xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision b5677b36047b601b9addaaa494a58ceae82c2a6c)
1 /*	$NetBSD: kernfs_vnops.c,v 1.136 2009/03/14 15:36:22 dsl 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 1.136 2009/03/14 15:36:22 dsl 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 int
767 kernfs_access(void *v)
768 {
769 	struct vop_access_args /* {
770 		struct vnode *a_vp;
771 		int a_mode;
772 		kauth_cred_t a_cred;
773 	} */ *ap = v;
774 	struct vattr va;
775 	int error;
776 
777 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
778 		return (error);
779 
780 	return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid,
781 	    ap->a_mode, ap->a_cred));
782 }
783 
784 static int
785 kernfs_default_fileop_getattr(void *v)
786 {
787 	struct vop_getattr_args /* {
788 		struct vnode *a_vp;
789 		struct vattr *a_vap;
790 		kauth_cred_t a_cred;
791 	} */ *ap = v;
792 	struct vattr *vap = ap->a_vap;
793 
794 	vap->va_nlink = 1;
795 	vap->va_bytes = vap->va_size = 0;
796 
797 	return 0;
798 }
799 
800 int
801 kernfs_getattr(void *v)
802 {
803 	struct vop_getattr_args /* {
804 		struct vnode *a_vp;
805 		struct vattr *a_vap;
806 		kauth_cred_t a_cred;
807 	} */ *ap = v;
808 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
809 	struct kernfs_subdir *ks;
810 	struct vattr *vap = ap->a_vap;
811 	int error = 0;
812 	char strbuf[KSTRING], *bf;
813 	size_t nread, total;
814 
815 	VATTR_NULL(vap);
816 	vap->va_type = ap->a_vp->v_type;
817 	vap->va_uid = 0;
818 	vap->va_gid = 0;
819 	vap->va_mode = kfs->kfs_mode;
820 	vap->va_fileid = kfs->kfs_fileno;
821 	vap->va_flags = 0;
822 	vap->va_size = 0;
823 	vap->va_blocksize = DEV_BSIZE;
824 	/* Make all times be current TOD, except for the "boottime" node. */
825 	if (kfs->kfs_kt->kt_namlen == 8 &&
826 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
827 		vap->va_ctime = boottime;
828 	} else {
829 		getnanotime(&vap->va_ctime);
830 	}
831 	vap->va_atime = vap->va_mtime = vap->va_ctime;
832 	vap->va_gen = 0;
833 	vap->va_flags = 0;
834 	vap->va_rdev = 0;
835 	vap->va_bytes = 0;
836 
837 	switch (kfs->kfs_type) {
838 	case KFSkern:
839 		vap->va_nlink = nkern_dirs;
840 		vap->va_bytes = vap->va_size = DEV_BSIZE;
841 		break;
842 
843 	case KFSroot:
844 		vap->va_nlink = 1;
845 		vap->va_bytes = vap->va_size = DEV_BSIZE;
846 		break;
847 
848 	case KFSsubdir:
849 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
850 		vap->va_nlink = ks->ks_dirs;
851 		vap->va_bytes = vap->va_size = DEV_BSIZE;
852 		break;
853 
854 	case KFSnull:
855 	case KFStime:
856 	case KFSint:
857 	case KFSstring:
858 	case KFShostname:
859 	case KFSavenrun:
860 	case KFSdevice:
861 	case KFSmsgbuf:
862 #ifdef IPSEC
863 	case KFSipsecsa:
864 	case KFSipsecsp:
865 #endif
866 		vap->va_nlink = 1;
867 		total = 0;
868 		do {
869 			bf = strbuf;
870 			error = kernfs_xread(kfs, total, &bf,
871 			    sizeof(strbuf), &nread);
872 			total += nread;
873 		} while (error == 0 && nread != 0);
874 		vap->va_bytes = vap->va_size = total;
875 		break;
876 
877 #ifdef IPSEC
878 	case KFSipsecsadir:
879 	case KFSipsecspdir:
880 		vap->va_nlink = 2;
881 		vap->va_bytes = vap->va_size = DEV_BSIZE;
882 		break;
883 #endif
884 
885 	default:
886 		error = kernfs_try_fileop(kfs->kfs_type,
887 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
888 		break;
889 	}
890 
891 	return (error);
892 }
893 
894 /*ARGSUSED*/
895 int
896 kernfs_setattr(void *v)
897 {
898 
899 	/*
900 	 * Silently ignore attribute changes.
901 	 * This allows for open with truncate to have no
902 	 * effect until some data is written.  I want to
903 	 * do it this way because all writes are atomic.
904 	 */
905 	return (0);
906 }
907 
908 int
909 kernfs_default_xread(void *v)
910 {
911 	struct vop_read_args /* {
912 		struct vnode *a_vp;
913 		struct uio *a_uio;
914 		int  a_ioflag;
915 		kauth_cred_t a_cred;
916 	} */ *ap = v;
917 	struct uio *uio = ap->a_uio;
918 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
919 	char strbuf[KSTRING], *bf;
920 	int off;
921 	size_t len;
922 	int error;
923 
924 	if (ap->a_vp->v_type == VDIR)
925 		return (EOPNOTSUPP);
926 
927 	off = (int)uio->uio_offset;
928 	/* Don't allow negative offsets */
929 	if (off < 0)
930 		return EINVAL;
931 
932 	bf = strbuf;
933 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
934 		error = uiomove(bf, len, uio);
935 	return (error);
936 }
937 
938 int
939 kernfs_read(void *v)
940 {
941 	struct vop_read_args /* {
942 		struct vnode *a_vp;
943 		struct uio *a_uio;
944 		int  a_ioflag;
945 		struct ucred *a_cred;
946 	} */ *ap = v;
947 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
948 
949 	if (kfs->kfs_type < KFSlasttype) {
950 		/* use default function */
951 		return kernfs_default_xread(v);
952 	}
953 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
954 	   EOPNOTSUPP);
955 }
956 
957 static int
958 kernfs_default_xwrite(void *v)
959 {
960 	struct vop_write_args /* {
961 		struct vnode *a_vp;
962 		struct uio *a_uio;
963 		int  a_ioflag;
964 		kauth_cred_t a_cred;
965 	} */ *ap = v;
966 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
967 	struct uio *uio = ap->a_uio;
968 	int error;
969 	size_t xlen;
970 	char strbuf[KSTRING];
971 
972 	if (uio->uio_offset != 0)
973 		return (EINVAL);
974 
975 	xlen = min(uio->uio_resid, KSTRING-1);
976 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
977 		return (error);
978 
979 	if (uio->uio_resid != 0)
980 		return (EIO);
981 
982 	strbuf[xlen] = '\0';
983 	xlen = strlen(strbuf);
984 	return (kernfs_xwrite(kfs, strbuf, xlen));
985 }
986 
987 int
988 kernfs_write(void *v)
989 {
990 	struct vop_write_args /* {
991 		struct vnode *a_vp;
992 		struct uio *a_uio;
993 		int  a_ioflag;
994 		kauth_cred_t a_cred;
995 	} */ *ap = v;
996 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
997 
998 	if (kfs->kfs_type < KFSlasttype) {
999 		/* use default function */
1000 		return kernfs_default_xwrite(v);
1001 	}
1002 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
1003 	    EOPNOTSUPP);
1004 }
1005 
1006 int
1007 kernfs_ioctl(void *v)
1008 {
1009 	struct vop_ioctl_args /* {
1010 		const struct vnodeop_desc *a_desc;
1011 		struct vnode *a_vp;
1012 		u_long a_command;
1013 		void *a_data;
1014 		int a_fflag;
1015 		kauth_cred_t a_cred;
1016 	} */ *ap = v;
1017 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1018 
1019 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1020 	    EPASSTHROUGH);
1021 }
1022 
1023 static int
1024 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1025     u_int32_t value, struct vop_readdir_args *ap)
1026 {
1027 	struct kernfs_node *kfs;
1028 	struct vnode *vp;
1029 	int error;
1030 
1031 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1032 	    value)) != 0)
1033 		return error;
1034 	if (kt->kt_tag == KFSdevice) {
1035 		struct vattr va;
1036 
1037 		error = VOP_GETATTR(vp, &va, ap->a_cred);
1038 		if (error != 0) {
1039 			return error;
1040 		}
1041 		d->d_fileno = va.va_fileid;
1042 	} else {
1043 		kfs = VTOKERN(vp);
1044 		d->d_fileno = kfs->kfs_fileno;
1045 	}
1046 	vput(vp);
1047 	return 0;
1048 }
1049 
1050 static int
1051 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1052     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1053     const struct kern_target *kt, struct vop_readdir_args *ap)
1054 {
1055 	const struct kern_target *ikt;
1056 	int error;
1057 
1058 	switch (entry) {
1059 	case 0:
1060 		d->d_fileno = thisdir_kfs->kfs_fileno;
1061 		return 0;
1062 	case 1:
1063 		ikt = parent_kt;
1064 		break;
1065 	default:
1066 		ikt = kt;
1067 		break;
1068 	}
1069 	if (ikt != thisdir_kfs->kfs_kt) {
1070 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1071 			return error;
1072 	} else
1073 		d->d_fileno = thisdir_kfs->kfs_fileno;
1074 	return 0;
1075 }
1076 
1077 int
1078 kernfs_readdir(void *v)
1079 {
1080 	struct vop_readdir_args /* {
1081 		struct vnode *a_vp;
1082 		struct uio *a_uio;
1083 		kauth_cred_t a_cred;
1084 		int *a_eofflag;
1085 		off_t **a_cookies;
1086 		int a_*ncookies;
1087 	} */ *ap = v;
1088 	struct uio *uio = ap->a_uio;
1089 	struct dirent d;
1090 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1091 	const struct kern_target *kt;
1092 	const struct dyn_kern_target *dkt = NULL;
1093 	const struct kernfs_subdir *ks;
1094 	off_t i, j;
1095 	int error;
1096 	off_t *cookies = NULL;
1097 	int ncookies = 0, n;
1098 #ifdef IPSEC
1099 	struct secasvar *sav, *sav2;
1100 	struct secpolicy *sp;
1101 #endif
1102 
1103 	if (uio->uio_resid < UIO_MX)
1104 		return (EINVAL);
1105 	if (uio->uio_offset < 0)
1106 		return (EINVAL);
1107 
1108 	error = 0;
1109 	i = uio->uio_offset;
1110 	memset(&d, 0, sizeof(d));
1111 	d.d_reclen = UIO_MX;
1112 	ncookies = uio->uio_resid / UIO_MX;
1113 
1114 	switch (kfs->kfs_type) {
1115 	case KFSkern:
1116 		if (i >= nkern_targets)
1117 			return (0);
1118 
1119 		if (ap->a_ncookies) {
1120 			ncookies = min(ncookies, (nkern_targets - i));
1121 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1122 			    M_WAITOK);
1123 			*ap->a_cookies = cookies;
1124 		}
1125 
1126 		n = 0;
1127 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1128 			if (i < static_nkern_targets)
1129 				kt = &kern_targets[i];
1130 			else {
1131 				if (dkt == NULL) {
1132 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1133 					for (j = static_nkern_targets; j < i &&
1134 						     dkt != NULL; j++)
1135 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1136 					if (j != i)
1137 						break;
1138 				} else {
1139 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1140 				}
1141 				if (dkt == NULL)
1142 					break;
1143 				kt = &dkt->dkt_kt;
1144 			}
1145 			if (kt->kt_tag == KFSdevice) {
1146 				dev_t *dp = kt->kt_data;
1147 				struct vnode *fvp;
1148 
1149 				if (*dp == NODEV ||
1150 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1151 					continue;
1152 			}
1153 			d.d_namlen = kt->kt_namlen;
1154 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1155 			    &kern_targets[0], kt, ap)) != 0)
1156 				break;
1157 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1158 			d.d_type = kt->kt_type;
1159 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1160 				break;
1161 			if (cookies)
1162 				*cookies++ = i + 1;
1163 			n++;
1164 		}
1165 		ncookies = n;
1166 		break;
1167 
1168 	case KFSroot:
1169 		if (i >= 2)
1170 			return 0;
1171 
1172 		if (ap->a_ncookies) {
1173 			ncookies = min(ncookies, (2 - i));
1174 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1175 			    M_WAITOK);
1176 			*ap->a_cookies = cookies;
1177 		}
1178 
1179 		n = 0;
1180 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
1181 			kt = &kern_targets[i];
1182 			d.d_namlen = kt->kt_namlen;
1183 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
1184 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1185 			d.d_type = kt->kt_type;
1186 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1187 				break;
1188 			if (cookies)
1189 				*cookies++ = i + 1;
1190 			n++;
1191 		}
1192 		ncookies = n;
1193 		break;
1194 
1195 	case KFSsubdir:
1196 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
1197 		if (i >= ks->ks_nentries)
1198 			return (0);
1199 
1200 		if (ap->a_ncookies) {
1201 			ncookies = min(ncookies, (ks->ks_nentries - i));
1202 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1203 			    M_WAITOK);
1204 			*ap->a_cookies = cookies;
1205 		}
1206 
1207 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1208 		for (j = 0; j < i && dkt != NULL; j++)
1209 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1210 		n = 0;
1211 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1212 			if (i < 2)
1213 				kt = &subdir_targets[i];
1214 			else {
1215 				/* check if ks_nentries lied to us */
1216 				if (dkt == NULL)
1217 					break;
1218 				kt = &dkt->dkt_kt;
1219 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1220 			}
1221 			if (kt->kt_tag == KFSdevice) {
1222 				dev_t *dp = kt->kt_data;
1223 				struct vnode *fvp;
1224 
1225 				if (*dp == NODEV ||
1226 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1227 					continue;
1228 			}
1229 			d.d_namlen = kt->kt_namlen;
1230 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1231 			    ks->ks_parent, kt, ap)) != 0)
1232 				break;
1233 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1234 			d.d_type = kt->kt_type;
1235 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1236 				break;
1237 			if (cookies)
1238 				*cookies++ = i + 1;
1239 			n++;
1240 		}
1241 		ncookies = n;
1242 		break;
1243 
1244 #ifdef IPSEC
1245 	case KFSipsecsadir:
1246 		/* count SA in the system */
1247 		n = 0;
1248 		if (&satailq == NULL)
1249 			return 0;
1250 		TAILQ_FOREACH(sav, &satailq, tailq) {
1251 			for (sav2 = TAILQ_FIRST(&satailq);
1252 			    sav2 != sav;
1253 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1254 				if (sav->spi == sav2->spi) {
1255 					/* multiple SA with same SPI */
1256 					break;
1257 				}
1258 			}
1259 			if (sav == sav2 || sav->spi != sav2->spi)
1260 				n++;
1261 		}
1262 
1263 		if (i >= nipsecsa_targets + n)
1264 			return (0);
1265 
1266 		if (ap->a_ncookies) {
1267 			ncookies = min(ncookies, (n - i));
1268 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1269 			    M_WAITOK);
1270 			*ap->a_cookies = cookies;
1271 		}
1272 
1273 		n = 0;
1274 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1275 			kt = &ipsecsa_targets[i];
1276 			d.d_namlen = kt->kt_namlen;
1277 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1278 			    &kern_targets[0], kt, ap)) != 0)
1279 				break;
1280 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1281 			d.d_type = kt->kt_type;
1282 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1283 				break;
1284 			if (cookies)
1285 				*cookies++ = i + 1;
1286 			n++;
1287 		}
1288 		if (error) {
1289 			ncookies = n;
1290 			break;
1291 		}
1292 
1293 		TAILQ_FOREACH(sav, &satailq, tailq) {
1294 			for (sav2 = TAILQ_FIRST(&satailq);
1295 			    sav2 != sav;
1296 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1297 				if (sav->spi == sav2->spi) {
1298 					/* multiple SA with same SPI */
1299 					break;
1300 				}
1301 			}
1302 			if (sav != sav2 && sav->spi == sav2->spi)
1303 				continue;
1304 			if (uio->uio_resid < UIO_MX)
1305 				break;
1306 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1307 			    sav->spi, ap)) != 0)
1308 				break;
1309 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1310 			    "%u", ntohl(sav->spi));
1311 			d.d_type = DT_REG;
1312 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1313 				break;
1314 			if (cookies)
1315 				*cookies++ = i + 1;
1316 			n++;
1317 			i++;
1318 		}
1319 		ncookies = n;
1320 		break;
1321 
1322 	case KFSipsecspdir:
1323 		/* count SP in the system */
1324 		if (&sptailq == NULL)
1325 			return 0;
1326 
1327 		n = 0;
1328 		TAILQ_FOREACH(sp, &sptailq, tailq)
1329 			n++;
1330 
1331 		if (i >= nipsecsp_targets + n)
1332 			return (0);
1333 
1334 		if (ap->a_ncookies) {
1335 			ncookies = min(ncookies, (n - i));
1336 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1337 			    M_WAITOK);
1338 			*ap->a_cookies = cookies;
1339 		}
1340 
1341 		n = 0;
1342 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1343 			kt = &ipsecsp_targets[i];
1344 			d.d_namlen = kt->kt_namlen;
1345 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1346 			    &kern_targets[0], kt, ap)) != 0)
1347 				break;
1348 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1349 			d.d_type = kt->kt_type;
1350 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1351 				break;
1352 			if (cookies)
1353 				*cookies++ = i + 1;
1354 			n++;
1355 		}
1356 		if (error) {
1357 			ncookies = n;
1358 			break;
1359 		}
1360 
1361 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1362 			if (uio->uio_resid < UIO_MX)
1363 				break;
1364 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1365 			    sp->id, ap)) != 0)
1366 				break;
1367 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1368 			    "%u", sp->id);
1369 			d.d_type = DT_REG;
1370 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1371 				break;
1372 			if (cookies)
1373 				*cookies++ = i + 1;
1374 			n++;
1375 			i++;
1376 		}
1377 		ncookies = n;
1378 		break;
1379 #endif
1380 
1381 	default:
1382 		error = ENOTDIR;
1383 		break;
1384 	}
1385 
1386 	if (ap->a_ncookies) {
1387 		if (error) {
1388 			if (cookies)
1389 				free(*ap->a_cookies, M_TEMP);
1390 			*ap->a_ncookies = 0;
1391 			*ap->a_cookies = NULL;
1392 		} else
1393 			*ap->a_ncookies = ncookies;
1394 	}
1395 
1396 	uio->uio_offset = i;
1397 	return (error);
1398 }
1399 
1400 int
1401 kernfs_inactive(void *v)
1402 {
1403 	struct vop_inactive_args /* {
1404 		struct vnode *a_vp;
1405 		bool *a_recycle;
1406 	} */ *ap = v;
1407 	struct vnode *vp = ap->a_vp;
1408 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1409 #ifdef IPSEC
1410 	struct mbuf *m;
1411 	struct secpolicy *sp;
1412 #endif
1413 
1414 	*ap->a_recycle = false;
1415 	switch (kfs->kfs_type) {
1416 #ifdef IPSEC
1417 	case KFSipsecsa:
1418 		if (key_setdumpsa_spi == NULL)
1419 			return 0;
1420 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1421 		if (m)
1422 			m_freem(m);
1423 		else
1424 			*ap->a_recycle = true;
1425 		break;
1426 	case KFSipsecsp:
1427 		if (key_getspbyid == NULL)
1428 			return 0;
1429 		sp = key_getspbyid(kfs->kfs_value);
1430 		if (sp)
1431 			key_freesp(sp);
1432 		else {
1433 			*ap->a_recycle = true;
1434 		}
1435 		break;
1436 #endif
1437 	default:
1438 		break;
1439 	}
1440 	VOP_UNLOCK(vp, 0);
1441 	return (0);
1442 }
1443 
1444 int
1445 kernfs_reclaim(void *v)
1446 {
1447 	struct vop_reclaim_args /* {
1448 		struct vnode *a_vp;
1449 	} */ *ap = v;
1450 
1451 	return (kernfs_freevp(ap->a_vp));
1452 }
1453 
1454 /*
1455  * Return POSIX pathconf information applicable to special devices.
1456  */
1457 int
1458 kernfs_pathconf(void *v)
1459 {
1460 	struct vop_pathconf_args /* {
1461 		struct vnode *a_vp;
1462 		int a_name;
1463 		register_t *a_retval;
1464 	} */ *ap = v;
1465 
1466 	switch (ap->a_name) {
1467 	case _PC_LINK_MAX:
1468 		*ap->a_retval = LINK_MAX;
1469 		return (0);
1470 	case _PC_MAX_CANON:
1471 		*ap->a_retval = MAX_CANON;
1472 		return (0);
1473 	case _PC_MAX_INPUT:
1474 		*ap->a_retval = MAX_INPUT;
1475 		return (0);
1476 	case _PC_PIPE_BUF:
1477 		*ap->a_retval = PIPE_BUF;
1478 		return (0);
1479 	case _PC_CHOWN_RESTRICTED:
1480 		*ap->a_retval = 1;
1481 		return (0);
1482 	case _PC_VDISABLE:
1483 		*ap->a_retval = _POSIX_VDISABLE;
1484 		return (0);
1485 	case _PC_SYNC_IO:
1486 		*ap->a_retval = 1;
1487 		return (0);
1488 	default:
1489 		return (EINVAL);
1490 	}
1491 	/* NOTREACHED */
1492 }
1493 
1494 /*
1495  * Print out the contents of a /dev/fd vnode.
1496  */
1497 /* ARGSUSED */
1498 int
1499 kernfs_print(void *v)
1500 {
1501 
1502 	printf("tag VT_KERNFS, kernfs vnode\n");
1503 	return (0);
1504 }
1505 
1506 int
1507 kernfs_link(void *v)
1508 {
1509 	struct vop_link_args /* {
1510 		struct vnode *a_dvp;
1511 		struct vnode *a_vp;
1512 		struct componentname *a_cnp;
1513 	} */ *ap = v;
1514 
1515 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1516 	vput(ap->a_dvp);
1517 	return (EROFS);
1518 }
1519 
1520 int
1521 kernfs_symlink(void *v)
1522 {
1523 	struct vop_symlink_args /* {
1524 		struct vnode *a_dvp;
1525 		struct vnode **a_vpp;
1526 		struct componentname *a_cnp;
1527 		struct vattr *a_vap;
1528 		char *a_target;
1529 	} */ *ap = v;
1530 
1531 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1532 	vput(ap->a_dvp);
1533 	return (EROFS);
1534 }
1535