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 1.42 2006/04/07 06:38:33 dillon 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 */
68 #include <vm/vm_extern.h>
70 #include <vfs/fifofs/fifo.h>
76 #include "ufs_extern.h"
81 static int ufs_access (struct vop_access_args *);
82 static int ufs_advlock (struct vop_advlock_args *);
83 static int ufs_chmod (struct vnode *, int, struct ucred *, struct thread *);
84 static int ufs_chown (struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
85 static int ufs_close (struct vop_close_args *);
86 static int ufs_create (struct vop_old_create_args *);
87 static int ufs_getattr (struct vop_getattr_args *);
88 static int ufs_link (struct vop_old_link_args *);
89 static int ufs_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
90 static int ufs_missingop (struct vop_generic_args *ap);
91 static int ufs_mkdir (struct vop_old_mkdir_args *);
92 static int ufs_mknod (struct vop_old_mknod_args *);
93 static int ufs_mmap (struct vop_mmap_args *);
94 static int ufs_open (struct vop_open_args *);
95 static int ufs_pathconf (struct vop_pathconf_args *);
96 static int ufs_print (struct vop_print_args *);
97 static int ufs_readdir (struct vop_readdir_args *);
98 static int ufs_readlink (struct vop_readlink_args *);
99 static int ufs_remove (struct vop_old_remove_args *);
100 static int ufs_rename (struct vop_old_rename_args *);
101 static int ufs_rmdir (struct vop_old_rmdir_args *);
102 static int ufs_setattr (struct vop_setattr_args *);
103 static int ufs_strategy (struct vop_strategy_args *);
104 static int ufs_symlink (struct vop_old_symlink_args *);
105 static int ufs_whiteout (struct vop_old_whiteout_args *);
106 static int ufsfifo_close (struct vop_close_args *);
107 static int ufsfifo_kqfilter (struct vop_kqfilter_args *);
108 static int ufsfifo_read (struct vop_read_args *);
109 static int ufsfifo_write (struct vop_write_args *);
110 static int ufsspec_close (struct vop_close_args *);
111 static int ufsspec_read (struct vop_read_args *);
112 static int ufsspec_write (struct vop_write_args *);
113 static int filt_ufsread (struct knote *kn, long hint);
114 static int filt_ufswrite (struct knote *kn, long hint);
115 static int filt_ufsvnode (struct knote *kn, long hint);
116 static void filt_ufsdetach (struct knote *kn);
117 static int ufs_kqfilter (struct vop_kqfilter_args *ap);
123 #define SETHIGH(q, h) { \
126 tmp.val[_QUAD_HIGHWORD] = (h); \
129 #define SETLOW(q, l) { \
132 tmp.val[_QUAD_LOWWORD] = (l); \
135 #define VN_KNOTE(vp, b) \
136 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
138 #define OFSFMT(vp) ((vp)->v_mount->mnt_maxsymlinklen <= 0)
141 * A virgin directory (no blushing please).
143 static struct dirtemplate mastertemplate = {
144 0, 12, DT_DIR, 1, ".",
145 0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
147 static struct odirtemplate omastertemplate = {
149 0, DIRBLKSIZ - 12, 2, ".."
153 ufs_itimes(struct vnode *vp)
159 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
161 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
162 ip->i_flag |= IN_LAZYMOD;
164 ip->i_flag |= IN_MODIFIED;
165 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
167 if (ip->i_flag & IN_ACCESS) {
168 ip->i_atime = ts.tv_sec;
169 ip->i_atimensec = ts.tv_nsec;
171 if (ip->i_flag & IN_UPDATE) {
172 ip->i_mtime = ts.tv_sec;
173 ip->i_mtimensec = ts.tv_nsec;
176 if (ip->i_flag & IN_CHANGE) {
177 ip->i_ctime = ts.tv_sec;
178 ip->i_ctimensec = ts.tv_nsec;
181 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
185 * Create a regular file
187 * ufs_create(struct vnode *a_dvp, struct vnode **a_vpp,
188 * struct componentname *a_cnp, struct vattr *a_vap)
192 ufs_create(struct vop_old_create_args *ap)
197 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
198 ap->a_dvp, ap->a_vpp, ap->a_cnp);
201 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
208 * ufs_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
209 * struct componentname *a_cnp, struct vattr *a_vap)
214 ufs_mknod(struct vop_old_mknod_args *ap)
216 struct vattr *vap = ap->a_vap;
217 struct vnode **vpp = ap->a_vpp;
222 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
223 ap->a_dvp, vpp, ap->a_cnp);
226 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
228 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
229 if (vap->va_rdev != VNOVAL) {
231 * Want to be able to use this to make badblock
232 * inodes, so don't truncate the dev number.
234 ip->i_rdev = vap->va_rdev;
237 * Remove inode, then reload it through VFS_VGET so it is
238 * checked to see if it is an alias of an existing entry in
241 (*vpp)->v_type = VNON;
242 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
245 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
258 * ufs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
259 * struct thread *a_td)
264 ufs_open(struct vop_open_args *ap)
266 struct vnode *vp = ap->a_vp;
269 * Files marked append-only must be opened for appending.
271 if ((VTOI(vp)->i_flags & APPEND) &&
272 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) {
276 return (vop_stdopen(ap));
282 * Update the times on the inode.
284 * ufs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
285 * struct thread *a_td)
290 ufs_close(struct vop_close_args *ap)
292 struct vnode *vp = ap->a_vp;
294 if (vp->v_usecount > 1)
296 return (vop_stdclose(ap));
300 * ufs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
301 * struct thread *a_td)
305 ufs_access(struct vop_access_args *ap)
307 struct vnode *vp = ap->a_vp;
308 struct inode *ip = VTOI(vp);
309 struct ucred *cred = ap->a_cred;
310 mode_t mask, mode = ap->a_mode;
318 * Disallow write attempts on read-only filesystems;
319 * unless the file is a socket, fifo, or a block or
320 * character device resident on the filesystem.
323 switch (vp->v_type) {
327 if (vp->v_mount->mnt_flag & MNT_RDONLY)
330 if ((error = ufs_getinoquota(ip)) != 0)
339 /* If immutable bit set, nobody gets to write it. */
340 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
343 /* Otherwise, user id 0 always gets access. */
344 if (cred->cr_uid == 0)
349 /* Otherwise, check the owner. */
350 if (cred->cr_uid == ip->i_uid) {
357 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
360 /* Otherwise, check the groups. */
361 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
362 if (ip->i_gid == *gp) {
369 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
372 /* Otherwise, check everyone else. */
379 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
383 * ufs_getattr(struct vnode *a_vp, struct vattr *a_vap,
384 * struct thread *a_td)
389 ufs_getattr(struct vop_getattr_args *ap)
391 struct vnode *vp = ap->a_vp;
392 struct inode *ip = VTOI(vp);
393 struct vattr *vap = ap->a_vap;
396 * This may cause i_fsmid to be updated even if no change (0)
397 * is returned, but we should only write out the inode if non-zero
398 * is returned and if the mount is read-write.
400 if (cache_check_fsmid_vp(vp, &ip->i_fsmid) &&
401 (vp->v_mount->mnt_flag & MNT_RDONLY) == 0
403 ip->i_flag |= IN_LAZYMOD;
408 * Copy from inode table
410 vap->va_fsid = dev2udev(ip->i_dev);
411 vap->va_fileid = ip->i_number;
412 vap->va_mode = ip->i_mode & ~IFMT;
413 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ?
414 ip->i_effnlink : ip->i_nlink;
415 vap->va_uid = ip->i_uid;
416 vap->va_gid = ip->i_gid;
417 vap->va_rdev = ip->i_rdev;
418 vap->va_size = ip->i_din.di_size;
419 vap->va_atime.tv_sec = ip->i_atime;
420 vap->va_atime.tv_nsec = ip->i_atimensec;
421 vap->va_mtime.tv_sec = ip->i_mtime;
422 vap->va_mtime.tv_nsec = ip->i_mtimensec;
423 vap->va_ctime.tv_sec = ip->i_ctime;
424 vap->va_ctime.tv_nsec = ip->i_ctimensec;
425 vap->va_flags = ip->i_flags;
426 vap->va_gen = ip->i_gen;
427 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
428 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
429 vap->va_type = IFTOVT(ip->i_mode);
430 vap->va_filerev = ip->i_modrev;
431 vap->va_fsmid = ip->i_fsmid;
436 * Set attribute vnode op. called from several syscalls
438 * ufs_setattr(struct vnode *a_vp, struct vattr *a_vap,
439 * struct ucred *a_cred, struct thread *a_td)
443 ufs_setattr(struct vop_setattr_args *ap)
445 struct vattr *vap = ap->a_vap;
446 struct vnode *vp = ap->a_vp;
447 struct inode *ip = VTOI(vp);
448 struct ucred *cred = ap->a_cred;
452 * Check for unsettable attributes.
454 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
455 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
456 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
457 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
460 if (vap->va_flags != VNOVAL) {
461 if (vp->v_mount->mnt_flag & MNT_RDONLY)
463 if (cred->cr_uid != ip->i_uid &&
464 (error = suser_cred(cred, PRISON_ROOT)))
467 * Note that a root chflags becomes a user chflags when
468 * we are jailed, unless the jail.chflags_allowed sysctl
471 if (cred->cr_uid == 0 &&
472 (!jailed(cred) || jail_chflags_allowed)) {
474 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
477 ip->i_flags = vap->va_flags;
480 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
481 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
483 ip->i_flags &= SF_SETTABLE;
484 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
486 ip->i_flag |= IN_CHANGE;
487 if (vap->va_flags & (IMMUTABLE | APPEND))
490 if (ip->i_flags & (IMMUTABLE | APPEND))
493 * Go through the fields and update iff not VNOVAL.
495 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
496 if (vp->v_mount->mnt_flag & MNT_RDONLY)
498 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
501 if (vap->va_size != VNOVAL) {
503 * Disallow write attempts on read-only filesystems;
504 * unless the file is a socket, fifo, or a block or
505 * character device resident on the filesystem.
507 switch (vp->v_type) {
512 if (vp->v_mount->mnt_flag & MNT_RDONLY)
518 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
522 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
523 if (vp->v_mount->mnt_flag & MNT_RDONLY)
525 if (cred->cr_uid != ip->i_uid &&
526 (error = suser_cred(cred, PRISON_ROOT)) &&
527 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
528 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
530 if (vap->va_atime.tv_sec != VNOVAL)
531 ip->i_flag |= IN_ACCESS;
532 if (vap->va_mtime.tv_sec != VNOVAL)
533 ip->i_flag |= IN_CHANGE | IN_UPDATE;
535 if (vap->va_atime.tv_sec != VNOVAL) {
536 ip->i_atime = vap->va_atime.tv_sec;
537 ip->i_atimensec = vap->va_atime.tv_nsec;
539 if (vap->va_mtime.tv_sec != VNOVAL) {
540 ip->i_mtime = vap->va_mtime.tv_sec;
541 ip->i_mtimensec = vap->va_mtime.tv_nsec;
543 error = UFS_UPDATE(vp, 0);
548 if (vap->va_mode != (mode_t)VNOVAL) {
549 if (vp->v_mount->mnt_flag & MNT_RDONLY)
551 error = ufs_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
553 VN_KNOTE(vp, NOTE_ATTRIB);
558 * Change the mode on a file.
559 * Inode must be locked before calling.
562 ufs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
564 struct inode *ip = VTOI(vp);
567 if (cred->cr_uid != ip->i_uid) {
568 error = suser_cred(cred, PRISON_ROOT);
573 if (vp->v_type != VDIR && (mode & S_ISTXT))
575 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
578 ip->i_mode &= ~ALLPERMS;
579 ip->i_mode |= (mode & ALLPERMS);
580 ip->i_flag |= IN_CHANGE;
585 * Perform chown operation on inode ip;
586 * inode must be locked prior to call.
589 ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
592 struct inode *ip = VTOI(vp);
601 if (uid == (uid_t)VNOVAL)
603 if (gid == (gid_t)VNOVAL)
606 * If we don't own the file, are trying to change the owner
607 * of the file, or are not a member of the target group,
608 * the caller must be superuser or the call fails.
610 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
611 (gid != ip->i_gid && !(cred->cr_gid == gid ||
612 groupmember((gid_t)gid, cred)))) &&
613 (error = suser_cred(cred, PRISON_ROOT)))
618 if ((error = ufs_getinoquota(ip)) != 0)
621 ufs_dqrele(vp, ip->i_dquot[USRQUOTA]);
622 ip->i_dquot[USRQUOTA] = NODQUOT;
625 ufs_dqrele(vp, ip->i_dquot[GRPQUOTA]);
626 ip->i_dquot[GRPQUOTA] = NODQUOT;
628 change = ip->i_blocks;
629 (void) ufs_chkdq(ip, -change, cred, CHOWN);
630 (void) ufs_chkiq(ip, -1, cred, CHOWN);
631 for (i = 0; i < MAXQUOTAS; i++) {
632 ufs_dqrele(vp, ip->i_dquot[i]);
633 ip->i_dquot[i] = NODQUOT;
639 if ((error = ufs_getinoquota(ip)) == 0) {
641 ufs_dqrele(vp, ip->i_dquot[USRQUOTA]);
642 ip->i_dquot[USRQUOTA] = NODQUOT;
645 ufs_dqrele(vp, ip->i_dquot[GRPQUOTA]);
646 ip->i_dquot[GRPQUOTA] = NODQUOT;
648 if ((error = ufs_chkdq(ip, change, cred, CHOWN)) == 0) {
649 if ((error = ufs_chkiq(ip, 1, cred, CHOWN)) == 0)
652 (void)ufs_chkdq(ip, -change, cred, CHOWN|FORCE);
654 for (i = 0; i < MAXQUOTAS; i++) {
655 ufs_dqrele(vp, ip->i_dquot[i]);
656 ip->i_dquot[i] = NODQUOT;
661 if (ufs_getinoquota(ip) == 0) {
663 ufs_dqrele(vp, ip->i_dquot[USRQUOTA]);
664 ip->i_dquot[USRQUOTA] = NODQUOT;
667 ufs_dqrele(vp, ip->i_dquot[GRPQUOTA]);
668 ip->i_dquot[GRPQUOTA] = NODQUOT;
670 (void) ufs_chkdq(ip, change, cred, FORCE|CHOWN);
671 (void) ufs_chkiq(ip, 1, cred, FORCE|CHOWN);
672 (void) ufs_getinoquota(ip);
676 if (ufs_getinoquota(ip))
677 panic("ufs_chown: lost quota");
679 ip->i_flag |= IN_CHANGE;
680 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
681 ip->i_mode &= ~(ISUID | ISGID);
688 * NB Currently unsupported.
690 * ufs_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
691 * struct thread *a_td)
696 ufs_mmap(struct vop_mmap_args *ap)
702 * ufs_remove(struct vnode *a_dvp, struct vnode *a_vp,
703 * struct componentname *a_cnp)
707 ufs_remove(struct vop_old_remove_args *ap)
710 struct vnode *vp = ap->a_vp;
711 struct vnode *dvp = ap->a_dvp;
715 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
716 (VTOI(dvp)->i_flags & APPEND)) {
720 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
721 VN_KNOTE(vp, NOTE_DELETE);
722 VN_KNOTE(dvp, NOTE_WRITE);
730 * ufs_link(struct vnode *a_tdvp, struct vnode *a_vp,
731 * struct componentname *a_cnp)
735 ufs_link(struct vop_old_link_args *ap)
737 struct vnode *vp = ap->a_vp;
738 struct vnode *tdvp = ap->a_tdvp;
739 struct componentname *cnp = ap->a_cnp;
740 struct thread *td = cnp->cn_td;
742 struct direct newdir;
745 if (tdvp->v_mount != vp->v_mount) {
749 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, td))) {
753 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
757 if (ip->i_flags & (IMMUTABLE | APPEND)) {
763 ip->i_flag |= IN_CHANGE;
764 if (DOINGSOFTDEP(vp))
765 softdep_change_linkcnt(ip);
766 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
768 ufs_makedirentry(ip, cnp, &newdir);
769 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
775 ip->i_flag |= IN_CHANGE;
776 if (DOINGSOFTDEP(vp))
777 softdep_change_linkcnt(ip);
781 VOP_UNLOCK(vp, 0, td);
783 VN_KNOTE(vp, NOTE_LINK);
784 VN_KNOTE(tdvp, NOTE_WRITE);
789 * whiteout vnode call
791 * ufs_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
795 ufs_whiteout(struct vop_old_whiteout_args *ap)
797 struct vnode *dvp = ap->a_dvp;
798 struct componentname *cnp = ap->a_cnp;
799 struct direct newdir;
802 switch (ap->a_flags) {
804 /* 4.4 format directories support whiteout operations */
805 if (dvp->v_mount->mnt_maxsymlinklen > 0)
810 /* create a new directory whiteout */
812 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
813 panic("ufs_whiteout: old format filesystem");
817 newdir.d_namlen = cnp->cn_namelen;
818 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
819 newdir.d_type = DT_WHT;
820 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
824 /* remove an existing directory whiteout */
826 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
827 panic("ufs_whiteout: old format filesystem");
830 cnp->cn_flags &= ~CNP_DOWHITEOUT;
831 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
834 panic("ufs_whiteout: unknown op");
840 * Rename system call.
841 * rename("foo", "bar");
844 * link("foo", "bar");
846 * but ``atomically''. Can't do full commit without saving state in the
847 * inode on disk which isn't feasible at this time. Best we can do is
848 * always guarantee the target exists.
850 * Basic algorithm is:
852 * 1) Bump link count on source while we're linking it to the
853 * target. This also ensure the inode won't be deleted out
854 * from underneath us while we work (it may be truncated by
855 * a concurrent `trunc' or `open' for creation).
856 * 2) Link source to destination. If destination already exists,
858 * 3) Unlink source reference to inode if still around. If a
859 * directory was moved and the parent of the destination
860 * is different from the source, patch the ".." entry in the
863 * ufs_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
864 * struct componentname *a_fcnp, struct vnode *a_tdvp,
865 * struct vnode *a_tvp, struct componentname *a_tcnp)
869 ufs_rename(struct vop_old_rename_args *ap)
871 struct vnode *tvp = ap->a_tvp;
872 struct vnode *tdvp = ap->a_tdvp;
873 struct vnode *fvp = ap->a_fvp;
874 struct vnode *fdvp = ap->a_fdvp;
875 struct componentname *tcnp = ap->a_tcnp;
876 struct componentname *fcnp = ap->a_fcnp;
877 struct thread *td = fcnp->cn_td;
878 struct inode *ip, *xp, *dp;
879 struct direct newdir;
880 ino_t oldparent = 0, newparent = 0;
881 int doingdirectory = 0;
882 int error = 0, ioflag;
885 * Check for cross-device rename.
887 if ((fvp->v_mount != tdvp->v_mount) ||
888 (tvp && (fvp->v_mount != tvp->v_mount))) {
902 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
903 (VTOI(tdvp)->i_flags & APPEND))) {
909 * Renaming a file to itself has no effect. The upper layers should
910 * not call us in that case. Temporarily just warn if they do.
913 printf("ufs_rename: fvp == tvp (can't happen)\n");
918 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
922 * Note: now that fvp is locked we have to be sure to unlock it before
923 * using the 'abortit' target.
927 if (ip->i_nlink >= LINK_MAX) {
928 VOP_UNLOCK(fvp, 0, td);
932 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
933 || (dp->i_flags & APPEND)) {
934 VOP_UNLOCK(fvp, 0, td);
938 if ((ip->i_mode & IFMT) == IFDIR) {
940 * Avoid ".", "..", and aliases of "." for obvious reasons.
942 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
943 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
944 (ip->i_flag & IN_RENAME)) {
945 VOP_UNLOCK(fvp, 0, td);
949 ip->i_flag |= IN_RENAME;
950 oldparent = dp->i_number;
953 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
956 * fvp still locked. ip->i_flag has IN_RENAME set if doingdirectory.
957 * Cleanup fvp requirements so we can unlock it.
959 * tvp and tdvp are locked. tvp may be NULL. Now that dp and xp
960 * is setup we can use the 'bad' target if we unlock fvp. We cannot
961 * use the abortit target anymore because of IN_RENAME.
970 * 1) Bump link count while we're moving stuff
971 * around. If we crash somewhere before
972 * completing our work, the link count
973 * may be wrong, but correctable.
977 ip->i_flag |= IN_CHANGE;
978 if (DOINGSOFTDEP(fvp))
979 softdep_change_linkcnt(ip);
980 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
981 DOINGASYNC(fvp)))) != 0) {
982 VOP_UNLOCK(fvp, 0, td);
987 * If ".." must be changed (ie the directory gets a new
988 * parent) then the source directory must not be in the
989 * directory heirarchy above the target, as this would
990 * orphan everything below the source directory. Also
991 * the user must have write permission in the source so
992 * as to be able to change "..". We must repeat the call
993 * to namei, as the parent directory is unlocked by the
994 * call to checkpath().
996 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
997 VOP_UNLOCK(fvp, 0, td);
1000 * We are now back to where we were in that fvp, fdvp are unlocked
1001 * and tvp, tdvp are locked. tvp may be NULL. IN_RENAME may be
1002 * set. Only the bad target or, if we clean up tvp and tdvp, the
1003 * out target, may be used.
1005 if (oldparent != dp->i_number)
1006 newparent = dp->i_number;
1007 if (doingdirectory && newparent) {
1008 if (error) /* write access check above */
1012 * Once we start messing with tvp and tdvp we cannot use the
1013 * 'bad' target, only finish cleaning tdvp and tvp up and
1014 * use the 'out' target.
1016 * This cleans up tvp.
1024 * This is a real mess. ufs_checkpath vput's the target
1025 * directory so retain an extra ref and note that tdvp will
1026 * lose its lock on return. This leaves us with one good
1027 * ref after ufs_checkpath returns.
1030 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
1031 tcnp->cn_flags |= CNP_PDIRUNLOCK;
1038 * relookup no longer messes with tdvp's refs. tdvp must be
1039 * unlocked on entry and will be locked on a successful
1042 error = relookup(tdvp, &tvp, tcnp);
1044 if (tcnp->cn_flags & CNP_PDIRUNLOCK)
1050 KKASSERT((tcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
1057 * We are back to fvp, fdvp unlocked, tvp, tdvp locked. tvp may
1058 * be NULL (xp will also be NULL in that case), and IN_RENAME will
1059 * be set if doingdirectory. This means we can use the 'bad' target
1064 * 2) If target doesn't exist, link the target
1065 * to the source and unlink the source.
1066 * Otherwise, rewrite the target directory
1067 * entry to reference the source inode and
1068 * expunge the original entry's existence.
1071 if (dp->i_dev != ip->i_dev)
1072 panic("ufs_rename: EXDEV");
1074 * Account for ".." in new directory.
1075 * When source and destination have the same
1076 * parent we don't fool with the link count.
1078 if (doingdirectory && newparent) {
1079 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1085 dp->i_flag |= IN_CHANGE;
1086 if (DOINGSOFTDEP(tdvp))
1087 softdep_change_linkcnt(dp);
1088 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1093 ufs_makedirentry(ip, tcnp, &newdir);
1094 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1096 if (doingdirectory && newparent) {
1099 dp->i_flag |= IN_CHANGE;
1100 if (DOINGSOFTDEP(tdvp))
1101 softdep_change_linkcnt(dp);
1102 (void)UFS_UPDATE(tdvp, 1);
1106 VN_KNOTE(tdvp, NOTE_WRITE);
1109 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1110 panic("ufs_rename: EXDEV");
1112 * Short circuit rename(foo, foo).
1114 if (xp->i_number == ip->i_number)
1115 panic("ufs_rename: same file");
1117 * If the parent directory is "sticky", then the user must
1118 * own the parent directory, or the destination of the rename,
1119 * otherwise the destination may not be changed (except by
1120 * root). This implements append-only directories.
1122 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1123 tcnp->cn_cred->cr_uid != dp->i_uid &&
1124 xp->i_uid != tcnp->cn_cred->cr_uid) {
1129 * Target must be empty if a directory and have no links
1130 * to it. Also, ensure source and target are compatible
1131 * (both directories, or both not directories).
1133 * Purge the file or directory being replaced from the
1136 if ((xp->i_mode&IFMT) == IFDIR) {
1137 if ((xp->i_effnlink > 2) ||
1138 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1142 if (!doingdirectory) {
1146 /* cache_purge removed - handled by VFS compat layer */
1147 } else if (doingdirectory == 0) {
1148 /* cache_purge removed - handled by VFS compat layer */
1154 * note: inode passed to ufs_dirrewrite() is 0 for a
1155 * non-directory file rename, 1 for a directory rename
1156 * in the same directory, and > 1 for an inode representing
1157 * the new directory.
1159 error = ufs_dirrewrite(dp, xp, ip->i_number,
1161 (doingdirectory && newparent) ?
1162 newparent : (ino_t)doingdirectory);
1165 if (doingdirectory) {
1168 if (DOINGSOFTDEP(tdvp))
1169 softdep_change_linkcnt(dp);
1172 if (DOINGSOFTDEP(tvp))
1173 softdep_change_linkcnt(xp);
1175 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1177 * Truncate inode. The only stuff left in the directory
1178 * is "." and "..". The "." reference is inconsequential
1179 * since we are quashing it. We have removed the "."
1180 * reference and the reference in the parent directory,
1181 * but there may be other hard links. The soft
1182 * dependency code will arrange to do these operations
1183 * after the parent directory entry has been deleted on
1184 * disk, so when running with that code we avoid doing
1189 dp->i_flag |= IN_CHANGE;
1192 xp->i_flag |= IN_CHANGE;
1193 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC;
1194 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
1195 tcnp->cn_cred, tcnp->cn_td)) != 0)
1198 VN_KNOTE(tdvp, NOTE_WRITE);
1200 VN_KNOTE(tvp, NOTE_DELETE);
1206 * tvp and tdvp have been cleaned up. only fvp and fdvp (both
1207 * unlocked) remain. We are about to overwrite fvp but we have to
1208 * keep 'ip' intact so we cannot release the old fvp, which is still
1209 * refd and accessible via ap->a_fvp.
1211 * This means we cannot use either 'bad' or 'out' to cleanup any
1216 * 3) Unlink the source.
1218 fcnp->cn_flags &= ~CNP_MODMASK;
1219 fcnp->cn_flags |= CNP_LOCKPARENT;
1220 error = relookup(fdvp, &fvp, fcnp);
1221 if (error || fvp == NULL) {
1223 * From name has disappeared. IN_RENAME will not be set if
1224 * we get past the panic so we don't have to clean it up.
1227 panic("ufs_rename: lost dir entry");
1229 if (fcnp->cn_flags & CNP_PDIRUNLOCK)
1235 KKASSERT((fcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
1238 * fdvp and fvp are locked.
1244 * Ensure that the directory entry still exists and has not
1245 * changed while the new name has been entered. If the source is
1246 * a file then the entry may have been unlinked or renamed. In
1247 * either case there is no further work to be done. If the source
1248 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1249 * flag ensures that it cannot be moved by another rename or removed
1250 * by a rmdir. Cleanup IN_RENAME.
1254 panic("ufs_rename: lost dir entry");
1257 * If the source is a directory with a
1258 * new parent, the link count of the old
1259 * parent directory must be decremented
1260 * and ".." set to point to the new parent.
1262 if (doingdirectory && newparent) {
1263 xp->i_offset = mastertemplate.dot_reclen;
1264 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
1265 /* cache_purge removed - handled by VFS compat layer */
1267 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1268 xp->i_flag &= ~IN_RENAME;
1271 VN_KNOTE(fvp, NOTE_RENAME);
1283 ip->i_flag &= ~IN_RENAME;
1284 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
1287 ip->i_flag |= IN_CHANGE;
1288 ip->i_flag &= ~IN_RENAME;
1289 if (DOINGSOFTDEP(fvp))
1290 softdep_change_linkcnt(ip);
1301 * ufs_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
1302 * struct componentname *a_cnp, struct vattr *a_vap)
1306 ufs_mkdir(struct vop_old_mkdir_args *ap)
1308 struct vnode *dvp = ap->a_dvp;
1309 struct vattr *vap = ap->a_vap;
1310 struct componentname *cnp = ap->a_cnp;
1311 struct inode *ip, *dp;
1314 struct dirtemplate dirtemplate, *dtp;
1315 struct direct newdir;
1320 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1324 dmode = vap->va_mode & 0777;
1327 * Must simulate part of ufs_makeinode here to acquire the inode,
1328 * but not have it entered in the parent directory. The entry is
1329 * made later after writing "." and ".." entries.
1331 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1335 ip->i_gid = dp->i_gid;
1339 struct ucred ucred, *ucp;
1343 * If we are hacking owners here, (only do this where told to)
1344 * and we are not giving it TO root, (would subvert quotas)
1345 * then go ahead and give it to the other user.
1346 * The new directory also inherits the SUID bit.
1347 * If user's UID and dir UID are the same,
1348 * 'give it away' so that the SUID is still forced on.
1350 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1351 (dp->i_mode & ISUID) && dp->i_uid) {
1353 ip->i_uid = dp->i_uid;
1355 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1357 * Make sure the correct user gets charged
1359 * Make a dummy credential for the victim.
1360 * XXX This seems to never be accessed out of
1361 * our context so a stack variable is ok.
1364 ucred.cr_uid = ip->i_uid;
1365 ucred.cr_ngroups = 1;
1366 ucred.cr_groups[0] = dp->i_gid;
1371 ip->i_uid = cnp->cn_cred->cr_uid;
1373 if ((error = ufs_getinoquota(ip)) ||
1374 (error = ufs_chkiq(ip, 1, ucp, 0))) {
1375 UFS_VFREE(tvp, ip->i_number, dmode);
1381 #else /* !SUIDDIR */
1382 ip->i_uid = cnp->cn_cred->cr_uid;
1384 if ((error = ufs_getinoquota(ip)) ||
1385 (error = ufs_chkiq(ip, 1, cnp->cn_cred, 0))) {
1386 UFS_VFREE(tvp, ip->i_number, dmode);
1391 #endif /* !SUIDDIR */
1392 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1394 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1397 if (DOINGSOFTDEP(tvp))
1398 softdep_change_linkcnt(ip);
1399 if (cnp->cn_flags & CNP_ISWHITEOUT)
1400 ip->i_flags |= UF_OPAQUE;
1403 * Bump link count in parent directory to reflect work done below.
1404 * Should be done before reference is created so cleanup is
1405 * possible if we crash.
1409 dp->i_flag |= IN_CHANGE;
1410 if (DOINGSOFTDEP(dvp))
1411 softdep_change_linkcnt(dp);
1412 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1417 * The vnode must have a VM object in order to issue buffer cache
1420 vinitvmio(tvp, DIRBLKSIZ);
1423 * Initialize directory with "." and ".." from static template.
1425 if (dvp->v_mount->mnt_maxsymlinklen > 0)
1426 dtp = &mastertemplate;
1428 dtp = (struct dirtemplate *)&omastertemplate;
1430 dirtemplate.dot_ino = ip->i_number;
1431 dirtemplate.dotdot_ino = dp->i_number;
1432 error = VOP_BALLOC(tvp, 0LL, DIRBLKSIZ, cnp->cn_cred, B_CLRBUF, &bp);
1435 ip->i_size = DIRBLKSIZ;
1436 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1437 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1438 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1439 if (DOINGSOFTDEP(tvp)) {
1441 * Ensure that the entire newly allocated block is a
1442 * valid directory so that future growth within the
1443 * block does not have to ensure that the block is
1444 * written before the inode.
1447 while (blkoff < bp->b_bcount) {
1449 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1450 blkoff += DIRBLKSIZ;
1453 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1454 DOINGASYNC(tvp)))) != 0) {
1455 VOP_BWRITE(bp->b_vp, bp);
1459 * Directory set up, now install its entry in the parent directory.
1461 * If we are not doing soft dependencies, then we must write out the
1462 * buffer containing the new directory body before entering the new
1463 * name in the parent. If we are doing soft dependencies, then the
1464 * buffer containing the new directory body will be passed to and
1465 * released in the soft dependency code after the code has attached
1466 * an appropriate ordering dependency to the buffer which ensures that
1467 * the buffer is written before the new name is written in the parent.
1469 if (DOINGASYNC(dvp))
1471 else if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp->b_vp, bp))))
1473 ufs_makedirentry(ip, cnp, &newdir);
1474 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1478 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1483 dp->i_flag |= IN_CHANGE;
1484 if (DOINGSOFTDEP(dvp))
1485 softdep_change_linkcnt(dp);
1487 * No need to do an explicit VOP_TRUNCATE here, vrele will
1488 * do this for us because we set the link count to 0.
1492 ip->i_flag |= IN_CHANGE;
1493 if (DOINGSOFTDEP(tvp))
1494 softdep_change_linkcnt(ip);
1502 * Rmdir system call.
1504 * ufs_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1505 * struct componentname *a_cnp)
1509 ufs_rmdir(struct vop_old_rmdir_args *ap)
1511 struct vnode *vp = ap->a_vp;
1512 struct vnode *dvp = ap->a_dvp;
1513 struct componentname *cnp = ap->a_cnp;
1514 struct inode *ip, *dp;
1521 * Do not remove a directory that is in the process of being renamed.
1522 * Verify the directory is empty (and valid). Rmdir ".." will not be
1523 * valid since ".." will contain a reference to the current directory
1524 * and thus be non-empty. Do not allow the removal of mounted on
1525 * directories (this can happen when an NFS exported filesystem
1526 * tries to remove a locally mounted on directory).
1529 if (ip->i_flag & IN_RENAME) {
1533 if (ip->i_effnlink != 2 ||
1534 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1538 if ((dp->i_flags & APPEND)
1539 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1543 if (vp->v_mountedhere != 0) {
1548 * Delete reference to directory before purging
1549 * inode. If we crash in between, the directory
1550 * will be reattached to lost+found,
1554 if (DOINGSOFTDEP(vp)) {
1555 softdep_change_linkcnt(dp);
1556 softdep_change_linkcnt(ip);
1558 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1562 if (DOINGSOFTDEP(vp)) {
1563 softdep_change_linkcnt(dp);
1564 softdep_change_linkcnt(ip);
1568 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1570 * Truncate inode. The only stuff left in the directory is "." and
1571 * "..". The "." reference is inconsequential since we are quashing
1572 * it. The soft dependency code will arrange to do these operations
1573 * after the parent directory entry has been deleted on disk, so
1574 * when running with that code we avoid doing them now.
1576 if (!DOINGSOFTDEP(vp)) {
1578 dp->i_flag |= IN_CHANGE;
1580 ip->i_flag |= IN_CHANGE;
1581 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC;
1582 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
1585 /* cache_purge removed - handled by VFS compat layer */
1587 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
1588 if (ip->i_dirhash != NULL)
1589 ufsdirhash_free(ip);
1592 VN_KNOTE(vp, NOTE_DELETE);
1597 * symlink -- make a symbolic link
1599 * ufs_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1600 * struct componentname *a_cnp, struct vattr *a_vap,
1605 ufs_symlink(struct vop_old_symlink_args *ap)
1607 struct vnode *vp, **vpp = ap->a_vpp;
1611 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1615 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1617 len = strlen(ap->a_target);
1618 if (len < vp->v_mount->mnt_maxsymlinklen) {
1620 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1622 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1625 * Make sure we have a VM object in order to use
1628 if (vp->v_object == NULL)
1630 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1631 UIO_SYSSPACE, IO_NODELOCKED,
1632 ap->a_cnp->cn_cred, NULL, NULL);
1640 * Vnode op for reading directories.
1642 * The routine below assumes that the on-disk format of a directory
1643 * is the same as that defined by <sys/dirent.h>. If the on-disk
1644 * format changes, then it will be necessary to do a conversion
1645 * from the on-disk format that read returns to the format defined
1646 * by <sys/dirent.h>.
1648 * ufs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
1649 * int *a_eofflag, int *ncookies, u_long **a_cookies)
1653 ufs_readdir(struct vop_readdir_args *ap)
1655 struct uio *uio = ap->a_uio;
1656 struct vnode *vp = ap->a_vp;
1657 struct direct *edp, *dp;
1661 int readcnt, retval;
1664 off_t startoffset = uio->uio_offset;
1666 count = uio->uio_resid;
1668 * Avoid complications for partial directory entries by adjusting
1669 * the i/o to end at a block boundary. Don't give up (like the old ufs
1670 * does) if the initial adjustment gives a negative count, since
1671 * many callers don't supply a large enough buffer. The correct
1672 * size is a little larger than DIRBLKSIZ to allow for expansion
1673 * of directory entries, but some callers just use 512.
1675 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1680 auio.uio_iov = &aiov;
1681 auio.uio_iovcnt = 1;
1682 auio.uio_resid = count;
1683 auio.uio_segflg = UIO_SYSSPACE;
1684 aiov.iov_len = count;
1685 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1686 aiov.iov_base = dirbuf;
1687 error = VOP_READ(vp, &auio, 0, ap->a_cred);
1689 readcnt = count - auio.uio_resid;
1690 edp = (struct direct *)&dirbuf[readcnt];
1692 for (dp = (struct direct *)dirbuf;
1693 !error && uio->uio_resid > 0 && dp < edp; ) {
1694 if (dp->d_reclen <= 0) {
1698 #if BYTE_ORDER == LITTLE_ENDIAN
1700 retval = vop_write_dirent(&error, uio,
1701 dp->d_ino, dp->d_namlen, dp->d_type,
1706 retval = vop_write_dirent(&error, uio,
1707 dp->d_ino, dp->d_type, dp->d_namlen,
1714 dp = (struct direct *)((char *)dp + dp->d_reclen);
1718 /* we need to correct uio_offset */
1719 uio->uio_offset = startoffset + (caddr_t)dp - dirbuf;
1721 if (!error && ap->a_ncookies != NULL) {
1722 u_long *cookiep, *cookies, *ecookies;
1725 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1726 panic("ufs_readdir: unexpected uio from NFS server");
1727 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1730 for (dp = (struct direct *)dirbuf,
1731 cookiep = cookies, ecookies = cookies + ncookies;
1733 dp = (struct direct *)((caddr_t) dp + dp->d_reclen)) {
1734 off += dp->d_reclen;
1735 *cookiep++ = (u_long) off;
1737 *ap->a_ncookies = ncookies;
1738 *ap->a_cookies = cookies;
1741 FREE(dirbuf, M_TEMP);
1743 *ap->a_eofflag = VTOI(vp)->i_size <= uio->uio_offset;
1748 * Return target name of a symbolic link
1750 * ufs_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1754 ufs_readlink(struct vop_readlink_args *ap)
1756 struct vnode *vp = ap->a_vp;
1757 struct inode *ip = VTOI(vp);
1761 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1762 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1763 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1768 * Perform the equivalent of an OPEN on vp so we can issue a
1771 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1775 * Calculate the logical to physical mapping if not done already,
1776 * then call the device strategy routine.
1778 * In order to be able to swap to a file, the VOP_BMAP operation may not
1779 * deadlock on memory. See ufs_bmap() for details.
1781 * ufs_strategy(struct vnode *a_vp, struct bio *a_bio)
1785 ufs_strategy(struct vop_strategy_args *ap)
1787 struct bio *bio = ap->a_bio;
1789 struct buf *bp = bio->bio_buf;
1790 struct vnode *vp = ap->a_vp;
1795 if (vp->v_type == VBLK || vp->v_type == VCHR)
1796 panic("ufs_strategy: spec");
1797 nbio = push_bio(bio);
1798 if (nbio->bio_offset == NOOFFSET) {
1799 error = VOP_BMAP(vp, bio->bio_offset, NULL, &nbio->bio_offset,
1802 bp->b_error = error;
1803 bp->b_flags |= B_ERROR;
1804 /* I/O was never started on nbio, must biodone(bio) */
1808 if (nbio->bio_offset == NOOFFSET)
1811 if (nbio->bio_offset == NOOFFSET) {
1812 /* I/O was never started on nbio, must biodone(bio) */
1816 vn_strategy(ip->i_devvp, nbio);
1821 * Print out the contents of an inode.
1823 * ufs_print(struct vnode *a_vp)
1827 ufs_print(struct vop_print_args *ap)
1829 struct vnode *vp = ap->a_vp;
1830 struct inode *ip = VTOI(vp);
1832 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)",
1833 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1835 if (vp->v_type == VFIFO)
1837 lockmgr_printinfo(&vp->v_lock);
1843 * Read wrapper for special devices.
1845 * ufsspec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1846 * struct ucred *a_cred)
1850 ufsspec_read(struct vop_read_args *ap)
1857 resid = uio->uio_resid;
1858 error = VOCALL(spec_vnode_vops, &ap->a_head);
1860 * The inode may have been revoked during the call, so it must not
1861 * be accessed blindly here or in the other wrapper functions.
1863 ip = VTOI(ap->a_vp);
1864 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1865 ip->i_flag |= IN_ACCESS;
1870 * Write wrapper for special devices.
1872 * ufsspec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1873 * struct ucred *a_cred)
1877 ufsspec_write(struct vop_write_args *ap)
1884 resid = uio->uio_resid;
1885 error = VOCALL(spec_vnode_vops, &ap->a_head);
1886 ip = VTOI(ap->a_vp);
1887 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1888 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1893 * Close wrapper for special devices.
1895 * Update the times on the inode then do device close.
1897 * ufsspec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1898 * struct thread *a_td)
1902 ufsspec_close(struct vop_close_args *ap)
1904 struct vnode *vp = ap->a_vp;
1906 if (vp->v_usecount > 1)
1908 return (VOCALL(spec_vnode_vops, &ap->a_head));
1912 * Read wrapper for fifos.
1914 * ufsfifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1915 * struct ucred *a_cred)
1919 ufsfifo_read(struct vop_read_args *ap)
1926 resid = uio->uio_resid;
1927 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1928 ip = VTOI(ap->a_vp);
1929 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1930 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1931 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1936 * Write wrapper for fifos.
1938 * ufsfifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1939 * struct ucred *a_cred)
1943 ufsfifo_write(struct vop_write_args *ap)
1950 resid = uio->uio_resid;
1951 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1952 ip = VTOI(ap->a_vp);
1953 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1954 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1959 * Close wrapper for fifos.
1961 * Update the times on the inode then do device close.
1963 * ufsfifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1964 * struct thread *a_td)
1968 ufsfifo_close(struct vop_close_args *ap)
1970 struct vnode *vp = ap->a_vp;
1972 if (vp->v_usecount > 1)
1974 return (VOCALL(fifo_vnode_vops, &ap->a_head));
1978 * Kqfilter wrapper for fifos.
1980 * Fall through to ufs kqfilter routines if needed
1984 ufsfifo_kqfilter(struct vop_kqfilter_args *ap)
1988 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1990 error = ufs_kqfilter(ap);
1995 * Return POSIX pathconf information applicable to ufs filesystems.
1997 * ufs_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
2001 ufs_pathconf(struct vop_pathconf_args *ap)
2003 switch (ap->a_name) {
2005 *ap->a_retval = LINK_MAX;
2008 *ap->a_retval = NAME_MAX;
2011 *ap->a_retval = PATH_MAX;
2014 *ap->a_retval = PIPE_BUF;
2016 case _PC_CHOWN_RESTRICTED:
2029 * Advisory record locking support
2031 * ufs_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
2036 ufs_advlock(struct vop_advlock_args *ap)
2038 struct inode *ip = VTOI(ap->a_vp);
2040 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
2044 * Initialize the vnode associated with a new inode, handle aliased
2047 * Make sure directories have their VM object now rather then later,
2048 * saving us from having to check on all the myrid directory VOPs
2049 * that might be executed without a VOP_OPEN being performed.
2052 ufs_vinit(struct mount *mntp, struct vnode **vpp)
2061 vp->v_type = IFTOVT(ip->i_mode);
2063 switch(vp->v_type) {
2066 vp->v_ops = &mntp->mnt_vn_spec_ops;
2067 addaliasu(vp, ip->i_rdev);
2070 vp->v_ops = &mntp->mnt_vn_fifo_ops;
2074 vinitvmio(vp, ip->i_size);
2077 if (ip->i_size >= vp->v_mount->mnt_maxsymlinklen)
2078 vinitvmio(vp, ip->i_size);
2085 if (ip->i_number == ROOTINO)
2086 vp->v_flag |= VROOT;
2088 * Initialize modrev times
2090 getmicrouptime(&tv);
2091 SETHIGH(ip->i_modrev, tv.tv_sec);
2092 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
2098 * Allocate a new inode.
2102 ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
2103 struct componentname *cnp)
2105 struct inode *ip, *pdir;
2106 struct direct newdir;
2112 if ((mode & IFMT) == 0)
2115 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
2119 ip->i_gid = pdir->i_gid;
2123 struct ucred ucred, *ucp;
2127 * If we are not the owner of the directory,
2128 * and we are hacking owners here, (only do this where told to)
2129 * and we are not giving it TO root, (would subvert quotas)
2130 * then go ahead and give it to the other user.
2131 * Note that this drops off the execute bits for security.
2133 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
2134 (pdir->i_mode & ISUID) &&
2135 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
2136 ip->i_uid = pdir->i_uid;
2140 * Make sure the correct user gets charged
2142 * Quickly knock up a dummy credential for the victim.
2143 * XXX This seems to never be accessed out of our
2144 * context so a stack variable is ok.
2147 ucred.cr_uid = ip->i_uid;
2148 ucred.cr_ngroups = 1;
2149 ucred.cr_groups[0] = pdir->i_gid;
2153 ip->i_uid = cnp->cn_cred->cr_uid;
2156 if ((error = ufs_getinoquota(ip)) ||
2157 (error = ufs_chkiq(ip, 1, ucp, 0))) {
2158 UFS_VFREE(tvp, ip->i_number, mode);
2164 #else /* !SUIDDIR */
2165 ip->i_uid = cnp->cn_cred->cr_uid;
2167 if ((error = ufs_getinoquota(ip)) ||
2168 (error = ufs_chkiq(ip, 1, cnp->cn_cred, 0))) {
2169 UFS_VFREE(tvp, ip->i_number, mode);
2174 #endif /* !SUIDDIR */
2175 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2177 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2180 if (DOINGSOFTDEP(tvp))
2181 softdep_change_linkcnt(ip);
2182 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
2183 suser_cred(cnp->cn_cred, 0)) {
2184 ip->i_mode &= ~ISGID;
2187 if (cnp->cn_flags & CNP_ISWHITEOUT)
2188 ip->i_flags |= UF_OPAQUE;
2191 * Regular files and directories need VM objects. Softlinks do
2192 * not (not immediately anyway).
2194 if (tvp->v_type == VREG || tvp->v_type == VDIR)
2198 * Make sure inode goes to disk before directory entry.
2200 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2203 ufs_makedirentry(ip, cnp, &newdir);
2204 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2212 * Write error occurred trying to update the inode
2213 * or the directory so must deallocate the inode.
2217 ip->i_flag |= IN_CHANGE;
2218 if (DOINGSOFTDEP(tvp))
2219 softdep_change_linkcnt(ip);
2225 ufs_missingop(struct vop_generic_args *ap)
2227 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2228 return (EOPNOTSUPP);
2231 static struct filterops ufsread_filtops =
2232 { 1, NULL, filt_ufsdetach, filt_ufsread };
2233 static struct filterops ufswrite_filtops =
2234 { 1, NULL, filt_ufsdetach, filt_ufswrite };
2235 static struct filterops ufsvnode_filtops =
2236 { 1, NULL, filt_ufsdetach, filt_ufsvnode };
2239 * ufs_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2242 ufs_kqfilter(struct vop_kqfilter_args *ap)
2244 struct vnode *vp = ap->a_vp;
2245 struct knote *kn = ap->a_kn;
2248 switch (kn->kn_filter) {
2250 kn->kn_fop = &ufsread_filtops;
2253 kn->kn_fop = &ufswrite_filtops;
2256 kn->kn_fop = &ufsvnode_filtops;
2262 kn->kn_hook = (caddr_t)vp;
2264 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2265 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
2266 lwkt_reltoken(&ilock);
2272 filt_ufsdetach(struct knote *kn)
2274 struct vnode *vp = (struct vnode *)kn->kn_hook;
2277 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2278 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2279 kn, knote, kn_selnext);
2280 lwkt_reltoken(&ilock);
2285 filt_ufsread(struct knote *kn, long hint)
2287 struct vnode *vp = (struct vnode *)kn->kn_hook;
2288 struct inode *ip = VTOI(vp);
2291 * filesystem is gone, so set the EOF flag and schedule
2292 * the knote for deletion.
2294 if (hint == NOTE_REVOKE) {
2295 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2299 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2300 return (kn->kn_data != 0);
2305 filt_ufswrite(struct knote *kn, long hint)
2308 * filesystem is gone, so set the EOF flag and schedule
2309 * the knote for deletion.
2311 if (hint == NOTE_REVOKE)
2312 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2319 filt_ufsvnode(struct knote *kn, long hint)
2321 if (kn->kn_sfflags & hint)
2322 kn->kn_fflags |= hint;
2323 if (hint == NOTE_REVOKE) {
2324 kn->kn_flags |= EV_EOF;
2327 return (kn->kn_fflags != 0);
2330 /* Global vfs data structures for ufs. */
2331 static struct vop_ops *ufs_vnode_vops;
2332 static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2333 { &vop_default_desc, vop_defaultop },
2334 { &vop_fsync_desc, (vnodeopv_entry_t) ufs_missingop },
2335 { &vop_read_desc, (vnodeopv_entry_t) ufs_missingop },
2336 { &vop_reallocblks_desc, (vnodeopv_entry_t) ufs_missingop },
2337 { &vop_write_desc, (vnodeopv_entry_t) ufs_missingop },
2338 { &vop_access_desc, (vnodeopv_entry_t) ufs_access },
2339 { &vop_advlock_desc, (vnodeopv_entry_t) ufs_advlock },
2340 { &vop_bmap_desc, (vnodeopv_entry_t) ufs_bmap },
2341 { &vop_old_lookup_desc, (vnodeopv_entry_t) ufs_lookup },
2342 { &vop_close_desc, (vnodeopv_entry_t) ufs_close },
2343 { &vop_old_create_desc, (vnodeopv_entry_t) ufs_create },
2344 { &vop_getattr_desc, (vnodeopv_entry_t) ufs_getattr },
2345 { &vop_inactive_desc, (vnodeopv_entry_t) ufs_inactive },
2346 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2347 { &vop_old_link_desc, (vnodeopv_entry_t) ufs_link },
2348 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2349 { &vop_old_mkdir_desc, (vnodeopv_entry_t) ufs_mkdir },
2350 { &vop_old_mknod_desc, (vnodeopv_entry_t) ufs_mknod },
2351 { &vop_mmap_desc, (vnodeopv_entry_t) ufs_mmap },
2352 { &vop_open_desc, (vnodeopv_entry_t) ufs_open },
2353 { &vop_pathconf_desc, (vnodeopv_entry_t) ufs_pathconf },
2354 { &vop_poll_desc, (vnodeopv_entry_t) vop_stdpoll },
2355 { &vop_kqfilter_desc, (vnodeopv_entry_t) ufs_kqfilter },
2356 { &vop_print_desc, (vnodeopv_entry_t) ufs_print },
2357 { &vop_readdir_desc, (vnodeopv_entry_t) ufs_readdir },
2358 { &vop_readlink_desc, (vnodeopv_entry_t) ufs_readlink },
2359 { &vop_reclaim_desc, (vnodeopv_entry_t) ufs_reclaim },
2360 { &vop_old_remove_desc, (vnodeopv_entry_t) ufs_remove },
2361 { &vop_old_rename_desc, (vnodeopv_entry_t) ufs_rename },
2362 { &vop_old_rmdir_desc, (vnodeopv_entry_t) ufs_rmdir },
2363 { &vop_setattr_desc, (vnodeopv_entry_t) ufs_setattr },
2364 { &vop_strategy_desc, (vnodeopv_entry_t) ufs_strategy },
2365 { &vop_old_symlink_desc, (vnodeopv_entry_t) ufs_symlink },
2366 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2367 { &vop_old_whiteout_desc, (vnodeopv_entry_t) ufs_whiteout },
2370 static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2371 { &ufs_vnode_vops, ufs_vnodeop_entries, VVF_SUPPORTS_FSMID };
2373 static struct vop_ops *ufs_spec_vops;
2374 static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2375 { &vop_default_desc, (vnodeopv_entry_t) spec_vnoperate },
2376 { &vop_fsync_desc, (vnodeopv_entry_t) ufs_missingop },
2377 { &vop_access_desc, (vnodeopv_entry_t) ufs_access },
2378 { &vop_close_desc, (vnodeopv_entry_t) ufsspec_close },
2379 { &vop_getattr_desc, (vnodeopv_entry_t) ufs_getattr },
2380 { &vop_inactive_desc, (vnodeopv_entry_t) ufs_inactive },
2381 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2382 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2383 { &vop_print_desc, (vnodeopv_entry_t) ufs_print },
2384 { &vop_read_desc, (vnodeopv_entry_t) ufsspec_read },
2385 { &vop_reclaim_desc, (vnodeopv_entry_t) ufs_reclaim },
2386 { &vop_setattr_desc, (vnodeopv_entry_t) ufs_setattr },
2387 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2388 { &vop_write_desc, (vnodeopv_entry_t) ufsspec_write },
2391 static struct vnodeopv_desc ufs_specop_opv_desc =
2392 { &ufs_spec_vops, ufs_specop_entries, VVF_SUPPORTS_FSMID };
2394 static struct vop_ops *ufs_fifo_vops;
2395 static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2396 { &vop_default_desc, (vnodeopv_entry_t) fifo_vnoperate },
2397 { &vop_fsync_desc, (vnodeopv_entry_t) ufs_missingop },
2398 { &vop_access_desc, (vnodeopv_entry_t) ufs_access },
2399 { &vop_close_desc, (vnodeopv_entry_t) ufsfifo_close },
2400 { &vop_getattr_desc, (vnodeopv_entry_t) ufs_getattr },
2401 { &vop_inactive_desc, (vnodeopv_entry_t) ufs_inactive },
2402 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2403 { &vop_kqfilter_desc, (vnodeopv_entry_t) ufsfifo_kqfilter },
2404 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2405 { &vop_print_desc, (vnodeopv_entry_t) ufs_print },
2406 { &vop_read_desc, (vnodeopv_entry_t) ufsfifo_read },
2407 { &vop_reclaim_desc, (vnodeopv_entry_t) ufs_reclaim },
2408 { &vop_setattr_desc, (vnodeopv_entry_t) ufs_setattr },
2409 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2410 { &vop_write_desc, (vnodeopv_entry_t) ufsfifo_write },
2413 static struct vnodeopv_desc ufs_fifoop_opv_desc =
2414 { &ufs_fifo_vops, ufs_fifoop_entries, VVF_SUPPORTS_FSMID };
2416 VNODEOP_SET(ufs_vnodeop_opv_desc);
2417 VNODEOP_SET(ufs_specop_opv_desc);
2418 VNODEOP_SET(ufs_fifoop_opv_desc);
2421 * ufs_vnoperate(struct vnodeop_desc *a_desc)
2424 ufs_vnoperate(struct vop_generic_args *ap)
2426 return (VOCALL(ufs_vnode_vops, ap));
2430 * ufs_vnoperatefifo(struct vnodeop_desc *a_desc)
2433 ufs_vnoperatefifo(struct vop_generic_args *ap)
2435 return (VOCALL(ufs_fifo_vops, ap));
2439 * ufs_vnoperatespec(struct vnodeop_desc *a_desc)
2442 ufs_vnoperatespec(struct vop_generic_args *ap)
2444 return (VOCALL(ufs_spec_vops, ap));