFSnull:
877 	case KFStime:
878 	case KFSint:
879 	case KFSstring:
880 	case KFShostname:
881 	case KFSavenrun:
882 	case KFSdevice:
883 	case KFSmsgbuf:
884 #ifdef IPSEC
885 	case KFSipsecsa:
886 	case KFSipsecsp:
887 #endif
888 		vap->va_nlink = 1;
889 		total = 0;
890 		do {
891 			bf = strbuf;
892 			error = kernfs_xread(kfs, total, &bf,
893 			    sizeof(strbuf), &nread);
894 			total += nread;
895 		} while (error == 0 && nread != 0);
896 		vap->va_bytes = vap->va_size = total;
897 		break;
898 
899 #ifdef IPSEC
900 	case KFSipsecsadir:
901 	case KFSipsecspdir:
902 		vap->va_nlink = 2;
903 		vap->va_bytes = vap->va_size = DEV_BSIZE;
904 		break;
905 #endif
906 
907 	default:
908 		error = kernfs_try_fileop(kfs->kfs_type,
909 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
910 		break;
911 	}
912 
913 	return (error);
914 }
915 
916 /*ARGSUSED*/
917 int
918 kernfs_setattr(void *v __unused)
919 {
920 
921 	/*
922 	 * Silently ignore attribute changes.
923 	 * This allows for open with truncate to have no
924 	 * effect until some data is written.  I want to
925 	 * do it this way because all writes are atomic.
926 	 */
927 	return (0);
928 }
929 
930 int
931 kernfs_default_xread(v)
932 	void *v;
933 {
934 	struct vop_read_args /* {
935 		struct vnode *a_vp;
936 		struct uio *a_uio;
937 		int  a_ioflag;
938 		kauth_cred_t a_cred;
939 	} */ *ap = v;
940 	struct uio *uio = ap->a_uio;
941 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
942 	char strbuf[KSTRING], *bf;
943 	int off;
944 	size_t len;
945 	int error;
946 
947 	if (ap->a_vp->v_type == VDIR)
948 		return (EOPNOTSUPP);
949 
950 	off = (int)uio->uio_offset;
951 	/* Don't allow negative offsets */
952 	if (off < 0)
953 		return EINVAL;
954 
955 	bf = strbuf;
956 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
957 		error = uiomove(bf, len, uio);
958 	return (error);
959 }
960 
961 int
962 kernfs_read(v)
963 	void *v;
964 {
965 	struct vop_read_args /* {
966 		struct vnode *a_vp;
967 		struct uio *a_uio;
968 		int  a_ioflag;
969 		struct ucred *a_cred;
970 	} */ *ap = v;
971 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
972 
973 	if (kfs->kfs_type < KFSlasttype) {
974 		/* use default function */
975 		return kernfs_default_xread(v);
976 	}
977 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
978 	   EOPNOTSUPP);
979 }
980 
981 static int
982 kernfs_default_xwrite(v)
983 	void *v;
984 {
985 	struct vop_write_args /* {
986 		struct vnode *a_vp;
987 		struct uio *a_uio;
988 		int  a_ioflag;
989 		kauth_cred_t a_cred;
990 	} */ *ap = v;
991 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
992 	struct uio *uio = ap->a_uio;
993 	int error, xlen;
994 	char strbuf[KSTRING];
995 
996 	if (uio->uio_offset != 0)
997 		return (EINVAL);
998 
999 	xlen = min(uio->uio_resid, KSTRING-1);
1000 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
1001 		return (error);
1002 
1003 	if (uio->uio_resid != 0)
1004 		return (EIO);
1005 
1006 	strbuf[xlen] = '\0';
1007 	xlen = strlen(strbuf);
1008 	return (kernfs_xwrite(kfs, strbuf, xlen));
1009 }
1010 
1011 int
1012 kernfs_write(v)
1013 	void *v;
1014 {
1015 	struct vop_write_args /* {
1016 		struct vnode *a_vp;
1017 		struct uio *a_uio;
1018 		int  a_ioflag;
1019 		kauth_cred_t a_cred;
1020 	} */ *ap = v;
1021 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1022 
1023 	if (kfs->kfs_type < KFSlasttype) {
1024 		/* use default function */
1025 		return kernfs_default_xwrite(v);
1026 	}
1027 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
1028 	    EOPNOTSUPP);
1029 }
1030 
1031 int
1032 kernfs_ioctl(v)
1033 	void *v;
1034 {
1035 	struct vop_ioctl_args /* {
1036 		const struct vnodeop_desc *a_desc;
1037 		struct vnode *a_vp;
1038 		u_long a_command;
1039 		void *a_data;
1040 		int a_fflag;
1041 		kauth_cred_t a_cred;
1042 		struct lwp *a_l;
1043 	} */ *ap = v;
1044 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1045 
1046 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1047 	    EPASSTHROUGH);
1048 }
1049 
1050 static int
1051 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1052     u_int32_t value, struct vop_readdir_args *ap)
1053 {
1054 	struct kernfs_node *kfs;
1055 	struct vnode *vp;
1056 	int error;
1057 
1058 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1059 	    value)) != 0)
1060 		return error;
1061 	if (kt->kt_tag == KFSdevice) {
1062 		struct vattr va;
1063 
1064 		error = VOP_GETATTR(vp, &va, ap->a_cred, curlwp);
1065 		if (error != 0) {
1066 			return error;
1067 		}
1068 		d->d_fileno = va.va_fileid;
1069 	} else {
1070 		kfs = VTOKERN(vp);
1071 		d->d_fileno = kfs->kfs_fileno;
1072 	}
1073 	vput(vp);
1074 	return 0;
1075 }
1076 
1077 static int
1078 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1079     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1080     const struct kern_target *kt, struct vop_readdir_args *ap)
1081 {
1082 	const struct kern_target *ikt;
1083 	int error;
1084 
1085 	switch (entry) {
1086 	case 0:
1087 		d->d_fileno = thisdir_kfs->kfs_fileno;
1088 		return 0;
1089 	case 1:
1090 		ikt = parent_kt;
1091 		break;
1092 	default:
1093 		ikt = kt;
1094 		break;
1095 	}
1096 	if (ikt != thisdir_kfs->kfs_kt) {
1097 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1098 			return error;
1099 	} else
1100 		d->d_fileno = thisdir_kfs->kfs_fileno;
1101 	return 0;
1102 }
1103 
1104 int
1105 kernfs_readdir(v)
1106 	void *v;
1107 {
1108 	struct vop_readdir_args /* {
1109 		struct vnode *a_vp;
1110 		struct uio *a_uio;
1111 		kauth_cred_t a_cred;
1112 		int *a_eofflag;
1113 		off_t **a_cookies;
1114 		int a_*ncookies;
1115 	} */ *ap = v;
1116 	struct uio *uio = ap->a_uio;
1117 	struct dirent d;
1118 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1119 	const struct kern_target *kt;
1120 	const struct dyn_kern_target *dkt = NULL;
1121 	const struct kernfs_subdir *ks;
1122 	off_t i, j;
1123 	int error;
1124 	off_t *cookies = NULL;
1125 	int ncookies = 0, n;
1126 #ifdef IPSEC
1127 	struct secasvar *sav, *sav2;
1128 	struct secpolicy *sp;
1129 #endif
1130 
1131 	if (uio->uio_resid < UIO_MX)
1132 		return (EINVAL);
1133 	if (uio->uio_offset < 0)
1134 		return (EINVAL);
1135 
1136 	error = 0;
1137 	i = uio->uio_offset;
1138 	memset(&d, 0, sizeof(d));
1139 	d.d_reclen = UIO_MX;
1140 	ncookies = uio->uio_resid / UIO_MX;
1141 
1142 	switch (kfs->kfs_type) {
1143 	case KFSkern:
1144 		if (i >= nkern_targets)
1145 			return (0);
1146 
1147 		if (ap->a_ncookies) {
1148 			ncookies = min(ncookies, (nkern_targets - i));
1149 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1150 			    M_WAITOK);
1151 			*ap->a_cookies = cookies;
1152 		}
1153 
1154 		n = 0;
1155 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1156 			if (i < static_nkern_targets)
1157 				kt = &kern_targets[i];
1158 			else {
1159 				if (dkt == NULL) {
1160 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1161 					for (j = static_nkern_targets; j < i &&
1162 						     dkt != NULL; j++)
1163 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1164 					if (j != i)
1165 						break;
1166 				} else {
1167 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1168 				}
1169 				if (dkt == NULL)
1170 					break;
1171 				kt = &dkt->dkt_kt;
1172 			}
1173 			if (kt->kt_tag == KFSdevice) {
1174 				dev_t *dp = kt->kt_data;
1175 				struct vnode *fvp;
1176 
1177 				if (*dp == NODEV ||
1178 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1179 					continue;
1180 			}
1181 			d.d_namlen = kt->kt_namlen;
1182 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1183 			    &kern_targets[0], kt, ap)) != 0)
1184 				break;
1185 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1186 			d.d_type = kt->kt_type;
1187 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1188 				break;
1189 			if (cookies)
1190 				*cookies++ = i + 1;
1191 			n++;
1192 		}
1193 		ncookies = n;
1194 		break;
1195 
1196 	case KFSroot:
1197 		if (i >= 2)
1198 			return 0;
1199 
1200 		if (ap->a_ncookies) {
1201 			ncookies = min(ncookies, (2 - i));
1202 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1203 			    M_WAITOK);
1204 			*ap->a_cookies = cookies;
1205 		}
1206 
1207 		n = 0;
1208 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1209 			kt = &kern_targets[i];
1210 			d.d_namlen = kt->kt_namlen;
1211 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
1212 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1213 			d.d_type = kt->kt_type;
1214 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1215 				break;
1216 			if (cookies)
1217 				*cookies++ = i + 1;
1218 			n++;
1219 		}
1220 		ncookies = n;
1221 		break;
1222 
1223 	case KFSsubdir:
1224 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
1225 		if (i >= ks->ks_nentries)
1226 			return (0);
1227 
1228 		if (ap->a_ncookies) {
1229 			ncookies = min(ncookies, (ks->ks_nentries - i));
1230 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1231 			    M_WAITOK);
1232 			*ap->a_cookies = cookies;
1233 		}
1234 
1235 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1236 		for (j = 0; j < i && dkt != NULL; j++)
1237 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1238 		n = 0;
1239 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1240 			if (i < 2)
1241 				kt = &subdir_targets[i];
1242 			else {
1243 				/* check if ks_nentries lied to us */
1244 				if (dkt == NULL)
1245 					break;
1246 				kt = &dkt->dkt_kt;
1247 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1248 			}
1249 			if (kt->kt_tag == KFSdevice) {
1250 				dev_t *dp = kt->kt_data;
1251 				struct vnode *fvp;
1252 
1253 				if (*dp == NODEV ||
1254 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1255 					continue;
1256 			}
1257 			d.d_namlen = kt->kt_namlen;
1258 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1259 			    ks->ks_parent, kt, ap)) != 0)
1260 				break;
1261 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1262 			d.d_type = kt->kt_type;
1263 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1264 				break;
1265 			if (cookies)
1266 				*cookies++ = i + 1;
1267 			n++;
1268 		}
1269 		ncookies = n;
1270 		break;
1271 
1272 #ifdef IPSEC
1273 	case KFSipsecsadir:
1274 		/* count SA in the system */
1275 		n = 0;
1276 		TAILQ_FOREACH(sav, &satailq, tailq) {
1277 			for (sav2 = TAILQ_FIRST(&satailq);
1278 			    sav2 != sav;
1279 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1280 				if (sav->spi == sav2->spi) {
1281 					/* multiple SA with same SPI */
1282 					break;
1283 				}
1284 			}
1285 			if (sav == sav2 || sav->spi != sav2->spi)
1286 				n++;
1287 		}
1288 
1289 		if (i >= nipsecsa_targets + n)
1290 			return (0);
1291 
1292 		if (ap->a_ncookies) {
1293 			ncookies = min(ncookies, (n - i));
1294 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1295 			    M_WAITOK);
1296 			*ap->a_cookies = cookies;
1297 		}
1298 
1299 		n = 0;
1300 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1301 			kt = &ipsecsa_targets[i];
1302 			d.d_namlen = kt->kt_namlen;
1303 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1304 			    &kern_targets[0], kt, ap)) != 0)
1305 				break;
1306 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1307 			d.d_type = kt->kt_type;
1308 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1309 				break;
1310 			if (cookies)
1311 				*cookies++ = i + 1;
1312 			n++;
1313 		}
1314 		if (error) {
1315 			ncookies = n;
1316 			break;
1317 		}
1318 
1319 		TAILQ_FOREACH(sav, &satailq, tailq) {
1320 			for (sav2 = TAILQ_FIRST(&satailq);
1321 			    sav2 != sav;
1322 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1323 				if (sav->spi == sav2->spi) {
1324 					/* multiple SA with same SPI */
1325 					break;
1326 				}
1327 			}
1328 			if (sav != sav2 && sav->spi == sav2->spi)
1329 				continue;
1330 			if (uio->uio_resid < UIO_MX)
1331 				break;
1332 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1333 			    sav->spi, ap)) != 0)
1334 				break;
1335 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1336 			    "%u", ntohl(sav->spi));
1337 			d.d_type = DT_REG;
1338 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1339 				break;
1340 			if (cookies)
1341 				*cookies++ = i + 1;
1342 			n++;
1343 			i++;
1344 		}
1345 		ncookies = n;
1346 		break;
1347 
1348 	case KFSipsecspdir:
1349 		/* count SP in the system */
1350 		n = 0;
1351 		TAILQ_FOREACH(sp, &sptailq, tailq)
1352 			n++;
1353 
1354 		if (i >= nipsecsp_targets + n)
1355 			return (0);
1356 
1357 		if (ap->a_ncookies) {
1358 			ncookies = min(ncookies, (n - i));
1359 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1360 			    M_WAITOK);
1361 			*ap->a_cookies = cookies;
1362 		}
1363 
1364 		n = 0;
1365 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1366 			kt = &ipsecsp_targets[i];
1367 			d.d_namlen = kt->kt_namlen;
1368 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1369 			    &kern_targets[0], kt, ap)) != 0)
1370 				break;
1371 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1372 			d.d_type = kt->kt_type;
1373 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1374 				break;
1375 			if (cookies)
1376 				*cookies++ = i + 1;
1377 			n++;
1378 		}
1379 		if (error) {
1380 			ncookies = n;
1381 			break;
1382 		}
1383 
1384 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1385 			if (uio->uio_resid < UIO_MX)
1386 				break;
1387 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1388 			    sp->id, ap)) != 0)
1389 				break;
1390 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1391 			    "%u", sp->id);
1392 			d.d_type = DT_REG;
1393 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1394 				break;
1395 			if (cookies)
1396 				*cookies++ = i + 1;
1397 			n++;
1398 			i++;
1399 		}
1400 		ncookies = n;
1401 		break;
1402 #endif
1403 
1404 	default:
1405 		error = ENOTDIR;
1406 		break;
1407 	}
1408 
1409 	if (ap->a_ncookies) {
1410 		if (error) {
1411 			if (cookies)
1412 				free(*ap->a_cookies, M_TEMP);
1413 			*ap->a_ncookies = 0;
1414 			*ap->a_cookies = NULL;
1415 		} else
1416 			*ap->a_ncookies = ncookies;
1417 	}
1418 
1419 	uio->uio_offset = i;
1420 	return (error);
1421 }
1422 
1423 int
1424 kernfs_inactive(v)
1425 	void *v;
1426 {
1427 	struct vop_inactive_args /* {
1428 		struct vnode *a_vp;
1429 		struct lwp *a_l;
1430 	} */ *ap = v;
1431 	struct vnode *vp = ap->a_vp;
1432 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1433 #ifdef IPSEC
1434 	struct mbuf *m;
1435 	struct secpolicy *sp;
1436 #endif
1437 
1438 	VOP_UNLOCK(vp, 0);
1439 	switch (kfs->kfs_type) {
1440 #ifdef IPSEC
1441 	case KFSipsecsa:
1442 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1443 		if (m)
1444 			m_freem(m);
1445 		else
1446 			vgone(vp);
1447 		break;
1448 	case KFSipsecsp:
1449 		sp = key_getspbyid(kfs->kfs_value);
1450 		if (sp)
1451 			key_freesp(sp);
1452 		else {
1453 			/* should never happen as we hold a refcnt */
1454 			vgone(vp);
1455 		}
1456 		break;
1457 #endif
1458 	default:
1459 		break;
1460 	}
1461 	return (0);
1462 }
1463 
1464 int
1465 kernfs_reclaim(v)
1466 	void *v;
1467 {
1468 	struct vop_reclaim_args /* {
1469 		struct vnode *a_vp;
1470 	} */ *ap = v;
1471 
1472 	return (kernfs_freevp(ap->a_vp));
1473 }
1474 
1475 /*
1476  * Return POSIX pathconf information applicable to special devices.
1477  */
1478 int
1479 kernfs_pathconf(v)
1480 	void *v;
1481 {
1482 	struct vop_pathconf_args /* {
1483 		struct vnode *a_vp;
1484 		int a_name;
1485 		register_t *a_retval;
1486 	} */ *ap = v;
1487 
1488 	switch (ap->a_name) {
1489 	case _PC_LINK_MAX:
1490 		*ap->a_retval = LINK_MAX;
1491 		return (0);
1492 	case _PC_MAX_CANON:
1493 		*ap->a_retval = MAX_CANON;
1494 		return (0);
1495 	case _PC_MAX_INPUT:
1496 		*ap->a_retval = MAX_INPUT;
1497 		return (0);
1498 	case _PC_PIPE_BUF:
1499 		*ap->a_retval = PIPE_BUF;
1500 		return (0);
1501 	case _PC_CHOWN_RESTRICTED:
1502 		*ap->a_retval = 1;
1503 		return (0);
1504 	case _PC_VDISABLE:
1505 		*ap->a_retval = _POSIX_VDISABLE;
1506 		return (0);
1507 	case _PC_SYNC_IO:
1508 		*ap->a_retval = 1;
1509 		return (0);
1510 	default:
1511 		return (EINVAL);
1512 	}
1513 	/* NOTREACHED */
1514 }
1515 
1516 /*
1517  * Print out the contents of a /dev/fd vnode.
1518  */
1519 /* ARGSUSED */
1520 int
1521 kernfs_print(void *v __unused)
1522 {
1523 
1524 	printf("tag VT_KERNFS, kernfs vnode\n");
1525 	return (0);
1526 }
1527 
1528 int
1529 kernfs_link(v)
1530 	void *v;
1531 {
1532 	struct vop_link_args /* {
1533 		struct vnode *a_dvp;
1534 		struct vnode *a_vp;
1535 		struct componentname *a_cnp;
1536 	} */ *ap = v;
1537 
1538 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1539 	vput(ap->a_dvp);
1540 	return (EROFS);
1541 }
1542 
1543 int
1544 kernfs_symlink(v)
1545 	void *v;
1546 {
1547 	struct vop_symlink_args /* {
1548 		struct vnode *a_dvp;
1549 		struct vnode **a_vpp;
1550 		struct componentname *a_cnp;
1551 		struct vattr *a_vap;
1552 		char *a_target;
1553 	} */ *ap = v;
1554 
1555 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1556 	vput(ap->a_dvp);
1557 	return (EROFS);
1558 }
1559