xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision 6dffe8d42bd46273f674d7ab834e7be9b1af990e)
1 /*	$NetBSD: kernfs_vnops.c,v 1.138 2009/07/03 21:17:41 elad 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 1.138 2009/07/03 21:17:41 elad 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(struct kernfs_node *kfs, int off, char **bufp, size_t len, size_t *wrlen)
384 {
385 	const struct kern_target *kt;
386 #ifdef IPSEC
387 	struct mbuf *m;
388 #endif
389 	int err;
390 
391 	kt = kfs->kfs_kt;
392 
393 	switch (kfs->kfs_type) {
394 	case KFStime: {
395 		struct timeval tv;
396 
397 		microtime(&tv);
398 		snprintf(*bufp, len, "%lld %ld\n", (long long)tv.tv_sec,
399 		    (long)tv.tv_usec);
400 		break;
401 	}
402 
403 	case KFSint: {
404 		int *ip = kt->kt_data;
405 
406 		snprintf(*bufp, len, "%d\n", *ip);
407 		break;
408 	}
409 
410 	case KFSstring: {
411 		char *cp = kt->kt_data;
412 
413 		*bufp = cp;
414 		break;
415 	}
416 
417 	case KFSmsgbuf: {
418 		long n;
419 
420 		/*
421 		 * deal with cases where the message buffer has
422 		 * become corrupted.
423 		 */
424 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
425 			msgbufenabled = 0;
426 			return (ENXIO);
427 		}
428 
429 		/*
430 		 * Note that reads of /kern/msgbuf won't necessarily yield
431 		 * consistent results, if the message buffer is modified
432 		 * while the read is in progress.  The worst that can happen
433 		 * is that incorrect data will be read.  There's no way
434 		 * that this can crash the system unless the values in the
435 		 * message buffer header are corrupted, but that'll cause
436 		 * the system to die anyway.
437 		 */
438 		if (off >= msgbufp->msg_bufs) {
439 			*wrlen = 0;
440 			return (0);
441 		}
442 		n = msgbufp->msg_bufx + off;
443 		if (n >= msgbufp->msg_bufs)
444 			n -= msgbufp->msg_bufs;
445 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
446 		*bufp = msgbufp->msg_bufc + n;
447 		*wrlen = len;
448 		return (0);
449 	}
450 
451 	case KFShostname: {
452 		char *cp = hostname;
453 		size_t xlen = hostnamelen;
454 
455 		if (xlen >= (len - 2))
456 			return (EINVAL);
457 
458 		memcpy(*bufp, cp, xlen);
459 		(*bufp)[xlen] = '\n';
460 		(*bufp)[xlen+1] = '\0';
461 		len = strlen(*bufp);
462 		break;
463 	}
464 
465 	case KFSavenrun:
466 		averunnable.fscale = FSCALE;
467 		snprintf(*bufp, len, "%d %d %d %ld\n",
468 		    averunnable.ldavg[0], averunnable.ldavg[1],
469 		    averunnable.ldavg[2], averunnable.fscale);
470 		break;
471 
472 #ifdef IPSEC
473 	case KFSipsecsa:
474 		if (key_setdumpsa_spi == NULL)
475 			return 0;
476 		/*
477 		 * Note that SA configuration could be changed during the
478 		 * read operation, resulting in garbled output.
479 		 */
480 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
481 		if (!m)
482 			return (ENOBUFS);
483 		if (off >= m->m_pkthdr.len) {
484 			*wrlen = 0;
485 			m_freem(m);
486 			return (0);
487 		}
488 		if (len > m->m_pkthdr.len - off)
489 			len = m->m_pkthdr.len - off;
490 		m_copydata(m, off, len, *bufp);
491 		*wrlen = len;
492 		m_freem(m);
493 		return (0);
494 
495 	case KFSipsecsp:
496 		/*
497 		 * Note that SP configuration could be changed during the
498 		 * read operation, resulting in garbled output.
499 		 */
500 		if (key_getspbyid == NULL)
501 			return 0;
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(const struct kernfs_node *kfs, char *bf, size_t len)
547 {
548 
549 	switch (kfs->kfs_type) {
550 	case KFShostname:
551 		if (bf[len-1] == '\n')
552 			--len;
553 		memcpy(hostname, bf, len);
554 		hostname[len] = '\0';
555 		hostnamelen = (size_t) len;
556 		return (0);
557 
558 	default:
559 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
560 	}
561 }
562 
563 
564 /*
565  * vp is the current namei directory
566  * ndp is the name to locate in that directory...
567  */
568 int
569 kernfs_lookup(void *v)
570 {
571 	struct vop_lookup_args /* {
572 		struct vnode * a_dvp;
573 		struct vnode ** a_vpp;
574 		struct componentname * a_cnp;
575 	} */ *ap = v;
576 	struct componentname *cnp = ap->a_cnp;
577 	struct vnode **vpp = ap->a_vpp;
578 	struct vnode *dvp = ap->a_dvp;
579 	const char *pname = cnp->cn_nameptr;
580 	const struct kernfs_node *kfs;
581 	const struct kern_target *kt;
582 	const struct dyn_kern_target *dkt;
583 	const struct kernfs_subdir *ks;
584 	int error, i;
585 #ifdef IPSEC
586 	char *ep;
587 	u_int32_t id;
588 #endif
589 
590 	*vpp = NULLVP;
591 
592 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
593 		return (EROFS);
594 
595 	if (cnp->cn_namelen == 1 && *pname == '.') {
596 		*vpp = dvp;
597 		VREF(dvp);
598 		return (0);
599 	}
600 
601 	kfs = VTOKERN(dvp);
602 	switch (kfs->kfs_type) {
603 	case KFSkern:
604 		/*
605 		 * Shouldn't get here with .. in the root node.
606 		 */
607 		if (cnp->cn_flags & ISDOTDOT)
608 			return (EIO);
609 
610 		for (i = 0; i < static_nkern_targets; i++) {
611 			kt = &kern_targets[i];
612 			if (cnp->cn_namelen == kt->kt_namlen &&
613 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
614 				goto found;
615 		}
616 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
617 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
618 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
619 				kt = &dkt->dkt_kt;
620 				goto found;
621 			}
622 		}
623 		break;
624 
625 	found:
626 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
627 		return (error);
628 
629 	case KFSsubdir:
630 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
631 		if (cnp->cn_flags & ISDOTDOT) {
632 			kt = ks->ks_parent;
633 			goto found;
634 		}
635 
636 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
637 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
638 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
639 				kt = &dkt->dkt_kt;
640 				goto found;
641 			}
642 		}
643 		break;
644 
645 #ifdef IPSEC
646 	case KFSipsecsadir:
647 		if (cnp->cn_flags & ISDOTDOT) {
648 			kt = &kern_targets[0];
649 			goto found;
650 		}
651 
652 		for (i = 2; i < nipsecsa_targets; i++) {
653 			kt = &ipsecsa_targets[i];
654 			if (cnp->cn_namelen == kt->kt_namlen &&
655 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
656 				goto found;
657 		}
658 
659 		ep = NULL;
660 		id = strtoul(pname, &ep, 10);
661 		if (!ep || *ep || ep == pname)
662 			break;
663 
664 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id);
665 		return (error);
666 
667 	case KFSipsecspdir:
668 		if (cnp->cn_flags & ISDOTDOT) {
669 			kt = &kern_targets[0];
670 			goto found;
671 		}
672 
673 		for (i = 2; i < nipsecsp_targets; i++) {
674 			kt = &ipsecsp_targets[i];
675 			if (cnp->cn_namelen == kt->kt_namlen &&
676 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
677 				goto found;
678 		}
679 
680 		ep = NULL;
681 		id = strtoul(pname, &ep, 10);
682 		if (!ep || *ep || ep == pname)
683 			break;
684 
685 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id);
686 		return (error);
687 #endif
688 
689 	default:
690 		return (ENOTDIR);
691 	}
692 
693 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
694 }
695 
696 int
697 kernfs_open(void *v)
698 {
699 	struct vop_open_args /* {
700 		struct vnode *a_vp;
701 		int a_mode;
702 		kauth_cred_t a_cred;
703 	} */ *ap = v;
704 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
705 #ifdef IPSEC
706 	struct mbuf *m;
707 	struct secpolicy *sp;
708 #endif
709 
710 	switch (kfs->kfs_type) {
711 #ifdef IPSEC
712 	case KFSipsecsa:
713 		if (key_setdumpsa_spi == NULL)
714 			return 0;
715 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
716 		if (m) {
717 			m_freem(m);
718 			return (0);
719 		} else
720 			return (ENOENT);
721 
722 	case KFSipsecsp:
723 		if (key_getspbyid == NULL)
724 			return 0;
725 		sp = key_getspbyid(kfs->kfs_value);
726 		if (sp) {
727 			kfs->kfs_v = sp;
728 			return (0);
729 		} else
730 			return (ENOENT);
731 #endif
732 
733 	default:
734 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN,
735 		    v, 0);
736 	}
737 }
738 
739 int
740 kernfs_close(void *v)
741 {
742 	struct vop_close_args /* {
743 		struct vnode *a_vp;
744 		int a_fflag;
745 		kauth_cred_t a_cred;
746 	} */ *ap = v;
747 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
748 
749 	switch (kfs->kfs_type) {
750 #ifdef IPSEC
751 	case KFSipsecsp:
752 		if (key_freesp == NULL)
753 			return 0;
754 		key_freesp((struct secpolicy *)kfs->kfs_v);
755 		break;
756 #endif
757 
758 	default:
759 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE,
760 		    v, 0);
761 	}
762 
763 	return (0);
764 }
765 
766 static int
767 kernfs_check_possible(struct vnode *vp, mode_t mode)
768 {
769 
770 	return 0;
771 }
772 
773 static int
774 kernfs_check_permitted(struct vattr *va, mode_t mode, kauth_cred_t cred)
775 {
776 
777 	return genfs_can_access(va->va_type, va->va_mode, va->va_uid, va->va_gid,
778 	    mode, cred);
779 }
780 
781 int
782 kernfs_access(void *v)
783 {
784 	struct vop_access_args /* {
785 		struct vnode *a_vp;
786 		int a_mode;
787 		kauth_cred_t a_cred;
788 	} */ *ap = v;
789 	struct vattr va;
790 	int error;
791 
792 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
793 		return (error);
794 
795 	error = kernfs_check_possible(ap->a_vp, ap->a_mode);
796 	if (error)
797 		return error;
798 
799 	error = kernfs_check_permitted(&va, ap->a_mode, ap->a_cred);
800 
801 	return error;
802 }
803 
804 static int
805 kernfs_default_fileop_getattr(void *v)
806 {
807 	struct vop_getattr_args /* {
808 		struct vnode *a_vp;
809 		struct vattr *a_vap;
810 		kauth_cred_t a_cred;
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(void *v)
822 {
823 	struct vop_getattr_args /* {
824 		struct vnode *a_vp;
825 		struct vattr *a_vap;
826 		kauth_cred_t a_cred;
827 	} */ *ap = v;
828 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
829 	struct kernfs_subdir *ks;
830 	struct vattr *vap = ap->a_vap;
831 	int error = 0;
832 	char strbuf[KSTRING], *bf;
833 	size_t nread, total;
834 
835 	VATTR_NULL(vap);
836 	vap->va_type = ap->a_vp->v_type;
837 	vap->va_uid = 0;
838 	vap->va_gid = 0;
839 	vap->va_mode = kfs->kfs_mode;
840 	vap->va_fileid = kfs->kfs_fileno;
841 	vap->va_flags = 0;
842 	vap->va_size = 0;
843 	vap->va_blocksize = DEV_BSIZE;
844 	/* Make all times be current TOD, except for the "boottime" node. */
845 	if (kfs->kfs_kt->kt_namlen == 8 &&
846 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
847 		vap->va_ctime = boottime;
848 	} else {
849 		getnanotime(&vap->va_ctime);
850 	}
851 	vap->va_atime = vap->va_mtime = vap->va_ctime;
852 	vap->va_gen = 0;
853 	vap->va_flags = 0;
854 	vap->va_rdev = 0;
855 	vap->va_bytes = 0;
856 
857 	switch (kfs->kfs_type) {
858 	case KFSkern:
859 		vap->va_nlink = nkern_dirs;
860 		vap->va_bytes = vap->va_size = DEV_BSIZE;
861 		break;
862 
863 	case KFSroot:
864 		vap->va_nlink = 1;
865 		vap->va_bytes = vap->va_size = DEV_BSIZE;
866 		break;
867 
868 	case KFSsubdir:
869 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
870 		vap->va_nlink = ks->ks_dirs;
871 		vap->va_bytes = vap->va_size = DEV_BSIZE;
872 		break;
873 
874 	case KFSnull:
875 	case KFStime:
876 	case KFSint:
877 	case KFSstring:
878 	case KFShostname:
879 	case KFSavenrun:
880 	case KFSdevice:
881 	case KFSmsgbuf:
882 #ifdef IPSEC
883 	case KFSipsecsa:
884 	case KFSipsecsp:
885 #endif
886 		vap->va_nlink = 1;
887 		total = 0;
888 		do {
889 			bf = strbuf;
890 			error = kernfs_xread(kfs, total, &bf,
891 			    sizeof(strbuf), &nread);
892 			total += nread;
893 		} while (error == 0 && nread != 0);
894 		vap->va_bytes = vap->va_size = total;
895 		break;
896 
897 #ifdef IPSEC
898 	case KFSipsecsadir:
899 	case KFSipsecspdir:
900 		vap->va_nlink = 2;
901 		vap->va_bytes = vap->va_size = DEV_BSIZE;
902 		break;
903 #endif
904 
905 	default:
906 		error = kernfs_try_fileop(kfs->kfs_type,
907 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
908 		break;
909 	}
910 
911 	return (error);
912 }
913 
914 /*ARGSUSED*/
915 int
916 kernfs_setattr(void *v)
917 {
918 
919 	/*
920 	 * Silently ignore attribute changes.
921 	 * This allows for open with truncate to have no
922 	 * effect until some data is written.  I want to
923 	 * do it this way because all writes are atomic.
924 	 */
925 	return (0);
926 }
927 
928 int
929 kernfs_default_xread(void *v)
930 {
931 	struct vop_read_args /* {
932 		struct vnode *a_vp;
933 		struct uio *a_uio;
934 		int  a_ioflag;
935 		kauth_cred_t a_cred;
936 	} */ *ap = v;
937 	struct uio *uio = ap->a_uio;
938 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
939 	char strbuf[KSTRING], *bf;
940 	int off;
941 	size_t len;
942 	int error;
943 
944 	if (ap->a_vp->v_type == VDIR)
945 		return (EOPNOTSUPP);
946 
947 	off = (int)uio->uio_offset;
948 	/* Don't allow negative offsets */
949 	if (off < 0)
950 		return EINVAL;
951 
952 	bf = strbuf;
953 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
954 		error = uiomove(bf, len, uio);
955 	return (error);
956 }
957 
958 int
959 kernfs_read(void *v)
960 {
961 	struct vop_read_args /* {
962 		struct vnode *a_vp;
963 		struct uio *a_uio;
964 		int  a_ioflag;
965 		struct ucred *a_cred;
966 	} */ *ap = v;
967 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
968 
969 	if (kfs->kfs_type < KFSlasttype) {
970 		/* use default function */
971 		return kernfs_default_xread(v);
972 	}
973 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
974 	   EOPNOTSUPP);
975 }
976 
977 static int
978 kernfs_default_xwrite(void *v)
979 {
980 	struct vop_write_args /* {
981 		struct vnode *a_vp;
982 		struct uio *a_uio;
983 		int  a_ioflag;
984 		kauth_cred_t a_cred;
985 	} */ *ap = v;
986 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
987 	struct uio *uio = ap->a_uio;
988 	int error;
989 	size_t xlen;
990 	char strbuf[KSTRING];
991 
992 	if (uio->uio_offset != 0)
993 		return (EINVAL);
994 
995 	xlen = min(uio->uio_resid, KSTRING-1);
996 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
997 		return (error);
998 
999 	if (uio->uio_resid != 0)
1000 		return (EIO);
1001 
1002 	strbuf[xlen] = '\0';
1003 	xlen = strlen(strbuf);
1004 	return (kernfs_xwrite(kfs, strbuf, xlen));
1005 }
1006 
1007 int
1008 kernfs_write(void *v)
1009 {
1010 	struct vop_write_args /* {
1011 		struct vnode *a_vp;
1012 		struct uio *a_uio;
1013 		int  a_ioflag;
1014 		kauth_cred_t a_cred;
1015 	} */ *ap = v;
1016 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1017 
1018 	if (kfs->kfs_type < KFSlasttype) {
1019 		/* use default function */
1020 		return kernfs_default_xwrite(v);
1021 	}
1022 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
1023 	    EOPNOTSUPP);
1024 }
1025 
1026 int
1027 kernfs_ioctl(void *v)
1028 {
1029 	struct vop_ioctl_args /* {
1030 		const struct vnodeop_desc *a_desc;
1031 		struct vnode *a_vp;
1032 		u_long a_command;
1033 		void *a_data;
1034 		int a_fflag;
1035 		kauth_cred_t a_cred;
1036 	} */ *ap = v;
1037 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1038 
1039 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1040 	    EPASSTHROUGH);
1041 }
1042 
1043 static int
1044 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1045     u_int32_t value, struct vop_readdir_args *ap)
1046 {
1047 	struct kernfs_node *kfs;
1048 	struct vnode *vp;
1049 	int error;
1050 
1051 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1052 	    value)) != 0)
1053 		return error;
1054 	if (kt->kt_tag == KFSdevice) {
1055 		struct vattr va;
1056 
1057 		error = VOP_GETATTR(vp, &va, ap->a_cred);
1058 		if (error != 0) {
1059 			return error;
1060 		}
1061 		d->d_fileno = va.va_fileid;
1062 	} else {
1063 		kfs = VTOKERN(vp);
1064 		d->d_fileno = kfs->kfs_fileno;
1065 	}
1066 	vput(vp);
1067 	return 0;
1068 }
1069 
1070 static int
1071 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1072     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1073     const struct kern_target *kt, struct vop_readdir_args *ap)
1074 {
1075 	const struct kern_target *ikt;
1076 	int error;
1077 
1078 	switch (entry) {
1079 	case 0:
1080 		d->d_fileno = thisdir_kfs->kfs_fileno;
1081 		return 0;
1082 	case 1:
1083 		ikt = parent_kt;
1084 		break;
1085 	default:
1086 		ikt = kt;
1087 		break;
1088 	}
1089 	if (ikt != thisdir_kfs->kfs_kt) {
1090 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1091 			return error;
1092 	} else
1093 		d->d_fileno = thisdir_kfs->kfs_fileno;
1094 	return 0;
1095 }
1096 
1097 int
1098 kernfs_readdir(void *v)
1099 {
1100 	struct vop_readdir_args /* {
1101 		struct vnode *a_vp;
1102 		struct uio *a_uio;
1103 		kauth_cred_t a_cred;
1104 		int *a_eofflag;
1105 		off_t **a_cookies;
1106 		int a_*ncookies;
1107 	} */ *ap = v;
1108 	struct uio *uio = ap->a_uio;
1109 	struct dirent d;
1110 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1111 	const struct kern_target *kt;
1112 	const struct dyn_kern_target *dkt = NULL;
1113 	const struct kernfs_subdir *ks;
1114 	off_t i, j;
1115 	int error;
1116 	off_t *cookies = NULL;
1117 	int ncookies = 0, n;
1118 #ifdef IPSEC
1119 	struct secasvar *sav, *sav2;
1120 	struct secpolicy *sp;
1121 #endif
1122 
1123 	if (uio->uio_resid < UIO_MX)
1124 		return (EINVAL);
1125 	if (uio->uio_offset < 0)
1126 		return (EINVAL);
1127 
1128 	error = 0;
1129 	i = uio->uio_offset;
1130 	memset(&d, 0, sizeof(d));
1131 	d.d_reclen = UIO_MX;
1132 	ncookies = uio->uio_resid / UIO_MX;
1133 
1134 	switch (kfs->kfs_type) {
1135 	case KFSkern:
1136 		if (i >= nkern_targets)
1137 			return (0);
1138 
1139 		if (ap->a_ncookies) {
1140 			ncookies = min(ncookies, (nkern_targets - i));
1141 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1142 			    M_WAITOK);
1143 			*ap->a_cookies = cookies;
1144 		}
1145 
1146 		n = 0;
1147 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1148 			if (i < static_nkern_targets)
1149 				kt = &kern_targets[i];
1150 			else {
1151 				if (dkt == NULL) {
1152 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1153 					for (j = static_nkern_targets; j < i &&
1154 						     dkt != NULL; j++)
1155 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1156 					if (j != i)
1157 						break;
1158 				} else {
1159 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1160 				}
1161 				if (dkt == NULL)
1162 					break;
1163 				kt = &dkt->dkt_kt;
1164 			}
1165 			if (kt->kt_tag == KFSdevice) {
1166 				dev_t *dp = kt->kt_data;
1167 				struct vnode *fvp;
1168 
1169 				if (*dp == NODEV ||
1170 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1171 					continue;
1172 			}
1173 			d.d_namlen = kt->kt_namlen;
1174 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1175 			    &kern_targets[0], kt, ap)) != 0)
1176 				break;
1177 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1178 			d.d_type = kt->kt_type;
1179 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1180 				break;
1181 			if (cookies)
1182 				*cookies++ = i + 1;
1183 			n++;
1184 		}
1185 		ncookies = n;
1186 		break;
1187 
1188 	case KFSroot:
1189 		if (i >= 2)
1190 			return 0;
1191 
1192 		if (ap->a_ncookies) {
1193 			ncookies = min(ncookies, (2 - i));
1194 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1195 			    M_WAITOK);
1196 			*ap->a_cookies = cookies;
1197 		}
1198 
1199 		n = 0;
1200 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1201 			kt = &kern_targets[i];
1202 			d.d_namlen = kt->kt_namlen;
1203 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
1204 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1205 			d.d_type = kt->kt_type;
1206 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1207 				break;
1208 			if (cookies)
1209 				*cookies++ = i + 1;
1210 			n++;
1211 		}
1212 		ncookies = n;
1213 		break;
1214 
1215 	case KFSsubdir:
1216 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
1217 		if (i >= ks->ks_nentries)
1218 			return (0);
1219 
1220 		if (ap->a_ncookies) {
1221 			ncookies = min(ncookies, (ks->ks_nentries - i));
1222 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1223 			    M_WAITOK);
1224 			*ap->a_cookies = cookies;
1225 		}
1226 
1227 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1228 		for (j = 0; j < i && dkt != NULL; j++)
1229 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1230 		n = 0;
1231 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1232 			if (i < 2)
1233 				kt = &subdir_targets[i];
1234 			else {
1235 				/* check if ks_nentries lied to us */
1236 				if (dkt == NULL)
1237 					break;
1238 				kt = &dkt->dkt_kt;
1239 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1240 			}
1241 			if (kt->kt_tag == KFSdevice) {
1242 				dev_t *dp = kt->kt_data;
1243 				struct vnode *fvp;
1244 
1245 				if (*dp == NODEV ||
1246 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1247 					continue;
1248 			}
1249 			d.d_namlen = kt->kt_namlen;
1250 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1251 			    ks->ks_parent, kt, ap)) != 0)
1252 				break;
1253 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1254 			d.d_type = kt->kt_type;
1255 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1256 				break;
1257 			if (cookies)
1258 				*cookies++ = i + 1;
1259 			n++;
1260 		}
1261 		ncookies = n;
1262 		break;
1263 
1264 #ifdef IPSEC
1265 	case KFSipsecsadir:
1266 		/* count SA in the system */
1267 		n = 0;
1268 		if (&satailq == NULL)
1269 			return 0;
1270 		TAILQ_FOREACH(sav, &satailq, tailq) {
1271 			for (sav2 = TAILQ_FIRST(&satailq);
1272 			    sav2 != sav;
1273 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1274 				if (sav->spi == sav2->spi) {
1275 					/* multiple SA with same SPI */
1276 					break;
1277 				}
1278 			}
1279 			if (sav == sav2 || sav->spi != sav2->spi)
1280 				n++;
1281 		}
1282 
1283 		if (i >= nipsecsa_targets + n)
1284 			return (0);
1285 
1286 		if (ap->a_ncookies) {
1287 			ncookies = min(ncookies, (n - i));
1288 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1289 			    M_WAITOK);
1290 			*ap->a_cookies = cookies;
1291 		}
1292 
1293 		n = 0;
1294 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1295 			kt = &ipsecsa_targets[i];
1296 			d.d_namlen = kt->kt_namlen;
1297 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1298 			    &kern_targets[0], kt, ap)) != 0)
1299 				break;
1300 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1301 			d.d_type = kt->kt_type;
1302 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1303 				break;
1304 			if (cookies)
1305 				*cookies++ = i + 1;
1306 			n++;
1307 		}
1308 		if (error) {
1309 			ncookies = n;
1310 			break;
1311 		}
1312 
1313 		TAILQ_FOREACH(sav, &satailq, tailq) {
1314 			for (sav2 = TAILQ_FIRST(&satailq);
1315 			    sav2 != sav;
1316 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1317 				if (sav->spi == sav2->spi) {
1318 					/* multiple SA with same SPI */
1319 					break;
1320 				}
1321 			}
1322 			if (sav != sav2 && sav->spi == sav2->spi)
1323 				continue;
1324 			if (uio->uio_resid < UIO_MX)
1325 				break;
1326 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1327 			    sav->spi, ap)) != 0)
1328 				break;
1329 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1330 			    "%u", ntohl(sav->spi));
1331 			d.d_type = DT_REG;
1332 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1333 				break;
1334 			if (cookies)
1335 				*cookies++ = i + 1;
1336 			n++;
1337 			i++;
1338 		}
1339 		ncookies = n;
1340 		break;
1341 
1342 	case KFSipsecspdir:
1343 		/* count SP in the system */
1344 		if (&sptailq == NULL)
1345 			return 0;
1346 
1347 		n = 0;
1348 		TAILQ_FOREACH(sp, &sptailq, tailq)
1349 			n++;
1350 
1351 		if (i >= nipsecsp_targets + n)
1352 			return (0);
1353 
1354 		if (ap->a_ncookies) {
1355 			ncookies = min(ncookies, (n - i));
1356 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1357 			    M_WAITOK);
1358 			*ap->a_cookies = cookies;
1359 		}
1360 
1361 		n = 0;
1362 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1363 			kt = &ipsecsp_targets[i];
1364 			d.d_namlen = kt->kt_namlen;
1365 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1366 			    &kern_targets[0], kt, ap)) != 0)
1367 				break;
1368 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1369 			d.d_type = kt->kt_type;
1370 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1371 				break;
1372 			if (cookies)
1373 				*cookies++ = i + 1;
1374 			n++;
1375 		}
1376 		if (error) {
1377 			ncookies = n;
1378 			break;
1379 		}
1380 
1381 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1382 			if (uio->uio_resid < UIO_MX)
1383 				break;
1384 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1385 			    sp->id, ap)) != 0)
1386 				break;
1387 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1388 			    "%u", sp->id);
1389 			d.d_type = DT_REG;
1390 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1391 				break;
1392 			if (cookies)
1393 				*cookies++ = i + 1;
1394 			n++;
1395 			i++;
1396 		}
1397 		ncookies = n;
1398 		break;
1399 #endif
1400 
1401 	default:
1402 		error = ENOTDIR;
1403 		break;
1404 	}
1405 
1406 	if (ap->a_ncookies) {
1407 		if (error) {
1408 			if (cookies)
1409 				free(*ap->a_cookies, M_TEMP);
1410 			*ap->a_ncookies = 0;
1411 			*ap->a_cookies = NULL;
1412 		} else
1413 			*ap->a_ncookies = ncookies;
1414 	}
1415 
1416 	uio->uio_offset = i;
1417 	return (error);
1418 }
1419 
1420 int
1421 kernfs_inactive(void *v)
1422 {
1423 	struct vop_inactive_args /* {
1424 		struct vnode *a_vp;
1425 		bool *a_recycle;
1426 	} */ *ap = v;
1427 	struct vnode *vp = ap->a_vp;
1428 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1429 #ifdef IPSEC
1430 	struct mbuf *m;
1431 	struct secpolicy *sp;
1432 #endif
1433 
1434 	*ap->a_recycle = false;
1435 	switch (kfs->kfs_type) {
1436 #ifdef IPSEC
1437 	case KFSipsecsa:
1438 		if (key_setdumpsa_spi == NULL)
1439 			return 0;
1440 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1441 		if (m)
1442 			m_freem(m);
1443 		else
1444 			*ap->a_recycle = true;
1445 		break;
1446 	case KFSipsecsp:
1447 		if (key_getspbyid == NULL)
1448 			return 0;
1449 		sp = key_getspbyid(kfs->kfs_value);
1450 		if (sp)
1451 			key_freesp(sp);
1452 		else {
1453 			*ap->a_recycle = true;
1454 		}
1455 		break;
1456 #endif
1457 	default:
1458 		break;
1459 	}
1460 	VOP_UNLOCK(vp, 0);
1461 	return (0);
1462 }
1463 
1464 int
1465 kernfs_reclaim(void *v)
1466 {
1467 	struct vop_reclaim_args /* {
1468 		struct vnode *a_vp;
1469 	} */ *ap = v;
1470 
1471 	return (kernfs_freevp(ap->a_vp));
1472 }
1473 
1474 /*
1475  * Return POSIX pathconf information applicable to special devices.
1476  */
1477 int
1478 kernfs_pathconf(void *v)
1479 {
1480 	struct vop_pathconf_args /* {
1481 		struct vnode *a_vp;
1482 		int a_name;
1483 		register_t *a_retval;
1484 	} */ *ap = v;
1485 
1486 	switch (ap->a_name) {
1487 	case _PC_LINK_MAX:
1488 		*ap->a_retval = LINK_MAX;
1489 		return (0);
1490 	case _PC_MAX_CANON:
1491 		*ap->a_retval = MAX_CANON;
1492 		return (0);
1493 	case _PC_MAX_INPUT:
1494 		*ap->a_retval = MAX_INPUT;
1495 		return (0);
1496 	case _PC_PIPE_BUF:
1497 		*ap->a_retval = PIPE_BUF;
1498 		return (0);
1499 	case _PC_CHOWN_RESTRICTED:
1500 		*ap->a_retval = 1;
1501 		return (0);
1502 	case _PC_VDISABLE:
1503 		*ap->a_retval = _POSIX_VDISABLE;
1504 		return (0);
1505 	case _PC_SYNC_IO:
1506 		*ap->a_retval = 1;
1507 		return (0);
1508 	default:
1509 		return (EINVAL);
1510 	}
1511 	/* NOTREACHED */
1512 }
1513 
1514 /*
1515  * Print out the contents of a /dev/fd vnode.
1516  */
1517 /* ARGSUSED */
1518 int
1519 kernfs_print(void *v)
1520 {
1521 
1522 	printf("tag VT_KERNFS, kernfs vnode\n");
1523 	return (0);
1524 }
1525 
1526 int
1527 kernfs_link(void *v)
1528 {
1529 	struct vop_link_args /* {
1530 		struct vnode *a_dvp;
1531 		struct vnode *a_vp;
1532 		struct componentname *a_cnp;
1533 	} */ *ap = v;
1534 
1535 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1536 	vput(ap->a_dvp);
1537 	return (EROFS);
1538 }
1539 
1540 int
1541 kernfs_symlink(void *v)
1542 {
1543 	struct vop_symlink_args /* {
1544 		struct vnode *a_dvp;
1545 		struct vnode **a_vpp;
1546 		struct componentname *a_cnp;
1547 		struct vattr *a_vap;
1548 		char *a_target;
1549 	} */ *ap = v;
1550 
1551 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1552 	vput(ap->a_dvp);
1553 	return (EROFS);
1554 }
1555