xref: /netbsd-src/sys/miscfs/kernfs/kernfs_vnops.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: kernfs_vnops.c,v 1.151 2014/04/08 17:56:10 christos Exp $	*/
2 
3 /*
4  * Copyright (c) 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software donated to Berkeley by
8  * Jan-Simon Pendry.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)kernfs_vnops.c	8.15 (Berkeley) 5/21/95
35  */
36 
37 /*
38  * Kernel parameter filesystem (/kern)
39  */
40 
41 #include <sys/cdefs.h>
42 __KERNEL_RCSID(0, "$NetBSD: kernfs_vnops.c,v 1.151 2014/04/08 17:56:10 christos 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 #define	kernfs_putpages	genfs_putpages
176 
177 static int	kernfs_xread(struct kernfs_node *, int, char **,
178 				size_t, size_t *);
179 static int	kernfs_xwrite(const struct kernfs_node *, char *, size_t);
180 
181 int (**kernfs_vnodeop_p)(void *);
182 const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = {
183 	{ &vop_default_desc, vn_default_error },
184 	{ &vop_lookup_desc, kernfs_lookup },		/* lookup */
185 	{ &vop_create_desc, kernfs_create },		/* create */
186 	{ &vop_mknod_desc, kernfs_mknod },		/* mknod */
187 	{ &vop_open_desc, kernfs_open },		/* open */
188 	{ &vop_close_desc, kernfs_close },		/* close */
189 	{ &vop_access_desc, kernfs_access },		/* access */
190 	{ &vop_getattr_desc, kernfs_getattr },		/* getattr */
191 	{ &vop_setattr_desc, kernfs_setattr },		/* setattr */
192 	{ &vop_read_desc, kernfs_read },		/* read */
193 	{ &vop_write_desc, kernfs_write },		/* write */
194 	{ &vop_fcntl_desc, kernfs_fcntl },		/* fcntl */
195 	{ &vop_ioctl_desc, kernfs_ioctl },		/* ioctl */
196 	{ &vop_poll_desc, kernfs_poll },		/* poll */
197 	{ &vop_revoke_desc, kernfs_revoke },		/* revoke */
198 	{ &vop_fsync_desc, kernfs_fsync },		/* fsync */
199 	{ &vop_seek_desc, kernfs_seek },		/* seek */
200 	{ &vop_remove_desc, kernfs_remove },		/* remove */
201 	{ &vop_link_desc, kernfs_link },		/* link */
202 	{ &vop_rename_desc, kernfs_rename },		/* rename */
203 	{ &vop_mkdir_desc, kernfs_mkdir },		/* mkdir */
204 	{ &vop_rmdir_desc, kernfs_rmdir },		/* rmdir */
205 	{ &vop_symlink_desc, kernfs_symlink },		/* symlink */
206 	{ &vop_readdir_desc, kernfs_readdir },		/* readdir */
207 	{ &vop_readlink_desc, kernfs_readlink },	/* readlink */
208 	{ &vop_abortop_desc, kernfs_abortop },		/* abortop */
209 	{ &vop_inactive_desc, kernfs_inactive },	/* inactive */
210 	{ &vop_reclaim_desc, kernfs_reclaim },		/* reclaim */
211 	{ &vop_lock_desc, kernfs_lock },		/* lock */
212 	{ &vop_unlock_desc, kernfs_unlock },		/* unlock */
213 	{ &vop_bmap_desc, kernfs_bmap },		/* bmap */
214 	{ &vop_strategy_desc, kernfs_strategy },	/* strategy */
215 	{ &vop_print_desc, kernfs_print },		/* print */
216 	{ &vop_islocked_desc, kernfs_islocked },	/* islocked */
217 	{ &vop_pathconf_desc, kernfs_pathconf },	/* pathconf */
218 	{ &vop_advlock_desc, kernfs_advlock },		/* advlock */
219 	{ &vop_bwrite_desc, kernfs_bwrite },		/* bwrite */
220 	{ &vop_putpages_desc, kernfs_putpages },	/* putpages */
221 	{ NULL, NULL }
222 };
223 const struct vnodeopv_desc kernfs_vnodeop_opv_desc =
224 	{ &kernfs_vnodeop_p, kernfs_vnodeop_entries };
225 
226 static inline int
227 kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b)
228 {
229 	if (a->kf_type < b->kf_type)
230 		return -1;
231 	if (a->kf_type > b->kf_type)
232 		return 1;
233 	if (a->kf_fileop < b->kf_fileop)
234 		return -1;
235 	if (a->kf_fileop > b->kf_fileop)
236 		return 1;
237 	return (0);
238 }
239 
240 SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree =
241 	SPLAY_INITIALIZER(kfsfileoptree);
242 SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
243 SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare);
244 
245 kfstype
246 kernfs_alloctype(int nkf, const struct kernfs_fileop *kf)
247 {
248 	static u_char nextfreetype = KFSlasttype;
249 	struct kernfs_fileop *dkf, *fkf, skf;
250 	int i;
251 
252 	/* XXX need to keep track of dkf's memory if we support
253            deallocating types */
254 	dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK);
255 	memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops));
256 
257 	for (i = 0; i < sizeof(kernfs_default_fileops) /
258 		     sizeof(kernfs_default_fileops[0]); i++) {
259 		dkf[i].kf_type = nextfreetype;
260 		SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]);
261 	}
262 
263 	for (i = 0; i < nkf; i++) {
264 		skf.kf_type = nextfreetype;
265 		skf.kf_fileop = kf[i].kf_fileop;
266 		if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
267 			fkf->kf_vop = kf[i].kf_vop;
268 	}
269 
270 	return nextfreetype++;
271 }
272 
273 int
274 kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error)
275 {
276 	struct kernfs_fileop *kf, skf;
277 
278 	skf.kf_type = type;
279 	skf.kf_fileop = fileop;
280 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
281 		if (kf->kf_vop)
282 			return kf->kf_vop(v);
283 	return error;
284 }
285 
286 int
287 kernfs_try_xread(kfstype type, const struct kernfs_node *kfs, char **bfp,
288     size_t len, int error)
289 {
290 	struct kernfs_fileop *kf, skf;
291 
292 	skf.kf_type = type;
293 	skf.kf_fileop = KERNFS_XREAD;
294 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
295 		if (kf->kf_xread)
296 			return kf->kf_xread(kfs, bfp, len);
297 	return error;
298 }
299 
300 int
301 kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf,
302     size_t len, int error)
303 {
304 	struct kernfs_fileop *kf, skf;
305 
306 	skf.kf_type = type;
307 	skf.kf_fileop = KERNFS_XWRITE;
308 	if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf)))
309 		if (kf->kf_xwrite)
310 			return kf->kf_xwrite(kfs, bf, len);
311 	return error;
312 }
313 
314 int
315 kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt)
316 {
317 	struct kernfs_subdir *ks, *parent;
318 
319 	if (pkt == NULL) {
320 		SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue);
321 		nkern_targets++;
322 		if (dkt->dkt_kt.kt_vtype == VDIR)
323 			nkern_dirs++;
324 	} else {
325 		parent = (struct kernfs_subdir *)pkt->kt_data;
326 		SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue);
327 		parent->ks_nentries++;
328 		if (dkt->dkt_kt.kt_vtype == VDIR)
329 			parent->ks_dirs++;
330 	}
331 	if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) {
332 		ks = malloc(sizeof(struct kernfs_subdir),
333 		    M_TEMP, M_WAITOK);
334 		SIMPLEQ_INIT(&ks->ks_entries);
335 		ks->ks_nentries = 2; /* . and .. */
336 		ks->ks_dirs = 2;
337 		ks->ks_parent = pkt ? pkt : &kern_targets[0];
338 		dkt->dkt_kt.kt_data = ks;
339 	}
340 	return 0;
341 }
342 
343 static int
344 kernfs_xread(struct kernfs_node *kfs, int off, char **bufp, size_t len, size_t *wrlen)
345 {
346 	const struct kern_target *kt;
347 	int err;
348 
349 	kt = kfs->kfs_kt;
350 
351 	switch (kfs->kfs_type) {
352 	case KFStime: {
353 		struct timeval tv;
354 
355 		microtime(&tv);
356 		snprintf(*bufp, len, "%lld %ld\n", (long long)tv.tv_sec,
357 		    (long)tv.tv_usec);
358 		break;
359 	}
360 
361 	case KFSint: {
362 		int *ip = kt->kt_data;
363 
364 		snprintf(*bufp, len, "%d\n", *ip);
365 		break;
366 	}
367 
368 	case KFSstring: {
369 		char *cp = kt->kt_data;
370 
371 		*bufp = cp;
372 		break;
373 	}
374 
375 	case KFSmsgbuf: {
376 		long n;
377 
378 		/*
379 		 * deal with cases where the message buffer has
380 		 * become corrupted.
381 		 */
382 		if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) {
383 			msgbufenabled = 0;
384 			return (ENXIO);
385 		}
386 
387 		/*
388 		 * Note that reads of /kern/msgbuf won't necessarily yield
389 		 * consistent results, if the message buffer is modified
390 		 * while the read is in progress.  The worst that can happen
391 		 * is that incorrect data will be read.  There's no way
392 		 * that this can crash the system unless the values in the
393 		 * message buffer header are corrupted, but that'll cause
394 		 * the system to die anyway.
395 		 */
396 		if (off >= msgbufp->msg_bufs) {
397 			*wrlen = 0;
398 			return (0);
399 		}
400 		n = msgbufp->msg_bufx + off;
401 		if (n >= msgbufp->msg_bufs)
402 			n -= msgbufp->msg_bufs;
403 		len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off);
404 		*bufp = msgbufp->msg_bufc + n;
405 		*wrlen = len;
406 		return (0);
407 	}
408 
409 	case KFShostname: {
410 		char *cp = hostname;
411 		size_t xlen = hostnamelen;
412 
413 		if (xlen >= (len - 2))
414 			return (EINVAL);
415 
416 		memcpy(*bufp, cp, xlen);
417 		(*bufp)[xlen] = '\n';
418 		(*bufp)[xlen+1] = '\0';
419 		break;
420 	}
421 
422 	case KFSavenrun:
423 		averunnable.fscale = FSCALE;
424 		snprintf(*bufp, len, "%d %d %d %ld\n",
425 		    averunnable.ldavg[0], averunnable.ldavg[1],
426 		    averunnable.ldavg[2], averunnable.fscale);
427 		break;
428 
429 	default:
430 		err = kernfs_try_xread(kfs->kfs_type, kfs, bufp, len,
431 		    EOPNOTSUPP);
432 		if (err)
433 			return err;
434 	}
435 
436 	len = strlen(*bufp);
437 	if (len <= off)
438 		*wrlen = 0;
439 	else {
440 		*bufp += off;
441 		*wrlen = len - off;
442 	}
443 	return (0);
444 }
445 
446 static int
447 kernfs_xwrite(const struct kernfs_node *kfs, char *bf, size_t len)
448 {
449 
450 	switch (kfs->kfs_type) {
451 	case KFShostname:
452 		if (bf[len-1] == '\n')
453 			--len;
454 		memcpy(hostname, bf, len);
455 		hostname[len] = '\0';
456 		hostnamelen = (size_t) len;
457 		return (0);
458 
459 	default:
460 		return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO);
461 	}
462 }
463 
464 
465 /*
466  * vp is the current namei directory
467  * ndp is the name to locate in that directory...
468  */
469 int
470 kernfs_lookup(void *v)
471 {
472 	struct vop_lookup_v2_args /* {
473 		struct vnode * a_dvp;
474 		struct vnode ** a_vpp;
475 		struct componentname * a_cnp;
476 	} */ *ap = v;
477 	struct componentname *cnp = ap->a_cnp;
478 	struct vnode **vpp = ap->a_vpp;
479 	struct vnode *dvp = ap->a_dvp;
480 	const char *pname = cnp->cn_nameptr;
481 	const struct kernfs_node *kfs;
482 	const struct kern_target *kt;
483 	const struct dyn_kern_target *dkt;
484 	const struct kernfs_subdir *ks;
485 	int error, i;
486 
487 	*vpp = NULLVP;
488 
489 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
490 		return (EROFS);
491 
492 	if (cnp->cn_namelen == 1 && *pname == '.') {
493 		*vpp = dvp;
494 		vref(dvp);
495 		return (0);
496 	}
497 
498 	kfs = VTOKERN(dvp);
499 	switch (kfs->kfs_type) {
500 	case KFSkern:
501 		/*
502 		 * Shouldn't get here with .. in the root node.
503 		 */
504 		if (cnp->cn_flags & ISDOTDOT)
505 			return (EIO);
506 
507 		for (i = 0; i < static_nkern_targets; i++) {
508 			kt = &kern_targets[i];
509 			if (cnp->cn_namelen == kt->kt_namlen &&
510 			    memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0)
511 				goto found;
512 		}
513 		SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) {
514 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
515 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
516 				kt = &dkt->dkt_kt;
517 				goto found;
518 			}
519 		}
520 		break;
521 
522 	found:
523 		error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0);
524 		if (error)
525 			return error;
526 		VOP_UNLOCK(*vpp);
527 		return 0;
528 
529 	case KFSsubdir:
530 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
531 		if (cnp->cn_flags & ISDOTDOT) {
532 			kt = ks->ks_parent;
533 			goto found;
534 		}
535 
536 		SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) {
537 			if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen &&
538 			    memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) {
539 				kt = &dkt->dkt_kt;
540 				goto found;
541 			}
542 		}
543 		break;
544 
545 	default:
546 		return (ENOTDIR);
547 	}
548 
549 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
550 }
551 
552 int
553 kernfs_open(void *v)
554 {
555 	struct vop_open_args /* {
556 		struct vnode *a_vp;
557 		int a_mode;
558 		kauth_cred_t a_cred;
559 	} */ *ap = v;
560 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
561 
562 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN, v, 0);
563 }
564 
565 int
566 kernfs_close(void *v)
567 {
568 	struct vop_close_args /* {
569 		struct vnode *a_vp;
570 		int a_fflag;
571 		kauth_cred_t a_cred;
572 	} */ *ap = v;
573 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
574 
575 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE, v, 0);
576 }
577 
578 int
579 kernfs_access(void *v)
580 {
581 	struct vop_access_args /* {
582 		struct vnode *a_vp;
583 		int a_mode;
584 		kauth_cred_t a_cred;
585 	} */ *ap = v;
586 	struct vattr va;
587 	int error;
588 
589 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred)) != 0)
590 		return (error);
591 
592 	return kauth_authorize_vnode(ap->a_cred,
593 	    KAUTH_ACCESS_ACTION(ap->a_mode, ap->a_vp->v_type, va.va_mode),
594 	    ap->a_vp, NULL, genfs_can_access(va.va_type, va.va_mode,
595 	    va.va_uid, va.va_gid, ap->a_mode, ap->a_cred));
596 }
597 
598 static int
599 kernfs_default_fileop_getattr(void *v)
600 {
601 	struct vop_getattr_args /* {
602 		struct vnode *a_vp;
603 		struct vattr *a_vap;
604 		kauth_cred_t a_cred;
605 	} */ *ap = v;
606 	struct vattr *vap = ap->a_vap;
607 
608 	vap->va_nlink = 1;
609 	vap->va_bytes = vap->va_size = 0;
610 
611 	return 0;
612 }
613 
614 int
615 kernfs_getattr(void *v)
616 {
617 	struct vop_getattr_args /* {
618 		struct vnode *a_vp;
619 		struct vattr *a_vap;
620 		kauth_cred_t a_cred;
621 	} */ *ap = v;
622 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
623 	struct kernfs_subdir *ks;
624 	struct vattr *vap = ap->a_vap;
625 	int error = 0;
626 	char strbuf[KSTRING], *bf;
627 	size_t nread, total;
628 
629 	vattr_null(vap);
630 	vap->va_type = ap->a_vp->v_type;
631 	vap->va_uid = 0;
632 	vap->va_gid = 0;
633 	vap->va_mode = kfs->kfs_mode;
634 	vap->va_fileid = kfs->kfs_fileno;
635 	vap->va_flags = 0;
636 	vap->va_size = 0;
637 	vap->va_blocksize = DEV_BSIZE;
638 	/* Make all times be current TOD, except for the "boottime" node. */
639 	if (kfs->kfs_kt->kt_namlen == 8 &&
640 	    !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) {
641 		vap->va_ctime = boottime;
642 	} else {
643 		getnanotime(&vap->va_ctime);
644 	}
645 	vap->va_atime = vap->va_mtime = vap->va_ctime;
646 	vap->va_gen = 0;
647 	vap->va_flags = 0;
648 	vap->va_rdev = 0;
649 	vap->va_bytes = 0;
650 
651 	switch (kfs->kfs_type) {
652 	case KFSkern:
653 		vap->va_nlink = nkern_dirs;
654 		vap->va_bytes = vap->va_size = DEV_BSIZE;
655 		break;
656 
657 	case KFSdevice:
658 		vap->va_nlink = 1;
659 		vap->va_rdev = ap->a_vp->v_rdev;
660 		break;
661 
662 	case KFSroot:
663 		vap->va_nlink = 1;
664 		vap->va_bytes = vap->va_size = DEV_BSIZE;
665 		break;
666 
667 	case KFSsubdir:
668 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
669 		vap->va_nlink = ks->ks_dirs;
670 		vap->va_bytes = vap->va_size = DEV_BSIZE;
671 		break;
672 
673 	case KFSnull:
674 	case KFStime:
675 	case KFSint:
676 	case KFSstring:
677 	case KFShostname:
678 	case KFSavenrun:
679 	case KFSmsgbuf:
680 		vap->va_nlink = 1;
681 		total = 0;
682 		do {
683 			bf = strbuf;
684 			error = kernfs_xread(kfs, total, &bf,
685 			    sizeof(strbuf), &nread);
686 			total += nread;
687 		} while (error == 0 && nread != 0);
688 		vap->va_bytes = vap->va_size = total;
689 		break;
690 
691 	default:
692 		error = kernfs_try_fileop(kfs->kfs_type,
693 		    KERNFS_FILEOP_GETATTR, v, EINVAL);
694 		break;
695 	}
696 
697 	return (error);
698 }
699 
700 /*ARGSUSED*/
701 int
702 kernfs_setattr(void *v)
703 {
704 
705 	/*
706 	 * Silently ignore attribute changes.
707 	 * This allows for open with truncate to have no
708 	 * effect until some data is written.  I want to
709 	 * do it this way because all writes are atomic.
710 	 */
711 	return (0);
712 }
713 
714 int
715 kernfs_default_xread(void *v)
716 {
717 	struct vop_read_args /* {
718 		struct vnode *a_vp;
719 		struct uio *a_uio;
720 		int  a_ioflag;
721 		kauth_cred_t a_cred;
722 	} */ *ap = v;
723 	struct uio *uio = ap->a_uio;
724 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
725 	char strbuf[KSTRING], *bf;
726 	int off;
727 	size_t len;
728 	int error;
729 
730 	if (ap->a_vp->v_type == VDIR)
731 		return EISDIR;
732 
733 	off = (int)uio->uio_offset;
734 	/* Don't allow negative offsets */
735 	if (off < 0)
736 		return EINVAL;
737 
738 	bf = strbuf;
739 	if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0)
740 		error = uiomove(bf, len, uio);
741 	return (error);
742 }
743 
744 int
745 kernfs_read(void *v)
746 {
747 	struct vop_read_args /* {
748 		struct vnode *a_vp;
749 		struct uio *a_uio;
750 		int  a_ioflag;
751 		struct ucred *a_cred;
752 	} */ *ap = v;
753 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
754 
755 	if (kfs->kfs_type < KFSlasttype) {
756 		/* use default function */
757 		return kernfs_default_xread(v);
758 	}
759 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v,
760 	   EOPNOTSUPP);
761 }
762 
763 static int
764 kernfs_default_xwrite(void *v)
765 {
766 	struct vop_write_args /* {
767 		struct vnode *a_vp;
768 		struct uio *a_uio;
769 		int  a_ioflag;
770 		kauth_cred_t a_cred;
771 	} */ *ap = v;
772 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
773 	struct uio *uio = ap->a_uio;
774 	int error;
775 	size_t xlen;
776 	char strbuf[KSTRING];
777 
778 	if (uio->uio_offset != 0)
779 		return (EINVAL);
780 
781 	xlen = min(uio->uio_resid, KSTRING-1);
782 	if ((error = uiomove(strbuf, xlen, uio)) != 0)
783 		return (error);
784 
785 	if (uio->uio_resid != 0)
786 		return (EIO);
787 
788 	strbuf[xlen] = '\0';
789 	xlen = strlen(strbuf);
790 	return (kernfs_xwrite(kfs, strbuf, xlen));
791 }
792 
793 int
794 kernfs_write(void *v)
795 {
796 	struct vop_write_args /* {
797 		struct vnode *a_vp;
798 		struct uio *a_uio;
799 		int  a_ioflag;
800 		kauth_cred_t a_cred;
801 	} */ *ap = v;
802 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
803 
804 	if (kfs->kfs_type < KFSlasttype) {
805 		/* use default function */
806 		return kernfs_default_xwrite(v);
807 	}
808 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v,
809 	    EOPNOTSUPP);
810 }
811 
812 int
813 kernfs_ioctl(void *v)
814 {
815 	struct vop_ioctl_args /* {
816 		const struct vnodeop_desc *a_desc;
817 		struct vnode *a_vp;
818 		u_long a_command;
819 		void *a_data;
820 		int a_fflag;
821 		kauth_cred_t a_cred;
822 	} */ *ap = v;
823 	struct kernfs_node *kfs = VTOKERN(ap->a_vp);
824 
825 	return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v,
826 	    EPASSTHROUGH);
827 }
828 
829 static int
830 kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt,
831     u_int32_t value, struct vop_readdir_args *ap)
832 {
833 	struct kernfs_node *kfs;
834 	struct vnode *vp;
835 	int error;
836 
837 	if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt,
838 	    value)) != 0)
839 		return error;
840 	kfs = VTOKERN(vp);
841 	d->d_fileno = kfs->kfs_fileno;
842 	vput(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, 0, 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 = min(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 (!msgbufenabled
948 				    || msgbufp->msg_magic != MSG_MAGIC) {
949 					continue;
950 				}
951 			}
952 			d.d_namlen = kt->kt_namlen;
953 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
954 			    &kern_targets[0], kt, ap)) != 0)
955 				break;
956 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
957 			d.d_type = kt->kt_type;
958 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
959 				break;
960 			if (cookies)
961 				*cookies++ = i + 1;
962 			n++;
963 		}
964 		ncookies = n;
965 		break;
966 
967 	case KFSroot:
968 		if (i >= 2)
969 			return 0;
970 
971 		if (ap->a_ncookies) {
972 			ncookies = min(ncookies, (2 - i));
973 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
974 			    M_WAITOK);
975 			*ap->a_cookies = cookies;
976 		}
977 
978 		n = 0;
979 		for (; i < 2 && uio->uio_resid >= UIO_MX; i++) {
980 			kt = &kern_targets[i];
981 			d.d_namlen = kt->kt_namlen;
982 			d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0);
983 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
984 			d.d_type = kt->kt_type;
985 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
986 				break;
987 			if (cookies)
988 				*cookies++ = i + 1;
989 			n++;
990 		}
991 		ncookies = n;
992 		break;
993 
994 	case KFSsubdir:
995 		ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data;
996 		if (i >= ks->ks_nentries)
997 			return (0);
998 
999 		if (ap->a_ncookies) {
1000 			ncookies = min(ncookies, (ks->ks_nentries - i));
1001 			cookies = malloc(ncookies * sizeof(off_t), M_TEMP,
1002 			    M_WAITOK);
1003 			*ap->a_cookies = cookies;
1004 		}
1005 
1006 		dkt = SIMPLEQ_FIRST(&ks->ks_entries);
1007 		for (j = 0; j < i && dkt != NULL; j++)
1008 			dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1009 		n = 0;
1010 		for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) {
1011 			if (i < 2)
1012 				kt = &subdir_targets[i];
1013 			else {
1014 				/* check if ks_nentries lied to us */
1015 				if (dkt == NULL)
1016 					break;
1017 				kt = &dkt->dkt_kt;
1018 				dkt = SIMPLEQ_NEXT(dkt, dkt_queue);
1019 			}
1020 			if (kt->kt_tag == KFSdevice) {
1021 				dev_t *dp = kt->kt_data;
1022 				struct vnode *fvp;
1023 
1024 				if (*dp == NODEV ||
1025 				    !vfinddev(*dp, kt->kt_vtype, &fvp))
1026 					continue;
1027 				vrele(fvp);
1028 			}
1029 			d.d_namlen = kt->kt_namlen;
1030 			if ((error = kernfs_setdirentfileno(&d, i, kfs,
1031 			    ks->ks_parent, kt, ap)) != 0)
1032 				break;
1033 			memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1);
1034 			d.d_type = kt->kt_type;
1035 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
1036 				break;
1037 			if (cookies)
1038 				*cookies++ = i + 1;
1039 			n++;
1040 		}
1041 		ncookies = n;
1042 		break;
1043 
1044 	default:
1045 		error = ENOTDIR;
1046 		break;
1047 	}
1048 
1049 	if (ap->a_ncookies) {
1050 		if (error) {
1051 			if (cookies)
1052 				free(*ap->a_cookies, M_TEMP);
1053 			*ap->a_ncookies = 0;
1054 			*ap->a_cookies = NULL;
1055 		} else
1056 			*ap->a_ncookies = ncookies;
1057 	}
1058 
1059 	uio->uio_offset = i;
1060 	return (error);
1061 }
1062 
1063 int
1064 kernfs_inactive(void *v)
1065 {
1066 	struct vop_inactive_args /* {
1067 		struct vnode *a_vp;
1068 		bool *a_recycle;
1069 	} */ *ap = v;
1070 	struct vnode *vp = ap->a_vp;
1071 
1072 	*ap->a_recycle = false;
1073 	VOP_UNLOCK(vp);
1074 	return (0);
1075 }
1076 
1077 int
1078 kernfs_reclaim(void *v)
1079 {
1080 	struct vop_reclaim_args /* {
1081 		struct vnode *a_vp;
1082 	} */ *ap = v;
1083 
1084 	return (kernfs_freevp(ap->a_vp));
1085 }
1086 
1087 /*
1088  * Return POSIX pathconf information applicable to special devices.
1089  */
1090 int
1091 kernfs_pathconf(void *v)
1092 {
1093 	struct vop_pathconf_args /* {
1094 		struct vnode *a_vp;
1095 		int a_name;
1096 		register_t *a_retval;
1097 	} */ *ap = v;
1098 
1099 	switch (ap->a_name) {
1100 	case _PC_LINK_MAX:
1101 		*ap->a_retval = LINK_MAX;
1102 		return (0);
1103 	case _PC_MAX_CANON:
1104 		*ap->a_retval = MAX_CANON;
1105 		return (0);
1106 	case _PC_MAX_INPUT:
1107 		*ap->a_retval = MAX_INPUT;
1108 		return (0);
1109 	case _PC_PIPE_BUF:
1110 		*ap->a_retval = PIPE_BUF;
1111 		return (0);
1112 	case _PC_CHOWN_RESTRICTED:
1113 		*ap->a_retval = 1;
1114 		return (0);
1115 	case _PC_VDISABLE:
1116 		*ap->a_retval = _POSIX_VDISABLE;
1117 		return (0);
1118 	case _PC_SYNC_IO:
1119 		*ap->a_retval = 1;
1120 		return (0);
1121 	default:
1122 		return (EINVAL);
1123 	}
1124 	/* NOTREACHED */
1125 }
1126 
1127 /*
1128  * Print out the contents of a /dev/fd vnode.
1129  */
1130 /* ARGSUSED */
1131 int
1132 kernfs_print(void *v)
1133 {
1134 
1135 	printf("tag VT_KERNFS, kernfs vnode\n");
1136 	return (0);
1137 }
1138 
1139 int
1140 kernfs_link(void *v)
1141 {
1142 	struct vop_link_args /* {
1143 		struct vnode *a_dvp;
1144 		struct vnode *a_vp;
1145 		struct componentname *a_cnp;
1146 	} */ *ap = v;
1147 
1148 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1149 	vput(ap->a_dvp);
1150 	return (EROFS);
1151 }
1152 
1153 int
1154 kernfs_symlink(void *v)
1155 {
1156 	struct vop_symlink_v3_args /* {
1157 		struct vnode *a_dvp;
1158 		struct vnode **a_vpp;
1159 		struct componentname *a_cnp;
1160 		struct vattr *a_vap;
1161 		char *a_target;
1162 	} */ *ap = v;
1163 
1164 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1165 	return (EROFS);
1166 }
1167