2 * Copyright (c) 1982, 1986, 1989, 1993, 1995
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 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
39 * $FreeBSD: src/sys/ufs/ufs/ufs_vnops.c,v 1.131.2.8 2003/01/02 17:26:19 bde Exp $
40 * $DragonFly: src/sys/vfs/ufs/ufs_vnops.c,v 1.13 2004/05/18 00:16:46 cpressey Exp $
43 #include "opt_quota.h"
44 #include "opt_suiddir.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/fcntl.h>
54 #include <sys/namei.h>
55 #include <sys/mount.h>
56 #include <sys/unistd.h>
57 #include <sys/vnode.h>
58 #include <sys/malloc.h>
59 #include <sys/dirent.h>
60 #include <sys/lockf.h>
61 #include <sys/event.h>
64 #include <sys/file.h> /* XXX */
67 #include <vm/vm_extern.h>
69 #include <vfs/fifofs/fifo.h>
75 #include "ufs_extern.h"
80 static int ufs_access (struct vop_access_args *);
81 static int ufs_advlock (struct vop_advlock_args *);
82 static int ufs_chmod (struct vnode *, int, struct ucred *, struct thread *);
83 static int ufs_chown (struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
84 static int ufs_close (struct vop_close_args *);
85 static int ufs_create (struct vop_create_args *);
86 static int ufs_getattr (struct vop_getattr_args *);
87 static int ufs_link (struct vop_link_args *);
88 static int ufs_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
89 static int ufs_missingop (struct vop_generic_args *ap);
90 static int ufs_mkdir (struct vop_mkdir_args *);
91 static int ufs_mknod (struct vop_mknod_args *);
92 static int ufs_mmap (struct vop_mmap_args *);
93 static int ufs_open (struct vop_open_args *);
94 static int ufs_pathconf (struct vop_pathconf_args *);
95 static int ufs_print (struct vop_print_args *);
96 static int ufs_readdir (struct vop_readdir_args *);
97 static int ufs_readlink (struct vop_readlink_args *);
98 static int ufs_remove (struct vop_remove_args *);
99 static int ufs_rename (struct vop_rename_args *);
100 static int ufs_rmdir (struct vop_rmdir_args *);
101 static int ufs_setattr (struct vop_setattr_args *);
102 static int ufs_strategy (struct vop_strategy_args *);
103 static int ufs_symlink (struct vop_symlink_args *);
104 static int ufs_whiteout (struct vop_whiteout_args *);
105 static int ufsfifo_close (struct vop_close_args *);
106 static int ufsfifo_kqfilter (struct vop_kqfilter_args *);
107 static int ufsfifo_read (struct vop_read_args *);
108 static int ufsfifo_write (struct vop_write_args *);
109 static int ufsspec_close (struct vop_close_args *);
110 static int ufsspec_read (struct vop_read_args *);
111 static int ufsspec_write (struct vop_write_args *);
112 static int filt_ufsread (struct knote *kn, long hint);
113 static int filt_ufswrite (struct knote *kn, long hint);
114 static int filt_ufsvnode (struct knote *kn, long hint);
115 static void filt_ufsdetach (struct knote *kn);
116 static int ufs_kqfilter (struct vop_kqfilter_args *ap);
122 #define SETHIGH(q, h) { \
125 tmp.val[_QUAD_HIGHWORD] = (h); \
128 #define SETLOW(q, l) { \
131 tmp.val[_QUAD_LOWWORD] = (l); \
134 #define VN_KNOTE(vp, b) \
135 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
138 * A virgin directory (no blushing please).
140 static struct dirtemplate mastertemplate = {
141 0, 12, DT_DIR, 1, ".",
142 0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
144 static struct odirtemplate omastertemplate = {
146 0, DIRBLKSIZ - 12, 2, ".."
150 ufs_itimes(struct vnode *vp)
156 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
158 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
159 ip->i_flag |= IN_LAZYMOD;
161 ip->i_flag |= IN_MODIFIED;
162 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
164 if (ip->i_flag & IN_ACCESS) {
165 ip->i_atime = ts.tv_sec;
166 ip->i_atimensec = ts.tv_nsec;
168 if (ip->i_flag & IN_UPDATE) {
169 ip->i_mtime = ts.tv_sec;
170 ip->i_mtimensec = ts.tv_nsec;
173 if (ip->i_flag & IN_CHANGE) {
174 ip->i_ctime = ts.tv_sec;
175 ip->i_ctimensec = ts.tv_nsec;
178 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
182 * Create a regular file
184 * ufs_create(struct vnode *a_dvp, struct vnode **a_vpp,
185 * struct componentname *a_cnp, struct vattr *a_vap)
188 ufs_create(struct vop_create_args *ap)
193 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
194 ap->a_dvp, ap->a_vpp, ap->a_cnp);
197 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
204 * ufs_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
205 * struct componentname *a_cnp, struct vattr *a_vap)
209 ufs_mknod(struct vop_mknod_args *ap)
211 struct vattr *vap = ap->a_vap;
212 struct vnode **vpp = ap->a_vpp;
217 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
218 ap->a_dvp, vpp, ap->a_cnp);
221 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
223 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
224 if (vap->va_rdev != VNOVAL) {
226 * Want to be able to use this to make badblock
227 * inodes, so don't truncate the dev number.
229 ip->i_rdev = vap->va_rdev;
232 * Remove inode, then reload it through VFS_VGET so it is
233 * checked to see if it is an alias of an existing entry in
237 (*vpp)->v_type = VNON;
238 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
240 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
253 * ufs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
254 * struct thread *a_td)
258 ufs_open(struct vop_open_args *ap)
261 * Files marked append-only must be opened for appending.
263 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
264 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
272 * Update the times on the inode.
274 * ufs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
275 * struct thread *a_td)
279 ufs_close(struct vop_close_args *ap)
281 struct vnode *vp = ap->a_vp;
284 lwkt_gettoken(&vlock, vp->v_interlock);
285 if (vp->v_usecount > 1)
287 lwkt_reltoken(&vlock);
292 * ufs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
293 * struct thread *a_td)
296 ufs_access(struct vop_access_args *ap)
298 struct vnode *vp = ap->a_vp;
299 struct inode *ip = VTOI(vp);
300 struct ucred *cred = ap->a_cred;
301 mode_t mask, mode = ap->a_mode;
309 * Disallow write attempts on read-only file systems;
310 * unless the file is a socket, fifo, or a block or
311 * character device resident on the file system.
314 switch (vp->v_type) {
318 if (vp->v_mount->mnt_flag & MNT_RDONLY)
321 if ((error = getinoquota(ip)) != 0)
330 /* If immutable bit set, nobody gets to write it. */
331 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
334 /* Otherwise, user id 0 always gets access. */
335 if (cred->cr_uid == 0)
340 /* Otherwise, check the owner. */
341 if (cred->cr_uid == ip->i_uid) {
348 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
351 /* Otherwise, check the groups. */
352 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
353 if (ip->i_gid == *gp) {
360 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
363 /* Otherwise, check everyone else. */
370 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
374 * ufs_getattr(struct vnode *a_vp, struct vattr *a_vap,
375 * struct thread *a_td)
379 ufs_getattr(struct vop_getattr_args *ap)
381 struct vnode *vp = ap->a_vp;
382 struct inode *ip = VTOI(vp);
383 struct vattr *vap = ap->a_vap;
387 * Copy from inode table
389 vap->va_fsid = dev2udev(ip->i_dev);
390 vap->va_fileid = ip->i_number;
391 vap->va_mode = ip->i_mode & ~IFMT;
392 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ?
393 ip->i_effnlink : ip->i_nlink;
394 vap->va_uid = ip->i_uid;
395 vap->va_gid = ip->i_gid;
396 vap->va_rdev = ip->i_rdev;
397 vap->va_size = ip->i_din.di_size;
398 vap->va_atime.tv_sec = ip->i_atime;
399 vap->va_atime.tv_nsec = ip->i_atimensec;
400 vap->va_mtime.tv_sec = ip->i_mtime;
401 vap->va_mtime.tv_nsec = ip->i_mtimensec;
402 vap->va_ctime.tv_sec = ip->i_ctime;
403 vap->va_ctime.tv_nsec = ip->i_ctimensec;
404 vap->va_flags = ip->i_flags;
405 vap->va_gen = ip->i_gen;
406 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
407 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
408 vap->va_type = IFTOVT(ip->i_mode);
409 vap->va_filerev = ip->i_modrev;
414 * Set attribute vnode op. called from several syscalls
416 * ufs_setattr(struct vnode *a_vp, struct vattr *a_vap,
417 * struct ucred *a_cred, struct thread *a_td)
420 ufs_setattr(struct vop_setattr_args *ap)
422 struct vattr *vap = ap->a_vap;
423 struct vnode *vp = ap->a_vp;
424 struct inode *ip = VTOI(vp);
425 struct ucred *cred = ap->a_cred;
429 * Check for unsettable attributes.
431 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
432 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
433 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
434 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
437 if (vap->va_flags != VNOVAL) {
438 if (vp->v_mount->mnt_flag & MNT_RDONLY)
440 if (cred->cr_uid != ip->i_uid &&
441 (error = suser_cred(cred, PRISON_ROOT)))
443 if ((cred->cr_uid == 0) && (cred->cr_prison == NULL)) {
445 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
448 ip->i_flags = vap->va_flags;
451 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
452 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
454 ip->i_flags &= SF_SETTABLE;
455 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
457 ip->i_flag |= IN_CHANGE;
458 if (vap->va_flags & (IMMUTABLE | APPEND))
461 if (ip->i_flags & (IMMUTABLE | APPEND))
464 * Go through the fields and update iff not VNOVAL.
466 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
467 if (vp->v_mount->mnt_flag & MNT_RDONLY)
469 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
472 if (vap->va_size != VNOVAL) {
474 * Disallow write attempts on read-only file systems;
475 * unless the file is a socket, fifo, or a block or
476 * character device resident on the file system.
478 switch (vp->v_type) {
483 if (vp->v_mount->mnt_flag & MNT_RDONLY)
489 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
493 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
494 if (vp->v_mount->mnt_flag & MNT_RDONLY)
496 if (cred->cr_uid != ip->i_uid &&
497 (error = suser_cred(cred, PRISON_ROOT)) &&
498 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
499 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
501 if (vap->va_atime.tv_sec != VNOVAL)
502 ip->i_flag |= IN_ACCESS;
503 if (vap->va_mtime.tv_sec != VNOVAL)
504 ip->i_flag |= IN_CHANGE | IN_UPDATE;
506 if (vap->va_atime.tv_sec != VNOVAL) {
507 ip->i_atime = vap->va_atime.tv_sec;
508 ip->i_atimensec = vap->va_atime.tv_nsec;
510 if (vap->va_mtime.tv_sec != VNOVAL) {
511 ip->i_mtime = vap->va_mtime.tv_sec;
512 ip->i_mtimensec = vap->va_mtime.tv_nsec;
514 error = UFS_UPDATE(vp, 0);
519 if (vap->va_mode != (mode_t)VNOVAL) {
520 if (vp->v_mount->mnt_flag & MNT_RDONLY)
522 error = ufs_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
524 VN_KNOTE(vp, NOTE_ATTRIB);
529 * Change the mode on a file.
530 * Inode must be locked before calling.
533 ufs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
535 struct inode *ip = VTOI(vp);
538 if (cred->cr_uid != ip->i_uid) {
539 error = suser_cred(cred, PRISON_ROOT);
544 if (vp->v_type != VDIR && (mode & S_ISTXT))
546 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
549 ip->i_mode &= ~ALLPERMS;
550 ip->i_mode |= (mode & ALLPERMS);
551 ip->i_flag |= IN_CHANGE;
556 * Perform chown operation on inode ip;
557 * inode must be locked prior to call.
560 ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
563 struct inode *ip = VTOI(vp);
572 if (uid == (uid_t)VNOVAL)
574 if (gid == (gid_t)VNOVAL)
577 * If we don't own the file, are trying to change the owner
578 * of the file, or are not a member of the target group,
579 * the caller must be superuser or the call fails.
581 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
582 (gid != ip->i_gid && !groupmember((gid_t)gid, cred))) &&
583 (error = suser_cred(cred, PRISON_ROOT)))
588 if ((error = getinoquota(ip)) != 0)
591 dqrele(vp, ip->i_dquot[USRQUOTA]);
592 ip->i_dquot[USRQUOTA] = NODQUOT;
595 dqrele(vp, ip->i_dquot[GRPQUOTA]);
596 ip->i_dquot[GRPQUOTA] = NODQUOT;
598 change = ip->i_blocks;
599 (void) chkdq(ip, -change, cred, CHOWN);
600 (void) chkiq(ip, -1, cred, CHOWN);
601 for (i = 0; i < MAXQUOTAS; i++) {
602 dqrele(vp, ip->i_dquot[i]);
603 ip->i_dquot[i] = NODQUOT;
609 if ((error = getinoquota(ip)) == 0) {
611 dqrele(vp, ip->i_dquot[USRQUOTA]);
612 ip->i_dquot[USRQUOTA] = NODQUOT;
615 dqrele(vp, ip->i_dquot[GRPQUOTA]);
616 ip->i_dquot[GRPQUOTA] = NODQUOT;
618 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
619 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
622 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
624 for (i = 0; i < MAXQUOTAS; i++) {
625 dqrele(vp, ip->i_dquot[i]);
626 ip->i_dquot[i] = NODQUOT;
631 if (getinoquota(ip) == 0) {
633 dqrele(vp, ip->i_dquot[USRQUOTA]);
634 ip->i_dquot[USRQUOTA] = NODQUOT;
637 dqrele(vp, ip->i_dquot[GRPQUOTA]);
638 ip->i_dquot[GRPQUOTA] = NODQUOT;
640 (void) chkdq(ip, change, cred, FORCE|CHOWN);
641 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
642 (void) getinoquota(ip);
647 panic("ufs_chown: lost quota");
649 ip->i_flag |= IN_CHANGE;
650 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
651 ip->i_mode &= ~(ISUID | ISGID);
658 * NB Currently unsupported.
660 * ufs_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
661 * struct thread *a_td)
665 ufs_mmap(struct vop_mmap_args *ap)
671 * ufs_remove(struct vnode *a_dvp, struct vnode *a_vp,
672 * struct componentname *a_cnp)
675 ufs_remove(struct vop_remove_args *ap)
678 struct vnode *vp = ap->a_vp;
679 struct vnode *dvp = ap->a_dvp;
683 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
684 (VTOI(dvp)->i_flags & APPEND)) {
688 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
689 VN_KNOTE(vp, NOTE_DELETE);
690 VN_KNOTE(dvp, NOTE_WRITE);
698 * ufs_link(struct vnode *a_tdvp, struct vnode *a_vp,
699 * struct componentname *a_cnp)
702 ufs_link(struct vop_link_args *ap)
704 struct vnode *vp = ap->a_vp;
705 struct vnode *tdvp = ap->a_tdvp;
706 struct componentname *cnp = ap->a_cnp;
707 struct thread *td = cnp->cn_td;
709 struct direct newdir;
713 if ((cnp->cn_flags & CNP_HASBUF) == 0)
714 panic("ufs_link: no name");
716 if (tdvp->v_mount != vp->v_mount) {
720 if (tdvp != vp && (error = vn_lock(vp, NULL, LK_EXCLUSIVE, td))) {
724 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
728 if (ip->i_flags & (IMMUTABLE | APPEND)) {
734 ip->i_flag |= IN_CHANGE;
735 if (DOINGSOFTDEP(vp))
736 softdep_change_linkcnt(ip);
737 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
739 ufs_makedirentry(ip, cnp, &newdir);
740 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
746 ip->i_flag |= IN_CHANGE;
747 if (DOINGSOFTDEP(vp))
748 softdep_change_linkcnt(ip);
752 VOP_UNLOCK(vp, NULL, 0, td);
754 VN_KNOTE(vp, NOTE_LINK);
755 VN_KNOTE(tdvp, NOTE_WRITE);
760 * whiteout vnode call
762 * ufs_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
765 ufs_whiteout(struct vop_whiteout_args *ap)
767 struct vnode *dvp = ap->a_dvp;
768 struct componentname *cnp = ap->a_cnp;
769 struct direct newdir;
772 switch (ap->a_flags) {
774 /* 4.4 format directories support whiteout operations */
775 if (dvp->v_mount->mnt_maxsymlinklen > 0)
780 /* create a new directory whiteout */
782 if ((cnp->cn_flags & CNP_SAVENAME) == 0)
783 panic("ufs_whiteout: missing name");
784 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
785 panic("ufs_whiteout: old format filesystem");
789 newdir.d_namlen = cnp->cn_namelen;
790 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
791 newdir.d_type = DT_WHT;
792 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
796 /* remove an existing directory whiteout */
798 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
799 panic("ufs_whiteout: old format filesystem");
802 cnp->cn_flags &= ~CNP_DOWHITEOUT;
803 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
806 panic("ufs_whiteout: unknown op");
812 * Rename system call.
813 * rename("foo", "bar");
816 * link("foo", "bar");
818 * but ``atomically''. Can't do full commit without saving state in the
819 * inode on disk which isn't feasible at this time. Best we can do is
820 * always guarantee the target exists.
822 * Basic algorithm is:
824 * 1) Bump link count on source while we're linking it to the
825 * target. This also ensure the inode won't be deleted out
826 * from underneath us while we work (it may be truncated by
827 * a concurrent `trunc' or `open' for creation).
828 * 2) Link source to destination. If destination already exists,
830 * 3) Unlink source reference to inode if still around. If a
831 * directory was moved and the parent of the destination
832 * is different from the source, patch the ".." entry in the
835 * ufs_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
836 * struct componentname *a_fcnp, struct vnode *a_tdvp,
837 * struct vnode *a_tvp, struct componentname *a_tcnp)
840 ufs_rename(struct vop_rename_args *ap)
842 struct vnode *tvp = ap->a_tvp;
843 struct vnode *tdvp = ap->a_tdvp;
844 struct vnode *fvp = ap->a_fvp;
845 struct vnode *fdvp = ap->a_fdvp;
846 struct componentname *tcnp = ap->a_tcnp;
847 struct componentname *fcnp = ap->a_fcnp;
848 struct thread *td = fcnp->cn_td;
849 struct inode *ip, *xp, *dp;
850 struct direct newdir;
851 int doingdirectory = 0, oldparent = 0, newparent = 0;
852 int error = 0, ioflag;
855 if ((tcnp->cn_flags & CNP_HASBUF) == 0 ||
856 (fcnp->cn_flags & CNP_HASBUF) == 0)
857 panic("ufs_rename: no name");
860 * Check for cross-device rename.
862 if ((fvp->v_mount != tdvp->v_mount) ||
863 (tvp && (fvp->v_mount != tvp->v_mount))) {
877 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
878 (VTOI(tdvp)->i_flags & APPEND))) {
884 * Renaming a file to itself has no effect. The upper layers should
885 * not call us in that case. Temporarily just warn if they do.
888 printf("ufs_rename: fvp == tvp (can't happen)\n");
893 if ((error = vn_lock(fvp, NULL, LK_EXCLUSIVE, td)) != 0)
897 if (ip->i_nlink >= LINK_MAX) {
898 VOP_UNLOCK(fvp, NULL, 0, td);
902 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
903 || (dp->i_flags & APPEND)) {
904 VOP_UNLOCK(fvp, NULL, 0, td);
908 if ((ip->i_mode & IFMT) == IFDIR) {
910 * Avoid ".", "..", and aliases of "." for obvious reasons.
912 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
913 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
914 (ip->i_flag & IN_RENAME)) {
915 VOP_UNLOCK(fvp, NULL, 0, td);
919 ip->i_flag |= IN_RENAME;
920 oldparent = dp->i_number;
923 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
927 * When the target exists, both the directory
928 * and target vnodes are returned locked.
936 * 1) Bump link count while we're moving stuff
937 * around. If we crash somewhere before
938 * completing our work, the link count
939 * may be wrong, but correctable.
943 ip->i_flag |= IN_CHANGE;
944 if (DOINGSOFTDEP(fvp))
945 softdep_change_linkcnt(ip);
946 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
947 DOINGASYNC(fvp)))) != 0) {
948 VOP_UNLOCK(fvp, NULL, 0, td);
953 * If ".." must be changed (ie the directory gets a new
954 * parent) then the source directory must not be in the
955 * directory heirarchy above the target, as this would
956 * orphan everything below the source directory. Also
957 * the user must have write permission in the source so
958 * as to be able to change "..". We must repeat the call
959 * to namei, as the parent directory is unlocked by the
960 * call to checkpath().
962 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
963 VOP_UNLOCK(fvp, NULL, 0, td);
964 if (oldparent != dp->i_number)
965 newparent = dp->i_number;
966 if (doingdirectory && newparent) {
967 if (error) /* write access check above */
971 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
974 if ((tcnp->cn_flags & CNP_SAVESTART) == 0)
975 panic("ufs_rename: lost to startdir");
977 error = relookup(tdvp, &tvp, tcnp);
987 * 2) If target doesn't exist, link the target
988 * to the source and unlink the source.
989 * Otherwise, rewrite the target directory
990 * entry to reference the source inode and
991 * expunge the original entry's existence.
994 if (dp->i_dev != ip->i_dev)
995 panic("ufs_rename: EXDEV");
997 * Account for ".." in new directory.
998 * When source and destination have the same
999 * parent we don't fool with the link count.
1001 if (doingdirectory && newparent) {
1002 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1008 dp->i_flag |= IN_CHANGE;
1009 if (DOINGSOFTDEP(tdvp))
1010 softdep_change_linkcnt(dp);
1011 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1016 ufs_makedirentry(ip, tcnp, &newdir);
1017 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1019 if (doingdirectory && newparent) {
1022 dp->i_flag |= IN_CHANGE;
1023 if (DOINGSOFTDEP(tdvp))
1024 softdep_change_linkcnt(dp);
1025 (void)UFS_UPDATE(tdvp, 1);
1029 VN_KNOTE(tdvp, NOTE_WRITE);
1032 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1033 panic("ufs_rename: EXDEV");
1035 * Short circuit rename(foo, foo).
1037 if (xp->i_number == ip->i_number)
1038 panic("ufs_rename: same file");
1040 * If the parent directory is "sticky", then the user must
1041 * own the parent directory, or the destination of the rename,
1042 * otherwise the destination may not be changed (except by
1043 * root). This implements append-only directories.
1045 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1046 tcnp->cn_cred->cr_uid != dp->i_uid &&
1047 xp->i_uid != tcnp->cn_cred->cr_uid) {
1052 * Target must be empty if a directory and have no links
1053 * to it. Also, ensure source and target are compatible
1054 * (both directories, or both not directories).
1056 if ((xp->i_mode&IFMT) == IFDIR) {
1057 if ((xp->i_effnlink > 2) ||
1058 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1062 if (!doingdirectory) {
1067 } else if (doingdirectory) {
1071 error = ufs_dirrewrite(dp, xp, ip->i_number,
1073 (doingdirectory && newparent) ? newparent : doingdirectory);
1076 if (doingdirectory) {
1079 if (DOINGSOFTDEP(tdvp))
1080 softdep_change_linkcnt(dp);
1083 if (DOINGSOFTDEP(tvp))
1084 softdep_change_linkcnt(xp);
1086 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1088 * Truncate inode. The only stuff left in the directory
1089 * is "." and "..". The "." reference is inconsequential
1090 * since we are quashing it. We have removed the "."
1091 * reference and the reference in the parent directory,
1092 * but there may be other hard links. The soft
1093 * dependency code will arrange to do these operations
1094 * after the parent directory entry has been deleted on
1095 * disk, so when running with that code we avoid doing
1100 dp->i_flag |= IN_CHANGE;
1103 xp->i_flag |= IN_CHANGE;
1104 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC;
1105 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
1106 tcnp->cn_cred, tcnp->cn_td)) != 0)
1109 VN_KNOTE(tdvp, NOTE_WRITE);
1111 VN_KNOTE(tvp, NOTE_DELETE);
1117 * 3) Unlink the source.
1119 fcnp->cn_flags &= ~CNP_MODMASK;
1120 fcnp->cn_flags |= CNP_LOCKPARENT | CNP_LOCKLEAF;
1121 if ((fcnp->cn_flags & CNP_SAVESTART) == 0)
1122 panic("ufs_rename: lost from startdir");
1124 error = relookup(fdvp, &fvp, fcnp);
1132 * From name has disappeared.
1135 panic("ufs_rename: lost dir entry");
1140 * Ensure that the directory entry still exists and has not
1141 * changed while the new name has been entered. If the source is
1142 * a file then the entry may have been unlinked or renamed. In
1143 * either case there is no further work to be done. If the source
1144 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1145 * flag ensures that it cannot be moved by another rename or removed
1150 panic("ufs_rename: lost dir entry");
1153 * If the source is a directory with a
1154 * new parent, the link count of the old
1155 * parent directory must be decremented
1156 * and ".." set to point to the new parent.
1158 if (doingdirectory && newparent) {
1159 xp->i_offset = mastertemplate.dot_reclen;
1160 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
1163 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1164 xp->i_flag &= ~IN_RENAME;
1166 VN_KNOTE(fvp, NOTE_RENAME);
1180 ip->i_flag &= ~IN_RENAME;
1181 if (vn_lock(fvp, NULL, LK_EXCLUSIVE, td) == 0) {
1184 ip->i_flag |= IN_CHANGE;
1185 ip->i_flag &= ~IN_RENAME;
1186 if (DOINGSOFTDEP(fvp))
1187 softdep_change_linkcnt(ip);
1197 * ufs_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
1198 * struct componentname *a_cnp, struct vattr *a_vap)
1201 ufs_mkdir(struct vop_mkdir_args *ap)
1203 struct vnode *dvp = ap->a_dvp;
1204 struct vattr *vap = ap->a_vap;
1205 struct componentname *cnp = ap->a_cnp;
1206 struct inode *ip, *dp;
1209 struct dirtemplate dirtemplate, *dtp;
1210 struct direct newdir;
1215 if ((cnp->cn_flags & CNP_HASBUF) == 0)
1216 panic("ufs_mkdir: no name");
1219 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1223 dmode = vap->va_mode & 0777;
1226 * Must simulate part of ufs_makeinode here to acquire the inode,
1227 * but not have it entered in the parent directory. The entry is
1228 * made later after writing "." and ".." entries.
1230 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1234 ip->i_gid = dp->i_gid;
1238 struct ucred ucred, *ucp;
1242 * If we are hacking owners here, (only do this where told to)
1243 * and we are not giving it TO root, (would subvert quotas)
1244 * then go ahead and give it to the other user.
1245 * The new directory also inherits the SUID bit.
1246 * If user's UID and dir UID are the same,
1247 * 'give it away' so that the SUID is still forced on.
1249 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1250 (dp->i_mode & ISUID) && dp->i_uid) {
1252 ip->i_uid = dp->i_uid;
1254 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1256 * Make sure the correct user gets charged
1258 * Make a dummy credential for the victim.
1259 * XXX This seems to never be accessed out of
1260 * our context so a stack variable is ok.
1263 ucred.cr_uid = ip->i_uid;
1264 ucred.cr_ngroups = 1;
1265 ucred.cr_groups[0] = dp->i_gid;
1270 ip->i_uid = cnp->cn_cred->cr_uid;
1272 if ((error = getinoquota(ip)) ||
1273 (error = chkiq(ip, 1, ucp, 0))) {
1274 UFS_VFREE(tvp, ip->i_number, dmode);
1280 #else /* !SUIDDIR */
1281 ip->i_uid = cnp->cn_cred->cr_uid;
1283 if ((error = getinoquota(ip)) ||
1284 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1285 UFS_VFREE(tvp, ip->i_number, dmode);
1290 #endif /* !SUIDDIR */
1291 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1293 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1296 if (DOINGSOFTDEP(tvp))
1297 softdep_change_linkcnt(ip);
1298 if (cnp->cn_flags & CNP_ISWHITEOUT)
1299 ip->i_flags |= UF_OPAQUE;
1302 * Bump link count in parent directory to reflect work done below.
1303 * Should be done before reference is created so cleanup is
1304 * possible if we crash.
1308 dp->i_flag |= IN_CHANGE;
1309 if (DOINGSOFTDEP(dvp))
1310 softdep_change_linkcnt(dp);
1311 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1316 * Initialize directory with "." and ".." from static template.
1318 if (dvp->v_mount->mnt_maxsymlinklen > 0
1320 dtp = &mastertemplate;
1322 dtp = (struct dirtemplate *)&omastertemplate;
1324 dirtemplate.dot_ino = ip->i_number;
1325 dirtemplate.dotdot_ino = dp->i_number;
1326 if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1327 B_CLRBUF, &bp)) != 0)
1329 ip->i_size = DIRBLKSIZ;
1330 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1331 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1332 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1333 if (DOINGSOFTDEP(tvp)) {
1335 * Ensure that the entire newly allocated block is a
1336 * valid directory so that future growth within the
1337 * block does not have to ensure that the block is
1338 * written before the inode.
1341 while (blkoff < bp->b_bcount) {
1343 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1344 blkoff += DIRBLKSIZ;
1347 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1348 DOINGASYNC(tvp)))) != 0) {
1349 (void)VOP_BWRITE(bp->b_vp, bp);
1353 * Directory set up, now install its entry in the parent directory.
1355 * If we are not doing soft dependencies, then we must write out the
1356 * buffer containing the new directory body before entering the new
1357 * name in the parent. If we are doing soft dependencies, then the
1358 * buffer containing the new directory body will be passed to and
1359 * released in the soft dependency code after the code has attached
1360 * an appropriate ordering dependency to the buffer which ensures that
1361 * the buffer is written before the new name is written in the parent.
1363 if (DOINGASYNC(dvp))
1365 else if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp->b_vp, bp))))
1367 ufs_makedirentry(ip, cnp, &newdir);
1368 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1372 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1377 dp->i_flag |= IN_CHANGE;
1378 if (DOINGSOFTDEP(dvp))
1379 softdep_change_linkcnt(dp);
1381 * No need to do an explicit VOP_TRUNCATE here, vrele will
1382 * do this for us because we set the link count to 0.
1386 ip->i_flag |= IN_CHANGE;
1387 if (DOINGSOFTDEP(tvp))
1388 softdep_change_linkcnt(ip);
1396 * Rmdir system call.
1398 * ufs_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1399 * struct componentname *a_cnp)
1402 ufs_rmdir(struct vop_rmdir_args *ap)
1404 struct vnode *vp = ap->a_vp;
1405 struct vnode *dvp = ap->a_dvp;
1406 struct componentname *cnp = ap->a_cnp;
1407 struct inode *ip, *dp;
1414 * Do not remove a directory that is in the process of being renamed.
1415 * Verify the directory is empty (and valid). Rmdir ".." will not be
1416 * valid since ".." will contain a reference to the current directory
1417 * and thus be non-empty. Do not allow the removal of mounted on
1418 * directories (this can happen when an NFS exported filesystem
1419 * tries to remove a locally mounted on directory).
1422 if (ip->i_flag & IN_RENAME) {
1426 if (ip->i_effnlink != 2 ||
1427 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1431 if ((dp->i_flags & APPEND)
1432 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1436 if (vp->v_mountedhere != 0) {
1441 * Delete reference to directory before purging
1442 * inode. If we crash in between, the directory
1443 * will be reattached to lost+found,
1447 if (DOINGSOFTDEP(vp)) {
1448 softdep_change_linkcnt(dp);
1449 softdep_change_linkcnt(ip);
1451 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1455 if (DOINGSOFTDEP(vp)) {
1456 softdep_change_linkcnt(dp);
1457 softdep_change_linkcnt(ip);
1461 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1464 * Truncate inode. The only stuff left in the directory is "." and
1465 * "..". The "." reference is inconsequential since we are quashing
1466 * it. The soft dependency code will arrange to do these operations
1467 * after the parent directory entry has been deleted on disk, so
1468 * when running with that code we avoid doing them now.
1470 if (!DOINGSOFTDEP(vp)) {
1472 dp->i_flag |= IN_CHANGE;
1474 ip->i_flag |= IN_CHANGE;
1475 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC;
1476 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
1481 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
1482 if (ip->i_dirhash != NULL)
1483 ufsdirhash_free(ip);
1486 VN_KNOTE(vp, NOTE_DELETE);
1491 * symlink -- make a symbolic link
1493 * ufs_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1494 * struct componentname *a_cnp, struct vattr *a_vap,
1498 ufs_symlink(struct vop_symlink_args *ap)
1500 struct vnode *vp, **vpp = ap->a_vpp;
1504 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1508 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1510 len = strlen(ap->a_target);
1511 if (len < vp->v_mount->mnt_maxsymlinklen) {
1513 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1515 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1517 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1518 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred,
1526 * Vnode op for reading directories.
1528 * The routine below assumes that the on-disk format of a directory
1529 * is the same as that defined by <sys/dirent.h>. If the on-disk
1530 * format changes, then it will be necessary to do a conversion
1531 * from the on-disk format that read returns to the format defined
1532 * by <sys/dirent.h>.
1534 * ufs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
1535 * int *a_eofflag, int *ncookies, u_long **a_cookies)
1538 ufs_readdir(struct vop_readdir_args *ap)
1540 struct uio *uio = ap->a_uio;
1545 if (ap->a_ncookies != NULL)
1547 * Ensure that the block is aligned. The caller can use
1548 * the cookies to determine where in the block to start.
1550 uio->uio_offset &= ~(DIRBLKSIZ - 1);
1551 off = uio->uio_offset;
1552 count = uio->uio_resid;
1553 /* Make sure we don't return partial entries. */
1554 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
1556 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1557 lost = uio->uio_resid - count;
1558 uio->uio_resid = count;
1559 uio->uio_iov->iov_len = count;
1560 # if (BYTE_ORDER == LITTLE_ENDIAN)
1561 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
1562 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1564 struct dirent *dp, *edp;
1572 auio.uio_iov = &aiov;
1573 auio.uio_iovcnt = 1;
1574 auio.uio_segflg = UIO_SYSSPACE;
1575 aiov.iov_len = count;
1576 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1577 aiov.iov_base = dirbuf;
1578 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
1580 readcnt = count - auio.uio_resid;
1581 edp = (struct dirent *)&dirbuf[readcnt];
1582 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
1584 dp->d_namlen = dp->d_type;
1586 if (dp->d_reclen > 0) {
1587 dp = (struct dirent *)
1588 ((char *)dp + dp->d_reclen);
1595 error = uiomove(dirbuf, readcnt, uio);
1597 FREE(dirbuf, M_TEMP);
1600 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1602 if (!error && ap->a_ncookies != NULL) {
1603 struct dirent* dpStart;
1604 struct dirent* dpEnd;
1610 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1611 panic("ufs_readdir: unexpected uio from NFS server");
1612 dpStart = (struct dirent *)
1613 (uio->uio_iov->iov_base - (uio->uio_offset - off));
1614 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
1615 for (dp = dpStart, ncookies = 0;
1617 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
1619 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1621 for (dp = dpStart, cookiep = cookies;
1623 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
1624 off += dp->d_reclen;
1625 *cookiep++ = (u_long) off;
1627 *ap->a_ncookies = ncookies;
1628 *ap->a_cookies = cookies;
1630 uio->uio_resid += lost;
1632 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
1637 * Return target name of a symbolic link
1639 * ufs_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1642 ufs_readlink(struct vop_readlink_args *ap)
1644 struct vnode *vp = ap->a_vp;
1645 struct inode *ip = VTOI(vp);
1649 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1650 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1651 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1654 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1658 * Calculate the logical to physical mapping if not done already,
1659 * then call the device strategy routine.
1661 * In order to be able to swap to a file, the VOP_BMAP operation may not
1662 * deadlock on memory. See ufs_bmap() for details.
1664 * ufs_strategy(struct vnode *a_vp, struct buf *a_bp)
1667 ufs_strategy(struct vop_strategy_args *ap)
1669 struct buf *bp = ap->a_bp;
1670 struct vnode *vp = ap->a_vp;
1675 if (vp->v_type == VBLK || vp->v_type == VCHR)
1676 panic("ufs_strategy: spec");
1677 if (bp->b_blkno == bp->b_lblkno) {
1678 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
1680 bp->b_error = error;
1681 bp->b_flags |= B_ERROR;
1685 if ((long)bp->b_blkno == -1)
1688 if ((long)bp->b_blkno == -1) {
1693 bp->b_dev = vp->v_rdev;
1694 VOP_STRATEGY(vp, bp);
1699 * Print out the contents of an inode.
1701 * ufs_print(struct vnode *a_vp)
1704 ufs_print(struct vop_print_args *ap)
1706 struct vnode *vp = ap->a_vp;
1707 struct inode *ip = VTOI(vp);
1709 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)",
1710 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1712 if (vp->v_type == VFIFO)
1714 lockmgr_printinfo(&ip->i_lock);
1720 * Read wrapper for special devices.
1722 * ufsspec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1723 * struct ucred *a_cred)
1726 ufsspec_read(struct vop_read_args *ap)
1733 resid = uio->uio_resid;
1734 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap);
1736 * The inode may have been revoked during the call, so it must not
1737 * be accessed blindly here or in the other wrapper functions.
1739 ip = VTOI(ap->a_vp);
1740 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1741 ip->i_flag |= IN_ACCESS;
1746 * Write wrapper for special devices.
1748 * ufsspec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1749 * struct ucred *a_cred)
1752 ufsspec_write(struct vop_write_args *ap)
1759 resid = uio->uio_resid;
1760 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap);
1761 ip = VTOI(ap->a_vp);
1762 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1763 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1768 * Close wrapper for special devices.
1770 * Update the times on the inode then do device close.
1772 * ufsspec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1773 * struct thread *a_td)
1776 ufsspec_close(struct vop_close_args *ap)
1778 struct vnode *vp = ap->a_vp;
1781 lwkt_gettoken(&vlock, vp->v_interlock);
1782 if (vp->v_usecount > 1)
1784 lwkt_reltoken(&vlock);
1785 return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
1789 * Read wrapper for fifos.
1791 * ufsfifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1792 * struct ucred *a_cred)
1795 ufsfifo_read(struct vop_read_args *ap)
1802 resid = uio->uio_resid;
1803 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap);
1804 ip = VTOI(ap->a_vp);
1805 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1806 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1807 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1812 * Write wrapper for fifos.
1814 * ufsfifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1815 * struct ucred *a_cred)
1818 ufsfifo_write(struct vop_write_args *ap)
1825 resid = uio->uio_resid;
1826 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap);
1827 ip = VTOI(ap->a_vp);
1828 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1829 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1834 * Close wrapper for fifos.
1836 * Update the times on the inode then do device close.
1838 * ufsfifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1839 * struct thread *a_td)
1842 ufsfifo_close(struct vop_close_args *ap)
1844 struct vnode *vp = ap->a_vp;
1847 lwkt_gettoken(&vlock, vp->v_interlock);
1848 if (vp->v_usecount > 1)
1850 lwkt_reltoken(&vlock);
1851 return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
1855 * Kqfilter wrapper for fifos.
1857 * Fall through to ufs kqfilter routines if needed
1860 ufsfifo_kqfilter(struct vop_kqfilter_args *ap)
1864 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_kqfilter), ap);
1866 error = ufs_kqfilter(ap);
1871 * Return POSIX pathconf information applicable to ufs filesystems.
1873 * ufs_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
1876 ufs_pathconf(struct vop_pathconf_args *ap)
1878 switch (ap->a_name) {
1880 *ap->a_retval = LINK_MAX;
1883 *ap->a_retval = NAME_MAX;
1886 *ap->a_retval = PATH_MAX;
1889 *ap->a_retval = PIPE_BUF;
1891 case _PC_CHOWN_RESTRICTED:
1904 * Advisory record locking support
1906 * ufs_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
1910 ufs_advlock(struct vop_advlock_args *ap)
1912 struct inode *ip = VTOI(ap->a_vp);
1914 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1918 * Initialize the vnode associated with a new inode, handle aliased
1922 ufs_vinit(struct mount *mntp, vop_t **specops, vop_t **fifoops,
1931 switch(vp->v_type = IFTOVT(ip->i_mode)) {
1935 addaliasu(vp, ip->i_rdev);
1944 if (ip->i_number == ROOTINO)
1945 vp->v_flag |= VROOT;
1947 * Initialize modrev times
1949 getmicrouptime(&tv);
1950 SETHIGH(ip->i_modrev, tv.tv_sec);
1951 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
1957 * Allocate a new inode.
1960 ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1961 struct componentname *cnp)
1963 struct inode *ip, *pdir;
1964 struct direct newdir;
1970 if ((cnp->cn_flags & CNP_HASBUF) == 0)
1971 panic("ufs_makeinode: no name");
1974 if ((mode & IFMT) == 0)
1977 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
1981 ip->i_gid = pdir->i_gid;
1985 struct ucred ucred, *ucp;
1989 * If we are not the owner of the directory,
1990 * and we are hacking owners here, (only do this where told to)
1991 * and we are not giving it TO root, (would subvert quotas)
1992 * then go ahead and give it to the other user.
1993 * Note that this drops off the execute bits for security.
1995 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1996 (pdir->i_mode & ISUID) &&
1997 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1998 ip->i_uid = pdir->i_uid;
2002 * Make sure the correct user gets charged
2004 * Quickly knock up a dummy credential for the victim.
2005 * XXX This seems to never be accessed out of our
2006 * context so a stack variable is ok.
2009 ucred.cr_uid = ip->i_uid;
2010 ucred.cr_ngroups = 1;
2011 ucred.cr_groups[0] = pdir->i_gid;
2015 ip->i_uid = cnp->cn_cred->cr_uid;
2018 if ((error = getinoquota(ip)) ||
2019 (error = chkiq(ip, 1, ucp, 0))) {
2020 UFS_VFREE(tvp, ip->i_number, mode);
2026 #else /* !SUIDDIR */
2027 ip->i_uid = cnp->cn_cred->cr_uid;
2029 if ((error = getinoquota(ip)) ||
2030 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2031 UFS_VFREE(tvp, ip->i_number, mode);
2036 #endif /* !SUIDDIR */
2037 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2039 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2042 if (DOINGSOFTDEP(tvp))
2043 softdep_change_linkcnt(ip);
2044 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
2045 suser_cred(cnp->cn_cred, 0)) {
2046 ip->i_mode &= ~ISGID;
2049 if (cnp->cn_flags & CNP_ISWHITEOUT)
2050 ip->i_flags |= UF_OPAQUE;
2053 * Make sure inode goes to disk before directory entry.
2055 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2058 ufs_makedirentry(ip, cnp, &newdir);
2059 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2067 * Write error occurred trying to update the inode
2068 * or the directory so must deallocate the inode.
2072 ip->i_flag |= IN_CHANGE;
2073 if (DOINGSOFTDEP(tvp))
2074 softdep_change_linkcnt(ip);
2080 ufs_missingop(struct vop_generic_args *ap)
2082 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2083 return (EOPNOTSUPP);
2086 static struct filterops ufsread_filtops =
2087 { 1, NULL, filt_ufsdetach, filt_ufsread };
2088 static struct filterops ufswrite_filtops =
2089 { 1, NULL, filt_ufsdetach, filt_ufswrite };
2090 static struct filterops ufsvnode_filtops =
2091 { 1, NULL, filt_ufsdetach, filt_ufsvnode };
2094 * ufs_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2097 ufs_kqfilter(struct vop_kqfilter_args *ap)
2099 struct vnode *vp = ap->a_vp;
2100 struct knote *kn = ap->a_kn;
2103 switch (kn->kn_filter) {
2105 kn->kn_fop = &ufsread_filtops;
2108 kn->kn_fop = &ufswrite_filtops;
2111 kn->kn_fop = &ufsvnode_filtops;
2117 kn->kn_hook = (caddr_t)vp;
2119 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2120 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
2121 lwkt_reltoken(&ilock);
2127 filt_ufsdetach(struct knote *kn)
2129 struct vnode *vp = (struct vnode *)kn->kn_hook;
2132 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2133 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2134 kn, knote, kn_selnext);
2135 lwkt_reltoken(&ilock);
2140 filt_ufsread(struct knote *kn, long hint)
2142 struct vnode *vp = (struct vnode *)kn->kn_hook;
2143 struct inode *ip = VTOI(vp);
2146 * filesystem is gone, so set the EOF flag and schedule
2147 * the knote for deletion.
2149 if (hint == NOTE_REVOKE) {
2150 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2154 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2155 return (kn->kn_data != 0);
2160 filt_ufswrite(struct knote *kn, long hint)
2163 * filesystem is gone, so set the EOF flag and schedule
2164 * the knote for deletion.
2166 if (hint == NOTE_REVOKE)
2167 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2174 filt_ufsvnode(struct knote *kn, long hint)
2176 if (kn->kn_sfflags & hint)
2177 kn->kn_fflags |= hint;
2178 if (hint == NOTE_REVOKE) {
2179 kn->kn_flags |= EV_EOF;
2182 return (kn->kn_fflags != 0);
2185 /* Global vfs data structures for ufs. */
2186 static vop_t **ufs_vnodeop_p;
2187 static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2188 { &vop_default_desc, (vop_t *) vop_defaultop },
2189 { &vop_fsync_desc, (vop_t *) ufs_missingop },
2190 { &vop_read_desc, (vop_t *) ufs_missingop },
2191 { &vop_reallocblks_desc, (vop_t *) ufs_missingop },
2192 { &vop_write_desc, (vop_t *) ufs_missingop },
2193 { &vop_access_desc, (vop_t *) ufs_access },
2194 { &vop_advlock_desc, (vop_t *) ufs_advlock },
2195 { &vop_bmap_desc, (vop_t *) ufs_bmap },
2196 { &vop_cachedlookup_desc, (vop_t *) ufs_lookup },
2197 { &vop_close_desc, (vop_t *) ufs_close },
2198 { &vop_create_desc, (vop_t *) ufs_create },
2199 { &vop_getattr_desc, (vop_t *) ufs_getattr },
2200 { &vop_inactive_desc, (vop_t *) ufs_inactive },
2201 { &vop_islocked_desc, (vop_t *) vop_stdislocked },
2202 { &vop_link_desc, (vop_t *) ufs_link },
2203 { &vop_lock_desc, (vop_t *) vop_stdlock },
2204 { &vop_lookup_desc, (vop_t *) vfs_cache_lookup },
2205 { &vop_mkdir_desc, (vop_t *) ufs_mkdir },
2206 { &vop_mknod_desc, (vop_t *) ufs_mknod },
2207 { &vop_mmap_desc, (vop_t *) ufs_mmap },
2208 { &vop_open_desc, (vop_t *) ufs_open },
2209 { &vop_pathconf_desc, (vop_t *) ufs_pathconf },
2210 { &vop_poll_desc, (vop_t *) vop_stdpoll },
2211 { &vop_kqfilter_desc, (vop_t *) ufs_kqfilter },
2212 { &vop_print_desc, (vop_t *) ufs_print },
2213 { &vop_readdir_desc, (vop_t *) ufs_readdir },
2214 { &vop_readlink_desc, (vop_t *) ufs_readlink },
2215 { &vop_reclaim_desc, (vop_t *) ufs_reclaim },
2216 { &vop_remove_desc, (vop_t *) ufs_remove },
2217 { &vop_rename_desc, (vop_t *) ufs_rename },
2218 { &vop_rmdir_desc, (vop_t *) ufs_rmdir },
2219 { &vop_setattr_desc, (vop_t *) ufs_setattr },
2220 { &vop_strategy_desc, (vop_t *) ufs_strategy },
2221 { &vop_symlink_desc, (vop_t *) ufs_symlink },
2222 { &vop_unlock_desc, (vop_t *) vop_stdunlock },
2223 { &vop_whiteout_desc, (vop_t *) ufs_whiteout },
2226 static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2227 { &ufs_vnodeop_p, ufs_vnodeop_entries };
2229 static vop_t **ufs_specop_p;
2230 static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2231 { &vop_default_desc, (vop_t *) spec_vnoperate },
2232 { &vop_fsync_desc, (vop_t *) ufs_missingop },
2233 { &vop_access_desc, (vop_t *) ufs_access },
2234 { &vop_close_desc, (vop_t *) ufsspec_close },
2235 { &vop_getattr_desc, (vop_t *) ufs_getattr },
2236 { &vop_inactive_desc, (vop_t *) ufs_inactive },
2237 { &vop_islocked_desc, (vop_t *) vop_stdislocked },
2238 { &vop_lock_desc, (vop_t *) vop_stdlock },
2239 { &vop_print_desc, (vop_t *) ufs_print },
2240 { &vop_read_desc, (vop_t *) ufsspec_read },
2241 { &vop_reclaim_desc, (vop_t *) ufs_reclaim },
2242 { &vop_setattr_desc, (vop_t *) ufs_setattr },
2243 { &vop_unlock_desc, (vop_t *) vop_stdunlock },
2244 { &vop_write_desc, (vop_t *) ufsspec_write },
2247 static struct vnodeopv_desc ufs_specop_opv_desc =
2248 { &ufs_specop_p, ufs_specop_entries };
2250 static vop_t **ufs_fifoop_p;
2251 static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2252 { &vop_default_desc, (vop_t *) fifo_vnoperate },
2253 { &vop_fsync_desc, (vop_t *) ufs_missingop },
2254 { &vop_access_desc, (vop_t *) ufs_access },
2255 { &vop_close_desc, (vop_t *) ufsfifo_close },
2256 { &vop_getattr_desc, (vop_t *) ufs_getattr },
2257 { &vop_inactive_desc, (vop_t *) ufs_inactive },
2258 { &vop_islocked_desc, (vop_t *) vop_stdislocked },
2259 { &vop_kqfilter_desc, (vop_t *) ufsfifo_kqfilter },
2260 { &vop_lock_desc, (vop_t *) vop_stdlock },
2261 { &vop_print_desc, (vop_t *) ufs_print },
2262 { &vop_read_desc, (vop_t *) ufsfifo_read },
2263 { &vop_reclaim_desc, (vop_t *) ufs_reclaim },
2264 { &vop_setattr_desc, (vop_t *) ufs_setattr },
2265 { &vop_unlock_desc, (vop_t *) vop_stdunlock },
2266 { &vop_write_desc, (vop_t *) ufsfifo_write },
2269 static struct vnodeopv_desc ufs_fifoop_opv_desc =
2270 { &ufs_fifoop_p, ufs_fifoop_entries };
2272 VNODEOP_SET(ufs_vnodeop_opv_desc);
2273 VNODEOP_SET(ufs_specop_opv_desc);
2274 VNODEOP_SET(ufs_fifoop_opv_desc);
2277 * ufs_vnoperate(struct vnodeop_desc *a_desc)
2280 ufs_vnoperate(struct vop_generic_args *ap)
2282 return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap));
2286 * ufs_vnoperatefifo(struct vnodeop_desc *a_desc)
2289 ufs_vnoperatefifo(struct vop_generic_args *ap)
2291 return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap));
2295 * ufs_vnoperatespec(struct vnodeop_desc *a_desc)
2298 ufs_vnoperatespec(struct vop_generic_args *ap)
2300 return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap));