xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision c5e820cae412164fcbee52f470436200af5358ea)
1 /*	$NetBSD: kernfs_vnops.c,v 1.145 2012/03/13 18:40:57 elad Exp $	*/
2 
3 /*
4  * Copyright (c) 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software donated to Berkeley by
8  * Jan-Simon Pendry.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)kernfs_vnops.c	8.15 (Berkeley) 5/21/95
35  */
36 
37 /*
38  * Kernel parameter filesystem (/kern)
39  */
40 
41 #include <sys/cdefs.h>
42 __KERNEL_RCSID(0, "$NetBSD: kernfs_vnops.c,v 1.145 2012/03/13 18:40:57 elad Exp $");
43 
44 #ifdef _KERNEL_OPT
45 #include "opt_ipsec.h"
46 #endif
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/kernel.h>
51 #include <sys/vmmeter.h>
52 #include <sys/time.h>
53 #include <sys/proc.h>
54 #include <sys/vnode.h>
55 #include <sys/malloc.h>
56 #include <sys/file.h>
57 #include <sys/stat.h>
58 #include <sys/mount.h>
59 #include <sys/namei.h>
60 #include <sys/buf.h>
61 #include <sys/dirent.h>
62 #include <sys/msgbuf.h>
63 
64 #include <miscfs/genfs/genfs.h>
65 #include <miscfs/kernfs/kernfs.h>
66 
67 #ifdef IPSEC
68 #include <sys/mbuf.h>
69 #include <net/route.h>
70 #include <netinet/in.h>
71 #include <netinet6/ipsec.h>
72 #include <netkey/key.h>
73 #endif
74 
75 #include <uvm/uvm_extern.h>
76 
77 #define KSTRING	256		/* Largest I/O available via this filesystem */
78 #define	UIO_MX 32
79 
80 #define	READ_MODE	(S_IRUSR|S_IRGRP|S_IROTH)
81 #define	WRITE_MODE	(S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH)
82 #define	UREAD_MODE	(S_IRUSR)
83 #define	DIR_MODE	(S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
84 #define	UDIR_MODE	(S_IRUSR|S_IXUSR)
85 
86 #define N(s) sizeof(s)-1, s
87 const struct kern_target kern_targets[] = {
88 /* NOTE: The name must be less than UIO_MX-16 chars in length */
89      /*        name            data          tag           type  ro/rw */
90      { DT_DIR, N("."),         0,            KFSkern,        VDIR, DIR_MODE   },
91      { DT_DIR, N(".."),        0,            KFSroot,        VDIR, DIR_MODE   },
92      { DT_REG, N("boottime"),  &boottime.tv_sec, KFSint,     VREG, READ_MODE  },
93 			/* XXXUNCONST */
94      { DT_REG, N("copyright"), __UNCONST(copyright),
95      					     KFSstring,      VREG, READ_MODE  },
96      { DT_REG, N("hostname"),  0,            KFShostname,    VREG, WRITE_MODE },
97      { DT_REG, N("hz"),        &hz,          KFSint,         VREG, READ_MODE  },
98 #ifdef IPSEC
99      { DT_DIR, N("ipsecsa"),   0,	     KFSipsecsadir,  VDIR, UDIR_MODE  },
100      { DT_DIR, N("ipsecsp"),   0,	     KFSipsecspdir,  VDIR, UDIR_MODE  },
101 #endif
102      { DT_REG, N("loadavg"),   0,            KFSavenrun,     VREG, READ_MODE  },
103      { DT_REG, N("msgbuf"),    0,	     KFSmsgbuf,      VREG, READ_MODE  },
104      { DT_REG, N("pagesize"),  &uvmexp.pagesize, KFSint,     VREG, READ_MODE  },
105      { DT_REG, N("physmem"),   &physmem,     KFSint,         VREG, READ_MODE  },
106 #if 0
107      { DT_DIR, N("root"),      0,            KFSnull,        VDIR, DIR_MODE   },
108 #endif
109      { DT_BLK, N("rootdev"),   &rootdev,     KFSdevice,      VBLK, READ_MODE  },
110      { DT_CHR, N("rrootdev"),  &rrootdev,    KFSdevice,      VCHR, READ_MODE  },
111      { DT_REG, N("time"),      0,            KFStime,        VREG, READ_MODE  },
112 			/* XXXUNCONST */
113      { DT_REG, N("version"),   __UNCONST(version),
114      					     KFSstring,      VREG, READ_MODE  },
115 };
116 const struct kern_target subdir_targets[] = {
117 /* NOTE: The name must be less than UIO_MX-16 chars in length */
118      /*        name            data          tag           type  ro/rw */
119      { DT_DIR, N("."),         0,            KFSsubdir,      VDIR, DIR_MODE   },
120      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
121 };
122 #ifdef IPSEC
123 const struct kern_target ipsecsa_targets[] = {
124 /* NOTE: The name must be less than UIO_MX-16 chars in length */
125      /*        name            data          tag           type  ro/rw */
126      { DT_DIR, N("."),         0,            KFSipsecsadir,  VDIR, DIR_MODE   },
127      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
128 };
129 const struct kern_target ipsecsp_targets[] = {
130 /* NOTE: The name must be less than UIO_MX-16 chars in length */
131      /*        name            data          tag           type  ro/rw */
132      { DT_DIR, N("."),         0,            KFSipsecspdir,  VDIR, DIR_MODE   },
133      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
134 };
135 const struct kern_target ipsecsa_kt =
136      { DT_DIR, N(""),          0,            KFSipsecsa,     VREG, UREAD_MODE };
137 const struct kern_target ipsecsp_kt =
138      { DT_DIR, N(""),          0,            KFSipsecsp,     VREG, UREAD_MODE };
139 #endif
140 #undef N
141 SIMPLEQ_HEAD(,dyn_kern_target) dyn_kern_targets =
142 	SIMPLEQ_HEAD_INITIALIZER(dyn_kern_targets);
143 int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
144 const int static_nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
145 #ifdef IPSEC
146 int nipsecsa_targets = sizeof(ipsecsa_targets) / sizeof(ipsecsa_targets[0]);
147 int nipsecsp_targets = sizeof(ipsecsp_targets) / sizeof(ipsecsp_targets[0]);
148 int nkern_dirs = 4; /* 2 extra subdirs */
149 #else
150 int nkern_dirs = 2;
151 #endif
152 
153 int kernfs_try_fileop(kfstype, kfsfileop, void *, int);
154 int kernfs_try_xread(kfstype, const struct kernfs_node *, char **,
155     size_t, int);
156 int kernfs_try_xwrite(kfstype, const struct kernfs_node *, char *,
157     size_t, int);
158 
159 static int kernfs_default_xread(void *v);
160 static int kernfs_default_xwrite(void *v);
161 static int kernfs_default_fileop_getattr(void *);
162 
163 /* must include all fileop's */
164 const struct kernfs_fileop kernfs_default_fileops[] = {
165   { .kf_fileop = KERNFS_XREAD },
166   { .kf_fileop = KERNFS_XWRITE },
167   { .kf_fileop = KERNFS_FILEOP_OPEN },
168   { .kf_fileop = KERNFS_FILEOP_GETATTR,
169     .kf_vop = kernfs_default_fileop_getattr },
170   { .kf_fileop = KERNFS_FILEOP_IOCTL },
171   { .kf_fileop = KERNFS_FILEOP_CLOSE },
172   { .kf_fileop = KERNFS_FILEOP_READ,
173     .kf_vop = kernfs_default_xread },
174   { .kf_fileop = KERNFS_FILEOP_WRITE,
175     .kf_vop = kernfs_default_xwrite },
176 };
177 
178 int	kernfs_lookup(void *);
179 #define	kernfs_create	genfs_eopnotsupp
180 #define	kernfs_mknod	genfs_eopnotsupp
181 int	kernfs_open(void *);
182 int	kernfs_close(void *);
183 int	kernfs_access(void *);
184 int	kernfs_getattr(void *);
185 int	kernfs_setattr(void *);
186 int	kernfs_read(void *);
187 int	kernfs_write(void *);
188 #define	kernfs_fcntl	genfs_fcntl
189 int	kernfs_ioctl(void *);
190 #define	kernfs_poll	genfs_poll
191 #define kernfs_revoke	genfs_revoke
192 #define	kernfs_fsync	genfs_nullop
193 #define	kernfs_seek	genfs_nullop
194 #define	kernfs_remove	genfs_eopnotsupp
195 int	kernfs_link(void *);
196 #define	kernfs_rename	genfs_eopnotsupp
197 #define	kernfs_mkdir	genfs_eopnotsupp
198 #define	kernfs_rmdir	genfs_eopnotsupp
199 int	kernfs_symlink(void *);
200 int	kernfs_readdir(void *);
201 #define	kernfs_readlink	genfs_eopnotsupp
202 #define	kernfs_abortop	genfs_abortop
203 int	kernfs_inactive(void *);
204 int	kernfs_reclaim(void *);
205 #define	kernfs_lock	genfs_lock
206 #define	kernfs_unlock	genfs_unlock
207 #define	kernfs_bmap	genfs_badop
208 #define	kernfs_strategy	genfs_badop
209 int	kernfs_print(void *);
210 #define	kernfs_islocked	genfs_islocked
211 int	kernfs_pathconf(void *);
212 #define	kernfs_advlock	genfs_einval
213 #define	kernfs_bwrite	genfs_eopnotsupp
214 #define	kernfs_putpages	genfs_putpages
215 
216 static int	kernfs_xread(struct kernfs_node *, int, char **,
217 				size_t, size_t *);
218 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
219 
220 int (**kernfs_vnodeop_p)(void *);
221 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
222 	{ &vop_default_desc, vn_default_error },
223 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
224 	{ &vop_create_desc, kernfs_create },		/* create */
225 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
226 	{ &vop_open_desc, kernfs_open },		/* open */
227 	{ &vop_close_desc, kernfs_close },		/* close */
228 	{ &vop_access_desc, kernfs_access },		/* access */
229 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
230 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
231 	{ &vop_read_desc, kernfs_read },		/* read */
232 	{ &vop_write_desc, kernfs_write },		/* write */
233 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
234 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
235 	{ &vop_poll_desc, kernfs_poll },		/* poll */
236 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
237 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
238 	{ &vop_seek_desc, kernfs_seek },		/* seek */
239 	{ &vop_remove_desc, kernfs_remove },		/* remove */
240 	{ &vop_link_desc, kernfs_link },		/* link */
241 	{ &vop_rename_desc, kernfs_rename },		/* rename */
242 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
243 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
244 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
245 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
246 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
247 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
248 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
249 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
250 	{ &vop_lock_desc, kernfs_lock },		/* lock */
251 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
252 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
253 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
254 	{ &vop_print_desc, kernfs_print },		/* print */
255 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
256 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
257 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
258 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
259 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
260 	{ NULL, NULL }
261 };
262 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
263 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
264 
265 static inline int
266 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
267 {
268 	if (a->kf_type < b->kf_type)
269 		return -1;
270 	if (a->kf_type > b->kf_type)
271 		return 1;
272 	if (a->kf_fileop < b->kf_fileop)
273 		return -1;
274 	if (a->kf_fileop > b->kf_fileop)
275 		return 1;
276 	return (0);
277 }
278 
279 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
280 	SPLAY_INITIALIZER(kfsfileoptree);
281 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
282 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
283 
284 kfstype
285 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
286 {
287 	static u_char nextfreetype = KFSlasttype;
288 	struct kernfs_fileop *dkf, *fkf, skf;
289 	int i;
290 
291 	/* XXX need to keep track of dkf's memory if we support
292            deallocating types */
293 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
294 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
295 
296 	for (i = 0; i < sizeof(kernfs_default_fileops) /
297 		     sizeof(kernfs_default_fileops[0]); i++) {
298 		dkf[i].kf_type = nextfreetype;
299 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
300 	}
301 
302 	for (i = 0; i < nkf; i++) {
303 		skf.kf_type = nextfreetype;
304 		skf.kf_fileop = kf[i].kf_fileop;
305 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
306 			fkf->kf_vop = kf[i].kf_vop;
307 	}
308 
309 	return nextfreetype++;
310 }
311 
312 int
313 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
314 {
315 	struct kernfs_fileop *kf, skf;
316 
317 	skf.kf_type = type;
318 	skf.kf_fileop = fileop;
319 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
320 		if (kf->kf_vop)
321 			return kf->kf_vop(v);
322 	return error;
323 }
324 
325 int
326 kernfs_try_xread(kfstype type, const struct kernfs_node *kfs, char **bfp,
327     size_t len, int error)
328 {
329 	struct kernfs_fileop *kf, skf;
330 
331 	skf.kf_type = type;
332 	skf.kf_fileop = KERNFS_XREAD;
333 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
334 		if (kf->kf_xread)
335 			return kf->kf_xread(kfs, bfp, len);
336 	return error;
337 }
338 
339 int
340 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
341     size_t len, int error)
342 {
343 	struct kernfs_fileop *kf, skf;
344 
345 	skf.kf_type = type;
346 	skf.kf_fileop = KERNFS_XWRITE;
347 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
348 		if (kf->kf_xwrite)
349 			return kf->kf_xwrite(kfs, bf, len);
350 	return error;
351 }
352 
353 int
354 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
355 {
356 	struct kernfs_subdir *ks, *parent;
357 
358 	if (pkt == NULL) {
359 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
360 		nkern_targets++;
361 		if (dkt->dkt_kt.kt_vtype == VDIR)
362 			nkern_dirs++;
363 	} else {
364 		parent = (struct kernfs_subdir *)pkt->kt_data;
365 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
366 		parent->ks_nentries++;
367 		if (dkt->dkt_kt.kt_vtype == VDIR)
368 			parent->ks_dirs++;
369 	}
370 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
371 		ks = malloc(sizeof(struct kernfs_subdir),
372 		    M_TEMP, M_WAITOK);
373 		SIMPLEQ_INIT(&ks->ks_entries);
374 		ks->ks_nentries = 2; /* . and .. */
375 		ks->ks_dirs = 2;
376 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
377 		dkt->dkt_kt.kt_data = ks;
378 	}
379 	return 0;
380 }
381 
382 static int
383 kernfs_xread(struct kernfs_node *kfs, int off, char **bufp, size_t len, size_t *wrlen)
384 {
385 	const struct kern_target *kt;
386 #ifdef IPSEC
387 	struct mbuf *m;
388 #endif
389 	int err;
390 
391 	kt = kfs->kfs_kt;
392 
393 	switch (kfs->kfs_type) {
394 	case KFStime: {
395 		struct timeval tv;
396 
397 		microtime(&tv);
398 		snprintf(*bufp, len, "%lld %ld\n", (long long)tv.tv_sec,
399 		    (long)tv.tv_usec);
400 		break;
401 	}
402 
403 	case KFSint: {
404 		int *ip = kt->kt_data;
405 
406 		snprintf(*bufp, len, "%d\n", *ip);
407 		break;
408 	}
409 
410 	case KFSstring: {
411 		char *cp = kt->kt_data;
412 
413 		*bufp = cp;
414 		break;
415 	}
416 
417 	case KFSmsgbuf: {
418 		long n;
419 
420 		/*
421 		 * deal with cases where the message buffer has
422 		 * become corrupted.
423 		 */
424 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
425 			msgbufenabled = 0;
426 			return (ENXIO);
427 		}
428 
429 		/*
430 		 * Note that reads of /kern/msgbuf won't necessarily yield
431 		 * consistent results, if the message buffer is modified
432 		 * while the read is in progress.  The worst that can happen
433 		 * is that incorrect data will be read.  There's no way
434 		 * that this can crash the system unless the values in the
435 		 * message buffer header are corrupted, but that'll cause
436 		 * the system to die anyway.
437 		 */
438 		if (off >= msgbufp->msg_bufs) {
439 			*wrlen = 0;
440 			return (0);
441 		}
442 		n = msgbufp->msg_bufx + off;
443 		if (n >= msgbufp->msg_bufs)
444 			n -= msgbufp->msg_bufs;
445 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
446 		*bufp = msgbufp->msg_bufc + n;
447 		*wrlen = len;
448 		return (0);
449 	}
450 
451 	case KFShostname: {
452 		char *cp = hostname;
453 		size_t xlen = hostnamelen;
454 
455 		if (xlen >= (len - 2))
456 			return (EINVAL);
457 
458 		memcpy(*bufp, cp, xlen);
459 		(*bufp)[xlen] = '\n';
460 		(*bufp)[xlen+1] = '\0';
461 		break;
462 	}
463 
464 	case KFSavenrun:
465 		averunnable.fscale = FSCALE;
466 		snprintf(*bufp, len, "%d %d %d %ld\n",
467 		    averunnable.ldavg[0], averunnable.ldavg[1],
468 		    averunnable.ldavg[2], averunnable.fscale);
469 		break;
470 
471 #ifdef IPSEC
472 	case KFSipsecsa:
473 		if (key_setdumpsa_spi == NULL)
474 			return 0;
475 		/*
476 		 * Note that SA configuration could be changed during the
477 		 * read operation, resulting in garbled output.
478 		 */
479 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
480 		if (!m)
481 			return (ENOBUFS);
482 		if (off >= m->m_pkthdr.len) {
483 			*wrlen = 0;
484 			m_freem(m);
485 			return (0);
486 		}
487 		if (len > m->m_pkthdr.len - off)
488 			len = m->m_pkthdr.len - off;
489 		m_copydata(m, off, len, *bufp);
490 		*wrlen = len;
491 		m_freem(m);
492 		return (0);
493 
494 	case KFSipsecsp:
495 		/*
496 		 * Note that SP configuration could be changed during the
497 		 * read operation, resulting in garbled output.
498 		 */
499 		if (key_getspbyid == NULL)
500 			return 0;
501 		if (!kfs->kfs_v) {
502 			struct secpolicy *sp;
503 
504 			sp = key_getspbyid(kfs->kfs_value);
505 			if (sp)
506 				kfs->kfs_v = sp;
507 			else
508 				return (ENOENT);
509 		}
510 		m = key_setdumpsp((struct secpolicy *)kfs->kfs_v,
511 		    SADB_X_SPDGET, 0, 0);
512 		if (!m)
513 			return (ENOBUFS);
514 		if (off >= m->m_pkthdr.len) {
515 			*wrlen = 0;
516 			m_freem(m);
517 			return (0);
518 		}
519 		if (len > m->m_pkthdr.len - off)
520 			len = m->m_pkthdr.len - off;
521 		m_copydata(m, off, len, *bufp);
522 		*wrlen = len;
523 		m_freem(m);
524 		return (0);
525 #endif
526 
527 	default:
528 		err = kernfs_try_xread(kfs->kfs_type, kfs, bufp, len,
529 		    EOPNOTSUPP);
530 		if (err)
531 			return err;
532 	}
533 
534 	len = strlen(*bufp);
535 	if (len <= off)
536 		*wrlen = 0;
537 	else {
538 		*bufp += off;
539 		*wrlen = len - off;
540 	}
541 	return (0);
542 }
543 
544 static int
545 kernfs_xwrite(const struct kernfs_node *kfs, char *bf, size_t len)
546 {
547 
548 	switch (kfs->kfs_type) {
549 	case KFShostname:
550 		if (bf[len-1] == '\n')
551 			--len;
552 		memcpy(hostname, bf, len);
553 		hostname[len] = '\0';
554 		hostnamelen = (size_t) len;
555 		return (0);
556 
557 	default:
558 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
559 	}
560 }
561 
562 
563 /*
564  * vp is the current namei directory
565  * ndp is the name to locate in that directory...
566  */
567 int
568 kernfs_lookup(void *v)
569 {
570 	struct vop_lookup_args /* {
571 		struct vnode * a_dvp;
572 		struct vnode ** a_vpp;
573 		struct componentname * a_cnp;
574 	} */ *ap = v;
575 	struct componentname *cnp = ap->a_cnp;
576 	struct vnode **vpp = ap->a_vpp;
577 	struct vnode *dvp = ap->a_dvp;
578 	const char *pname = cnp->cn_nameptr;
579 	const struct kernfs_node *kfs;
580 	const struct kern_target *kt;
581 	const struct dyn_kern_target *dkt;
582 	const struct kernfs_subdir *ks;
583 	int error, i;
584 #ifdef IPSEC
585 	char *ep;
586 	u_int32_t id;
587 #endif
588 
589 	*vpp = NULLVP;
590 
591 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
592 		return (EROFS);
593 
594 	if (cnp->cn_namelen == 1 && *pname == '.') {
595 		*vpp = dvp;
596 		vref(dvp);
597 		return (0);
598 	}
599 
600 	kfs = VTOKERN(dvp);
601 	switch (kfs->kfs_type) {
602 	case KFSkern:
603 		/*
604 		 * Shouldn't get here with .. in the root node.
605 		 */
606 		if (cnp->cn_flags & ISDOTDOT)
607 			return (EIO);
608 
609 		for (i = 0; i < static_nkern_targets; i++) {
610 			kt = &kern_targets[i];
611 			if (cnp->cn_namelen == kt->kt_namlen &&
612 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
613 				goto found;
614 		}
615 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
616 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
617 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
618 				kt = &dkt->dkt_kt;
619 				goto found;
620 			}
621 		}
622 		break;
623 
624 	found:
625 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
626 		return (error);
627 
628 	case KFSsubdir:
629 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
630 		if (cnp->cn_flags & ISDOTDOT) {
631 			kt = ks->ks_parent;
632 			goto found;
633 		}
634 
635 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
636 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
637 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
638 				kt = &dkt->dkt_kt;
639 				goto found;
640 			}
641 		}
642 		break;
643 
644 #ifdef IPSEC
645 	case KFSipsecsadir:
646 		if (cnp->cn_flags & ISDOTDOT) {
647 			kt = &kern_targets[0];
648 			goto found;
649 		}
650 
651 		for (i = 2; i < nipsecsa_targets; i++) {
652 			kt = &ipsecsa_targets[i];
653 			if (cnp->cn_namelen == kt->kt_namlen &&
654 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
655 				goto found;
656 		}
657 
658 		ep = NULL;
659 		id = strtoul(pname, &ep, 10);
660 		if (!ep || *ep || ep == pname)
661 			break;
662 
663 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id);
664 		return (error);
665 
666 	case KFSipsecspdir:
667 		if (cnp->cn_flags & ISDOTDOT) {
668 			kt = &kern_targets[0];
669 			goto found;
670 		}
671 
672 		for (i = 2; i < nipsecsp_targets; i++) {
673 			kt = &ipsecsp_targets[i];
674 			if (cnp->cn_namelen == kt->kt_namlen &&
675 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
676 				goto found;
677 		}
678 
679 		ep = NULL;
680 		id = strtoul(pname, &ep, 10);
681 		if (!ep || *ep || ep == pname)
682 			break;
683 
684 		error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id);
685 		return (error);
686 #endif
687 
688 	default:
689 		return (ENOTDIR);
690 	}
691 
692 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
693 }
694 
695 int
696 kernfs_open(void *v)
697 {
698 	struct vop_open_args /* {
699 		struct vnode *a_vp;
700 		int a_mode;
701 		kauth_cred_t a_cred;
702 	} */ *ap = v;
703 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
704 #ifdef IPSEC
705 	struct mbuf *m;
706 	struct secpolicy *sp;
707 #endif
708 
709 	switch (kfs->kfs_type) {
710 #ifdef IPSEC
711 	case KFSipsecsa:
712 		if (key_setdumpsa_spi == NULL)
713 			return 0;
714 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
715 		if (m) {
716 			m_freem(m);
717 			return (0);
718 		} else
719 			return (ENOENT);
720 
721 	case KFSipsecsp:
722 		if (key_getspbyid == NULL)
723 			return 0;
724 		sp = key_getspbyid(kfs->kfs_value);
725 		if (sp) {
726 			kfs->kfs_v = sp;
727 			return (0);
728 		} else
729 			return (ENOENT);
730 #endif
731 
732 	default:
733 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN,
734 		    v, 0);
735 	}
736 }
737 
738 int
739 kernfs_close(void *v)
740 {
741 	struct vop_close_args /* {
742 		struct vnode *a_vp;
743 		int a_fflag;
744 		kauth_cred_t a_cred;
745 	} */ *ap = v;
746 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
747 
748 	switch (kfs->kfs_type) {
749 #ifdef IPSEC
750 	case KFSipsecsp:
751 		if (key_freesp == NULL)
752 			return 0;
753 		key_freesp((struct secpolicy *)kfs->kfs_v);
754 		break;
755 #endif
756 
757 	default:
758 		return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE,
759 		    v, 0);
760 	}
761 
762 	return (0);
763 }
764 
765 int
766 kernfs_access(void *v)
767 {
768 	struct vop_access_args /* {
769 		struct vnode *a_vp;
770 		int a_mode;
771 		kauth_cred_t a_cred;
772 	} */ *ap = v;
773 	struct vattr va;
774 	int error;
775 
776 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
777 		return (error);
778 
779 	return kauth_authorize_vnode(ap->a_cred,
780 	    kauth_access_action(ap->a_mode, ap->a_vp->v_type, va.va_mode),
781 	    ap->a_vp, NULL, genfs_can_access(va.va_type, va.va_mode,
782 	    va.va_uid, va.va_gid, ap->a_mode, ap->a_cred));
783 }
784 
785 static int
786 kernfs_default_fileop_getattr(void *v)
787 {
788 	struct vop_getattr_args /* {
789 		struct vnode *a_vp;
790 		struct vattr *a_vap;
791 		kauth_cred_t a_cred;
792 	} */ *ap = v;
793 	struct vattr *vap = ap->a_vap;
794 
795 	vap->va_nlink = 1;
796 	vap->va_bytes = vap->va_size = 0;
797 
798 	return 0;
799 }
800 
801 int
802 kernfs_getattr(void *v)
803 {
804 	struct vop_getattr_args /* {
805 		struct vnode *a_vp;
806 		struct vattr *a_vap;
807 		kauth_cred_t a_cred;
808 	} */ *ap = v;
809 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
810 	struct kernfs_subdir *ks;
811 	struct vattr *vap = ap->a_vap;
812 	int error = 0;
813 	char strbuf[KSTRING], *bf;
814 	size_t nread, total;
815 
816 	vattr_null(vap);
817 	vap->va_type = ap->a_vp->v_type;
818 	vap->va_uid = 0;
819 	vap->va_gid = 0;
820 	vap->va_mode = kfs->kfs_mode;
821 	vap->va_fileid = kfs->kfs_fileno;
822 	vap->va_flags = 0;
823 	vap->va_size = 0;
824 	vap->va_blocksize = DEV_BSIZE;
825 	/* Make all times be current TOD, except for the "boottime" node. */
826 	if (kfs->kfs_kt->kt_namlen == 8 &&
827 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
828 		vap->va_ctime = boottime;
829 	} else {
830 		getnanotime(&vap->va_ctime);
831 	}
832 	vap->va_atime = vap->va_mtime = vap->va_ctime;
833 	vap->va_gen = 0;
834 	vap->va_flags = 0;
835 	vap->va_rdev = 0;
836 	vap->va_bytes = 0;
837 
838 	switch (kfs->kfs_type) {
839 	case KFSkern:
840 		vap->va_nlink = nkern_dirs;
841 		vap->va_bytes = vap->va_size = DEV_BSIZE;
842 		break;
843 
844 	case KFSroot:
845 		vap->va_nlink = 1;
846 		vap->va_bytes = vap->va_size = DEV_BSIZE;
847 		break;
848 
849 	case KFSsubdir:
850 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
851 		vap->va_nlink = ks->ks_dirs;
852 		vap->va_bytes = vap->va_size = DEV_BSIZE;
853 		break;
854 
855 	case KFSnull:
856 	case KFStime:
857 	case KFSint:
858 	case KFSstring:
859 	case KFShostname:
860 	case KFSavenrun:
861 	case KFSdevice:
862 	case KFSmsgbuf:
863 #ifdef IPSEC
864 	case KFSipsecsa:
865 	case KFSipsecsp:
866 #endif
867 		vap->va_nlink = 1;
868 		total = 0;
869 		do {
870 			bf = strbuf;
871 			error = kernfs_xread(kfs, total, &bf,
872 			    sizeof(strbuf), &nread);
873 			total += nread;
874 		} while (error == 0 && nread != 0);
875 		vap->va_bytes = vap->va_size = total;
876 		break;
877 
878 #ifdef IPSEC
879 	case KFSipsecsadir:
880 	case KFSipsecspdir:
881 		vap->va_nlink = 2;
882 		vap->va_bytes = vap->va_size = DEV_BSIZE;
883 		break;
884 #endif
885 
886 	default:
887 		error = kernfs_try_fileop(kfs->kfs_type,
888 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
889 		break;
890 	}
891 
892 	return (error);
893 }
894 
895 /*ARGSUSED*/
896 int
897 kernfs_setattr(void *v)
898 {
899 
900 	/*
901 	 * Silently ignore attribute changes.
902 	 * This allows for open with truncate to have no
903 	 * effect until some data is written.  I want to
904 	 * do it this way because all writes are atomic.
905 	 */
906 	return (0);
907 }
908 
909 int
910 kernfs_default_xread(void *v)
911 {
912 	struct vop_read_args /* {
913 		struct vnode *a_vp;
914 		struct uio *a_uio;
915 		int  a_ioflag;
916 		kauth_cred_t a_cred;
917 	} */ *ap = v;
918 	struct uio *uio = ap->a_uio;
919 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
920 	char strbuf[KSTRING], *bf;
921 	int off;
922 	size_t len;
923 	int error;
924 
925 	if (ap->a_vp->v_type == VDIR)
926 		return EISDIR;
927 
928 	off = (int)uio->uio_offset;
929 	/* Don't allow negative offsets */
930 	if (off < 0)
931 		return EINVAL;
932 
933 	bf = strbuf;
934 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
935 		error = uiomove(bf, len, uio);
936 	return (error);
937 }
938 
939 int
940 kernfs_read(void *v)
941 {
942 	struct vop_read_args /* {
943 		struct vnode *a_vp;
944 		struct uio *a_uio;
945 		int  a_ioflag;
946 		struct ucred *a_cred;
947 	} */ *ap = v;
948 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
949 
950 	if (kfs->kfs_type < KFSlasttype) {
951 		/* use default function */
952 		return kernfs_default_xread(v);
953 	}
954 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
955 	   EOPNOTSUPP);
956 }
957 
958 static int
959 kernfs_default_xwrite(void *v)
960 {
961 	struct vop_write_args /* {
962 		struct vnode *a_vp;
963 		struct uio *a_uio;
964 		int  a_ioflag;
965 		kauth_cred_t a_cred;
966 	} */ *ap = v;
967 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
968 	struct uio *uio = ap->a_uio;
969 	int error;
970 	size_t xlen;
971 	char strbuf[KSTRING];
972 
973 	if (uio->uio_offset != 0)
974 		return (EINVAL);
975 
976 	xlen = min(uio->uio_resid, KSTRING-1);
977 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
978 		return (error);
979 
980 	if (uio->uio_resid != 0)
981 		return (EIO);
982 
983 	strbuf[xlen] = '\0';
984 	xlen = strlen(strbuf);
985 	return (kernfs_xwrite(kfs, strbuf, xlen));
986 }
987 
988 int
989 kernfs_write(void *v)
990 {
991 	struct vop_write_args /* {
992 		struct vnode *a_vp;
993 		struct uio *a_uio;
994 		int  a_ioflag;
995 		kauth_cred_t a_cred;
996 	} */ *ap = v;
997 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
998 
999 	if (kfs->kfs_type < KFSlasttype) {
1000 		/* use default function */
1001 		return kernfs_default_xwrite(v);
1002 	}
1003 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
1004 	    EOPNOTSUPP);
1005 }
1006 
1007 int
1008 kernfs_ioctl(void *v)
1009 {
1010 	struct vop_ioctl_args /* {
1011 		const struct vnodeop_desc *a_desc;
1012 		struct vnode *a_vp;
1013 		u_long a_command;
1014 		void *a_data;
1015 		int a_fflag;
1016 		kauth_cred_t a_cred;
1017 	} */ *ap = v;
1018 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1019 
1020 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
1021 	    EPASSTHROUGH);
1022 }
1023 
1024 static int
1025 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
1026     u_int32_t value, struct vop_readdir_args *ap)
1027 {
1028 	struct kernfs_node *kfs;
1029 	struct vnode *vp;
1030 	int error;
1031 
1032 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
1033 	    value)) != 0)
1034 		return error;
1035 	if (kt->kt_tag == KFSdevice) {
1036 		struct vattr va;
1037 
1038 		error = VOP_GETATTR(vp, &va, ap->a_cred);
1039 		if (error != 0) {
1040 			return error;
1041 		}
1042 		d->d_fileno = va.va_fileid;
1043 	} else {
1044 		kfs = VTOKERN(vp);
1045 		d->d_fileno = kfs->kfs_fileno;
1046 	}
1047 	vput(vp);
1048 	return 0;
1049 }
1050 
1051 static int
1052 kernfs_setdirentfileno(struct dirent *d, off_t entry,
1053     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
1054     const struct kern_target *kt, struct vop_readdir_args *ap)
1055 {
1056 	const struct kern_target *ikt;
1057 	int error;
1058 
1059 	switch (entry) {
1060 	case 0:
1061 		d->d_fileno = thisdir_kfs->kfs_fileno;
1062 		return 0;
1063 	case 1:
1064 		ikt = parent_kt;
1065 		break;
1066 	default:
1067 		ikt = kt;
1068 		break;
1069 	}
1070 	if (ikt != thisdir_kfs->kfs_kt) {
1071 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
1072 			return error;
1073 	} else
1074 		d->d_fileno = thisdir_kfs->kfs_fileno;
1075 	return 0;
1076 }
1077 
1078 int
1079 kernfs_readdir(void *v)
1080 {
1081 	struct vop_readdir_args /* {
1082 		struct vnode *a_vp;
1083 		struct uio *a_uio;
1084 		kauth_cred_t a_cred;
1085 		int *a_eofflag;
1086 		off_t **a_cookies;
1087 		int a_*ncookies;
1088 	} */ *ap = v;
1089 	struct uio *uio = ap->a_uio;
1090 	struct dirent d;
1091 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1092 	const struct kern_target *kt;
1093 	const struct dyn_kern_target *dkt = NULL;
1094 	const struct kernfs_subdir *ks;
1095 	off_t i, j;
1096 	int error;
1097 	off_t *cookies = NULL;
1098 	int ncookies = 0, n;
1099 #ifdef IPSEC
1100 	struct secasvar *sav, *sav2;
1101 	struct secpolicy *sp;
1102 #endif
1103 
1104 	if (uio->uio_resid < UIO_MX)
1105 		return (EINVAL);
1106 	if (uio->uio_offset < 0)
1107 		return (EINVAL);
1108 
1109 	error = 0;
1110 	i = uio->uio_offset;
1111 	memset(&d, 0, sizeof(d));
1112 	d.d_reclen = UIO_MX;
1113 	ncookies = uio->uio_resid / UIO_MX;
1114 
1115 	switch (kfs->kfs_type) {
1116 	case KFSkern:
1117 		if (i >= nkern_targets)
1118 			return (0);
1119 
1120 		if (ap->a_ncookies) {
1121 			ncookies = min(ncookies, (nkern_targets - i));
1122 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1123 			    M_WAITOK);
1124 			*ap->a_cookies = cookies;
1125 		}
1126 
1127 		n = 0;
1128 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
1129 			if (i < static_nkern_targets)
1130 				kt = &kern_targets[i];
1131 			else {
1132 				if (dkt == NULL) {
1133 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
1134 					for (j = static_nkern_targets; j < i &&
1135 						     dkt != NULL; j++)
1136 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1137 					if (j != i)
1138 						break;
1139 				} else {
1140 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1141 				}
1142 				if (dkt == NULL)
1143 					break;
1144 				kt = &dkt->dkt_kt;
1145 			}
1146 			if (kt->kt_tag == KFSdevice) {
1147 				dev_t *dp = kt->kt_data;
1148 				struct vnode *fvp;
1149 
1150 				if (*dp == NODEV ||
1151 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1152 					continue;
1153 				vrele(fvp);
1154 			}
1155 			if (kt->kt_tag == KFSmsgbuf) {
1156 				if (!msgbufenabled
1157 				    || msgbufp->msg_magic != MSG_MAGIC) {
1158 					continue;
1159 				}
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 				vrele(fvp);
1237 			}
1238 			d.d_namlen = kt->kt_namlen;
1239 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1240 			    ks->ks_parent, kt, ap)) != 0)
1241 				break;
1242 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1243 			d.d_type = kt->kt_type;
1244 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1245 				break;
1246 			if (cookies)
1247 				*cookies++ = i + 1;
1248 			n++;
1249 		}
1250 		ncookies = n;
1251 		break;
1252 
1253 #ifdef IPSEC
1254 	case KFSipsecsadir:
1255 		/* count SA in the system */
1256 		n = 0;
1257 		if (&satailq == NULL)
1258 			return 0;
1259 		TAILQ_FOREACH(sav, &satailq, tailq) {
1260 			for (sav2 = TAILQ_FIRST(&satailq);
1261 			    sav2 != sav;
1262 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1263 				if (sav->spi == sav2->spi) {
1264 					/* multiple SA with same SPI */
1265 					break;
1266 				}
1267 			}
1268 			if (sav == sav2 || sav->spi != sav2->spi)
1269 				n++;
1270 		}
1271 
1272 		if (i >= nipsecsa_targets + n)
1273 			return (0);
1274 
1275 		if (ap->a_ncookies) {
1276 			ncookies = min(ncookies, (n - i));
1277 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1278 			    M_WAITOK);
1279 			*ap->a_cookies = cookies;
1280 		}
1281 
1282 		n = 0;
1283 		for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) {
1284 			kt = &ipsecsa_targets[i];
1285 			d.d_namlen = kt->kt_namlen;
1286 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1287 			    &kern_targets[0], kt, ap)) != 0)
1288 				break;
1289 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1290 			d.d_type = kt->kt_type;
1291 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1292 				break;
1293 			if (cookies)
1294 				*cookies++ = i + 1;
1295 			n++;
1296 		}
1297 		if (error) {
1298 			ncookies = n;
1299 			break;
1300 		}
1301 
1302 		TAILQ_FOREACH(sav, &satailq, tailq) {
1303 			for (sav2 = TAILQ_FIRST(&satailq);
1304 			    sav2 != sav;
1305 			    sav2 = TAILQ_NEXT(sav2, tailq)) {
1306 				if (sav->spi == sav2->spi) {
1307 					/* multiple SA with same SPI */
1308 					break;
1309 				}
1310 			}
1311 			if (sav != sav2 && sav->spi == sav2->spi)
1312 				continue;
1313 			if (uio->uio_resid < UIO_MX)
1314 				break;
1315 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt,
1316 			    sav->spi, ap)) != 0)
1317 				break;
1318 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1319 			    "%u", ntohl(sav->spi));
1320 			d.d_type = DT_REG;
1321 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1322 				break;
1323 			if (cookies)
1324 				*cookies++ = i + 1;
1325 			n++;
1326 			i++;
1327 		}
1328 		ncookies = n;
1329 		break;
1330 
1331 	case KFSipsecspdir:
1332 		/* count SP in the system */
1333 		if (&sptailq == NULL)
1334 			return 0;
1335 
1336 		n = 0;
1337 		TAILQ_FOREACH(sp, &sptailq, tailq)
1338 			n++;
1339 
1340 		if (i >= nipsecsp_targets + n)
1341 			return (0);
1342 
1343 		if (ap->a_ncookies) {
1344 			ncookies = min(ncookies, (n - i));
1345 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1346 			    M_WAITOK);
1347 			*ap->a_cookies = cookies;
1348 		}
1349 
1350 		n = 0;
1351 		for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) {
1352 			kt = &ipsecsp_targets[i];
1353 			d.d_namlen = kt->kt_namlen;
1354 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1355 			    &kern_targets[0], kt, ap)) != 0)
1356 				break;
1357 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1358 			d.d_type = kt->kt_type;
1359 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1360 				break;
1361 			if (cookies)
1362 				*cookies++ = i + 1;
1363 			n++;
1364 		}
1365 		if (error) {
1366 			ncookies = n;
1367 			break;
1368 		}
1369 
1370 		TAILQ_FOREACH(sp, &sptailq, tailq) {
1371 			if (uio->uio_resid < UIO_MX)
1372 				break;
1373 			if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt,
1374 			    sp->id, ap)) != 0)
1375 				break;
1376 			d.d_namlen = snprintf(d.d_name, sizeof(d.d_name),
1377 			    "%u", sp->id);
1378 			d.d_type = DT_REG;
1379 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1380 				break;
1381 			if (cookies)
1382 				*cookies++ = i + 1;
1383 			n++;
1384 			i++;
1385 		}
1386 		ncookies = n;
1387 		break;
1388 #endif
1389 
1390 	default:
1391 		error = ENOTDIR;
1392 		break;
1393 	}
1394 
1395 	if (ap->a_ncookies) {
1396 		if (error) {
1397 			if (cookies)
1398 				free(*ap->a_cookies, M_TEMP);
1399 			*ap->a_ncookies = 0;
1400 			*ap->a_cookies = NULL;
1401 		} else
1402 			*ap->a_ncookies = ncookies;
1403 	}
1404 
1405 	uio->uio_offset = i;
1406 	return (error);
1407 }
1408 
1409 int
1410 kernfs_inactive(void *v)
1411 {
1412 	struct vop_inactive_args /* {
1413 		struct vnode *a_vp;
1414 		bool *a_recycle;
1415 	} */ *ap = v;
1416 	struct vnode *vp = ap->a_vp;
1417 	const struct kernfs_node *kfs = VTOKERN(ap->a_vp);
1418 #ifdef IPSEC
1419 	struct mbuf *m;
1420 	struct secpolicy *sp;
1421 #endif
1422 
1423 	*ap->a_recycle = false;
1424 	switch (kfs->kfs_type) {
1425 #ifdef IPSEC
1426 	case KFSipsecsa:
1427 		if (key_setdumpsa_spi == NULL)
1428 			return 0;
1429 		m = key_setdumpsa_spi(htonl(kfs->kfs_value));
1430 		if (m)
1431 			m_freem(m);
1432 		else
1433 			*ap->a_recycle = true;
1434 		break;
1435 	case KFSipsecsp:
1436 		if (key_getspbyid == NULL)
1437 			return 0;
1438 		sp = key_getspbyid(kfs->kfs_value);
1439 		if (sp)
1440 			key_freesp(sp);
1441 		else {
1442 			*ap->a_recycle = true;
1443 		}
1444 		break;
1445 #endif
1446 	default:
1447 		break;
1448 	}
1449 	VOP_UNLOCK(vp);
1450 	return (0);
1451 }
1452 
1453 int
1454 kernfs_reclaim(void *v)
1455 {
1456 	struct vop_reclaim_args /* {
1457 		struct vnode *a_vp;
1458 	} */ *ap = v;
1459 
1460 	return (kernfs_freevp(ap->a_vp));
1461 }
1462 
1463 /*
1464  * Return POSIX pathconf information applicable to special devices.
1465  */
1466 int
1467 kernfs_pathconf(void *v)
1468 {
1469 	struct vop_pathconf_args /* {
1470 		struct vnode *a_vp;
1471 		int a_name;
1472 		register_t *a_retval;
1473 	} */ *ap = v;
1474 
1475 	switch (ap->a_name) {
1476 	case _PC_LINK_MAX:
1477 		*ap->a_retval = LINK_MAX;
1478 		return (0);
1479 	case _PC_MAX_CANON:
1480 		*ap->a_retval = MAX_CANON;
1481 		return (0);
1482 	case _PC_MAX_INPUT:
1483 		*ap->a_retval = MAX_INPUT;
1484 		return (0);
1485 	case _PC_PIPE_BUF:
1486 		*ap->a_retval = PIPE_BUF;
1487 		return (0);
1488 	case _PC_CHOWN_RESTRICTED:
1489 		*ap->a_retval = 1;
1490 		return (0);
1491 	case _PC_VDISABLE:
1492 		*ap->a_retval = _POSIX_VDISABLE;
1493 		return (0);
1494 	case _PC_SYNC_IO:
1495 		*ap->a_retval = 1;
1496 		return (0);
1497 	default:
1498 		return (EINVAL);
1499 	}
1500 	/* NOTREACHED */
1501 }
1502 
1503 /*
1504  * Print out the contents of a /dev/fd vnode.
1505  */
1506 /* ARGSUSED */
1507 int
1508 kernfs_print(void *v)
1509 {
1510 
1511 	printf("tag VT_KERNFS, kernfs vnode\n");
1512 	return (0);
1513 }
1514 
1515 int
1516 kernfs_link(void *v)
1517 {
1518 	struct vop_link_args /* {
1519 		struct vnode *a_dvp;
1520 		struct vnode *a_vp;
1521 		struct componentname *a_cnp;
1522 	} */ *ap = v;
1523 
1524 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1525 	vput(ap->a_dvp);
1526 	return (EROFS);
1527 }
1528 
1529 int
1530 kernfs_symlink(void *v)
1531 {
1532 	struct vop_symlink_args /* {
1533 		struct vnode *a_dvp;
1534 		struct vnode **a_vpp;
1535 		struct componentname *a_cnp;
1536 		struct vattr *a_vap;
1537 		char *a_target;
1538 	} */ *ap = v;
1539 
1540 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1541 	vput(ap->a_dvp);
1542 	return (EROFS);
1543 }
1544