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 1.54 2007/11/02 19:52:25 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
)
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
;
163 * NORMAL OPEN FILE CASE
172 * split case to allow us to re-resolve and retry the ncp in case
176 if (fmode
& O_CREAT
) {
177 if (nd
->nl_nch
.ncp
->nc_vp
== NULL
) {
178 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
180 if ((dvp
= nd
->nl_nch
.ncp
->nc_parent
->nc_vp
) == NULL
)
182 /* vhold(dvp); - dvp can't go away */
185 vap
->va_mode
= cmode
;
187 vap
->va_vaflags
|= VA_EXCLUSIVE
;
188 error
= VOP_NCREATE(&nd
->nl_nch
, dvp
, &vp
,
194 /* locked vnode is returned */
196 if (fmode
& O_EXCL
) {
199 error
= cache_vget(&nd
->nl_nch
, cred
,
207 error
= cache_vget(&nd
->nl_nch
, cred
, LK_EXCLUSIVE
, &vp
);
213 * We have a locked vnode and ncp now. Note that the ncp will
214 * be cleaned up by the caller if nd->nl_nch is left intact.
216 if (vp
->v_type
== VLNK
) {
220 if (vp
->v_type
== VSOCK
) {
224 if ((fmode
& O_CREAT
) == 0) {
226 if (fmode
& (FWRITE
| O_TRUNC
)) {
227 if (vp
->v_type
== VDIR
) {
231 error
= vn_writechk(vp
, &nd
->nl_nch
);
234 * Special stale handling, re-resolve the
237 if (error
== ESTALE
) {
240 cache_setunresolved(&nd
->nl_nch
);
241 error
= cache_resolve(&nd
->nl_nch
, cred
);
252 error
= VOP_ACCESS(vp
, mode
, cred
);
255 * Special stale handling, re-resolve the
258 if (error
== ESTALE
) {
261 cache_setunresolved(&nd
->nl_nch
);
262 error
= cache_resolve(&nd
->nl_nch
, cred
);
270 if (fmode
& O_TRUNC
) {
271 vn_unlock(vp
); /* XXX */
272 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
); /* XXX */
275 error
= VOP_SETATTR(vp
, vap
, cred
);
281 * Setup the fp so VOP_OPEN can override it. No descriptor has been
282 * associated with the fp yet so we own it clean.
284 * f_nchandle inherits nl_nch. This used to be necessary only for
285 * directories but now we do it unconditionally so f*() ops
286 * such as fchmod() can access the actual namespace that was
287 * used to open the file.
290 fp
->f_nchandle
= nd
->nl_nch
;
291 cache_zero(&nd
->nl_nch
);
292 cache_unlock(&fp
->f_nchandle
);
296 * Get rid of nl_nch. vn_open does not return it (it returns the
297 * vnode or the file pointer). Note: we can't leave nl_nch locked
298 * through the VOP_OPEN anyway since the VOP_OPEN may block, e.g.
302 cache_put(&nd
->nl_nch
);
304 error
= VOP_OPEN(vp
, fmode
, cred
, fp
);
307 * setting f_ops to &badfileops will prevent the descriptor
308 * code from trying to close and release the vnode, since
309 * the open failed we do not want to call close.
313 fp
->f_ops
= &badfileops
;
320 * Assert that VREG files have been setup for vmio.
322 KASSERT(vp
->v_type
!= VREG
|| vp
->v_object
!= NULL
,
323 ("vn_open: regular file was not VMIO enabled!"));
327 * Return the vnode. XXX needs some cleaning up. The vnode is
328 * only returned in the fp == NULL case.
332 nd
->nl_vp_fmode
= fmode
;
333 if ((nd
->nl_flags
& NLC_LOCKVP
) == 0)
346 vn_opendisk(const char *devname
, int fmode
, struct vnode
**vpp
)
351 if (strncmp(devname
, "/dev/", 5) == 0)
353 if ((vp
= getsynthvnode(devname
)) == NULL
) {
356 error
= VOP_OPEN(vp
, fmode
, proc0
.p_ucred
, NULL
);
368 * Check for write permissions on the specified vnode. nch may be NULL.
371 vn_writechk(struct vnode
*vp
, struct nchandle
*nch
)
374 * If there's shared text associated with
375 * the vnode, try to free it up once. If
376 * we fail, we can't allow writing.
378 if (vp
->v_flag
& VTEXT
)
382 * If the vnode represents a regular file, check the mount
383 * point via the nch. This may be a different mount point
384 * then the one embedded in the vnode (e.g. nullfs).
386 * We can still write to non-regular files (e.g. devices)
387 * via read-only mounts.
389 if (nch
&& nch
->ncp
&& vp
->v_type
== VREG
)
390 return (ncp_writechk(nch
));
395 * Check whether the underlying mount is read-only. The mount point
396 * referenced by the namecache may be different from the mount point
397 * used by the underlying vnode in the case of NULLFS, so a separate
401 ncp_writechk(struct nchandle
*nch
)
403 if (nch
->mount
&& (nch
->mount
->mnt_flag
& MNT_RDONLY
))
412 vn_close(struct vnode
*vp
, int flags
)
416 if ((error
= vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
)) == 0) {
417 error
= VOP_CLOSE(vp
, flags
);
426 sequential_heuristic(struct uio
*uio
, struct file
*fp
)
429 * Sequential heuristic - detect sequential operation
431 if ((uio
->uio_offset
== 0 && fp
->f_seqcount
> 0) ||
432 uio
->uio_offset
== fp
->f_nextoff
) {
433 int tmpseq
= fp
->f_seqcount
;
435 * XXX we assume that the filesystem block size is
436 * the default. Not true, but still gives us a pretty
437 * good indicator of how sequential the read operations
440 tmpseq
+= (uio
->uio_resid
+ BKVASIZE
- 1) / BKVASIZE
;
441 if (tmpseq
> IO_SEQMAX
)
443 fp
->f_seqcount
= tmpseq
;
444 return(fp
->f_seqcount
<< IO_SEQSHIFT
);
448 * Not sequential, quick draw-down of seqcount
450 if (fp
->f_seqcount
> 1)
458 * Package up an I/O request on a vnode into a uio and do it.
461 vn_rdwr(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
462 off_t offset
, enum uio_seg segflg
, int ioflg
,
463 struct ucred
*cred
, int *aresid
)
467 struct ccms_lock ccms_lock
;
470 if ((ioflg
& IO_NODELOCKED
) == 0)
471 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
472 auio
.uio_iov
= &aiov
;
474 aiov
.iov_base
= base
;
476 auio
.uio_resid
= len
;
477 auio
.uio_offset
= offset
;
478 auio
.uio_segflg
= segflg
;
480 auio
.uio_td
= curthread
;
481 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, &auio
);
482 if (rw
== UIO_READ
) {
483 error
= VOP_READ(vp
, &auio
, ioflg
, cred
);
485 error
= VOP_WRITE(vp
, &auio
, ioflg
, cred
);
487 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
489 *aresid
= auio
.uio_resid
;
491 if (auio
.uio_resid
&& error
== 0)
493 if ((ioflg
& IO_NODELOCKED
) == 0)
499 * Package up an I/O request on a vnode into a uio and do it. The I/O
500 * request is split up into smaller chunks and we try to avoid saturating
501 * the buffer cache while potentially holding a vnode locked, so we
502 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield()
503 * to give other processes a chance to lock the vnode (either other processes
504 * core'ing the same binary, or unrelated processes scanning the directory).
507 vn_rdwr_inchunks(enum uio_rw rw
, struct vnode
*vp
, caddr_t base
, int len
,
508 off_t offset
, enum uio_seg segflg
, int ioflg
,
509 struct ucred
*cred
, int *aresid
)
517 * Force `offset' to a multiple of MAXBSIZE except possibly
518 * for the first chunk, so that filesystems only need to
519 * write full blocks except possibly for the first and last
522 chunk
= MAXBSIZE
- (uoff_t
)offset
% MAXBSIZE
;
526 if (rw
!= UIO_READ
&& vp
->v_type
== VREG
)
528 error
= vn_rdwr(rw
, vp
, base
, chunk
, offset
, segflg
,
529 ioflg
, cred
, aresid
);
530 len
-= chunk
; /* aresid calc already includes length */
543 * MPALMOSTSAFE - acquires mplock
546 vn_read(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
548 struct ccms_lock ccms_lock
;
553 KASSERT(uio
->uio_td
== curthread
,
554 ("uio_td %p is not td %p", uio
->uio_td
, curthread
));
555 vp
= (struct vnode
*)fp
->f_data
;
558 if (flags
& O_FBLOCKING
) {
559 /* ioflag &= ~IO_NDELAY; */
560 } else if (flags
& O_FNONBLOCKING
) {
562 } else if (fp
->f_flag
& FNONBLOCK
) {
565 if (flags
& O_FBUFFERED
) {
566 /* ioflag &= ~IO_DIRECT; */
567 } else if (flags
& O_FUNBUFFERED
) {
569 } else if (fp
->f_flag
& O_DIRECT
) {
572 vn_lock(vp
, LK_SHARED
| LK_RETRY
);
573 if ((flags
& O_FOFFSET
) == 0)
574 uio
->uio_offset
= fp
->f_offset
;
575 ioflag
|= sequential_heuristic(uio
, fp
);
577 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, uio
);
578 error
= VOP_READ(vp
, uio
, ioflag
, cred
);
579 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
580 if ((flags
& O_FOFFSET
) == 0)
581 fp
->f_offset
= uio
->uio_offset
;
582 fp
->f_nextoff
= uio
->uio_offset
;
589 * Device-optimized file table vnode read routine.
591 * This bypasses the VOP table and talks directly to the device. Most
592 * filesystems just route to specfs and can make this optimization.
594 * MPALMOSTSAFE - acquires mplock
597 svn_read(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
605 KASSERT(uio
->uio_td
== curthread
,
606 ("uio_td %p is not td %p", uio
->uio_td
, curthread
));
608 vp
= (struct vnode
*)fp
->f_data
;
609 if (vp
== NULL
|| vp
->v_type
== VBAD
) {
614 if ((dev
= vp
->v_rdev
) == NULL
) {
620 if (uio
->uio_resid
== 0) {
624 if ((flags
& O_FOFFSET
) == 0)
625 uio
->uio_offset
= fp
->f_offset
;
628 if (flags
& O_FBLOCKING
) {
629 /* ioflag &= ~IO_NDELAY; */
630 } else if (flags
& O_FNONBLOCKING
) {
632 } else if (fp
->f_flag
& FNONBLOCK
) {
635 if (flags
& O_FBUFFERED
) {
636 /* ioflag &= ~IO_DIRECT; */
637 } else if (flags
& O_FUNBUFFERED
) {
639 } else if (fp
->f_flag
& O_DIRECT
) {
642 ioflag
|= sequential_heuristic(uio
, fp
);
644 error
= dev_dread(dev
, uio
, ioflag
);
647 if ((flags
& O_FOFFSET
) == 0)
648 fp
->f_offset
= uio
->uio_offset
;
649 fp
->f_nextoff
= uio
->uio_offset
;
656 * MPALMOSTSAFE - acquires mplock
659 vn_write(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
661 struct ccms_lock ccms_lock
;
666 KASSERT(uio
->uio_td
== curthread
,
667 ("uio_td %p is not p %p", uio
->uio_td
, curthread
));
668 vp
= (struct vnode
*)fp
->f_data
;
669 if (vp
->v_type
== VREG
)
671 vp
= (struct vnode
*)fp
->f_data
; /* XXX needed? */
674 if (vp
->v_type
== VREG
&&
675 ((fp
->f_flag
& O_APPEND
) || (flags
& O_FAPPEND
))) {
679 if (flags
& O_FBLOCKING
) {
680 /* ioflag &= ~IO_NDELAY; */
681 } else if (flags
& O_FNONBLOCKING
) {
683 } else if (fp
->f_flag
& FNONBLOCK
) {
686 if (flags
& O_FBUFFERED
) {
687 /* ioflag &= ~IO_DIRECT; */
688 } else if (flags
& O_FUNBUFFERED
) {
690 } else if (fp
->f_flag
& O_DIRECT
) {
693 if (flags
& O_FASYNCWRITE
) {
694 /* ioflag &= ~IO_SYNC; */
695 } else if (flags
& O_FSYNCWRITE
) {
697 } else if (fp
->f_flag
& O_FSYNC
) {
701 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_SYNCHRONOUS
))
703 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
704 if ((flags
& O_FOFFSET
) == 0)
705 uio
->uio_offset
= fp
->f_offset
;
706 ioflag
|= sequential_heuristic(uio
, fp
);
707 ccms_lock_get_uio(&vp
->v_ccms
, &ccms_lock
, uio
);
708 error
= VOP_WRITE(vp
, uio
, ioflag
, cred
);
709 ccms_lock_put(&vp
->v_ccms
, &ccms_lock
);
710 if ((flags
& O_FOFFSET
) == 0)
711 fp
->f_offset
= uio
->uio_offset
;
712 fp
->f_nextoff
= uio
->uio_offset
;
719 * Device-optimized file table vnode write routine.
721 * This bypasses the VOP table and talks directly to the device. Most
722 * filesystems just route to specfs and can make this optimization.
724 * MPALMOSTSAFE - acquires mplock
727 svn_write(struct file
*fp
, struct uio
*uio
, struct ucred
*cred
, int flags
)
735 KASSERT(uio
->uio_td
== curthread
,
736 ("uio_td %p is not p %p", uio
->uio_td
, curthread
));
738 vp
= (struct vnode
*)fp
->f_data
;
739 if (vp
== NULL
|| vp
->v_type
== VBAD
) {
743 if (vp
->v_type
== VREG
)
745 vp
= (struct vnode
*)fp
->f_data
; /* XXX needed? */
747 if ((dev
= vp
->v_rdev
) == NULL
) {
753 if ((flags
& O_FOFFSET
) == 0)
754 uio
->uio_offset
= fp
->f_offset
;
757 if (vp
->v_type
== VREG
&&
758 ((fp
->f_flag
& O_APPEND
) || (flags
& O_FAPPEND
))) {
762 if (flags
& O_FBLOCKING
) {
763 /* ioflag &= ~IO_NDELAY; */
764 } else if (flags
& O_FNONBLOCKING
) {
766 } else if (fp
->f_flag
& FNONBLOCK
) {
769 if (flags
& O_FBUFFERED
) {
770 /* ioflag &= ~IO_DIRECT; */
771 } else if (flags
& O_FUNBUFFERED
) {
773 } else if (fp
->f_flag
& O_DIRECT
) {
776 if (flags
& O_FASYNCWRITE
) {
777 /* ioflag &= ~IO_SYNC; */
778 } else if (flags
& O_FSYNCWRITE
) {
780 } else if (fp
->f_flag
& O_FSYNC
) {
784 if (vp
->v_mount
&& (vp
->v_mount
->mnt_flag
& MNT_SYNCHRONOUS
))
786 ioflag
|= sequential_heuristic(uio
, fp
);
788 error
= dev_dwrite(dev
, uio
, ioflag
);
791 if ((flags
& O_FOFFSET
) == 0)
792 fp
->f_offset
= uio
->uio_offset
;
793 fp
->f_nextoff
= uio
->uio_offset
;
800 * MPALMOSTSAFE - acquires mplock
803 vn_statfile(struct file
*fp
, struct stat
*sb
, struct ucred
*cred
)
809 vp
= (struct vnode
*)fp
->f_data
;
810 error
= vn_stat(vp
, sb
, cred
);
816 vn_stat(struct vnode
*vp
, struct stat
*sb
, struct ucred
*cred
)
825 error
= VOP_GETATTR(vp
, vap
);
830 * Zero the spare stat fields
836 * Copy from vattr table
838 if (vap
->va_fsid
!= VNOVAL
)
839 sb
->st_dev
= vap
->va_fsid
;
841 sb
->st_dev
= vp
->v_mount
->mnt_stat
.f_fsid
.val
[0];
842 sb
->st_ino
= vap
->va_fileid
;
844 switch (vap
->va_type
) {
862 /* This is a cosmetic change, symlinks do not have a mode. */
863 if (vp
->v_mount
->mnt_flag
& MNT_NOSYMFOLLOW
)
864 sb
->st_mode
&= ~ACCESSPERMS
; /* 0000 */
866 sb
->st_mode
|= ACCESSPERMS
; /* 0777 */
878 if (vap
->va_nlink
> (nlink_t
)-1)
879 sb
->st_nlink
= (nlink_t
)-1;
881 sb
->st_nlink
= vap
->va_nlink
;
882 sb
->st_uid
= vap
->va_uid
;
883 sb
->st_gid
= vap
->va_gid
;
884 sb
->st_rdev
= makeudev(vap
->va_rmajor
, vap
->va_rminor
);
885 sb
->st_size
= vap
->va_size
;
886 sb
->st_atimespec
= vap
->va_atime
;
887 sb
->st_mtimespec
= vap
->va_mtime
;
888 sb
->st_ctimespec
= vap
->va_ctime
;
891 * A VCHR and VBLK device may track the last access and last modified
892 * time independantly of the filesystem. This is particularly true
893 * because device read and write calls may bypass the filesystem.
895 if (vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
) {
896 if ((dev
= vp
->v_rdev
) != NULL
) {
897 if (dev
->si_lastread
) {
898 sb
->st_atimespec
.tv_sec
= dev
->si_lastread
;
899 sb
->st_atimespec
.tv_nsec
= 0;
901 if (dev
->si_lastwrite
) {
902 sb
->st_atimespec
.tv_sec
= dev
->si_lastwrite
;
903 sb
->st_atimespec
.tv_nsec
= 0;
909 * According to www.opengroup.org, the meaning of st_blksize is
910 * "a filesystem-specific preferred I/O block size for this
911 * object. In some filesystem types, this may vary from file
913 * Default to PAGE_SIZE after much discussion.
916 if (vap
->va_type
== VREG
) {
917 sb
->st_blksize
= vap
->va_blocksize
;
918 } else if (vn_isdisk(vp
, NULL
)) {
920 * XXX this is broken. If the device is not yet open (aka
921 * stat() call, aka v_rdev == NULL), how are we supposed
922 * to get a valid block size out of it?
926 if ((dev
= vp
->v_rdev
) == NULL
) {
927 if (vp
->v_type
== VCHR
)
928 dev
= get_dev(vp
->v_umajor
, vp
->v_uminor
);
930 sb
->st_blksize
= dev
->si_bsize_best
;
931 if (sb
->st_blksize
< dev
->si_bsize_phys
)
932 sb
->st_blksize
= dev
->si_bsize_phys
;
933 if (sb
->st_blksize
< BLKDEV_IOSIZE
)
934 sb
->st_blksize
= BLKDEV_IOSIZE
;
936 sb
->st_blksize
= PAGE_SIZE
;
939 sb
->st_flags
= vap
->va_flags
;
940 if (suser_cred(cred
, 0))
943 sb
->st_gen
= (u_int32_t
)vap
->va_gen
;
945 #if (S_BLKSIZE == 512)
946 /* Optimize this case */
947 sb
->st_blocks
= vap
->va_bytes
>> 9;
949 sb
->st_blocks
= vap
->va_bytes
/ S_BLKSIZE
;
951 sb
->st_fsmid
= vap
->va_fsmid
;
956 * MPALMOSTSAFE - acquires mplock
959 vn_ioctl(struct file
*fp
, u_long com
, caddr_t data
, struct ucred
*ucred
)
961 struct vnode
*vp
= ((struct vnode
*)fp
->f_data
);
968 switch (vp
->v_type
) {
971 if (com
== FIONREAD
) {
972 if ((error
= VOP_GETATTR(vp
, &vattr
)) != 0)
974 *(int *)data
= vattr
.va_size
- fp
->f_offset
;
978 if (com
== FIOASYNC
) { /* XXX */
990 if (com
== FIODTYPE
) {
991 if (vp
->v_type
!= VCHR
&& vp
->v_type
!= VBLK
) {
995 *(int *)data
= dev_dflags(vp
->v_rdev
) & D_TYPEMASK
;
999 error
= VOP_IOCTL(vp
, com
, data
, fp
->f_flag
, ucred
);
1000 if (error
== 0 && com
== TIOCSCTTY
) {
1001 struct proc
*p
= curthread
->td_proc
;
1002 struct session
*sess
;
1009 sess
= p
->p_session
;
1010 /* Do nothing if reassigning same control tty */
1011 if (sess
->s_ttyvp
== vp
) {
1016 /* Get rid of reference to old control tty */
1017 ovp
= sess
->s_ttyvp
;
1030 * MPALMOSTSAFE - acquires mplock
1033 vn_poll(struct file
*fp
, int events
, struct ucred
*cred
)
1038 error
= VOP_POLL(((struct vnode
*)fp
->f_data
), events
, cred
);
1044 * Check that the vnode is still valid, and if so
1045 * acquire requested lock.
1049 vn_lock(struct vnode
*vp
, int flags
)
1051 debug_vn_lock(struct vnode
*vp
, int flags
, const char *filename
, int line
)
1058 vp
->filename
= filename
;
1060 error
= debuglockmgr(&vp
->v_lock
, flags
,
1061 "vn_lock", filename
, line
);
1063 error
= lockmgr(&vp
->v_lock
, flags
);
1067 } while (flags
& LK_RETRY
);
1070 * Because we (had better!) have a ref on the vnode, once it
1071 * goes to VRECLAIMED state it will not be recycled until all
1072 * refs go away. So we can just check the flag.
1074 if (error
== 0 && (vp
->v_flag
& VRECLAIMED
)) {
1075 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1082 vn_unlock(struct vnode
*vp
)
1084 lockmgr(&vp
->v_lock
, LK_RELEASE
);
1088 vn_islocked(struct vnode
*vp
)
1090 return (lockstatus(&vp
->v_lock
, curthread
));
1094 * MPALMOSTSAFE - acquires mplock
1097 vn_closefile(struct file
*fp
)
1102 fp
->f_ops
= &badfileops
;
1103 error
= vn_close(((struct vnode
*)fp
->f_data
), fp
->f_flag
);
1109 * MPALMOSTSAFE - acquires mplock
1112 vn_kqfilter(struct file
*fp
, struct knote
*kn
)
1117 error
= VOP_KQFILTER(((struct vnode
*)fp
->f_data
), kn
);