xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision a536ee5124e62c9a0051a252f7833dc8f50f44c9)
1 /*	$NetBSD: kernfs_vnops.c,v 1.146 2012/03/22 20:34:38 drochner 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 1.146 2012/03/22 20:34:38 drochner Exp $");
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/vmmeter.h>
48 #include <sys/time.h>
49 #include <sys/proc.h>
50 #include <sys/vnode.h>
51 #include <sys/malloc.h>
52 #include <sys/file.h>
53 #include <sys/stat.h>
54 #include <sys/mount.h>
55 #include <sys/namei.h>
56 #include <sys/buf.h>
57 #include <sys/dirent.h>
58 #include <sys/msgbuf.h>
59 
60 #include <miscfs/genfs/genfs.h>
61 #include <miscfs/kernfs/kernfs.h>
62 
63 #include <uvm/uvm_extern.h>
64 
65 #define KSTRING	256		/* Largest I/O available via this filesystem */
66 #define	UIO_MX 32
67 
68 #define	READ_MODE	(S_IRUSR|S_IRGRP|S_IROTH)
69 #define	WRITE_MODE	(S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH)
70 #define	UREAD_MODE	(S_IRUSR)
71 #define	DIR_MODE	(S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
72 #define	UDIR_MODE	(S_IRUSR|S_IXUSR)
73 
74 #define N(s) sizeof(s)-1, s
75 const struct kern_target kern_targets[] = {
76 /* NOTE: The name must be less than UIO_MX-16 chars in length */
77      /*        name            data          tag           type  ro/rw */
78      { DT_DIR, N("."),         0,            KFSkern,        VDIR, DIR_MODE   },
79      { DT_DIR, N(".."),        0,            KFSroot,        VDIR, DIR_MODE   },
80      { DT_REG, N("boottime"),  &boottime.tv_sec, KFSint,     VREG, READ_MODE  },
81 			/* XXXUNCONST */
82      { DT_REG, N("copyright"), __UNCONST(copyright),
83      					     KFSstring,      VREG, READ_MODE  },
84      { DT_REG, N("hostname"),  0,            KFShostname,    VREG, WRITE_MODE },
85      { DT_REG, N("hz"),        &hz,          KFSint,         VREG, READ_MODE  },
86      { DT_REG, N("loadavg"),   0,            KFSavenrun,     VREG, READ_MODE  },
87      { DT_REG, N("msgbuf"),    0,	     KFSmsgbuf,      VREG, READ_MODE  },
88      { DT_REG, N("pagesize"),  &uvmexp.pagesize, KFSint,     VREG, READ_MODE  },
89      { DT_REG, N("physmem"),   &physmem,     KFSint,         VREG, READ_MODE  },
90 #if 0
91      { DT_DIR, N("root"),      0,            KFSnull,        VDIR, DIR_MODE   },
92 #endif
93      { DT_BLK, N("rootdev"),   &rootdev,     KFSdevice,      VBLK, READ_MODE  },
94      { DT_CHR, N("rrootdev"),  &rrootdev,    KFSdevice,      VCHR, READ_MODE  },
95      { DT_REG, N("time"),      0,            KFStime,        VREG, READ_MODE  },
96 			/* XXXUNCONST */
97      { DT_REG, N("version"),   __UNCONST(version),
98      					     KFSstring,      VREG, READ_MODE  },
99 };
100 const struct kern_target subdir_targets[] = {
101 /* NOTE: The name must be less than UIO_MX-16 chars in length */
102      /*        name            data          tag           type  ro/rw */
103      { DT_DIR, N("."),         0,            KFSsubdir,      VDIR, DIR_MODE   },
104      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
105 };
106 #undef N
107 SIMPLEQ_HEAD(,dyn_kern_target) dyn_kern_targets =
108 	SIMPLEQ_HEAD_INITIALIZER(dyn_kern_targets);
109 int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
110 const int static_nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
111 int nkern_dirs = 2;
112 
113 int kernfs_try_fileop(kfstype, kfsfileop, void *, int);
114 int kernfs_try_xread(kfstype, const struct kernfs_node *, char **,
115     size_t, int);
116 int kernfs_try_xwrite(kfstype, const struct kernfs_node *, char *,
117     size_t, int);
118 
119 static int kernfs_default_xread(void *v);
120 static int kernfs_default_xwrite(void *v);
121 static int kernfs_default_fileop_getattr(void *);
122 
123 /* must include all fileop's */
124 const struct kernfs_fileop kernfs_default_fileops[] = {
125   { .kf_fileop = KERNFS_XREAD },
126   { .kf_fileop = KERNFS_XWRITE },
127   { .kf_fileop = KERNFS_FILEOP_OPEN },
128   { .kf_fileop = KERNFS_FILEOP_GETATTR,
129     .kf_vop = kernfs_default_fileop_getattr },
130   { .kf_fileop = KERNFS_FILEOP_IOCTL },
131   { .kf_fileop = KERNFS_FILEOP_CLOSE },
132   { .kf_fileop = KERNFS_FILEOP_READ,
133     .kf_vop = kernfs_default_xread },
134   { .kf_fileop = KERNFS_FILEOP_WRITE,
135     .kf_vop = kernfs_default_xwrite },
136 };
137 
138 int	kernfs_lookup(void *);
139 #define	kernfs_create	genfs_eopnotsupp
140 #define	kernfs_mknod	genfs_eopnotsupp
141 int	kernfs_open(void *);
142 int	kernfs_close(void *);
143 int	kernfs_access(void *);
144 int	kernfs_getattr(void *);
145 int	kernfs_setattr(void *);
146 int	kernfs_read(void *);
147 int	kernfs_write(void *);
148 #define	kernfs_fcntl	genfs_fcntl
149 int	kernfs_ioctl(void *);
150 #define	kernfs_poll	genfs_poll
151 #define kernfs_revoke	genfs_revoke
152 #define	kernfs_fsync	genfs_nullop
153 #define	kernfs_seek	genfs_nullop
154 #define	kernfs_remove	genfs_eopnotsupp
155 int	kernfs_link(void *);
156 #define	kernfs_rename	genfs_eopnotsupp
157 #define	kernfs_mkdir	genfs_eopnotsupp
158 #define	kernfs_rmdir	genfs_eopnotsupp
159 int	kernfs_symlink(void *);
160 int	kernfs_readdir(void *);
161 #define	kernfs_readlink	genfs_eopnotsupp
162 #define	kernfs_abortop	genfs_abortop
163 int	kernfs_inactive(void *);
164 int	kernfs_reclaim(void *);
165 #define	kernfs_lock	genfs_lock
166 #define	kernfs_unlock	genfs_unlock
167 #define	kernfs_bmap	genfs_badop
168 #define	kernfs_strategy	genfs_badop
169 int	kernfs_print(void *);
170 #define	kernfs_islocked	genfs_islocked
171 int	kernfs_pathconf(void *);
172 #define	kernfs_advlock	genfs_einval
173 #define	kernfs_bwrite	genfs_eopnotsupp
174 #define	kernfs_putpages	genfs_putpages
175 
176 static int	kernfs_xread(struct kernfs_node *, int, char **,
177 				size_t, size_t *);
178 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
179 
180 int (**kernfs_vnodeop_p)(void *);
181 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
182 	{ &vop_default_desc, vn_default_error },
183 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
184 	{ &vop_create_desc, kernfs_create },		/* create */
185 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
186 	{ &vop_open_desc, kernfs_open },		/* open */
187 	{ &vop_close_desc, kernfs_close },		/* close */
188 	{ &vop_access_desc, kernfs_access },		/* access */
189 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
190 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
191 	{ &vop_read_desc, kernfs_read },		/* read */
192 	{ &vop_write_desc, kernfs_write },		/* write */
193 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
194 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
195 	{ &vop_poll_desc, kernfs_poll },		/* poll */
196 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
197 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
198 	{ &vop_seek_desc, kernfs_seek },		/* seek */
199 	{ &vop_remove_desc, kernfs_remove },		/* remove */
200 	{ &vop_link_desc, kernfs_link },		/* link */
201 	{ &vop_rename_desc, kernfs_rename },		/* rename */
202 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
203 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
204 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
205 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
206 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
207 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
208 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
209 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
210 	{ &vop_lock_desc, kernfs_lock },		/* lock */
211 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
212 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
213 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
214 	{ &vop_print_desc, kernfs_print },		/* print */
215 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
216 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
217 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
218 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
219 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
220 	{ NULL, NULL }
221 };
222 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
223 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
224 
225 static inline int
226 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
227 {
228 	if (a->kf_type < b->kf_type)
229 		return -1;
230 	if (a->kf_type > b->kf_type)
231 		return 1;
232 	if (a->kf_fileop < b->kf_fileop)
233 		return -1;
234 	if (a->kf_fileop > b->kf_fileop)
235 		return 1;
236 	return (0);
237 }
238 
239 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
240 	SPLAY_INITIALIZER(kfsfileoptree);
241 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
242 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
243 
244 kfstype
245 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
246 {
247 	static u_char nextfreetype = KFSlasttype;
248 	struct kernfs_fileop *dkf, *fkf, skf;
249 	int i;
250 
251 	/* XXX need to keep track of dkf's memory if we support
252            deallocating types */
253 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
254 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
255 
256 	for (i = 0; i < sizeof(kernfs_default_fileops) /
257 		     sizeof(kernfs_default_fileops[0]); i++) {
258 		dkf[i].kf_type = nextfreetype;
259 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
260 	}
261 
262 	for (i = 0; i < nkf; i++) {
263 		skf.kf_type = nextfreetype;
264 		skf.kf_fileop = kf[i].kf_fileop;
265 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
266 			fkf->kf_vop = kf[i].kf_vop;
267 	}
268 
269 	return nextfreetype++;
270 }
271 
272 int
273 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
274 {
275 	struct kernfs_fileop *kf, skf;
276 
277 	skf.kf_type = type;
278 	skf.kf_fileop = fileop;
279 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
280 		if (kf->kf_vop)
281 			return kf->kf_vop(v);
282 	return error;
283 }
284 
285 int
286 kernfs_try_xread(kfstype type, const struct kernfs_node *kfs, char **bfp,
287     size_t len, int error)
288 {
289 	struct kernfs_fileop *kf, skf;
290 
291 	skf.kf_type = type;
292 	skf.kf_fileop = KERNFS_XREAD;
293 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
294 		if (kf->kf_xread)
295 			return kf->kf_xread(kfs, bfp, len);
296 	return error;
297 }
298 
299 int
300 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
301     size_t len, int error)
302 {
303 	struct kernfs_fileop *kf, skf;
304 
305 	skf.kf_type = type;
306 	skf.kf_fileop = KERNFS_XWRITE;
307 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
308 		if (kf->kf_xwrite)
309 			return kf->kf_xwrite(kfs, bf, len);
310 	return error;
311 }
312 
313 int
314 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
315 {
316 	struct kernfs_subdir *ks, *parent;
317 
318 	if (pkt == NULL) {
319 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
320 		nkern_targets++;
321 		if (dkt->dkt_kt.kt_vtype == VDIR)
322 			nkern_dirs++;
323 	} else {
324 		parent = (struct kernfs_subdir *)pkt->kt_data;
325 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
326 		parent->ks_nentries++;
327 		if (dkt->dkt_kt.kt_vtype == VDIR)
328 			parent->ks_dirs++;
329 	}
330 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
331 		ks = malloc(sizeof(struct kernfs_subdir),
332 		    M_TEMP, M_WAITOK);
333 		SIMPLEQ_INIT(&ks->ks_entries);
334 		ks->ks_nentries = 2; /* . and .. */
335 		ks->ks_dirs = 2;
336 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
337 		dkt->dkt_kt.kt_data = ks;
338 	}
339 	return 0;
340 }
341 
342 static int
343 kernfs_xread(struct kernfs_node *kfs, int off, char **bufp, size_t len, size_t *wrlen)
344 {
345 	const struct kern_target *kt;
346 	int err;
347 
348 	kt = kfs->kfs_kt;
349 
350 	switch (kfs->kfs_type) {
351 	case KFStime: {
352 		struct timeval tv;
353 
354 		microtime(&tv);
355 		snprintf(*bufp, len, "%lld %ld\n", (long long)tv.tv_sec,
356 		    (long)tv.tv_usec);
357 		break;
358 	}
359 
360 	case KFSint: {
361 		int *ip = kt->kt_data;
362 
363 		snprintf(*bufp, len, "%d\n", *ip);
364 		break;
365 	}
366 
367 	case KFSstring: {
368 		char *cp = kt->kt_data;
369 
370 		*bufp = cp;
371 		break;
372 	}
373 
374 	case KFSmsgbuf: {
375 		long n;
376 
377 		/*
378 		 * deal with cases where the message buffer has
379 		 * become corrupted.
380 		 */
381 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
382 			msgbufenabled = 0;
383 			return (ENXIO);
384 		}
385 
386 		/*
387 		 * Note that reads of /kern/msgbuf won't necessarily yield
388 		 * consistent results, if the message buffer is modified
389 		 * while the read is in progress.  The worst that can happen
390 		 * is that incorrect data will be read.  There's no way
391 		 * that this can crash the system unless the values in the
392 		 * message buffer header are corrupted, but that'll cause
393 		 * the system to die anyway.
394 		 */
395 		if (off >= msgbufp->msg_bufs) {
396 			*wrlen = 0;
397 			return (0);
398 		}
399 		n = msgbufp->msg_bufx + off;
400 		if (n >= msgbufp->msg_bufs)
401 			n -= msgbufp->msg_bufs;
402 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
403 		*bufp = msgbufp->msg_bufc + n;
404 		*wrlen = len;
405 		return (0);
406 	}
407 
408 	case KFShostname: {
409 		char *cp = hostname;
410 		size_t xlen = hostnamelen;
411 
412 		if (xlen >= (len - 2))
413 			return (EINVAL);
414 
415 		memcpy(*bufp, cp, xlen);
416 		(*bufp)[xlen] = '\n';
417 		(*bufp)[xlen+1] = '\0';
418 		break;
419 	}
420 
421 	case KFSavenrun:
422 		averunnable.fscale = FSCALE;
423 		snprintf(*bufp, len, "%d %d %d %ld\n",
424 		    averunnable.ldavg[0], averunnable.ldavg[1],
425 		    averunnable.ldavg[2], averunnable.fscale);
426 		break;
427 
428 	default:
429 		err = kernfs_try_xread(kfs->kfs_type, kfs, bufp, len,
430 		    EOPNOTSUPP);
431 		if (err)
432 			return err;
433 	}
434 
435 	len = strlen(*bufp);
436 	if (len <= off)
437 		*wrlen = 0;
438 	else {
439 		*bufp += off;
440 		*wrlen = len - off;
441 	}
442 	return (0);
443 }
444 
445 static int
446 kernfs_xwrite(const struct kernfs_node *kfs, char *bf, size_t len)
447 {
448 
449 	switch (kfs->kfs_type) {
450 	case KFShostname:
451 		if (bf[len-1] == '\n')
452 			--len;
453 		memcpy(hostname, bf, len);
454 		hostname[len] = '\0';
455 		hostnamelen = (size_t) len;
456 		return (0);
457 
458 	default:
459 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
460 	}
461 }
462 
463 
464 /*
465  * vp is the current namei directory
466  * ndp is the name to locate in that directory...
467  */
468 int
469 kernfs_lookup(void *v)
470 {
471 	struct vop_lookup_args /* {
472 		struct vnode * a_dvp;
473 		struct vnode ** a_vpp;
474 		struct componentname * a_cnp;
475 	} */ *ap = v;
476 	struct componentname *cnp = ap->a_cnp;
477 	struct vnode **vpp = ap->a_vpp;
478 	struct vnode *dvp = ap->a_dvp;
479 	const char *pname = cnp->cn_nameptr;
480 	const struct kernfs_node *kfs;
481 	const struct kern_target *kt;
482 	const struct dyn_kern_target *dkt;
483 	const struct kernfs_subdir *ks;
484 	int error, i;
485 
486 	*vpp = NULLVP;
487 
488 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
489 		return (EROFS);
490 
491 	if (cnp->cn_namelen == 1 && *pname == '.') {
492 		*vpp = dvp;
493 		vref(dvp);
494 		return (0);
495 	}
496 
497 	kfs = VTOKERN(dvp);
498 	switch (kfs->kfs_type) {
499 	case KFSkern:
500 		/*
501 		 * Shouldn't get here with .. in the root node.
502 		 */
503 		if (cnp->cn_flags & ISDOTDOT)
504 			return (EIO);
505 
506 		for (i = 0; i < static_nkern_targets; i++) {
507 			kt = &kern_targets[i];
508 			if (cnp->cn_namelen == kt->kt_namlen &&
509 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
510 				goto found;
511 		}
512 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
513 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
514 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
515 				kt = &dkt->dkt_kt;
516 				goto found;
517 			}
518 		}
519 		break;
520 
521 	found:
522 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
523 		return (error);
524 
525 	case KFSsubdir:
526 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
527 		if (cnp->cn_flags & ISDOTDOT) {
528 			kt = ks->ks_parent;
529 			goto found;
530 		}
531 
532 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
533 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
534 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
535 				kt = &dkt->dkt_kt;
536 				goto found;
537 			}
538 		}
539 		break;
540 
541 	default:
542 		return (ENOTDIR);
543 	}
544 
545 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
546 }
547 
548 int
549 kernfs_open(void *v)
550 {
551 	struct vop_open_args /* {
552 		struct vnode *a_vp;
553 		int a_mode;
554 		kauth_cred_t a_cred;
555 	} */ *ap = v;
556 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
557 
558 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN, v, 0);
559 }
560 
561 int
562 kernfs_close(void *v)
563 {
564 	struct vop_close_args /* {
565 		struct vnode *a_vp;
566 		int a_fflag;
567 		kauth_cred_t a_cred;
568 	} */ *ap = v;
569 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
570 
571 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE, v, 0);
572 }
573 
574 int
575 kernfs_access(void *v)
576 {
577 	struct vop_access_args /* {
578 		struct vnode *a_vp;
579 		int a_mode;
580 		kauth_cred_t a_cred;
581 	} */ *ap = v;
582 	struct vattr va;
583 	int error;
584 
585 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
586 		return (error);
587 
588 	return kauth_authorize_vnode(ap->a_cred,
589 	    kauth_access_action(ap->a_mode, ap->a_vp->v_type, va.va_mode),
590 	    ap->a_vp, NULL, genfs_can_access(va.va_type, va.va_mode,
591 	    va.va_uid, va.va_gid, ap->a_mode, ap->a_cred));
592 }
593 
594 static int
595 kernfs_default_fileop_getattr(void *v)
596 {
597 	struct vop_getattr_args /* {
598 		struct vnode *a_vp;
599 		struct vattr *a_vap;
600 		kauth_cred_t a_cred;
601 	} */ *ap = v;
602 	struct vattr *vap = ap->a_vap;
603 
604 	vap->va_nlink = 1;
605 	vap->va_bytes = vap->va_size = 0;
606 
607 	return 0;
608 }
609 
610 int
611 kernfs_getattr(void *v)
612 {
613 	struct vop_getattr_args /* {
614 		struct vnode *a_vp;
615 		struct vattr *a_vap;
616 		kauth_cred_t a_cred;
617 	} */ *ap = v;
618 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
619 	struct kernfs_subdir *ks;
620 	struct vattr *vap = ap->a_vap;
621 	int error = 0;
622 	char strbuf[KSTRING], *bf;
623 	size_t nread, total;
624 
625 	vattr_null(vap);
626 	vap->va_type = ap->a_vp->v_type;
627 	vap->va_uid = 0;
628 	vap->va_gid = 0;
629 	vap->va_mode = kfs->kfs_mode;
630 	vap->va_fileid = kfs->kfs_fileno;
631 	vap->va_flags = 0;
632 	vap->va_size = 0;
633 	vap->va_blocksize = DEV_BSIZE;
634 	/* Make all times be current TOD, except for the "boottime" node. */
635 	if (kfs->kfs_kt->kt_namlen == 8 &&
636 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
637 		vap->va_ctime = boottime;
638 	} else {
639 		getnanotime(&vap->va_ctime);
640 	}
641 	vap->va_atime = vap->va_mtime = vap->va_ctime;
642 	vap->va_gen = 0;
643 	vap->va_flags = 0;
644 	vap->va_rdev = 0;
645 	vap->va_bytes = 0;
646 
647 	switch (kfs->kfs_type) {
648 	case KFSkern:
649 		vap->va_nlink = nkern_dirs;
650 		vap->va_bytes = vap->va_size = DEV_BSIZE;
651 		break;
652 
653 	case KFSroot:
654 		vap->va_nlink = 1;
655 		vap->va_bytes = vap->va_size = DEV_BSIZE;
656 		break;
657 
658 	case KFSsubdir:
659 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
660 		vap->va_nlink = ks->ks_dirs;
661 		vap->va_bytes = vap->va_size = DEV_BSIZE;
662 		break;
663 
664 	case KFSnull:
665 	case KFStime:
666 	case KFSint:
667 	case KFSstring:
668 	case KFShostname:
669 	case KFSavenrun:
670 	case KFSdevice:
671 	case KFSmsgbuf:
672 		vap->va_nlink = 1;
673 		total = 0;
674 		do {
675 			bf = strbuf;
676 			error = kernfs_xread(kfs, total, &bf,
677 			    sizeof(strbuf), &nread);
678 			total += nread;
679 		} while (error == 0 && nread != 0);
680 		vap->va_bytes = vap->va_size = total;
681 		break;
682 
683 	default:
684 		error = kernfs_try_fileop(kfs->kfs_type,
685 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
686 		break;
687 	}
688 
689 	return (error);
690 }
691 
692 /*ARGSUSED*/
693 int
694 kernfs_setattr(void *v)
695 {
696 
697 	/*
698 	 * Silently ignore attribute changes.
699 	 * This allows for open with truncate to have no
700 	 * effect until some data is written.  I want to
701 	 * do it this way because all writes are atomic.
702 	 */
703 	return (0);
704 }
705 
706 int
707 kernfs_default_xread(void *v)
708 {
709 	struct vop_read_args /* {
710 		struct vnode *a_vp;
711 		struct uio *a_uio;
712 		int  a_ioflag;
713 		kauth_cred_t a_cred;
714 	} */ *ap = v;
715 	struct uio *uio = ap->a_uio;
716 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
717 	char strbuf[KSTRING], *bf;
718 	int off;
719 	size_t len;
720 	int error;
721 
722 	if (ap->a_vp->v_type == VDIR)
723 		return EISDIR;
724 
725 	off = (int)uio->uio_offset;
726 	/* Don't allow negative offsets */
727 	if (off < 0)
728 		return EINVAL;
729 
730 	bf = strbuf;
731 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
732 		error = uiomove(bf, len, uio);
733 	return (error);
734 }
735 
736 int
737 kernfs_read(void *v)
738 {
739 	struct vop_read_args /* {
740 		struct vnode *a_vp;
741 		struct uio *a_uio;
742 		int  a_ioflag;
743 		struct ucred *a_cred;
744 	} */ *ap = v;
745 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
746 
747 	if (kfs->kfs_type < KFSlasttype) {
748 		/* use default function */
749 		return kernfs_default_xread(v);
750 	}
751 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
752 	   EOPNOTSUPP);
753 }
754 
755 static int
756 kernfs_default_xwrite(void *v)
757 {
758 	struct vop_write_args /* {
759 		struct vnode *a_vp;
760 		struct uio *a_uio;
761 		int  a_ioflag;
762 		kauth_cred_t a_cred;
763 	} */ *ap = v;
764 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
765 	struct uio *uio = ap->a_uio;
766 	int error;
767 	size_t xlen;
768 	char strbuf[KSTRING];
769 
770 	if (uio->uio_offset != 0)
771 		return (EINVAL);
772 
773 	xlen = min(uio->uio_resid, KSTRING-1);
774 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
775 		return (error);
776 
777 	if (uio->uio_resid != 0)
778 		return (EIO);
779 
780 	strbuf[xlen] = '\0';
781 	xlen = strlen(strbuf);
782 	return (kernfs_xwrite(kfs, strbuf, xlen));
783 }
784 
785 int
786 kernfs_write(void *v)
787 {
788 	struct vop_write_args /* {
789 		struct vnode *a_vp;
790 		struct uio *a_uio;
791 		int  a_ioflag;
792 		kauth_cred_t a_cred;
793 	} */ *ap = v;
794 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
795 
796 	if (kfs->kfs_type < KFSlasttype) {
797 		/* use default function */
798 		return kernfs_default_xwrite(v);
799 	}
800 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
801 	    EOPNOTSUPP);
802 }
803 
804 int
805 kernfs_ioctl(void *v)
806 {
807 	struct vop_ioctl_args /* {
808 		const struct vnodeop_desc *a_desc;
809 		struct vnode *a_vp;
810 		u_long a_command;
811 		void *a_data;
812 		int a_fflag;
813 		kauth_cred_t a_cred;
814 	} */ *ap = v;
815 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
816 
817 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
818 	    EPASSTHROUGH);
819 }
820 
821 static int
822 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
823     u_int32_t value, struct vop_readdir_args *ap)
824 {
825 	struct kernfs_node *kfs;
826 	struct vnode *vp;
827 	int error;
828 
829 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
830 	    value)) != 0)
831 		return error;
832 	if (kt->kt_tag == KFSdevice) {
833 		struct vattr va;
834 
835 		error = VOP_GETATTR(vp, &va, ap->a_cred);
836 		if (error != 0) {
837 			return error;
838 		}
839 		d->d_fileno = va.va_fileid;
840 	} else {
841 		kfs = VTOKERN(vp);
842 		d->d_fileno = kfs->kfs_fileno;
843 	}
844 	vput(vp);
845 	return 0;
846 }
847 
848 static int
849 kernfs_setdirentfileno(struct dirent *d, off_t entry,
850     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
851     const struct kern_target *kt, struct vop_readdir_args *ap)
852 {
853 	const struct kern_target *ikt;
854 	int error;
855 
856 	switch (entry) {
857 	case 0:
858 		d->d_fileno = thisdir_kfs->kfs_fileno;
859 		return 0;
860 	case 1:
861 		ikt = parent_kt;
862 		break;
863 	default:
864 		ikt = kt;
865 		break;
866 	}
867 	if (ikt != thisdir_kfs->kfs_kt) {
868 		if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0)
869 			return error;
870 	} else
871 		d->d_fileno = thisdir_kfs->kfs_fileno;
872 	return 0;
873 }
874 
875 int
876 kernfs_readdir(void *v)
877 {
878 	struct vop_readdir_args /* {
879 		struct vnode *a_vp;
880 		struct uio *a_uio;
881 		kauth_cred_t a_cred;
882 		int *a_eofflag;
883 		off_t **a_cookies;
884 		int a_*ncookies;
885 	} */ *ap = v;
886 	struct uio *uio = ap->a_uio;
887 	struct dirent d;
888 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
889 	const struct kern_target *kt;
890 	const struct dyn_kern_target *dkt = NULL;
891 	const struct kernfs_subdir *ks;
892 	off_t i, j;
893 	int error;
894 	off_t *cookies = NULL;
895 	int ncookies = 0, n;
896 
897 	if (uio->uio_resid < UIO_MX)
898 		return (EINVAL);
899 	if (uio->uio_offset < 0)
900 		return (EINVAL);
901 
902 	error = 0;
903 	i = uio->uio_offset;
904 	memset(&d, 0, sizeof(d));
905 	d.d_reclen = UIO_MX;
906 	ncookies = uio->uio_resid / UIO_MX;
907 
908 	switch (kfs->kfs_type) {
909 	case KFSkern:
910 		if (i >= nkern_targets)
911 			return (0);
912 
913 		if (ap->a_ncookies) {
914 			ncookies = min(ncookies, (nkern_targets - i));
915 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
916 			    M_WAITOK);
917 			*ap->a_cookies = cookies;
918 		}
919 
920 		n = 0;
921 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
922 			if (i < static_nkern_targets)
923 				kt = &kern_targets[i];
924 			else {
925 				if (dkt == NULL) {
926 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
927 					for (j = static_nkern_targets; j < i &&
928 						     dkt != NULL; j++)
929 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
930 					if (j != i)
931 						break;
932 				} else {
933 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
934 				}
935 				if (dkt == NULL)
936 					break;
937 				kt = &dkt->dkt_kt;
938 			}
939 			if (kt->kt_tag == KFSdevice) {
940 				dev_t *dp = kt->kt_data;
941 				struct vnode *fvp;
942 
943 				if (*dp == NODEV ||
944 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
945 					continue;
946 				vrele(fvp);
947 			}
948 			if (kt->kt_tag == KFSmsgbuf) {
949 				if (!msgbufenabled
950 				    || msgbufp->msg_magic != MSG_MAGIC) {
951 					continue;
952 				}
953 			}
954 			d.d_namlen = kt->kt_namlen;
955 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
956 			    &kern_targets[0], kt, ap)) != 0)
957 				break;
958 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
959 			d.d_type = kt->kt_type;
960 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
961 				break;
962 			if (cookies)
963 				*cookies++ = i + 1;
964 			n++;
965 		}
966 		ncookies = n;
967 		break;
968 
969 	case KFSroot:
970 		if (i >= 2)
971 			return 0;
972 
973 		if (ap->a_ncookies) {
974 			ncookies = min(ncookies, (2 - i));
975 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
976 			    M_WAITOK);
977 			*ap->a_cookies = cookies;
978 		}
979 
980 		n = 0;
981 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
982 			kt = &kern_targets[i];
983 			d.d_namlen = kt->kt_namlen;
984 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
985 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
986 			d.d_type = kt->kt_type;
987 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
988 				break;
989 			if (cookies)
990 				*cookies++ = i + 1;
991 			n++;
992 		}
993 		ncookies = n;
994 		break;
995 
996 	case KFSsubdir:
997 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
998 		if (i >= ks->ks_nentries)
999 			return (0);
1000 
1001 		if (ap->a_ncookies) {
1002 			ncookies = min(ncookies, (ks->ks_nentries - i));
1003 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1004 			    M_WAITOK);
1005 			*ap->a_cookies = cookies;
1006 		}
1007 
1008 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1009 		for (j = 0; j < i && dkt != NULL; j++)
1010 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1011 		n = 0;
1012 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1013 			if (i < 2)
1014 				kt = &subdir_targets[i];
1015 			else {
1016 				/* check if ks_nentries lied to us */
1017 				if (dkt == NULL)
1018 					break;
1019 				kt = &dkt->dkt_kt;
1020 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1021 			}
1022 			if (kt->kt_tag == KFSdevice) {
1023 				dev_t *dp = kt->kt_data;
1024 				struct vnode *fvp;
1025 
1026 				if (*dp == NODEV ||
1027 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1028 					continue;
1029 				vrele(fvp);
1030 			}
1031 			d.d_namlen = kt->kt_namlen;
1032 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1033 			    ks->ks_parent, kt, ap)) != 0)
1034 				break;
1035 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1036 			d.d_type = kt->kt_type;
1037 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1038 				break;
1039 			if (cookies)
1040 				*cookies++ = i + 1;
1041 			n++;
1042 		}
1043 		ncookies = n;
1044 		break;
1045 
1046 	default:
1047 		error = ENOTDIR;
1048 		break;
1049 	}
1050 
1051 	if (ap->a_ncookies) {
1052 		if (error) {
1053 			if (cookies)
1054 				free(*ap->a_cookies, M_TEMP);
1055 			*ap->a_ncookies = 0;
1056 			*ap->a_cookies = NULL;
1057 		} else
1058 			*ap->a_ncookies = ncookies;
1059 	}
1060 
1061 	uio->uio_offset = i;
1062 	return (error);
1063 }
1064 
1065 int
1066 kernfs_inactive(void *v)
1067 {
1068 	struct vop_inactive_args /* {
1069 		struct vnode *a_vp;
1070 		bool *a_recycle;
1071 	} */ *ap = v;
1072 	struct vnode *vp = ap->a_vp;
1073 
1074 	*ap->a_recycle = false;
1075 	VOP_UNLOCK(vp);
1076 	return (0);
1077 }
1078 
1079 int
1080 kernfs_reclaim(void *v)
1081 {
1082 	struct vop_reclaim_args /* {
1083 		struct vnode *a_vp;
1084 	} */ *ap = v;
1085 
1086 	return (kernfs_freevp(ap->a_vp));
1087 }
1088 
1089 /*
1090  * Return POSIX pathconf information applicable to special devices.
1091  */
1092 int
1093 kernfs_pathconf(void *v)
1094 {
1095 	struct vop_pathconf_args /* {
1096 		struct vnode *a_vp;
1097 		int a_name;
1098 		register_t *a_retval;
1099 	} */ *ap = v;
1100 
1101 	switch (ap->a_name) {
1102 	case _PC_LINK_MAX:
1103 		*ap->a_retval = LINK_MAX;
1104 		return (0);
1105 	case _PC_MAX_CANON:
1106 		*ap->a_retval = MAX_CANON;
1107 		return (0);
1108 	case _PC_MAX_INPUT:
1109 		*ap->a_retval = MAX_INPUT;
1110 		return (0);
1111 	case _PC_PIPE_BUF:
1112 		*ap->a_retval = PIPE_BUF;
1113 		return (0);
1114 	case _PC_CHOWN_RESTRICTED:
1115 		*ap->a_retval = 1;
1116 		return (0);
1117 	case _PC_VDISABLE:
1118 		*ap->a_retval = _POSIX_VDISABLE;
1119 		return (0);
1120 	case _PC_SYNC_IO:
1121 		*ap->a_retval = 1;
1122 		return (0);
1123 	default:
1124 		return (EINVAL);
1125 	}
1126 	/* NOTREACHED */
1127 }
1128 
1129 /*
1130  * Print out the contents of a /dev/fd vnode.
1131  */
1132 /* ARGSUSED */
1133 int
1134 kernfs_print(void *v)
1135 {
1136 
1137 	printf("tag VT_KERNFS, kernfs vnode\n");
1138 	return (0);
1139 }
1140 
1141 int
1142 kernfs_link(void *v)
1143 {
1144 	struct vop_link_args /* {
1145 		struct vnode *a_dvp;
1146 		struct vnode *a_vp;
1147 		struct componentname *a_cnp;
1148 	} */ *ap = v;
1149 
1150 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1151 	vput(ap->a_dvp);
1152 	return (EROFS);
1153 }
1154 
1155 int
1156 kernfs_symlink(void *v)
1157 {
1158 	struct vop_symlink_args /* {
1159 		struct vnode *a_dvp;
1160 		struct vnode **a_vpp;
1161 		struct componentname *a_cnp;
1162 		struct vattr *a_vap;
1163 		char *a_target;
1164 	} */ *ap = v;
1165 
1166 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1167 	vput(ap->a_dvp);
1168 	return (EROFS);
1169 }
1170