xref: /netbsd-src/sys/kern/vfs_vnops.c (revision 5b84b3983f71fd20a534cfa5d1556623a8aaa717)
1 /*	$NetBSD: vfs_vnops.c,v 1.94 2005/07/23 18:19:51 erh Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_vnops.c	8.14 (Berkeley) 6/15/95
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.94 2005/07/23 18:19:51 erh Exp $");
41 
42 #include "opt_verified_exec.h"
43 
44 #include "fs_union.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/file.h>
50 #include <sys/stat.h>
51 #include <sys/buf.h>
52 #include <sys/proc.h>
53 #include <sys/malloc.h>
54 #include <sys/mount.h>
55 #include <sys/namei.h>
56 #include <sys/vnode.h>
57 #include <sys/ioctl.h>
58 #include <sys/tty.h>
59 #include <sys/poll.h>
60 
61 #include <miscfs/specfs/specdev.h>
62 
63 #include <uvm/uvm_extern.h>
64 
65 #ifdef UNION
66 #include <fs/union/union.h>
67 #endif
68 
69 #if defined(LKM) || defined(UNION)
70 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *);
71 #endif
72 
73 #ifdef VERIFIED_EXEC
74 #include <sys/verified_exec.h>
75 #endif
76 
77 static int vn_read(struct file *fp, off_t *offset, struct uio *uio,
78 	    struct ucred *cred, int flags);
79 static int vn_write(struct file *fp, off_t *offset, struct uio *uio,
80 	    struct ucred *cred, int flags);
81 static int vn_closefile(struct file *fp, struct proc *p);
82 static int vn_poll(struct file *fp, int events, struct proc *p);
83 static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p);
84 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p);
85 static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p);
86 
87 const struct fileops vnops = {
88 	vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
89 	vn_statfile, vn_closefile, vn_kqfilter
90 };
91 
92 /*
93  * Common code for vnode open operations.
94  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
95  */
96 int
97 vn_open(struct nameidata *ndp, int fmode, int cmode)
98 {
99 	struct vnode *vp;
100 	struct mount *mp;
101 	struct proc *p = ndp->ni_cnd.cn_proc;
102 	struct ucred *cred = p->p_ucred;
103 	struct vattr va;
104 	int error;
105 #ifdef VERIFIED_EXEC
106 	struct veriexec_hash_entry *vhe = NULL;
107 #endif /* VERIFIED_EXEC */
108 
109 restart:
110 	if (fmode & O_CREAT) {
111 		ndp->ni_cnd.cn_nameiop = CREATE;
112 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
113 		if ((fmode & O_EXCL) == 0 &&
114 		    ((fmode & O_NOFOLLOW) == 0))
115 			ndp->ni_cnd.cn_flags |= FOLLOW;
116 		if ((error = namei(ndp)) != 0)
117 			return (error);
118 		if (ndp->ni_vp == NULL) {
119 #ifdef VERIFIED_EXEC
120 			/* Lockdown mode: Prevent creation of new files. */
121 			if (veriexec_strict >= 3) {
122 				VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
123 
124 				printf("Veriexec: vn_open: Preventing "
125 				       "new file creation in %s.\n",
126 				       ndp->ni_dirp);
127 
128 				vp = ndp->ni_dvp;
129 				error = EPERM;
130 				goto bad;
131 			}
132 #endif /* VERIFIED_EXEC */
133 
134 			VATTR_NULL(&va);
135 			va.va_type = VREG;
136 			va.va_mode = cmode;
137 			if (fmode & O_EXCL)
138 				 va.va_vaflags |= VA_EXCLUSIVE;
139 			if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
140 				VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
141 				vput(ndp->ni_dvp);
142 				if ((error = vn_start_write(NULL, &mp,
143 				    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
144 					return (error);
145 				goto restart;
146 			}
147 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
148 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
149 					   &ndp->ni_cnd, &va);
150 			vn_finished_write(mp, 0);
151 			if (error)
152 				return (error);
153 			fmode &= ~O_TRUNC;
154 			vp = ndp->ni_vp;
155 		} else {
156 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
157 			if (ndp->ni_dvp == ndp->ni_vp)
158 				vrele(ndp->ni_dvp);
159 			else
160 				vput(ndp->ni_dvp);
161 			ndp->ni_dvp = NULL;
162 			vp = ndp->ni_vp;
163 			if (fmode & O_EXCL) {
164 				error = EEXIST;
165 				goto bad;
166 			}
167 			fmode &= ~O_CREAT;
168 		}
169 	} else {
170 		ndp->ni_cnd.cn_nameiop = LOOKUP;
171 		ndp->ni_cnd.cn_flags = LOCKLEAF;
172 		if ((fmode & O_NOFOLLOW) == 0)
173 			ndp->ni_cnd.cn_flags |= FOLLOW;
174 		if ((error = namei(ndp)) != 0)
175 			return (error);
176 		vp = ndp->ni_vp;
177 	}
178 	if (vp->v_type == VSOCK) {
179 		error = EOPNOTSUPP;
180 		goto bad;
181 	}
182 	if (ndp->ni_vp->v_type == VLNK) {
183 		error = EFTYPE;
184 		goto bad;
185 	}
186 
187 #ifdef VERIFIED_EXEC
188 	if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0)
189 		goto bad;
190 #endif
191 
192 	if ((fmode & O_CREAT) == 0) {
193 #ifdef VERIFIED_EXEC
194 		/* XXX may need pathbuf instead */
195 		if ((error = veriexec_verify(p, vp, &va, ndp->ni_dirp,
196 					     VERIEXEC_FILE, &vhe)) != 0)
197 			goto bad;
198 #endif
199 
200 		if (fmode & FREAD) {
201 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
202 				goto bad;
203 		}
204 
205 		if (fmode & (FWRITE | O_TRUNC)) {
206 			if (vp->v_type == VDIR) {
207 				error = EISDIR;
208 				goto bad;
209 			}
210 			if ((error = vn_writechk(vp)) != 0 ||
211 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
212 				goto bad;
213 #ifdef VERIFIED_EXEC
214 			if (vhe != NULL) {
215 				veriexec_report("Write access request.",
216 						ndp->ni_dirp, &va, p,
217 						REPORT_NOVERBOSE,
218 						REPORT_ALARM,
219 						REPORT_NOPANIC);
220 
221 				/* IPS mode: Deny writing to monitored files. */
222 				if (veriexec_strict >= 2) {
223 					error = EPERM;
224 					goto bad;
225 				} else {
226 					vhe->status = FINGERPRINT_NOTEVAL;
227 				}
228 			}
229 #endif
230 		}
231 	}
232 
233 	if (fmode & O_TRUNC) {
234 		VOP_UNLOCK(vp, 0);			/* XXX */
235 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
236 			vput(vp);
237 			return (error);
238 		}
239 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
240 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
241 		VATTR_NULL(&va);
242 		va.va_size = 0;
243 		error = VOP_SETATTR(vp, &va, cred, p);
244 		vn_finished_write(mp, 0);
245 		if (error != 0)
246 			goto bad;
247 	}
248 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
249 		goto bad;
250 	if (vp->v_type == VREG &&
251 	    uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
252 		error = EIO;
253 		goto bad;
254 	}
255 	if (fmode & FWRITE)
256 		vp->v_writecount++;
257 
258 	return (0);
259 bad:
260 	vput(vp);
261 	return (error);
262 }
263 
264 /*
265  * Check for write permissions on the specified vnode.
266  * Prototype text segments cannot be written.
267  */
268 int
269 vn_writechk(struct vnode *vp)
270 {
271 
272 	/*
273 	 * If the vnode is in use as a process's text,
274 	 * we can't allow writing.
275 	 */
276 	if (vp->v_flag & VTEXT)
277 		return (ETXTBSY);
278 	return (0);
279 }
280 
281 /*
282  * Mark a vnode as having executable mappings.
283  */
284 void
285 vn_markexec(struct vnode *vp)
286 {
287 	if ((vp->v_flag & VEXECMAP) == 0) {
288 		uvmexp.filepages -= vp->v_uobj.uo_npages;
289 		uvmexp.execpages += vp->v_uobj.uo_npages;
290 	}
291 	vp->v_flag |= VEXECMAP;
292 }
293 
294 /*
295  * Mark a vnode as being the text of a process.
296  * Fail if the vnode is currently writable.
297  */
298 int
299 vn_marktext(struct vnode *vp)
300 {
301 
302 	if (vp->v_writecount != 0) {
303 		KASSERT((vp->v_flag & VTEXT) == 0);
304 		return (ETXTBSY);
305 	}
306 	vp->v_flag |= VTEXT;
307 	vn_markexec(vp);
308 	return (0);
309 }
310 
311 /*
312  * Vnode close call
313  *
314  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
315  */
316 int
317 vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p)
318 {
319 	int error;
320 
321 	if (flags & FWRITE)
322 		vp->v_writecount--;
323 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
324 	error = VOP_CLOSE(vp, flags, cred, p);
325 	vput(vp);
326 	return (error);
327 }
328 
329 /*
330  * Package up an I/O request on a vnode into a uio and do it.
331  */
332 int
333 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset,
334     enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid,
335     struct proc *p)
336 {
337 	struct uio auio;
338 	struct iovec aiov;
339 	struct mount *mp;
340 	int error;
341 
342 	if ((ioflg & IO_NODELOCKED) == 0) {
343 		if (rw == UIO_READ) {
344 			vn_lock(vp, LK_SHARED | LK_RETRY);
345 		} else /* UIO_WRITE */ {
346 			if (vp->v_type != VCHR &&
347 			    (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH))
348 			    != 0)
349 				return (error);
350 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
351 		}
352 	}
353 	auio.uio_iov = &aiov;
354 	auio.uio_iovcnt = 1;
355 	aiov.iov_base = base;
356 	aiov.iov_len = len;
357 	auio.uio_resid = len;
358 	auio.uio_offset = offset;
359 	auio.uio_segflg = segflg;
360 	auio.uio_rw = rw;
361 	auio.uio_procp = p;
362 	if (rw == UIO_READ) {
363 		error = VOP_READ(vp, &auio, ioflg, cred);
364 	} else {
365 		error = VOP_WRITE(vp, &auio, ioflg, cred);
366 	}
367 	if (aresid)
368 		*aresid = auio.uio_resid;
369 	else
370 		if (auio.uio_resid && error == 0)
371 			error = EIO;
372 	if ((ioflg & IO_NODELOCKED) == 0) {
373 		if (rw == UIO_WRITE)
374 			vn_finished_write(mp, 0);
375 		VOP_UNLOCK(vp, 0);
376 	}
377 	return (error);
378 }
379 
380 int
381 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done,
382     struct proc *p, off_t **cookies, int *ncookies)
383 {
384 	struct vnode *vp = (struct vnode *)fp->f_data;
385 	struct iovec aiov;
386 	struct uio auio;
387 	int error, eofflag;
388 
389 unionread:
390 	if (vp->v_type != VDIR)
391 		return (EINVAL);
392 	aiov.iov_base = bf;
393 	aiov.iov_len = count;
394 	auio.uio_iov = &aiov;
395 	auio.uio_iovcnt = 1;
396 	auio.uio_rw = UIO_READ;
397 	auio.uio_segflg = segflg;
398 	auio.uio_procp = p;
399 	auio.uio_resid = count;
400 	vn_lock(vp, LK_SHARED | LK_RETRY);
401 	auio.uio_offset = fp->f_offset;
402 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
403 		    ncookies);
404 	fp->f_offset = auio.uio_offset;
405 	VOP_UNLOCK(vp, 0);
406 	if (error)
407 		return (error);
408 
409 #if defined(UNION) || defined(LKM)
410 	if (count == auio.uio_resid && vn_union_readdir_hook) {
411 		struct vnode *ovp = vp;
412 
413 		error = (*vn_union_readdir_hook)(&vp, fp, p);
414 		if (error)
415 			return (error);
416 		if (vp != ovp)
417 			goto unionread;
418 	}
419 #endif /* UNION || LKM */
420 
421 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
422 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
423 		struct vnode *tvp = vp;
424 		vp = vp->v_mount->mnt_vnodecovered;
425 		VREF(vp);
426 		fp->f_data = vp;
427 		fp->f_offset = 0;
428 		vrele(tvp);
429 		goto unionread;
430 	}
431 	*done = count - auio.uio_resid;
432 	return error;
433 }
434 
435 /*
436  * File table vnode read routine.
437  */
438 static int
439 vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred,
440     int flags)
441 {
442 	struct vnode *vp = (struct vnode *)fp->f_data;
443 	int count, error, ioflag = 0;
444 
445 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
446 	if (fp->f_flag & FNONBLOCK)
447 		ioflag |= IO_NDELAY;
448 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
449 		ioflag |= IO_SYNC;
450 	if (fp->f_flag & FALTIO)
451 		ioflag |= IO_ALTSEMANTICS;
452 	vn_lock(vp, LK_SHARED | LK_RETRY);
453 	uio->uio_offset = *offset;
454 	count = uio->uio_resid;
455 	error = VOP_READ(vp, uio, ioflag, cred);
456 	if (flags & FOF_UPDATE_OFFSET)
457 		*offset += count - uio->uio_resid;
458 	VOP_UNLOCK(vp, 0);
459 	return (error);
460 }
461 
462 /*
463  * File table vnode write routine.
464  */
465 static int
466 vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred,
467     int flags)
468 {
469 	struct vnode *vp = (struct vnode *)fp->f_data;
470 	struct mount *mp;
471 	int count, error, ioflag = IO_UNIT;
472 
473 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
474 		ioflag |= IO_APPEND;
475 	if (fp->f_flag & FNONBLOCK)
476 		ioflag |= IO_NDELAY;
477 	if (fp->f_flag & FFSYNC ||
478 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
479 		ioflag |= IO_SYNC;
480 	else if (fp->f_flag & FDSYNC)
481 		ioflag |= IO_DSYNC;
482 	if (fp->f_flag & FALTIO)
483 		ioflag |= IO_ALTSEMANTICS;
484 	mp = NULL;
485 	if (vp->v_type != VCHR &&
486 	    (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
487 		return (error);
488 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
489 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
490 	uio->uio_offset = *offset;
491 	count = uio->uio_resid;
492 	error = VOP_WRITE(vp, uio, ioflag, cred);
493 	if (flags & FOF_UPDATE_OFFSET) {
494 		if (ioflag & IO_APPEND)
495 			*offset = uio->uio_offset;
496 		else
497 			*offset += count - uio->uio_resid;
498 	}
499 	VOP_UNLOCK(vp, 0);
500 	vn_finished_write(mp, 0);
501 	return (error);
502 }
503 
504 /*
505  * File table vnode stat routine.
506  */
507 static int
508 vn_statfile(struct file *fp, struct stat *sb, struct proc *p)
509 {
510 	struct vnode *vp = (struct vnode *)fp->f_data;
511 
512 	return vn_stat(vp, sb, p);
513 }
514 
515 int
516 vn_stat(struct vnode *vp, struct stat *sb, struct proc *p)
517 {
518 	struct vattr va;
519 	int error;
520 	mode_t mode;
521 
522 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
523 	if (error)
524 		return (error);
525 	/*
526 	 * Copy from vattr table
527 	 */
528 	sb->st_dev = va.va_fsid;
529 	sb->st_ino = va.va_fileid;
530 	mode = va.va_mode;
531 	switch (vp->v_type) {
532 	case VREG:
533 		mode |= S_IFREG;
534 		break;
535 	case VDIR:
536 		mode |= S_IFDIR;
537 		break;
538 	case VBLK:
539 		mode |= S_IFBLK;
540 		break;
541 	case VCHR:
542 		mode |= S_IFCHR;
543 		break;
544 	case VLNK:
545 		mode |= S_IFLNK;
546 		break;
547 	case VSOCK:
548 		mode |= S_IFSOCK;
549 		break;
550 	case VFIFO:
551 		mode |= S_IFIFO;
552 		break;
553 	default:
554 		return (EBADF);
555 	};
556 	sb->st_mode = mode;
557 	sb->st_nlink = va.va_nlink;
558 	sb->st_uid = va.va_uid;
559 	sb->st_gid = va.va_gid;
560 	sb->st_rdev = va.va_rdev;
561 	sb->st_size = va.va_size;
562 	sb->st_atimespec = va.va_atime;
563 	sb->st_mtimespec = va.va_mtime;
564 	sb->st_ctimespec = va.va_ctime;
565 	sb->st_birthtimespec = va.va_birthtime;
566 	sb->st_blksize = va.va_blocksize;
567 	sb->st_flags = va.va_flags;
568 	sb->st_gen = 0;
569 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
570 	return (0);
571 }
572 
573 /*
574  * File table vnode fcntl routine.
575  */
576 static int
577 vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p)
578 {
579 	struct vnode *vp = ((struct vnode *)fp->f_data);
580 	int error;
581 
582 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
583 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
584 	VOP_UNLOCK(vp, 0);
585 	return (error);
586 }
587 
588 /*
589  * File table vnode ioctl routine.
590  */
591 static int
592 vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p)
593 {
594 	struct vnode *vp = ((struct vnode *)fp->f_data);
595 	struct vattr vattr;
596 	int error;
597 
598 	switch (vp->v_type) {
599 
600 	case VREG:
601 	case VDIR:
602 		if (com == FIONREAD) {
603 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
604 			if (error)
605 				return (error);
606 			*(int *)data = vattr.va_size - fp->f_offset;
607 			return (0);
608 		}
609 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
610 			/*
611 			 * Files don't have send queues, so there never
612 			 * are any bytes in them, nor is there any
613 			 * open space in them.
614 			 */
615 			*(int *)data = 0;
616 			return (0);
617 		}
618 		if (com == FIOGETBMAP) {
619 			daddr_t *block;
620 
621 			if (*(daddr_t *)data < 0)
622 				return (EINVAL);
623 			block = (daddr_t *)data;
624 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
625 		}
626 		if (com == OFIOGETBMAP) {
627 			daddr_t ibn, obn;
628 
629 			if (*(int32_t *)data < 0)
630 				return (EINVAL);
631 			ibn = (daddr_t)*(int32_t *)data;
632 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
633 			*(int32_t *)data = (int32_t)obn;
634 			return error;
635 		}
636 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
637 			return (0);			/* XXX */
638 		/* fall into ... */
639 	case VFIFO:
640 	case VCHR:
641 	case VBLK:
642 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
643 		if (error == 0 && com == TIOCSCTTY) {
644 			if (p->p_session->s_ttyvp)
645 				vrele(p->p_session->s_ttyvp);
646 			p->p_session->s_ttyvp = vp;
647 			VREF(vp);
648 		}
649 		return (error);
650 
651 	default:
652 		return (EPASSTHROUGH);
653 	}
654 }
655 
656 /*
657  * File table vnode poll routine.
658  */
659 static int
660 vn_poll(struct file *fp, int events, struct proc *p)
661 {
662 
663 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
664 }
665 
666 /*
667  * File table vnode kqfilter routine.
668  */
669 int
670 vn_kqfilter(struct file *fp, struct knote *kn)
671 {
672 
673 	return (VOP_KQFILTER((struct vnode *)fp->f_data, kn));
674 }
675 
676 /*
677  * Check that the vnode is still valid, and if so
678  * acquire requested lock.
679  */
680 int
681 vn_lock(struct vnode *vp, int flags)
682 {
683 	int error;
684 
685 #if 0
686 	KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
687 	    || (vp->v_flag & VONWORKLST) != 0);
688 #endif
689 
690 	do {
691 		if ((flags & LK_INTERLOCK) == 0)
692 			simple_lock(&vp->v_interlock);
693 		if (vp->v_flag & VXLOCK) {
694 			if (flags & LK_NOWAIT) {
695 				simple_unlock(&vp->v_interlock);
696 				return EBUSY;
697 			}
698 			vp->v_flag |= VXWANT;
699 			ltsleep(vp, PINOD | PNORELOCK,
700 			    "vn_lock", 0, &vp->v_interlock);
701 			error = ENOENT;
702 		} else {
703 			error = VOP_LOCK(vp,
704 			    (flags & ~LK_RETRY) | LK_INTERLOCK);
705 			if (error == 0 || error == EDEADLK || error == EBUSY)
706 				return (error);
707 		}
708 		flags &= ~LK_INTERLOCK;
709 	} while (flags & LK_RETRY);
710 	return (error);
711 }
712 
713 /*
714  * File table vnode close routine.
715  */
716 static int
717 vn_closefile(struct file *fp, struct proc *p)
718 {
719 
720 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
721 		fp->f_cred, p));
722 }
723 
724 /*
725  * Enable LK_CANRECURSE on lock. Return prior status.
726  */
727 u_int
728 vn_setrecurse(struct vnode *vp)
729 {
730 	struct lock *lkp = &vp->v_lock;
731 	u_int retval = lkp->lk_flags & LK_CANRECURSE;
732 
733 	lkp->lk_flags |= LK_CANRECURSE;
734 	return retval;
735 }
736 
737 /*
738  * Called when done with locksetrecurse.
739  */
740 void
741 vn_restorerecurse(struct vnode *vp, u_int flags)
742 {
743 	struct lock *lkp = &vp->v_lock;
744 
745 	lkp->lk_flags &= ~LK_CANRECURSE;
746 	lkp->lk_flags |= flags;
747 }
748 
749 int
750 vn_cow_establish(struct vnode *vp,
751     int (*func)(void *, struct buf *), void *cookie)
752 {
753 	int s;
754 	struct spec_cow_entry *e;
755 
756 	MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry),
757 	    M_DEVBUF, M_WAITOK);
758 	e->ce_func = func;
759 	e->ce_cookie = cookie;
760 
761 	SPEC_COW_LOCK(vp->v_specinfo, s);
762 	vp->v_spec_cow_req++;
763 	while (vp->v_spec_cow_count > 0)
764 		ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0,
765 		    &vp->v_spec_cow_slock);
766 
767 	SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list);
768 
769 	vp->v_spec_cow_req--;
770 	if (vp->v_spec_cow_req == 0)
771 		wakeup(&vp->v_spec_cow_req);
772 	SPEC_COW_UNLOCK(vp->v_specinfo, s);
773 
774 	return 0;
775 }
776 
777 int
778 vn_cow_disestablish(struct vnode *vp,
779     int (*func)(void *, struct buf *), void *cookie)
780 {
781 	int s;
782 	struct spec_cow_entry *e;
783 
784 	SPEC_COW_LOCK(vp->v_specinfo, s);
785 	vp->v_spec_cow_req++;
786 	while (vp->v_spec_cow_count > 0)
787 		ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0,
788 		    &vp->v_spec_cow_slock);
789 
790 	SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list)
791 		if (e->ce_func == func && e->ce_cookie == cookie) {
792 			SLIST_REMOVE(&vp->v_spec_cow_head, e,
793 			    spec_cow_entry, ce_list);
794 			FREE(e, M_DEVBUF);
795 			break;
796 		}
797 
798 	vp->v_spec_cow_req--;
799 	if (vp->v_spec_cow_req == 0)
800 		wakeup(&vp->v_spec_cow_req);
801 	SPEC_COW_UNLOCK(vp->v_specinfo, s);
802 
803 	return e ? 0 : EINVAL;
804 }
805 
806 /*
807  * Simplified in-kernel wrapper calls for extended attribute access.
808  * Both calls pass in a NULL credential, authorizing a "kernel" access.
809  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
810  */
811 int
812 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
813     const char *attrname, size_t *buflen, void *bf, struct proc *p)
814 {
815 	struct uio auio;
816 	struct iovec aiov;
817 	int error;
818 
819 	aiov.iov_len = *buflen;
820 	aiov.iov_base = bf;
821 
822 	auio.uio_iov = &aiov;
823 	auio.uio_iovcnt = 1;
824 	auio.uio_rw = UIO_READ;
825 	auio.uio_segflg = UIO_SYSSPACE;
826 	auio.uio_procp = p;
827 	auio.uio_offset = 0;
828 	auio.uio_resid = *buflen;
829 
830 	if ((ioflg & IO_NODELOCKED) == 0)
831 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
832 
833 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL,
834 	    p);
835 
836 	if ((ioflg & IO_NODELOCKED) == 0)
837 		VOP_UNLOCK(vp, 0);
838 
839 	if (error == 0)
840 		*buflen = *buflen - auio.uio_resid;
841 
842 	return (error);
843 }
844 
845 /*
846  * XXX Failure mode if partially written?
847  */
848 int
849 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
850     const char *attrname, size_t buflen, const void *bf, struct proc *p)
851 {
852 	struct uio auio;
853 	struct iovec aiov;
854 	struct mount *mp;
855 	int error;
856 
857 	aiov.iov_len = buflen;
858 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
859 
860 	auio.uio_iov = &aiov;
861 	auio.uio_iovcnt = 1;
862 	auio.uio_rw = UIO_WRITE;
863 	auio.uio_segflg = UIO_SYSSPACE;
864 	auio.uio_procp = p;
865 	auio.uio_offset = 0;
866 	auio.uio_resid = buflen;
867 
868 	if ((ioflg & IO_NODELOCKED) == 0) {
869 		if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
870 			return (error);
871 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
872 	}
873 
874 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p);
875 
876 	if ((ioflg & IO_NODELOCKED) == 0) {
877 		vn_finished_write(mp, 0);
878 		VOP_UNLOCK(vp, 0);
879 	}
880 
881 	return (error);
882 }
883 
884 int
885 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
886     const char *attrname, struct proc *p)
887 {
888 	struct mount *mp;
889 	int error;
890 
891 	if ((ioflg & IO_NODELOCKED) == 0) {
892 		if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
893 			return (error);
894 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
895 	}
896 
897 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p);
898 	if (error == EOPNOTSUPP)
899 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
900 		    NULL, p);
901 
902 	if ((ioflg & IO_NODELOCKED) == 0) {
903 		vn_finished_write(mp, 0);
904 		VOP_UNLOCK(vp, 0);
905 	}
906 
907 	return (error);
908 }
909