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 vop_advlock_t ext2_advlock;
87 static int ext2_chmod(struct vnode *, int, struct ucred *, struct thread *);
88 static int ext2_chown(struct vnode *, uid_t, gid_t, struct ucred *,
90 static vop_close_t ext2_close;
91 static vop_create_t ext2_create;
92 static vop_fsync_t ext2_fsync;
93 static vop_getattr_t ext2_getattr;
94 static vop_kqfilter_t ext2_kqfilter;
95 static vop_link_t ext2_link;
96 static vop_mkdir_t ext2_mkdir;
97 static vop_mknod_t ext2_mknod;
98 static vop_open_t ext2_open;
99 static vop_pathconf_t ext2_pathconf;
100 static vop_print_t ext2_print;
101 static vop_read_t ext2_read;
102 static vop_readlink_t ext2_readlink;
103 static vop_remove_t ext2_remove;
104 static vop_rename_t ext2_rename;
105 static vop_rmdir_t ext2_rmdir;
106 static vop_setattr_t ext2_setattr;
107 static vop_strategy_t ext2_strategy;
108 static vop_symlink_t ext2_symlink;
109 static vop_write_t ext2_write;
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_advlock = ext2_advlock,
122 .vop_bmap = ext2_bmap,
123 .vop_cachedlookup = ext2_lookup,
124 .vop_close = ext2_close,
125 .vop_create = ext2_create,
126 .vop_fsync = ext2_fsync,
127 .vop_getattr = ext2_getattr,
128 .vop_inactive = ext2_inactive,
129 .vop_link = ext2_link,
130 .vop_lookup = vfs_cache_lookup,
131 .vop_mkdir = ext2_mkdir,
132 .vop_mknod = ext2_mknod,
133 .vop_open = ext2_open,
134 .vop_pathconf = ext2_pathconf,
135 .vop_poll = vop_stdpoll,
136 .vop_kqfilter = ext2_kqfilter,
137 .vop_print = ext2_print,
138 .vop_read = ext2_read,
139 .vop_readdir = ext2_readdir,
140 .vop_readlink = ext2_readlink,
141 .vop_reallocblks = ext2_reallocblks,
142 .vop_reclaim = ext2_reclaim,
143 .vop_remove = ext2_remove,
144 .vop_rename = ext2_rename,
145 .vop_rmdir = ext2_rmdir,
146 .vop_setattr = ext2_setattr,
147 .vop_strategy = ext2_strategy,
148 .vop_symlink = ext2_symlink,
149 .vop_write = ext2_write,
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,
167 #include <gnu/fs/ext2fs/ext2_readwrite.c>
170 * A virgin directory (no blushing please).
171 * Note that the type and namlen fields are reversed relative to ext2.
172 * Also, we don't use `struct odirtemplate', since it would just cause
173 * endianness problems.
175 static struct dirtemplate mastertemplate = {
176 0, 12, 1, EXT2_FT_DIR, ".",
177 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
179 static struct dirtemplate omastertemplate = {
180 0, 12, 1, EXT2_FT_UNKNOWN, ".",
181 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
192 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
194 if ((vp->v_type == VBLK || vp->v_type == VCHR))
195 ip->i_flag |= IN_LAZYMOD;
197 ip->i_flag |= IN_MODIFIED;
198 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
200 if (ip->i_flag & IN_ACCESS) {
201 ip->i_atime = ts.tv_sec;
202 ip->i_atimensec = ts.tv_nsec;
204 if (ip->i_flag & IN_UPDATE) {
205 ip->i_mtime = ts.tv_sec;
206 ip->i_mtimensec = ts.tv_nsec;
209 if (ip->i_flag & IN_CHANGE) {
210 ip->i_ctime = ts.tv_sec;
211 ip->i_ctimensec = ts.tv_nsec;
214 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
218 * Create a regular file
222 struct vop_create_args /* {
224 struct vnode **a_vpp;
225 struct componentname *a_cnp;
232 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
233 ap->a_dvp, ap->a_vpp, ap->a_cnp);
241 struct vop_open_args /* {
244 struct ucred *a_cred;
249 if (ap->a_vp->v_type == VBLK || ap->a_vp->v_type == VCHR)
253 * Files marked append-only must be opened for appending.
255 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
256 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
259 vnode_create_vobject(ap->a_vp, VTOI(ap->a_vp)->i_size, ap->a_td);
267 * Update the times on the inode.
271 struct vop_close_args /* {
274 struct ucred *a_cred;
278 struct vnode *vp = ap->a_vp;
281 if (vp->v_usecount > 1)
289 struct vop_access_args /* {
292 struct ucred *a_cred;
296 struct vnode *vp = ap->a_vp;
297 struct inode *ip = VTOI(vp);
298 mode_t mode = ap->a_mode;
301 if (vp->v_type == VBLK || vp->v_type == VCHR)
305 * Disallow write attempts on read-only file systems;
306 * unless the file is a socket, fifo, or a block or
307 * character device resident on the file system.
310 switch (vp->v_type) {
314 if (vp->v_mount->mnt_flag & MNT_RDONLY)
322 /* If immutable bit set, nobody gets to write it. */
323 if ((mode & VWRITE) && (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
326 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
327 ap->a_mode, ap->a_cred, NULL);
333 struct vop_getattr_args /* {
336 struct ucred *a_cred;
340 struct vnode *vp = ap->a_vp;
341 struct inode *ip = VTOI(vp);
342 struct vattr *vap = ap->a_vap;
346 * Copy from inode table
348 vap->va_fsid = dev2udev(ip->i_devvp->v_rdev);
349 vap->va_fileid = ip->i_number;
350 vap->va_mode = ip->i_mode & ~IFMT;
351 vap->va_nlink = ip->i_nlink;
352 vap->va_uid = ip->i_uid;
353 vap->va_gid = ip->i_gid;
354 vap->va_rdev = ip->i_rdev;
355 vap->va_size = ip->i_size;
356 vap->va_atime.tv_sec = ip->i_atime;
357 vap->va_atime.tv_nsec = ip->i_atimensec;
358 vap->va_mtime.tv_sec = ip->i_mtime;
359 vap->va_mtime.tv_nsec = ip->i_mtimensec;
360 vap->va_ctime.tv_sec = ip->i_ctime;
361 vap->va_ctime.tv_nsec = ip->i_ctimensec;
362 vap->va_flags = ip->i_flags;
363 vap->va_gen = ip->i_gen;
364 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
365 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
366 vap->va_type = IFTOVT(ip->i_mode);
367 vap->va_filerev = ip->i_modrev;
372 * Set attribute vnode op. called from several syscalls
376 struct vop_setattr_args /* {
379 struct ucred *a_cred;
383 struct vattr *vap = ap->a_vap;
384 struct vnode *vp = ap->a_vp;
385 struct inode *ip = VTOI(vp);
386 struct ucred *cred = ap->a_cred;
387 struct thread *td = ap->a_td;
391 * Check for unsettable attributes.
393 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
394 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
395 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
396 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
399 if (vap->va_flags != VNOVAL) {
400 if (vp->v_mount->mnt_flag & MNT_RDONLY)
403 * Callers may only modify the file flags on objects they
404 * have VADMIN rights for.
406 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
409 * Unprivileged processes and privileged processes in
410 * jail() are not permitted to unset system flags, or
411 * modify flags if any system flags are set.
412 * Privileged non-jail processes may not modify system flags
413 * if securelevel > 0 and any existing system flags are set.
415 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS,
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,
540 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
541 error = priv_check_cred(cred, PRIV_VFS_SETGID,
546 ip->i_mode &= ~ALLPERMS;
547 ip->i_mode |= (mode & ALLPERMS);
548 ip->i_flag |= IN_CHANGE;
553 * Perform chown operation on inode ip;
554 * inode must be locked prior to call.
557 ext2_chown(vp, uid, gid, cred, td)
564 struct inode *ip = VTOI(vp);
569 if (uid == (uid_t)VNOVAL)
571 if (gid == (gid_t)VNOVAL)
574 * To modify the ownership of a file, must possess VADMIN
577 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
580 * To change the owner of a file, or change the group of a file
581 * to a group of which we are not a member, the caller must
584 if (uid != ip->i_uid || (gid != ip->i_gid &&
585 !groupmember(gid, cred))) {
586 error = priv_check_cred(cred, PRIV_VFS_CHOWN,
595 ip->i_flag |= IN_CHANGE;
596 if (ouid != uid || ogid != gid) {
597 if (priv_check_cred(cred, PRIV_VFS_CLEARSUGID,
598 SUSER_ALLOWJAIL) != 0)
599 ip->i_mode &= ~(ISUID | ISGID);
605 * Synch an open file.
610 struct vop_fsync_args /* {
612 struct ucred *a_cred;
618 * Flush all dirty buffers associated with a vnode.
620 ext2_discard_prealloc(VTOI(ap->a_vp));
624 return (ext2_update(ap->a_vp, ap->a_waitfor == MNT_WAIT));
633 struct vop_mknod_args /* {
635 struct vnode **a_vpp;
636 struct componentname *a_cnp;
640 struct vattr *vap = ap->a_vap;
641 struct vnode **vpp = ap->a_vpp;
646 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
647 ap->a_dvp, vpp, ap->a_cnp);
651 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
652 if (vap->va_rdev != VNOVAL) {
654 * Want to be able to use this to make badblock
655 * inodes, so don't truncate the dev number.
657 ip->i_rdev = vap->va_rdev;
660 * Remove inode, then reload it through VFS_VGET so it is
661 * checked to see if it is an alias of an existing entry in
662 * the inode cache. XXX I don't believe this is necessary now.
664 (*vpp)->v_type = VNON;
665 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
668 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
678 struct vop_remove_args /* {
681 struct componentname *a_cnp;
685 struct vnode *vp = ap->a_vp;
686 struct vnode *dvp = ap->a_dvp;
690 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
691 (VTOI(dvp)->i_flags & APPEND)) {
695 error = ext2_dirremove(dvp, ap->a_cnp);
698 ip->i_flag |= IN_CHANGE;
709 struct vop_link_args /* {
710 struct vnode *a_tdvp;
712 struct componentname *a_cnp;
715 struct vnode *vp = ap->a_vp;
716 struct vnode *tdvp = ap->a_tdvp;
717 struct componentname *cnp = ap->a_cnp;
722 if ((cnp->cn_flags & HASBUF) == 0)
723 panic("ext2_link: no name");
725 if (tdvp->v_mount != vp->v_mount) {
730 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
734 if (ip->i_flags & (IMMUTABLE | APPEND)) {
739 ip->i_flag |= IN_CHANGE;
740 error = ext2_update(vp, 1);
742 error = ext2_direnter(ip, tdvp, cnp);
745 ip->i_flag |= IN_CHANGE;
752 * Rename system call.
753 * See comments in sys/ufs/ufs/ufs_vnops.c
757 struct vop_rename_args /* {
758 struct vnode *a_fdvp;
760 struct componentname *a_fcnp;
761 struct vnode *a_tdvp;
763 struct componentname *a_tcnp;
766 struct vnode *tvp = ap->a_tvp;
767 struct vnode *tdvp = ap->a_tdvp;
768 struct vnode *fvp = ap->a_fvp;
769 struct vnode *fdvp = ap->a_fdvp;
770 struct componentname *tcnp = ap->a_tcnp;
771 struct componentname *fcnp = ap->a_fcnp;
772 struct thread *td = fcnp->cn_thread;
773 struct inode *ip, *xp, *dp;
774 struct dirtemplate dirbuf;
775 int doingdirectory = 0, oldparent = 0, newparent = 0;
780 if ((tcnp->cn_flags & HASBUF) == 0 ||
781 (fcnp->cn_flags & HASBUF) == 0)
782 panic("ext2_rename: no name");
785 * Check for cross-device rename.
787 if ((fvp->v_mount != tdvp->v_mount) ||
788 (tvp && (fvp->v_mount != tvp->v_mount))) {
802 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
803 (VTOI(tdvp)->i_flags & APPEND))) {
809 * Renaming a file to itself has no effect. The upper layers should
810 * not call us in that case. Temporarily just warn if they do.
813 printf("ext2_rename: fvp == tvp (can't happen)\n");
818 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
822 if (ip->i_nlink >= LINK_MAX) {
823 VOP_UNLOCK(fvp, 0, td);
827 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
828 || (dp->i_flags & APPEND)) {
829 VOP_UNLOCK(fvp, 0, td);
833 if ((ip->i_mode & IFMT) == IFDIR) {
835 * Avoid ".", "..", and aliases of "." for obvious reasons.
837 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
838 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
839 (ip->i_flag & IN_RENAME)) {
840 VOP_UNLOCK(fvp, 0, td);
844 ip->i_flag |= IN_RENAME;
845 oldparent = dp->i_number;
851 * When the target exists, both the directory
852 * and target vnodes are returned locked.
860 * 1) Bump link count while we're moving stuff
861 * around. If we crash somewhere before
862 * completing our work, the link count
863 * may be wrong, but correctable.
866 ip->i_flag |= IN_CHANGE;
867 if ((error = ext2_update(fvp, 1)) != 0) {
868 VOP_UNLOCK(fvp, 0, td);
873 * If ".." must be changed (ie the directory gets a new
874 * parent) then the source directory must not be in the
875 * directory heirarchy above the target, as this would
876 * orphan everything below the source directory. Also
877 * the user must have write permission in the source so
878 * as to be able to change "..". We must repeat the call
879 * to namei, as the parent directory is unlocked by the
880 * call to checkpath().
882 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
883 VOP_UNLOCK(fvp, 0, td);
884 if (oldparent != dp->i_number)
885 newparent = dp->i_number;
886 if (doingdirectory && newparent) {
887 if (error) /* write access check above */
891 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
895 error = relookup(tdvp, &tvp, tcnp);
905 * 2) If target doesn't exist, link the target
906 * to the source and unlink the source.
907 * Otherwise, rewrite the target directory
908 * entry to reference the source inode and
909 * expunge the original entry's existence.
912 if (dp->i_devvp != ip->i_devvp)
913 panic("ext2_rename: EXDEV");
915 * Account for ".." in new directory.
916 * When source and destination have the same
917 * parent we don't fool with the link count.
919 if (doingdirectory && newparent) {
920 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
925 dp->i_flag |= IN_CHANGE;
926 error = ext2_update(tdvp, 1);
930 error = ext2_direnter(ip, tdvp, tcnp);
932 if (doingdirectory && newparent) {
934 dp->i_flag |= IN_CHANGE;
935 (void)ext2_update(tdvp, 1);
941 if (xp->i_devvp != dp->i_devvp || xp->i_devvp != ip->i_devvp)
942 panic("ext2_rename: EXDEV");
944 * Short circuit rename(foo, foo).
946 if (xp->i_number == ip->i_number)
947 panic("ext2_rename: same file");
949 * If the parent directory is "sticky", then the user must
950 * own the parent directory, or the destination of the rename,
951 * otherwise the destination may not be changed (except by
952 * root). This implements append-only directories.
954 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
955 tcnp->cn_cred->cr_uid != dp->i_uid &&
956 xp->i_uid != tcnp->cn_cred->cr_uid) {
961 * Target must be empty if a directory and have no links
962 * to it. Also, ensure source and target are compatible
963 * (both directories, or both not directories).
965 if ((xp->i_mode&IFMT) == IFDIR) {
966 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
971 if (!doingdirectory) {
976 } else if (doingdirectory) {
980 error = ext2_dirrewrite(dp, ip, tcnp);
984 * If the target directory is in the same
985 * directory as the source directory,
986 * decrement the link count on the parent
987 * of the target directory.
989 if (doingdirectory && !newparent) {
991 dp->i_flag |= IN_CHANGE;
995 * Adjust the link count of the target to
996 * reflect the dirrewrite above. If this is
997 * a directory it is empty and there are
998 * no links to it, so we can squash the inode and
999 * any space associated with it. We disallowed
1000 * renaming over top of a directory with links to
1001 * it above, as the remaining link would point to
1002 * a directory without "." or ".." entries.
1005 if (doingdirectory) {
1006 if (--xp->i_nlink != 0)
1007 panic("ext2_rename: linked directory");
1008 error = ext2_truncate(tvp, (off_t)0, IO_SYNC,
1009 tcnp->cn_cred, tcnp->cn_thread);
1011 xp->i_flag |= IN_CHANGE;
1017 * 3) Unlink the source.
1019 fcnp->cn_flags &= ~MODMASK;
1020 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1022 error = relookup(fdvp, &fvp, fcnp);
1030 * From name has disappeared.
1033 panic("ext2_rename: lost dir entry");
1038 * Ensure that the directory entry still exists and has not
1039 * changed while the new name has been entered. If the source is
1040 * a file then the entry may have been unlinked or renamed. In
1041 * either case there is no further work to be done. If the source
1042 * is a directory then it cannot have been rmdir'ed; its link
1043 * count of three would cause a rmdir to fail with ENOTEMPTY.
1044 * The IN_RENAME flag ensures that it cannot be moved by another
1049 panic("ext2_rename: lost dir entry");
1052 * If the source is a directory with a
1053 * new parent, the link count of the old
1054 * parent directory must be decremented
1055 * and ".." set to point to the new parent.
1057 if (doingdirectory && newparent) {
1059 dp->i_flag |= IN_CHANGE;
1060 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
1061 sizeof (struct dirtemplate), (off_t)0,
1062 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1063 tcnp->cn_cred, NOCRED, (int *)0,
1064 (struct thread *)0);
1066 /* Like ufs little-endian: */
1067 namlen = dirbuf.dotdot_type;
1069 dirbuf.dotdot_name[0] != '.' ||
1070 dirbuf.dotdot_name[1] != '.') {
1071 ext2_dirbad(xp, (doff_t)12,
1072 "rename: mangled dir");
1074 dirbuf.dotdot_ino = newparent;
1075 (void) vn_rdwr(UIO_WRITE, fvp,
1077 sizeof (struct dirtemplate),
1078 (off_t)0, UIO_SYSSPACE,
1079 IO_NODELOCKED | IO_SYNC |
1080 IO_NOMACCHECK, tcnp->cn_cred,
1082 (struct thread *)0);
1087 error = ext2_dirremove(fdvp, fcnp);
1090 xp->i_flag |= IN_CHANGE;
1092 xp->i_flag &= ~IN_RENAME;
1107 ip->i_flag &= ~IN_RENAME;
1108 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
1110 ip->i_flag |= IN_CHANGE;
1111 ip->i_flag &= ~IN_RENAME;
1123 struct vop_mkdir_args /* {
1124 struct vnode *a_dvp;
1125 struct vnode **a_vpp;
1126 struct componentname *a_cnp;
1127 struct vattr *a_vap;
1130 struct vnode *dvp = ap->a_dvp;
1131 struct vattr *vap = ap->a_vap;
1132 struct componentname *cnp = ap->a_cnp;
1133 struct inode *ip, *dp;
1135 struct dirtemplate dirtemplate, *dtp;
1139 if ((cnp->cn_flags & HASBUF) == 0)
1140 panic("ext2_mkdir: no name");
1143 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1147 dmode = vap->va_mode & 0777;
1150 * Must simulate part of ext2_makeinode here to acquire the inode,
1151 * but not have it entered in the parent directory. The entry is
1152 * made later after writing "." and ".." entries.
1154 error = ext2_valloc(dvp, dmode, cnp->cn_cred, &tvp);
1158 ip->i_gid = dp->i_gid;
1162 * if we are hacking owners here, (only do this where told to)
1163 * and we are not giving it TOO root, (would subvert quotas)
1164 * then go ahead and give it to the other user.
1165 * The new directory also inherits the SUID bit.
1166 * If user's UID and dir UID are the same,
1167 * 'give it away' so that the SUID is still forced on.
1169 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1170 (dp->i_mode & ISUID) && dp->i_uid) {
1172 ip->i_uid = dp->i_uid;
1174 ip->i_uid = cnp->cn_cred->cr_uid;
1178 ip->i_uid = cnp->cn_cred->cr_uid;
1180 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1182 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1184 if (cnp->cn_flags & ISWHITEOUT)
1185 ip->i_flags |= UF_OPAQUE;
1186 error = ext2_update(tvp, 1);
1189 * Bump link count in parent directory
1190 * to reflect work done below. Should
1191 * be done before reference is created
1192 * so reparation is possible if we crash.
1195 dp->i_flag |= IN_CHANGE;
1196 error = ext2_update(dvp, 1);
1200 /* Initialize directory with "." and ".." from static template. */
1201 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
1202 EXT2_FEATURE_INCOMPAT_FILETYPE))
1203 dtp = &mastertemplate;
1205 dtp = &omastertemplate;
1207 dirtemplate.dot_ino = ip->i_number;
1208 dirtemplate.dotdot_ino = dp->i_number;
1209 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
1210 * so let's just redefine it - for this function only
1213 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->s_blocksize
1214 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
1215 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
1216 sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
1217 IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
1218 (int *)0, (struct thread *)0);
1221 dp->i_flag |= IN_CHANGE;
1224 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
1225 /* XXX should grow with balloc() */
1226 panic("ext2_mkdir: blksize");
1228 ip->i_size = DIRBLKSIZ;
1229 ip->i_flag |= IN_CHANGE;
1232 /* Directory set up, now install its entry in the parent directory. */
1233 error = ext2_direnter(ip, dvp, cnp);
1236 dp->i_flag |= IN_CHANGE;
1240 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
1241 * for us because we set the link count to 0.
1245 ip->i_flag |= IN_CHANGE;
1252 #define DIRBLKSIZ DEV_BSIZE
1256 * Rmdir system call.
1260 struct vop_rmdir_args /* {
1261 struct vnode *a_dvp;
1263 struct componentname *a_cnp;
1266 struct vnode *vp = ap->a_vp;
1267 struct vnode *dvp = ap->a_dvp;
1268 struct componentname *cnp = ap->a_cnp;
1269 struct thread *td = cnp->cn_thread;
1270 struct inode *ip, *dp;
1277 * Verify the directory is empty (and valid).
1278 * (Rmdir ".." won't be valid since
1279 * ".." will contain a reference to
1280 * the current directory and thus be
1284 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1288 if ((dp->i_flags & APPEND)
1289 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1294 * Delete reference to directory before purging
1295 * inode. If we crash in between, the directory
1296 * will be reattached to lost+found,
1298 error = ext2_dirremove(dvp, cnp);
1302 dp->i_flag |= IN_CHANGE;
1304 VOP_UNLOCK(dvp, 0, td);
1306 * Truncate inode. The only stuff left
1307 * in the directory is "." and "..". The
1308 * "." reference is inconsequential since
1309 * we're quashing it. The ".." reference
1310 * has already been adjusted above. We've
1311 * removed the "." reference and the reference
1312 * in the parent directory, but there may be
1313 * other hard links so decrement by 2 and
1314 * worry about them later.
1317 error = ext2_truncate(vp, (off_t)0, IO_SYNC, cnp->cn_cred, td);
1318 cache_purge(ITOV(ip));
1319 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
1325 * symlink -- make a symbolic link
1329 struct vop_symlink_args /* {
1330 struct vnode *a_dvp;
1331 struct vnode **a_vpp;
1332 struct componentname *a_cnp;
1333 struct vattr *a_vap;
1337 struct vnode *vp, **vpp = ap->a_vpp;
1341 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1346 len = strlen(ap->a_target);
1347 if (len < vp->v_mount->mnt_maxsymlinklen) {
1349 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1351 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1353 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1354 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1355 ap->a_cnp->cn_cred, NOCRED, (int *)0, (struct thread *)0);
1362 * Return target name of a symbolic link
1366 struct vop_readlink_args /* {
1369 struct ucred *a_cred;
1372 struct vnode *vp = ap->a_vp;
1373 struct inode *ip = VTOI(vp);
1377 if (isize < vp->v_mount->mnt_maxsymlinklen) {
1378 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1381 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1385 * Calculate the logical to physical mapping if not done already,
1386 * then call the device strategy routine.
1388 * In order to be able to swap to a file, the ext2_bmaparray() operation may not
1389 * deadlock on memory. See ext2_bmap() for details.
1393 struct vop_strategy_args /* {
1398 struct buf *bp = ap->a_bp;
1399 struct vnode *vp = ap->a_vp;
1406 if (vp->v_type == VBLK || vp->v_type == VCHR)
1407 panic("ext2_strategy: spec");
1408 if (bp->b_blkno == bp->b_lblkno) {
1409 error = ext2_bmaparray(vp, bp->b_lblkno, &blkno, NULL, NULL);
1410 bp->b_blkno = blkno;
1412 bp->b_error = error;
1413 bp->b_ioflags |= BIO_ERROR;
1417 if ((long)bp->b_blkno == -1)
1420 if ((long)bp->b_blkno == -1) {
1424 bp->b_iooffset = dbtob(bp->b_blkno);
1425 bo = VFSTOEXT2(vp->v_mount)->um_bo;
1426 BO_STRATEGY(bo, bp);
1431 * Print out the contents of an inode.
1435 struct vop_print_args /* {
1439 struct vnode *vp = ap->a_vp;
1440 struct inode *ip = VTOI(vp);
1442 vn_printf(ip->i_devvp, "\tino %lu", (u_long)ip->i_number);
1443 if (vp->v_type == VFIFO)
1450 * Close wrapper for fifos.
1452 * Update the times on the inode then do device close.
1456 struct vop_close_args /* {
1459 struct ucred *a_cred;
1460 struct thread *a_td;
1463 struct vnode *vp = ap->a_vp;
1466 if (vp->v_usecount > 1)
1469 return (fifo_specops.vop_close(ap));
1473 * Kqfilter wrapper for fifos.
1475 * Fall through to ext2 kqfilter routines if needed
1478 ext2fifo_kqfilter(ap)
1479 struct vop_kqfilter_args *ap;
1483 error = fifo_specops.vop_kqfilter(ap);
1485 error = ext2_kqfilter(ap);
1490 * Return POSIX pathconf information applicable to ext2 filesystems.
1494 struct vop_pathconf_args /* {
1501 switch (ap->a_name) {
1503 *ap->a_retval = LINK_MAX;
1506 *ap->a_retval = NAME_MAX;
1509 *ap->a_retval = PATH_MAX;
1512 *ap->a_retval = PIPE_BUF;
1514 case _PC_CHOWN_RESTRICTED:
1527 * Advisory record locking support
1531 struct vop_advlock_args /* {
1539 struct inode *ip = VTOI(ap->a_vp);
1541 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
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_CLEARSUGID,
1628 ip->i_mode &= ~ISGID;
1631 if (cnp->cn_flags & ISWHITEOUT)
1632 ip->i_flags |= UF_OPAQUE;
1635 * Make sure inode goes to disk before directory entry.
1637 error = ext2_update(tvp, 1);
1640 error = ext2_direnter(ip, dvp, cnp);
1649 * Write error occurred trying to update the inode
1650 * or the directory so must deallocate the inode.
1653 ip->i_flag |= IN_CHANGE;
1658 static struct filterops ext2read_filtops =
1659 { 1, NULL, filt_ext2detach, filt_ext2read };
1660 static struct filterops ext2write_filtops =
1661 { 1, NULL, filt_ext2detach, filt_ext2write };
1662 static struct filterops ext2vnode_filtops =
1663 { 1, NULL, filt_ext2detach, filt_ext2vnode };
1667 struct vop_kqfilter_args /* {
1672 struct vnode *vp = ap->a_vp;
1673 struct knote *kn = ap->a_kn;
1675 switch (kn->kn_filter) {
1677 kn->kn_fop = &ext2read_filtops;
1680 kn->kn_fop = &ext2write_filtops;
1683 kn->kn_fop = &ext2vnode_filtops;
1689 kn->kn_hook = (caddr_t)vp;
1691 if (vp->v_pollinfo == NULL)
1693 if (vp->v_pollinfo == NULL)
1695 knlist_add(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
1701 filt_ext2detach(struct knote *kn)
1703 struct vnode *vp = (struct vnode *)kn->kn_hook;
1705 KASSERT(vp->v_pollinfo != NULL, ("Mising v_pollinfo"));
1706 knlist_remove(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
1711 filt_ext2read(struct knote *kn, long hint)
1713 struct vnode *vp = (struct vnode *)kn->kn_hook;
1714 struct inode *ip = VTOI(vp);
1717 * filesystem is gone, so set the EOF flag and schedule
1718 * the knote for deletion.
1720 if (hint == NOTE_REVOKE) {
1721 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1725 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
1726 return (kn->kn_data != 0);
1731 filt_ext2write(struct knote *kn, long hint)
1735 * filesystem is gone, so set the EOF flag and schedule
1736 * the knote for deletion.
1738 if (hint == NOTE_REVOKE)
1739 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1746 filt_ext2vnode(struct knote *kn, long hint)
1749 if (kn->kn_sfflags & hint)
1750 kn->kn_fflags |= hint;
1751 if (hint == NOTE_REVOKE) {
1752 kn->kn_flags |= EV_EOF;
1755 return (kn->kn_fflags != 0);