xref: /dragonfly/sys/kern/vfs_vnops.c (revision 43b4d1bd)
1 /*
2  * Copyright (c) 1982, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
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. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)vfs_vnops.c	8.2 (Berkeley) 1/21/94
39  * $FreeBSD: src/sys/kern/vfs_vnops.c,v 1.87.2.13 2002/12/29 18:19:53 dillon Exp $
40  * $DragonFly: src/sys/kern/vfs_vnops.c,v 1.22 2004/06/15 00:30:53 dillon Exp $
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/fcntl.h>
46 #include <sys/file.h>
47 #include <sys/stat.h>
48 #include <sys/proc.h>
49 #include <sys/mount.h>
50 #include <sys/namei.h>
51 #include <sys/vnode.h>
52 #include <sys/buf.h>
53 #include <sys/filio.h>
54 #include <sys/ttycom.h>
55 #include <sys/conf.h>
56 #include <sys/syslog.h>
57 
58 static int vn_closefile (struct file *fp, struct thread *td);
59 static int vn_ioctl (struct file *fp, u_long com, caddr_t data,
60 		struct thread *td);
61 static int vn_read (struct file *fp, struct uio *uio,
62 		struct ucred *cred, int flags, struct thread *td);
63 static int vn_poll (struct file *fp, int events, struct ucred *cred,
64 		struct thread *td);
65 static int vn_kqfilter (struct file *fp, struct knote *kn);
66 static int vn_statfile (struct file *fp, struct stat *sb, struct thread *td);
67 static int vn_write (struct file *fp, struct uio *uio,
68 		struct ucred *cred, int flags, struct thread *td);
69 
70 struct 	fileops vnops = {
71 	NULL,	/* port */
72 	NULL,	/* clone */
73 	vn_read, vn_write, vn_ioctl, vn_poll, vn_kqfilter,
74 	vn_statfile, vn_closefile
75 };
76 
77 /*
78  * Common code for vnode open operations.
79  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
80  *
81  * Note that this does NOT free nameidata for the successful case,
82  * due to the NDINIT being done elsewhere.
83  */
84 int
85 vn_open(ndp, fmode, cmode)
86 	struct nameidata *ndp;
87 	int fmode, cmode;
88 {
89 	struct vnode *vp;
90 	struct thread *td = ndp->ni_cnd.cn_td;
91 	struct ucred *cred = ndp->ni_cnd.cn_cred;
92 	struct vattr vat;
93 	struct vattr *vap = &vat;
94 	int mode, error;
95 
96 	KKASSERT(cred == td->td_proc->p_ucred);
97 
98 	if (fmode & O_CREAT) {
99 		ndp->ni_cnd.cn_nameiop = NAMEI_CREATE;
100 		ndp->ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF;
101 		if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
102 			ndp->ni_cnd.cn_flags |= CNP_FOLLOW;
103 		bwillwrite();
104 		error = namei(ndp);
105 		if (error)
106 			return (error);
107 		if (ndp->ni_vp == NULL) {
108 			VATTR_NULL(vap);
109 			vap->va_type = VREG;
110 			vap->va_mode = cmode;
111 			if (fmode & O_EXCL)
112 				vap->va_vaflags |= VA_EXCLUSIVE;
113 			VOP_LEASE(ndp->ni_dvp, td, cred, LEASE_WRITE);
114 			error = VOP_CREATE(ndp->ni_dvp, NCPNULL, &ndp->ni_vp,
115 					   &ndp->ni_cnd, vap);
116 			if (error) {
117 				NDFREE(ndp, NDF_ONLY_PNBUF);
118 				vput(ndp->ni_dvp);
119 				return (error);
120 			}
121 			vput(ndp->ni_dvp);
122 			ASSERT_VOP_UNLOCKED(ndp->ni_dvp, "create");
123 			ASSERT_VOP_LOCKED(ndp->ni_vp, "create");
124 			fmode &= ~O_TRUNC;
125 			vp = ndp->ni_vp;
126 		} else {
127 			if (ndp->ni_dvp == ndp->ni_vp)
128 				vrele(ndp->ni_dvp);
129 			else
130 				vput(ndp->ni_dvp);
131 			ndp->ni_dvp = NULL;
132 			vp = ndp->ni_vp;
133 			if (fmode & O_EXCL) {
134 				error = EEXIST;
135 				goto bad;
136 			}
137 			fmode &= ~O_CREAT;
138 		}
139 	} else {
140 		ndp->ni_cnd.cn_nameiop = NAMEI_LOOKUP;
141 		ndp->ni_cnd.cn_flags = CNP_LOCKLEAF |
142 		    ((fmode & O_NOFOLLOW) ? 0 : CNP_FOLLOW);
143 		error = namei(ndp);
144 		if (error)
145 			return (error);
146 		vp = ndp->ni_vp;
147 	}
148 	if (vp->v_type == VLNK) {
149 		error = EMLINK;
150 		goto bad;
151 	}
152 	if (vp->v_type == VSOCK) {
153 		error = EOPNOTSUPP;
154 		goto bad;
155 	}
156 	if ((fmode & O_CREAT) == 0) {
157 		mode = 0;
158 		if (fmode & (FWRITE | O_TRUNC)) {
159 			if (vp->v_type == VDIR) {
160 				error = EISDIR;
161 				goto bad;
162 			}
163 			error = vn_writechk(vp);
164 			if (error)
165 				goto bad;
166 			mode |= VWRITE;
167 		}
168 		if (fmode & FREAD)
169 			mode |= VREAD;
170 		if (mode) {
171 		        error = VOP_ACCESS(vp, mode, cred, td);
172 			if (error)
173 				goto bad;
174 		}
175 	}
176 	if (fmode & O_TRUNC) {
177 		VOP_UNLOCK(vp, NULL, 0, td);			/* XXX */
178 		VOP_LEASE(vp, td, cred, LEASE_WRITE);
179 		vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);	/* XXX */
180 		VATTR_NULL(vap);
181 		vap->va_size = 0;
182 		error = VOP_SETATTR(vp, vap, cred, td);
183 		if (error)
184 			goto bad;
185 	}
186 	error = VOP_OPEN(vp, fmode, cred, td);
187 	if (error)
188 		goto bad;
189 	/*
190 	 * Make sure that a VM object is created for VMIO support.
191 	 */
192 	if (vn_canvmio(vp) == TRUE) {
193 		if ((error = vfs_object_create(vp, td)) != 0)
194 			goto bad;
195 	}
196 
197 	if (fmode & FWRITE)
198 		vp->v_writecount++;
199 	return (0);
200 bad:
201 	NDFREE(ndp, NDF_ONLY_PNBUF);
202 	vput(vp);
203 	return (error);
204 }
205 
206 /*
207  * Check for write permissions on the specified vnode.
208  * Prototype text segments cannot be written.
209  */
210 int
211 vn_writechk(vp)
212 	struct vnode *vp;
213 {
214 
215 	/*
216 	 * If there's shared text associated with
217 	 * the vnode, try to free it up once.  If
218 	 * we fail, we can't allow writing.
219 	 */
220 	if (vp->v_flag & VTEXT)
221 		return (ETXTBSY);
222 	return (0);
223 }
224 
225 /*
226  * Vnode close call
227  */
228 int
229 vn_close(struct vnode *vp, int flags, struct thread *td)
230 {
231 	int error;
232 
233 	if (flags & FWRITE)
234 		vp->v_writecount--;
235 	error = VOP_CLOSE(vp, flags, td);
236 	vrele(vp);
237 	return (error);
238 }
239 
240 static __inline
241 int
242 sequential_heuristic(struct uio *uio, struct file *fp)
243 {
244 	/*
245 	 * Sequential heuristic - detect sequential operation
246 	 */
247 	if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
248 	    uio->uio_offset == fp->f_nextoff) {
249 		int tmpseq = fp->f_seqcount;
250 		/*
251 		 * XXX we assume that the filesystem block size is
252 		 * the default.  Not true, but still gives us a pretty
253 		 * good indicator of how sequential the read operations
254 		 * are.
255 		 */
256 		tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
257 		if (tmpseq > IO_SEQMAX)
258 			tmpseq = IO_SEQMAX;
259 		fp->f_seqcount = tmpseq;
260 		return(fp->f_seqcount << IO_SEQSHIFT);
261 	}
262 
263 	/*
264 	 * Not sequential, quick draw-down of seqcount
265 	 */
266 	if (fp->f_seqcount > 1)
267 		fp->f_seqcount = 1;
268 	else
269 		fp->f_seqcount = 0;
270 	return(0);
271 }
272 
273 /*
274  * Package up an I/O request on a vnode into a uio and do it.
275  *
276  * We are going to assume the caller has done the appropriate
277  * VOP_LEASE() call before calling vn_rdwr()
278  */
279 int
280 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, td)
281 	enum uio_rw rw;
282 	struct vnode *vp;
283 	caddr_t base;
284 	int len;
285 	off_t offset;
286 	enum uio_seg segflg;
287 	int ioflg;
288 	struct ucred *cred;
289 	int *aresid;
290 	struct thread *td;
291 {
292 	struct uio auio;
293 	struct iovec aiov;
294 	int error;
295 
296 	if ((ioflg & IO_NODELOCKED) == 0)
297 		vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
298 	auio.uio_iov = &aiov;
299 	auio.uio_iovcnt = 1;
300 	aiov.iov_base = base;
301 	aiov.iov_len = len;
302 	auio.uio_resid = len;
303 	auio.uio_offset = offset;
304 	auio.uio_segflg = segflg;
305 	auio.uio_rw = rw;
306 	auio.uio_td = td;
307 	if (rw == UIO_READ) {
308 		error = VOP_READ(vp, &auio, ioflg, cred);
309 	} else {
310 		error = VOP_WRITE(vp, &auio, ioflg, cred);
311 	}
312 	if (aresid)
313 		*aresid = auio.uio_resid;
314 	else
315 		if (auio.uio_resid && error == 0)
316 			error = EIO;
317 	if ((ioflg & IO_NODELOCKED) == 0)
318 		VOP_UNLOCK(vp, NULL, 0, td);
319 	return (error);
320 }
321 
322 /*
323  * Package up an I/O request on a vnode into a uio and do it.  The I/O
324  * request is split up into smaller chunks and we try to avoid saturating
325  * the buffer cache while potentially holding a vnode locked, so we
326  * check bwillwrite() before calling vn_rdwr().  We also call uio_yield()
327  * to give other processes a chance to lock the vnode (either other processes
328  * core'ing the same binary, or unrelated processes scanning the directory).
329  */
330 int
331 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, td)
332 	enum uio_rw rw;
333 	struct vnode *vp;
334 	caddr_t base;
335 	int len;
336 	off_t offset;
337 	enum uio_seg segflg;
338 	int ioflg;
339 	struct ucred *cred;
340 	int *aresid;
341 	struct thread *td;
342 {
343 	int error = 0;
344 
345 	do {
346 		int chunk;
347 
348 		/*
349 		 * Force `offset' to a multiple of MAXBSIZE except possibly
350 		 * for the first chunk, so that filesystems only need to
351 		 * write full blocks except possibly for the first and last
352 		 * chunks.
353 		 */
354 		chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
355 
356 		if (chunk > len)
357 			chunk = len;
358 		if (rw != UIO_READ && vp->v_type == VREG)
359 			bwillwrite();
360 		error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
361 			    ioflg, cred, aresid, td);
362 		len -= chunk;	/* aresid calc already includes length */
363 		if (error)
364 			break;
365 		offset += chunk;
366 		base += chunk;
367 		uio_yield();
368 	} while (len);
369 	if (aresid)
370 		*aresid += len;
371 	return (error);
372 }
373 
374 /*
375  * File table vnode read routine.
376  */
377 static int
378 vn_read(fp, uio, cred, flags, td)
379 	struct file *fp;
380 	struct uio *uio;
381 	struct ucred *cred;
382 	struct thread *td;
383 	int flags;
384 {
385 	struct vnode *vp;
386 	int error, ioflag;
387 
388 	KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td));
389 	vp = (struct vnode *)fp->f_data;
390 	ioflag = 0;
391 	if (fp->f_flag & FNONBLOCK)
392 		ioflag |= IO_NDELAY;
393 	if (fp->f_flag & O_DIRECT)
394 		ioflag |= IO_DIRECT;
395 	VOP_LEASE(vp, td, cred, LEASE_READ);
396 	vn_lock(vp, NULL, LK_SHARED | LK_NOPAUSE | LK_RETRY, td);
397 	if ((flags & FOF_OFFSET) == 0)
398 		uio->uio_offset = fp->f_offset;
399 
400 	ioflag |= sequential_heuristic(uio, fp);
401 
402 	error = VOP_READ(vp, uio, ioflag, cred);
403 	if ((flags & FOF_OFFSET) == 0)
404 		fp->f_offset = uio->uio_offset;
405 	fp->f_nextoff = uio->uio_offset;
406 	VOP_UNLOCK(vp, NULL, 0, td);
407 	return (error);
408 }
409 
410 /*
411  * File table vnode write routine.
412  */
413 static int
414 vn_write(fp, uio, cred, flags, td)
415 	struct file *fp;
416 	struct uio *uio;
417 	struct ucred *cred;
418 	struct thread *td;
419 	int flags;
420 {
421 	struct vnode *vp;
422 	int error, ioflag;
423 
424 	KASSERT(uio->uio_td == td, ("uio_procp %p is not p %p",
425 	    uio->uio_td, td));
426 	vp = (struct vnode *)fp->f_data;
427 	if (vp->v_type == VREG)
428 		bwillwrite();
429 	vp = (struct vnode *)fp->f_data;	/* XXX needed? */
430 	ioflag = IO_UNIT;
431 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
432 		ioflag |= IO_APPEND;
433 	if (fp->f_flag & FNONBLOCK)
434 		ioflag |= IO_NDELAY;
435 	if (fp->f_flag & O_DIRECT)
436 		ioflag |= IO_DIRECT;
437 	if ((fp->f_flag & O_FSYNC) ||
438 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
439 		ioflag |= IO_SYNC;
440 	VOP_LEASE(vp, td, cred, LEASE_WRITE);
441 	vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
442 	if ((flags & FOF_OFFSET) == 0)
443 		uio->uio_offset = fp->f_offset;
444 	ioflag |= sequential_heuristic(uio, fp);
445 	error = VOP_WRITE(vp, uio, ioflag, cred);
446 	if ((flags & FOF_OFFSET) == 0)
447 		fp->f_offset = uio->uio_offset;
448 	fp->f_nextoff = uio->uio_offset;
449 	VOP_UNLOCK(vp, NULL, 0, td);
450 	return (error);
451 }
452 
453 /*
454  * File table vnode stat routine.
455  */
456 static int
457 vn_statfile(struct file *fp, struct stat *sb, struct thread *td)
458 {
459 	struct vnode *vp = (struct vnode *)fp->f_data;
460 
461 	return vn_stat(vp, sb, td);
462 }
463 
464 int
465 vn_stat(struct vnode *vp, struct stat *sb, struct thread *td)
466 {
467 	struct vattr vattr;
468 	struct vattr *vap;
469 	int error;
470 	u_short mode;
471 
472 	vap = &vattr;
473 	error = VOP_GETATTR(vp, vap, td);
474 	if (error)
475 		return (error);
476 
477 	/*
478 	 * Zero the spare stat fields
479 	 */
480 	sb->st_lspare = 0;
481 	sb->st_qspare[0] = 0;
482 	sb->st_qspare[1] = 0;
483 
484 	/*
485 	 * Copy from vattr table
486 	 */
487 	if (vap->va_fsid != VNOVAL)
488 		sb->st_dev = vap->va_fsid;
489 	else
490 		sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
491 	sb->st_ino = vap->va_fileid;
492 	mode = vap->va_mode;
493 	switch (vap->va_type) {
494 	case VREG:
495 		mode |= S_IFREG;
496 		break;
497 	case VDIR:
498 		mode |= S_IFDIR;
499 		break;
500 	case VBLK:
501 		mode |= S_IFBLK;
502 		break;
503 	case VCHR:
504 		mode |= S_IFCHR;
505 		break;
506 	case VLNK:
507 		mode |= S_IFLNK;
508 		/* This is a cosmetic change, symlinks do not have a mode. */
509 		if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW)
510 			sb->st_mode &= ~ACCESSPERMS;	/* 0000 */
511 		else
512 			sb->st_mode |= ACCESSPERMS;	/* 0777 */
513 		break;
514 	case VSOCK:
515 		mode |= S_IFSOCK;
516 		break;
517 	case VFIFO:
518 		mode |= S_IFIFO;
519 		break;
520 	default:
521 		return (EBADF);
522 	};
523 	sb->st_mode = mode;
524 	sb->st_nlink = vap->va_nlink;
525 	sb->st_uid = vap->va_uid;
526 	sb->st_gid = vap->va_gid;
527 	sb->st_rdev = vap->va_rdev;
528 	sb->st_size = vap->va_size;
529 	sb->st_atimespec = vap->va_atime;
530 	sb->st_mtimespec = vap->va_mtime;
531 	sb->st_ctimespec = vap->va_ctime;
532 
533         /*
534 	 * According to www.opengroup.org, the meaning of st_blksize is
535 	 *   "a filesystem-specific preferred I/O block size for this
536 	 *    object.  In some filesystem types, this may vary from file
537 	 *    to file"
538 	 * Default to PAGE_SIZE after much discussion.
539 	 */
540 
541 	if (vap->va_type == VREG) {
542 		sb->st_blksize = vap->va_blocksize;
543 	} else if (vn_isdisk(vp, NULL)) {
544 		/*
545 		 * XXX this is broken.  If the device is not yet open (aka
546 		 * stat() call, aka v_rdev == NULL), how are we supposed
547 		 * to get a valid block size out of it?
548 		 */
549 		dev_t dev;
550 
551 		if ((dev = vp->v_rdev) == NULL)
552 			dev = udev2dev(vp->v_udev, vp->v_type == VBLK);
553 		sb->st_blksize = dev->si_bsize_best;
554 		if (sb->st_blksize < dev->si_bsize_phys)
555 			sb->st_blksize = dev->si_bsize_phys;
556 		if (sb->st_blksize < BLKDEV_IOSIZE)
557 			sb->st_blksize = BLKDEV_IOSIZE;
558 	} else {
559 		sb->st_blksize = PAGE_SIZE;
560 	}
561 
562 	sb->st_flags = vap->va_flags;
563 	if (suser(td))
564 		sb->st_gen = 0;
565 	else
566 		sb->st_gen = vap->va_gen;
567 
568 #if (S_BLKSIZE == 512)
569 	/* Optimize this case */
570 	sb->st_blocks = vap->va_bytes >> 9;
571 #else
572 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
573 #endif
574 	return (0);
575 }
576 
577 /*
578  * File table vnode ioctl routine.
579  */
580 static int
581 vn_ioctl(struct file *fp, u_long com, caddr_t data, struct thread *td)
582 {
583 	struct vnode *vp = ((struct vnode *)fp->f_data);
584 	struct vnode *ovp;
585 	struct ucred *ucred;
586 	struct vattr vattr;
587 	int error;
588 
589 	KKASSERT(td->td_proc != NULL);
590 	ucred = td->td_proc->p_ucred;
591 
592 	switch (vp->v_type) {
593 	case VREG:
594 	case VDIR:
595 		if (com == FIONREAD) {
596 			error = VOP_GETATTR(vp, &vattr, td);
597 			if (error)
598 				return (error);
599 			*(int *)data = vattr.va_size - fp->f_offset;
600 			return (0);
601 		}
602 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
603 			return (0);			/* XXX */
604 		/* fall into ... */
605 	default:
606 #if 0
607 		return (ENOTTY);
608 #endif
609 	case VFIFO:
610 	case VCHR:
611 	case VBLK:
612 		if (com == FIODTYPE) {
613 			if (vp->v_type != VCHR && vp->v_type != VBLK)
614 				return (ENOTTY);
615 			*(int *)data = dev_dflags(vp->v_rdev) & D_TYPEMASK;
616 			return (0);
617 		}
618 		error = VOP_IOCTL(vp, com, data, fp->f_flag, ucred, td);
619 		if (error == 0 && com == TIOCSCTTY) {
620 			struct session *sess = td->td_proc->p_session;
621 
622 			/* Do nothing if reassigning same control tty */
623 			if (sess->s_ttyvp == vp)
624 				return (0);
625 
626 			/* Get rid of reference to old control tty */
627 			ovp = sess->s_ttyvp;
628 			vref(vp);
629 			sess->s_ttyvp = vp;
630 			if (ovp)
631 				vrele(ovp);
632 		}
633 		return (error);
634 	}
635 }
636 
637 /*
638  * File table vnode poll routine.
639  */
640 static int
641 vn_poll(struct file *fp, int events, struct ucred *cred, struct thread *td)
642 {
643 	return (VOP_POLL(((struct vnode *)fp->f_data), events, cred, td));
644 }
645 
646 /*
647  * Check that the vnode is still valid, and if so
648  * acquire requested lock.
649  */
650 int
651 #ifndef	DEBUG_LOCKS
652 vn_lock(struct vnode *vp, lwkt_tokref_t vlock, int flags, struct thread *td)
653 #else
654 debug_vn_lock(struct vnode *vp, lwkt_tokref_t vlock, int flags,
655 		struct thread *td, const char *filename, int line)
656 #endif
657 {
658 	int error;
659 	lwkt_tokref vvlock;
660 
661 	do {
662 		if ((flags & LK_INTERLOCK) == 0) {
663 			lwkt_gettoken(&vvlock, vp->v_interlock);
664 			vlock = &vvlock;
665 		}
666 		if ((vp->v_flag & VXLOCK) && vp->v_vxthread != curthread) {
667 			vp->v_flag |= VXWANT;
668 			lwkt_reltoken(vlock);
669 			tsleep((caddr_t)vp, 0, "vn_lock", 0);
670 			error = ENOENT;
671 		} else {
672 #if 0
673 			/* this can now occur in normal operation */
674 			if (vp->v_vxthread != NULL)
675 				log(LOG_INFO, "VXLOCK interlock avoided in vn_lock\n");
676 #endif
677 #ifdef	DEBUG_LOCKS
678 			vp->filename = filename;
679 			vp->line = line;
680 #endif
681 			error = VOP_LOCK(vp, vlock,
682 				    flags | LK_NOPAUSE | LK_INTERLOCK, td);
683 			if (error == 0)
684 				return (0);
685 		}
686 		flags &= ~LK_INTERLOCK;
687 	} while (flags & LK_RETRY);
688 	return (error);
689 }
690 
691 /*
692  * File table vnode close routine.
693  */
694 static int
695 vn_closefile(struct file *fp, struct thread *td)
696 {
697 	int err;
698 
699 	fp->f_ops = &badfileops;
700 	err = vn_close(((struct vnode *)fp->f_data), fp->f_flag, td);
701 	return(err);
702 }
703 
704 static int
705 vn_kqfilter(struct file *fp, struct knote *kn)
706 {
707 
708 	return (VOP_KQFILTER(((struct vnode *)fp->f_data), kn));
709 }
710