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. 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.
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
34 * @(#)vfs_vnops.c 8.2 (Berkeley) 1/21/94
35 * $FreeBSD: src/sys/kern/vfs_vnops.c,v 1.87.2.13 2002/12/29 18:19:53 dillon Exp $
38 #include <sys/param.h>
39 #include <sys/systm.h>
41 #include <sys/fcntl.h>
46 #include <sys/mount.h>
47 #include <sys/nlookup.h>
48 #include <sys/vnode.h>
50 #include <sys/filio.h>
51 #include <sys/ttycom.h>
53 #include <sys/sysctl.h>
54 #include <sys/syslog.h>
56 #include <sys/mplock2.h>
58 static int vn_closefile (struct file
*fp
);
59 static int vn_ioctl (struct file
*fp
, u_long com
, caddr_t data
,
60 struct ucred
*cred
, struct sysmsg
*msg
);
61 static int vn_read (struct file
*fp
, struct uio
*uio
,
62 struct ucred
*cred
, int flags
);
63 static int vn_kqfilter (struct file
*fp
, struct knote
*kn
);
64 static int vn_statfile (struct file
*fp
, struct stat
*sb
, struct ucred
*cred
);
65 static int vn_write (struct file
*fp
, struct uio
*uio
,
66 struct ucred
*cred
, int flags
);
68 struct fileops vnode_fileops
= {
72 .fo_kqfilter
= vn_kqfilter
,
73 .fo_stat
= vn_statfile
,
74 .fo_close
= vn_closefile
,
75 .fo_shutdown
= nofo_shutdown
79 * Common code for vnode open operations. Check permissions, and call
80 * the VOP_NOPEN or VOP_NCREATE routine.
82 * The caller is responsible for setting up nd with nlookup_init() and
83 * for cleaning it up with nlookup_done(), whether we return an error
86 * On success nd->nl_open_vp will hold a referenced and, if requested,
87 * locked vnode. A locked vnode is requested via NLC_LOCKVP. If fp
88 * is non-NULL the vnode will be installed in the file pointer.
90 * NOTE: If the caller wishes the namecache entry to be operated with
91 * a shared lock it must use NLC_SHAREDLOCK. If NLC_LOCKVP is set
92 * then the vnode lock will also be shared.
94 * NOTE: The vnode is referenced just once on return whether or not it
95 * is also installed in the file pointer.
98 vn_open(struct nlookupdata
*nd
, struct file
*fp
, int fmode
, int cmode
)
101 struct ucred
*cred
= nd
->nl_cred
;
103 struct vattr
*vap
= &vat
;
111 * Certain combinations are illegal
113 if ((fmode
& (FWRITE
| O_TRUNC
)) == O_TRUNC
)
117 * Lookup the path and create or obtain the vnode. After a
118 * successful lookup a locked nd->nl_nch will be returned.
120 * The result of this section should be a locked vnode.
122 * XXX with only a little work we should be able to avoid locking
123 * the vnode if FWRITE, O_CREAT, and O_TRUNC are *not* set.
125 nd
->nl_flags
|= NLC_OPEN
;
126 if (fmode
& O_APPEND
)
127 nd
->nl_flags
|= NLC_APPEND
;
129 nd
->nl_flags
|= NLC_TRUNCATE
;
131 nd
->nl_flags
|= NLC_READ
;
133 nd
->nl_flags
|= NLC_WRITE
;
134 if ((fmode
& O_EXCL
) == 0 && (fmode
& O_NOFOLLOW
) == 0)
135 nd
->nl_flags
|= NLC_FOLLOW
;
137 if (fmode
& O_CREAT
) {
139 * CONDITIONAL CREATE FILE CASE
141 * Setting NLC_CREATE causes a negative hit to store
142 * the negative hit ncp and not return an error. Then
143 * nc_error or nc_vp may be checked to see if the ncp
144 * represents a negative hit. NLC_CREATE also requires
145 * write permission on the governing directory or EPERM
148 nd
->nl_flags
|= NLC_CREATE
;
149 nd
->nl_flags
|= NLC_REFDVP
;
154 * NORMAL OPEN FILE CASE
163 * split case to allow us to re-resolve and retry the ncp in case
166 * (error is 0 on entry / retry)
170 * Checks for (likely) filesystem-modifying cases and allows
171 * the filesystem to stall the front-end.
173 if ((fmode
& (FWRITE
| O_TRUNC
)) ||
174 ((fmode
& O_CREAT
) && nd
->nl_nch
.ncp
->nc_vp
== NULL
)) {
175 error
= ncp_writechk(&nd
->nl_nch
);
181 if (fmode
& O_CREAT
) {
182 if (nd
->nl_nch
.ncp
->nc_vp
== NULL
) {
185 vap
->va_mode
= cmode
;
186 vap
->va_fuseflags
= fmode
; /* FUSE */
188 vap
->va_vaflags
|= VA_EXCLUSIVE
;
189 error
= VOP_NCREATE(&nd
->nl_nch
, nd
->nl_dvp
, &vp
,
194 /* locked vnode is returned */
196 if (fmode
& O_EXCL
) {
199 error
= cache_vget(&nd
->nl_nch
, cred
,
208 * In most other cases a shared lock on the vnode is
209 * sufficient. However, the O_RDWR case needs an
210 * exclusive lock if the vnode is executable. The
211 * NLC_EXCLLOCK_IFEXEC and NCF_NOTX flags help resolve
214 * NOTE: If NCF_NOTX is not set, we do not know the
215 * the state of the 'x' bits and have to get
216 * an exclusive lock for the EXCLLOCK_IFEXEC case.
218 if ((nd
->nl_flags
& NLC_SHAREDLOCK
) &&
219 ((nd
->nl_flags
& NLC_EXCLLOCK_IFEXEC
) == 0 ||
220 nd
->nl_nch
.ncp
->nc_flag
& NCF_NOTX
)) {
221 error
= cache_vget(&nd
->nl_nch
, cred
, LK_SHARED
, &vp
);
224 error
= cache_vget(&nd
->nl_nch
, cred
,
232 * We have a locked vnode and ncp now. Note that the ncp will
233 * be cleaned up by the caller if nd->nl_nch is left intact.
235 if (vp
->v_type
== VLNK
) {
239 if (vp
->v_type
== VSOCK
) {
243 if (vp
->v_type
!= VDIR
&& (fmode
& O_DIRECTORY
)) {
247 if ((fmode
& O_CREAT
) == 0) {
248 if (fmode
& (FWRITE
| O_TRUNC
)) {
249 if (vp
->v_type
== VDIR
) {
255 * Additional checks on vnode (does not substitute
256 * for ncp_writechk()).
258 error
= vn_writechk(vp
);
261 * Special stale handling, re-resolve the
264 if (error
== ESTALE
) {
268 cache_unlock(&nd
->nl_nch
);
269 cache_lock(&nd
->nl_nch
);
271 cache_setunresolved(&nd
->nl_nch
);
272 error
= cache_resolve(&nd
->nl_nch
,
281 if (fmode
& O_TRUNC
) {
282 vn_unlock(vp
); /* XXX */
283 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
); /* XXX */
284 osize
= vp
->v_filesize
;
287 error
= VOP_SETATTR_FP(vp
, vap
, cred
, fp
);
290 error
= VOP_GETATTR(vp
, vap
);
294 VFS_ACCOUNT(mp
, vap
->va_uid
, vap
->va_gid
, -osize
);
298 * Set or clear VNSWAPCACHE on the vp based on nd->nl_nch.ncp->nc_flag.
299 * These particular bits a tracked all the way from the root.
301 * NOTE: Might not work properly on NFS servers due to the
302 * disconnected namecache.
304 flags
= nd
->nl_nch
.ncp
->nc_flag
;
305 if ((flags
& (NCF_UF_CACHE
| NCF_UF_PCACHE
)) &&
306 (flags
& (NCF_SF_NOCACHE
| NCF_SF_PNOCACHE
)) == 0) {
307 vsetflags(vp
, VSWAPCACHE
);
309 vclrflags(vp
, VSWAPCACHE
);
313 * Setup the fp so VOP_OPEN can override it. No descriptor has been
314 * associated with the fp yet so we own it clean.
316 * f_nchandle inherits nl_nch. This used to be necessary only for
317 * directories but now we do it unconditionally so f*() ops
318 * such as fchmod() can access the actual namespace that was
319 * used to open the file.
322 if (nd
->nl_flags
& NLC_APPENDONLY
)
323 fmode
|= FAPPENDONLY
;
324 fp
->f_nchandle
= nd
->nl_nch
;
325 cache_zero(&nd
->nl_nch
);
326 cache_unlock(&fp
->f_nchandle
);
330 * Get rid of nl_nch. vn_open does not return it (it returns the
331 * vnode or the file pointer). Note: we can't leave nl_nch locked
332 * through the VOP_OPEN anyway since the VOP_OPEN may block, e.g.
336 cache_put(&nd
->nl_nch
);
338 error
= VOP_OPEN(vp
, fmode
, cred
, fp
);
341 * setting f_ops to &badfileops will prevent the descriptor
342 * code from trying to close and release the vnode, since
343 * the open failed we do not want to call close.
347 fp
->f_ops
= &badfileops
;
354 * Assert that VREG files have been setup for vmio.
356 KASSERT(vp
->v_type
!= VREG
|| vp
->v_object
!= NULL
,
357 ("vn_open: regular file was not VMIO enabled!"));
361 * Return the vnode. XXX needs some cleaning up. The vnode is
362 * only returned in the fp == NULL case.
366 nd
->nl_vp_fmode
= fmode
;
367 if ((nd
->nl_flags
& NLC_LOCKVP
) == 0)
380 vn_opendisk(const char *devname
, int fmode
, struct vnode
**vpp
)
385 if (strncmp(devname
, "/dev/", 5) == 0)
387 if ((vp
= getsynthvnode(devname
)) == NULL
) {
390 error
= VOP_OPEN(vp
, fmode
, proc0
.p_ucred
, NULL
);
402 * Checks for special conditions on the vnode which might prevent writing
403 * after the vnode has (likely) been locked. The vnode might or might not
404 * be locked as of this call, but will be at least referenced.
406 * Also re-checks the mount RDONLY flag that ncp_writechk() checked prior
407 * to the vnode being locked.
410 vn_writechk(struct vnode
*vp
)
413 * If there's shared text associated with
414 * the vnode, try to free it up once. If
415 * we fail, we can't allow writing.
417 if (vp
->v_flag
& VTEXT
)
419 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_RDONLY
))
425 * Check whether the underlying mount is read-only. The mount point
426 * referenced by the namecache may be different from the mount point
427 * used by the underlying vnode in the case of NULLFS, so a separate
430 * Must be called PRIOR to any vnodes being locked.
433 ncp_writechk(struct nchandle
*nch
)
437 if ((mp
= nch
->mount
) != NULL
) {
438 if (mp
->mnt_flag
& MNT_RDONLY
)
440 if (mp
->mnt_op
->vfs_modifying
!= vfs_stdmodifying
)
452 vn_close(struct vnode
*vp
, int flags
, struct file
*fp
)
456 error
= vn_lock(vp
, LK_SHARED
| LK_RETRY
| LK_FAILRECLAIM
);
458 error
= VOP_CLOSE(vp
, flags
, fp
);
466 * Sequential heuristic.
468 * MPSAFE (f_seqcount and f_nextoff are allowed to race)
472 sequential_heuristic(struct uio
*uio
, struct file
*fp
)
475 * Sequential heuristic - detect sequential operation
477 * NOTE: SMP: We allow f_seqcount updates to race.
479 if ((uio
->uio_offset
== 0 && fp
->f_seqcount
> 0) ||
480 uio
->uio_offset
== fp
->f_nextoff
) {
481 int tmpseq
= fp
->f_seqcount
;
483 tmpseq
+= (uio
->uio_resid
+ MAXBSIZE
- 1) / MAXBSIZE
;
484 if (tmpseq
> IO_SEQMAX
)
486 fp
->f_seqcount
= tmpseq
;
487 return(fp
->f_seqcount
<< IO_SEQSHIFT
);
491 * Not sequential, quick draw-down of seqcount
493 * NOTE: SMP: We allow f_seqcount updates to race.
495 if (fp
->f_seqcount
> 1)
503 * get - lock and return the f_offset field.
504 * set - set and unlock the f_offset field.
506 * These routines serve the dual purpose of serializing access to the
507 * f_offset field (at least on x86) and guaranteeing operational integrity
508 * when multiple read()ers and write()ers are present on the same fp.
512 static __inline off_t
513 vn_get_fpf_offset(struct file
*fp
)
519 * Shortcut critical path.
521 flags
= fp
->f_flag
& ~FOFFSETLOCK
;
522 if (atomic_cmpset_int(&fp
->f_flag
, flags
, flags
| FOFFSETLOCK
))
523 return(fp
->f_offset
);
530 if (flags
& FOFFSETLOCK
) {
531 nflags
= flags
| FOFFSETWAKE
;
532 tsleep_interlock(&fp
->f_flag
, 0);
533 if (atomic_cmpset_int(&fp
->f_flag
, flags
, nflags
))
534 tsleep(&fp
->f_flag
, PINTERLOCKED
, "fpoff", 0);
536 nflags
= flags
| FOFFSETLOCK
;
537 if (atomic_cmpset_int(&fp
->f_flag
, flags
, nflags
))
541 return(fp
->f_offset
);
548 vn_set_fpf_offset(struct file
*fp
, off_t offset
)
554 * We hold the lock so we can set the offset without interference.
556 fp
->f_offset
= offset
;
559 * Normal release is already a reasonably critical path.
563 nflags
= flags
& ~(FOFFSETLOCK
| FOFFSETWAKE
);
564 if (atomic_cmpset_int(&fp
->f_flag
, flags
, nflags
)) {
565 if (flags
& FOFFSETWAKE
)
575 static __inline off_t
576 vn_poll_fpf_offset(struct file
*fp
)
578 #if defined(__x86_64__)
579 return(fp
->f_offset
);
581 off_t off
= vn_get_fpf_offset(fp
);
582 vn_set_fpf_offset(fp
, off
);
588 * Package up an I/O request on a vnode into a uio and do it.
593 vn_rdwr(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
594 off_t offset
, enum uio_seg segflg
, int ioflg
,
595 struct ucred
*cred
, int *aresid
)
601 if ((ioflg
& IO_NODELOCKED
) == 0)
602 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
603 auio
.uio_iov
= &aiov
;
605 aiov
.iov_base
= base
;
607 auio
.uio_resid
= len
;
608 auio
.uio_offset
= offset
;
609 auio
.uio_segflg
= segflg
;
611 auio
.uio_td
= curthread
;
612 if (rw
== UIO_READ
) {
613 error
= VOP_READ(vp
, &auio
, ioflg
, cred
);
615 error
= VOP_WRITE(vp
, &auio
, ioflg
, cred
);
618 *aresid
= auio
.uio_resid
;
620 if (auio
.uio_resid
&& error
== 0)
622 if ((ioflg
& IO_NODELOCKED
) == 0)
628 * Package up an I/O request on a vnode into a uio and do it. The I/O
629 * request is split up into smaller chunks and we try to avoid saturating
630 * the buffer cache while potentially holding a vnode locked, so we
631 * check bwillwrite() before calling vn_rdwr(). We also call lwkt_user_yield()
632 * to give other processes a chance to lock the vnode (either other processes
633 * core'ing the same binary, or unrelated processes scanning the directory).
638 vn_rdwr_inchunks(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
639 off_t offset
, enum uio_seg segflg
, int ioflg
,
640 struct ucred
*cred
, int *aresid
)
648 * Force `offset' to a multiple of MAXBSIZE except possibly
649 * for the first chunk, so that filesystems only need to
650 * write full blocks except possibly for the first and last
653 chunk
= MAXBSIZE
- (uoff_t
)offset
% MAXBSIZE
;
657 if (vp
->v_type
== VREG
&& (ioflg
& IO_RECURSE
) == 0) {
667 error
= vn_rdwr(rw
, vp
, base
, chunk
, offset
, segflg
,
668 ioflg
, cred
, aresid
);
669 len
-= chunk
; /* aresid calc already includes length */
682 * File pointers can no longer get ripped up by revoke so
683 * we don't need to lock access to the vp.
685 * f_offset updates are not guaranteed against multiple readers
688 vn_read(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
693 KASSERT(uio
->uio_td
== curthread
,
694 ("uio_td %p is not td %p", uio
->uio_td
, curthread
));
695 vp
= (struct vnode
*)fp
->f_data
;
698 if (flags
& O_FBLOCKING
) {
699 /* ioflag &= ~IO_NDELAY; */
700 } else if (flags
& O_FNONBLOCKING
) {
702 } else if (fp
->f_flag
& FNONBLOCK
) {
705 if (fp
->f_flag
& O_DIRECT
) {
708 if ((flags
& O_FOFFSET
) == 0 && (vp
->v_flag
& VNOTSEEKABLE
) == 0)
709 uio
->uio_offset
= vn_get_fpf_offset(fp
);
710 vn_lock(vp
, LK_SHARED
| LK_RETRY
);
711 ioflag
|= sequential_heuristic(uio
, fp
);
713 error
= VOP_READ_FP(vp
, uio
, ioflag
, cred
, fp
);
714 fp
->f_nextoff
= uio
->uio_offset
;
716 if ((flags
& O_FOFFSET
) == 0 && (vp
->v_flag
& VNOTSEEKABLE
) == 0)
717 vn_set_fpf_offset(fp
, uio
->uio_offset
);
725 vn_write(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
730 KASSERT(uio
->uio_td
== curthread
,
731 ("uio_td %p is not p %p", uio
->uio_td
, curthread
));
732 vp
= (struct vnode
*)fp
->f_data
;
735 if (vp
->v_type
== VREG
&&
736 ((fp
->f_flag
& O_APPEND
) || (flags
& O_FAPPEND
))) {
740 if (flags
& O_FBLOCKING
) {
741 /* ioflag &= ~IO_NDELAY; */
742 } else if (flags
& O_FNONBLOCKING
) {
744 } else if (fp
->f_flag
& FNONBLOCK
) {
747 if (fp
->f_flag
& O_DIRECT
) {
750 if (flags
& O_FASYNCWRITE
) {
751 /* ioflag &= ~IO_SYNC; */
752 } else if (flags
& O_FSYNCWRITE
) {
754 } else if (fp
->f_flag
& O_FSYNC
) {
758 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_SYNCHRONOUS
))
760 if ((flags
& O_FOFFSET
) == 0)
761 uio
->uio_offset
= vn_get_fpf_offset(fp
);
763 VFS_MODIFYING(vp
->v_mount
);
764 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
765 ioflag
|= sequential_heuristic(uio
, fp
);
766 error
= VOP_WRITE_FP(vp
, uio
, ioflag
, cred
, fp
);
767 fp
->f_nextoff
= uio
->uio_offset
;
769 if ((flags
& O_FOFFSET
) == 0)
770 vn_set_fpf_offset(fp
, uio
->uio_offset
);
778 vn_statfile(struct file
*fp
, struct stat
*sb
, struct ucred
*cred
)
783 vp
= (struct vnode
*)fp
->f_data
;
784 error
= vn_stat(vp
, sb
, cred
);
792 vn_stat(struct vnode
*vp
, struct stat
*sb
, struct ucred
*cred
)
801 * vp already has a ref and is validated, can call unlocked.
804 error
= VOP_GETATTR(vp
, vap
);
809 * Zero the spare stat fields
815 * Copy from vattr table
817 if (vap
->va_fsid
!= VNOVAL
)
818 sb
->st_dev
= vap
->va_fsid
;
820 sb
->st_dev
= vp
->v_mount
->mnt_stat
.f_fsid
.val
[0];
821 sb
->st_ino
= vap
->va_fileid
;
823 switch (vap
->va_type
) {
841 /* This is a cosmetic change, symlinks do not have a mode. */
842 if (vp
->v_mount
->mnt_flag
& MNT_NOSYMFOLLOW
)
843 sb
->st_mode
&= ~ACCESSPERMS
; /* 0000 */
845 sb
->st_mode
|= ACCESSPERMS
; /* 0777 */
857 if (vap
->va_nlink
> (nlink_t
)-1)
858 sb
->st_nlink
= (nlink_t
)-1;
860 sb
->st_nlink
= vap
->va_nlink
;
861 sb
->st_uid
= vap
->va_uid
;
862 sb
->st_gid
= vap
->va_gid
;
863 sb
->st_rdev
= devid_from_dev(vp
->v_rdev
);
864 sb
->st_size
= vap
->va_size
;
865 sb
->st_atimespec
= vap
->va_atime
;
866 sb
->st_mtimespec
= vap
->va_mtime
;
867 sb
->st_ctimespec
= vap
->va_ctime
;
870 * A VCHR and VBLK device may track the last access and last modified
871 * time independantly of the filesystem. This is particularly true
872 * because device read and write calls may bypass the filesystem.
874 if (vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
) {
877 if (dev
->si_lastread
) {
878 sb
->st_atimespec
.tv_sec
= time_second
+
881 sb
->st_atimespec
.tv_nsec
= 0;
883 if (dev
->si_lastwrite
) {
884 sb
->st_atimespec
.tv_sec
= time_second
+
887 sb
->st_atimespec
.tv_nsec
= 0;
893 * According to www.opengroup.org, the meaning of st_blksize is
894 * "a filesystem-specific preferred I/O block size for this
895 * object. In some filesystem types, this may vary from file
897 * Default to PAGE_SIZE after much discussion.
900 if (vap
->va_type
== VREG
) {
901 sb
->st_blksize
= vap
->va_blocksize
;
902 } else if (vn_isdisk(vp
, NULL
)) {
904 * XXX this is broken. If the device is not yet open (aka
905 * stat() call, aka v_rdev == NULL), how are we supposed
906 * to get a valid block size out of it?
910 sb
->st_blksize
= dev
->si_bsize_best
;
911 if (sb
->st_blksize
< dev
->si_bsize_phys
)
912 sb
->st_blksize
= dev
->si_bsize_phys
;
913 if (sb
->st_blksize
< BLKDEV_IOSIZE
)
914 sb
->st_blksize
= BLKDEV_IOSIZE
;
916 sb
->st_blksize
= PAGE_SIZE
;
919 sb
->st_flags
= vap
->va_flags
;
921 error
= priv_check_cred(cred
, PRIV_VFS_GENERATION
, 0);
925 sb
->st_gen
= (u_int32_t
)vap
->va_gen
;
927 sb
->st_blocks
= vap
->va_bytes
/ S_BLKSIZE
;
930 * This is for ABI compatibility <= 5.7 (for ABI change made in
933 sb
->__old_st_blksize
= sb
->st_blksize
;
939 * MPALMOSTSAFE - acquires mplock
942 vn_ioctl(struct file
*fp
, u_long com
, caddr_t data
, struct ucred
*ucred
,
945 struct vnode
*vp
= ((struct vnode
*)fp
->f_data
);
951 switch (vp
->v_type
) {
954 if (com
== FIONREAD
) {
955 error
= VOP_GETATTR(vp
, &vattr
);
958 size
= vattr
.va_size
;
959 if ((vp
->v_flag
& VNOTSEEKABLE
) == 0)
960 size
-= vn_poll_fpf_offset(fp
);
961 if (size
> 0x7FFFFFFF)
967 if (com
== FIOASYNC
) { /* XXX */
979 if (com
== FIODTYPE
) {
980 if (vp
->v_type
!= VCHR
&& vp
->v_type
!= VBLK
) {
984 *(int *)data
= dev_dflags(vp
->v_rdev
) & D_TYPEMASK
;
988 error
= VOP_IOCTL(vp
, com
, data
, fp
->f_flag
, ucred
, msg
);
989 if (error
== 0 && com
== TIOCSCTTY
) {
990 struct proc
*p
= curthread
->td_proc
;
991 struct session
*sess
;
1000 /* Do nothing if reassigning same control tty */
1001 if (sess
->s_ttyvp
== vp
) {
1007 /* Get rid of reference to old control tty */
1008 ovp
= sess
->s_ttyvp
;
1021 * Obtain the requested vnode lock
1023 * LK_RETRY Automatically retry on timeout
1024 * LK_FAILRECLAIM Fail if the vnode is being reclaimed
1026 * Failures will occur if the vnode is undergoing recyclement, but not
1027 * all callers expect that the function will fail so the caller must pass
1028 * LK_FAILOK if it wants to process an error code.
1030 * Errors can occur for other reasons if you pass in other LK_ flags,
1031 * regardless of whether you pass in LK_FAILRECLAIM
1034 vn_lock(struct vnode
*vp
, int flags
)
1039 error
= lockmgr(&vp
->v_lock
, flags
);
1042 } while (flags
& LK_RETRY
);
1045 * Because we (had better!) have a ref on the vnode, once it
1046 * goes to VRECLAIMED state it will not be recycled until all
1047 * refs go away. So we can just check the flag.
1049 if (error
== 0 && (vp
->v_flag
& VRECLAIMED
)) {
1050 if (flags
& LK_FAILRECLAIM
) {
1051 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1059 vn_relock(struct vnode
*vp
, int flags
)
1064 error
= lockmgr(&vp
->v_lock
, flags
);
1067 } while (flags
& LK_RETRY
);
1072 #ifdef DEBUG_VN_UNLOCK
1075 debug_vn_unlock(struct vnode
*vp
, const char *filename
, int line
)
1077 kprintf("vn_unlock from %s:%d\n", filename
, line
);
1078 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1084 vn_unlock(struct vnode
*vp
)
1086 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1095 vn_islocked(struct vnode
*vp
)
1097 return (lockstatus(&vp
->v_lock
, curthread
));
1101 * Return the lock status of a vnode and unlock the vnode
1102 * if we owned the lock. This is not a boolean, if the
1103 * caller cares what the lock status is the caller must
1104 * check the various possible values.
1106 * This only unlocks exclusive locks held by the caller,
1107 * it will NOT unlock shared locks (there is no way to
1108 * tell who the shared lock belongs to).
1113 vn_islocked_unlock(struct vnode
*vp
)
1117 vpls
= lockstatus(&vp
->v_lock
, curthread
);
1118 if (vpls
== LK_EXCLUSIVE
)
1119 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1124 * Restore a vnode lock that we previously released via
1125 * vn_islocked_unlock(). This is a NOP if we did not
1126 * own the original lock.
1131 vn_islocked_relock(struct vnode
*vp
, int vpls
)
1135 if (vpls
== LK_EXCLUSIVE
)
1136 error
= lockmgr(&vp
->v_lock
, vpls
);
1143 vn_closefile(struct file
*fp
)
1147 fp
->f_ops
= &badfileops
;
1148 error
= vn_close(((struct vnode
*)fp
->f_data
), fp
->f_flag
, fp
);
1156 vn_kqfilter(struct file
*fp
, struct knote
*kn
)
1160 error
= VOP_KQFILTER(((struct vnode
*)fp
->f_data
), kn
);