2 * modified for EXT2FS support in Lites 1.1
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
8 * Copyright (c) 1982, 1986, 1989, 1993
9 * The Regents of the University of California. All rights reserved.
10 * (c) UNIX System Laboratories, Inc.
11 * All or some portions of this file are derived from material licensed
12 * to the University of California by American Telephone and Telegraph
13 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
14 * the permission of UNIX System Laboratories, Inc.
16 * Redistribution and use in source and binary forms, with or without
17 * modification, are permitted provided that the following conditions
19 * 1. Redistributions of source code must retain the above copyright
20 * notice, this list of conditions and the following disclaimer.
21 * 2. Redistributions in binary form must reproduce the above copyright
22 * notice, this list of conditions and the following disclaimer in the
23 * documentation and/or other materials provided with the distribution.
24 * 4. Neither the name of the University nor the names of its contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
40 * @(#)ufs_vnops.c 8.7 (Berkeley) 2/3/94
41 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
45 #include "opt_suiddir.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/resourcevar.h>
50 #include <sys/kernel.h>
51 #include <sys/fcntl.h>
57 #include <sys/mount.h>
58 #include <sys/unistd.h>
60 #include <sys/vnode.h>
61 #include <sys/namei.h>
62 #include <sys/lockf.h>
63 #include <sys/event.h>
68 #include <vm/vm_extern.h>
69 #include <vm/vnode_pager.h>
71 #include <fs/fifofs/fifo.h>
73 #include <sys/signalvar.h>
74 #include <ufs/ufs/dir.h>
76 #include <gnu/fs/ext2fs/inode.h>
77 #include <gnu/fs/ext2fs/ext2_mount.h>
78 #include <gnu/fs/ext2fs/ext2_fs_sb.h>
79 #include <gnu/fs/ext2fs/fs.h>
80 #include <gnu/fs/ext2fs/ext2_extern.h>
81 #include <gnu/fs/ext2fs/ext2_fs.h>
83 static int ext2_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
85 static vop_access_t ext2_access;
86 static int ext2_chmod(struct vnode *, int, struct ucred *, struct thread *);
87 static int ext2_chown(struct vnode *, uid_t, gid_t, struct ucred *,
89 static vop_close_t ext2_close;
90 static vop_create_t ext2_create;
91 static vop_fsync_t ext2_fsync;
92 static vop_getattr_t ext2_getattr;
93 static vop_kqfilter_t ext2_kqfilter;
94 static vop_link_t ext2_link;
95 static vop_mkdir_t ext2_mkdir;
96 static vop_mknod_t ext2_mknod;
97 static vop_open_t ext2_open;
98 static vop_pathconf_t ext2_pathconf;
99 static vop_print_t ext2_print;
100 static vop_read_t ext2_read;
101 static vop_readlink_t ext2_readlink;
102 static vop_remove_t ext2_remove;
103 static vop_rename_t ext2_rename;
104 static vop_rmdir_t ext2_rmdir;
105 static vop_setattr_t ext2_setattr;
106 static vop_strategy_t ext2_strategy;
107 static vop_symlink_t ext2_symlink;
108 static vop_write_t ext2_write;
109 static vop_vptofh_t ext2_vptofh;
110 static vop_close_t ext2fifo_close;
111 static vop_kqfilter_t ext2fifo_kqfilter;
112 static int filt_ext2read(struct knote *kn, long hint);
113 static int filt_ext2write(struct knote *kn, long hint);
114 static int filt_ext2vnode(struct knote *kn, long hint);
115 static void filt_ext2detach(struct knote *kn);
117 /* Global vfs data structures for ext2. */
118 struct vop_vector ext2_vnodeops = {
119 .vop_default = &default_vnodeops,
120 .vop_access = ext2_access,
121 .vop_bmap = ext2_bmap,
122 .vop_cachedlookup = ext2_lookup,
123 .vop_close = ext2_close,
124 .vop_create = ext2_create,
125 .vop_fsync = ext2_fsync,
126 .vop_getattr = ext2_getattr,
127 .vop_inactive = ext2_inactive,
128 .vop_link = ext2_link,
129 .vop_lookup = vfs_cache_lookup,
130 .vop_mkdir = ext2_mkdir,
131 .vop_mknod = ext2_mknod,
132 .vop_open = ext2_open,
133 .vop_pathconf = ext2_pathconf,
134 .vop_poll = vop_stdpoll,
135 .vop_kqfilter = ext2_kqfilter,
136 .vop_print = ext2_print,
137 .vop_read = ext2_read,
138 .vop_readdir = ext2_readdir,
139 .vop_readlink = ext2_readlink,
140 .vop_reallocblks = ext2_reallocblks,
141 .vop_reclaim = ext2_reclaim,
142 .vop_remove = ext2_remove,
143 .vop_rename = ext2_rename,
144 .vop_rmdir = ext2_rmdir,
145 .vop_setattr = ext2_setattr,
146 .vop_strategy = ext2_strategy,
147 .vop_symlink = ext2_symlink,
148 .vop_write = ext2_write,
149 .vop_vptofh = ext2_vptofh,
152 struct vop_vector ext2_fifoops = {
153 .vop_default = &fifo_specops,
154 .vop_access = ext2_access,
155 .vop_close = ext2fifo_close,
156 .vop_fsync = ext2_fsync,
157 .vop_getattr = ext2_getattr,
158 .vop_inactive = ext2_inactive,
159 .vop_kqfilter = ext2fifo_kqfilter,
160 .vop_print = ext2_print,
161 .vop_read = VOP_PANIC,
162 .vop_reclaim = ext2_reclaim,
163 .vop_setattr = ext2_setattr,
164 .vop_write = VOP_PANIC,
165 .vop_vptofh = ext2_vptofh,
168 #include <gnu/fs/ext2fs/ext2_readwrite.c>
171 * A virgin directory (no blushing please).
172 * Note that the type and namlen fields are reversed relative to ext2.
173 * Also, we don't use `struct odirtemplate', since it would just cause
174 * endianness problems.
176 static struct dirtemplate mastertemplate = {
177 0, 12, 1, EXT2_FT_DIR, ".",
178 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
180 static struct dirtemplate omastertemplate = {
181 0, 12, 1, EXT2_FT_UNKNOWN, ".",
182 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
193 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
195 if ((vp->v_type == VBLK || vp->v_type == VCHR))
196 ip->i_flag |= IN_LAZYMOD;
198 ip->i_flag |= IN_MODIFIED;
199 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
201 if (ip->i_flag & IN_ACCESS) {
202 ip->i_atime = ts.tv_sec;
203 ip->i_atimensec = ts.tv_nsec;
205 if (ip->i_flag & IN_UPDATE) {
206 ip->i_mtime = ts.tv_sec;
207 ip->i_mtimensec = ts.tv_nsec;
210 if (ip->i_flag & IN_CHANGE) {
211 ip->i_ctime = ts.tv_sec;
212 ip->i_ctimensec = ts.tv_nsec;
215 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
219 * Create a regular file
223 struct vop_create_args /* {
225 struct vnode **a_vpp;
226 struct componentname *a_cnp;
233 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
234 ap->a_dvp, ap->a_vpp, ap->a_cnp);
242 struct vop_open_args /* {
245 struct ucred *a_cred;
250 if (ap->a_vp->v_type == VBLK || ap->a_vp->v_type == VCHR)
254 * Files marked append-only must be opened for appending.
256 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
257 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
260 vnode_create_vobject(ap->a_vp, VTOI(ap->a_vp)->i_size, ap->a_td);
268 * Update the times on the inode.
272 struct vop_close_args /* {
275 struct ucred *a_cred;
279 struct vnode *vp = ap->a_vp;
282 if (vp->v_usecount > 1)
290 struct vop_access_args /* {
293 struct ucred *a_cred;
297 struct vnode *vp = ap->a_vp;
298 struct inode *ip = VTOI(vp);
299 mode_t mode = ap->a_mode;
302 if (vp->v_type == VBLK || vp->v_type == VCHR)
306 * Disallow write attempts on read-only file systems;
307 * unless the file is a socket, fifo, or a block or
308 * character device resident on the file system.
311 switch (vp->v_type) {
315 if (vp->v_mount->mnt_flag & MNT_RDONLY)
323 /* If immutable bit set, nobody gets to write it. */
324 if ((mode & VWRITE) && (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
327 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
328 ap->a_mode, ap->a_cred, NULL);
334 struct vop_getattr_args /* {
337 struct ucred *a_cred;
341 struct vnode *vp = ap->a_vp;
342 struct inode *ip = VTOI(vp);
343 struct vattr *vap = ap->a_vap;
347 * Copy from inode table
349 vap->va_fsid = dev2udev(ip->i_devvp->v_rdev);
350 vap->va_fileid = ip->i_number;
351 vap->va_mode = ip->i_mode & ~IFMT;
352 vap->va_nlink = ip->i_nlink;
353 vap->va_uid = ip->i_uid;
354 vap->va_gid = ip->i_gid;
355 vap->va_rdev = ip->i_rdev;
356 vap->va_size = ip->i_size;
357 vap->va_atime.tv_sec = ip->i_atime;
358 vap->va_atime.tv_nsec = ip->i_atimensec;
359 vap->va_mtime.tv_sec = ip->i_mtime;
360 vap->va_mtime.tv_nsec = ip->i_mtimensec;
361 vap->va_ctime.tv_sec = ip->i_ctime;
362 vap->va_ctime.tv_nsec = ip->i_ctimensec;
363 vap->va_flags = ip->i_flags;
364 vap->va_gen = ip->i_gen;
365 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
366 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
367 vap->va_type = IFTOVT(ip->i_mode);
368 vap->va_filerev = ip->i_modrev;
373 * Set attribute vnode op. called from several syscalls
377 struct vop_setattr_args /* {
380 struct ucred *a_cred;
384 struct vattr *vap = ap->a_vap;
385 struct vnode *vp = ap->a_vp;
386 struct inode *ip = VTOI(vp);
387 struct ucred *cred = ap->a_cred;
388 struct thread *td = ap->a_td;
392 * Check for unsettable attributes.
394 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
395 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
396 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
397 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
400 if (vap->va_flags != VNOVAL) {
401 if (vp->v_mount->mnt_flag & MNT_RDONLY)
404 * Callers may only modify the file flags on objects they
405 * have VADMIN rights for.
407 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
410 * Unprivileged processes and privileged processes in
411 * jail() are not permitted to unset system flags, or
412 * modify flags if any system flags are set.
413 * Privileged non-jail processes may not modify system flags
414 * if securelevel > 0 and any existing system flags are set.
416 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
418 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
419 error = securelevel_gt(cred, 0);
423 ip->i_flags = vap->va_flags;
426 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
427 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
429 ip->i_flags &= SF_SETTABLE;
430 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
432 ip->i_flag |= IN_CHANGE;
433 if (vap->va_flags & (IMMUTABLE | APPEND))
436 if (ip->i_flags & (IMMUTABLE | APPEND))
439 * Go through the fields and update iff not VNOVAL.
441 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
442 if (vp->v_mount->mnt_flag & MNT_RDONLY)
444 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred,
448 if (vap->va_size != VNOVAL) {
450 * Disallow write attempts on read-only file systems;
451 * unless the file is a socket, fifo, or a block or
452 * character device resident on the file system.
454 switch (vp->v_type) {
459 if (vp->v_mount->mnt_flag & MNT_RDONLY)
465 if ((error = ext2_truncate(vp, vap->va_size, 0, cred, td)) != 0)
468 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
469 if (vp->v_mount->mnt_flag & MNT_RDONLY)
473 * If times is NULL, ... The caller must be the owner of
474 * the file, have permission to write the file, or be the
476 * If times is non-NULL, ... The caller must be the owner of
477 * the file or be the super-user.
479 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
480 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
481 (error = VOP_ACCESS(vp, VWRITE, cred, td))))
483 if (vap->va_atime.tv_sec != VNOVAL)
484 ip->i_flag |= IN_ACCESS;
485 if (vap->va_mtime.tv_sec != VNOVAL)
486 ip->i_flag |= IN_CHANGE | IN_UPDATE;
488 if (vap->va_atime.tv_sec != VNOVAL) {
489 ip->i_atime = vap->va_atime.tv_sec;
490 ip->i_atimensec = vap->va_atime.tv_nsec;
492 if (vap->va_mtime.tv_sec != VNOVAL) {
493 ip->i_mtime = vap->va_mtime.tv_sec;
494 ip->i_mtimensec = vap->va_mtime.tv_nsec;
496 error = ext2_update(vp, 0);
501 if (vap->va_mode != (mode_t)VNOVAL) {
502 if (vp->v_mount->mnt_flag & MNT_RDONLY)
504 error = ext2_chmod(vp, (int)vap->va_mode, cred, td);
510 * Change the mode on a file.
511 * Inode must be locked before calling.
514 ext2_chmod(vp, mode, cred, td)
520 struct inode *ip = VTOI(vp);
524 * To modify the permissions on a file, must possess VADMIN
527 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
530 * Privileged processes may set the sticky bit on non-directories,
531 * as well as set the setgid bit on a file with a group that the
532 * process is not a member of.
534 if (vp->v_type != VDIR && (mode & S_ISTXT)) {
535 error = priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0);
539 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
540 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
544 ip->i_mode &= ~ALLPERMS;
545 ip->i_mode |= (mode & ALLPERMS);
546 ip->i_flag |= IN_CHANGE;
551 * Perform chown operation on inode ip;
552 * inode must be locked prior to call.
555 ext2_chown(vp, uid, gid, cred, td)
562 struct inode *ip = VTOI(vp);
567 if (uid == (uid_t)VNOVAL)
569 if (gid == (gid_t)VNOVAL)
572 * To modify the ownership of a file, must possess VADMIN
575 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
578 * To change the owner of a file, or change the group of a file
579 * to a group of which we are not a member, the caller must
582 if (uid != ip->i_uid || (gid != ip->i_gid &&
583 !groupmember(gid, cred))) {
584 error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0);
592 ip->i_flag |= IN_CHANGE;
593 if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
594 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0) != 0)
595 ip->i_mode &= ~(ISUID | ISGID);
601 * Synch an open file.
606 struct vop_fsync_args /* {
608 struct ucred *a_cred;
614 * Flush all dirty buffers associated with a vnode.
616 ext2_discard_prealloc(VTOI(ap->a_vp));
620 return (ext2_update(ap->a_vp, ap->a_waitfor == MNT_WAIT));
629 struct vop_mknod_args /* {
631 struct vnode **a_vpp;
632 struct componentname *a_cnp;
636 struct vattr *vap = ap->a_vap;
637 struct vnode **vpp = ap->a_vpp;
642 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
643 ap->a_dvp, vpp, ap->a_cnp);
647 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
648 if (vap->va_rdev != VNOVAL) {
650 * Want to be able to use this to make badblock
651 * inodes, so don't truncate the dev number.
653 ip->i_rdev = vap->va_rdev;
656 * Remove inode, then reload it through VFS_VGET so it is
657 * checked to see if it is an alias of an existing entry in
658 * the inode cache. XXX I don't believe this is necessary now.
660 (*vpp)->v_type = VNON;
661 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
664 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
674 struct vop_remove_args /* {
677 struct componentname *a_cnp;
681 struct vnode *vp = ap->a_vp;
682 struct vnode *dvp = ap->a_dvp;
686 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
687 (VTOI(dvp)->i_flags & APPEND)) {
691 error = ext2_dirremove(dvp, ap->a_cnp);
694 ip->i_flag |= IN_CHANGE;
705 struct vop_link_args /* {
706 struct vnode *a_tdvp;
708 struct componentname *a_cnp;
711 struct vnode *vp = ap->a_vp;
712 struct vnode *tdvp = ap->a_tdvp;
713 struct componentname *cnp = ap->a_cnp;
718 if ((cnp->cn_flags & HASBUF) == 0)
719 panic("ext2_link: no name");
721 if (tdvp->v_mount != vp->v_mount) {
726 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
730 if (ip->i_flags & (IMMUTABLE | APPEND)) {
735 ip->i_flag |= IN_CHANGE;
736 error = ext2_update(vp, 1);
738 error = ext2_direnter(ip, tdvp, cnp);
741 ip->i_flag |= IN_CHANGE;
748 * Rename system call.
749 * See comments in sys/ufs/ufs/ufs_vnops.c
753 struct vop_rename_args /* {
754 struct vnode *a_fdvp;
756 struct componentname *a_fcnp;
757 struct vnode *a_tdvp;
759 struct componentname *a_tcnp;
762 struct vnode *tvp = ap->a_tvp;
763 struct vnode *tdvp = ap->a_tdvp;
764 struct vnode *fvp = ap->a_fvp;
765 struct vnode *fdvp = ap->a_fdvp;
766 struct componentname *tcnp = ap->a_tcnp;
767 struct componentname *fcnp = ap->a_fcnp;
768 struct thread *td = fcnp->cn_thread;
769 struct inode *ip, *xp, *dp;
770 struct dirtemplate dirbuf;
771 int doingdirectory = 0, oldparent = 0, newparent = 0;
776 if ((tcnp->cn_flags & HASBUF) == 0 ||
777 (fcnp->cn_flags & HASBUF) == 0)
778 panic("ext2_rename: no name");
781 * Check for cross-device rename.
783 if ((fvp->v_mount != tdvp->v_mount) ||
784 (tvp && (fvp->v_mount != tvp->v_mount))) {
798 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
799 (VTOI(tdvp)->i_flags & APPEND))) {
805 * Renaming a file to itself has no effect. The upper layers should
806 * not call us in that case. Temporarily just warn if they do.
809 printf("ext2_rename: fvp == tvp (can't happen)\n");
814 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
818 if (ip->i_nlink >= LINK_MAX) {
819 VOP_UNLOCK(fvp, 0, td);
823 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
824 || (dp->i_flags & APPEND)) {
825 VOP_UNLOCK(fvp, 0, td);
829 if ((ip->i_mode & IFMT) == IFDIR) {
831 * Avoid ".", "..", and aliases of "." for obvious reasons.
833 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
834 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
835 (ip->i_flag & IN_RENAME)) {
836 VOP_UNLOCK(fvp, 0, td);
840 ip->i_flag |= IN_RENAME;
841 oldparent = dp->i_number;
847 * When the target exists, both the directory
848 * and target vnodes are returned locked.
856 * 1) Bump link count while we're moving stuff
857 * around. If we crash somewhere before
858 * completing our work, the link count
859 * may be wrong, but correctable.
862 ip->i_flag |= IN_CHANGE;
863 if ((error = ext2_update(fvp, 1)) != 0) {
864 VOP_UNLOCK(fvp, 0, td);
869 * If ".." must be changed (ie the directory gets a new
870 * parent) then the source directory must not be in the
871 * directory heirarchy above the target, as this would
872 * orphan everything below the source directory. Also
873 * the user must have write permission in the source so
874 * as to be able to change "..". We must repeat the call
875 * to namei, as the parent directory is unlocked by the
876 * call to checkpath().
878 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
879 VOP_UNLOCK(fvp, 0, td);
880 if (oldparent != dp->i_number)
881 newparent = dp->i_number;
882 if (doingdirectory && newparent) {
883 if (error) /* write access check above */
887 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
891 error = relookup(tdvp, &tvp, tcnp);
901 * 2) If target doesn't exist, link the target
902 * to the source and unlink the source.
903 * Otherwise, rewrite the target directory
904 * entry to reference the source inode and
905 * expunge the original entry's existence.
908 if (dp->i_devvp != ip->i_devvp)
909 panic("ext2_rename: EXDEV");
911 * Account for ".." in new directory.
912 * When source and destination have the same
913 * parent we don't fool with the link count.
915 if (doingdirectory && newparent) {
916 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
921 dp->i_flag |= IN_CHANGE;
922 error = ext2_update(tdvp, 1);
926 error = ext2_direnter(ip, tdvp, tcnp);
928 if (doingdirectory && newparent) {
930 dp->i_flag |= IN_CHANGE;
931 (void)ext2_update(tdvp, 1);
937 if (xp->i_devvp != dp->i_devvp || xp->i_devvp != ip->i_devvp)
938 panic("ext2_rename: EXDEV");
940 * Short circuit rename(foo, foo).
942 if (xp->i_number == ip->i_number)
943 panic("ext2_rename: same file");
945 * If the parent directory is "sticky", then the user must
946 * own the parent directory, or the destination of the rename,
947 * otherwise the destination may not be changed (except by
948 * root). This implements append-only directories.
950 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
951 tcnp->cn_cred->cr_uid != dp->i_uid &&
952 xp->i_uid != tcnp->cn_cred->cr_uid) {
957 * Target must be empty if a directory and have no links
958 * to it. Also, ensure source and target are compatible
959 * (both directories, or both not directories).
961 if ((xp->i_mode&IFMT) == IFDIR) {
962 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
967 if (!doingdirectory) {
972 } else if (doingdirectory) {
976 error = ext2_dirrewrite(dp, ip, tcnp);
980 * If the target directory is in the same
981 * directory as the source directory,
982 * decrement the link count on the parent
983 * of the target directory.
985 if (doingdirectory && !newparent) {
987 dp->i_flag |= IN_CHANGE;
991 * Adjust the link count of the target to
992 * reflect the dirrewrite above. If this is
993 * a directory it is empty and there are
994 * no links to it, so we can squash the inode and
995 * any space associated with it. We disallowed
996 * renaming over top of a directory with links to
997 * it above, as the remaining link would point to
998 * a directory without "." or ".." entries.
1001 if (doingdirectory) {
1002 if (--xp->i_nlink != 0)
1003 panic("ext2_rename: linked directory");
1004 error = ext2_truncate(tvp, (off_t)0, IO_SYNC,
1005 tcnp->cn_cred, tcnp->cn_thread);
1007 xp->i_flag |= IN_CHANGE;
1013 * 3) Unlink the source.
1015 fcnp->cn_flags &= ~MODMASK;
1016 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1018 error = relookup(fdvp, &fvp, fcnp);
1026 * From name has disappeared.
1029 panic("ext2_rename: lost dir entry");
1034 * Ensure that the directory entry still exists and has not
1035 * changed while the new name has been entered. If the source is
1036 * a file then the entry may have been unlinked or renamed. In
1037 * either case there is no further work to be done. If the source
1038 * is a directory then it cannot have been rmdir'ed; its link
1039 * count of three would cause a rmdir to fail with ENOTEMPTY.
1040 * The IN_RENAME flag ensures that it cannot be moved by another
1045 panic("ext2_rename: lost dir entry");
1048 * If the source is a directory with a
1049 * new parent, the link count of the old
1050 * parent directory must be decremented
1051 * and ".." set to point to the new parent.
1053 if (doingdirectory && newparent) {
1055 dp->i_flag |= IN_CHANGE;
1056 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
1057 sizeof (struct dirtemplate), (off_t)0,
1058 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1059 tcnp->cn_cred, NOCRED, (int *)0,
1060 (struct thread *)0);
1062 /* Like ufs little-endian: */
1063 namlen = dirbuf.dotdot_type;
1065 dirbuf.dotdot_name[0] != '.' ||
1066 dirbuf.dotdot_name[1] != '.') {
1067 ext2_dirbad(xp, (doff_t)12,
1068 "rename: mangled dir");
1070 dirbuf.dotdot_ino = newparent;
1071 (void) vn_rdwr(UIO_WRITE, fvp,
1073 sizeof (struct dirtemplate),
1074 (off_t)0, UIO_SYSSPACE,
1075 IO_NODELOCKED | IO_SYNC |
1076 IO_NOMACCHECK, tcnp->cn_cred,
1078 (struct thread *)0);
1083 error = ext2_dirremove(fdvp, fcnp);
1086 xp->i_flag |= IN_CHANGE;
1088 xp->i_flag &= ~IN_RENAME;
1103 ip->i_flag &= ~IN_RENAME;
1104 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
1106 ip->i_flag |= IN_CHANGE;
1107 ip->i_flag &= ~IN_RENAME;
1119 struct vop_mkdir_args /* {
1120 struct vnode *a_dvp;
1121 struct vnode **a_vpp;
1122 struct componentname *a_cnp;
1123 struct vattr *a_vap;
1126 struct vnode *dvp = ap->a_dvp;
1127 struct vattr *vap = ap->a_vap;
1128 struct componentname *cnp = ap->a_cnp;
1129 struct inode *ip, *dp;
1131 struct dirtemplate dirtemplate, *dtp;
1135 if ((cnp->cn_flags & HASBUF) == 0)
1136 panic("ext2_mkdir: no name");
1139 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1143 dmode = vap->va_mode & 0777;
1146 * Must simulate part of ext2_makeinode here to acquire the inode,
1147 * but not have it entered in the parent directory. The entry is
1148 * made later after writing "." and ".." entries.
1150 error = ext2_valloc(dvp, dmode, cnp->cn_cred, &tvp);
1154 ip->i_gid = dp->i_gid;
1158 * if we are hacking owners here, (only do this where told to)
1159 * and we are not giving it TOO root, (would subvert quotas)
1160 * then go ahead and give it to the other user.
1161 * The new directory also inherits the SUID bit.
1162 * If user's UID and dir UID are the same,
1163 * 'give it away' so that the SUID is still forced on.
1165 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1166 (dp->i_mode & ISUID) && dp->i_uid) {
1168 ip->i_uid = dp->i_uid;
1170 ip->i_uid = cnp->cn_cred->cr_uid;
1174 ip->i_uid = cnp->cn_cred->cr_uid;
1176 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1178 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1180 if (cnp->cn_flags & ISWHITEOUT)
1181 ip->i_flags |= UF_OPAQUE;
1182 error = ext2_update(tvp, 1);
1185 * Bump link count in parent directory
1186 * to reflect work done below. Should
1187 * be done before reference is created
1188 * so reparation is possible if we crash.
1191 dp->i_flag |= IN_CHANGE;
1192 error = ext2_update(dvp, 1);
1196 /* Initialize directory with "." and ".." from static template. */
1197 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
1198 EXT2_FEATURE_INCOMPAT_FILETYPE))
1199 dtp = &mastertemplate;
1201 dtp = &omastertemplate;
1203 dirtemplate.dot_ino = ip->i_number;
1204 dirtemplate.dotdot_ino = dp->i_number;
1205 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
1206 * so let's just redefine it - for this function only
1209 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->s_blocksize
1210 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
1211 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
1212 sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
1213 IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
1214 (int *)0, (struct thread *)0);
1217 dp->i_flag |= IN_CHANGE;
1220 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
1221 /* XXX should grow with balloc() */
1222 panic("ext2_mkdir: blksize");
1224 ip->i_size = DIRBLKSIZ;
1225 ip->i_flag |= IN_CHANGE;
1228 /* Directory set up, now install its entry in the parent directory. */
1229 error = ext2_direnter(ip, dvp, cnp);
1232 dp->i_flag |= IN_CHANGE;
1236 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
1237 * for us because we set the link count to 0.
1241 ip->i_flag |= IN_CHANGE;
1248 #define DIRBLKSIZ DEV_BSIZE
1252 * Rmdir system call.
1256 struct vop_rmdir_args /* {
1257 struct vnode *a_dvp;
1259 struct componentname *a_cnp;
1262 struct vnode *vp = ap->a_vp;
1263 struct vnode *dvp = ap->a_dvp;
1264 struct componentname *cnp = ap->a_cnp;
1265 struct thread *td = cnp->cn_thread;
1266 struct inode *ip, *dp;
1273 * Verify the directory is empty (and valid).
1274 * (Rmdir ".." won't be valid since
1275 * ".." will contain a reference to
1276 * the current directory and thus be
1280 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1284 if ((dp->i_flags & APPEND)
1285 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1290 * Delete reference to directory before purging
1291 * inode. If we crash in between, the directory
1292 * will be reattached to lost+found,
1294 error = ext2_dirremove(dvp, cnp);
1298 dp->i_flag |= IN_CHANGE;
1300 VOP_UNLOCK(dvp, 0, td);
1302 * Truncate inode. The only stuff left
1303 * in the directory is "." and "..". The
1304 * "." reference is inconsequential since
1305 * we're quashing it. The ".." reference
1306 * has already been adjusted above. We've
1307 * removed the "." reference and the reference
1308 * in the parent directory, but there may be
1309 * other hard links so decrement by 2 and
1310 * worry about them later.
1313 error = ext2_truncate(vp, (off_t)0, IO_SYNC, cnp->cn_cred, td);
1314 cache_purge(ITOV(ip));
1315 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
1321 * symlink -- make a symbolic link
1325 struct vop_symlink_args /* {
1326 struct vnode *a_dvp;
1327 struct vnode **a_vpp;
1328 struct componentname *a_cnp;
1329 struct vattr *a_vap;
1333 struct vnode *vp, **vpp = ap->a_vpp;
1337 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1342 len = strlen(ap->a_target);
1343 if (len < vp->v_mount->mnt_maxsymlinklen) {
1345 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1347 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1349 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1350 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1351 ap->a_cnp->cn_cred, NOCRED, (int *)0, (struct thread *)0);
1358 * Return target name of a symbolic link
1362 struct vop_readlink_args /* {
1365 struct ucred *a_cred;
1368 struct vnode *vp = ap->a_vp;
1369 struct inode *ip = VTOI(vp);
1373 if (isize < vp->v_mount->mnt_maxsymlinklen) {
1374 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1377 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1381 * Calculate the logical to physical mapping if not done already,
1382 * then call the device strategy routine.
1384 * In order to be able to swap to a file, the ext2_bmaparray() operation may not
1385 * deadlock on memory. See ext2_bmap() for details.
1389 struct vop_strategy_args /* {
1394 struct buf *bp = ap->a_bp;
1395 struct vnode *vp = ap->a_vp;
1402 if (vp->v_type == VBLK || vp->v_type == VCHR)
1403 panic("ext2_strategy: spec");
1404 if (bp->b_blkno == bp->b_lblkno) {
1405 error = ext2_bmaparray(vp, bp->b_lblkno, &blkno, NULL, NULL);
1406 bp->b_blkno = blkno;
1408 bp->b_error = error;
1409 bp->b_ioflags |= BIO_ERROR;
1413 if ((long)bp->b_blkno == -1)
1416 if ((long)bp->b_blkno == -1) {
1420 bp->b_iooffset = dbtob(bp->b_blkno);
1421 bo = VFSTOEXT2(vp->v_mount)->um_bo;
1422 BO_STRATEGY(bo, bp);
1427 * Print out the contents of an inode.
1431 struct vop_print_args /* {
1435 struct vnode *vp = ap->a_vp;
1436 struct inode *ip = VTOI(vp);
1438 vn_printf(ip->i_devvp, "\tino %lu", (u_long)ip->i_number);
1439 if (vp->v_type == VFIFO)
1446 * Close wrapper for fifos.
1448 * Update the times on the inode then do device close.
1452 struct vop_close_args /* {
1455 struct ucred *a_cred;
1456 struct thread *a_td;
1459 struct vnode *vp = ap->a_vp;
1462 if (vp->v_usecount > 1)
1465 return (fifo_specops.vop_close(ap));
1469 * Kqfilter wrapper for fifos.
1471 * Fall through to ext2 kqfilter routines if needed
1474 ext2fifo_kqfilter(ap)
1475 struct vop_kqfilter_args *ap;
1479 error = fifo_specops.vop_kqfilter(ap);
1481 error = ext2_kqfilter(ap);
1486 * Return POSIX pathconf information applicable to ext2 filesystems.
1490 struct vop_pathconf_args /* {
1497 switch (ap->a_name) {
1499 *ap->a_retval = LINK_MAX;
1502 *ap->a_retval = NAME_MAX;
1505 *ap->a_retval = PATH_MAX;
1508 *ap->a_retval = PIPE_BUF;
1510 case _PC_CHOWN_RESTRICTED:
1523 * Vnode pointer to File handle
1528 struct vop_vptofh_args /* {
1536 ip = VTOI(ap->a_vp);
1537 ufhp = (struct ufid *)ap->a_fhp;
1538 ufhp->ufid_len = sizeof(struct ufid);
1539 ufhp->ufid_ino = ip->i_number;
1540 ufhp->ufid_gen = ip->i_gen;
1545 * Initialize the vnode associated with a new inode, handle aliased
1549 ext2_vinit(mntp, fifoops, vpp)
1551 struct vop_vector *fifoops;
1559 vp->v_type = IFTOVT(ip->i_mode);
1560 if (vp->v_type == VFIFO)
1563 if (ip->i_number == ROOTINO)
1564 vp->v_vflag |= VV_ROOT;
1565 ip->i_modrev = init_va_filerev();
1571 * Allocate a new inode.
1574 ext2_makeinode(mode, dvp, vpp, cnp)
1578 struct componentname *cnp;
1580 struct inode *ip, *pdir;
1586 if ((cnp->cn_flags & HASBUF) == 0)
1587 panic("ext2_makeinode: no name");
1590 if ((mode & IFMT) == 0)
1593 error = ext2_valloc(dvp, mode, cnp->cn_cred, &tvp);
1598 ip->i_gid = pdir->i_gid;
1603 * not the owner of the directory,
1604 * and we are hacking owners here, (only do this where told to)
1605 * and we are not giving it TOO root, (would subvert quotas)
1606 * then go ahead and give it to the other user.
1607 * Note that this drops off the execute bits for security.
1609 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1610 (pdir->i_mode & ISUID) &&
1611 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1612 ip->i_uid = pdir->i_uid;
1615 ip->i_uid = cnp->cn_cred->cr_uid;
1619 ip->i_uid = cnp->cn_cred->cr_uid;
1621 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1623 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1625 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred)) {
1626 if (priv_check_cred(cnp->cn_cred, PRIV_VFS_RETAINSUGID, 0))
1627 ip->i_mode &= ~ISGID;
1630 if (cnp->cn_flags & ISWHITEOUT)
1631 ip->i_flags |= UF_OPAQUE;
1634 * Make sure inode goes to disk before directory entry.
1636 error = ext2_update(tvp, 1);
1639 error = ext2_direnter(ip, dvp, cnp);
1648 * Write error occurred trying to update the inode
1649 * or the directory so must deallocate the inode.
1652 ip->i_flag |= IN_CHANGE;
1657 static struct filterops ext2read_filtops =
1658 { 1, NULL, filt_ext2detach, filt_ext2read };
1659 static struct filterops ext2write_filtops =
1660 { 1, NULL, filt_ext2detach, filt_ext2write };
1661 static struct filterops ext2vnode_filtops =
1662 { 1, NULL, filt_ext2detach, filt_ext2vnode };
1666 struct vop_kqfilter_args /* {
1671 struct vnode *vp = ap->a_vp;
1672 struct knote *kn = ap->a_kn;
1674 switch (kn->kn_filter) {
1676 kn->kn_fop = &ext2read_filtops;
1679 kn->kn_fop = &ext2write_filtops;
1682 kn->kn_fop = &ext2vnode_filtops;
1688 kn->kn_hook = (caddr_t)vp;
1690 if (vp->v_pollinfo == NULL)
1692 if (vp->v_pollinfo == NULL)
1694 knlist_add(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
1700 filt_ext2detach(struct knote *kn)
1702 struct vnode *vp = (struct vnode *)kn->kn_hook;
1704 KASSERT(vp->v_pollinfo != NULL, ("Mising v_pollinfo"));
1705 knlist_remove(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
1710 filt_ext2read(struct knote *kn, long hint)
1712 struct vnode *vp = (struct vnode *)kn->kn_hook;
1713 struct inode *ip = VTOI(vp);
1716 * filesystem is gone, so set the EOF flag and schedule
1717 * the knote for deletion.
1719 if (hint == NOTE_REVOKE) {
1720 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1724 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
1725 return (kn->kn_data != 0);
1730 filt_ext2write(struct knote *kn, long hint)
1734 * filesystem is gone, so set the EOF flag and schedule
1735 * the knote for deletion.
1737 if (hint == NOTE_REVOKE)
1738 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1745 filt_ext2vnode(struct knote *kn, long hint)
1748 if (kn->kn_sfflags & hint)
1749 kn->kn_fflags |= hint;
1750 if (hint == NOTE_REVOKE) {
1751 kn->kn_flags |= EV_EOF;
1754 return (kn->kn_fflags != 0);