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 * 3. 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/kernel.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
55 #include <sys/endian.h>
57 #include <sys/rwlock.h>
58 #include <sys/mount.h>
59 #include <sys/unistd.h>
61 #include <sys/vnode.h>
62 #include <sys/namei.h>
63 #include <sys/lockf.h>
64 #include <sys/event.h>
67 #include <sys/extattr.h>
68 #include <sys/vmmeter.h>
71 #include <vm/vm_param.h>
72 #include <vm/vm_extern.h>
73 #include <vm/vm_object.h>
74 #include <vm/vm_page.h>
75 #include <vm/vm_pager.h>
76 #include <vm/vnode_pager.h>
78 #include "opt_directio.h"
80 #include <ufs/ufs/dir.h>
82 #include <fs/ext2fs/fs.h>
83 #include <fs/ext2fs/inode.h>
84 #include <fs/ext2fs/ext2_acl.h>
85 #include <fs/ext2fs/ext2_extern.h>
86 #include <fs/ext2fs/ext2fs.h>
87 #include <fs/ext2fs/ext2_dinode.h>
88 #include <fs/ext2fs/ext2_dir.h>
89 #include <fs/ext2fs/ext2_mount.h>
90 #include <fs/ext2fs/ext2_extattr.h>
92 static int ext2_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
93 static void ext2_itimes_locked(struct vnode *);
94 static int ext4_ext_read(struct vop_read_args *);
95 static int ext2_ind_read(struct vop_read_args *);
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_print = ext2_print,
184 .vop_read = VOP_PANIC,
185 .vop_reclaim = ext2_reclaim,
186 .vop_setattr = ext2_setattr,
187 .vop_write = VOP_PANIC,
188 .vop_vptofh = ext2_vptofh,
192 * A virgin directory (no blushing please).
193 * Note that the type and namlen fields are reversed relative to ext2.
194 * Also, we don't use `struct odirtemplate', since it would just cause
195 * endianness problems.
197 static struct dirtemplate mastertemplate = {
198 0, 12, 1, EXT2_FT_DIR, ".",
199 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
201 static struct dirtemplate omastertemplate = {
202 0, 12, 1, EXT2_FT_UNKNOWN, ".",
203 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
207 ext2_itimes_locked(struct vnode *vp)
212 ASSERT_VI_LOCKED(vp, __func__);
215 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
217 if ((vp->v_type == VBLK || vp->v_type == VCHR))
218 ip->i_flag |= IN_LAZYMOD;
220 ip->i_flag |= IN_MODIFIED;
221 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
223 if (ip->i_flag & IN_ACCESS) {
224 ip->i_atime = ts.tv_sec;
225 ip->i_atimensec = ts.tv_nsec;
227 if (ip->i_flag & IN_UPDATE) {
228 ip->i_mtime = ts.tv_sec;
229 ip->i_mtimensec = ts.tv_nsec;
232 if (ip->i_flag & IN_CHANGE) {
233 ip->i_ctime = ts.tv_sec;
234 ip->i_ctimensec = ts.tv_nsec;
237 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
241 ext2_itimes(struct vnode *vp)
245 ext2_itimes_locked(vp);
250 * Create a regular file
253 ext2_create(struct vop_create_args *ap)
258 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
259 ap->a_dvp, ap->a_vpp, ap->a_cnp);
262 if ((ap->a_cnp->cn_flags & MAKEENTRY) != 0)
263 cache_enter(ap->a_dvp, *ap->a_vpp, ap->a_cnp);
268 ext2_open(struct vop_open_args *ap)
271 if (ap->a_vp->v_type == VBLK || ap->a_vp->v_type == VCHR)
275 * Files marked append-only must be opened for appending.
277 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
278 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
281 vnode_create_vobject(ap->a_vp, VTOI(ap->a_vp)->i_size, ap->a_td);
289 * Update the times on the inode.
292 ext2_close(struct vop_close_args *ap)
294 struct vnode *vp = ap->a_vp;
297 if (vp->v_usecount > 1)
298 ext2_itimes_locked(vp);
304 ext2_access(struct vop_access_args *ap)
306 struct vnode *vp = ap->a_vp;
307 struct inode *ip = VTOI(vp);
308 accmode_t accmode = ap->a_accmode;
311 if (vp->v_type == VBLK || vp->v_type == VCHR)
315 * Disallow write attempts on read-only file systems;
316 * unless the file is a socket, fifo, or a block or
317 * character device resident on the file system.
319 if (accmode & VWRITE) {
320 switch (vp->v_type) {
324 if (vp->v_mount->mnt_flag & MNT_RDONLY)
332 /* If immutable bit set, nobody gets to write it. */
333 if ((accmode & VWRITE) && (ip->i_flags & (SF_IMMUTABLE | SF_SNAPSHOT)))
336 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
337 ap->a_accmode, ap->a_cred, NULL);
342 ext2_getattr(struct vop_getattr_args *ap)
344 struct vnode *vp = ap->a_vp;
345 struct inode *ip = VTOI(vp);
346 struct vattr *vap = ap->a_vap;
350 * Copy from inode table
352 vap->va_fsid = dev2udev(ip->i_devvp->v_rdev);
353 vap->va_fileid = ip->i_number;
354 vap->va_mode = ip->i_mode & ~IFMT;
355 vap->va_nlink = ip->i_nlink;
356 vap->va_uid = ip->i_uid;
357 vap->va_gid = ip->i_gid;
358 vap->va_rdev = ip->i_rdev;
359 vap->va_size = ip->i_size;
360 vap->va_atime.tv_sec = ip->i_atime;
361 vap->va_atime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_atimensec : 0;
362 vap->va_mtime.tv_sec = ip->i_mtime;
363 vap->va_mtime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_mtimensec : 0;
364 vap->va_ctime.tv_sec = ip->i_ctime;
365 vap->va_ctime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_ctimensec : 0;
366 if E2DI_HAS_XTIME(ip) {
367 vap->va_birthtime.tv_sec = ip->i_birthtime;
368 vap->va_birthtime.tv_nsec = ip->i_birthnsec;
370 vap->va_flags = ip->i_flags;
371 vap->va_gen = ip->i_gen;
372 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
373 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
374 vap->va_type = IFTOVT(ip->i_mode);
375 vap->va_filerev = ip->i_modrev;
380 * Set attribute vnode op. called from several syscalls
383 ext2_setattr(struct vop_setattr_args *ap)
385 struct vattr *vap = ap->a_vap;
386 struct vnode *vp = ap->a_vp;
387 struct inode *ip = VTOI(vp);
388 struct ucred *cred = ap->a_cred;
389 struct thread *td = curthread;
393 * Check for unsettable attributes.
395 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
396 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
397 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
398 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
401 if (vap->va_flags != VNOVAL) {
402 /* Disallow flags not supported by ext2fs. */
403 if (vap->va_flags & ~(SF_APPEND | SF_IMMUTABLE | UF_NODUMP))
406 if (vp->v_mount->mnt_flag & MNT_RDONLY)
409 * Callers may only modify the file flags on objects they
410 * have VADMIN rights for.
412 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
415 * Unprivileged processes and privileged processes in
416 * jail() are not permitted to unset system flags, or
417 * modify flags if any system flags are set.
418 * Privileged non-jail processes may not modify system flags
419 * if securelevel > 0 and any existing system flags are set.
421 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
422 if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) {
423 error = securelevel_gt(cred, 0);
428 if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND) ||
429 ((vap->va_flags ^ ip->i_flags) & SF_SETTABLE))
432 ip->i_flags = vap->va_flags;
433 ip->i_flag |= IN_CHANGE;
434 if (ip->i_flags & (IMMUTABLE | APPEND))
437 if (ip->i_flags & (IMMUTABLE | APPEND))
440 * Go through the fields and update iff not VNOVAL.
442 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
443 if (vp->v_mount->mnt_flag & MNT_RDONLY)
445 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred,
449 if (vap->va_size != VNOVAL) {
451 * Disallow write attempts on read-only file systems;
452 * unless the file is a socket, fifo, or a block or
453 * character device resident on the file system.
455 switch (vp->v_type) {
460 if (vp->v_mount->mnt_flag & MNT_RDONLY)
466 if ((error = ext2_truncate(vp, vap->va_size, 0, cred, td)) != 0)
469 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
470 if (vp->v_mount->mnt_flag & MNT_RDONLY)
474 * If times is NULL, ... The caller must be the owner of
475 * the file, have permission to write the file, or be the
477 * If times is non-NULL, ... The caller must be the owner of
478 * the file or be the super-user.
480 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
481 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
482 (error = VOP_ACCESS(vp, VWRITE, cred, td))))
484 ip->i_flag |= IN_CHANGE | IN_MODIFIED;
485 if (vap->va_atime.tv_sec != VNOVAL) {
486 ip->i_flag &= ~IN_ACCESS;
487 ip->i_atime = vap->va_atime.tv_sec;
488 ip->i_atimensec = vap->va_atime.tv_nsec;
490 if (vap->va_mtime.tv_sec != VNOVAL) {
491 ip->i_flag &= ~IN_UPDATE;
492 ip->i_mtime = vap->va_mtime.tv_sec;
493 ip->i_mtimensec = vap->va_mtime.tv_nsec;
495 ip->i_birthtime = vap->va_birthtime.tv_sec;
496 ip->i_birthnsec = vap->va_birthtime.tv_nsec;
497 error = ext2_update(vp, 0);
502 if (vap->va_mode != (mode_t)VNOVAL) {
503 if (vp->v_mount->mnt_flag & MNT_RDONLY)
505 error = ext2_chmod(vp, (int)vap->va_mode, cred, td);
511 * Change the mode on a file.
512 * Inode must be locked before calling.
515 ext2_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
517 struct inode *ip = VTOI(vp);
521 * To modify the permissions on a file, must possess VADMIN
524 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
527 * Privileged processes may set the sticky bit on non-directories,
528 * as well as set the setgid bit on a file with a group that the
529 * process is not a member of.
531 if (vp->v_type != VDIR && (mode & S_ISTXT)) {
532 error = priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0);
536 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
537 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
541 ip->i_mode &= ~ALLPERMS;
542 ip->i_mode |= (mode & ALLPERMS);
543 ip->i_flag |= IN_CHANGE;
548 * Perform chown operation on inode ip;
549 * inode must be locked prior to call.
552 ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
555 struct inode *ip = VTOI(vp);
560 if (uid == (uid_t)VNOVAL)
562 if (gid == (gid_t)VNOVAL)
565 * To modify the ownership of a file, must possess VADMIN
568 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
571 * To change the owner of a file, or change the group of a file
572 * to a group of which we are not a member, the caller must
575 if (uid != ip->i_uid || (gid != ip->i_gid &&
576 !groupmember(gid, cred))) {
577 error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0);
585 ip->i_flag |= IN_CHANGE;
586 if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
587 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0) != 0)
588 ip->i_mode &= ~(ISUID | ISGID);
594 * Synch an open file.
598 ext2_fsync(struct vop_fsync_args *ap)
601 * Flush all dirty buffers associated with a vnode.
606 return (ext2_update(ap->a_vp, ap->a_waitfor == MNT_WAIT));
614 ext2_mknod(struct vop_mknod_args *ap)
616 struct vattr *vap = ap->a_vap;
617 struct vnode **vpp = ap->a_vpp;
622 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
623 ap->a_dvp, vpp, ap->a_cnp);
627 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
628 if (vap->va_rdev != VNOVAL) {
630 * Want to be able to use this to make badblock
631 * inodes, so don't truncate the dev number.
633 ip->i_rdev = vap->va_rdev;
636 * Remove inode, then reload it through VFS_VGET so it is
637 * checked to see if it is an alias of an existing entry in
638 * the inode cache. XXX I don't believe this is necessary now.
640 (*vpp)->v_type = VNON;
641 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
644 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
653 ext2_remove(struct vop_remove_args *ap)
656 struct vnode *vp = ap->a_vp;
657 struct vnode *dvp = ap->a_dvp;
661 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
662 (VTOI(dvp)->i_flags & APPEND)) {
666 error = ext2_dirremove(dvp, ap->a_cnp);
669 ip->i_flag |= IN_CHANGE;
675 static unsigned short
676 ext2_max_nlink(struct inode *ip)
682 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_DIR_NLINK))
683 return (EXT4_LINK_MAX);
685 return (EXT2_LINK_MAX);
692 ext2_link(struct vop_link_args *ap)
694 struct vnode *vp = ap->a_vp;
695 struct vnode *tdvp = ap->a_tdvp;
696 struct componentname *cnp = ap->a_cnp;
701 if ((cnp->cn_flags & HASBUF) == 0)
702 panic("ext2_link: no name");
705 if ((nlink_t)ip->i_nlink >= ext2_max_nlink(ip)) {
709 if (ip->i_flags & (IMMUTABLE | APPEND)) {
714 ip->i_flag |= IN_CHANGE;
715 error = ext2_update(vp, !DOINGASYNC(vp));
717 error = ext2_direnter(ip, tdvp, cnp);
720 ip->i_flag |= IN_CHANGE;
727 ext2_inc_nlink(struct inode *ip)
732 if (ext2_htree_has_idx(ip) && ip->i_nlink > 1) {
733 if (ip->i_nlink >= ext2_max_nlink(ip) || ip->i_nlink == 2)
735 } else if (ip->i_nlink > ext2_max_nlink(ip)) {
744 ext2_dec_nlink(struct inode *ip)
747 if (!S_ISDIR(ip->i_mode) || ip->i_nlink > 2)
752 * Rename system call.
753 * rename("foo", "bar");
756 * link("foo", "bar");
758 * but ``atomically''. Can't do full commit without saving state in the
759 * inode on disk which isn't feasible at this time. Best we can do is
760 * always guarantee the target exists.
762 * Basic algorithm is:
764 * 1) Bump link count on source while we're linking it to the
765 * target. This also ensure the inode won't be deleted out
766 * from underneath us while we work (it may be truncated by
767 * a concurrent `trunc' or `open' for creation).
768 * 2) Link source to destination. If destination already exists,
770 * 3) Unlink source reference to inode if still around. If a
771 * directory was moved and the parent of the destination
772 * is different from the source, patch the ".." entry in the
776 ext2_rename(struct vop_rename_args *ap)
778 struct vnode *tvp = ap->a_tvp;
779 struct vnode *tdvp = ap->a_tdvp;
780 struct vnode *fvp = ap->a_fvp;
781 struct vnode *fdvp = ap->a_fdvp;
782 struct componentname *tcnp = ap->a_tcnp;
783 struct componentname *fcnp = ap->a_fcnp;
784 struct inode *ip, *xp, *dp;
785 struct dirtemplate dirbuf;
786 int doingdirectory = 0, oldparent = 0, newparent = 0;
791 if ((tcnp->cn_flags & HASBUF) == 0 ||
792 (fcnp->cn_flags & HASBUF) == 0)
793 panic("ext2_rename: no name");
796 * Check for cross-device rename.
798 if ((fvp->v_mount != tdvp->v_mount) ||
799 (tvp && (fvp->v_mount != tvp->v_mount))) {
813 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
814 (VTOI(tdvp)->i_flags & APPEND))) {
820 * Renaming a file to itself has no effect. The upper layers should
821 * not call us in that case. Temporarily just warn if they do.
824 printf("ext2_rename: fvp == tvp (can't happen)\n");
829 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
833 if (ip->i_nlink >= ext2_max_nlink(ip) && !ext2_htree_has_idx(ip)) {
838 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
839 || (dp->i_flags & APPEND)) {
844 if ((ip->i_mode & IFMT) == IFDIR) {
846 * Avoid ".", "..", and aliases of "." for obvious reasons.
848 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
849 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
850 (ip->i_flag & IN_RENAME)) {
855 ip->i_flag |= IN_RENAME;
856 oldparent = dp->i_number;
862 * When the target exists, both the directory
863 * and target vnodes are returned locked.
871 * 1) Bump link count while we're moving stuff
872 * around. If we crash somewhere before
873 * completing our work, the link count
874 * may be wrong, but correctable.
877 ip->i_flag |= IN_CHANGE;
878 if ((error = ext2_update(fvp, !DOINGASYNC(fvp))) != 0) {
884 * If ".." must be changed (ie the directory gets a new
885 * parent) then the source directory must not be in the
886 * directory hierarchy above the target, as this would
887 * orphan everything below the source directory. Also
888 * the user must have write permission in the source so
889 * as to be able to change "..". We must repeat the call
890 * to namei, as the parent directory is unlocked by the
891 * call to checkpath().
893 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
895 if (oldparent != dp->i_number)
896 newparent = dp->i_number;
897 if (doingdirectory && newparent) {
898 if (error) /* write access check above */
902 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
906 error = relookup(tdvp, &tvp, tcnp);
916 * 2) If target doesn't exist, link the target
917 * to the source and unlink the source.
918 * Otherwise, rewrite the target directory
919 * entry to reference the source inode and
920 * expunge the original entry's existence.
923 if (dp->i_devvp != ip->i_devvp)
924 panic("ext2_rename: EXDEV");
926 * Account for ".." in new directory.
927 * When source and destination have the same
928 * parent we don't fool with the link count.
930 if (doingdirectory && newparent) {
931 error = ext2_inc_nlink(dp);
935 dp->i_flag |= IN_CHANGE;
936 error = ext2_update(tdvp, !DOINGASYNC(tdvp));
940 error = ext2_direnter(ip, tdvp, tcnp);
942 if (doingdirectory && newparent) {
944 dp->i_flag |= IN_CHANGE;
945 (void)ext2_update(tdvp, 1);
951 if (xp->i_devvp != dp->i_devvp || xp->i_devvp != ip->i_devvp)
952 panic("ext2_rename: EXDEV");
954 * Short circuit rename(foo, foo).
956 if (xp->i_number == ip->i_number)
957 panic("ext2_rename: same file");
959 * If the parent directory is "sticky", then the user must
960 * own the parent directory, or the destination of the rename,
961 * otherwise the destination may not be changed (except by
962 * root). This implements append-only directories.
964 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
965 tcnp->cn_cred->cr_uid != dp->i_uid &&
966 xp->i_uid != tcnp->cn_cred->cr_uid) {
971 * Target must be empty if a directory and have no links
972 * to it. Also, ensure source and target are compatible
973 * (both directories, or both not directories).
975 if ((xp->i_mode & IFMT) == IFDIR) {
976 if (!ext2_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
980 if (!doingdirectory) {
985 } else if (doingdirectory) {
989 error = ext2_dirrewrite(dp, ip, tcnp);
993 * If the target directory is in the same
994 * directory as the source directory,
995 * decrement the link count on the parent
996 * of the target directory.
998 if (doingdirectory && !newparent) {
1000 dp->i_flag |= IN_CHANGE;
1004 * Adjust the link count of the target to
1005 * reflect the dirrewrite above. If this is
1006 * a directory it is empty and there are
1007 * no links to it, so we can squash the inode and
1008 * any space associated with it. We disallowed
1009 * renaming over top of a directory with links to
1010 * it above, as the remaining link would point to
1011 * a directory without "." or ".." entries.
1014 if (doingdirectory) {
1015 if (--xp->i_nlink != 0)
1016 panic("ext2_rename: linked directory");
1017 error = ext2_truncate(tvp, (off_t)0, IO_SYNC,
1018 tcnp->cn_cred, tcnp->cn_thread);
1020 xp->i_flag |= IN_CHANGE;
1026 * 3) Unlink the source.
1028 fcnp->cn_flags &= ~MODMASK;
1029 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1031 error = relookup(fdvp, &fvp, fcnp);
1039 * From name has disappeared. IN_RENAME is not sufficient
1040 * to protect against directory races due to timing windows,
1041 * so we can't panic here.
1047 * Ensure that the directory entry still exists and has not
1048 * changed while the new name has been entered. If the source is
1049 * a file then the entry may have been unlinked or renamed. In
1050 * either case there is no further work to be done. If the source
1051 * is a directory then it cannot have been rmdir'ed; its link
1052 * count of three would cause a rmdir to fail with ENOTEMPTY.
1053 * The IN_RENAME flag ensures that it cannot be moved by another
1058 * From name resolves to a different inode. IN_RENAME is
1059 * not sufficient protection against timing window races
1060 * so we can't panic here.
1064 * If the source is a directory with a
1065 * new parent, the link count of the old
1066 * parent directory must be decremented
1067 * and ".." set to point to the new parent.
1069 if (doingdirectory && newparent) {
1071 dp->i_flag |= IN_CHANGE;
1072 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
1073 sizeof(struct dirtemplate), (off_t)0,
1074 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1075 tcnp->cn_cred, NOCRED, NULL, NULL);
1077 /* Like ufs little-endian: */
1078 namlen = dirbuf.dotdot_type;
1080 dirbuf.dotdot_name[0] != '.' ||
1081 dirbuf.dotdot_name[1] != '.') {
1082 ext2_dirbad(xp, (doff_t)12,
1083 "rename: mangled dir");
1085 dirbuf.dotdot_ino = newparent;
1086 (void)vn_rdwr(UIO_WRITE, fvp,
1088 sizeof(struct dirtemplate),
1089 (off_t)0, UIO_SYSSPACE,
1090 IO_NODELOCKED | IO_SYNC |
1091 IO_NOMACCHECK, tcnp->cn_cred,
1092 NOCRED, NULL, NULL);
1097 error = ext2_dirremove(fdvp, fcnp);
1100 xp->i_flag |= IN_CHANGE;
1102 xp->i_flag &= ~IN_RENAME;
1117 ip->i_flag &= ~IN_RENAME;
1118 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
1120 ip->i_flag |= IN_CHANGE;
1121 ip->i_flag &= ~IN_RENAME;
1130 ext2_do_posix1e_acl_inheritance_dir(struct vnode *dvp, struct vnode *tvp,
1131 mode_t dmode, struct ucred *cred, struct thread *td)
1134 struct inode *ip = VTOI(tvp);
1135 struct acl *dacl, *acl;
1137 acl = acl_alloc(M_WAITOK);
1138 dacl = acl_alloc(M_WAITOK);
1141 * Retrieve default ACL from parent, if any.
1143 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
1147 * Retrieved a default ACL, so merge mode and ACL if
1148 * necessary. If the ACL is empty, fall through to
1149 * the "not defined or available" case.
1151 if (acl->acl_cnt != 0) {
1152 dmode = acl_posix1e_newfilemode(dmode, acl);
1155 ext2_sync_acl_from_inode(ip, acl);
1162 * Just use the mode as-is.
1172 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
1174 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, cred, td);
1181 * XXX: This should not happen, as EOPNOTSUPP above
1182 * was supposed to free acl.
1185 printf("ext2_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
1201 ext2_do_posix1e_acl_inheritance_file(struct vnode *dvp, struct vnode *tvp,
1202 mode_t mode, struct ucred *cred, struct thread *td)
1205 struct inode *ip = VTOI(tvp);
1208 acl = acl_alloc(M_WAITOK);
1211 * Retrieve default ACL for parent, if any.
1213 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
1217 * Retrieved a default ACL, so merge mode and ACL if
1220 if (acl->acl_cnt != 0) {
1222 * Two possible ways for default ACL to not
1223 * be present. First, the EA can be
1224 * undefined, or second, the default ACL can
1225 * be blank. If it's blank, fall through to
1226 * the it's not defined case.
1228 mode = acl_posix1e_newfilemode(mode, acl);
1230 ext2_sync_acl_from_inode(ip, acl);
1237 * Just use the mode as-is.
1247 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
1254 * XXX: This should not happen, as EOPNOTSUPP above was
1255 * supposed to free acl.
1257 printf("ufs_do_posix1e_acl_inheritance_file: VOP_GETACL() "
1258 "but no VOP_SETACL()\n");
1259 /* panic("ufs_do_posix1e_acl_inheritance_file: VOP_GETACL() "
1260 "but no VOP_SETACL()"); */
1273 #endif /* UFS_ACL */
1279 ext2_mkdir(struct vop_mkdir_args *ap)
1281 struct vnode *dvp = ap->a_dvp;
1282 struct vattr *vap = ap->a_vap;
1283 struct componentname *cnp = ap->a_cnp;
1284 struct inode *ip, *dp;
1286 struct dirtemplate dirtemplate, *dtp;
1290 if ((cnp->cn_flags & HASBUF) == 0)
1291 panic("ext2_mkdir: no name");
1294 if ((nlink_t)dp->i_nlink >= ext2_max_nlink(dp) &&
1295 !ext2_htree_has_idx(dp)) {
1299 dmode = vap->va_mode & 0777;
1302 * Must simulate part of ext2_makeinode here to acquire the inode,
1303 * but not have it entered in the parent directory. The entry is
1304 * made later after writing "." and ".." entries.
1306 error = ext2_valloc(dvp, dmode, cnp->cn_cred, &tvp);
1310 ip->i_gid = dp->i_gid;
1314 * if we are hacking owners here, (only do this where told to)
1315 * and we are not giving it TOO root, (would subvert quotas)
1316 * then go ahead and give it to the other user.
1317 * The new directory also inherits the SUID bit.
1318 * If user's UID and dir UID are the same,
1319 * 'give it away' so that the SUID is still forced on.
1321 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1322 (dp->i_mode & ISUID) && dp->i_uid) {
1324 ip->i_uid = dp->i_uid;
1326 ip->i_uid = cnp->cn_cred->cr_uid;
1330 ip->i_uid = cnp->cn_cred->cr_uid;
1332 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1334 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1336 if (cnp->cn_flags & ISWHITEOUT)
1337 ip->i_flags |= UF_OPAQUE;
1338 error = ext2_update(tvp, 1);
1341 * Bump link count in parent directory
1342 * to reflect work done below. Should
1343 * be done before reference is created
1344 * so reparation is possible if we crash.
1347 dp->i_flag |= IN_CHANGE;
1348 error = ext2_update(dvp, !DOINGASYNC(dvp));
1352 /* Initialize directory with "." and ".." from static template. */
1353 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs,
1354 EXT2F_INCOMPAT_FTYPE))
1355 dtp = &mastertemplate;
1357 dtp = &omastertemplate;
1359 dirtemplate.dot_ino = ip->i_number;
1360 dirtemplate.dotdot_ino = dp->i_number;
1362 * note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE so let's
1363 * just redefine it - for this function only
1366 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->e2fs_bsize
1367 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
1368 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
1369 sizeof(dirtemplate), (off_t)0, UIO_SYSSPACE,
1370 IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
1374 dp->i_flag |= IN_CHANGE;
1377 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
1378 /* XXX should grow with balloc() */
1379 panic("ext2_mkdir: blksize");
1381 ip->i_size = DIRBLKSIZ;
1382 ip->i_flag |= IN_CHANGE;
1386 if (dvp->v_mount->mnt_flag & MNT_ACLS) {
1387 error = ext2_do_posix1e_acl_inheritance_dir(dvp, tvp, dmode,
1388 cnp->cn_cred, cnp->cn_thread);
1393 #endif /* UFS_ACL */
1395 /* Directory set up, now install its entry in the parent directory. */
1396 error = ext2_direnter(ip, dvp, cnp);
1399 dp->i_flag |= IN_CHANGE;
1403 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
1404 * for us because we set the link count to 0.
1408 ip->i_flag |= IN_CHANGE;
1415 #define DIRBLKSIZ DEV_BSIZE
1419 * Rmdir system call.
1422 ext2_rmdir(struct vop_rmdir_args *ap)
1424 struct vnode *vp = ap->a_vp;
1425 struct vnode *dvp = ap->a_dvp;
1426 struct componentname *cnp = ap->a_cnp;
1427 struct inode *ip, *dp;
1434 * Verify the directory is empty (and valid).
1435 * (Rmdir ".." won't be valid since
1436 * ".." will contain a reference to
1437 * the current directory and thus be
1440 if (!ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1444 if ((dp->i_flags & APPEND)
1445 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1450 * Delete reference to directory before purging
1451 * inode. If we crash in between, the directory
1452 * will be reattached to lost+found,
1454 error = ext2_dirremove(dvp, cnp);
1458 dp->i_flag |= IN_CHANGE;
1462 * Truncate inode. The only stuff left
1463 * in the directory is "." and "..".
1466 error = ext2_truncate(vp, (off_t)0, IO_SYNC, cnp->cn_cred,
1468 cache_purge(ITOV(ip));
1469 if (vn_lock(dvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
1471 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
1472 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1479 * symlink -- make a symbolic link
1482 ext2_symlink(struct vop_symlink_args *ap)
1484 struct vnode *vp, **vpp = ap->a_vpp;
1488 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1493 len = strlen(ap->a_target);
1494 if (len < vp->v_mount->mnt_maxsymlinklen) {
1496 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1498 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1500 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1501 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1502 ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
1509 * Return target name of a symbolic link
1512 ext2_readlink(struct vop_readlink_args *ap)
1514 struct vnode *vp = ap->a_vp;
1515 struct inode *ip = VTOI(vp);
1519 if (isize < vp->v_mount->mnt_maxsymlinklen) {
1520 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1523 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1527 * Calculate the logical to physical mapping if not done already,
1528 * then call the device strategy routine.
1530 * In order to be able to swap to a file, the ext2_bmaparray() operation may not
1531 * deadlock on memory. See ext2_bmap() for details.
1534 ext2_strategy(struct vop_strategy_args *ap)
1536 struct buf *bp = ap->a_bp;
1537 struct vnode *vp = ap->a_vp;
1542 if (vp->v_type == VBLK || vp->v_type == VCHR)
1543 panic("ext2_strategy: spec");
1544 if (bp->b_blkno == bp->b_lblkno) {
1545 error = ext2_bmaparray(vp, bp->b_lblkno, &blkno, NULL, NULL);
1546 bp->b_blkno = blkno;
1548 bp->b_error = error;
1549 bp->b_ioflags |= BIO_ERROR;
1553 if ((long)bp->b_blkno == -1)
1556 if ((long)bp->b_blkno == -1) {
1560 bp->b_iooffset = dbtob(bp->b_blkno);
1561 bo = VFSTOEXT2(vp->v_mount)->um_bo;
1562 BO_STRATEGY(bo, bp);
1567 * Print out the contents of an inode.
1570 ext2_print(struct vop_print_args *ap)
1572 struct vnode *vp = ap->a_vp;
1573 struct inode *ip = VTOI(vp);
1575 vn_printf(ip->i_devvp, "\tino %ju", (uintmax_t)ip->i_number);
1576 if (vp->v_type == VFIFO)
1583 * Close wrapper for fifos.
1585 * Update the times on the inode then do device close.
1588 ext2fifo_close(struct vop_close_args *ap)
1590 struct vnode *vp = ap->a_vp;
1593 if (vp->v_usecount > 1)
1594 ext2_itimes_locked(vp);
1596 return (fifo_specops.vop_close(ap));
1600 * Kqfilter wrapper for fifos.
1602 * Fall through to ext2 kqfilter routines if needed
1605 ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1609 error = fifo_specops.vop_kqfilter(ap);
1611 error = vfs_kqfilter(ap);
1616 * Return POSIX pathconf information applicable to ext2 filesystems.
1619 ext2_pathconf(struct vop_pathconf_args *ap)
1623 switch (ap->a_name) {
1625 if (ext2_htree_has_idx(VTOI(ap->a_vp)))
1626 *ap->a_retval = INT_MAX;
1628 *ap->a_retval = ext2_max_nlink(VTOI(ap->a_vp));
1635 case _PC_ACL_EXTENDED:
1636 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
1641 case _PC_ACL_PATH_MAX:
1642 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
1643 *ap->a_retval = ACL_MAX_ENTRIES;
1647 #endif /* UFS_ACL */
1649 case _PC_MIN_HOLE_SIZE:
1650 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1658 case _PC_ALLOC_SIZE_MIN:
1659 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
1661 case _PC_FILESIZEBITS:
1664 case _PC_REC_INCR_XFER_SIZE:
1665 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1667 case _PC_REC_MAX_XFER_SIZE:
1668 *ap->a_retval = -1; /* means ``unlimited'' */
1670 case _PC_REC_MIN_XFER_SIZE:
1671 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1673 case _PC_REC_XFER_ALIGN:
1674 *ap->a_retval = PAGE_SIZE;
1676 case _PC_SYMLINK_MAX:
1677 *ap->a_retval = MAXPATHLEN;
1681 error = vop_stdpathconf(ap);
1688 * Vnode operation to remove a named attribute.
1691 ext2_deleteextattr(struct vop_deleteextattr_args *ap)
1694 struct m_ext2fs *fs;
1697 ip = VTOI(ap->a_vp);
1700 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1701 return (EOPNOTSUPP);
1703 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1704 return (EOPNOTSUPP);
1706 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1707 ap->a_cred, ap->a_td, VWRITE);
1713 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1714 error = ext2_extattr_inode_delete(ip, ap->a_attrnamespace, ap->a_name);
1715 if (error != ENOATTR)
1720 error = ext2_extattr_block_delete(ip, ap->a_attrnamespace, ap->a_name);
1726 * Vnode operation to retrieve a named extended attribute.
1729 ext2_getextattr(struct vop_getextattr_args *ap)
1732 struct m_ext2fs *fs;
1735 ip = VTOI(ap->a_vp);
1738 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1739 return (EOPNOTSUPP);
1741 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1742 return (EOPNOTSUPP);
1744 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1745 ap->a_cred, ap->a_td, VREAD);
1749 if (ap->a_size != NULL)
1754 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1755 error = ext2_extattr_inode_get(ip, ap->a_attrnamespace,
1756 ap->a_name, ap->a_uio, ap->a_size);
1757 if (error != ENOATTR)
1762 error = ext2_extattr_block_get(ip, ap->a_attrnamespace,
1763 ap->a_name, ap->a_uio, ap->a_size);
1769 * Vnode operation to retrieve extended attributes on a vnode.
1772 ext2_listextattr(struct vop_listextattr_args *ap)
1775 struct m_ext2fs *fs;
1778 ip = VTOI(ap->a_vp);
1781 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1782 return (EOPNOTSUPP);
1784 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1785 return (EOPNOTSUPP);
1787 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1788 ap->a_cred, ap->a_td, VREAD);
1792 if (ap->a_size != NULL)
1795 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1796 error = ext2_extattr_inode_list(ip, ap->a_attrnamespace,
1797 ap->a_uio, ap->a_size);
1803 error = ext2_extattr_block_list(ip, ap->a_attrnamespace,
1804 ap->a_uio, ap->a_size);
1810 * Vnode operation to set a named attribute.
1813 ext2_setextattr(struct vop_setextattr_args *ap)
1816 struct m_ext2fs *fs;
1819 ip = VTOI(ap->a_vp);
1822 if (!EXT2_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_EXT_ATTR))
1823 return (EOPNOTSUPP);
1825 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
1826 return (EOPNOTSUPP);
1828 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
1829 ap->a_cred, ap->a_td, VWRITE);
1833 error = ext2_extattr_valid_attrname(ap->a_attrnamespace, ap->a_name);
1837 if (EXT2_INODE_SIZE(fs) != E2FS_REV0_INODE_SIZE) {
1838 error = ext2_extattr_inode_set(ip, ap->a_attrnamespace,
1839 ap->a_name, ap->a_uio);
1840 if (error != ENOSPC)
1844 error = ext2_extattr_block_set(ip, ap->a_attrnamespace,
1845 ap->a_name, ap->a_uio);
1851 * Vnode pointer to File handle
1855 ext2_vptofh(struct vop_vptofh_args *ap)
1860 ip = VTOI(ap->a_vp);
1861 ufhp = (struct ufid *)ap->a_fhp;
1862 ufhp->ufid_len = sizeof(struct ufid);
1863 ufhp->ufid_ino = ip->i_number;
1864 ufhp->ufid_gen = ip->i_gen;
1869 * Initialize the vnode associated with a new inode, handle aliased
1873 ext2_vinit(struct mount *mntp, struct vop_vector *fifoops, struct vnode **vpp)
1880 vp->v_type = IFTOVT(ip->i_mode);
1881 if (vp->v_type == VFIFO)
1884 if (ip->i_number == EXT2_ROOTINO)
1885 vp->v_vflag |= VV_ROOT;
1886 ip->i_modrev = init_va_filerev();
1892 * Allocate a new inode.
1895 ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1896 struct componentname *cnp)
1898 struct inode *ip, *pdir;
1904 if ((cnp->cn_flags & HASBUF) == 0)
1905 panic("ext2_makeinode: no name");
1908 if ((mode & IFMT) == 0)
1911 error = ext2_valloc(dvp, mode, cnp->cn_cred, &tvp);
1916 ip->i_gid = pdir->i_gid;
1921 * not the owner of the directory,
1922 * and we are hacking owners here, (only do this where told to)
1923 * and we are not giving it TOO root, (would subvert quotas)
1924 * then go ahead and give it to the other user.
1925 * Note that this drops off the execute bits for security.
1927 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1928 (pdir->i_mode & ISUID) &&
1929 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1930 ip->i_uid = pdir->i_uid;
1933 ip->i_uid = cnp->cn_cred->cr_uid;
1937 ip->i_uid = cnp->cn_cred->cr_uid;
1939 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1941 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1943 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred)) {
1944 if (priv_check_cred(cnp->cn_cred, PRIV_VFS_RETAINSUGID, 0))
1945 ip->i_mode &= ~ISGID;
1948 if (cnp->cn_flags & ISWHITEOUT)
1949 ip->i_flags |= UF_OPAQUE;
1952 * Make sure inode goes to disk before directory entry.
1954 error = ext2_update(tvp, !DOINGASYNC(tvp));
1959 if (dvp->v_mount->mnt_flag & MNT_ACLS) {
1960 error = ext2_do_posix1e_acl_inheritance_file(dvp, tvp, mode,
1961 cnp->cn_cred, cnp->cn_thread);
1965 #endif /* UFS_ACL */
1967 error = ext2_direnter(ip, dvp, cnp);
1976 * Write error occurred trying to update the inode
1977 * or the directory so must deallocate the inode.
1980 ip->i_flag |= IN_CHANGE;
1986 * Vnode op for reading.
1989 ext2_read(struct vop_read_args *ap)
1998 /* EXT4_EXT_LOCK(ip); */
1999 if (ip->i_flag & IN_E4EXTENTS)
2000 error = ext4_ext_read(ap);
2002 error = ext2_ind_read(ap);
2003 /* EXT4_EXT_UNLOCK(ip); */
2008 * Vnode op for reading.
2011 ext2_ind_read(struct vop_read_args *ap)
2016 struct m_ext2fs *fs;
2018 daddr_t lbn, nextlbn;
2020 long size, xfersize, blkoffset;
2021 int error, orig_resid, seqcount;
2026 ioflag = ap->a_ioflag;
2028 seqcount = ap->a_ioflag >> IO_SEQSHIFT;
2032 if (uio->uio_rw != UIO_READ)
2033 panic("%s: mode", "ext2_read");
2035 if (vp->v_type == VLNK) {
2036 if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
2037 panic("%s: short symlink", "ext2_read");
2038 } else if (vp->v_type != VREG && vp->v_type != VDIR)
2039 panic("%s: type %d", "ext2_read", vp->v_type);
2041 orig_resid = uio->uio_resid;
2042 KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
2043 if (orig_resid == 0)
2045 KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
2047 if (uio->uio_offset < ip->i_size &&
2048 uio->uio_offset >= fs->e2fs_maxfilesize)
2051 for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
2052 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
2054 lbn = lblkno(fs, uio->uio_offset);
2056 size = blksize(fs, ip, lbn);
2057 blkoffset = blkoff(fs, uio->uio_offset);
2059 xfersize = fs->e2fs_fsize - blkoffset;
2060 if (uio->uio_resid < xfersize)
2061 xfersize = uio->uio_resid;
2062 if (bytesinfile < xfersize)
2063 xfersize = bytesinfile;
2065 if (lblktosize(fs, nextlbn) >= ip->i_size)
2066 error = bread(vp, lbn, size, NOCRED, &bp);
2067 else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
2068 error = cluster_read(vp, ip->i_size, lbn, size,
2069 NOCRED, blkoffset + uio->uio_resid, seqcount,
2071 } else if (seqcount > 1) {
2072 u_int nextsize = blksize(fs, ip, nextlbn);
2074 error = breadn(vp, lbn,
2075 size, &nextlbn, &nextsize, 1, NOCRED, &bp);
2077 error = bread(vp, lbn, size, NOCRED, &bp);
2085 * We should only get non-zero b_resid when an I/O error
2086 * has occurred, which should cause us to break above.
2087 * However, if the short read did not cause an error,
2088 * then we want to ensure that we do not uiomove bad
2089 * or uninitialized data.
2091 size -= bp->b_resid;
2092 if (size < xfersize) {
2097 error = uiomove((char *)bp->b_data + blkoffset,
2098 (int)xfersize, uio);
2101 vfs_bio_brelse(bp, ioflag);
2105 * This can only happen in the case of an error because the loop
2106 * above resets bp to NULL on each iteration and on normal
2107 * completion has not set a new value into it. so it must have come
2108 * from a 'break' statement
2111 vfs_bio_brelse(bp, ioflag);
2113 if ((error == 0 || uio->uio_resid != orig_resid) &&
2114 (vp->v_mount->mnt_flag & (MNT_NOATIME | MNT_RDONLY)) == 0)
2115 ip->i_flag |= IN_ACCESS;
2120 ext2_ioctl(struct vop_ioctl_args *ap)
2123 switch (ap->a_command) {
2126 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
2127 (off_t *)ap->a_data, ap->a_cred));
2134 * this function handles ext4 extents block mapping
2137 ext4_ext_read(struct vop_read_args *ap)
2139 static unsigned char zeroes[EXT2_MAX_BLOCK_SIZE];
2143 struct m_ext2fs *fs;
2145 struct ext4_extent nex, *ep;
2146 struct ext4_extent_path path;
2147 daddr_t lbn, newblk;
2152 long size, xfersize, blkoffset;
2157 memset(&path, 0, sizeof(path));
2159 orig_resid = uio->uio_resid;
2160 KASSERT(orig_resid >= 0, ("%s: uio->uio_resid < 0", __func__));
2161 if (orig_resid == 0)
2163 KASSERT(uio->uio_offset >= 0, ("%s: uio->uio_offset < 0", __func__));
2165 if (uio->uio_offset < ip->i_size && uio->uio_offset >= fs->e2fs_maxfilesize)
2168 while (uio->uio_resid > 0) {
2169 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
2171 lbn = lblkno(fs, uio->uio_offset);
2172 size = blksize(fs, ip, lbn);
2173 blkoffset = blkoff(fs, uio->uio_offset);
2175 xfersize = fs->e2fs_fsize - blkoffset;
2176 xfersize = MIN(xfersize, uio->uio_resid);
2177 xfersize = MIN(xfersize, bytesinfile);
2179 /* get block from ext4 extent cache */
2180 cache_type = ext4_ext_in_cache(ip, lbn, &nex);
2181 switch (cache_type) {
2182 case EXT4_EXT_CACHE_NO:
2183 ext4_ext_find_extent(fs, ip, lbn, &path);
2184 if (path.ep_is_sparse)
2185 ep = &path.ep_sparse_ext;
2191 ext4_ext_put_cache(ip, ep,
2192 path.ep_is_sparse ? EXT4_EXT_CACHE_GAP : EXT4_EXT_CACHE_IN);
2194 newblk = lbn - ep->e_blk + (ep->e_start_lo |
2195 (daddr_t)ep->e_start_hi << 32);
2197 if (path.ep_bp != NULL) {
2203 case EXT4_EXT_CACHE_GAP:
2204 /* block has not been allocated yet */
2207 case EXT4_EXT_CACHE_IN:
2208 newblk = lbn - nex.e_blk + (nex.e_start_lo |
2209 (daddr_t)nex.e_start_hi << 32);
2213 panic("%s: invalid cache type", __func__);
2216 if (cache_type == EXT4_EXT_CACHE_GAP ||
2217 (cache_type == EXT4_EXT_CACHE_NO && path.ep_is_sparse)) {
2218 if (xfersize > sizeof(zeroes))
2219 xfersize = sizeof(zeroes);
2220 error = uiomove(zeroes, xfersize, uio);
2224 error = bread(ip->i_devvp, fsbtodb(fs, newblk), size,
2231 size -= bp->b_resid;
2232 if (size < xfersize) {
2239 error = uiomove(bp->b_data + blkoffset, xfersize, uio);
2250 * Vnode op for writing.
2253 ext2_write(struct vop_write_args *ap)
2258 struct m_ext2fs *fs;
2262 int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
2264 ioflag = ap->a_ioflag;
2268 seqcount = ioflag >> IO_SEQSHIFT;
2272 if (uio->uio_rw != UIO_WRITE)
2273 panic("%s: mode", "ext2_write");
2276 switch (vp->v_type) {
2278 if (ioflag & IO_APPEND)
2279 uio->uio_offset = ip->i_size;
2280 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
2286 /* XXX differs from ffs -- this is called from ext2_mkdir(). */
2287 if ((ioflag & IO_SYNC) == 0)
2288 panic("ext2_write: nonsync dir write");
2291 panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
2292 vp->v_type, (intmax_t)uio->uio_offset,
2293 (intmax_t)uio->uio_resid);
2296 KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
2297 KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
2299 if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
2302 * Maybe this should be above the vnode op call, but so long as
2303 * file servers have no limits, I don't think it matters.
2305 if (vn_rlimit_fsize(vp, uio, uio->uio_td))
2308 resid = uio->uio_resid;
2310 if (seqcount > BA_SEQMAX)
2311 flags = BA_SEQMAX << BA_SEQSHIFT;
2313 flags = seqcount << BA_SEQSHIFT;
2314 if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
2317 for (error = 0; uio->uio_resid > 0;) {
2318 lbn = lblkno(fs, uio->uio_offset);
2319 blkoffset = blkoff(fs, uio->uio_offset);
2320 xfersize = fs->e2fs_fsize - blkoffset;
2321 if (uio->uio_resid < xfersize)
2322 xfersize = uio->uio_resid;
2323 if (uio->uio_offset + xfersize > ip->i_size)
2324 vnode_pager_setsize(vp, uio->uio_offset + xfersize);
2327 * We must perform a read-before-write if the transfer size
2328 * does not cover the entire buffer.
2330 if (fs->e2fs_bsize > xfersize)
2333 flags &= ~BA_CLRBUF;
2334 error = ext2_balloc(ip, lbn, blkoffset + xfersize,
2335 ap->a_cred, &bp, flags);
2339 if ((ioflag & (IO_SYNC | IO_INVAL)) == (IO_SYNC | IO_INVAL))
2340 bp->b_flags |= B_NOCACHE;
2341 if (uio->uio_offset + xfersize > ip->i_size)
2342 ip->i_size = uio->uio_offset + xfersize;
2343 size = blksize(fs, ip, lbn) - bp->b_resid;
2344 if (size < xfersize)
2348 uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
2350 * If the buffer is not already filled and we encounter an
2351 * error while trying to fill it, we have to clear out any
2352 * garbage data from the pages instantiated for the buffer.
2353 * If we do not, a failed uiomove() during a write can leave
2354 * the prior contents of the pages exposed to a userland mmap.
2356 * Note that we need only clear buffers with a transfer size
2357 * equal to the block size because buffers with a shorter
2358 * transfer size were cleared above by the call to ext2_balloc()
2359 * with the BA_CLRBUF flag set.
2361 * If the source region for uiomove identically mmaps the
2362 * buffer, uiomove() performed the NOP copy, and the buffer
2363 * content remains valid because the page fault handler
2364 * validated the pages.
2366 if (error != 0 && (bp->b_flags & B_CACHE) == 0 &&
2367 fs->e2fs_bsize == xfersize)
2370 vfs_bio_set_flags(bp, ioflag);
2373 * If IO_SYNC each buffer is written synchronously. Otherwise
2374 * if we have a severe page deficiency write the buffer
2375 * asynchronously. Otherwise try to cluster, and if that
2376 * doesn't do it then either do an async write (if O_DIRECT),
2377 * or a delayed write (if not).
2379 if (ioflag & IO_SYNC) {
2381 } else if (vm_page_count_severe() ||
2382 buf_dirty_count_severe() ||
2383 (ioflag & IO_ASYNC)) {
2384 bp->b_flags |= B_CLUSTEROK;
2386 } else if (xfersize + blkoffset == fs->e2fs_fsize) {
2387 if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
2388 bp->b_flags |= B_CLUSTEROK;
2389 cluster_write(vp, bp, ip->i_size, seqcount, 0);
2393 } else if (ioflag & IO_DIRECT) {
2394 bp->b_flags |= B_CLUSTEROK;
2397 bp->b_flags |= B_CLUSTEROK;
2400 if (error || xfersize == 0)
2404 * If we successfully wrote any data, and we are not the superuser
2405 * we clear the setuid and setgid bits as a precaution against
2408 if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid &&
2410 if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
2411 ip->i_mode &= ~(ISUID | ISGID);
2414 if (ioflag & IO_UNIT) {
2415 (void)ext2_truncate(vp, osize,
2416 ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
2417 uio->uio_offset -= resid - uio->uio_resid;
2418 uio->uio_resid = resid;
2421 if (uio->uio_resid != resid) {
2422 ip->i_flag |= IN_CHANGE | IN_UPDATE;
2423 if (ioflag & IO_SYNC)
2424 error = ext2_update(vp, 1);