xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision 76c7fc5f6b13ed0b1508e6b313e88e59977ed78e)
1 /*	$NetBSD: kernfs_vnops.c,v 1.161 2019/08/29 06:43:13 hannken 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 1.161 2019/08/29 06:43:13 hannken 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 #include <miscfs/specfs/specdev.h>
63 
64 #include <uvm/uvm_extern.h>
65 
66 #define KSTRING	256		/* Largest I/O available via this filesystem */
67 #define	UIO_MX 32
68 
69 #define	READ_MODE	(S_IRUSR|S_IRGRP|S_IROTH)
70 #define	WRITE_MODE	(S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH)
71 #define	UREAD_MODE	(S_IRUSR)
72 #define	DIR_MODE	(S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH)
73 #define	UDIR_MODE	(S_IRUSR|S_IXUSR)
74 
75 #define N(s) sizeof(s)-1, s
76 const struct kern_target kern_targets[] = {
77 /* NOTE: The name must be less than UIO_MX-16 chars in length */
78      /*        name            data          tag           type  ro/rw */
79      { DT_DIR, N("."),         0,            KFSkern,        VDIR, DIR_MODE   },
80      { DT_DIR, N(".."),        0,            KFSroot,        VDIR, DIR_MODE   },
81      { DT_REG, N("boottime"),  &boottime.tv_sec, KFSint,     VREG, READ_MODE  },
82 			/* XXXUNCONST */
83      { DT_REG, N("copyright"), __UNCONST(copyright),
84      					     KFSstring,      VREG, READ_MODE  },
85      { DT_REG, N("hostname"),  0,            KFShostname,    VREG, WRITE_MODE },
86      { DT_REG, N("hz"),        &hz,          KFSint,         VREG, READ_MODE  },
87      { DT_REG, N("loadavg"),   0,            KFSavenrun,     VREG, READ_MODE  },
88      { DT_REG, N("msgbuf"),    0,	     KFSmsgbuf,      VREG, READ_MODE  },
89      { DT_REG, N("pagesize"),  &uvmexp.pagesize, KFSint,     VREG, READ_MODE  },
90      { DT_REG, N("physmem"),   &physmem,     KFSint,         VREG, READ_MODE  },
91 #if 0
92      { DT_DIR, N("root"),      0,            KFSnull,        VDIR, DIR_MODE   },
93 #endif
94      { DT_BLK, N("rootdev"),   &rootdev,     KFSdevice,      VBLK, READ_MODE  },
95      { DT_CHR, N("rrootdev"),  &rrootdev,    KFSdevice,      VCHR, READ_MODE  },
96      { DT_REG, N("time"),      0,            KFStime,        VREG, READ_MODE  },
97 			/* XXXUNCONST */
98      { DT_REG, N("version"),   __UNCONST(version),
99      					     KFSstring,      VREG, READ_MODE  },
100 };
101 const struct kern_target subdir_targets[] = {
102 /* NOTE: The name must be less than UIO_MX-16 chars in length */
103      /*        name            data          tag           type  ro/rw */
104      { DT_DIR, N("."),         0,            KFSsubdir,      VDIR, DIR_MODE   },
105      { DT_DIR, N(".."),        0,            KFSkern,        VDIR, DIR_MODE   },
106 };
107 #undef N
108 SIMPLEQ_HEAD(,dyn_kern_target) dyn_kern_targets =
109 	SIMPLEQ_HEAD_INITIALIZER(dyn_kern_targets);
110 int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
111 const int static_nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]);
112 int nkern_dirs = 2;
113 
114 int kernfs_try_fileop(kfstype, kfsfileop, void *, int);
115 int kernfs_try_xread(kfstype, const struct kernfs_node *, char **,
116     size_t, int);
117 int kernfs_try_xwrite(kfstype, const struct kernfs_node *, char *,
118     size_t, int);
119 
120 static int kernfs_default_xread(void *v);
121 static int kernfs_default_xwrite(void *v);
122 static int kernfs_default_fileop_getattr(void *);
123 
124 /* must include all fileop's */
125 const struct kernfs_fileop kernfs_default_fileops[] = {
126   { .kf_fileop = KERNFS_XREAD },
127   { .kf_fileop = KERNFS_XWRITE },
128   { .kf_fileop = KERNFS_FILEOP_OPEN },
129   { .kf_fileop = KERNFS_FILEOP_GETATTR,
130     .kf_vop = kernfs_default_fileop_getattr },
131   { .kf_fileop = KERNFS_FILEOP_IOCTL },
132   { .kf_fileop = KERNFS_FILEOP_CLOSE },
133   { .kf_fileop = KERNFS_FILEOP_READ,
134     .kf_vop = kernfs_default_xread },
135   { .kf_fileop = KERNFS_FILEOP_WRITE,
136     .kf_vop = kernfs_default_xwrite },
137 };
138 
139 int	kernfs_lookup(void *);
140 #define	kernfs_create	genfs_eopnotsupp
141 #define	kernfs_mknod	genfs_eopnotsupp
142 int	kernfs_open(void *);
143 int	kernfs_close(void *);
144 int	kernfs_access(void *);
145 int	kernfs_getattr(void *);
146 int	kernfs_setattr(void *);
147 int	kernfs_read(void *);
148 int	kernfs_write(void *);
149 #define	kernfs_fcntl	genfs_fcntl
150 int	kernfs_ioctl(void *);
151 #define	kernfs_poll	genfs_poll
152 #define kernfs_revoke	genfs_revoke
153 #define	kernfs_fsync	genfs_nullop
154 #define	kernfs_seek	genfs_nullop
155 #define	kernfs_remove	genfs_eopnotsupp
156 int	kernfs_link(void *);
157 #define	kernfs_rename	genfs_eopnotsupp
158 #define	kernfs_mkdir	genfs_eopnotsupp
159 #define	kernfs_rmdir	genfs_eopnotsupp
160 int	kernfs_symlink(void *);
161 int	kernfs_readdir(void *);
162 #define	kernfs_readlink	genfs_eopnotsupp
163 #define	kernfs_abortop	genfs_abortop
164 int	kernfs_inactive(void *);
165 int	kernfs_reclaim(void *);
166 #define	kernfs_lock	genfs_lock
167 #define	kernfs_unlock	genfs_unlock
168 #define	kernfs_bmap	genfs_badop
169 #define	kernfs_strategy	genfs_badop
170 int	kernfs_print(void *);
171 #define	kernfs_islocked	genfs_islocked
172 int	kernfs_pathconf(void *);
173 #define	kernfs_advlock	genfs_einval
174 #define	kernfs_bwrite	genfs_eopnotsupp
175 int	kernfs_getpages(void *);
176 #define	kernfs_putpages	genfs_putpages
177 
178 static int	kernfs_xread(struct kernfs_node *, int, char **,
179 				size_t, size_t *);
180 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
181 
182 int (**kernfs_vnodeop_p)(void *);
183 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
184 	{ &vop_default_desc, vn_default_error },
185 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
186 	{ &vop_create_desc, kernfs_create },		/* create */
187 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
188 	{ &vop_open_desc, kernfs_open },		/* open */
189 	{ &vop_close_desc, kernfs_close },		/* close */
190 	{ &vop_access_desc, kernfs_access },		/* access */
191 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
192 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
193 	{ &vop_read_desc, kernfs_read },		/* read */
194 	{ &vop_write_desc, kernfs_write },		/* write */
195 	{ &vop_fallocate_desc, genfs_eopnotsupp },	/* fallocate */
196 	{ &vop_fdiscard_desc, genfs_eopnotsupp },	/* fdiscard */
197 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
198 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
199 	{ &vop_poll_desc, kernfs_poll },		/* poll */
200 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
201 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
202 	{ &vop_seek_desc, kernfs_seek },		/* seek */
203 	{ &vop_remove_desc, kernfs_remove },		/* remove */
204 	{ &vop_link_desc, kernfs_link },		/* link */
205 	{ &vop_rename_desc, kernfs_rename },		/* rename */
206 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
207 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
208 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
209 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
210 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
211 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
212 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
213 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
214 	{ &vop_lock_desc, kernfs_lock },		/* lock */
215 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
216 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
217 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
218 	{ &vop_print_desc, kernfs_print },		/* print */
219 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
220 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
221 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
222 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
223 	{ &vop_getpages_desc, kernfs_getpages },	/* getpages */
224 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
225 	{ NULL, NULL }
226 };
227 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
228 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
229 
230 static inline int
231 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
232 {
233 	if (a->kf_type < b->kf_type)
234 		return -1;
235 	if (a->kf_type > b->kf_type)
236 		return 1;
237 	if (a->kf_fileop < b->kf_fileop)
238 		return -1;
239 	if (a->kf_fileop > b->kf_fileop)
240 		return 1;
241 	return (0);
242 }
243 
244 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
245 	SPLAY_INITIALIZER(kfsfileoptree);
246 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
247 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
248 
249 kfstype
250 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
251 {
252 	static u_char nextfreetype = KFSlasttype;
253 	struct kernfs_fileop *dkf, *fkf, skf;
254 	int i;
255 
256 	/* XXX need to keep track of dkf's memory if we support
257            deallocating types */
258 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
259 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
260 
261 	for (i = 0; i < sizeof(kernfs_default_fileops) /
262 		     sizeof(kernfs_default_fileops[0]); i++) {
263 		dkf[i].kf_type = nextfreetype;
264 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
265 	}
266 
267 	for (i = 0; i < nkf; i++) {
268 		skf.kf_type = nextfreetype;
269 		skf.kf_fileop = kf[i].kf_fileop;
270 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
271 			fkf->kf_vop = kf[i].kf_vop;
272 	}
273 
274 	return nextfreetype++;
275 }
276 
277 int
278 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
279 {
280 	struct kernfs_fileop *kf, skf;
281 
282 	skf.kf_type = type;
283 	skf.kf_fileop = fileop;
284 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
285 		if (kf->kf_vop)
286 			return kf->kf_vop(v);
287 	return error;
288 }
289 
290 int
291 kernfs_try_xread(kfstype type, const struct kernfs_node *kfs, char **bfp,
292     size_t len, int error)
293 {
294 	struct kernfs_fileop *kf, skf;
295 
296 	skf.kf_type = type;
297 	skf.kf_fileop = KERNFS_XREAD;
298 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
299 		if (kf->kf_xread)
300 			return kf->kf_xread(kfs, bfp, len);
301 	return error;
302 }
303 
304 int
305 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
306     size_t len, int error)
307 {
308 	struct kernfs_fileop *kf, skf;
309 
310 	skf.kf_type = type;
311 	skf.kf_fileop = KERNFS_XWRITE;
312 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
313 		if (kf->kf_xwrite)
314 			return kf->kf_xwrite(kfs, bf, len);
315 	return error;
316 }
317 
318 int
319 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
320 {
321 	struct kernfs_subdir *ks, *parent;
322 
323 	if (pkt == NULL) {
324 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
325 		nkern_targets++;
326 		if (dkt->dkt_kt.kt_vtype == VDIR)
327 			nkern_dirs++;
328 	} else {
329 		parent = (struct kernfs_subdir *)pkt->kt_data;
330 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
331 		parent->ks_nentries++;
332 		if (dkt->dkt_kt.kt_vtype == VDIR)
333 			parent->ks_dirs++;
334 	}
335 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
336 		ks = malloc(sizeof(struct kernfs_subdir),
337 		    M_TEMP, M_WAITOK);
338 		SIMPLEQ_INIT(&ks->ks_entries);
339 		ks->ks_nentries = 2; /* . and .. */
340 		ks->ks_dirs = 2;
341 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
342 		dkt->dkt_kt.kt_data = ks;
343 	}
344 	return 0;
345 }
346 
347 static int
348 kernfs_xread(struct kernfs_node *kfs, int off, char **bufp, size_t len, size_t *wrlen)
349 {
350 	const struct kern_target *kt;
351 	int err;
352 
353 	kt = kfs->kfs_kt;
354 
355 	switch (kfs->kfs_type) {
356 	case KFStime: {
357 		struct timeval tv;
358 
359 		microtime(&tv);
360 		snprintf(*bufp, len, "%lld %ld\n", (long long)tv.tv_sec,
361 		    (long)tv.tv_usec);
362 		break;
363 	}
364 
365 	case KFSint: {
366 		int *ip = kt->kt_data;
367 
368 		snprintf(*bufp, len, "%d\n", *ip);
369 		break;
370 	}
371 
372 	case KFSstring: {
373 		char *cp = kt->kt_data;
374 
375 		*bufp = cp;
376 		break;
377 	}
378 
379 	case KFSmsgbuf: {
380 		long n;
381 
382 		/*
383 		 * deal with cases where the message buffer has
384 		 * become corrupted.
385 		 */
386 		if (!logenabled(msgbufp)) {
387 			msgbufenabled = 0;
388 			return (ENXIO);
389 		}
390 
391 		/*
392 		 * Note that reads of /kern/msgbuf won't necessarily yield
393 		 * consistent results, if the message buffer is modified
394 		 * while the read is in progress.  The worst that can happen
395 		 * is that incorrect data will be read.  There's no way
396 		 * that this can crash the system unless the values in the
397 		 * message buffer header are corrupted, but that'll cause
398 		 * the system to die anyway.
399 		 */
400 		if (off >= msgbufp->msg_bufs) {
401 			*wrlen = 0;
402 			return (0);
403 		}
404 		n = msgbufp->msg_bufx + off;
405 		if (n >= msgbufp->msg_bufs)
406 			n -= msgbufp->msg_bufs;
407 		len = uimin(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
408 		*bufp = msgbufp->msg_bufc + n;
409 		*wrlen = len;
410 		return (0);
411 	}
412 
413 	case KFShostname: {
414 		char *cp = hostname;
415 		size_t xlen = hostnamelen;
416 
417 		if (xlen >= (len - 2))
418 			return (EINVAL);
419 
420 		memcpy(*bufp, cp, xlen);
421 		(*bufp)[xlen] = '\n';
422 		(*bufp)[xlen+1] = '\0';
423 		break;
424 	}
425 
426 	case KFSavenrun:
427 		averunnable.fscale = FSCALE;
428 		snprintf(*bufp, len, "%d %d %d %ld\n",
429 		    averunnable.ldavg[0], averunnable.ldavg[1],
430 		    averunnable.ldavg[2], averunnable.fscale);
431 		break;
432 
433 	default:
434 		err = kernfs_try_xread(kfs->kfs_type, kfs, bufp, len,
435 		    EOPNOTSUPP);
436 		if (err)
437 			return err;
438 	}
439 
440 	len = strlen(*bufp);
441 	if (len <= off)
442 		*wrlen = 0;
443 	else {
444 		*bufp += off;
445 		*wrlen = len - off;
446 	}
447 	return (0);
448 }
449 
450 static int
451 kernfs_xwrite(const struct kernfs_node *kfs, char *bf, size_t len)
452 {
453 
454 	switch (kfs->kfs_type) {
455 	case KFShostname:
456 		if (bf[len-1] == '\n')
457 			--len;
458 		memcpy(hostname, bf, len);
459 		hostname[len] = '\0';
460 		hostnamelen = (size_t) len;
461 		return (0);
462 
463 	default:
464 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
465 	}
466 }
467 
468 
469 /*
470  * vp is the current namei directory
471  * ndp is the name to locate in that directory...
472  */
473 int
474 kernfs_lookup(void *v)
475 {
476 	struct vop_lookup_v2_args /* {
477 		struct vnode * a_dvp;
478 		struct vnode ** a_vpp;
479 		struct componentname * a_cnp;
480 	} */ *ap = v;
481 	struct componentname *cnp = ap->a_cnp;
482 	struct vnode **vpp = ap->a_vpp;
483 	struct vnode *dvp = ap->a_dvp;
484 	const char *pname = cnp->cn_nameptr;
485 	const struct kernfs_node *kfs;
486 	const struct kern_target *kt;
487 	const struct dyn_kern_target *dkt;
488 	const struct kernfs_subdir *ks;
489 	int error, i;
490 
491 	*vpp = NULLVP;
492 
493 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
494 		return (EROFS);
495 
496 	if (cnp->cn_namelen == 1 && *pname == '.') {
497 		*vpp = dvp;
498 		vref(dvp);
499 		return (0);
500 	}
501 
502 	kfs = VTOKERN(dvp);
503 	switch (kfs->kfs_type) {
504 	case KFSkern:
505 		/*
506 		 * Shouldn't get here with .. in the root node.
507 		 */
508 		if (cnp->cn_flags & ISDOTDOT)
509 			return (EIO);
510 
511 		for (i = 0; i < static_nkern_targets; i++) {
512 			kt = &kern_targets[i];
513 			if (cnp->cn_namelen == kt->kt_namlen &&
514 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
515 				goto found;
516 		}
517 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
518 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
519 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
520 				kt = &dkt->dkt_kt;
521 				goto found;
522 			}
523 		}
524 		break;
525 
526 	found:
527 		error = vcache_get(dvp->v_mount, &kt, sizeof(kt), vpp);
528 		return error;
529 
530 	case KFSsubdir:
531 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
532 		if (cnp->cn_flags & ISDOTDOT) {
533 			kt = ks->ks_parent;
534 			goto found;
535 		}
536 
537 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
538 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
539 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
540 				kt = &dkt->dkt_kt;
541 				goto found;
542 			}
543 		}
544 		break;
545 
546 	default:
547 		return (ENOTDIR);
548 	}
549 
550 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
551 }
552 
553 int
554 kernfs_open(void *v)
555 {
556 	struct vop_open_args /* {
557 		struct vnode *a_vp;
558 		int a_mode;
559 		kauth_cred_t a_cred;
560 	} */ *ap = v;
561 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
562 
563 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN, v, 0);
564 }
565 
566 int
567 kernfs_close(void *v)
568 {
569 	struct vop_close_args /* {
570 		struct vnode *a_vp;
571 		int a_fflag;
572 		kauth_cred_t a_cred;
573 	} */ *ap = v;
574 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
575 
576 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE, v, 0);
577 }
578 
579 int
580 kernfs_access(void *v)
581 {
582 	struct vop_access_args /* {
583 		struct vnode *a_vp;
584 		int a_mode;
585 		kauth_cred_t a_cred;
586 	} */ *ap = v;
587 	struct vattr va;
588 	int error;
589 
590 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
591 		return (error);
592 
593 	return kauth_authorize_vnode(ap->a_cred,
594 	    KAUTH_ACCESS_ACTION(ap->a_mode, ap->a_vp->v_type, va.va_mode),
595 	    ap->a_vp, NULL, genfs_can_access(va.va_type, va.va_mode,
596 	    va.va_uid, va.va_gid, ap->a_mode, ap->a_cred));
597 }
598 
599 static int
600 kernfs_default_fileop_getattr(void *v)
601 {
602 	struct vop_getattr_args /* {
603 		struct vnode *a_vp;
604 		struct vattr *a_vap;
605 		kauth_cred_t a_cred;
606 	} */ *ap = v;
607 	struct vattr *vap = ap->a_vap;
608 
609 	vap->va_nlink = 1;
610 	vap->va_bytes = vap->va_size = 0;
611 
612 	return 0;
613 }
614 
615 int
616 kernfs_getattr(void *v)
617 {
618 	struct vop_getattr_args /* {
619 		struct vnode *a_vp;
620 		struct vattr *a_vap;
621 		kauth_cred_t a_cred;
622 	} */ *ap = v;
623 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
624 	struct kernfs_subdir *ks;
625 	struct vattr *vap = ap->a_vap;
626 	int error = 0;
627 	char strbuf[KSTRING], *bf;
628 	size_t nread, total;
629 
630 	vattr_null(vap);
631 	vap->va_type = ap->a_vp->v_type;
632 	vap->va_uid = 0;
633 	vap->va_gid = 0;
634 	vap->va_mode = kfs->kfs_mode;
635 	vap->va_fileid = kfs->kfs_fileno;
636 	vap->va_flags = 0;
637 	vap->va_size = 0;
638 	vap->va_blocksize = DEV_BSIZE;
639 	/* Make all times be current TOD, except for the "boottime" node. */
640 	if (kfs->kfs_kt->kt_namlen == 8 &&
641 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
642 		vap->va_ctime = boottime;
643 	} else {
644 		getnanotime(&vap->va_ctime);
645 	}
646 	vap->va_atime = vap->va_mtime = vap->va_ctime;
647 	vap->va_gen = 0;
648 	vap->va_flags = 0;
649 	vap->va_rdev = 0;
650 	vap->va_bytes = 0;
651 
652 	switch (kfs->kfs_type) {
653 	case KFSkern:
654 		vap->va_nlink = nkern_dirs;
655 		vap->va_bytes = vap->va_size = DEV_BSIZE;
656 		break;
657 
658 	case KFSdevice:
659 		vap->va_nlink = 1;
660 		vap->va_rdev = ap->a_vp->v_rdev;
661 		break;
662 
663 	case KFSroot:
664 		vap->va_nlink = 1;
665 		vap->va_bytes = vap->va_size = DEV_BSIZE;
666 		break;
667 
668 	case KFSsubdir:
669 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
670 		vap->va_nlink = ks->ks_dirs;
671 		vap->va_bytes = vap->va_size = DEV_BSIZE;
672 		break;
673 
674 	case KFSnull:
675 	case KFStime:
676 	case KFSint:
677 	case KFSstring:
678 	case KFShostname:
679 	case KFSavenrun:
680 	case KFSmsgbuf:
681 		vap->va_nlink = 1;
682 		total = 0;
683 		do {
684 			bf = strbuf;
685 			error = kernfs_xread(kfs, total, &bf,
686 			    sizeof(strbuf), &nread);
687 			total += nread;
688 		} while (error == 0 && nread != 0);
689 		vap->va_bytes = vap->va_size = total;
690 		break;
691 
692 	default:
693 		error = kernfs_try_fileop(kfs->kfs_type,
694 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
695 		break;
696 	}
697 
698 	return (error);
699 }
700 
701 /*ARGSUSED*/
702 int
703 kernfs_setattr(void *v)
704 {
705 
706 	/*
707 	 * Silently ignore attribute changes.
708 	 * This allows for open with truncate to have no
709 	 * effect until some data is written.  I want to
710 	 * do it this way because all writes are atomic.
711 	 */
712 	return (0);
713 }
714 
715 int
716 kernfs_default_xread(void *v)
717 {
718 	struct vop_read_args /* {
719 		struct vnode *a_vp;
720 		struct uio *a_uio;
721 		int  a_ioflag;
722 		kauth_cred_t a_cred;
723 	} */ *ap = v;
724 	struct uio *uio = ap->a_uio;
725 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
726 	char strbuf[KSTRING], *bf;
727 	int off;
728 	size_t len;
729 	int error;
730 
731 	if (ap->a_vp->v_type == VDIR)
732 		return EISDIR;
733 
734 	off = (int)uio->uio_offset;
735 	/* Don't allow negative offsets */
736 	if (off < 0)
737 		return EINVAL;
738 
739 	bf = strbuf;
740 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
741 		error = uiomove(bf, len, uio);
742 	return (error);
743 }
744 
745 int
746 kernfs_read(void *v)
747 {
748 	struct vop_read_args /* {
749 		struct vnode *a_vp;
750 		struct uio *a_uio;
751 		int  a_ioflag;
752 		struct ucred *a_cred;
753 	} */ *ap = v;
754 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
755 
756 	if (kfs->kfs_type < KFSlasttype) {
757 		/* use default function */
758 		return kernfs_default_xread(v);
759 	}
760 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
761 	   EOPNOTSUPP);
762 }
763 
764 static int
765 kernfs_default_xwrite(void *v)
766 {
767 	struct vop_write_args /* {
768 		struct vnode *a_vp;
769 		struct uio *a_uio;
770 		int  a_ioflag;
771 		kauth_cred_t a_cred;
772 	} */ *ap = v;
773 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
774 	struct uio *uio = ap->a_uio;
775 	int error;
776 	size_t xlen;
777 	char strbuf[KSTRING];
778 
779 	if (uio->uio_offset != 0)
780 		return (EINVAL);
781 
782 	xlen = uimin(uio->uio_resid, KSTRING-1);
783 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
784 		return (error);
785 
786 	if (uio->uio_resid != 0)
787 		return (EIO);
788 
789 	strbuf[xlen] = '\0';
790 	xlen = strlen(strbuf);
791 	return (kernfs_xwrite(kfs, strbuf, xlen));
792 }
793 
794 int
795 kernfs_write(void *v)
796 {
797 	struct vop_write_args /* {
798 		struct vnode *a_vp;
799 		struct uio *a_uio;
800 		int  a_ioflag;
801 		kauth_cred_t a_cred;
802 	} */ *ap = v;
803 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
804 
805 	if (kfs->kfs_type < KFSlasttype) {
806 		/* use default function */
807 		return kernfs_default_xwrite(v);
808 	}
809 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
810 	    EOPNOTSUPP);
811 }
812 
813 int
814 kernfs_ioctl(void *v)
815 {
816 	struct vop_ioctl_args /* {
817 		const struct vnodeop_desc *a_desc;
818 		struct vnode *a_vp;
819 		u_long a_command;
820 		void *a_data;
821 		int a_fflag;
822 		kauth_cred_t a_cred;
823 	} */ *ap = v;
824 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
825 
826 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
827 	    EPASSTHROUGH);
828 }
829 
830 static int
831 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
832     struct vop_readdir_args *ap)
833 {
834 	struct kernfs_node *kfs;
835 	struct vnode *vp;
836 	int error;
837 
838 	if ((error = vcache_get(ap->a_vp->v_mount, &kt, sizeof(kt), &vp)) != 0)
839 		return error;
840 	kfs = VTOKERN(vp);
841 	d->d_fileno = kfs->kfs_fileno;
842 	vrele(vp);
843 	return 0;
844 }
845 
846 static int
847 kernfs_setdirentfileno(struct dirent *d, off_t entry,
848     struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt,
849     const struct kern_target *kt, struct vop_readdir_args *ap)
850 {
851 	const struct kern_target *ikt;
852 	int error;
853 
854 	switch (entry) {
855 	case 0:
856 		d->d_fileno = thisdir_kfs->kfs_fileno;
857 		return 0;
858 	case 1:
859 		ikt = parent_kt;
860 		break;
861 	default:
862 		ikt = kt;
863 		break;
864 	}
865 	if (ikt != thisdir_kfs->kfs_kt) {
866 		if ((error = kernfs_setdirentfileno_kt(d, ikt, ap)) != 0)
867 			return error;
868 	} else
869 		d->d_fileno = thisdir_kfs->kfs_fileno;
870 	return 0;
871 }
872 
873 int
874 kernfs_readdir(void *v)
875 {
876 	struct vop_readdir_args /* {
877 		struct vnode *a_vp;
878 		struct uio *a_uio;
879 		kauth_cred_t a_cred;
880 		int *a_eofflag;
881 		off_t **a_cookies;
882 		int a_*ncookies;
883 	} */ *ap = v;
884 	struct uio *uio = ap->a_uio;
885 	struct dirent d;
886 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
887 	const struct kern_target *kt;
888 	const struct dyn_kern_target *dkt = NULL;
889 	const struct kernfs_subdir *ks;
890 	off_t i, j;
891 	int error;
892 	off_t *cookies = NULL;
893 	int ncookies = 0, n;
894 
895 	if (uio->uio_resid < UIO_MX)
896 		return (EINVAL);
897 	if (uio->uio_offset < 0)
898 		return (EINVAL);
899 
900 	error = 0;
901 	i = uio->uio_offset;
902 	memset(&d, 0, sizeof(d));
903 	d.d_reclen = UIO_MX;
904 	ncookies = uio->uio_resid / UIO_MX;
905 
906 	switch (kfs->kfs_type) {
907 	case KFSkern:
908 		if (i >= nkern_targets)
909 			return (0);
910 
911 		if (ap->a_ncookies) {
912 			ncookies = uimin(ncookies, (nkern_targets - i));
913 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
914 			    M_WAITOK);
915 			*ap->a_cookies = cookies;
916 		}
917 
918 		n = 0;
919 		for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) {
920 			if (i < static_nkern_targets)
921 				kt = &kern_targets[i];
922 			else {
923 				if (dkt == NULL) {
924 					dkt = SIMPLEQ_FIRST(&dyn_kern_targets);
925 					for (j = static_nkern_targets; j < i &&
926 						     dkt != NULL; j++)
927 						dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
928 					if (j != i)
929 						break;
930 				} else {
931 					dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
932 				}
933 				if (dkt == NULL)
934 					break;
935 				kt = &dkt->dkt_kt;
936 			}
937 			if (kt->kt_tag == KFSdevice) {
938 				dev_t *dp = kt->kt_data;
939 				struct vnode *fvp;
940 
941 				if (*dp == NODEV ||
942 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
943 					continue;
944 				vrele(fvp);
945 			}
946 			if (kt->kt_tag == KFSmsgbuf) {
947 				if (!logenabled(msgbufp)) {
948 					continue;
949 				}
950 			}
951 			d.d_namlen = kt->kt_namlen;
952 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
953 			    &kern_targets[0], kt, ap)) != 0)
954 				break;
955 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
956 			d.d_type = kt->kt_type;
957 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
958 				break;
959 			if (cookies)
960 				*cookies++ = i + 1;
961 			n++;
962 		}
963 		ncookies = n;
964 		break;
965 
966 	case KFSroot:
967 		if (i >= 2)
968 			return 0;
969 
970 		if (ap->a_ncookies) {
971 			ncookies = uimin(ncookies, (2 - i));
972 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
973 			    M_WAITOK);
974 			*ap->a_cookies = cookies;
975 		}
976 
977 		n = 0;
978 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
979 			kt = &kern_targets[i];
980 			d.d_namlen = kt->kt_namlen;
981 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
982 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
983 			d.d_type = kt->kt_type;
984 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
985 				break;
986 			if (cookies)
987 				*cookies++ = i + 1;
988 			n++;
989 		}
990 		ncookies = n;
991 		break;
992 
993 	case KFSsubdir:
994 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
995 		if (i >= ks->ks_nentries)
996 			return (0);
997 
998 		if (ap->a_ncookies) {
999 			ncookies = uimin(ncookies, (ks->ks_nentries - i));
1000 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1001 			    M_WAITOK);
1002 			*ap->a_cookies = cookies;
1003 		}
1004 
1005 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1006 		for (j = 0; j < i && dkt != NULL; j++)
1007 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1008 		n = 0;
1009 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1010 			if (i < 2)
1011 				kt = &subdir_targets[i];
1012 			else {
1013 				/* check if ks_nentries lied to us */
1014 				if (dkt == NULL)
1015 					break;
1016 				kt = &dkt->dkt_kt;
1017 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1018 			}
1019 			if (kt->kt_tag == KFSdevice) {
1020 				dev_t *dp = kt->kt_data;
1021 				struct vnode *fvp;
1022 
1023 				if (*dp == NODEV ||
1024 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1025 					continue;
1026 				vrele(fvp);
1027 			}
1028 			d.d_namlen = kt->kt_namlen;
1029 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1030 			    ks->ks_parent, kt, ap)) != 0)
1031 				break;
1032 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1033 			d.d_type = kt->kt_type;
1034 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1035 				break;
1036 			if (cookies)
1037 				*cookies++ = i + 1;
1038 			n++;
1039 		}
1040 		ncookies = n;
1041 		break;
1042 
1043 	default:
1044 		error = ENOTDIR;
1045 		break;
1046 	}
1047 
1048 	if (ap->a_ncookies) {
1049 		if (error) {
1050 			if (cookies)
1051 				free(*ap->a_cookies, M_TEMP);
1052 			*ap->a_ncookies = 0;
1053 			*ap->a_cookies = NULL;
1054 		} else
1055 			*ap->a_ncookies = ncookies;
1056 	}
1057 
1058 	uio->uio_offset = i;
1059 	return (error);
1060 }
1061 
1062 int
1063 kernfs_inactive(void *v)
1064 {
1065 	struct vop_inactive_v2_args /* {
1066 		struct vnode *a_vp;
1067 		bool *a_recycle;
1068 	} */ *ap = v;
1069 
1070 	*ap->a_recycle = false;
1071 
1072 	return (0);
1073 }
1074 
1075 int
1076 kernfs_reclaim(void *v)
1077 {
1078 	struct vop_reclaim_v2_args /* {
1079 		struct vnode *a_vp;
1080 	} */ *ap = v;
1081 	struct vnode *vp = ap->a_vp;
1082 	struct kernfs_node *kfs = VTOKERN(vp);
1083 
1084 	VOP_UNLOCK(vp);
1085 
1086 	vp->v_data = NULL;
1087 	mutex_enter(&kfs_lock);
1088 	TAILQ_REMOVE(&VFSTOKERNFS(vp->v_mount)->nodelist, kfs, kfs_list);
1089 	mutex_exit(&kfs_lock);
1090 	kmem_free(kfs, sizeof(struct kernfs_node));
1091 
1092 	return 0;
1093 }
1094 
1095 /*
1096  * Return POSIX pathconf information applicable to special devices.
1097  */
1098 int
1099 kernfs_pathconf(void *v)
1100 {
1101 	struct vop_pathconf_args /* {
1102 		struct vnode *a_vp;
1103 		int a_name;
1104 		register_t *a_retval;
1105 	} */ *ap = v;
1106 
1107 	switch (ap->a_name) {
1108 	case _PC_LINK_MAX:
1109 		*ap->a_retval = LINK_MAX;
1110 		return (0);
1111 	case _PC_MAX_CANON:
1112 		*ap->a_retval = MAX_CANON;
1113 		return (0);
1114 	case _PC_MAX_INPUT:
1115 		*ap->a_retval = MAX_INPUT;
1116 		return (0);
1117 	case _PC_PIPE_BUF:
1118 		*ap->a_retval = PIPE_BUF;
1119 		return (0);
1120 	case _PC_CHOWN_RESTRICTED:
1121 		*ap->a_retval = 1;
1122 		return (0);
1123 	case _PC_VDISABLE:
1124 		*ap->a_retval = _POSIX_VDISABLE;
1125 		return (0);
1126 	case _PC_SYNC_IO:
1127 		*ap->a_retval = 1;
1128 		return (0);
1129 	default:
1130 		return (EINVAL);
1131 	}
1132 	/* NOTREACHED */
1133 }
1134 
1135 /*
1136  * Print out the contents of a /dev/fd vnode.
1137  */
1138 /* ARGSUSED */
1139 int
1140 kernfs_print(void *v)
1141 {
1142 
1143 	printf("tag VT_KERNFS, kernfs vnode\n");
1144 	return (0);
1145 }
1146 
1147 int
1148 kernfs_link(void *v)
1149 {
1150 	struct vop_link_v2_args /* {
1151 		struct vnode *a_dvp;
1152 		struct vnode *a_vp;
1153 		struct componentname *a_cnp;
1154 	} */ *ap = v;
1155 
1156 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1157 	return (EROFS);
1158 }
1159 
1160 int
1161 kernfs_symlink(void *v)
1162 {
1163 	struct vop_symlink_v3_args /* {
1164 		struct vnode *a_dvp;
1165 		struct vnode **a_vpp;
1166 		struct componentname *a_cnp;
1167 		struct vattr *a_vap;
1168 		char *a_target;
1169 	} */ *ap = v;
1170 
1171 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1172 	return (EROFS);
1173 }
1174 
1175 int
1176 kernfs_getpages(void *v)
1177 {
1178 	struct vop_getpages_args /* {
1179 		struct vnode *a_vp;
1180 		voff_t a_offset;
1181 		struct vm_page **a_m;
1182 		int *a_count;
1183 		int a_centeridx;
1184 		vm_prot_t a_access_type;
1185 		int a_advice;
1186 		int a_flags;
1187 	} */ *ap = v;
1188 
1189 	if ((ap->a_flags & PGO_LOCKED) == 0)
1190 		mutex_exit(ap->a_vp->v_interlock);
1191 
1192 	return (EFAULT);
1193 }
1194