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 1.58 2008/06/28 17:59:49 dillon Exp $
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/fcntl.h>
50 #include <sys/mount.h>
51 #include <sys/nlookup.h>
52 #include <sys/vnode.h>
54 #include <sys/filio.h>
55 #include <sys/ttycom.h>
57 #include <sys/sysctl.h>
58 #include <sys/syslog.h>
60 #include <sys/thread2.h>
61 #include <sys/mplock2.h>
63 static int vn_closefile (struct file
*fp
);
64 static int vn_ioctl (struct file
*fp
, u_long com
, caddr_t data
,
65 struct ucred
*cred
, struct sysmsg
*msg
);
66 static int vn_read (struct file
*fp
, struct uio
*uio
,
67 struct ucred
*cred
, int flags
);
68 static int vn_poll (struct file
*fp
, int events
, struct ucred
*cred
);
69 static int vn_kqfilter (struct file
*fp
, struct knote
*kn
);
70 static int vn_statfile (struct file
*fp
, struct stat
*sb
, struct ucred
*cred
);
71 static int vn_write (struct file
*fp
, struct uio
*uio
,
72 struct ucred
*cred
, int flags
);
75 static int read_mpsafe
= 0;
76 SYSCTL_INT(_vfs
, OID_AUTO
, read_mpsafe
, CTLFLAG_RW
, &read_mpsafe
, 0, "");
77 static int write_mpsafe
= 0;
78 SYSCTL_INT(_vfs
, OID_AUTO
, write_mpsafe
, CTLFLAG_RW
, &write_mpsafe
, 0, "");
79 static int getattr_mpsafe
= 0;
80 SYSCTL_INT(_vfs
, OID_AUTO
, getattr_mpsafe
, CTLFLAG_RW
, &getattr_mpsafe
, 0, "");
83 #define write_mpsafe 0
84 #define getattr_mpsafe 0
87 struct fileops vnode_fileops
= {
92 .fo_kqfilter
= vn_kqfilter
,
93 .fo_stat
= vn_statfile
,
94 .fo_close
= vn_closefile
,
95 .fo_shutdown
= nofo_shutdown
99 * Common code for vnode open operations. Check permissions, and call
100 * the VOP_NOPEN or VOP_NCREATE routine.
102 * The caller is responsible for setting up nd with nlookup_init() and
103 * for cleaning it up with nlookup_done(), whether we return an error
106 * On success nd->nl_open_vp will hold a referenced and, if requested,
107 * locked vnode. A locked vnode is requested via NLC_LOCKVP. If fp
108 * is non-NULL the vnode will be installed in the file pointer.
110 * NOTE: The vnode is referenced just once on return whether or not it
111 * is also installed in the file pointer.
114 vn_open(struct nlookupdata
*nd
, struct file
*fp
, int fmode
, int cmode
)
117 struct ucred
*cred
= nd
->nl_cred
;
119 struct vattr
*vap
= &vat
;
123 * Certain combinations are illegal
125 if ((fmode
& (FWRITE
| O_TRUNC
)) == O_TRUNC
)
129 * Lookup the path and create or obtain the vnode. After a
130 * successful lookup a locked nd->nl_nch will be returned.
132 * The result of this section should be a locked vnode.
134 * XXX with only a little work we should be able to avoid locking
135 * the vnode if FWRITE, O_CREAT, and O_TRUNC are *not* set.
137 nd
->nl_flags
|= NLC_OPEN
;
138 if (fmode
& O_APPEND
)
139 nd
->nl_flags
|= NLC_APPEND
;
141 nd
->nl_flags
|= NLC_TRUNCATE
;
143 nd
->nl_flags
|= NLC_READ
;
145 nd
->nl_flags
|= NLC_WRITE
;
146 if ((fmode
& O_EXCL
) == 0 && (fmode
& O_NOFOLLOW
) == 0)
147 nd
->nl_flags
|= NLC_FOLLOW
;
149 if (fmode
& O_CREAT
) {
151 * CONDITIONAL CREATE FILE CASE
153 * Setting NLC_CREATE causes a negative hit to store
154 * the negative hit ncp and not return an error. Then
155 * nc_error or nc_vp may be checked to see if the ncp
156 * represents a negative hit. NLC_CREATE also requires
157 * write permission on the governing directory or EPERM
160 nd
->nl_flags
|= NLC_CREATE
;
161 nd
->nl_flags
|= NLC_REFDVP
;
166 * NORMAL OPEN FILE CASE
175 * split case to allow us to re-resolve and retry the ncp in case
179 if (fmode
& O_CREAT
) {
180 if (nd
->nl_nch
.ncp
->nc_vp
== NULL
) {
181 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
185 vap
->va_mode
= cmode
;
187 vap
->va_vaflags
|= VA_EXCLUSIVE
;
188 error
= VOP_NCREATE(&nd
->nl_nch
, nd
->nl_dvp
, &vp
,
193 /* locked vnode is returned */
195 if (fmode
& O_EXCL
) {
198 error
= cache_vget(&nd
->nl_nch
, cred
,
206 error
= cache_vget(&nd
->nl_nch
, cred
, LK_EXCLUSIVE
, &vp
);
212 * We have a locked vnode and ncp now. Note that the ncp will
213 * be cleaned up by the caller if nd->nl_nch is left intact.
215 if (vp
->v_type
== VLNK
) {
219 if (vp
->v_type
== VSOCK
) {
223 if ((fmode
& O_CREAT
) == 0) {
224 if (fmode
& (FWRITE
| O_TRUNC
)) {
225 if (vp
->v_type
== VDIR
) {
229 error
= vn_writechk(vp
, &nd
->nl_nch
);
232 * Special stale handling, re-resolve the
235 if (error
== ESTALE
) {
238 cache_setunresolved(&nd
->nl_nch
);
239 error
= cache_resolve(&nd
->nl_nch
, cred
);
247 if (fmode
& O_TRUNC
) {
248 vn_unlock(vp
); /* XXX */
249 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
); /* XXX */
252 error
= VOP_SETATTR(vp
, vap
, cred
);
258 * Setup the fp so VOP_OPEN can override it. No descriptor has been
259 * associated with the fp yet so we own it clean.
261 * f_nchandle inherits nl_nch. This used to be necessary only for
262 * directories but now we do it unconditionally so f*() ops
263 * such as fchmod() can access the actual namespace that was
264 * used to open the file.
267 if (nd
->nl_flags
& NLC_APPENDONLY
)
268 fmode
|= FAPPENDONLY
;
269 fp
->f_nchandle
= nd
->nl_nch
;
270 cache_zero(&nd
->nl_nch
);
271 cache_unlock(&fp
->f_nchandle
);
275 * Get rid of nl_nch. vn_open does not return it (it returns the
276 * vnode or the file pointer). Note: we can't leave nl_nch locked
277 * through the VOP_OPEN anyway since the VOP_OPEN may block, e.g.
281 cache_put(&nd
->nl_nch
);
283 error
= VOP_OPEN(vp
, fmode
, cred
, fp
);
286 * setting f_ops to &badfileops will prevent the descriptor
287 * code from trying to close and release the vnode, since
288 * the open failed we do not want to call close.
292 fp
->f_ops
= &badfileops
;
299 * Assert that VREG files have been setup for vmio.
301 KASSERT(vp
->v_type
!= VREG
|| vp
->v_object
!= NULL
,
302 ("vn_open: regular file was not VMIO enabled!"));
306 * Return the vnode. XXX needs some cleaning up. The vnode is
307 * only returned in the fp == NULL case.
311 nd
->nl_vp_fmode
= fmode
;
312 if ((nd
->nl_flags
& NLC_LOCKVP
) == 0)
325 vn_opendisk(const char *devname
, int fmode
, struct vnode
**vpp
)
330 if (strncmp(devname
, "/dev/", 5) == 0)
332 if ((vp
= getsynthvnode(devname
)) == NULL
) {
335 error
= VOP_OPEN(vp
, fmode
, proc0
.p_ucred
, NULL
);
347 * Check for write permissions on the specified vnode. nch may be NULL.
350 vn_writechk(struct vnode
*vp
, struct nchandle
*nch
)
353 * If there's shared text associated with
354 * the vnode, try to free it up once. If
355 * we fail, we can't allow writing.
357 if (vp
->v_flag
& VTEXT
)
361 * If the vnode represents a regular file, check the mount
362 * point via the nch. This may be a different mount point
363 * then the one embedded in the vnode (e.g. nullfs).
365 * We can still write to non-regular files (e.g. devices)
366 * via read-only mounts.
368 if (nch
&& nch
->ncp
&& vp
->v_type
== VREG
)
369 return (ncp_writechk(nch
));
374 * Check whether the underlying mount is read-only. The mount point
375 * referenced by the namecache may be different from the mount point
376 * used by the underlying vnode in the case of NULLFS, so a separate
380 ncp_writechk(struct nchandle
*nch
)
382 if (nch
->mount
&& (nch
->mount
->mnt_flag
& MNT_RDONLY
))
391 vn_close(struct vnode
*vp
, int flags
)
395 error
= vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
397 error
= VOP_CLOSE(vp
, flags
);
406 sequential_heuristic(struct uio
*uio
, struct file
*fp
)
409 * Sequential heuristic - detect sequential operation
411 * NOTE: SMP: We allow f_seqcount updates to race.
413 if ((uio
->uio_offset
== 0 && fp
->f_seqcount
> 0) ||
414 uio
->uio_offset
== fp
->f_nextoff
) {
415 int tmpseq
= fp
->f_seqcount
;
417 * XXX we assume that the filesystem block size is
418 * the default. Not true, but still gives us a pretty
419 * good indicator of how sequential the read operations
422 tmpseq
+= (uio
->uio_resid
+ BKVASIZE
- 1) / BKVASIZE
;
423 if (tmpseq
> IO_SEQMAX
)
425 fp
->f_seqcount
= tmpseq
;
426 return(fp
->f_seqcount
<< IO_SEQSHIFT
);
430 * Not sequential, quick draw-down of seqcount
432 * NOTE: SMP: We allow f_seqcount updates to race.
434 if (fp
->f_seqcount
> 1)
442 * get - lock and return the f_offset field.
443 * set - set and unlock the f_offset field.
445 * These routines serve the dual purpose of serializing access to the
446 * f_offset field (at least on i386) and guaranteeing operational integrity
447 * when multiple read()ers and write()ers are present on the same fp.
449 static __inline off_t
450 vn_get_fpf_offset(struct file
*fp
)
456 * Shortcut critical path.
458 flags
= fp
->f_flag
& ~FOFFSETLOCK
;
459 if (atomic_cmpset_int(&fp
->f_flag
, flags
, flags
| FOFFSETLOCK
))
460 return(fp
->f_offset
);
467 if (flags
& FOFFSETLOCK
) {
468 nflags
= flags
| FOFFSETWAKE
;
469 tsleep_interlock(&fp
->f_flag
, 0);
470 if (atomic_cmpset_int(&fp
->f_flag
, flags
, nflags
))
471 tsleep(&fp
->f_flag
, PINTERLOCKED
, "fpoff", 0);
473 nflags
= flags
| FOFFSETLOCK
;
474 if (atomic_cmpset_int(&fp
->f_flag
, flags
, nflags
))
478 return(fp
->f_offset
);
482 vn_set_fpf_offset(struct file
*fp
, off_t offset
)
488 * We hold the lock so we can set the offset without interference.
490 fp
->f_offset
= offset
;
493 * Normal release is already a reasonably critical path.
497 nflags
= flags
& ~(FOFFSETLOCK
| FOFFSETWAKE
);
498 if (atomic_cmpset_int(&fp
->f_flag
, flags
, nflags
)) {
499 if (flags
& FOFFSETWAKE
)
506 static __inline off_t
507 vn_poll_fpf_offset(struct file
*fp
)
509 #if defined(__x86_64__) || !defined(SMP)
510 return(fp
->f_offset
);
512 off_t off
= vn_get_fpf_offset(fp
);
513 vn_set_fpf_offset(fp
, off
);
519 * Package up an I/O request on a vnode into a uio and do it.
522 vn_rdwr(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
523 off_t offset
, enum uio_seg segflg
, int ioflg
,
524 struct ucred
*cred
, int *aresid
)
528 struct ccms_lock ccms_lock
;
531 if ((ioflg
& IO_NODELOCKED
) == 0)
532 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
533 auio
.uio_iov
= &aiov
;
535 aiov
.iov_base
= base
;
537 auio
.uio_resid
= len
;
538 auio
.uio_offset
= offset
;
539 auio
.uio_segflg
= segflg
;
541 auio
.uio_td
= curthread
;
542 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, &auio
);
543 if (rw
== UIO_READ
) {
544 error
= VOP_READ(vp
, &auio
, ioflg
, cred
);
546 error
= VOP_WRITE(vp
, &auio
, ioflg
, cred
);
548 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
550 *aresid
= auio
.uio_resid
;
552 if (auio
.uio_resid
&& error
== 0)
554 if ((ioflg
& IO_NODELOCKED
) == 0)
560 * Package up an I/O request on a vnode into a uio and do it. The I/O
561 * request is split up into smaller chunks and we try to avoid saturating
562 * the buffer cache while potentially holding a vnode locked, so we
563 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
564 * to give other processes a chance to lock the vnode (either other processes
565 * core'ing the same binary, or unrelated processes scanning the directory).
568 vn_rdwr_inchunks(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
569 off_t offset
, enum uio_seg segflg
, int ioflg
,
570 struct ucred
*cred
, int *aresid
)
578 * Force `offset' to a multiple of MAXBSIZE except possibly
579 * for the first chunk, so that filesystems only need to
580 * write full blocks except possibly for the first and last
583 chunk
= MAXBSIZE
- (uoff_t
)offset
% MAXBSIZE
;
587 if (vp
->v_type
== VREG
) {
597 error
= vn_rdwr(rw
, vp
, base
, chunk
, offset
, segflg
,
598 ioflg
, cred
, aresid
);
599 len
-= chunk
; /* aresid calc already includes length */
612 * MPALMOSTSAFE - acquires mplock
614 * File pointers can no longer get ripped up by revoke so
615 * we don't need to lock access to the vp.
617 * f_offset updates are not guaranteed against multiple readers
620 vn_read(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
622 struct ccms_lock ccms_lock
;
626 KASSERT(uio
->uio_td
== curthread
,
627 ("uio_td %p is not td %p", uio
->uio_td
, curthread
));
628 vp
= (struct vnode
*)fp
->f_data
;
631 if (flags
& O_FBLOCKING
) {
632 /* ioflag &= ~IO_NDELAY; */
633 } else if (flags
& O_FNONBLOCKING
) {
635 } else if (fp
->f_flag
& FNONBLOCK
) {
638 if (flags
& O_FBUFFERED
) {
639 /* ioflag &= ~IO_DIRECT; */
640 } else if (flags
& O_FUNBUFFERED
) {
642 } else if (fp
->f_flag
& O_DIRECT
) {
645 if ((flags
& O_FOFFSET
) == 0 && (vp
->v_flag
& VNOTSEEKABLE
) == 0)
646 uio
->uio_offset
= vn_get_fpf_offset(fp
);
647 vn_lock(vp
, LK_SHARED
| LK_RETRY
);
648 ioflag
|= sequential_heuristic(uio
, fp
);
650 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, uio
);
651 if (read_mpsafe
&& (vp
->v_flag
& VMP_READ
)) {
652 error
= VOP_READ(vp
, uio
, ioflag
, cred
);
655 error
= VOP_READ(vp
, uio
, ioflag
, cred
);
658 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
659 fp
->f_nextoff
= uio
->uio_offset
;
661 if ((flags
& O_FOFFSET
) == 0 && (vp
->v_flag
& VNOTSEEKABLE
) == 0)
662 vn_set_fpf_offset(fp
, uio
->uio_offset
);
667 * MPALMOSTSAFE - acquires mplock
670 vn_write(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
672 struct ccms_lock ccms_lock
;
676 KASSERT(uio
->uio_td
== curthread
,
677 ("uio_td %p is not p %p", uio
->uio_td
, curthread
));
678 vp
= (struct vnode
*)fp
->f_data
;
681 if (vp
->v_type
== VREG
&&
682 ((fp
->f_flag
& O_APPEND
) || (flags
& O_FAPPEND
))) {
686 if (flags
& O_FBLOCKING
) {
687 /* ioflag &= ~IO_NDELAY; */
688 } else if (flags
& O_FNONBLOCKING
) {
690 } else if (fp
->f_flag
& FNONBLOCK
) {
693 if (flags
& O_FBUFFERED
) {
694 /* ioflag &= ~IO_DIRECT; */
695 } else if (flags
& O_FUNBUFFERED
) {
697 } else if (fp
->f_flag
& O_DIRECT
) {
700 if (flags
& O_FASYNCWRITE
) {
701 /* ioflag &= ~IO_SYNC; */
702 } else if (flags
& O_FSYNCWRITE
) {
704 } else if (fp
->f_flag
& O_FSYNC
) {
708 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_SYNCHRONOUS
))
710 if ((flags
& O_FOFFSET
) == 0)
711 uio
->uio_offset
= vn_get_fpf_offset(fp
);
712 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
713 ioflag
|= sequential_heuristic(uio
, fp
);
714 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, uio
);
715 if (write_mpsafe
&& (vp
->v_flag
& VMP_WRITE
)) {
716 error
= VOP_WRITE(vp
, uio
, ioflag
, cred
);
719 error
= VOP_WRITE(vp
, uio
, ioflag
, cred
);
722 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
723 fp
->f_nextoff
= uio
->uio_offset
;
725 if ((flags
& O_FOFFSET
) == 0)
726 vn_set_fpf_offset(fp
, uio
->uio_offset
);
734 vn_statfile(struct file
*fp
, struct stat
*sb
, struct ucred
*cred
)
739 vp
= (struct vnode
*)fp
->f_data
;
740 error
= vn_stat(vp
, sb
, cred
);
745 * MPSAFE (if vnode has VMP_GETATTR)
748 vn_stat(struct vnode
*vp
, struct stat
*sb
, struct ucred
*cred
)
757 if (getattr_mpsafe
&& (vp
->v_flag
& VMP_GETATTR
)) {
758 error
= VOP_GETATTR(vp
, vap
);
761 error
= VOP_GETATTR(vp
, vap
);
768 * Zero the spare stat fields
775 * Copy from vattr table
777 if (vap
->va_fsid
!= VNOVAL
)
778 sb
->st_dev
= vap
->va_fsid
;
780 sb
->st_dev
= vp
->v_mount
->mnt_stat
.f_fsid
.val
[0];
781 sb
->st_ino
= vap
->va_fileid
;
783 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 if (vap
->va_nlink
> (nlink_t
)-1)
818 sb
->st_nlink
= (nlink_t
)-1;
820 sb
->st_nlink
= vap
->va_nlink
;
821 sb
->st_uid
= vap
->va_uid
;
822 sb
->st_gid
= vap
->va_gid
;
823 sb
->st_rdev
= dev2udev(vp
->v_rdev
);
824 sb
->st_size
= vap
->va_size
;
825 sb
->st_atimespec
= vap
->va_atime
;
826 sb
->st_mtimespec
= vap
->va_mtime
;
827 sb
->st_ctimespec
= vap
->va_ctime
;
830 * A VCHR and VBLK device may track the last access and last modified
831 * time independantly of the filesystem. This is particularly true
832 * because device read and write calls may bypass the filesystem.
834 if (vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
) {
837 if (dev
->si_lastread
) {
838 sb
->st_atimespec
.tv_sec
= dev
->si_lastread
;
839 sb
->st_atimespec
.tv_nsec
= 0;
841 if (dev
->si_lastwrite
) {
842 sb
->st_atimespec
.tv_sec
= dev
->si_lastwrite
;
843 sb
->st_atimespec
.tv_nsec
= 0;
849 * According to www.opengroup.org, the meaning of st_blksize is
850 * "a filesystem-specific preferred I/O block size for this
851 * object. In some filesystem types, this may vary from file
853 * Default to PAGE_SIZE after much discussion.
856 if (vap
->va_type
== VREG
) {
857 sb
->st_blksize
= vap
->va_blocksize
;
858 } else if (vn_isdisk(vp
, NULL
)) {
860 * XXX this is broken. If the device is not yet open (aka
861 * stat() call, aka v_rdev == NULL), how are we supposed
862 * to get a valid block size out of it?
866 sb
->st_blksize
= dev
->si_bsize_best
;
867 if (sb
->st_blksize
< dev
->si_bsize_phys
)
868 sb
->st_blksize
= dev
->si_bsize_phys
;
869 if (sb
->st_blksize
< BLKDEV_IOSIZE
)
870 sb
->st_blksize
= BLKDEV_IOSIZE
;
872 sb
->st_blksize
= PAGE_SIZE
;
875 sb
->st_flags
= vap
->va_flags
;
877 error
= priv_check_cred(cred
, PRIV_VFS_GENERATION
, 0);
881 sb
->st_gen
= (u_int32_t
)vap
->va_gen
;
883 sb
->st_blocks
= vap
->va_bytes
/ S_BLKSIZE
;
888 * MPALMOSTSAFE - acquires mplock
891 vn_ioctl(struct file
*fp
, u_long com
, caddr_t data
, struct ucred
*ucred
,
894 struct vnode
*vp
= ((struct vnode
*)fp
->f_data
);
902 switch (vp
->v_type
) {
905 if (com
== FIONREAD
) {
906 error
= VOP_GETATTR(vp
, &vattr
);
909 size
= vattr
.va_size
;
910 if ((vp
->v_flag
& VNOTSEEKABLE
) == 0)
911 size
-= vn_poll_fpf_offset(fp
);
912 if (size
> 0x7FFFFFFF)
918 if (com
== FIOASYNC
) { /* XXX */
930 if (com
== FIODTYPE
) {
931 if (vp
->v_type
!= VCHR
&& vp
->v_type
!= VBLK
) {
935 *(int *)data
= dev_dflags(vp
->v_rdev
) & D_TYPEMASK
;
939 error
= VOP_IOCTL(vp
, com
, data
, fp
->f_flag
, ucred
, msg
);
940 if (error
== 0 && com
== TIOCSCTTY
) {
941 struct proc
*p
= curthread
->td_proc
;
942 struct session
*sess
;
950 /* Do nothing if reassigning same control tty */
951 if (sess
->s_ttyvp
== vp
) {
956 /* Get rid of reference to old control tty */
970 * MPALMOSTSAFE - acquires mplock
973 vn_poll(struct file
*fp
, int events
, struct ucred
*cred
)
978 error
= VOP_POLL(((struct vnode
*)fp
->f_data
), events
, cred
);
984 * Check that the vnode is still valid, and if so
985 * acquire requested lock.
989 vn_lock(struct vnode
*vp
, int flags
)
991 debug_vn_lock(struct vnode
*vp
, int flags
, const char *filename
, int line
)
998 vp
->filename
= filename
;
1000 error
= debuglockmgr(&vp
->v_lock
, flags
,
1001 "vn_lock", filename
, line
);
1003 error
= lockmgr(&vp
->v_lock
, flags
);
1007 } while (flags
& LK_RETRY
);
1010 * Because we (had better!) have a ref on the vnode, once it
1011 * goes to VRECLAIMED state it will not be recycled until all
1012 * refs go away. So we can just check the flag.
1014 if (error
== 0 && (vp
->v_flag
& VRECLAIMED
)) {
1015 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1022 vn_unlock(struct vnode
*vp
)
1024 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1028 vn_islocked(struct vnode
*vp
)
1030 return (lockstatus(&vp
->v_lock
, curthread
));
1034 * MPALMOSTSAFE - acquires mplock
1037 vn_closefile(struct file
*fp
)
1042 fp
->f_ops
= &badfileops
;
1043 error
= vn_close(((struct vnode
*)fp
->f_data
), fp
->f_flag
);
1049 * MPALMOSTSAFE - acquires mplock
1052 vn_kqfilter(struct file
*fp
, struct knote
*kn
)
1057 error
= VOP_KQFILTER(((struct vnode
*)fp
->f_data
), kn
);