xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: kernfs_vnops.c,v 1.106 2004/10/27 06:55:53 skrll 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 1.106 2004/10/27 06:55:53 skrll 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 			/* XXX cast away const */
94      { DT_REG, N("copyright"), (void *)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 			/* XXX cast away const */
113      { DT_REG, N("version"),   (void *)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_xwrite(kfstype, const struct kernfs_node *, char *,
155     size_t, int);
156 
157 static int kernfs_default_xwrite(void *v);
158 static int kernfs_default_fileop_getattr(void *);
159 
160 /* must include all fileop's */
161 const struct kernfs_fileop kernfs_default_fileops[] = {
162   { .kf_fileop = KERNFS_XWRITE },
163   { .kf_fileop = KERNFS_FILEOP_OPEN },
164   { .kf_fileop = KERNFS_FILEOP_GETATTR,
165     .kf_genop = {kernfs_default_fileop_getattr} },
166   { .kf_fileop = KERNFS_FILEOP_IOCTL },
167   { .kf_fileop = KERNFS_FILEOP_MMAP },
168   { .kf_fileop = KERNFS_FILEOP_CLOSE },
169   { .kf_fileop = KERNFS_FILEOP_WRITE, .kf_genop = {kernfs_default_xwrite} },
170 };
171 
172 int	kernfs_lookup	__P((void *));
173 #define	kernfs_create	genfs_eopnotsupp
174 #define	kernfs_mknod	genfs_eopnotsupp
175 int	kernfs_open	__P((void *));
176 int	kernfs_close	__P((void *));
177 int	kernfs_access	__P((void *));
178 int	kernfs_getattr	__P((void *));
179 int	kernfs_setattr	__P((void *));
180 int	kernfs_read	__P((void *));
181 int	kernfs_write	__P((void *));
182 #define	kernfs_fcntl	genfs_fcntl
183 int	kernfs_ioctl	__P((void *));
184 #define	kernfs_poll	genfs_poll
185 #define kernfs_revoke	genfs_revoke
186 int	kernfs_mmap	__P((void *));
187 #define	kernfs_fsync	genfs_nullop
188 #define	kernfs_seek	genfs_nullop
189 #define	kernfs_remove	genfs_eopnotsupp
190 int	kernfs_link	__P((void *));
191 #define	kernfs_rename	genfs_eopnotsupp
192 #define	kernfs_mkdir	genfs_eopnotsupp
193 #define	kernfs_rmdir	genfs_eopnotsupp
194 int	kernfs_symlink	__P((void *));
195 int	kernfs_readdir	__P((void *));
196 #define	kernfs_readlink	genfs_eopnotsupp
197 #define	kernfs_abortop	genfs_abortop
198 int	kernfs_inactive	__P((void *));
199 int	kernfs_reclaim	__P((void *));
200 #define	kernfs_lock	genfs_lock
201 #define	kernfs_unlock	genfs_unlock
202 #define	kernfs_bmap	genfs_badop
203 #define	kernfs_strategy	genfs_badop
204 int	kernfs_print	__P((void *));
205 #define	kernfs_islocked	genfs_islocked
206 int	kernfs_pathconf	__P((void *));
207 #define	kernfs_advlock	genfs_einval
208 #define	kernfs_blkatoff	genfs_eopnotsupp
209 #define	kernfs_valloc	genfs_eopnotsupp
210 #define	kernfs_vfree	genfs_nullop
211 #define	kernfs_truncate	genfs_eopnotsupp
212 #define	kernfs_update	genfs_nullop
213 #define	kernfs_bwrite	genfs_eopnotsupp
214 #define	kernfs_putpages	genfs_putpages
215 
216 static int	kernfs_xread __P((struct kernfs_node *, int, char **, size_t, size_t *));
217 static int	kernfs_xwrite __P((const struct kernfs_node *, char *, size_t));
218 
219 int (**kernfs_vnodeop_p) __P((void *));
220 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
221 	{ &vop_default_desc, vn_default_error },
222 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
223 	{ &vop_create_desc, kernfs_create },		/* create */
224 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
225 	{ &vop_open_desc, kernfs_open },		/* open */
226 	{ &vop_close_desc, kernfs_close },		/* close */
227 	{ &vop_access_desc, kernfs_access },		/* access */
228 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
229 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
230 	{ &vop_read_desc, kernfs_read },		/* read */
231 	{ &vop_write_desc, kernfs_write },		/* write */
232 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
233 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
234 	{ &vop_poll_desc, kernfs_poll },		/* poll */
235 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
236 	{ &vop_mmap_desc, kernfs_mmap },		/* mmap */
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_blkatoff_desc, kernfs_blkatoff },	/* blkatoff */
259 	{ &vop_valloc_desc, kernfs_valloc },		/* valloc */
260 	{ &vop_vfree_desc, kernfs_vfree },		/* vfree */
261 	{ &vop_truncate_desc, kernfs_truncate },	/* truncate */
262 	{ &vop_update_desc, kernfs_update },		/* update */
263 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
264 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
265 	{ NULL, NULL }
266 };
267 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
268 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
269 
270 static __inline int
271 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
272 {
273 	if (a->kf_type < b->kf_type)
274 		return -1;
275 	if (a->kf_type > b->kf_type)
276 		return 1;
277 	if (a->kf_fileop < b->kf_fileop)
278 		return -1;
279 	if (a->kf_fileop > b->kf_fileop)
280 		return 1;
281 	return (0);
282 }
283 
284 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
285 	SPLAY_INITIALIZER(kfsfileoptree);
286 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
287 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
288 
289 kfstype
290 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
291 {
292 	static u_char nextfreetype = KFSlasttype;
293 	struct kernfs_fileop *dkf, *fkf, skf;
294 	int i;
295 
296 	/* XXX need to keep track of dkf's memory if we support
297            deallocating types */
298 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
299 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
300 
301 	for (i = 0; i < sizeof(kernfs_default_fileops) /
302 		     sizeof(kernfs_default_fileops[0]); i++) {
303 		dkf[i].kf_type = nextfreetype;
304 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
305 	}
306 
307 	for (i = 0; i < nkf; i++) {
308 		skf.kf_type = nextfreetype;
309 		skf.kf_fileop = kf[i].kf_fileop;
310 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
311 			fkf->kf_genop = kf[i].kf_genop;
312 	}
313 
314 	return nextfreetype++;
315 }
316 
317 int
318 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
319 {
320 	struct kernfs_fileop *kf, skf;
321 
322 	skf.kf_type = type;
323 	skf.kf_fileop = fileop;
324 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
325 		if (kf->kf_vop)
326 			return kf->kf_vop(v);
327 	return error;
328 }
329 
330 int
331 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *buf,
332     size_t len, int error)
333 {
334 	struct kernfs_fileop *kf, skf;
335 
336 	skf.kf_type = type;
337 	skf.kf_fileop = KERNFS_XWRITE;
338 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
339 		if (kf->kf_xwrite)
340 			return kf->kf_xwrite(kfs, buf, len);
341 	return error;
342 }
343 
344 int
345 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
346 {
347 	struct kernfs_subdir *ks, *parent;
348 
349 	if (pkt == NULL) {
350 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
351 		nkern_targets++;
352 		if (dkt->dkt_kt.kt_vtype == VDIR)
353 			nkern_dirs++;
354 	} else {
355 		parent = (struct kernfs_subdir *)pkt->kt_data;
356 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
357 		parent->ks_nentries++;
358 		if (dkt->dkt_kt.kt_vtype == VDIR)
359 			parent->ks_dirs++;
360 	}
361 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
362 		ks = malloc(sizeof(struct kernfs_subdir),
363 		    M_TEMP, M_WAITOK);
364 		SIMPLEQ_INIT(&ks->ks_entries);
365 		ks->ks_nentries = 2; /* . and .. */
366 		ks->ks_dirs = 2;
367 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
368 		dkt->dkt_kt.kt_data = ks;
369 	}
370 	return 0;
371 }
372 
373 static int
374 kernfs_xread(kfs, off, bufp, len, wrlen)
375 	struct kernfs_node *kfs;
376 	int off;
377 	char **bufp;
378 	size_t len;
379 	size_t *wrlen;
380 {
381 	const struct kern_target *kt;
382 #ifdef IPSEC
383 	struct mbuf *m;
384 #endif
385 
386 	kt = kfs->kfs_kt;
387 
388 	switch (kfs->kfs_type) {
389 	case KFStime: {
390 		struct timeval tv;
391 
392 		microtime(&tv);
393 		snprintf(*bufp, len, "%ld %ld\n", tv.tv_sec, tv.tv_usec);
394 		break;
395 	}
396 
397 	case KFSint: {
398 		int *ip = kt->kt_data;
399 
400 		snprintf(*bufp, len, "%d\n", *ip);
401 		break;
402 	}
403 
404 	case KFSstring: {
405 		char *cp = kt->kt_data;
406 
407 		*bufp = cp;
408 		break;
409 	}
410 
411 	case KFSmsgbuf: {
412 		long n;
413 
414 		/*
415 		 * deal with cases where the message buffer has
416 		 * become corrupted.
417 		 */
418 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
419 			msgbufenabled = 0;
420 			return (ENXIO);
421 		}
422 
423 		/*
424 		 * Note that reads of /kern/msgbuf won't necessarily yield
425 		 * consistent results, if the message buffer is modified
426 		 * while the read is in progress.  The worst that can happen
427 		 * is that incorrect data will be read.  There's no way
428 		 * that this can crash the system unless the values in the
429 		 * message buffer header are corrupted, but that'll cause
430 		 * the system to die anyway.
431 		 */
432 		if (off >= msgbufp->msg_bufs) {
433 			*wrlen = 0;
434 			return (0);
435 		}
436 		n = msgbufp->msg_bufx + off;
437 		if (n >= msgbufp->msg_bufs)
438 			n -= msgbufp->msg_bufs;
439 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
440 		*bufp = msgbufp->msg_bufc + n;
441 		*wrlen = len;
442 		return (0);
443 	}
444 
445 	case KFShostname: {
446 		char *cp = hostname;
447 		int xlen = hostnamelen;
448 
449 		if (xlen >= (len - 2))
450 			return (EINVAL);
451 
452 		memcpy(*bufp, cp, xlen);
453 		(*bufp)[xlen] = '\n';
454 		(*bufp)[xlen+1] = '\0';
455 		len = strlen(*bufp);
456 		break;
457 	}
458 
459 	case KFSavenrun:
460 		averunnable.fscale = FSCALE;
461 		snprintf(*bufp, len, "%d %d %d %ld\n",
462 		    averunnable.ldavg[0], averunnable.ldavg[1],
463 		    averunnable.ldavg[2], averunnable.fscale);
464 		break;
465 
466 #ifdef IPSEC
467 	case KFSipsecsa:
468 		/*
469 		 * Note that SA configuration could be changed during the
470 		 * read operation, resulting in garbled output.
471 		 */
472 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
473 		if (!m)
474 			return (ENOBUFS);
475 		if (off >= m->m_pkthdr.len) {
476 			*wrlen = 0;
477 			m_freem(m);
478 			return (0);
479 		}
480 		if (len > m->m_pkthdr.len - off)
481 			len = m->m_pkthdr.len - off;
482 		m_copydata(m, off, len, *bufp);
483 		*wrlen = len;
484 		m_freem(m);
485 		return (0);
486 
487 	case KFSipsecsp:
488 		/*
489 		 * Note that SP configuration could be changed during the
490 		 * read operation, resulting in garbled output.
491 		 */
492 		if (!kfs->kfs_v) {
493 			struct secpolicy *sp;
494 
495 			sp = key_getspbyid(kfs->kfs_value);
496 			if (sp)
497 				kfs->kfs_v = sp;
498 			else
499 				return (ENOENT);
500 		}
501 		m = key_setdumpsp((struct secpolicy *)kfs->kfs_v,
502 		    SADB_X_SPDGET, 0, 0);
503 		if (!m)
504 			return (ENOBUFS);
505 		if (off >= m->m_pkthdr.len) {
506 			*wrlen = 0;
507 			m_freem(m);
508 			return (0);
509 		}
510 		if (len > m->m_pkthdr.len - off)
511 			len = m->m_pkthdr.len - off;
512 		m_copydata(m, off, len, *bufp);
513 		*wrlen = len;
514 		m_freem(m);
515 		return (0);
516 #endif
517 
518 	default:
519 		*wrlen = 0;
520 		return (0);
521 	}
522 
523 	len = strlen(*bufp);
524 	if (len <= off)
525 		*wrlen = 0;
526 	else {
527 		*bufp += off;
528 		*wrlen = len - off;
529 	}
530 	return (0);
531 }
532 
533 static int
534 kernfs_xwrite(kfs, buf, len)
535 	const struct kernfs_node *kfs;
536 	char *buf;
537 	size_t len;
538 {
539 
540 	switch (kfs->kfs_type) {
541 	case KFShostname:
542 		if (buf[len-1] == '\n')
543 			--len;
544 		memcpy(hostname, buf, len);
545 		hostname[len] = '\0';
546 		hostnamelen = (size_t) len;
547 		return (0);
548 
549 	default:
550 		return kernfs_try_xwrite(kfs->kfs_type, kfs, buf, len, EIO);
551 	}
552 }
553 
554 
555 /*
556  * vp is the current namei directory
557  * ndp is the name to locate in that directory...
558  */
559 int
560 kernfs_lookup(v)
561 	void *v;
562 {
563 	struct vop_lookup_args /* {
564 		struct vnode * a_dvp;
565 		struct vnode ** a_vpp;
566 		struct componentname * a_cnp;
567 	} */ *ap = v;
568 	struct componentname *cnp = ap->a_cnp;
569 	struct vnode **vpp = ap->a_vpp;
570 	struct vnode *dvp = ap->a_dvp;
571 	const char *pname = cnp->cn_nameptr;
572 	const struct kernfs_node *kfs;
573 	const struct kern_target *kt;
574 	const struct dyn_kern_target *dkt;
575 	const struct kernfs_subdir *ks;
576 	int error, i, wantpunlock;
577 #ifdef IPSEC
578 	char *ep;
579 	u_int32_t id;
580 #endif
581 
582 	*vpp = NULLVP;
583 	cnp->cn_flags &= ~PDIRUNLOCK;
584 
585 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
586 		return (EROFS);
587 
588 	if (cnp->cn_namelen == 1 && *pname == '.') {
589 		*vpp = dvp;
590 		VREF(dvp);
591 		return (0);
592 	}
593 
594 	wantpunlock = (~cnp->cn_flags & (LOCKPARENT | ISLASTCN));
595 	kfs = VTOKERN(dvp);
596 	switch (kfs->kfs_type) {
597 	case KFSkern:
598 		/*
599 		 * Shouldn't get here with .. in the root node.
600 		 */
601 		if (cnp->cn_flags & ISDOTDOT)
602 			return (EIO);
603 
604 		for (i = 0; i < static_nkern_targets; i++) {
605 			kt = &kern_targets[i];
606 			if (cnp->cn_namelen == kt->kt_namlen &&
607 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
608 				goto found;
609 		}
610 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
611 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
612 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
613 				kt = &dkt->dkt_kt;
614 				goto found;
615 			}
616 		}
617 		break;
618 
619 	found:
620 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
621 		if ((error == 0) && wantpunlock) {
622 			VOP_UNLOCK(dvp, 0);
623 			cnp->cn_flags |= PDIRUNLOCK;
624 		}
625 		return (error);
626 
627 	case KFSsubdir:
628 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
629 		if (cnp->cn_flags & ISDOTDOT) {
630 			kt = ks->ks_parent;
631 			goto found;
632 		}
633 
634 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
635 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
636 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
637 				kt = &dkt->dkt_kt;
638 				goto found;
639 			}
640 		}
641 		break;
642 
643 #ifdef IPSEC
644 	case KFSipsecsadir:
645 		if (cnp->cn_flags & ISDOTDOT) {
646 			kt = &kern_targets[0];
647 			goto found;
648 		}
649 
650 		for (i = 2; i < nipsecsa_targets; i++) {
651 			kt = &ipsecsa_targets[i];
652 			if (cnp->cn_namelen == kt->kt_namlen &&
653 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
654 				goto found;
655 		}
656 
657 		ep = NULL;
658 		id = strtoul(pname, &ep, 10);
659 		if (!ep || *ep || ep == pname)
660 			break;
661 
662 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id);
663 		if ((error == 0) && wantpunlock) {
664 			VOP_UNLOCK(dvp, 0);
665 			cnp->cn_flags |= PDIRUNLOCK;
666 		}
667 		return (error);
668 
669 	case KFSipsecspdir:
670 		if (cnp->cn_flags & ISDOTDOT) {
671 			kt = &kern_targets[0];
672 			goto found;
673 		}
674 
675 		for (i = 2; i < nipsecsp_targets; i++) {
676 			kt = &ipsecsp_targets[i];
677 			if (cnp->cn_namelen == kt->kt_namlen &&
678 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
679 				goto found;
680 		}
681 
682 		ep = NULL;
683 		id = strtoul(pname, &ep, 10);
684 		if (!ep || *ep || ep == pname)
685 			break;
686 
687 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id);
688 		if ((error == 0) && wantpunlock) {
689 			VOP_UNLOCK(dvp, 0);
690 			cnp->cn_flags |= PDIRUNLOCK;
691 		}
692 		return (error);
693 #endif
694 
695 	default:
696 		return (ENOTDIR);
697 	}
698 
699 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
700 }
701 
702 int
703 kernfs_open(v)
704 	void *v;
705 {
706 	struct vop_open_args /* {
707 		struct vnode *a_vp;
708 		int a_mode;
709 		struct ucred *a_cred;
710 		struct proc *a_p;
711 	} */ *ap = v;
712 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
713 #ifdef IPSEC
714 	struct mbuf *m;
715 	struct secpolicy *sp;
716 #endif
717 
718 	switch (kfs->kfs_type) {
719 #ifdef IPSEC
720 	case KFSipsecsa:
721 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
722 		if (m) {
723 			m_freem(m);
724 			return (0);
725 		} else
726 			return (ENOENT);
727 
728 	case KFSipsecsp:
729 		sp = key_getspbyid(kfs->kfs_value);
730 		if (sp) {
731 			kfs->kfs_v = sp;
732 			return (0);
733 		} else
734 			return (ENOENT);
735 #endif
736 
737 	default:
738 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN,
739 		    v, 0);
740 	}
741 }
742 
743 int
744 kernfs_close(v)
745 	void *v;
746 {
747 	struct vop_close_args /* {
748 		struct vnode *a_vp;
749 		int a_fflag;
750 		struct ucred *a_cred;
751 		struct proc *a_p;
752 	} */ *ap = v;
753 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
754 
755 	switch (kfs->kfs_type) {
756 #ifdef IPSEC
757 	case KFSipsecsp:
758 		key_freesp((struct secpolicy *)kfs->kfs_v);
759 		break;
760 #endif
761 
762 	default:
763 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE,
764 		    v, 0);
765 	}
766 
767 	return (0);
768 }
769 
770 int
771 kernfs_access(v)
772 	void *v;
773 {
774 	struct vop_access_args /* {
775 		struct vnode *a_vp;
776 		int a_mode;
777 		struct ucred *a_cred;
778 		struct proc *a_p;
779 	} */ *ap = v;
780 	struct vattr va;
781 	int error;
782 
783 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_p)) != 0)
784 		return (error);
785 
786 	return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid,
787 	    ap->a_mode, ap->a_cred));
788 }
789 
790 static int
791 kernfs_default_fileop_getattr(v)
792 	void *v;
793 {
794 	struct vop_getattr_args /* {
795 		struct vnode *a_vp;
796 		struct vattr *a_vap;
797 		struct ucred *a_cred;
798 		struct proc *a_p;
799 	} */ *ap = v;
800 	struct vattr *vap = ap->a_vap;
801 
802 	vap->va_nlink = 1;
803 	vap->va_bytes = vap->va_size = 0;
804 
805 	return 0;
806 }
807 
808 int
809 kernfs_getattr(v)
810 	void *v;
811 {
812 	struct vop_getattr_args /* {
813 		struct vnode *a_vp;
814 		struct vattr *a_vap;
815 		struct ucred *a_cred;
816 		struct proc *a_p;
817 	} */ *ap = v;
818 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
819 	struct kernfs_subdir *ks;
820 	struct vattr *vap = ap->a_vap;
821 	int error = 0;
822 	char strbuf[KSTRING], *buf;
823 	size_t nread, total;
824 
825 	VATTR_NULL(vap);
826 	vap->va_type = ap->a_vp->v_type;
827 	vap->va_uid = 0;
828 	vap->va_gid = 0;
829 	vap->va_mode = kfs->kfs_mode;
830 	vap->va_fileid = kfs->kfs_fileno;
831 	vap->va_flags = 0;
832 	vap->va_size = 0;
833 	vap->va_blocksize = DEV_BSIZE;
834 	/*
835 	 * Make all times be current TOD, except for the "boottime" node.
836 	 * Avoid microtime(9), it's slow.
837 	 * We don't guard the read from time(9) with splclock(9) since we
838 	 * don't actually need to be THAT sure the access is atomic.
839 	 */
840 	if (kfs->kfs_kt && kfs->kfs_kt->kt_namlen == 8 &&
841 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
842 		TIMEVAL_TO_TIMESPEC(&boottime, &vap->va_ctime);
843 	} else {
844 		TIMEVAL_TO_TIMESPEC(&time, &vap->va_ctime);
845 	}
846 	vap->va_atime = vap->va_mtime = vap->va_ctime;
847 	vap->va_gen = 0;
848 	vap->va_flags = 0;
849 	vap->va_rdev = 0;
850 	vap->va_bytes = 0;
851 
852 	switch (kfs->kfs_type) {
853 	case KFSkern:
854 		vap->va_nlink = nkern_dirs;
855 		vap->va_bytes = vap->va_size = DEV_BSIZE;
856 		break;
857 
858 	case KFSroot:
859 		vap->va_nlink = 1;
860 		vap->va_bytes = vap->va_size = DEV_BSIZE;
861 		break;
862 
863 	case KFSsubdir:
864 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
865 		vap->va_nlink = ks->ks_dirs;
866 		vap->va_bytes = vap->va_size = DEV_BSIZE;
867 		break;
868 
869 	case KFSnull:
870 	case KFStime:
871 	case KFSint:
872 	case KFSstring:
873 	case KFShostname:
874 	case KFSavenrun:
875 	case KFSdevice:
876 	case KFSmsgbuf:
877 #ifdef IPSEC
878 	case KFSipsecsa:
879 	case KFSipsecsp:
880 #endif
881 		vap->va_nlink = 1;
882 		total = 0;
883 		do {
884 			buf = strbuf;
885 			error = kernfs_xread(kfs, total, &buf,
886 			    sizeof(strbuf), &nread);
887 			total += nread;
888 		} while (error == 0 && nread != 0);
889 		vap->va_bytes = vap->va_size = total;
890 		break;
891 
892 #ifdef IPSEC
893 	case KFSipsecsadir:
894 	case KFSipsecspdir:
895 		vap->va_nlink = 2;
896 		vap->va_bytes = vap->va_size = DEV_BSIZE;
897 		break;
898 #endif
899 
900 	default:
901 		error = kernfs_try_fileop(kfs->kfs_type,
902 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
903 		break;
904 	}
905 
906 	return (error);
907 }
908 
909 /*ARGSUSED*/
910 int
911 kernfs_setattr(v)
912 	void *v;
913 {
914 
915 	/*
916 	 * Silently ignore attribute changes.
917 	 * This allows for open with truncate to have no
918 	 * effect until some data is written.  I want to
919 	 * do it this way because all writes are atomic.
920 	 */
921 	return (0);
922 }
923 
924 int
925 kernfs_read(v)
926 	void *v;
927 {
928 	struct vop_read_args /* {
929 		struct vnode *a_vp;
930 		struct uio *a_uio;
931 		int  a_ioflag;
932 		struct ucred *a_cred;
933 	} */ *ap = v;
934 	struct uio *uio = ap->a_uio;
935 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
936 	char strbuf[KSTRING], *buf;
937 	off_t off;
938 	size_t len;
939 	int error;
940 
941 	if (ap->a_vp->v_type == VDIR)
942 		return (EOPNOTSUPP);
943 
944 	off = uio->uio_offset;
945 	buf = strbuf;
946 	if ((error = kernfs_xread(kfs, off, &buf, sizeof(strbuf), &len)) == 0)
947 		error = uiomove(buf, len, uio);
948 	return (error);
949 }
950 
951 static int
952 kernfs_default_xwrite(v)
953 	void *v;
954 {
955 	struct vop_write_args /* {
956 		struct vnode *a_vp;
957 		struct uio *a_uio;
958 		int  a_ioflag;
959 		struct ucred *a_cred;
960 	} */ *ap = v;
961 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
962 	struct uio *uio = ap->a_uio;
963 	int error, xlen;
964 	char strbuf[KSTRING];
965 
966 	if (uio->uio_offset != 0)
967 		return (EINVAL);
968 
969 	xlen = min(uio->uio_resid, KSTRING-1);
970 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
971 		return (error);
972 
973 	if (uio->uio_resid != 0)
974 		return (EIO);
975 
976 	strbuf[xlen] = '\0';
977 	xlen = strlen(strbuf);
978 	return (kernfs_xwrite(kfs, strbuf, xlen));
979 }
980 
981 int
982 kernfs_write(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 		struct ucred *a_cred;
990 	} */ *ap = v;
991 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
992 
993 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v, 0);
994 }
995 
996 int
997 kernfs_ioctl(v)
998 	void *v;
999 {
1000 	struct vop_ioctl_args /* {
1001 		const struct vnodeop_desc *a_desc;
1002 		struct vnode *a_vp;
1003 		u_long a_command;
1004 		void *a_data;
1005 		int a_fflag;
1006 		struct ucred *a_cred;
1007 		struct proc *a_p;
1008 	} */ *ap = v;
1009 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1010 
1011 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1012 	    EPASSTHROUGH);
1013 }
1014 
1015 int
1016 kernfs_mmap(v)
1017 	void *v;
1018 {
1019 	struct vop_mmap_args /* {
1020 		const struct vnodeop_desc *a_desc;
1021 		struct vnode *a_vp;
1022 		int a_fflags;
1023 		struct ucred *a_cred;
1024 		struct proc *a_p;
1025 	} */ *ap = v;
1026 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1027 
1028 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_MMAP, v, 0);
1029 }
1030 
1031 static int
1032 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1033     u_int32_t value, struct vop_readdir_args *ap)
1034 {
1035 	struct kernfs_node *kfs;
1036 	struct vnode *vp;
1037 	int error;
1038 
1039 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1040 	    value)) != 0)
1041 		return error;
1042 	if (kt->kt_tag == KFSdevice) {
1043 		struct vattr va;
1044 		if ((error = VOP_GETATTR(vp, &va, ap->a_cred,
1045 		    ap->a_uio->uio_segflg == UIO_USERSPACE ?
1046 		    ap->a_uio->uio_procp : &proc0)) != 0)
1047 			return (error);
1048 		d->d_fileno = va.va_fileid;
1049 	} else {
1050 		kfs = VTOKERN(vp);
1051 		d->d_fileno = kfs->kfs_fileno;
1052 	}
1053 	vput(vp);
1054 	return 0;
1055 }
1056 
1057 static int
1058 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1059     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1060     const struct kern_target *kt, struct vop_readdir_args *ap)
1061 {
1062 	const struct kern_target *ikt;
1063 	int error;
1064 
1065 	switch (entry) {
1066 	case 0:
1067 		d->d_fileno = thisdir_kfs->kfs_fileno;
1068 		return 0;
1069 	case 1:
1070 		ikt = parent_kt;
1071 		break;
1072 	default:
1073 		ikt = kt;
1074 		break;
1075 	}
1076 	if (ikt != thisdir_kfs->kfs_kt) {
1077 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1078 			return error;
1079 	} else
1080 		d->d_fileno = thisdir_kfs->kfs_fileno;
1081 	return 0;
1082 }
1083 
1084 int
1085 kernfs_readdir(v)
1086 	void *v;
1087 {
1088 	struct vop_readdir_args /* {
1089 		struct vnode *a_vp;
1090 		struct uio *a_uio;
1091 		struct ucred *a_cred;
1092 		int *a_eofflag;
1093 		off_t **a_cookies;
1094 		int a_*ncookies;
1095 	} */ *ap = v;
1096 	struct uio *uio = ap->a_uio;
1097 	struct dirent d;
1098 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1099 	const struct kern_target *kt;
1100 	const struct dyn_kern_target *dkt = NULL;
1101 	const struct kernfs_subdir *ks;
1102 	off_t i, j;
1103 	int error;
1104 	off_t *cookies = NULL;
1105 	int ncookies = 0, n;
1106 #ifdef IPSEC
1107 	struct secasvar *sav, *sav2;
1108 	struct secpolicy *sp;
1109 #endif
1110 
1111 	if (uio->uio_resid < UIO_MX)
1112 		return (EINVAL);
1113 	if (uio->uio_offset < 0)
1114 		return (EINVAL);
1115 
1116 	error = 0;
1117 	i = uio->uio_offset;
1118 	memset(&d, 0, sizeof(d));
1119 	d.d_reclen = UIO_MX;
1120 	ncookies = uio->uio_resid / UIO_MX;
1121 
1122 	switch (kfs->kfs_type) {
1123 	case KFSkern:
1124 		if (i >= nkern_targets)
1125 			return (0);
1126 
1127 		if (ap->a_ncookies) {
1128 			ncookies = min(ncookies, (nkern_targets - i));
1129 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1130 			    M_WAITOK);
1131 			*ap->a_cookies = cookies;
1132 		}
1133 
1134 		n = 0;
1135 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1136 			if (i < static_nkern_targets)
1137 				kt = &kern_targets[i];
1138 			else {
1139 				if (dkt == NULL) {
1140 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1141 					for (j = static_nkern_targets; j < i &&
1142 						     dkt != NULL; j++)
1143 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1144 					if (j != i)
1145 						break;
1146 				} else {
1147 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1148 					if (dkt == NULL)
1149 						break;
1150 				}
1151 				kt = &dkt->dkt_kt;
1152 			}
1153 			if (kt->kt_tag == KFSdevice) {
1154 				dev_t *dp = kt->kt_data;
1155 				struct vnode *fvp;
1156 
1157 				if (*dp == NODEV ||
1158 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1159 					continue;
1160 			}
1161 			d.d_namlen = kt->kt_namlen;
1162 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1163 			    &kern_targets[0], kt, ap)) != 0)
1164 				break;
1165 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1166 			d.d_type = kt->kt_type;
1167 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1168 				break;
1169 			if (cookies)
1170 				*cookies++ = i + 1;
1171 			n++;
1172 		}
1173 		ncookies = n;
1174 		break;
1175 
1176 	case KFSroot:
1177 		if (i >= 2)
1178 			return 0;
1179 
1180 		if (ap->a_ncookies) {
1181 			ncookies = min(ncookies, (2 - i));
1182 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1183 			    M_WAITOK);
1184 			*ap->a_cookies = cookies;
1185 		}
1186 
1187 		n = 0;
1188 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1189 			kt = &kern_targets[i];
1190 			d.d_namlen = kt->kt_namlen;
1191 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
1192 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1193 			d.d_type = kt->kt_type;
1194 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1195 				break;
1196 			if (cookies)
1197 				*cookies++ = i + 1;
1198 			n++;
1199 		}
1200 		ncookies = n;
1201 		break;
1202 
1203 	case KFSsubdir:
1204 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
1205 		if (i >= ks->ks_nentries)
1206 			return (0);
1207 
1208 		if (ap->a_ncookies) {
1209 			ncookies = min(ncookies, (ks->ks_nentries - i));
1210 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1211 			    M_WAITOK);
1212 			*ap->a_cookies = cookies;
1213 		}
1214 
1215 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1216 		for (j = 0; j < i && dkt != NULL; j++)
1217 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1218 		n = 0;
1219 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1220 			if (i < 2)
1221 				kt = &subdir_targets[i];
1222 			else {
1223 				/* check if ks_nentries lied to us */
1224 				if (dkt == NULL)
1225 					break;
1226 				kt = &dkt->dkt_kt;
1227 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1228 			}
1229 			if (kt->kt_tag == KFSdevice) {
1230 				dev_t *dp = kt->kt_data;
1231 				struct vnode *fvp;
1232 
1233 				if (*dp == NODEV ||
1234 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1235 					continue;
1236 			}
1237 			d.d_namlen = kt->kt_namlen;
1238 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1239 			    ks->ks_parent, kt, ap)) != 0)
1240 				break;
1241 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1242 			d.d_type = kt->kt_type;
1243 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1244 				break;
1245 			if (cookies)
1246 				*cookies++ = i + 1;
1247 			n++;
1248 		}
1249 		ncookies = n;
1250 		break;
1251 
1252 #ifdef IPSEC
1253 	case KFSipsecsadir:
1254 		/* count SA in the system */
1255 		n = 0;
1256 		TAILQ_FOREACH(sav, &satailq, tailq) {
1257 			for (sav2 = TAILQ_FIRST(&satailq);
1258 			    sav2 != sav;
1259 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1260 				if (sav->spi == sav2->spi) {
1261 					/* multiple SA with same SPI */
1262 					break;
1263 				}
1264 			}
1265 			if (sav == sav2 || sav->spi != sav2->spi)
1266 				n++;
1267 		}
1268 
1269 		if (i >= nipsecsa_targets + n)
1270 			return (0);
1271 
1272 		if (ap->a_ncookies) {
1273 			ncookies = min(ncookies, (n - i));
1274 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1275 			    M_WAITOK);
1276 			*ap->a_cookies = cookies;
1277 		}
1278 
1279 		n = 0;
1280 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1281 			kt = &ipsecsa_targets[i];
1282 			d.d_namlen = kt->kt_namlen;
1283 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1284 			    &kern_targets[0], kt, ap)) != 0)
1285 				break;
1286 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1287 			d.d_type = kt->kt_type;
1288 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1289 				break;
1290 			if (cookies)
1291 				*cookies++ = i + 1;
1292 			n++;
1293 		}
1294 		if (error) {
1295 			ncookies = n;
1296 			break;
1297 		}
1298 
1299 		TAILQ_FOREACH(sav, &satailq, tailq) {
1300 			for (sav2 = TAILQ_FIRST(&satailq);
1301 			    sav2 != sav;
1302 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1303 				if (sav->spi == sav2->spi) {
1304 					/* multiple SA with same SPI */
1305 					break;
1306 				}
1307 			}
1308 			if (sav != sav2 && sav->spi == sav2->spi)
1309 				continue;
1310 			if (uio->uio_resid < UIO_MX)
1311 				break;
1312 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1313 			    sav->spi, ap)) != 0)
1314 				break;
1315 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1316 			    "%u", ntohl(sav->spi));
1317 			d.d_type = DT_REG;
1318 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1319 				break;
1320 			if (cookies)
1321 				*cookies++ = i + 1;
1322 			n++;
1323 			i++;
1324 		}
1325 		ncookies = n;
1326 		break;
1327 
1328 	case KFSipsecspdir:
1329 		/* count SP in the system */
1330 		n = 0;
1331 		TAILQ_FOREACH(sp, &sptailq, tailq)
1332 			n++;
1333 
1334 		if (i >= nipsecsp_targets + n)
1335 			return (0);
1336 
1337 		if (ap->a_ncookies) {
1338 			ncookies = min(ncookies, (n - i));
1339 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1340 			    M_WAITOK);
1341 			*ap->a_cookies = cookies;
1342 		}
1343 
1344 		n = 0;
1345 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1346 			kt = &ipsecsp_targets[i];
1347 			d.d_namlen = kt->kt_namlen;
1348 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1349 			    &kern_targets[0], kt, ap)) != 0)
1350 				break;
1351 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1352 			d.d_type = kt->kt_type;
1353 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1354 				break;
1355 			if (cookies)
1356 				*cookies++ = i + 1;
1357 			n++;
1358 		}
1359 		if (error) {
1360 			ncookies = n;
1361 			break;
1362 		}
1363 
1364 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1365 			if (uio->uio_resid < UIO_MX)
1366 				break;
1367 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1368 			    sp->id, ap)) != 0)
1369 				break;
1370 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1371 			    "%u", sp->id);
1372 			d.d_type = DT_REG;
1373 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1374 				break;
1375 			if (cookies)
1376 				*cookies++ = i + 1;
1377 			n++;
1378 			i++;
1379 		}
1380 		ncookies = n;
1381 		break;
1382 #endif
1383 
1384 	default:
1385 		error = ENOTDIR;
1386 		break;
1387 	}
1388 
1389 	if (ap->a_ncookies) {
1390 		if (error) {
1391 			if (cookies)
1392 				free(*ap->a_cookies, M_TEMP);
1393 			*ap->a_ncookies = 0;
1394 			*ap->a_cookies = NULL;
1395 		} else
1396 			*ap->a_ncookies = ncookies;
1397 	}
1398 
1399 	uio->uio_offset = i;
1400 	return (error);
1401 }
1402 
1403 int
1404 kernfs_inactive(v)
1405 	void *v;
1406 {
1407 	struct vop_inactive_args /* {
1408 		struct vnode *a_vp;
1409 		struct proc *a_p;
1410 	} */ *ap = v;
1411 	struct vnode *vp = ap->a_vp;
1412 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1413 #ifdef IPSEC
1414 	struct mbuf *m;
1415 	struct secpolicy *sp;
1416 #endif
1417 
1418 	VOP_UNLOCK(vp, 0);
1419 	switch (kfs->kfs_type) {
1420 #ifdef IPSEC
1421 	case KFSipsecsa:
1422 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1423 		if (m)
1424 			m_freem(m);
1425 		else
1426 			vgone(vp);
1427 		break;
1428 	case KFSipsecsp:
1429 		sp = key_getspbyid(kfs->kfs_value);
1430 		if (sp)
1431 			key_freesp(sp);
1432 		else {
1433 			/* should never happen as we hold a refcnt */
1434 			vgone(vp);
1435 		}
1436 		break;
1437 #endif
1438 	default:
1439 		break;
1440 	}
1441 	return (0);
1442 }
1443 
1444 int
1445 kernfs_reclaim(v)
1446 	void *v;
1447 {
1448 	struct vop_reclaim_args /* {
1449 		struct vnode *a_vp;
1450 	} */ *ap = v;
1451 
1452 	return (kernfs_freevp(ap->a_vp));
1453 }
1454 
1455 /*
1456  * Return POSIX pathconf information applicable to special devices.
1457  */
1458 int
1459 kernfs_pathconf(v)
1460 	void *v;
1461 {
1462 	struct vop_pathconf_args /* {
1463 		struct vnode *a_vp;
1464 		int a_name;
1465 		register_t *a_retval;
1466 	} */ *ap = v;
1467 
1468 	switch (ap->a_name) {
1469 	case _PC_LINK_MAX:
1470 		*ap->a_retval = LINK_MAX;
1471 		return (0);
1472 	case _PC_MAX_CANON:
1473 		*ap->a_retval = MAX_CANON;
1474 		return (0);
1475 	case _PC_MAX_INPUT:
1476 		*ap->a_retval = MAX_INPUT;
1477 		return (0);
1478 	case _PC_PIPE_BUF:
1479 		*ap->a_retval = PIPE_BUF;
1480 		return (0);
1481 	case _PC_CHOWN_RESTRICTED:
1482 		*ap->a_retval = 1;
1483 		return (0);
1484 	case _PC_VDISABLE:
1485 		*ap->a_retval = _POSIX_VDISABLE;
1486 		return (0);
1487 	case _PC_SYNC_IO:
1488 		*ap->a_retval = 1;
1489 		return (0);
1490 	default:
1491 		return (EINVAL);
1492 	}
1493 	/* NOTREACHED */
1494 }
1495 
1496 /*
1497  * Print out the contents of a /dev/fd vnode.
1498  */
1499 /* ARGSUSED */
1500 int
1501 kernfs_print(v)
1502 	void *v;
1503 {
1504 
1505 	printf("tag VT_KERNFS, kernfs vnode\n");
1506 	return (0);
1507 }
1508 
1509 int
1510 kernfs_link(v)
1511 	void *v;
1512 {
1513 	struct vop_link_args /* {
1514 		struct vnode *a_dvp;
1515 		struct vnode *a_vp;
1516 		struct componentname *a_cnp;
1517 	} */ *ap = v;
1518 
1519 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1520 	vput(ap->a_dvp);
1521 	return (EROFS);
1522 }
1523 
1524 int
1525 kernfs_symlink(v)
1526 	void *v;
1527 {
1528 	struct vop_symlink_args /* {
1529 		struct vnode *a_dvp;
1530 		struct vnode **a_vpp;
1531 		struct componentname *a_cnp;
1532 		struct vattr *a_vap;
1533 		char *a_target;
1534 	} */ *ap = v;
1535 
1536 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1537 	vput(ap->a_dvp);
1538 	return (EROFS);
1539 }
1540