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 * SPDX-License-Identifier: BSD-3-Clause
10 * Copyright (c) 1982, 1986, 1989, 1993
11 * The Regents of the University of California. All rights reserved.
12 * (c) UNIX System Laboratories, Inc.
13 * All or some portions of this file are derived from material licensed
14 * to the University of California by American Telephone and Telegraph
15 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
16 * the permission of UNIX System Laboratories, Inc.
18 * Redistribution and use in source and binary forms, with or without
19 * modification, are permitted provided that the following conditions
21 * 1. Redistributions of source code must retain the above copyright
22 * notice, this list of conditions and the following disclaimer.
23 * 2. Redistributions in binary form must reproduce the above copyright
24 * notice, this list of conditions and the following disclaimer in the
25 * documentation and/or other materials provided with the distribution.
26 * 3. Neither the name of the University nor the names of its contributors
27 * may be used to endorse or promote products derived from this software
28 * without specific prior written permission.
30 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
31 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
32 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
33 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
34 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
35 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
36 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
37 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
38 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
39 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
42 * @(#)ufs_vnops.c 8.7 (Berkeley) 2/3/94
43 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
47 #include "opt_suiddir.h"
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/kernel.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
57 #include <sys/endian.h>
59 #include <sys/rwlock.h>
60 #include <sys/mount.h>
61 #include <sys/unistd.h>
63 #include <sys/vnode.h>
64 #include <sys/namei.h>
65 #include <sys/lockf.h>
66 #include <sys/event.h>
69 #include <sys/extattr.h>
70 #include <sys/vmmeter.h>
73 #include <vm/vm_param.h>
74 #include <vm/vm_extern.h>
75 #include <vm/vm_object.h>
76 #include <vm/vm_page.h>
77 #include <vm/vm_pager.h>
78 #include <vm/vnode_pager.h>
80 #include "opt_directio.h"
82 #include <ufs/ufs/dir.h>
84 #include <fs/ext2fs/fs.h>
85 #include <fs/ext2fs/inode.h>
86 #include <fs/ext2fs/ext2_acl.h>
87 #include <fs/ext2fs/ext2fs.h>
88 #include <fs/ext2fs/ext2_extern.h>
89 #include <fs/ext2fs/ext2_dinode.h>
90 #include <fs/ext2fs/ext2_dir.h>
91 #include <fs/ext2fs/ext2_mount.h>
92 #include <fs/ext2fs/ext2_extattr.h>
94 static int ext2_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
95 static void ext2_itimes_locked(struct vnode *);
97 static vop_access_t ext2_access;
98 static int ext2_chmod(struct vnode *, int, struct ucred *, struct thread *);
99 static int ext2_chown(struct vnode *, uid_t, gid_t, struct ucred *,
101 static vop_close_t ext2_close;
102 static vop_create_t ext2_create;
103 static vop_fsync_t ext2_fsync;
104 static vop_getattr_t ext2_getattr;
105 static vop_ioctl_t ext2_ioctl;
106 static vop_link_t ext2_link;
107 static vop_mkdir_t ext2_mkdir;
108 static vop_mknod_t ext2_mknod;
109 static vop_open_t ext2_open;
110 static vop_pathconf_t ext2_pathconf;
111 static vop_print_t ext2_print;
112 static vop_read_t ext2_read;
113 static vop_readlink_t ext2_readlink;
114 static vop_remove_t ext2_remove;
115 static vop_rename_t ext2_rename;
116 static vop_rmdir_t ext2_rmdir;
117 static vop_setattr_t ext2_setattr;
118 static vop_strategy_t ext2_strategy;
119 static vop_symlink_t ext2_symlink;
120 static vop_write_t ext2_write;
121 static vop_deleteextattr_t ext2_deleteextattr;
122 static vop_getextattr_t ext2_getextattr;
123 static vop_listextattr_t ext2_listextattr;
124 static vop_setextattr_t ext2_setextattr;
125 static vop_vptofh_t ext2_vptofh;
126 static vop_close_t ext2fifo_close;
127 static vop_kqfilter_t ext2fifo_kqfilter;
129 /* Global vfs data structures for ext2. */
130 struct vop_vector ext2_vnodeops = {
131 .vop_default = &default_vnodeops,
132 .vop_access = ext2_access,
133 .vop_bmap = ext2_bmap,
134 .vop_cachedlookup = ext2_lookup,
135 .vop_close = ext2_close,
136 .vop_create = ext2_create,
137 .vop_fsync = ext2_fsync,
138 .vop_getpages = vnode_pager_local_getpages,
139 .vop_getpages_async = vnode_pager_local_getpages_async,
140 .vop_getattr = ext2_getattr,
141 .vop_inactive = ext2_inactive,
142 .vop_ioctl = ext2_ioctl,
143 .vop_link = ext2_link,
144 .vop_lookup = vfs_cache_lookup,
145 .vop_mkdir = ext2_mkdir,
146 .vop_mknod = ext2_mknod,
147 .vop_open = ext2_open,
148 .vop_pathconf = ext2_pathconf,
149 .vop_poll = vop_stdpoll,
150 .vop_print = ext2_print,
151 .vop_read = ext2_read,
152 .vop_readdir = ext2_readdir,
153 .vop_readlink = ext2_readlink,
154 .vop_reallocblks = ext2_reallocblks,
155 .vop_reclaim = ext2_reclaim,
156 .vop_remove = ext2_remove,
157 .vop_rename = ext2_rename,
158 .vop_rmdir = ext2_rmdir,
159 .vop_setattr = ext2_setattr,
160 .vop_strategy = ext2_strategy,
161 .vop_symlink = ext2_symlink,
162 .vop_write = ext2_write,
163 .vop_deleteextattr = ext2_deleteextattr,
164 .vop_getextattr = ext2_getextattr,
165 .vop_listextattr = ext2_listextattr,
166 .vop_setextattr = ext2_setextattr,
168 .vop_getacl = ext2_getacl,
169 .vop_setacl = ext2_setacl,
170 .vop_aclcheck = ext2_aclcheck,
172 .vop_vptofh = ext2_vptofh,
175 struct vop_vector ext2_fifoops = {
176 .vop_default = &fifo_specops,
177 .vop_access = ext2_access,
178 .vop_close = ext2fifo_close,
179 .vop_fsync = ext2_fsync,
180 .vop_getattr = ext2_getattr,
181 .vop_inactive = ext2_inactive,
182 .vop_kqfilter = ext2fifo_kqfilter,
183 .vop_pathconf = ext2_pathconf,
184 .vop_print = ext2_print,
185 .vop_read = VOP_PANIC,
186 .vop_reclaim = ext2_reclaim,
187 .vop_setattr = ext2_setattr,
188 .vop_write = VOP_PANIC,
189 .vop_vptofh = ext2_vptofh,
193 * A virgin directory (no blushing please).
194 * Note that the type and namlen fields are reversed relative to ext2.
195 * Also, we don't use `struct odirtemplate', since it would just cause
196 * endianness problems.
198 static struct dirtemplate mastertemplate = {
199 0, 12, 1, EXT2_FT_DIR, ".",
200 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
202 static struct dirtemplate omastertemplate = {
203 0, 12, 1, EXT2_FT_UNKNOWN, ".",
204 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
208 ext2_itimes_locked(struct vnode *vp)
213 ASSERT_VI_LOCKED(vp, __func__);
216 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
218 if ((vp->v_type == VBLK || vp->v_type == VCHR))
219 ip->i_flag |= IN_LAZYMOD;
221 ip->i_flag |= IN_MODIFIED;
222 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
224 if (ip->i_flag & IN_ACCESS) {
225 ip->i_atime = ts.tv_sec;
226 ip->i_atimensec = ts.tv_nsec;
228 if (ip->i_flag & IN_UPDATE) {
229 ip->i_mtime = ts.tv_sec;
230 ip->i_mtimensec = ts.tv_nsec;
233 if (ip->i_flag & IN_CHANGE) {
234 ip->i_ctime = ts.tv_sec;
235 ip->i_ctimensec = ts.tv_nsec;
238 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
242 ext2_itimes(struct vnode *vp)
246 ext2_itimes_locked(vp);
251 * Create a regular file
254 ext2_create(struct vop_create_args *ap)
259 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
260 ap->a_dvp, ap->a_vpp, ap->a_cnp);
263 if ((ap->a_cnp->cn_flags & MAKEENTRY) != 0)
264 cache_enter(ap->a_dvp, *ap->a_vpp, ap->a_cnp);
269 ext2_open(struct vop_open_args *ap)
272 if (ap->a_vp->v_type == VBLK || ap->a_vp->v_type == VCHR)
276 * Files marked append-only must be opened for appending.
278 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
279 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
282 vnode_create_vobject(ap->a_vp, VTOI(ap->a_vp)->i_size, ap->a_td);
290 * Update the times on the inode.
293 ext2_close(struct vop_close_args *ap)
295 struct vnode *vp = ap->a_vp;
298 if (vp->v_usecount > 1)
299 ext2_itimes_locked(vp);
305 ext2_access(struct vop_access_args *ap)
307 struct vnode *vp = ap->a_vp;
308 struct inode *ip = VTOI(vp);
309 accmode_t accmode = ap->a_accmode;
312 if (vp->v_type == VBLK || vp->v_type == VCHR)
316 * Disallow write attempts on read-only file systems;
317 * unless the file is a socket, fifo, or a block or
318 * character device resident on the file system.
320 if (accmode & VWRITE) {
321 switch (vp->v_type) {
325 if (vp->v_mount->mnt_flag & MNT_RDONLY)
333 /* If immutable bit set, nobody gets to write it. */
334 if ((accmode & VWRITE) && (ip->i_flags & (SF_IMMUTABLE | SF_SNAPSHOT)))
337 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
338 ap->a_accmode, ap->a_cred, NULL);
343 ext2_getattr(struct vop_getattr_args *ap)
345 struct vnode *vp = ap->a_vp;
346 struct inode *ip = VTOI(vp);
347 struct vattr *vap = ap->a_vap;
351 * Copy from inode table
353 vap->va_fsid = dev2udev(ip->i_devvp->v_rdev);
354 vap->va_fileid = ip->i_number;
355 vap->va_mode = ip->i_mode & ~IFMT;
356 vap->va_nlink = ip->i_nlink;
357 vap->va_uid = ip->i_uid;
358 vap->va_gid = ip->i_gid;
359 vap->va_rdev = ip->i_rdev;
360 vap->va_size = ip->i_size;
361 vap->va_atime.tv_sec = ip->i_atime;
362 vap->va_atime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_atimensec : 0;
363 vap->va_mtime.tv_sec = ip->i_mtime;
364 vap->va_mtime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_mtimensec : 0;
365 vap->va_ctime.tv_sec = ip->i_ctime;
366 vap->va_ctime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_ctimensec : 0;
367 if E2DI_HAS_XTIME(ip) {
368 vap->va_birthtime.tv_sec = ip->i_birthtime;
369 vap->va_birthtime.tv_nsec = ip->i_birthnsec;
371 vap->va_flags = ip->i_flags;
372 vap->va_gen = ip->i_gen;
373 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
374 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
375 vap->va_type = IFTOVT(ip->i_mode);
376 vap->va_filerev = ip->i_modrev;
381 * Set attribute vnode op. called from several syscalls
384 ext2_setattr(struct vop_setattr_args *ap)
386 struct vattr *vap = ap->a_vap;
387 struct vnode *vp = ap->a_vp;
388 struct inode *ip = VTOI(vp);
389 struct ucred *cred = ap->a_cred;
390 struct thread *td = curthread;
394 * Check for unsettable attributes.
396 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
397 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
398 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
399 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
402 if (vap->va_flags != VNOVAL) {
403 /* Disallow flags not supported by ext2fs. */
404 if (vap->va_flags & ~(SF_APPEND | SF_IMMUTABLE | UF_NODUMP))
407 if (vp->v_mount->mnt_flag & MNT_RDONLY)
410 * Callers may only modify the file flags on objects they
411 * have VADMIN rights for.
413 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
416 * Unprivileged processes and privileged processes in
417 * jail() are not permitted to unset system flags, or
418 * modify flags if any system flags are set.
419 * Privileged non-jail processes may not modify system flags
420 * if securelevel > 0 and any existing system flags are set.
422 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
423 if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) {
424 error = securelevel_gt(cred, 0);
429 if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND) ||
430 ((vap->va_flags ^ ip->i_flags) & SF_SETTABLE))
433 ip->i_flags = vap->va_flags;
434 ip->i_flag |= IN_CHANGE;
435 if (ip->i_flags & (IMMUTABLE | APPEND))
438 if (ip->i_flags & (IMMUTABLE | APPEND))
441 * Go through the fields and update iff not VNOVAL.
443 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
444 if (vp->v_mount->mnt_flag & MNT_RDONLY)
446 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred,
450 if (vap->va_size != VNOVAL) {
452 * Disallow write attempts on read-only file systems;
453 * unless the file is a socket, fifo, or a block or
454 * character device resident on the file system.
456 switch (vp->v_type) {
461 if (vp->v_mount->mnt_flag & MNT_RDONLY)
467 if ((error = ext2_truncate(vp, vap->va_size, 0, cred, td)) != 0)
470 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
471 if (vp->v_mount->mnt_flag & MNT_RDONLY)
475 * If times is NULL, ... The caller must be the owner of
476 * the file, have permission to write the file, or be the
478 * If times is non-NULL, ... The caller must be the owner of
479 * the file or be the super-user.
481 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
482 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
483 (error = VOP_ACCESS(vp, VWRITE, cred, td))))
485 ip->i_flag |= IN_CHANGE | IN_MODIFIED;
486 if (vap->va_atime.tv_sec != VNOVAL) {
487 ip->i_flag &= ~IN_ACCESS;
488 ip->i_atime = vap->va_atime.tv_sec;
489 ip->i_atimensec = vap->va_atime.tv_nsec;
491 if (vap->va_mtime.tv_sec != VNOVAL) {
492 ip->i_flag &= ~IN_UPDATE;
493 ip->i_mtime = vap->va_mtime.tv_sec;
494 ip->i_mtimensec = vap->va_mtime.tv_nsec;
496 ip->i_birthtime = vap->va_birthtime.tv_sec;
497 ip->i_birthnsec = vap->va_birthtime.tv_nsec;
498 error = ext2_update(vp, 0);
503 if (vap->va_mode != (mode_t)VNOVAL) {
504 if (vp->v_mount->mnt_flag & MNT_RDONLY)
506 error = ext2_chmod(vp, (int)vap->va_mode, cred, td);
512 * Change the mode on a file.
513 * Inode must be locked before calling.
516 ext2_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
518 struct inode *ip = VTOI(vp);
522 * To modify the permissions on a file, must possess VADMIN
525 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
528 * Privileged processes may set the sticky bit on non-directories,
529 * as well as set the setgid bit on a file with a group that the
530 * process is not a member of.
532 if (vp->v_type != VDIR && (mode & S_ISTXT)) {
533 error = priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0);
537 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
538 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
542 ip->i_mode &= ~ALLPERMS;
543 ip->i_mode |= (mode & ALLPERMS);
544 ip->i_flag |= IN_CHANGE;
549 * Perform chown operation on inode ip;
550 * inode must be locked prior to call.
553 ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
556 struct inode *ip = VTOI(vp);
561 if (uid == (uid_t)VNOVAL)
563 if (gid == (gid_t)VNOVAL)
566 * To modify the ownership of a file, must possess VADMIN
569 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
572 * To change the owner of a file, or change the group of a file
573 * to a group of which we are not a member, the caller must
576 if (uid != ip->i_uid || (gid != ip->i_gid &&
577 !groupmember(gid, cred))) {
578 error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0);
586 ip->i_flag |= IN_CHANGE;
587 if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
588 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0) != 0)
589 ip->i_mode &= ~(ISUID | ISGID);
595 * Synch an open file.
599 ext2_fsync(struct vop_fsync_args *ap)
602 * Flush all dirty buffers associated with a vnode.
607 return (ext2_update(ap->a_vp, ap->a_waitfor == MNT_WAIT));
615 ext2_mknod(struct vop_mknod_args *ap)
617 struct vattr *vap = ap->a_vap;
618 struct vnode **vpp = ap->a_vpp;
623 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
624 ap->a_dvp, vpp, ap->a_cnp);
628 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
629 if (vap->va_rdev != VNOVAL) {
631 * Want to be able to use this to make badblock
632 * inodes, so don't truncate the dev number.
634 if (!(ip->i_flag & IN_E4EXTENTS))
635 ip->i_rdev = vap->va_rdev;
638 * Remove inode, then reload it through VFS_VGET so it is
639 * checked to see if it is an alias of an existing entry in
640 * the inode cache. XXX I don't believe this is necessary now.
642 (*vpp)->v_type = VNON;
643 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
646 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
655 ext2_remove(struct vop_remove_args *ap)
658 struct vnode *vp = ap->a_vp;
659 struct vnode *dvp = ap->a_dvp;
663 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
664 (VTOI(dvp)->i_flags & APPEND)) {
668 error = ext2_dirremove(dvp, ap->a_cnp);
671 ip->i_flag |= IN_CHANGE;
677 static unsigned short
678 ext2_max_nlink(struct inode *ip)
684 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_DIR_NLINK))
685 return (EXT4_LINK_MAX);
687 return (EXT2_LINK_MAX);
694 ext2_link(struct vop_link_args *ap)
696 struct vnode *vp = ap->a_vp;
697 struct vnode *tdvp = ap->a_tdvp;
698 struct componentname *cnp = ap->a_cnp;
703 if ((cnp->cn_flags & HASBUF) == 0)
704 panic("ext2_link: no name");
707 if ((nlink_t)ip->i_nlink >= ext2_max_nlink(ip)) {
711 if (ip->i_flags & (IMMUTABLE | APPEND)) {
716 ip->i_flag |= IN_CHANGE;
717 error = ext2_update(vp, !DOINGASYNC(vp));
719 error = ext2_direnter(ip, tdvp, cnp);
722 ip->i_flag |= IN_CHANGE;
729 ext2_inc_nlink(struct inode *ip)
734 if (ext2_htree_has_idx(ip) && ip->i_nlink > 1) {
735 if (ip->i_nlink >= ext2_max_nlink(ip) || ip->i_nlink == 2)
737 } else if (ip->i_nlink > ext2_max_nlink(ip)) {
746 ext2_dec_nlink(struct inode *ip)
749 if (!S_ISDIR(ip->i_mode) || ip->i_nlink > 2)
754 * Rename system call.
755 * rename("foo", "bar");
758 * link("foo", "bar");
760 * but ``atomically''. Can't do full commit without saving state in the
761 * inode on disk which isn't feasible at this time. Best we can do is
762 * always guarantee the target exists.
764 * Basic algorithm is:
766 * 1) Bump link count on source while we're linking it to the
767 * target. This also ensure the inode won't be deleted out
768 * from underneath us while we work (it may be truncated by
769 * a concurrent `trunc' or `open' for creation).
770 * 2) Link source to destination. If destination already exists,
772 * 3) Unlink source reference to inode if still around. If a
773 * directory was moved and the parent of the destination
774 * is different from the source, patch the ".." entry in the
778 ext2_rename(struct vop_rename_args *ap)
780 struct vnode *tvp = ap->a_tvp;
781 struct vnode *tdvp = ap->a_tdvp;
782 struct vnode *fvp = ap->a_fvp;
783 struct vnode *fdvp = ap->a_fdvp;
784 struct componentname *tcnp = ap->a_tcnp;
785 struct componentname *fcnp = ap->a_fcnp;
786 struct inode *ip, *xp, *dp;
787 struct dirtemplate dirbuf;
788 int doingdirectory = 0, oldparent = 0, newparent = 0;
793 if ((tcnp->cn_flags & HASBUF) == 0 ||
794 (fcnp->cn_flags & HASBUF) == 0)
795 panic("ext2_rename: no name");
798 * Check for cross-device rename.
800 if ((fvp->v_mount != tdvp->v_mount) ||
801 (tvp && (fvp->v_mount != tvp->v_mount))) {
815 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
816 (VTOI(tdvp)->i_flags & APPEND))) {
822 * Renaming a file to itself has no effect. The upper layers should
823 * not call us in that case. Temporarily just warn if they do.
826 printf("ext2_rename: fvp == tvp (can't happen)\n");
831 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
835 if (ip->i_nlink >= ext2_max_nlink(ip) && !ext2_htree_has_idx(ip)) {
840 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
841 || (dp->i_flags & APPEND)) {
846 if ((ip->i_mode & IFMT) == IFDIR) {
848 * Avoid ".", "..", and aliases of "." for obvious reasons.
850 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
851 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
852 (ip->i_flag & IN_RENAME)) {
857 ip->i_flag |= IN_RENAME;
858 oldparent = dp->i_number;
864 * When the target exists, both the directory
865 * and target vnodes are returned locked.
873 * 1) Bump link count while we're moving stuff
874 * around. If we crash somewhere before
875 * completing our work, the link count
876 * may be wrong, but correctable.
879 ip->i_flag |= IN_CHANGE;
880 if ((error = ext2_update(fvp, !DOINGASYNC(fvp))) != 0) {
886 * If ".." must be changed (ie the directory gets a new
887 * parent) then the source directory must not be in the
888 * directory hierarchy above the target, as this would
889 * orphan everything below the source directory. Also
890 * the user must have write permission in the source so
891 * as to be able to change "..". We must repeat the call
892 * to namei, as the parent directory is unlocked by the
893 * call to checkpath().
895 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
897 if (oldparent != dp->i_number)
898 newparent = dp->i_number;
899 if (doingdirectory && newparent) {
900 if (error) /* write access check above */
904 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
908 error = relookup(tdvp, &tvp, tcnp);
918 * 2) If target doesn't exist, link the target
919 * to the source and unlink the source.
920 * Otherwise, rewrite the target directory
921 * entry to reference the source inode and
922 * expunge the original entry's existence.
925 if (dp->i_devvp != ip->i_devvp)
926 panic("ext2_rename: EXDEV");
928 * Account for ".." in new directory.
929 * When source and destination have the same
930 * parent we don't fool with the link count.
932 if (doingdirectory && newparent) {
933 error = ext2_inc_nlink(dp);
937 dp->i_flag |= IN_CHANGE;
938 error = ext2_update(tdvp, !DOINGASYNC(tdvp));
942 error = ext2_direnter(ip, tdvp, tcnp);
944 if (doingdirectory && newparent) {
946 dp->i_flag |= IN_CHANGE;
947 (void)ext2_update(tdvp, 1);
953 if (xp->i_devvp != dp->i_devvp || xp->i_devvp != ip->i_devvp)
954 panic("ext2_rename: EXDEV");
956 * Short circuit rename(foo, foo).
958 if (xp->i_number == ip->i_number)
959 panic("ext2_rename: same file");
961 * If the parent directory is "sticky", then the user must
962 * own the parent directory, or the destination of the rename,
963 * otherwise the destination may not be changed (except by
964 * root). This implements append-only directories.
966 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
967 tcnp->cn_cred->cr_uid != dp->i_uid &&
968 xp->i_uid != tcnp->cn_cred->cr_uid) {
973 * Target must be empty if a directory and have no links
974 * to it. Also, ensure source and target are compatible
975 * (both directories, or both not directories).
977 if ((xp->i_mode & IFMT) == IFDIR) {
978 if (!ext2_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
982 if (!doingdirectory) {
987 } else if (doingdirectory) {
991 error = ext2_dirrewrite(dp, ip, tcnp);
995 * If the target directory is in the same
996 * directory as the source directory,
997 * decrement the link count on the parent
998 * of the target directory.
1000 if (doingdirectory && !newparent) {
1002 dp->i_flag |= IN_CHANGE;
1006 * Adjust the link count of the target to
1007 * reflect the dirrewrite above. If this is
1008 * a directory it is empty and there are
1009 * no links to it, so we can squash the inode and
1010 * any space associated with it. We disallowed
1011 * renaming over top of a directory with links to
1012 * it above, as the remaining link would point to
1013 * a directory without "." or ".." entries.
1016 if (doingdirectory) {
1017 if (--xp->i_nlink != 0)
1018 panic("ext2_rename: linked directory");
1019 error = ext2_truncate(tvp, (off_t)0, IO_SYNC,
1020 tcnp->cn_cred, tcnp->cn_thread);
1022 xp->i_flag |= IN_CHANGE;
1028 * 3) Unlink the source.
1030 fcnp->cn_flags &= ~MODMASK;
1031 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1033 error = relookup(fdvp, &fvp, fcnp);
1041 * From name has disappeared. IN_RENAME is not sufficient
1042 * to protect against directory races due to timing windows,
1043 * so we can't panic here.
1049 * Ensure that the directory entry still exists and has not
1050 * changed while the new name has been entered. If the source is
1051 * a file then the entry may have been unlinked or renamed. In
1052 * either case there is no further work to be done. If the source
1053 * is a directory then it cannot have been rmdir'ed; its link
1054 * count of three would cause a rmdir to fail with ENOTEMPTY.
1055 * The IN_RENAME flag ensures that it cannot be moved by another
1060 * From name resolves to a different inode. IN_RENAME is
1061 * not sufficient protection against timing window races
1062 * so we can't panic here.
1066 * If the source is a directory with a
1067 * new parent, the link count of the old
1068 * parent directory must be decremented
1069 * and ".." set to point to the new parent.
1071 if (doingdirectory && newparent) {
1073 dp->i_flag |= IN_CHANGE;
1074 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
1075 sizeof(struct dirtemplate), (off_t)0,
1076 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1077 tcnp->cn_cred, NOCRED, NULL, NULL);
1079 /* Like ufs little-endian: */
1080 namlen = dirbuf.dotdot_type;
1082 dirbuf.dotdot_name[0] != '.' ||
1083 dirbuf.dotdot_name[1] != '.') {
1084 ext2_dirbad(xp, (doff_t)12,
1085 "rename: mangled dir");
1087 dirbuf.dotdot_ino = newparent;
1088 (void)vn_rdwr(UIO_WRITE, fvp,
1090 sizeof(struct dirtemplate),
1091 (off_t)0, UIO_SYSSPACE,
1092 IO_NODELOCKED | IO_SYNC |
1093 IO_NOMACCHECK, tcnp->cn_cred,
1094 NOCRED, NULL, NULL);
1099 error = ext2_dirremove(fdvp, fcnp);
1102 xp->i_flag |= IN_CHANGE;
1104 xp->i_flag &= ~IN_RENAME;
1119 ip->i_flag &= ~IN_RENAME;
1120 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
1122 ip->i_flag |= IN_CHANGE;
1123 ip->i_flag &= ~IN_RENAME;
1132 ext2_do_posix1e_acl_inheritance_dir(struct vnode *dvp, struct vnode *tvp,
1133 mode_t dmode, struct ucred *cred, struct thread *td)
1136 struct inode *ip = VTOI(tvp);
1137 struct acl *dacl, *acl;
1139 acl = acl_alloc(M_WAITOK);
1140 dacl = acl_alloc(M_WAITOK);
1143 * Retrieve default ACL from parent, if any.
1145 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
1149 * Retrieved a default ACL, so merge mode and ACL if
1150 * necessary. If the ACL is empty, fall through to
1151 * the "not defined or available" case.
1153 if (acl->acl_cnt != 0) {
1154 dmode = acl_posix1e_newfilemode(dmode, acl);
1157 ext2_sync_acl_from_inode(ip, acl);
1164 * Just use the mode as-is.
1174 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
1176 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, cred, td);
1183 * XXX: This should not happen, as EOPNOTSUPP above
1184 * was supposed to free acl.
1187 printf("ext2_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
1203 ext2_do_posix1e_acl_inheritance_file(struct vnode *dvp, struct vnode *tvp,
1204 mode_t mode, struct ucred *cred, struct thread *td)
1207 struct inode *ip = VTOI(tvp);
1210 acl = acl_alloc(M_WAITOK);
1213 * Retrieve default ACL for parent, if any.
1215 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
1219 * Retrieved a default ACL, so merge mode and ACL if
1222 if (acl->acl_cnt != 0) {
1224 * Two possible ways for default ACL to not
1225 * be present. First, the EA can be
1226 * undefined, or second, the default ACL can
1227 * be blank. If it's blank, fall through to
1228 * the it's not defined case.
1230 mode = acl_posix1e_newfilemode(mode, acl);
1232 ext2_sync_acl_from_inode(ip, acl);
1239 * Just use the mode as-is.
1249 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
1256 * XXX: This should not happen, as EOPNOTSUPP above was
1257 * supposed to free acl.
1259 printf("ufs_do_posix1e_acl_inheritance_file: VOP_GETACL() "
1260 "but no VOP_SETACL()\n");
1261 /* panic("ufs_do_posix1e_acl_inheritance_file: VOP_GETACL() "
1262 "but no VOP_SETACL()"); */
1275 #endif /* UFS_ACL */
1281 ext2_mkdir(struct vop_mkdir_args *ap)
1283 struct vnode *dvp = ap->a_dvp;
1284 struct vattr *vap = ap->a_vap;
1285 struct componentname *cnp = ap->a_cnp;
1286 struct inode *ip, *dp;
1288 struct dirtemplate dirtemplate, *dtp;
1292 if ((cnp->cn_flags & HASBUF) == 0)
1293 panic("ext2_mkdir: no name");
1296 if ((nlink_t)dp->i_nlink >= ext2_max_nlink(dp) &&
1297 !ext2_htree_has_idx(dp)) {
1301 dmode = vap->va_mode & 0777;
1304 * Must simulate part of ext2_makeinode here to acquire the inode,
1305 * but not have it entered in the parent directory. The entry is
1306 * made later after writing "." and ".." entries.
1308 error = ext2_valloc(dvp, dmode, cnp->cn_cred, &tvp);
1312 ip->i_gid = dp->i_gid;
1316 * if we are hacking owners here, (only do this where told to)
1317 * and we are not giving it TOO root, (would subvert quotas)
1318 * then go ahead and give it to the other user.
1319 * The new directory also inherits the SUID bit.
1320 * If user's UID and dir UID are the same,
1321 * 'give it away' so that the SUID is still forced on.
1323 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1324 (dp->i_mode & ISUID) && dp->i_uid) {
1326 ip->i_uid = dp->i_uid;
1328 ip->i_uid = cnp->cn_cred->cr_uid;
1332 ip->i_uid = cnp->cn_cred->cr_uid;
1334 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1336 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1338 if (cnp->cn_flags & ISWHITEOUT)
1339 ip->i_flags |= UF_OPAQUE;
1340 error = ext2_update(tvp, 1);
1343 * Bump link count in parent directory
1344 * to reflect work done below. Should
1345 * be done before reference is created
1346 * so reparation is possible if we crash.
1349 dp->i_flag |= IN_CHANGE;
1350 error = ext2_update(dvp, !DOINGASYNC(dvp));
1354 /* Initialize directory with "." and ".." from static template. */
1355 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs,
1356 EXT2F_INCOMPAT_FTYPE))
1357 dtp = &mastertemplate;
1359 dtp = &omastertemplate;
1361 dirtemplate.dot_ino = ip->i_number;
1362 dirtemplate.dotdot_ino = dp->i_number;
1364 * note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE so let's
1365 * just redefine it - for this function only
1368 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->e2fs_bsize
1369 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
1370 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
1371 sizeof(dirtemplate), (off_t)0, UIO_SYSSPACE,
1372 IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
1376 dp->i_flag |= IN_CHANGE;
1379 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
1380 /* XXX should grow with balloc() */
1381 panic("ext2_mkdir: blksize");
1383 ip->i_size = DIRBLKSIZ;
1384 ip->i_flag |= IN_CHANGE;
1388 if (dvp->v_mount->mnt_flag & MNT_ACLS) {
1389 error = ext2_do_posix1e_acl_inheritance_dir(dvp, tvp, dmode,
1390 cnp->cn_cred, cnp->cn_thread);
1395 #endif /* UFS_ACL */
1397 /* Directory set up, now install its entry in the parent directory. */
1398 error = ext2_direnter(ip, dvp, cnp);
1401 dp->i_flag |= IN_CHANGE;
1405 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
1406 * for us because we set the link count to 0.
1410 ip->i_flag |= IN_CHANGE;
1417 #define DIRBLKSIZ DEV_BSIZE
1421 * Rmdir system call.
1424 ext2_rmdir(struct vop_rmdir_args *ap)
1426 struct vnode *vp = ap->a_vp;
1427 struct vnode *dvp = ap->a_dvp;
1428 struct componentname *cnp = ap->a_cnp;
1429 struct inode *ip, *dp;
1436 * Verify the directory is empty (and valid).
1437 * (Rmdir ".." won't be valid since
1438 * ".." will contain a reference to
1439 * the current directory and thus be
1442 if (!ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1446 if ((dp->i_flags & APPEND)
1447 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1452 * Delete reference to directory before purging
1453 * inode. If we crash in between, the directory
1454 * will be reattached to lost+found,
1456 error = ext2_dirremove(dvp, cnp);
1460 dp->i_flag |= IN_CHANGE;
1464 * Truncate inode. The only stuff left
1465 * in the directory is "." and "..".
1468 error = ext2_truncate(vp, (off_t)0, IO_SYNC, cnp->cn_cred,
1470 cache_purge(ITOV(ip));
1471 if (vn_lock(dvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
1473 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
1474 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1481 * symlink -- make a symbolic link
1484 ext2_symlink(struct vop_symlink_args *ap)
1486 struct vnode *vp, **vpp = ap->a_vpp;
1490 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1495 len = strlen(ap->a_target);
1496 if (len < vp->v_mount->mnt_maxsymlinklen) {
1498 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1500 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1502 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1503 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1504 ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
1511 * Return target name of a symbolic link
1514 ext2_readlink(struct vop_readlink_args *ap)
1516 struct vnode *vp = ap->a_vp;
1517 struct inode *ip = VTOI(vp);
1521 if (isize < vp->v_mount->mnt_maxsymlinklen) {
1522 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1525 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1529 * Calculate the logical to physical mapping if not done already,
1530 * then call the device strategy routine.
1532 * In order to be able to swap to a file, the ext2_bmaparray() operation may not
1533 * deadlock on memory. See ext2_bmap() for details.
1536 ext2_strategy(struct vop_strategy_args *ap)
1538 struct buf *bp = ap->a_bp;
1539 struct vnode *vp = ap->a_vp;
1544 if (vp->v_type == VBLK || vp->v_type == VCHR)
1545 panic("ext2_strategy: spec");
1546 if (bp->b_blkno == bp->b_lblkno) {
1548 if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS)
1549 error = ext4_bmapext(vp, bp->b_lblkno, &blkno, NULL, NULL);
1551 error = ext2_bmaparray(vp, bp->b_lblkno, &blkno, NULL, NULL);
1553 bp->b_blkno = blkno;
1555 bp->b_error = error;
1556 bp->b_ioflags |= BIO_ERROR;
1560 if ((long)bp->b_blkno == -1)
1563 if ((long)bp->b_blkno == -1) {
1567 bp->b_iooffset = dbtob(bp->b_blkno);
1568 bo = VFSTOEXT2(vp->v_mount)->um_bo;
1569 BO_STRATEGY(bo, bp);
1574 * Print out the contents of an inode.
1577 ext2_print(struct vop_print_args *ap)
1579 struct vnode *vp = ap->a_vp;
1580 struct inode *ip = VTOI(vp);
1582 vn_printf(ip->i_devvp, "\tino %ju", (uintmax_t)ip->i_number);
1583 if (vp->v_type == VFIFO)
1590 * Close wrapper for fifos.
1592 * Update the times on the inode then do device close.
1595 ext2fifo_close(struct vop_close_args *ap)
1597 struct vnode *vp = ap->a_vp;
1600 if (vp->v_usecount > 1)
1601 ext2_itimes_locked(vp);
1603 return (fifo_specops.vop_close(ap));
1607 * Kqfilter wrapper for fifos.
1609 * Fall through to ext2 kqfilter routines if needed
1612 ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1616 error = fifo_specops.vop_kqfilter(ap);
1618 error = vfs_kqfilter(ap);
1623 * Return POSIX pathconf information applicable to ext2 filesystems.
1626 ext2_pathconf(struct vop_pathconf_args *ap)
1630 switch (ap->a_name) {
1632 if (ext2_htree_has_idx(VTOI(ap->a_vp)))
1633 *ap->a_retval = INT_MAX;
1635 *ap->a_retval = ext2_max_nlink(VTOI(ap->a_vp));
1638 *ap->a_retval = NAME_MAX;
1641 if (ap->a_vp->v_type == VDIR || ap->a_vp->v_type == VFIFO)
1642 *ap->a_retval = PIPE_BUF;
1646 case _PC_CHOWN_RESTRICTED:
1654 case _PC_ACL_EXTENDED:
1655 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
1660 case _PC_ACL_PATH_MAX:
1661 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
1662 *ap->a_retval = ACL_MAX_ENTRIES;
1666 #endif /* UFS_ACL */
1668 case _PC_MIN_HOLE_SIZE:
1669 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1677 case _PC_ALLOC_SIZE_MIN:
1678 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
1680 case _PC_FILESIZEBITS:
1683 case _PC_REC_INCR_XFER_SIZE:
1684 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1686 case _PC_REC_MAX_XFER_SIZE:
1687 *ap->a_retval = -1; /* means ``unlimited'' */
1689 case _PC_REC_MIN_XFER_SIZE:
1690 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1692 case _PC_REC_XFER_ALIGN:
1693 *ap->a_retval = PAGE_SIZE;
1695 case _PC_SYMLINK_MAX:
1696 *ap->a_retval = MAXPATHLEN;
1700 error = vop_stdpathconf(ap);
1707 * Vnode operation to remove a named attribute.
1710 ext2_deleteextattr(struct vop_deleteextattr_args *ap)
1713 struct m_ext2fs *fs;
1716 ip = VTOI(ap->a_vp);
1719 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1720 return (EOPNOTSUPP);
1722 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1723 return (EOPNOTSUPP);
1725 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1726 ap->a_cred, ap->a_td, VWRITE);
1732 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1733 error = ext2_extattr_inode_delete(ip, ap->a_attrnamespace, ap->a_name);
1734 if (error != ENOATTR)
1739 error = ext2_extattr_block_delete(ip, ap->a_attrnamespace, ap->a_name);
1745 * Vnode operation to retrieve a named extended attribute.
1748 ext2_getextattr(struct vop_getextattr_args *ap)
1751 struct m_ext2fs *fs;
1754 ip = VTOI(ap->a_vp);
1757 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1758 return (EOPNOTSUPP);
1760 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1761 return (EOPNOTSUPP);
1763 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1764 ap->a_cred, ap->a_td, VREAD);
1768 if (ap->a_size != NULL)
1773 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1774 error = ext2_extattr_inode_get(ip, ap->a_attrnamespace,
1775 ap->a_name, ap->a_uio, ap->a_size);
1776 if (error != ENOATTR)
1781 error = ext2_extattr_block_get(ip, ap->a_attrnamespace,
1782 ap->a_name, ap->a_uio, ap->a_size);
1788 * Vnode operation to retrieve extended attributes on a vnode.
1791 ext2_listextattr(struct vop_listextattr_args *ap)
1794 struct m_ext2fs *fs;
1797 ip = VTOI(ap->a_vp);
1800 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1801 return (EOPNOTSUPP);
1803 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1804 return (EOPNOTSUPP);
1806 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1807 ap->a_cred, ap->a_td, VREAD);
1811 if (ap->a_size != NULL)
1814 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1815 error = ext2_extattr_inode_list(ip, ap->a_attrnamespace,
1816 ap->a_uio, ap->a_size);
1822 error = ext2_extattr_block_list(ip, ap->a_attrnamespace,
1823 ap->a_uio, ap->a_size);
1829 * Vnode operation to set a named attribute.
1832 ext2_setextattr(struct vop_setextattr_args *ap)
1835 struct m_ext2fs *fs;
1838 ip = VTOI(ap->a_vp);
1841 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1842 return (EOPNOTSUPP);
1844 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1845 return (EOPNOTSUPP);
1847 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1848 ap->a_cred, ap->a_td, VWRITE);
1852 error = ext2_extattr_valid_attrname(ap->a_attrnamespace, ap->a_name);
1856 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1857 error = ext2_extattr_inode_set(ip, ap->a_attrnamespace,
1858 ap->a_name, ap->a_uio);
1859 if (error != ENOSPC)
1863 error = ext2_extattr_block_set(ip, ap->a_attrnamespace,
1864 ap->a_name, ap->a_uio);
1870 * Vnode pointer to File handle
1874 ext2_vptofh(struct vop_vptofh_args *ap)
1879 ip = VTOI(ap->a_vp);
1880 ufhp = (struct ufid *)ap->a_fhp;
1881 ufhp->ufid_len = sizeof(struct ufid);
1882 ufhp->ufid_ino = ip->i_number;
1883 ufhp->ufid_gen = ip->i_gen;
1888 * Initialize the vnode associated with a new inode, handle aliased
1892 ext2_vinit(struct mount *mntp, struct vop_vector *fifoops, struct vnode **vpp)
1899 vp->v_type = IFTOVT(ip->i_mode);
1900 if (vp->v_type == VFIFO)
1903 if (ip->i_number == EXT2_ROOTINO)
1904 vp->v_vflag |= VV_ROOT;
1905 ip->i_modrev = init_va_filerev();
1911 * Allocate a new inode.
1914 ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1915 struct componentname *cnp)
1917 struct inode *ip, *pdir;
1923 if ((cnp->cn_flags & HASBUF) == 0)
1924 panic("ext2_makeinode: no name");
1927 if ((mode & IFMT) == 0)
1930 error = ext2_valloc(dvp, mode, cnp->cn_cred, &tvp);
1935 ip->i_gid = pdir->i_gid;
1940 * not the owner of the directory,
1941 * and we are hacking owners here, (only do this where told to)
1942 * and we are not giving it TOO root, (would subvert quotas)
1943 * then go ahead and give it to the other user.
1944 * Note that this drops off the execute bits for security.
1946 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1947 (pdir->i_mode & ISUID) &&
1948 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1949 ip->i_uid = pdir->i_uid;
1952 ip->i_uid = cnp->cn_cred->cr_uid;
1956 ip->i_uid = cnp->cn_cred->cr_uid;
1958 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1960 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1962 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred)) {
1963 if (priv_check_cred(cnp->cn_cred, PRIV_VFS_RETAINSUGID, 0))
1964 ip->i_mode &= ~ISGID;
1967 if (cnp->cn_flags & ISWHITEOUT)
1968 ip->i_flags |= UF_OPAQUE;
1971 * Make sure inode goes to disk before directory entry.
1973 error = ext2_update(tvp, !DOINGASYNC(tvp));
1978 if (dvp->v_mount->mnt_flag & MNT_ACLS) {
1979 error = ext2_do_posix1e_acl_inheritance_file(dvp, tvp, mode,
1980 cnp->cn_cred, cnp->cn_thread);
1984 #endif /* UFS_ACL */
1986 error = ext2_direnter(ip, dvp, cnp);
1995 * Write error occurred trying to update the inode
1996 * or the directory so must deallocate the inode.
1999 ip->i_flag |= IN_CHANGE;
2005 * Vnode op for reading.
2008 ext2_read(struct vop_read_args *ap)
2013 struct m_ext2fs *fs;
2015 daddr_t lbn, nextlbn;
2017 long size, xfersize, blkoffset;
2018 int error, orig_resid, seqcount;
2023 ioflag = ap->a_ioflag;
2025 seqcount = ap->a_ioflag >> IO_SEQSHIFT;
2029 if (uio->uio_rw != UIO_READ)
2030 panic("%s: mode", "ext2_read");
2032 if (vp->v_type == VLNK) {
2033 if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
2034 panic("%s: short symlink", "ext2_read");
2035 } else if (vp->v_type != VREG && vp->v_type != VDIR)
2036 panic("%s: type %d", "ext2_read", vp->v_type);
2038 orig_resid = uio->uio_resid;
2039 KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
2040 if (orig_resid == 0)
2042 KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
2044 if (uio->uio_offset < ip->i_size &&
2045 uio->uio_offset >= fs->e2fs_maxfilesize)
2048 for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
2049 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
2051 lbn = lblkno(fs, uio->uio_offset);
2053 size = blksize(fs, ip, lbn);
2054 blkoffset = blkoff(fs, uio->uio_offset);
2056 xfersize = fs->e2fs_fsize - blkoffset;
2057 if (uio->uio_resid < xfersize)
2058 xfersize = uio->uio_resid;
2059 if (bytesinfile < xfersize)
2060 xfersize = bytesinfile;
2062 if (lblktosize(fs, nextlbn) >= ip->i_size)
2063 error = bread(vp, lbn, size, NOCRED, &bp);
2064 else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
2065 error = cluster_read(vp, ip->i_size, lbn, size,
2066 NOCRED, blkoffset + uio->uio_resid, seqcount,
2068 } else if (seqcount > 1) {
2069 u_int nextsize = blksize(fs, ip, nextlbn);
2071 error = breadn(vp, lbn,
2072 size, &nextlbn, &nextsize, 1, NOCRED, &bp);
2074 error = bread(vp, lbn, size, NOCRED, &bp);
2082 * We should only get non-zero b_resid when an I/O error
2083 * has occurred, which should cause us to break above.
2084 * However, if the short read did not cause an error,
2085 * then we want to ensure that we do not uiomove bad
2086 * or uninitialized data.
2088 size -= bp->b_resid;
2089 if (size < xfersize) {
2094 error = uiomove((char *)bp->b_data + blkoffset,
2095 (int)xfersize, uio);
2098 vfs_bio_brelse(bp, ioflag);
2102 * This can only happen in the case of an error because the loop
2103 * above resets bp to NULL on each iteration and on normal
2104 * completion has not set a new value into it. so it must have come
2105 * from a 'break' statement
2108 vfs_bio_brelse(bp, ioflag);
2110 if ((error == 0 || uio->uio_resid != orig_resid) &&
2111 (vp->v_mount->mnt_flag & (MNT_NOATIME | MNT_RDONLY)) == 0)
2112 ip->i_flag |= IN_ACCESS;
2117 ext2_ioctl(struct vop_ioctl_args *ap)
2120 switch (ap->a_command) {
2123 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
2124 (off_t *)ap->a_data, ap->a_cred));
2131 * Vnode op for writing.
2134 ext2_write(struct vop_write_args *ap)
2139 struct m_ext2fs *fs;
2143 int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
2145 ioflag = ap->a_ioflag;
2149 seqcount = ioflag >> IO_SEQSHIFT;
2153 if (uio->uio_rw != UIO_WRITE)
2154 panic("%s: mode", "ext2_write");
2157 switch (vp->v_type) {
2159 if (ioflag & IO_APPEND)
2160 uio->uio_offset = ip->i_size;
2161 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
2167 /* XXX differs from ffs -- this is called from ext2_mkdir(). */
2168 if ((ioflag & IO_SYNC) == 0)
2169 panic("ext2_write: nonsync dir write");
2172 panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
2173 vp->v_type, (intmax_t)uio->uio_offset,
2174 (intmax_t)uio->uio_resid);
2177 KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
2178 KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
2180 if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
2183 * Maybe this should be above the vnode op call, but so long as
2184 * file servers have no limits, I don't think it matters.
2186 if (vn_rlimit_fsize(vp, uio, uio->uio_td))
2189 resid = uio->uio_resid;
2191 if (seqcount > BA_SEQMAX)
2192 flags = BA_SEQMAX << BA_SEQSHIFT;
2194 flags = seqcount << BA_SEQSHIFT;
2195 if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
2198 for (error = 0; uio->uio_resid > 0;) {
2199 lbn = lblkno(fs, uio->uio_offset);
2200 blkoffset = blkoff(fs, uio->uio_offset);
2201 xfersize = fs->e2fs_fsize - blkoffset;
2202 if (uio->uio_resid < xfersize)
2203 xfersize = uio->uio_resid;
2204 if (uio->uio_offset + xfersize > ip->i_size)
2205 vnode_pager_setsize(vp, uio->uio_offset + xfersize);
2208 * We must perform a read-before-write if the transfer size
2209 * does not cover the entire buffer.
2211 if (fs->e2fs_bsize > xfersize)
2214 flags &= ~BA_CLRBUF;
2215 error = ext2_balloc(ip, lbn, blkoffset + xfersize,
2216 ap->a_cred, &bp, flags);
2220 if ((ioflag & (IO_SYNC | IO_INVAL)) == (IO_SYNC | IO_INVAL))
2221 bp->b_flags |= B_NOCACHE;
2222 if (uio->uio_offset + xfersize > ip->i_size)
2223 ip->i_size = uio->uio_offset + xfersize;
2224 size = blksize(fs, ip, lbn) - bp->b_resid;
2225 if (size < xfersize)
2229 uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
2231 * If the buffer is not already filled and we encounter an
2232 * error while trying to fill it, we have to clear out any
2233 * garbage data from the pages instantiated for the buffer.
2234 * If we do not, a failed uiomove() during a write can leave
2235 * the prior contents of the pages exposed to a userland mmap.
2237 * Note that we need only clear buffers with a transfer size
2238 * equal to the block size because buffers with a shorter
2239 * transfer size were cleared above by the call to ext2_balloc()
2240 * with the BA_CLRBUF flag set.
2242 * If the source region for uiomove identically mmaps the
2243 * buffer, uiomove() performed the NOP copy, and the buffer
2244 * content remains valid because the page fault handler
2245 * validated the pages.
2247 if (error != 0 && (bp->b_flags & B_CACHE) == 0 &&
2248 fs->e2fs_bsize == xfersize)
2251 vfs_bio_set_flags(bp, ioflag);
2254 * If IO_SYNC each buffer is written synchronously. Otherwise
2255 * if we have a severe page deficiency write the buffer
2256 * asynchronously. Otherwise try to cluster, and if that
2257 * doesn't do it then either do an async write (if O_DIRECT),
2258 * or a delayed write (if not).
2260 if (ioflag & IO_SYNC) {
2262 } else if (vm_page_count_severe() ||
2263 buf_dirty_count_severe() ||
2264 (ioflag & IO_ASYNC)) {
2265 bp->b_flags |= B_CLUSTEROK;
2267 } else if (xfersize + blkoffset == fs->e2fs_fsize) {
2268 if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
2269 bp->b_flags |= B_CLUSTEROK;
2270 cluster_write(vp, bp, ip->i_size, seqcount, 0);
2274 } else if (ioflag & IO_DIRECT) {
2275 bp->b_flags |= B_CLUSTEROK;
2278 bp->b_flags |= B_CLUSTEROK;
2281 if (error || xfersize == 0)
2285 * If we successfully wrote any data, and we are not the superuser
2286 * we clear the setuid and setgid bits as a precaution against
2289 if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid &&
2291 if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
2292 ip->i_mode &= ~(ISUID | ISGID);
2295 if (ioflag & IO_UNIT) {
2296 (void)ext2_truncate(vp, osize,
2297 ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
2298 uio->uio_offset -= resid - uio->uio_resid;
2299 uio->uio_resid = resid;
2302 if (uio->uio_resid != resid) {
2303 ip->i_flag |= IN_CHANGE | IN_UPDATE;
2304 if (ioflag & IO_SYNC)
2305 error = ext2_update(vp, 1);