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.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
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.
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
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 1.44 2006/08/12 00:26:20 dillon Exp $
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/fcntl.h>
49 #include <sys/mount.h>
50 #include <sys/nlookup.h>
51 #include <sys/vnode.h>
53 #include <sys/filio.h>
54 #include <sys/ttycom.h>
56 #include <sys/syslog.h>
58 static int vn_closefile (struct file *fp);
59 static int vn_ioctl (struct file *fp, u_long com, caddr_t data,
61 static int vn_read (struct file *fp, struct uio *uio,
62 struct ucred *cred, int flags);
63 static int svn_read (struct file *fp, struct uio *uio,
64 struct ucred *cred, int flags);
65 static int vn_poll (struct file *fp, int events, struct ucred *cred);
66 static int vn_kqfilter (struct file *fp, struct knote *kn);
67 static int vn_statfile (struct file *fp, struct stat *sb, struct ucred *cred);
68 static int vn_write (struct file *fp, struct uio *uio,
69 struct ucred *cred, int flags);
70 static int svn_write (struct file *fp, struct uio *uio,
71 struct ucred *cred, int flags);
73 struct fileops vnode_fileops = {
78 .fo_kqfilter = vn_kqfilter,
79 .fo_stat = vn_statfile,
80 .fo_close = vn_closefile,
81 .fo_shutdown = nofo_shutdown
84 struct fileops specvnode_fileops = {
86 .fo_write = svn_write,
89 .fo_kqfilter = vn_kqfilter,
90 .fo_stat = vn_statfile,
91 .fo_close = vn_closefile,
92 .fo_shutdown = nofo_shutdown
96 * Shortcut the device read/write. This avoids a lot of vnode junk.
97 * Basically the specfs vnops for read and write take the locked vnode,
98 * unlock it (because we can't hold the vnode locked while reading or writing
99 * a device which may block indefinitely), issues the device operation, then
100 * relock the vnode before returning, plus other junk. This bypasses all
101 * of that and just does the device operation.
104 vn_setspecops(struct file *fp)
106 if (vfs_fastdev && fp->f_ops == &vnode_fileops) {
107 fp->f_ops = &specvnode_fileops;
112 * Common code for vnode open operations. Check permissions, and call
113 * the VOP_NOPEN or VOP_NCREATE routine.
115 * The caller is responsible for setting up nd with nlookup_init() and
116 * for cleaning it up with nlookup_done(), whether we return an error
119 * On success nd->nl_open_vp will hold a referenced and, if requested,
120 * locked vnode. A locked vnode is requested via NLC_LOCKVP. If fp
121 * is non-NULL the vnode will be installed in the file pointer.
123 * NOTE: The vnode is referenced just once on return whether or not it
124 * is also installed in the file pointer.
127 vn_open(struct nlookupdata *nd, struct file *fp, int fmode, int cmode)
130 struct ucred *cred = nd->nl_cred;
132 struct vattr *vap = &vat;
133 struct namecache *ncp;
137 * Lookup the path and create or obtain the vnode. After a
138 * successful lookup a locked nd->nl_ncp will be returned.
140 * The result of this section should be a locked vnode.
142 * XXX with only a little work we should be able to avoid locking
143 * the vnode if FWRITE, O_CREAT, and O_TRUNC are *not* set.
145 if (fmode & O_CREAT) {
147 * CONDITIONAL CREATE FILE CASE
149 * Setting NLC_CREATE causes a negative hit to store
150 * the negative hit ncp and not return an error. Then
151 * nc_error or nc_vp may be checked to see if the ncp
152 * represents a negative hit. NLC_CREATE also requires
153 * write permission on the governing directory or EPERM
156 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
157 nd->nl_flags |= NLC_FOLLOW;
158 nd->nl_flags |= NLC_CREATE;
163 * NORMAL OPEN FILE CASE
173 * split case to allow us to re-resolve and retry the ncp in case
177 if (fmode & O_CREAT) {
178 if (ncp->nc_vp == NULL) {
181 vap->va_mode = cmode;
183 vap->va_vaflags |= VA_EXCLUSIVE;
184 error = VOP_NCREATE(ncp, &vp, nd->nl_cred, vap);
188 /* locked vnode is returned */
190 if (fmode & O_EXCL) {
193 error = cache_vget(ncp, cred,
201 error = cache_vget(ncp, cred, LK_EXCLUSIVE, &vp);
207 * We have a locked vnode and ncp now. Note that the ncp will
208 * be cleaned up by the caller if nd->nl_ncp is left intact.
210 if (vp->v_type == VLNK) {
214 if (vp->v_type == VSOCK) {
218 if ((fmode & O_CREAT) == 0) {
220 if (fmode & (FWRITE | O_TRUNC)) {
221 if (vp->v_type == VDIR) {
225 error = vn_writechk(vp);
228 * Special stale handling, re-resolve the
231 if (error == ESTALE) {
234 cache_setunresolved(ncp);
235 error = cache_resolve(ncp, cred);
246 error = VOP_ACCESS(vp, mode, cred);
249 * Special stale handling, re-resolve the
252 if (error == ESTALE) {
255 cache_setunresolved(ncp);
256 error = cache_resolve(ncp, cred);
264 if (fmode & O_TRUNC) {
265 vn_unlock(vp); /* XXX */
266 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
269 error = VOP_SETATTR(vp, vap, cred);
275 * Setup the fp so VOP_OPEN can override it. No descriptor has been
276 * associated with the fp yet so we own it clean. f_ncp inherits
280 if (vp->v_type == VDIR) {
281 fp->f_ncp = nd->nl_ncp;
283 cache_unlock(fp->f_ncp);
288 * Get rid of nl_ncp. vn_open does not return it (it returns the
289 * vnode or the file pointer). Note: we can't leave nl_ncp locked
290 * through the VOP_OPEN anyway since the VOP_OPEN may block, e.g.
294 cache_put(nd->nl_ncp);
298 error = VOP_OPEN(vp, fmode, cred, fp);
301 * setting f_ops to &badfileops will prevent the descriptor
302 * code from trying to close and release the vnode, since
303 * the open failed we do not want to call close.
307 fp->f_ops = &badfileops;
314 * Assert that VREG files have been setup for vmio.
316 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
317 ("vn_open: regular file was not VMIO enabled!"));
321 * Return the vnode. XXX needs some cleaning up. The vnode is
322 * only returned in the fp == NULL case.
326 nd->nl_vp_fmode = fmode;
327 if ((nd->nl_flags & NLC_LOCKVP) == 0)
340 * Check for write permissions on the specified vnode.
341 * Prototype text segments cannot be written.
349 * If there's shared text associated with
350 * the vnode, try to free it up once. If
351 * we fail, we can't allow writing.
353 if (vp->v_flag & VTEXT)
362 vn_close(struct vnode *vp, int flags)
366 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) == 0) {
367 error = VOP_CLOSE(vp, flags);
376 sequential_heuristic(struct uio *uio, struct file *fp)
379 * Sequential heuristic - detect sequential operation
381 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
382 uio->uio_offset == fp->f_nextoff) {
383 int tmpseq = fp->f_seqcount;
385 * XXX we assume that the filesystem block size is
386 * the default. Not true, but still gives us a pretty
387 * good indicator of how sequential the read operations
390 tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
391 if (tmpseq > IO_SEQMAX)
393 fp->f_seqcount = tmpseq;
394 return(fp->f_seqcount << IO_SEQSHIFT);
398 * Not sequential, quick draw-down of seqcount
400 if (fp->f_seqcount > 1)
408 * Package up an I/O request on a vnode into a uio and do it.
411 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len,
412 off_t offset, enum uio_seg segflg, int ioflg,
413 struct ucred *cred, int *aresid)
419 if ((ioflg & IO_NODELOCKED) == 0)
420 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
421 auio.uio_iov = &aiov;
423 aiov.iov_base = base;
425 auio.uio_resid = len;
426 auio.uio_offset = offset;
427 auio.uio_segflg = segflg;
429 auio.uio_td = curthread;
430 if (rw == UIO_READ) {
431 error = VOP_READ(vp, &auio, ioflg, cred);
433 error = VOP_WRITE(vp, &auio, ioflg, cred);
436 *aresid = auio.uio_resid;
438 if (auio.uio_resid && error == 0)
440 if ((ioflg & IO_NODELOCKED) == 0)
446 * Package up an I/O request on a vnode into a uio and do it. The I/O
447 * request is split up into smaller chunks and we try to avoid saturating
448 * the buffer cache while potentially holding a vnode locked, so we
449 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
450 * to give other processes a chance to lock the vnode (either other processes
451 * core'ing the same binary, or unrelated processes scanning the directory).
454 vn_rdwr_inchunks(enum uio_rw rw, struct vnode *vp, caddr_t base, int len,
455 off_t offset, enum uio_seg segflg, int ioflg,
456 struct ucred *cred, int *aresid)
464 * Force `offset' to a multiple of MAXBSIZE except possibly
465 * for the first chunk, so that filesystems only need to
466 * write full blocks except possibly for the first and last
469 chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
473 if (rw != UIO_READ && vp->v_type == VREG)
475 error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
476 ioflg, cred, aresid);
477 len -= chunk; /* aresid calc already includes length */
490 * MPALMOSTSAFE - acquires mplock
493 vn_read(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
499 KASSERT(uio->uio_td == curthread,
500 ("uio_td %p is not td %p", uio->uio_td, curthread));
501 vp = (struct vnode *)fp->f_data;
504 if (flags & O_FBLOCKING) {
505 /* ioflag &= ~IO_NDELAY; */
506 } else if (flags & O_FNONBLOCKING) {
508 } else if (fp->f_flag & FNONBLOCK) {
511 if (flags & O_FBUFFERED) {
512 /* ioflag &= ~IO_DIRECT; */
513 } else if (flags & O_FUNBUFFERED) {
515 } else if (fp->f_flag & O_DIRECT) {
518 vn_lock(vp, LK_SHARED | LK_RETRY);
519 if ((flags & O_FOFFSET) == 0)
520 uio->uio_offset = fp->f_offset;
522 ioflag |= sequential_heuristic(uio, fp);
524 error = VOP_READ(vp, uio, ioflag, cred);
525 if ((flags & O_FOFFSET) == 0)
526 fp->f_offset = uio->uio_offset;
527 fp->f_nextoff = uio->uio_offset;
534 * Device-optimized file table vnode read routine.
536 * This bypasses the VOP table and talks directly to the device. Most
537 * filesystems just route to specfs and can make this optimization.
539 * MPALMOSTSAFE - acquires mplock
542 svn_read(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
550 KASSERT(uio->uio_td == curthread,
551 ("uio_td %p is not td %p", uio->uio_td, curthread));
553 vp = (struct vnode *)fp->f_data;
554 if (vp == NULL || vp->v_type == VBAD) {
559 if ((dev = vp->v_rdev) == NULL) {
565 if (uio->uio_resid == 0) {
569 if ((flags & O_FOFFSET) == 0)
570 uio->uio_offset = fp->f_offset;
573 if (flags & O_FBLOCKING) {
574 /* ioflag &= ~IO_NDELAY; */
575 } else if (flags & O_FNONBLOCKING) {
577 } else if (fp->f_flag & FNONBLOCK) {
580 if (flags & O_FBUFFERED) {
581 /* ioflag &= ~IO_DIRECT; */
582 } else if (flags & O_FUNBUFFERED) {
584 } else if (fp->f_flag & O_DIRECT) {
587 ioflag |= sequential_heuristic(uio, fp);
589 error = dev_dread(dev, uio, ioflag);
592 if ((flags & O_FOFFSET) == 0)
593 fp->f_offset = uio->uio_offset;
594 fp->f_nextoff = uio->uio_offset;
601 * MPALMOSTSAFE - acquires mplock
604 vn_write(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
610 KASSERT(uio->uio_td == curthread,
611 ("uio_procp %p is not p %p", uio->uio_td, curthread));
612 vp = (struct vnode *)fp->f_data;
613 if (vp->v_type == VREG)
615 vp = (struct vnode *)fp->f_data; /* XXX needed? */
618 if (vp->v_type == VREG &&
619 ((fp->f_flag & O_APPEND) || (flags & O_FAPPEND))) {
623 if (flags & O_FBLOCKING) {
624 /* ioflag &= ~IO_NDELAY; */
625 } else if (flags & O_FNONBLOCKING) {
627 } else if (fp->f_flag & FNONBLOCK) {
630 if (flags & O_FBUFFERED) {
631 /* ioflag &= ~IO_DIRECT; */
632 } else if (flags & O_FUNBUFFERED) {
634 } else if (fp->f_flag & O_DIRECT) {
637 if (flags & O_FASYNCWRITE) {
638 /* ioflag &= ~IO_SYNC; */
639 } else if (flags & O_FSYNCWRITE) {
641 } else if (fp->f_flag & O_FSYNC) {
645 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))
647 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
648 if ((flags & O_FOFFSET) == 0)
649 uio->uio_offset = fp->f_offset;
650 ioflag |= sequential_heuristic(uio, fp);
651 error = VOP_WRITE(vp, uio, ioflag, cred);
652 if ((flags & O_FOFFSET) == 0)
653 fp->f_offset = uio->uio_offset;
654 fp->f_nextoff = uio->uio_offset;
661 * Device-optimized file table vnode write routine.
663 * This bypasses the VOP table and talks directly to the device. Most
664 * filesystems just route to specfs and can make this optimization.
666 * MPALMOSTSAFE - acquires mplock
669 svn_write(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
677 KASSERT(uio->uio_td == curthread,
678 ("uio_procp %p is not p %p", uio->uio_td, curthread));
680 vp = (struct vnode *)fp->f_data;
681 if (vp == NULL || vp->v_type == VBAD) {
685 if (vp->v_type == VREG)
687 vp = (struct vnode *)fp->f_data; /* XXX needed? */
689 if ((dev = vp->v_rdev) == NULL) {
695 if ((flags & O_FOFFSET) == 0)
696 uio->uio_offset = fp->f_offset;
699 if (vp->v_type == VREG &&
700 ((fp->f_flag & O_APPEND) || (flags & O_FAPPEND))) {
704 if (flags & O_FBLOCKING) {
705 /* ioflag &= ~IO_NDELAY; */
706 } else if (flags & O_FNONBLOCKING) {
708 } else if (fp->f_flag & FNONBLOCK) {
711 if (flags & O_FBUFFERED) {
712 /* ioflag &= ~IO_DIRECT; */
713 } else if (flags & O_FUNBUFFERED) {
715 } else if (fp->f_flag & O_DIRECT) {
718 if (flags & O_FASYNCWRITE) {
719 /* ioflag &= ~IO_SYNC; */
720 } else if (flags & O_FSYNCWRITE) {
722 } else if (fp->f_flag & O_FSYNC) {
726 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))
728 ioflag |= sequential_heuristic(uio, fp);
730 error = dev_dwrite(dev, uio, ioflag);
733 if ((flags & O_FOFFSET) == 0)
734 fp->f_offset = uio->uio_offset;
735 fp->f_nextoff = uio->uio_offset;
742 * MPALMOSTSAFE - acquires mplock
745 vn_statfile(struct file *fp, struct stat *sb, struct ucred *cred)
751 vp = (struct vnode *)fp->f_data;
752 error = vn_stat(vp, sb, cred);
758 vn_stat(struct vnode *vp, struct stat *sb, struct ucred *cred)
767 error = VOP_GETATTR(vp, vap);
772 * Zero the spare stat fields
778 * Copy from vattr table
780 if (vap->va_fsid != VNOVAL)
781 sb->st_dev = vap->va_fsid;
783 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
784 sb->st_ino = vap->va_fileid;
786 switch (vap->va_type) {
801 /* This is a cosmetic change, symlinks do not have a mode. */
802 if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW)
803 sb->st_mode &= ~ACCESSPERMS; /* 0000 */
805 sb->st_mode |= ACCESSPERMS; /* 0777 */
817 sb->st_nlink = vap->va_nlink;
818 sb->st_uid = vap->va_uid;
819 sb->st_gid = vap->va_gid;
820 sb->st_rdev = vap->va_rdev;
821 sb->st_size = vap->va_size;
822 sb->st_atimespec = vap->va_atime;
823 sb->st_mtimespec = vap->va_mtime;
824 sb->st_ctimespec = vap->va_ctime;
827 * A VCHR and VBLK device may track the last access and last modified
828 * time independantly of the filesystem. This is particularly true
829 * because device read and write calls may bypass the filesystem.
831 if (vp->v_type == VCHR || vp->v_type == VBLK) {
832 if ((dev = vp->v_rdev) != NULL) {
833 if (dev->si_lastread) {
834 sb->st_atimespec.tv_sec = dev->si_lastread;
835 sb->st_atimespec.tv_nsec = 0;
837 if (dev->si_lastwrite) {
838 sb->st_atimespec.tv_sec = dev->si_lastwrite;
839 sb->st_atimespec.tv_nsec = 0;
845 * According to www.opengroup.org, the meaning of st_blksize is
846 * "a filesystem-specific preferred I/O block size for this
847 * object. In some filesystem types, this may vary from file
849 * Default to PAGE_SIZE after much discussion.
852 if (vap->va_type == VREG) {
853 sb->st_blksize = vap->va_blocksize;
854 } else if (vn_isdisk(vp, NULL)) {
856 * XXX this is broken. If the device is not yet open (aka
857 * stat() call, aka v_rdev == NULL), how are we supposed
858 * to get a valid block size out of it?
862 if ((dev = vp->v_rdev) == NULL)
863 dev = udev2dev(vp->v_udev, vp->v_type == VBLK);
864 sb->st_blksize = dev->si_bsize_best;
865 if (sb->st_blksize < dev->si_bsize_phys)
866 sb->st_blksize = dev->si_bsize_phys;
867 if (sb->st_blksize < BLKDEV_IOSIZE)
868 sb->st_blksize = BLKDEV_IOSIZE;
870 sb->st_blksize = PAGE_SIZE;
873 sb->st_flags = vap->va_flags;
874 if (suser_cred(cred, 0))
877 sb->st_gen = vap->va_gen;
879 #if (S_BLKSIZE == 512)
880 /* Optimize this case */
881 sb->st_blocks = vap->va_bytes >> 9;
883 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
885 sb->st_fsmid = vap->va_fsmid;
890 * MPALMOSTSAFE - acquires mplock
893 vn_ioctl(struct file *fp, u_long com, caddr_t data, struct ucred *ucred)
895 struct vnode *vp = ((struct vnode *)fp->f_data);
902 switch (vp->v_type) {
905 if (com == FIONREAD) {
906 if ((error = VOP_GETATTR(vp, &vattr)) != 0)
908 *(int *)data = vattr.va_size - fp->f_offset;
912 if (com == FIOASYNC) { /* XXX */
924 if (com == FIODTYPE) {
925 if (vp->v_type != VCHR && vp->v_type != VBLK) {
929 *(int *)data = dev_dflags(vp->v_rdev) & D_TYPEMASK;
933 error = VOP_IOCTL(vp, com, data, fp->f_flag, ucred);
934 if (error == 0 && com == TIOCSCTTY) {
935 struct proc *p = curthread->td_proc;
936 struct session *sess;
944 /* Do nothing if reassigning same control tty */
945 if (sess->s_ttyvp == vp) {
950 /* Get rid of reference to old control tty */
964 * MPALMOSTSAFE - acquires mplock
967 vn_poll(struct file *fp, int events, struct ucred *cred)
972 error = VOP_POLL(((struct vnode *)fp->f_data), events, cred);
978 * Check that the vnode is still valid, and if so
979 * acquire requested lock.
983 vn_lock(struct vnode *vp, int flags)
985 debug_vn_lock(struct vnode *vp, int flags, const char *filename, int line)
992 vp->filename = filename;
994 error = debuglockmgr(&vp->v_lock, flags,
995 "vn_lock", filename, line);
997 error = lockmgr(&vp->v_lock, flags);
1001 } while (flags & LK_RETRY);
1004 * Because we (had better!) have a ref on the vnode, once it
1005 * goes to VRECLAIMED state it will not be recycled until all
1006 * refs go away. So we can just check the flag.
1008 if (error == 0 && (vp->v_flag & VRECLAIMED)) {
1009 lockmgr(&vp->v_lock, LK_RELEASE);
1016 vn_unlock(struct vnode *vp)
1018 lockmgr(&vp->v_lock, LK_RELEASE);
1022 vn_islocked(struct vnode *vp)
1024 return (lockstatus(&vp->v_lock, curthread));
1028 * MPALMOSTSAFE - acquires mplock
1031 vn_closefile(struct file *fp)
1036 fp->f_ops = &badfileops;
1037 error = vn_close(((struct vnode *)fp->f_data), fp->f_flag);
1043 * MPALMOSTSAFE - acquires mplock
1046 vn_kqfilter(struct file *fp, struct knote *kn)
1051 error = VOP_KQFILTER(((struct vnode *)fp->f_data), kn);