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/syslog.h>
59 static int vn_closefile (struct file
*fp
);
60 static int vn_ioctl (struct file
*fp
, u_long com
, caddr_t data
,
62 static int vn_read (struct file
*fp
, struct uio
*uio
,
63 struct ucred
*cred
, int flags
);
64 static int svn_read (struct file
*fp
, struct uio
*uio
,
65 struct ucred
*cred
, int flags
);
66 static int vn_poll (struct file
*fp
, int events
, struct ucred
*cred
);
67 static int vn_kqfilter (struct file
*fp
, struct knote
*kn
);
68 static int vn_statfile (struct file
*fp
, struct stat
*sb
, struct ucred
*cred
);
69 static int vn_write (struct file
*fp
, struct uio
*uio
,
70 struct ucred
*cred
, int flags
);
71 static int svn_write (struct file
*fp
, struct uio
*uio
,
72 struct ucred
*cred
, int flags
);
74 struct fileops vnode_fileops
= {
79 .fo_kqfilter
= vn_kqfilter
,
80 .fo_stat
= vn_statfile
,
81 .fo_close
= vn_closefile
,
82 .fo_shutdown
= nofo_shutdown
85 struct fileops specvnode_fileops
= {
87 .fo_write
= svn_write
,
90 .fo_kqfilter
= vn_kqfilter
,
91 .fo_stat
= vn_statfile
,
92 .fo_close
= vn_closefile
,
93 .fo_shutdown
= nofo_shutdown
97 * Shortcut the device read/write. This avoids a lot of vnode junk.
98 * Basically the specfs vnops for read and write take the locked vnode,
99 * unlock it (because we can't hold the vnode locked while reading or writing
100 * a device which may block indefinitely), issues the device operation, then
101 * relock the vnode before returning, plus other junk. This bypasses all
102 * of that and just does the device operation.
105 vn_setspecops(struct file
*fp
)
107 if (vfs_fastdev
&& fp
->f_ops
== &vnode_fileops
) {
108 fp
->f_ops
= &specvnode_fileops
;
113 * Common code for vnode open operations. Check permissions, and call
114 * the VOP_NOPEN or VOP_NCREATE routine.
116 * The caller is responsible for setting up nd with nlookup_init() and
117 * for cleaning it up with nlookup_done(), whether we return an error
120 * On success nd->nl_open_vp will hold a referenced and, if requested,
121 * locked vnode. A locked vnode is requested via NLC_LOCKVP. If fp
122 * is non-NULL the vnode will be installed in the file pointer.
124 * NOTE: The vnode is referenced just once on return whether or not it
125 * is also installed in the file pointer.
128 vn_open(struct nlookupdata
*nd
, struct file
*fp
, int fmode
, int cmode
)
131 struct ucred
*cred
= nd
->nl_cred
;
133 struct vattr
*vap
= &vat
;
137 * Lookup the path and create or obtain the vnode. After a
138 * successful lookup a locked nd->nl_nch 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
;
159 nd
->nl_flags
|= NLC_REFDVP
;
164 * 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 (nd
->nl_nch
.ncp
->nc_vp
== NULL
) {
179 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
183 vap
->va_mode
= cmode
;
185 vap
->va_vaflags
|= VA_EXCLUSIVE
;
186 error
= VOP_NCREATE(&nd
->nl_nch
, nd
->nl_dvp
, &vp
,
191 /* locked vnode is returned */
193 if (fmode
& O_EXCL
) {
196 error
= cache_vget(&nd
->nl_nch
, cred
,
204 error
= cache_vget(&nd
->nl_nch
, cred
, LK_EXCLUSIVE
, &vp
);
210 * We have a locked vnode and ncp now. Note that the ncp will
211 * be cleaned up by the caller if nd->nl_nch is left intact.
213 if (vp
->v_type
== VLNK
) {
217 if (vp
->v_type
== VSOCK
) {
221 if ((fmode
& O_CREAT
) == 0) {
223 if (fmode
& (FWRITE
| O_TRUNC
)) {
224 if (vp
->v_type
== VDIR
) {
228 error
= vn_writechk(vp
, &nd
->nl_nch
);
231 * Special stale handling, re-resolve the
234 if (error
== ESTALE
) {
237 cache_setunresolved(&nd
->nl_nch
);
238 error
= cache_resolve(&nd
->nl_nch
, cred
);
249 error
= VOP_ACCESS(vp
, mode
, cred
);
252 * Special stale handling, re-resolve the
255 if (error
== ESTALE
) {
258 cache_setunresolved(&nd
->nl_nch
);
259 error
= cache_resolve(&nd
->nl_nch
, cred
);
267 if (fmode
& O_TRUNC
) {
268 vn_unlock(vp
); /* XXX */
269 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
); /* XXX */
272 error
= VOP_SETATTR(vp
, vap
, cred
);
278 * Setup the fp so VOP_OPEN can override it. No descriptor has been
279 * associated with the fp yet so we own it clean.
281 * f_nchandle inherits nl_nch. This used to be necessary only for
282 * directories but now we do it unconditionally so f*() ops
283 * such as fchmod() can access the actual namespace that was
284 * used to open the file.
287 fp
->f_nchandle
= nd
->nl_nch
;
288 cache_zero(&nd
->nl_nch
);
289 cache_unlock(&fp
->f_nchandle
);
293 * Get rid of nl_nch. vn_open does not return it (it returns the
294 * vnode or the file pointer). Note: we can't leave nl_nch locked
295 * through the VOP_OPEN anyway since the VOP_OPEN may block, e.g.
299 cache_put(&nd
->nl_nch
);
301 error
= VOP_OPEN(vp
, fmode
, cred
, fp
);
304 * setting f_ops to &badfileops will prevent the descriptor
305 * code from trying to close and release the vnode, since
306 * the open failed we do not want to call close.
310 fp
->f_ops
= &badfileops
;
317 * Assert that VREG files have been setup for vmio.
319 KASSERT(vp
->v_type
!= VREG
|| vp
->v_object
!= NULL
,
320 ("vn_open: regular file was not VMIO enabled!"));
324 * Return the vnode. XXX needs some cleaning up. The vnode is
325 * only returned in the fp == NULL case.
329 nd
->nl_vp_fmode
= fmode
;
330 if ((nd
->nl_flags
& NLC_LOCKVP
) == 0)
343 vn_opendisk(const char *devname
, int fmode
, struct vnode
**vpp
)
348 if (strncmp(devname
, "/dev/", 5) == 0)
350 if ((vp
= getsynthvnode(devname
)) == NULL
) {
353 error
= VOP_OPEN(vp
, fmode
, proc0
.p_ucred
, NULL
);
365 * Check for write permissions on the specified vnode. nch may be NULL.
368 vn_writechk(struct vnode
*vp
, struct nchandle
*nch
)
371 * If there's shared text associated with
372 * the vnode, try to free it up once. If
373 * we fail, we can't allow writing.
375 if (vp
->v_flag
& VTEXT
)
379 * If the vnode represents a regular file, check the mount
380 * point via the nch. This may be a different mount point
381 * then the one embedded in the vnode (e.g. nullfs).
383 * We can still write to non-regular files (e.g. devices)
384 * via read-only mounts.
386 if (nch
&& nch
->ncp
&& vp
->v_type
== VREG
)
387 return (ncp_writechk(nch
));
392 * Check whether the underlying mount is read-only. The mount point
393 * referenced by the namecache may be different from the mount point
394 * used by the underlying vnode in the case of NULLFS, so a separate
398 ncp_writechk(struct nchandle
*nch
)
400 if (nch
->mount
&& (nch
->mount
->mnt_flag
& MNT_RDONLY
))
409 vn_close(struct vnode
*vp
, int flags
)
413 error
= vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
415 error
= VOP_CLOSE(vp
, flags
);
424 sequential_heuristic(struct uio
*uio
, struct file
*fp
)
427 * Sequential heuristic - detect sequential operation
429 if ((uio
->uio_offset
== 0 && fp
->f_seqcount
> 0) ||
430 uio
->uio_offset
== fp
->f_nextoff
) {
431 int tmpseq
= fp
->f_seqcount
;
433 * XXX we assume that the filesystem block size is
434 * the default. Not true, but still gives us a pretty
435 * good indicator of how sequential the read operations
438 tmpseq
+= (uio
->uio_resid
+ BKVASIZE
- 1) / BKVASIZE
;
439 if (tmpseq
> IO_SEQMAX
)
441 fp
->f_seqcount
= tmpseq
;
442 return(fp
->f_seqcount
<< IO_SEQSHIFT
);
446 * Not sequential, quick draw-down of seqcount
448 if (fp
->f_seqcount
> 1)
456 * Package up an I/O request on a vnode into a uio and do it.
459 vn_rdwr(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
460 off_t offset
, enum uio_seg segflg
, int ioflg
,
461 struct ucred
*cred
, int *aresid
)
465 struct ccms_lock ccms_lock
;
468 if ((ioflg
& IO_NODELOCKED
) == 0)
469 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
470 auio
.uio_iov
= &aiov
;
472 aiov
.iov_base
= base
;
474 auio
.uio_resid
= len
;
475 auio
.uio_offset
= offset
;
476 auio
.uio_segflg
= segflg
;
478 auio
.uio_td
= curthread
;
479 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, &auio
);
480 if (rw
== UIO_READ
) {
481 error
= VOP_READ(vp
, &auio
, ioflg
, cred
);
483 error
= VOP_WRITE(vp
, &auio
, ioflg
, cred
);
485 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
487 *aresid
= auio
.uio_resid
;
489 if (auio
.uio_resid
&& error
== 0)
491 if ((ioflg
& IO_NODELOCKED
) == 0)
497 * Package up an I/O request on a vnode into a uio and do it. The I/O
498 * request is split up into smaller chunks and we try to avoid saturating
499 * the buffer cache while potentially holding a vnode locked, so we
500 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
501 * to give other processes a chance to lock the vnode (either other processes
502 * core'ing the same binary, or unrelated processes scanning the directory).
505 vn_rdwr_inchunks(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
506 off_t offset
, enum uio_seg segflg
, int ioflg
,
507 struct ucred
*cred
, int *aresid
)
515 * Force `offset' to a multiple of MAXBSIZE except possibly
516 * for the first chunk, so that filesystems only need to
517 * write full blocks except possibly for the first and last
520 chunk
= MAXBSIZE
- (uoff_t
)offset
% MAXBSIZE
;
524 if (vp
->v_type
== VREG
) {
534 error
= vn_rdwr(rw
, vp
, base
, chunk
, offset
, segflg
,
535 ioflg
, cred
, aresid
);
536 len
-= chunk
; /* aresid calc already includes length */
549 * MPALMOSTSAFE - acquires mplock
552 vn_read(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
554 struct ccms_lock ccms_lock
;
559 KASSERT(uio
->uio_td
== curthread
,
560 ("uio_td %p is not td %p", uio
->uio_td
, curthread
));
561 vp
= (struct vnode
*)fp
->f_data
;
564 if (flags
& O_FBLOCKING
) {
565 /* ioflag &= ~IO_NDELAY; */
566 } else if (flags
& O_FNONBLOCKING
) {
568 } else if (fp
->f_flag
& FNONBLOCK
) {
571 if (flags
& O_FBUFFERED
) {
572 /* ioflag &= ~IO_DIRECT; */
573 } else if (flags
& O_FUNBUFFERED
) {
575 } else if (fp
->f_flag
& O_DIRECT
) {
578 vn_lock(vp
, LK_SHARED
| LK_RETRY
);
579 if ((flags
& O_FOFFSET
) == 0)
580 uio
->uio_offset
= fp
->f_offset
;
581 ioflag
|= sequential_heuristic(uio
, fp
);
583 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, uio
);
584 error
= VOP_READ(vp
, uio
, ioflag
, cred
);
585 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
586 if ((flags
& O_FOFFSET
) == 0)
587 fp
->f_offset
= uio
->uio_offset
;
588 fp
->f_nextoff
= uio
->uio_offset
;
595 * Device-optimized file table vnode read routine.
597 * This bypasses the VOP table and talks directly to the device. Most
598 * filesystems just route to specfs and can make this optimization.
600 * MPALMOSTSAFE - acquires mplock
603 svn_read(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
611 KASSERT(uio
->uio_td
== curthread
,
612 ("uio_td %p is not td %p", uio
->uio_td
, curthread
));
614 vp
= (struct vnode
*)fp
->f_data
;
615 if (vp
== NULL
|| vp
->v_type
== VBAD
) {
620 if ((dev
= vp
->v_rdev
) == NULL
) {
626 if (uio
->uio_resid
== 0) {
630 if ((flags
& O_FOFFSET
) == 0)
631 uio
->uio_offset
= fp
->f_offset
;
634 if (flags
& O_FBLOCKING
) {
635 /* ioflag &= ~IO_NDELAY; */
636 } else if (flags
& O_FNONBLOCKING
) {
638 } else if (fp
->f_flag
& FNONBLOCK
) {
641 if (flags
& O_FBUFFERED
) {
642 /* ioflag &= ~IO_DIRECT; */
643 } else if (flags
& O_FUNBUFFERED
) {
645 } else if (fp
->f_flag
& O_DIRECT
) {
648 ioflag
|= sequential_heuristic(uio
, fp
);
650 error
= dev_dread(dev
, uio
, ioflag
);
653 if ((flags
& O_FOFFSET
) == 0)
654 fp
->f_offset
= uio
->uio_offset
;
655 fp
->f_nextoff
= uio
->uio_offset
;
662 * MPALMOSTSAFE - acquires mplock
665 vn_write(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
667 struct ccms_lock ccms_lock
;
672 KASSERT(uio
->uio_td
== curthread
,
673 ("uio_td %p is not p %p", uio
->uio_td
, curthread
));
674 vp
= (struct vnode
*)fp
->f_data
;
676 /* VOP_WRITE should handle this now */
677 if (vp
->v_type
== VREG
|| vp
->v_type
== VDATABASE
)
680 vp
= (struct vnode
*)fp
->f_data
; /* XXX needed? */
683 if (vp
->v_type
== VREG
&&
684 ((fp
->f_flag
& O_APPEND
) || (flags
& O_FAPPEND
))) {
688 if (flags
& O_FBLOCKING
) {
689 /* ioflag &= ~IO_NDELAY; */
690 } else if (flags
& O_FNONBLOCKING
) {
692 } else if (fp
->f_flag
& FNONBLOCK
) {
695 if (flags
& O_FBUFFERED
) {
696 /* ioflag &= ~IO_DIRECT; */
697 } else if (flags
& O_FUNBUFFERED
) {
699 } else if (fp
->f_flag
& O_DIRECT
) {
702 if (flags
& O_FASYNCWRITE
) {
703 /* ioflag &= ~IO_SYNC; */
704 } else if (flags
& O_FSYNCWRITE
) {
706 } else if (fp
->f_flag
& O_FSYNC
) {
710 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_SYNCHRONOUS
))
712 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
713 if ((flags
& O_FOFFSET
) == 0)
714 uio
->uio_offset
= fp
->f_offset
;
715 ioflag
|= sequential_heuristic(uio
, fp
);
716 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, uio
);
717 error
= VOP_WRITE(vp
, uio
, ioflag
, cred
);
718 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
719 if ((flags
& O_FOFFSET
) == 0)
720 fp
->f_offset
= uio
->uio_offset
;
721 fp
->f_nextoff
= uio
->uio_offset
;
728 * Device-optimized file table vnode write routine.
730 * This bypasses the VOP table and talks directly to the device. Most
731 * filesystems just route to specfs and can make this optimization.
733 * MPALMOSTSAFE - acquires mplock
736 svn_write(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
744 KASSERT(uio
->uio_td
== curthread
,
745 ("uio_td %p is not p %p", uio
->uio_td
, curthread
));
747 vp
= (struct vnode
*)fp
->f_data
;
748 if (vp
== NULL
|| vp
->v_type
== VBAD
) {
752 if (vp
->v_type
== VREG
)
753 bwillwrite(uio
->uio_resid
);
754 vp
= (struct vnode
*)fp
->f_data
; /* XXX needed? */
756 if ((dev
= vp
->v_rdev
) == NULL
) {
762 if ((flags
& O_FOFFSET
) == 0)
763 uio
->uio_offset
= fp
->f_offset
;
766 if (vp
->v_type
== VREG
&&
767 ((fp
->f_flag
& O_APPEND
) || (flags
& O_FAPPEND
))) {
771 if (flags
& O_FBLOCKING
) {
772 /* ioflag &= ~IO_NDELAY; */
773 } else if (flags
& O_FNONBLOCKING
) {
775 } else if (fp
->f_flag
& FNONBLOCK
) {
778 if (flags
& O_FBUFFERED
) {
779 /* ioflag &= ~IO_DIRECT; */
780 } else if (flags
& O_FUNBUFFERED
) {
782 } else if (fp
->f_flag
& O_DIRECT
) {
785 if (flags
& O_FASYNCWRITE
) {
786 /* ioflag &= ~IO_SYNC; */
787 } else if (flags
& O_FSYNCWRITE
) {
789 } else if (fp
->f_flag
& O_FSYNC
) {
793 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_SYNCHRONOUS
))
795 ioflag
|= sequential_heuristic(uio
, fp
);
797 error
= dev_dwrite(dev
, uio
, ioflag
);
800 if ((flags
& O_FOFFSET
) == 0)
801 fp
->f_offset
= uio
->uio_offset
;
802 fp
->f_nextoff
= uio
->uio_offset
;
809 * MPALMOSTSAFE - acquires mplock
812 vn_statfile(struct file
*fp
, struct stat
*sb
, struct ucred
*cred
)
818 vp
= (struct vnode
*)fp
->f_data
;
819 error
= vn_stat(vp
, sb
, cred
);
825 vn_stat(struct vnode
*vp
, struct stat
*sb
, struct ucred
*cred
)
834 error
= VOP_GETATTR(vp
, vap
);
839 * Zero the spare stat fields
845 * Copy from vattr table
847 if (vap
->va_fsid
!= VNOVAL
)
848 sb
->st_dev
= vap
->va_fsid
;
850 sb
->st_dev
= vp
->v_mount
->mnt_stat
.f_fsid
.val
[0];
851 sb
->st_ino
= vap
->va_fileid
;
853 switch (vap
->va_type
) {
871 /* This is a cosmetic change, symlinks do not have a mode. */
872 if (vp
->v_mount
->mnt_flag
& MNT_NOSYMFOLLOW
)
873 sb
->st_mode
&= ~ACCESSPERMS
; /* 0000 */
875 sb
->st_mode
|= ACCESSPERMS
; /* 0777 */
887 if (vap
->va_nlink
> (nlink_t
)-1)
888 sb
->st_nlink
= (nlink_t
)-1;
890 sb
->st_nlink
= vap
->va_nlink
;
891 sb
->st_uid
= vap
->va_uid
;
892 sb
->st_gid
= vap
->va_gid
;
893 sb
->st_rdev
= makeudev(vap
->va_rmajor
, vap
->va_rminor
);
894 sb
->st_size
= vap
->va_size
;
895 sb
->st_atimespec
= vap
->va_atime
;
896 sb
->st_mtimespec
= vap
->va_mtime
;
897 sb
->st_ctimespec
= vap
->va_ctime
;
900 * A VCHR and VBLK device may track the last access and last modified
901 * time independantly of the filesystem. This is particularly true
902 * because device read and write calls may bypass the filesystem.
904 if (vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
) {
907 if (dev
->si_lastread
) {
908 sb
->st_atimespec
.tv_sec
= dev
->si_lastread
;
909 sb
->st_atimespec
.tv_nsec
= 0;
911 if (dev
->si_lastwrite
) {
912 sb
->st_atimespec
.tv_sec
= dev
->si_lastwrite
;
913 sb
->st_atimespec
.tv_nsec
= 0;
919 * According to www.opengroup.org, the meaning of st_blksize is
920 * "a filesystem-specific preferred I/O block size for this
921 * object. In some filesystem types, this may vary from file
923 * Default to PAGE_SIZE after much discussion.
926 if (vap
->va_type
== VREG
) {
927 sb
->st_blksize
= vap
->va_blocksize
;
928 } else if (vn_isdisk(vp
, NULL
)) {
930 * XXX this is broken. If the device is not yet open (aka
931 * stat() call, aka v_rdev == NULL), how are we supposed
932 * to get a valid block size out of it?
935 if (dev
== NULL
&& vp
->v_type
== VCHR
) {
936 dev
= get_dev(vp
->v_umajor
, vp
->v_uminor
);
938 sb
->st_blksize
= dev
->si_bsize_best
;
939 if (sb
->st_blksize
< dev
->si_bsize_phys
)
940 sb
->st_blksize
= dev
->si_bsize_phys
;
941 if (sb
->st_blksize
< BLKDEV_IOSIZE
)
942 sb
->st_blksize
= BLKDEV_IOSIZE
;
944 sb
->st_blksize
= PAGE_SIZE
;
947 sb
->st_flags
= vap
->va_flags
;
949 error
= priv_check_cred(cred
, PRIV_VFS_GENERATION
, 0);
953 sb
->st_gen
= (u_int32_t
)vap
->va_gen
;
955 sb
->st_blocks
= vap
->va_bytes
/ S_BLKSIZE
;
956 sb
->st_fsmid
= vap
->va_fsmid
;
961 * MPALMOSTSAFE - acquires mplock
964 vn_ioctl(struct file
*fp
, u_long com
, caddr_t data
, struct ucred
*ucred
)
966 struct vnode
*vp
= ((struct vnode
*)fp
->f_data
);
973 switch (vp
->v_type
) {
976 if (com
== FIONREAD
) {
977 error
= VOP_GETATTR(vp
, &vattr
);
980 *(int *)data
= vattr
.va_size
- fp
->f_offset
;
984 if (com
== FIOASYNC
) { /* XXX */
996 if (com
== FIODTYPE
) {
997 if (vp
->v_type
!= VCHR
&& vp
->v_type
!= VBLK
) {
1001 *(int *)data
= dev_dflags(vp
->v_rdev
) & D_TYPEMASK
;
1005 error
= VOP_IOCTL(vp
, com
, data
, fp
->f_flag
, ucred
);
1006 if (error
== 0 && com
== TIOCSCTTY
) {
1007 struct proc
*p
= curthread
->td_proc
;
1008 struct session
*sess
;
1015 sess
= p
->p_session
;
1016 /* Do nothing if reassigning same control tty */
1017 if (sess
->s_ttyvp
== vp
) {
1022 /* Get rid of reference to old control tty */
1023 ovp
= sess
->s_ttyvp
;
1036 * MPALMOSTSAFE - acquires mplock
1039 vn_poll(struct file
*fp
, int events
, struct ucred
*cred
)
1044 error
= VOP_POLL(((struct vnode
*)fp
->f_data
), events
, cred
);
1050 * Check that the vnode is still valid, and if so
1051 * acquire requested lock.
1055 vn_lock(struct vnode
*vp
, int flags
)
1057 debug_vn_lock(struct vnode
*vp
, int flags
, const char *filename
, int line
)
1064 vp
->filename
= filename
;
1066 error
= debuglockmgr(&vp
->v_lock
, flags
,
1067 "vn_lock", filename
, line
);
1069 error
= lockmgr(&vp
->v_lock
, flags
);
1073 } while (flags
& LK_RETRY
);
1076 * Because we (had better!) have a ref on the vnode, once it
1077 * goes to VRECLAIMED state it will not be recycled until all
1078 * refs go away. So we can just check the flag.
1080 if (error
== 0 && (vp
->v_flag
& VRECLAIMED
)) {
1081 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1088 vn_unlock(struct vnode
*vp
)
1090 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1094 vn_islocked(struct vnode
*vp
)
1096 return (lockstatus(&vp
->v_lock
, curthread
));
1100 * MPALMOSTSAFE - acquires mplock
1103 vn_closefile(struct file
*fp
)
1108 fp
->f_ops
= &badfileops
;
1109 error
= vn_close(((struct vnode
*)fp
->f_data
), fp
->f_flag
);
1115 * MPALMOSTSAFE - acquires mplock
1118 vn_kqfilter(struct file
*fp
, struct knote
*kn
)
1123 error
= VOP_KQFILTER(((struct vnode
*)fp
->f_data
), kn
);