xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision e5548b402ae4c44fb816de42c7bba9581ce23ef5)
1 /*	$NetBSD: kernfs_vnops.c,v 1.115 2005/12/11 12:24:51 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 1.115 2005/12/11 12:24:51 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_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_CLOSE },
168   { .kf_fileop = KERNFS_FILEOP_WRITE, .kf_genop = {kernfs_default_xwrite} },
169 };
170 
171 int	kernfs_lookup(void *);
172 #define	kernfs_create	genfs_eopnotsupp
173 #define	kernfs_mknod	genfs_eopnotsupp
174 int	kernfs_open(void *);
175 int	kernfs_close(void *);
176 int	kernfs_access(void *);
177 int	kernfs_getattr(void *);
178 int	kernfs_setattr(void *);
179 int	kernfs_read(void *);
180 int	kernfs_write(void *);
181 #define	kernfs_fcntl	genfs_fcntl
182 int	kernfs_ioctl(void *);
183 #define	kernfs_poll	genfs_poll
184 #define kernfs_revoke	genfs_revoke
185 #define	kernfs_fsync	genfs_nullop
186 #define	kernfs_seek	genfs_nullop
187 #define	kernfs_remove	genfs_eopnotsupp
188 int	kernfs_link(void *);
189 #define	kernfs_rename	genfs_eopnotsupp
190 #define	kernfs_mkdir	genfs_eopnotsupp
191 #define	kernfs_rmdir	genfs_eopnotsupp
192 int	kernfs_symlink(void *);
193 int	kernfs_readdir(void *);
194 #define	kernfs_readlink	genfs_eopnotsupp
195 #define	kernfs_abortop	genfs_abortop
196 int	kernfs_inactive(void *);
197 int	kernfs_reclaim(void *);
198 #define	kernfs_lock	genfs_lock
199 #define	kernfs_unlock	genfs_unlock
200 #define	kernfs_bmap	genfs_badop
201 #define	kernfs_strategy	genfs_badop
202 int	kernfs_print(void *);
203 #define	kernfs_islocked	genfs_islocked
204 int	kernfs_pathconf(void *);
205 #define	kernfs_advlock	genfs_einval
206 #define	kernfs_bwrite	genfs_eopnotsupp
207 #define	kernfs_putpages	genfs_putpages
208 
209 static int	kernfs_xread(struct kernfs_node *, int, char **,
210 				size_t, size_t *);
211 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
212 
213 int (**kernfs_vnodeop_p)(void *);
214 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
215 	{ &vop_default_desc, vn_default_error },
216 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
217 	{ &vop_create_desc, kernfs_create },		/* create */
218 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
219 	{ &vop_open_desc, kernfs_open },		/* open */
220 	{ &vop_close_desc, kernfs_close },		/* close */
221 	{ &vop_access_desc, kernfs_access },		/* access */
222 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
223 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
224 	{ &vop_read_desc, kernfs_read },		/* read */
225 	{ &vop_write_desc, kernfs_write },		/* write */
226 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
227 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
228 	{ &vop_poll_desc, kernfs_poll },		/* poll */
229 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
230 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
231 	{ &vop_seek_desc, kernfs_seek },		/* seek */
232 	{ &vop_remove_desc, kernfs_remove },		/* remove */
233 	{ &vop_link_desc, kernfs_link },		/* link */
234 	{ &vop_rename_desc, kernfs_rename },		/* rename */
235 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
236 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
237 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
238 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
239 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
240 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
241 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
242 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
243 	{ &vop_lock_desc, kernfs_lock },		/* lock */
244 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
245 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
246 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
247 	{ &vop_print_desc, kernfs_print },		/* print */
248 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
249 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
250 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
251 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
252 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
253 	{ NULL, NULL }
254 };
255 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
256 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
257 
258 static __inline int
259 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
260 {
261 	if (a->kf_type < b->kf_type)
262 		return -1;
263 	if (a->kf_type > b->kf_type)
264 		return 1;
265 	if (a->kf_fileop < b->kf_fileop)
266 		return -1;
267 	if (a->kf_fileop > b->kf_fileop)
268 		return 1;
269 	return (0);
270 }
271 
272 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
273 	SPLAY_INITIALIZER(kfsfileoptree);
274 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
275 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
276 
277 kfstype
278 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
279 {
280 	static u_char nextfreetype = KFSlasttype;
281 	struct kernfs_fileop *dkf, *fkf, skf;
282 	int i;
283 
284 	/* XXX need to keep track of dkf's memory if we support
285            deallocating types */
286 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
287 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
288 
289 	for (i = 0; i < sizeof(kernfs_default_fileops) /
290 		     sizeof(kernfs_default_fileops[0]); i++) {
291 		dkf[i].kf_type = nextfreetype;
292 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
293 	}
294 
295 	for (i = 0; i < nkf; i++) {
296 		skf.kf_type = nextfreetype;
297 		skf.kf_fileop = kf[i].kf_fileop;
298 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
299 			fkf->kf_genop = kf[i].kf_genop;
300 	}
301 
302 	return nextfreetype++;
303 }
304 
305 int
306 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
307 {
308 	struct kernfs_fileop *kf, skf;
309 
310 	skf.kf_type = type;
311 	skf.kf_fileop = fileop;
312 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
313 		if (kf->kf_vop)
314 			return kf->kf_vop(v);
315 	return error;
316 }
317 
318 int
319 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
320     size_t len, int error)
321 {
322 	struct kernfs_fileop *kf, skf;
323 
324 	skf.kf_type = type;
325 	skf.kf_fileop = KERNFS_XWRITE;
326 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
327 		if (kf->kf_xwrite)
328 			return kf->kf_xwrite(kfs, bf, len);
329 	return error;
330 }
331 
332 int
333 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
334 {
335 	struct kernfs_subdir *ks, *parent;
336 
337 	if (pkt == NULL) {
338 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
339 		nkern_targets++;
340 		if (dkt->dkt_kt.kt_vtype == VDIR)
341 			nkern_dirs++;
342 	} else {
343 		parent = (struct kernfs_subdir *)pkt->kt_data;
344 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
345 		parent->ks_nentries++;
346 		if (dkt->dkt_kt.kt_vtype == VDIR)
347 			parent->ks_dirs++;
348 	}
349 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
350 		ks = malloc(sizeof(struct kernfs_subdir),
351 		    M_TEMP, M_WAITOK);
352 		SIMPLEQ_INIT(&ks->ks_entries);
353 		ks->ks_nentries = 2; /* . and .. */
354 		ks->ks_dirs = 2;
355 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
356 		dkt->dkt_kt.kt_data = ks;
357 	}
358 	return 0;
359 }
360 
361 static int
362 kernfs_xread(kfs, off, bufp, len, wrlen)
363 	struct kernfs_node *kfs;
364 	int off;
365 	char **bufp;
366 	size_t len;
367 	size_t *wrlen;
368 {
369 	const struct kern_target *kt;
370 #ifdef IPSEC
371 	struct mbuf *m;
372 #endif
373 
374 	kt = kfs->kfs_kt;
375 
376 	switch (kfs->kfs_type) {
377 	case KFStime: {
378 		struct timeval tv;
379 
380 		microtime(&tv);
381 		snprintf(*bufp, len, "%ld %ld\n", tv.tv_sec, tv.tv_usec);
382 		break;
383 	}
384 
385 	case KFSint: {
386 		int *ip = kt->kt_data;
387 
388 		snprintf(*bufp, len, "%d\n", *ip);
389 		break;
390 	}
391 
392 	case KFSstring: {
393 		char *cp = kt->kt_data;
394 
395 		*bufp = cp;
396 		break;
397 	}
398 
399 	case KFSmsgbuf: {
400 		long n;
401 
402 		/*
403 		 * deal with cases where the message buffer has
404 		 * become corrupted.
405 		 */
406 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
407 			msgbufenabled = 0;
408 			return (ENXIO);
409 		}
410 
411 		/*
412 		 * Note that reads of /kern/msgbuf won't necessarily yield
413 		 * consistent results, if the message buffer is modified
414 		 * while the read is in progress.  The worst that can happen
415 		 * is that incorrect data will be read.  There's no way
416 		 * that this can crash the system unless the values in the
417 		 * message buffer header are corrupted, but that'll cause
418 		 * the system to die anyway.
419 		 */
420 		if (off >= msgbufp->msg_bufs) {
421 			*wrlen = 0;
422 			return (0);
423 		}
424 		n = msgbufp->msg_bufx + off;
425 		if (n >= msgbufp->msg_bufs)
426 			n -= msgbufp->msg_bufs;
427 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
428 		*bufp = msgbufp->msg_bufc + n;
429 		*wrlen = len;
430 		return (0);
431 	}
432 
433 	case KFShostname: {
434 		char *cp = hostname;
435 		int xlen = hostnamelen;
436 
437 		if (xlen >= (len - 2))
438 			return (EINVAL);
439 
440 		memcpy(*bufp, cp, xlen);
441 		(*bufp)[xlen] = '\n';
442 		(*bufp)[xlen+1] = '\0';
443 		len = strlen(*bufp);
444 		break;
445 	}
446 
447 	case KFSavenrun:
448 		averunnable.fscale = FSCALE;
449 		snprintf(*bufp, len, "%d %d %d %ld\n",
450 		    averunnable.ldavg[0], averunnable.ldavg[1],
451 		    averunnable.ldavg[2], averunnable.fscale);
452 		break;
453 
454 #ifdef IPSEC
455 	case KFSipsecsa:
456 		/*
457 		 * Note that SA configuration could be changed during the
458 		 * read operation, resulting in garbled output.
459 		 */
460 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
461 		if (!m)
462 			return (ENOBUFS);
463 		if (off >= m->m_pkthdr.len) {
464 			*wrlen = 0;
465 			m_freem(m);
466 			return (0);
467 		}
468 		if (len > m->m_pkthdr.len - off)
469 			len = m->m_pkthdr.len - off;
470 		m_copydata(m, off, len, *bufp);
471 		*wrlen = len;
472 		m_freem(m);
473 		return (0);
474 
475 	case KFSipsecsp:
476 		/*
477 		 * Note that SP configuration could be changed during the
478 		 * read operation, resulting in garbled output.
479 		 */
480 		if (!kfs->kfs_v) {
481 			struct secpolicy *sp;
482 
483 			sp = key_getspbyid(kfs->kfs_value);
484 			if (sp)
485 				kfs->kfs_v = sp;
486 			else
487 				return (ENOENT);
488 		}
489 		m = key_setdumpsp((struct secpolicy *)kfs->kfs_v,
490 		    SADB_X_SPDGET, 0, 0);
491 		if (!m)
492 			return (ENOBUFS);
493 		if (off >= m->m_pkthdr.len) {
494 			*wrlen = 0;
495 			m_freem(m);
496 			return (0);
497 		}
498 		if (len > m->m_pkthdr.len - off)
499 			len = m->m_pkthdr.len - off;
500 		m_copydata(m, off, len, *bufp);
501 		*wrlen = len;
502 		m_freem(m);
503 		return (0);
504 #endif
505 
506 	default:
507 		*wrlen = 0;
508 		return (0);
509 	}
510 
511 	len = strlen(*bufp);
512 	if (len <= off)
513 		*wrlen = 0;
514 	else {
515 		*bufp += off;
516 		*wrlen = len - off;
517 	}
518 	return (0);
519 }
520 
521 static int
522 kernfs_xwrite(kfs, bf, len)
523 	const struct kernfs_node *kfs;
524 	char *bf;
525 	size_t len;
526 {
527 
528 	switch (kfs->kfs_type) {
529 	case KFShostname:
530 		if (bf[len-1] == '\n')
531 			--len;
532 		memcpy(hostname, bf, len);
533 		hostname[len] = '\0';
534 		hostnamelen = (size_t) len;
535 		return (0);
536 
537 	default:
538 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
539 	}
540 }
541 
542 
543 /*
544  * vp is the current namei directory
545  * ndp is the name to locate in that directory...
546  */
547 int
548 kernfs_lookup(v)
549 	void *v;
550 {
551 	struct vop_lookup_args /* {
552 		struct vnode * a_dvp;
553 		struct vnode ** a_vpp;
554 		struct componentname * a_cnp;
555 	} */ *ap = v;
556 	struct componentname *cnp = ap->a_cnp;
557 	struct vnode **vpp = ap->a_vpp;
558 	struct vnode *dvp = ap->a_dvp;
559 	const char *pname = cnp->cn_nameptr;
560 	const struct kernfs_node *kfs;
561 	const struct kern_target *kt;
562 	const struct dyn_kern_target *dkt;
563 	const struct kernfs_subdir *ks;
564 	int error, i, wantpunlock;
565 #ifdef IPSEC
566 	char *ep;
567 	u_int32_t id;
568 #endif
569 
570 	*vpp = NULLVP;
571 	cnp->cn_flags &= ~PDIRUNLOCK;
572 
573 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
574 		return (EROFS);
575 
576 	if (cnp->cn_namelen == 1 && *pname == '.') {
577 		*vpp = dvp;
578 		VREF(dvp);
579 		return (0);
580 	}
581 
582 	wantpunlock = (~cnp->cn_flags & (LOCKPARENT | ISLASTCN));
583 	kfs = VTOKERN(dvp);
584 	switch (kfs->kfs_type) {
585 	case KFSkern:
586 		/*
587 		 * Shouldn't get here with .. in the root node.
588 		 */
589 		if (cnp->cn_flags & ISDOTDOT)
590 			return (EIO);
591 
592 		for (i = 0; i < static_nkern_targets; i++) {
593 			kt = &kern_targets[i];
594 			if (cnp->cn_namelen == kt->kt_namlen &&
595 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
596 				goto found;
597 		}
598 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
599 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
600 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
601 				kt = &dkt->dkt_kt;
602 				goto found;
603 			}
604 		}
605 		break;
606 
607 	found:
608 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
609 		if ((error == 0) && wantpunlock) {
610 			VOP_UNLOCK(dvp, 0);
611 			cnp->cn_flags |= PDIRUNLOCK;
612 		}
613 		return (error);
614 
615 	case KFSsubdir:
616 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
617 		if (cnp->cn_flags & ISDOTDOT) {
618 			kt = ks->ks_parent;
619 			goto found;
620 		}
621 
622 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, 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 #ifdef IPSEC
632 	case KFSipsecsadir:
633 		if (cnp->cn_flags & ISDOTDOT) {
634 			kt = &kern_targets[0];
635 			goto found;
636 		}
637 
638 		for (i = 2; i < nipsecsa_targets; i++) {
639 			kt = &ipsecsa_targets[i];
640 			if (cnp->cn_namelen == kt->kt_namlen &&
641 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
642 				goto found;
643 		}
644 
645 		ep = NULL;
646 		id = strtoul(pname, &ep, 10);
647 		if (!ep || *ep || ep == pname)
648 			break;
649 
650 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id);
651 		if ((error == 0) && wantpunlock) {
652 			VOP_UNLOCK(dvp, 0);
653 			cnp->cn_flags |= PDIRUNLOCK;
654 		}
655 		return (error);
656 
657 	case KFSipsecspdir:
658 		if (cnp->cn_flags & ISDOTDOT) {
659 			kt = &kern_targets[0];
660 			goto found;
661 		}
662 
663 		for (i = 2; i < nipsecsp_targets; i++) {
664 			kt = &ipsecsp_targets[i];
665 			if (cnp->cn_namelen == kt->kt_namlen &&
666 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
667 				goto found;
668 		}
669 
670 		ep = NULL;
671 		id = strtoul(pname, &ep, 10);
672 		if (!ep || *ep || ep == pname)
673 			break;
674 
675 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id);
676 		if ((error == 0) && wantpunlock) {
677 			VOP_UNLOCK(dvp, 0);
678 			cnp->cn_flags |= PDIRUNLOCK;
679 		}
680 		return (error);
681 #endif
682 
683 	default:
684 		return (ENOTDIR);
685 	}
686 
687 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
688 }
689 
690 int
691 kernfs_open(v)
692 	void *v;
693 {
694 	struct vop_open_args /* {
695 		struct vnode *a_vp;
696 		int a_mode;
697 		struct ucred *a_cred;
698 		struct lwp *a_l;
699 	} */ *ap = v;
700 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
701 #ifdef IPSEC
702 	struct mbuf *m;
703 	struct secpolicy *sp;
704 #endif
705 
706 	switch (kfs->kfs_type) {
707 #ifdef IPSEC
708 	case KFSipsecsa:
709 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
710 		if (m) {
711 			m_freem(m);
712 			return (0);
713 		} else
714 			return (ENOENT);
715 
716 	case KFSipsecsp:
717 		sp = key_getspbyid(kfs->kfs_value);
718 		if (sp) {
719 			kfs->kfs_v = sp;
720 			return (0);
721 		} else
722 			return (ENOENT);
723 #endif
724 
725 	default:
726 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN,
727 		    v, 0);
728 	}
729 }
730 
731 int
732 kernfs_close(v)
733 	void *v;
734 {
735 	struct vop_close_args /* {
736 		struct vnode *a_vp;
737 		int a_fflag;
738 		struct ucred *a_cred;
739 		struct lwp *a_l;
740 	} */ *ap = v;
741 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
742 
743 	switch (kfs->kfs_type) {
744 #ifdef IPSEC
745 	case KFSipsecsp:
746 		key_freesp((struct secpolicy *)kfs->kfs_v);
747 		break;
748 #endif
749 
750 	default:
751 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE,
752 		    v, 0);
753 	}
754 
755 	return (0);
756 }
757 
758 int
759 kernfs_access(v)
760 	void *v;
761 {
762 	struct vop_access_args /* {
763 		struct vnode *a_vp;
764 		int a_mode;
765 		struct ucred *a_cred;
766 		struct lwp *a_l;
767 	} */ *ap = v;
768 	struct vattr va;
769 	int error;
770 
771 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_l)) != 0)
772 		return (error);
773 
774 	return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid,
775 	    ap->a_mode, ap->a_cred));
776 }
777 
778 static int
779 kernfs_default_fileop_getattr(v)
780 	void *v;
781 {
782 	struct vop_getattr_args /* {
783 		struct vnode *a_vp;
784 		struct vattr *a_vap;
785 		struct ucred *a_cred;
786 		struct lwp *a_l;
787 	} */ *ap = v;
788 	struct vattr *vap = ap->a_vap;
789 
790 	vap->va_nlink = 1;
791 	vap->va_bytes = vap->va_size = 0;
792 
793 	return 0;
794 }
795 
796 int
797 kernfs_getattr(v)
798 	void *v;
799 {
800 	struct vop_getattr_args /* {
801 		struct vnode *a_vp;
802 		struct vattr *a_vap;
803 		struct ucred *a_cred;
804 		struct lwp *a_l;
805 	} */ *ap = v;
806 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
807 	struct kernfs_subdir *ks;
808 	struct vattr *vap = ap->a_vap;
809 	int error = 0;
810 	char strbuf[KSTRING], *bf;
811 	size_t nread, total;
812 
813 	VATTR_NULL(vap);
814 	vap->va_type = ap->a_vp->v_type;
815 	vap->va_uid = 0;
816 	vap->va_gid = 0;
817 	vap->va_mode = kfs->kfs_mode;
818 	vap->va_fileid = kfs->kfs_fileno;
819 	vap->va_flags = 0;
820 	vap->va_size = 0;
821 	vap->va_blocksize = DEV_BSIZE;
822 	/*
823 	 * Make all times be current TOD, except for the "boottime" node.
824 	 * Avoid microtime(9), it's slow.
825 	 * We don't guard the read from time(9) with splclock(9) since we
826 	 * don't actually need to be THAT sure the access is atomic.
827 	 */
828 	if (kfs->kfs_kt && kfs->kfs_kt->kt_namlen == 8 &&
829 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
830 		TIMEVAL_TO_TIMESPEC(&boottime, &vap->va_ctime);
831 	} else {
832 		TIMEVAL_TO_TIMESPEC(&time, &vap->va_ctime);
833 	}
834 	vap->va_atime = vap->va_mtime = vap->va_ctime;
835 	vap->va_gen = 0;
836 	vap->va_flags = 0;
837 	vap->va_rdev = 0;
838 	vap->va_bytes = 0;
839 
840 	switch (kfs->kfs_type) {
841 	case KFSkern:
842 		vap->va_nlink = nkern_dirs;
843 		vap->va_bytes = vap->va_size = DEV_BSIZE;
844 		break;
845 
846 	case KFSroot:
847 		vap->va_nlink = 1;
848 		vap->va_bytes = vap->va_size = DEV_BSIZE;
849 		break;
850 
851 	case KFSsubdir:
852 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
853 		vap->va_nlink = ks->ks_dirs;
854 		vap->va_bytes = vap->va_size = DEV_BSIZE;
855 		break;
856 
857 	case KFSnull:
858 	case KFStime:
859 	case KFSint:
860 	case KFSstring:
861 	case KFShostname:
862 	case KFSavenrun:
863 	case KFSdevice:
864 	case KFSmsgbuf:
865 #ifdef IPSEC
866 	case KFSipsecsa:
867 	case KFSipsecsp:
868 #endif
869 		vap->va_nlink = 1;
870 		total = 0;
871 		do {
872 			bf = strbuf;
873 			error = kernfs_xread(kfs, total, &bf,
874 			    sizeof(strbuf), &nread);
875 			total += nread;
876 		} while (error == 0 && nread != 0);
877 		vap->va_bytes = vap->va_size = total;
878 		break;
879 
880 #ifdef IPSEC
881 	case KFSipsecsadir:
882 	case KFSipsecspdir:
883 		vap->va_nlink = 2;
884 		vap->va_bytes = vap->va_size = DEV_BSIZE;
885 		break;
886 #endif
887 
888 	default:
889 		error = kernfs_try_fileop(kfs->kfs_type,
890 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
891 		break;
892 	}
893 
894 	return (error);
895 }
896 
897 /*ARGSUSED*/
898 int
899 kernfs_setattr(v)
900 	void *v;
901 {
902 
903 	/*
904 	 * Silently ignore attribute changes.
905 	 * This allows for open with truncate to have no
906 	 * effect until some data is written.  I want to
907 	 * do it this way because all writes are atomic.
908 	 */
909 	return (0);
910 }
911 
912 int
913 kernfs_read(v)
914 	void *v;
915 {
916 	struct vop_read_args /* {
917 		struct vnode *a_vp;
918 		struct uio *a_uio;
919 		int  a_ioflag;
920 		struct ucred *a_cred;
921 	} */ *ap = v;
922 	struct uio *uio = ap->a_uio;
923 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
924 	char strbuf[KSTRING], *bf;
925 	int off;
926 	size_t len;
927 	int error;
928 
929 	if (ap->a_vp->v_type == VDIR)
930 		return (EOPNOTSUPP);
931 
932 	off = (int)uio->uio_offset;
933 	/* Don't allow negative offsets */
934 	if (off < 0)
935 		return EINVAL;
936 
937 	bf = strbuf;
938 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
939 		error = uiomove(bf, len, uio);
940 	return (error);
941 }
942 
943 static int
944 kernfs_default_xwrite(v)
945 	void *v;
946 {
947 	struct vop_write_args /* {
948 		struct vnode *a_vp;
949 		struct uio *a_uio;
950 		int  a_ioflag;
951 		struct ucred *a_cred;
952 	} */ *ap = v;
953 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
954 	struct uio *uio = ap->a_uio;
955 	int error, xlen;
956 	char strbuf[KSTRING];
957 
958 	if (uio->uio_offset != 0)
959 		return (EINVAL);
960 
961 	xlen = min(uio->uio_resid, KSTRING-1);
962 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
963 		return (error);
964 
965 	if (uio->uio_resid != 0)
966 		return (EIO);
967 
968 	strbuf[xlen] = '\0';
969 	xlen = strlen(strbuf);
970 	return (kernfs_xwrite(kfs, strbuf, xlen));
971 }
972 
973 int
974 kernfs_write(v)
975 	void *v;
976 {
977 	struct vop_write_args /* {
978 		struct vnode *a_vp;
979 		struct uio *a_uio;
980 		int  a_ioflag;
981 		struct ucred *a_cred;
982 	} */ *ap = v;
983 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
984 
985 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v, 0);
986 }
987 
988 int
989 kernfs_ioctl(v)
990 	void *v;
991 {
992 	struct vop_ioctl_args /* {
993 		const struct vnodeop_desc *a_desc;
994 		struct vnode *a_vp;
995 		u_long a_command;
996 		void *a_data;
997 		int a_fflag;
998 		struct ucred *a_cred;
999 		struct lwp *a_l;
1000 	} */ *ap = v;
1001 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1002 
1003 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1004 	    EPASSTHROUGH);
1005 }
1006 
1007 static int
1008 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1009     u_int32_t value, struct vop_readdir_args *ap)
1010 {
1011 	struct kernfs_node *kfs;
1012 	struct vnode *vp;
1013 	int error;
1014 
1015 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1016 	    value)) != 0)
1017 		return error;
1018 	if (kt->kt_tag == KFSdevice) {
1019 		struct vattr va;
1020 		if ((error = VOP_GETATTR(vp, &va, ap->a_cred,
1021 		    ap->a_uio->uio_segflg == UIO_USERSPACE ?
1022 		    ap->a_uio->uio_lwp : &lwp0)) != 0)
1023 			return (error);
1024 		d->d_fileno = va.va_fileid;
1025 	} else {
1026 		kfs = VTOKERN(vp);
1027 		d->d_fileno = kfs->kfs_fileno;
1028 	}
1029 	vput(vp);
1030 	return 0;
1031 }
1032 
1033 static int
1034 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1035     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1036     const struct kern_target *kt, struct vop_readdir_args *ap)
1037 {
1038 	const struct kern_target *ikt;
1039 	int error;
1040 
1041 	switch (entry) {
1042 	case 0:
1043 		d->d_fileno = thisdir_kfs->kfs_fileno;
1044 		return 0;
1045 	case 1:
1046 		ikt = parent_kt;
1047 		break;
1048 	default:
1049 		ikt = kt;
1050 		break;
1051 	}
1052 	if (ikt != thisdir_kfs->kfs_kt) {
1053 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1054 			return error;
1055 	} else
1056 		d->d_fileno = thisdir_kfs->kfs_fileno;
1057 	return 0;
1058 }
1059 
1060 int
1061 kernfs_readdir(v)
1062 	void *v;
1063 {
1064 	struct vop_readdir_args /* {
1065 		struct vnode *a_vp;
1066 		struct uio *a_uio;
1067 		struct ucred *a_cred;
1068 		int *a_eofflag;
1069 		off_t **a_cookies;
1070 		int a_*ncookies;
1071 	} */ *ap = v;
1072 	struct uio *uio = ap->a_uio;
1073 	struct dirent d;
1074 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1075 	const struct kern_target *kt;
1076 	const struct dyn_kern_target *dkt = NULL;
1077 	const struct kernfs_subdir *ks;
1078 	off_t i, j;
1079 	int error;
1080 	off_t *cookies = NULL;
1081 	int ncookies = 0, n;
1082 #ifdef IPSEC
1083 	struct secasvar *sav, *sav2;
1084 	struct secpolicy *sp;
1085 #endif
1086 
1087 	if (uio->uio_resid < UIO_MX)
1088 		return (EINVAL);
1089 	if (uio->uio_offset < 0)
1090 		return (EINVAL);
1091 
1092 	error = 0;
1093 	i = uio->uio_offset;
1094 	memset(&d, 0, sizeof(d));
1095 	d.d_reclen = UIO_MX;
1096 	ncookies = uio->uio_resid / UIO_MX;
1097 
1098 	switch (kfs->kfs_type) {
1099 	case KFSkern:
1100 		if (i >= nkern_targets)
1101 			return (0);
1102 
1103 		if (ap->a_ncookies) {
1104 			ncookies = min(ncookies, (nkern_targets - i));
1105 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1106 			    M_WAITOK);
1107 			*ap->a_cookies = cookies;
1108 		}
1109 
1110 		n = 0;
1111 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1112 			if (i < static_nkern_targets)
1113 				kt = &kern_targets[i];
1114 			else {
1115 				if (dkt == NULL) {
1116 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1117 					for (j = static_nkern_targets; j < i &&
1118 						     dkt != NULL; j++)
1119 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1120 					if (j != i)
1121 						break;
1122 				} else {
1123 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1124 					if (dkt == NULL)
1125 						break;
1126 				}
1127 				kt = &dkt->dkt_kt;
1128 			}
1129 			if (kt->kt_tag == KFSdevice) {
1130 				dev_t *dp = kt->kt_data;
1131 				struct vnode *fvp;
1132 
1133 				if (*dp == NODEV ||
1134 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1135 					continue;
1136 			}
1137 			d.d_namlen = kt->kt_namlen;
1138 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1139 			    &kern_targets[0], kt, ap)) != 0)
1140 				break;
1141 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1142 			d.d_type = kt->kt_type;
1143 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1144 				break;
1145 			if (cookies)
1146 				*cookies++ = i + 1;
1147 			n++;
1148 		}
1149 		ncookies = n;
1150 		break;
1151 
1152 	case KFSroot:
1153 		if (i >= 2)
1154 			return 0;
1155 
1156 		if (ap->a_ncookies) {
1157 			ncookies = min(ncookies, (2 - i));
1158 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1159 			    M_WAITOK);
1160 			*ap->a_cookies = cookies;
1161 		}
1162 
1163 		n = 0;
1164 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1165 			kt = &kern_targets[i];
1166 			d.d_namlen = kt->kt_namlen;
1167 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
1168 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1169 			d.d_type = kt->kt_type;
1170 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1171 				break;
1172 			if (cookies)
1173 				*cookies++ = i + 1;
1174 			n++;
1175 		}
1176 		ncookies = n;
1177 		break;
1178 
1179 	case KFSsubdir:
1180 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
1181 		if (i >= ks->ks_nentries)
1182 			return (0);
1183 
1184 		if (ap->a_ncookies) {
1185 			ncookies = min(ncookies, (ks->ks_nentries - i));
1186 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1187 			    M_WAITOK);
1188 			*ap->a_cookies = cookies;
1189 		}
1190 
1191 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1192 		for (j = 0; j < i && dkt != NULL; j++)
1193 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1194 		n = 0;
1195 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1196 			if (i < 2)
1197 				kt = &subdir_targets[i];
1198 			else {
1199 				/* check if ks_nentries lied to us */
1200 				if (dkt == NULL)
1201 					break;
1202 				kt = &dkt->dkt_kt;
1203 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1204 			}
1205 			if (kt->kt_tag == KFSdevice) {
1206 				dev_t *dp = kt->kt_data;
1207 				struct vnode *fvp;
1208 
1209 				if (*dp == NODEV ||
1210 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1211 					continue;
1212 			}
1213 			d.d_namlen = kt->kt_namlen;
1214 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1215 			    ks->ks_parent, kt, ap)) != 0)
1216 				break;
1217 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1218 			d.d_type = kt->kt_type;
1219 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1220 				break;
1221 			if (cookies)
1222 				*cookies++ = i + 1;
1223 			n++;
1224 		}
1225 		ncookies = n;
1226 		break;
1227 
1228 #ifdef IPSEC
1229 	case KFSipsecsadir:
1230 		/* count SA in the system */
1231 		n = 0;
1232 		TAILQ_FOREACH(sav, &satailq, tailq) {
1233 			for (sav2 = TAILQ_FIRST(&satailq);
1234 			    sav2 != sav;
1235 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1236 				if (sav->spi == sav2->spi) {
1237 					/* multiple SA with same SPI */
1238 					break;
1239 				}
1240 			}
1241 			if (sav == sav2 || sav->spi != sav2->spi)
1242 				n++;
1243 		}
1244 
1245 		if (i >= nipsecsa_targets + n)
1246 			return (0);
1247 
1248 		if (ap->a_ncookies) {
1249 			ncookies = min(ncookies, (n - i));
1250 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1251 			    M_WAITOK);
1252 			*ap->a_cookies = cookies;
1253 		}
1254 
1255 		n = 0;
1256 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1257 			kt = &ipsecsa_targets[i];
1258 			d.d_namlen = kt->kt_namlen;
1259 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1260 			    &kern_targets[0], kt, ap)) != 0)
1261 				break;
1262 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1263 			d.d_type = kt->kt_type;
1264 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1265 				break;
1266 			if (cookies)
1267 				*cookies++ = i + 1;
1268 			n++;
1269 		}
1270 		if (error) {
1271 			ncookies = n;
1272 			break;
1273 		}
1274 
1275 		TAILQ_FOREACH(sav, &satailq, tailq) {
1276 			for (sav2 = TAILQ_FIRST(&satailq);
1277 			    sav2 != sav;
1278 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1279 				if (sav->spi == sav2->spi) {
1280 					/* multiple SA with same SPI */
1281 					break;
1282 				}
1283 			}
1284 			if (sav != sav2 && sav->spi == sav2->spi)
1285 				continue;
1286 			if (uio->uio_resid < UIO_MX)
1287 				break;
1288 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1289 			    sav->spi, ap)) != 0)
1290 				break;
1291 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1292 			    "%u", ntohl(sav->spi));
1293 			d.d_type = DT_REG;
1294 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1295 				break;
1296 			if (cookies)
1297 				*cookies++ = i + 1;
1298 			n++;
1299 			i++;
1300 		}
1301 		ncookies = n;
1302 		break;
1303 
1304 	case KFSipsecspdir:
1305 		/* count SP in the system */
1306 		n = 0;
1307 		TAILQ_FOREACH(sp, &sptailq, tailq)
1308 			n++;
1309 
1310 		if (i >= nipsecsp_targets + n)
1311 			return (0);
1312 
1313 		if (ap->a_ncookies) {
1314 			ncookies = min(ncookies, (n - i));
1315 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1316 			    M_WAITOK);
1317 			*ap->a_cookies = cookies;
1318 		}
1319 
1320 		n = 0;
1321 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1322 			kt = &ipsecsp_targets[i];
1323 			d.d_namlen = kt->kt_namlen;
1324 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1325 			    &kern_targets[0], kt, ap)) != 0)
1326 				break;
1327 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1328 			d.d_type = kt->kt_type;
1329 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1330 				break;
1331 			if (cookies)
1332 				*cookies++ = i + 1;
1333 			n++;
1334 		}
1335 		if (error) {
1336 			ncookies = n;
1337 			break;
1338 		}
1339 
1340 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1341 			if (uio->uio_resid < UIO_MX)
1342 				break;
1343 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1344 			    sp->id, ap)) != 0)
1345 				break;
1346 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1347 			    "%u", sp->id);
1348 			d.d_type = DT_REG;
1349 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1350 				break;
1351 			if (cookies)
1352 				*cookies++ = i + 1;
1353 			n++;
1354 			i++;
1355 		}
1356 		ncookies = n;
1357 		break;
1358 #endif
1359 
1360 	default:
1361 		error = ENOTDIR;
1362 		break;
1363 	}
1364 
1365 	if (ap->a_ncookies) {
1366 		if (error) {
1367 			if (cookies)
1368 				free(*ap->a_cookies, M_TEMP);
1369 			*ap->a_ncookies = 0;
1370 			*ap->a_cookies = NULL;
1371 		} else
1372 			*ap->a_ncookies = ncookies;
1373 	}
1374 
1375 	uio->uio_offset = i;
1376 	return (error);
1377 }
1378 
1379 int
1380 kernfs_inactive(v)
1381 	void *v;
1382 {
1383 	struct vop_inactive_args /* {
1384 		struct vnode *a_vp;
1385 		struct lwp *a_l;
1386 	} */ *ap = v;
1387 	struct vnode *vp = ap->a_vp;
1388 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1389 #ifdef IPSEC
1390 	struct mbuf *m;
1391 	struct secpolicy *sp;
1392 #endif
1393 
1394 	VOP_UNLOCK(vp, 0);
1395 	switch (kfs->kfs_type) {
1396 #ifdef IPSEC
1397 	case KFSipsecsa:
1398 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1399 		if (m)
1400 			m_freem(m);
1401 		else
1402 			vgone(vp);
1403 		break;
1404 	case KFSipsecsp:
1405 		sp = key_getspbyid(kfs->kfs_value);
1406 		if (sp)
1407 			key_freesp(sp);
1408 		else {
1409 			/* should never happen as we hold a refcnt */
1410 			vgone(vp);
1411 		}
1412 		break;
1413 #endif
1414 	default:
1415 		break;
1416 	}
1417 	return (0);
1418 }
1419 
1420 int
1421 kernfs_reclaim(v)
1422 	void *v;
1423 {
1424 	struct vop_reclaim_args /* {
1425 		struct vnode *a_vp;
1426 	} */ *ap = v;
1427 
1428 	return (kernfs_freevp(ap->a_vp));
1429 }
1430 
1431 /*
1432  * Return POSIX pathconf information applicable to special devices.
1433  */
1434 int
1435 kernfs_pathconf(v)
1436 	void *v;
1437 {
1438 	struct vop_pathconf_args /* {
1439 		struct vnode *a_vp;
1440 		int a_name;
1441 		register_t *a_retval;
1442 	} */ *ap = v;
1443 
1444 	switch (ap->a_name) {
1445 	case _PC_LINK_MAX:
1446 		*ap->a_retval = LINK_MAX;
1447 		return (0);
1448 	case _PC_MAX_CANON:
1449 		*ap->a_retval = MAX_CANON;
1450 		return (0);
1451 	case _PC_MAX_INPUT:
1452 		*ap->a_retval = MAX_INPUT;
1453 		return (0);
1454 	case _PC_PIPE_BUF:
1455 		*ap->a_retval = PIPE_BUF;
1456 		return (0);
1457 	case _PC_CHOWN_RESTRICTED:
1458 		*ap->a_retval = 1;
1459 		return (0);
1460 	case _PC_VDISABLE:
1461 		*ap->a_retval = _POSIX_VDISABLE;
1462 		return (0);
1463 	case _PC_SYNC_IO:
1464 		*ap->a_retval = 1;
1465 		return (0);
1466 	default:
1467 		return (EINVAL);
1468 	}
1469 	/* NOTREACHED */
1470 }
1471 
1472 /*
1473  * Print out the contents of a /dev/fd vnode.
1474  */
1475 /* ARGSUSED */
1476 int
1477 kernfs_print(v)
1478 	void *v;
1479 {
1480 
1481 	printf("tag VT_KERNFS, kernfs vnode\n");
1482 	return (0);
1483 }
1484 
1485 int
1486 kernfs_link(v)
1487 	void *v;
1488 {
1489 	struct vop_link_args /* {
1490 		struct vnode *a_dvp;
1491 		struct vnode *a_vp;
1492 		struct componentname *a_cnp;
1493 	} */ *ap = v;
1494 
1495 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1496 	vput(ap->a_dvp);
1497 	return (EROFS);
1498 }
1499 
1500 int
1501 kernfs_symlink(v)
1502 	void *v;
1503 {
1504 	struct vop_symlink_args /* {
1505 		struct vnode *a_dvp;
1506 		struct vnode **a_vpp;
1507 		struct componentname *a_cnp;
1508 		struct vattr *a_vap;
1509 		char *a_target;
1510 	} */ *ap = v;
1511 
1512 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1513 	vput(ap->a_dvp);
1514 	return (EROFS);
1515 }
1516