2 * modified for EXT2FS support in Lites 1.1
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
8 * Copyright (c) 1982, 1986, 1989, 1993
9 * The Regents of the University of California. All rights reserved.
10 * (c) UNIX System Laboratories, Inc.
11 * All or some portions of this file are derived from material licensed
12 * to the University of California by American Telephone and Telegraph
13 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
14 * the permission of UNIX System Laboratories, Inc.
16 * Redistribution and use in source and binary forms, with or without
17 * modification, are permitted provided that the following conditions
19 * 1. Redistributions of source code must retain the above copyright
20 * notice, this list of conditions and the following disclaimer.
21 * 2. Redistributions in binary form must reproduce the above copyright
22 * notice, this list of conditions and the following disclaimer in the
23 * documentation and/or other materials provided with the distribution.
24 * 4. Neither the name of the University nor the names of its contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
40 * @(#)ufs_vnops.c 8.7 (Berkeley) 2/3/94
41 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
45 #include "opt_suiddir.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
55 #include <sys/endian.h>
57 #include <sys/mount.h>
58 #include <sys/unistd.h>
60 #include <sys/vnode.h>
61 #include <sys/namei.h>
62 #include <sys/lockf.h>
63 #include <sys/event.h>
68 #include <vm/vm_page.h>
69 #include <vm/vm_object.h>
70 #include <vm/vm_extern.h>
71 #include <vm/vnode_pager.h>
73 #include "opt_directio.h"
75 #include <fs/fifofs/fifo.h>
77 #include <ufs/ufs/dir.h>
79 #include <fs/ext2fs/fs.h>
80 #include <fs/ext2fs/inode.h>
81 #include <fs/ext2fs/ext2_extern.h>
82 #include <fs/ext2fs/ext2fs.h>
83 #include <fs/ext2fs/ext2_dinode.h>
84 #include <fs/ext2fs/ext2_dir.h>
85 #include <fs/ext2fs/ext2_mount.h>
87 static int ext2_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
88 static void ext2_itimes_locked(struct vnode *);
89 static int ext4_ext_read(struct vop_read_args *);
90 static int ext2_ind_read(struct vop_read_args *);
92 static vop_access_t ext2_access;
93 static int ext2_chmod(struct vnode *, int, struct ucred *, struct thread *);
94 static int ext2_chown(struct vnode *, uid_t, gid_t, struct ucred *,
96 static vop_close_t ext2_close;
97 static vop_create_t ext2_create;
98 static vop_fsync_t ext2_fsync;
99 static vop_getattr_t ext2_getattr;
100 static vop_ioctl_t ext2_ioctl;
101 static vop_link_t ext2_link;
102 static vop_mkdir_t ext2_mkdir;
103 static vop_mknod_t ext2_mknod;
104 static vop_open_t ext2_open;
105 static vop_pathconf_t ext2_pathconf;
106 static vop_print_t ext2_print;
107 static vop_read_t ext2_read;
108 static vop_readlink_t ext2_readlink;
109 static vop_remove_t ext2_remove;
110 static vop_rename_t ext2_rename;
111 static vop_rmdir_t ext2_rmdir;
112 static vop_setattr_t ext2_setattr;
113 static vop_strategy_t ext2_strategy;
114 static vop_symlink_t ext2_symlink;
115 static vop_write_t ext2_write;
116 static vop_vptofh_t ext2_vptofh;
117 static vop_close_t ext2fifo_close;
118 static vop_kqfilter_t ext2fifo_kqfilter;
120 /* Global vfs data structures for ext2. */
121 struct vop_vector ext2_vnodeops = {
122 .vop_default = &default_vnodeops,
123 .vop_access = ext2_access,
124 .vop_bmap = ext2_bmap,
125 .vop_cachedlookup = ext2_lookup,
126 .vop_close = ext2_close,
127 .vop_create = ext2_create,
128 .vop_fsync = ext2_fsync,
129 .vop_getattr = ext2_getattr,
130 .vop_inactive = ext2_inactive,
131 .vop_ioctl = ext2_ioctl,
132 .vop_link = ext2_link,
133 .vop_lookup = vfs_cache_lookup,
134 .vop_mkdir = ext2_mkdir,
135 .vop_mknod = ext2_mknod,
136 .vop_open = ext2_open,
137 .vop_pathconf = ext2_pathconf,
138 .vop_poll = vop_stdpoll,
139 .vop_print = ext2_print,
140 .vop_read = ext2_read,
141 .vop_readdir = ext2_readdir,
142 .vop_readlink = ext2_readlink,
143 .vop_reallocblks = ext2_reallocblks,
144 .vop_reclaim = ext2_reclaim,
145 .vop_remove = ext2_remove,
146 .vop_rename = ext2_rename,
147 .vop_rmdir = ext2_rmdir,
148 .vop_setattr = ext2_setattr,
149 .vop_strategy = ext2_strategy,
150 .vop_symlink = ext2_symlink,
151 .vop_write = ext2_write,
152 .vop_vptofh = ext2_vptofh,
155 struct vop_vector ext2_fifoops = {
156 .vop_default = &fifo_specops,
157 .vop_access = ext2_access,
158 .vop_close = ext2fifo_close,
159 .vop_fsync = ext2_fsync,
160 .vop_getattr = ext2_getattr,
161 .vop_inactive = ext2_inactive,
162 .vop_kqfilter = ext2fifo_kqfilter,
163 .vop_print = ext2_print,
164 .vop_read = VOP_PANIC,
165 .vop_reclaim = ext2_reclaim,
166 .vop_setattr = ext2_setattr,
167 .vop_write = VOP_PANIC,
168 .vop_vptofh = ext2_vptofh,
172 * A virgin directory (no blushing please).
173 * Note that the type and namlen fields are reversed relative to ext2.
174 * Also, we don't use `struct odirtemplate', since it would just cause
175 * endianness problems.
177 static struct dirtemplate mastertemplate = {
178 0, 12, 1, EXT2_FT_DIR, ".",
179 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
181 static struct dirtemplate omastertemplate = {
182 0, 12, 1, EXT2_FT_UNKNOWN, ".",
183 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
187 ext2_itimes_locked(struct vnode *vp)
192 ASSERT_VI_LOCKED(vp, __func__);
195 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
197 if ((vp->v_type == VBLK || vp->v_type == VCHR))
198 ip->i_flag |= IN_LAZYMOD;
200 ip->i_flag |= IN_MODIFIED;
201 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
203 if (ip->i_flag & IN_ACCESS) {
204 ip->i_atime = ts.tv_sec;
205 ip->i_atimensec = ts.tv_nsec;
207 if (ip->i_flag & IN_UPDATE) {
208 ip->i_mtime = ts.tv_sec;
209 ip->i_mtimensec = ts.tv_nsec;
212 if (ip->i_flag & IN_CHANGE) {
213 ip->i_ctime = ts.tv_sec;
214 ip->i_ctimensec = ts.tv_nsec;
217 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
221 ext2_itimes(struct vnode *vp)
225 ext2_itimes_locked(vp);
230 * Create a regular file
233 ext2_create(struct vop_create_args *ap)
238 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
239 ap->a_dvp, ap->a_vpp, ap->a_cnp);
246 ext2_open(struct vop_open_args *ap)
249 if (ap->a_vp->v_type == VBLK || ap->a_vp->v_type == VCHR)
253 * Files marked append-only must be opened for appending.
255 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
256 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
259 vnode_create_vobject(ap->a_vp, VTOI(ap->a_vp)->i_size, ap->a_td);
267 * Update the times on the inode.
270 ext2_close(struct vop_close_args *ap)
272 struct vnode *vp = ap->a_vp;
275 if (vp->v_usecount > 1)
276 ext2_itimes_locked(vp);
282 ext2_access(struct vop_access_args *ap)
284 struct vnode *vp = ap->a_vp;
285 struct inode *ip = VTOI(vp);
286 accmode_t accmode = ap->a_accmode;
289 if (vp->v_type == VBLK || vp->v_type == VCHR)
293 * Disallow write attempts on read-only file systems;
294 * unless the file is a socket, fifo, or a block or
295 * character device resident on the file system.
297 if (accmode & VWRITE) {
298 switch (vp->v_type) {
302 if (vp->v_mount->mnt_flag & MNT_RDONLY)
310 /* If immutable bit set, nobody gets to write it. */
311 if ((accmode & VWRITE) && (ip->i_flags & (SF_IMMUTABLE | SF_SNAPSHOT)))
314 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
315 ap->a_accmode, ap->a_cred, NULL);
320 ext2_getattr(struct vop_getattr_args *ap)
322 struct vnode *vp = ap->a_vp;
323 struct inode *ip = VTOI(vp);
324 struct vattr *vap = ap->a_vap;
328 * Copy from inode table
330 vap->va_fsid = dev2udev(ip->i_devvp->v_rdev);
331 vap->va_fileid = ip->i_number;
332 vap->va_mode = ip->i_mode & ~IFMT;
333 vap->va_nlink = ip->i_nlink;
334 vap->va_uid = ip->i_uid;
335 vap->va_gid = ip->i_gid;
336 vap->va_rdev = ip->i_rdev;
337 vap->va_size = ip->i_size;
338 vap->va_atime.tv_sec = ip->i_atime;
339 vap->va_atime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_atimensec : 0;
340 vap->va_mtime.tv_sec = ip->i_mtime;
341 vap->va_mtime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_mtimensec : 0;
342 vap->va_ctime.tv_sec = ip->i_ctime;
343 vap->va_ctime.tv_nsec = E2DI_HAS_XTIME(ip) ? ip->i_ctimensec : 0;
344 if E2DI_HAS_XTIME(ip) {
345 vap->va_birthtime.tv_sec = ip->i_birthtime;
346 vap->va_birthtime.tv_nsec = ip->i_birthnsec;
348 vap->va_flags = ip->i_flags;
349 vap->va_gen = ip->i_gen;
350 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
351 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
352 vap->va_type = IFTOVT(ip->i_mode);
353 vap->va_filerev = ip->i_modrev;
358 * Set attribute vnode op. called from several syscalls
361 ext2_setattr(struct vop_setattr_args *ap)
363 struct vattr *vap = ap->a_vap;
364 struct vnode *vp = ap->a_vp;
365 struct inode *ip = VTOI(vp);
366 struct ucred *cred = ap->a_cred;
367 struct thread *td = curthread;
371 * Check for unsettable attributes.
373 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
374 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
375 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
376 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
379 if (vap->va_flags != VNOVAL) {
380 /* Disallow flags not supported by ext2fs. */
381 if(vap->va_flags & ~(SF_APPEND | SF_IMMUTABLE | UF_NODUMP))
384 if (vp->v_mount->mnt_flag & MNT_RDONLY)
387 * Callers may only modify the file flags on objects they
388 * have VADMIN rights for.
390 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
393 * Unprivileged processes and privileged processes in
394 * jail() are not permitted to unset system flags, or
395 * modify flags if any system flags are set.
396 * Privileged non-jail processes may not modify system flags
397 * if securelevel > 0 and any existing system flags are set.
399 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
401 (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
402 error = securelevel_gt(cred, 0);
406 ip->i_flags = vap->va_flags;
409 (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
410 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
412 ip->i_flags &= SF_SETTABLE;
413 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
415 ip->i_flag |= IN_CHANGE;
416 if (ip->i_flags & (IMMUTABLE | APPEND))
419 if (ip->i_flags & (IMMUTABLE | APPEND))
422 * Go through the fields and update iff not VNOVAL.
424 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
425 if (vp->v_mount->mnt_flag & MNT_RDONLY)
427 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred,
431 if (vap->va_size != VNOVAL) {
433 * Disallow write attempts on read-only file systems;
434 * unless the file is a socket, fifo, or a block or
435 * character device resident on the file system.
437 switch (vp->v_type) {
442 if (vp->v_mount->mnt_flag & MNT_RDONLY)
448 if ((error = ext2_truncate(vp, vap->va_size, 0, cred, td)) != 0)
451 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
452 if (vp->v_mount->mnt_flag & MNT_RDONLY)
456 * If times is NULL, ... The caller must be the owner of
457 * the file, have permission to write the file, or be the
459 * If times is non-NULL, ... The caller must be the owner of
460 * the file or be the super-user.
462 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
463 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
464 (error = VOP_ACCESS(vp, VWRITE, cred, td))))
466 if (vap->va_atime.tv_sec != VNOVAL)
467 ip->i_flag |= IN_ACCESS;
468 if (vap->va_mtime.tv_sec != VNOVAL)
469 ip->i_flag |= IN_CHANGE | IN_UPDATE;
471 if (vap->va_atime.tv_sec != VNOVAL) {
472 ip->i_atime = vap->va_atime.tv_sec;
473 ip->i_atimensec = vap->va_atime.tv_nsec;
475 if (vap->va_mtime.tv_sec != VNOVAL) {
476 ip->i_mtime = vap->va_mtime.tv_sec;
477 ip->i_mtimensec = vap->va_mtime.tv_nsec;
479 ip->i_birthtime = vap->va_birthtime.tv_sec;
480 ip->i_birthnsec = vap->va_birthtime.tv_nsec;
481 error = ext2_update(vp, 0);
486 if (vap->va_mode != (mode_t)VNOVAL) {
487 if (vp->v_mount->mnt_flag & MNT_RDONLY)
489 error = ext2_chmod(vp, (int)vap->va_mode, cred, td);
495 * Change the mode on a file.
496 * Inode must be locked before calling.
499 ext2_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
501 struct inode *ip = VTOI(vp);
505 * To modify the permissions on a file, must possess VADMIN
508 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
511 * Privileged processes may set the sticky bit on non-directories,
512 * as well as set the setgid bit on a file with a group that the
513 * process is not a member of.
515 if (vp->v_type != VDIR && (mode & S_ISTXT)) {
516 error = priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0);
520 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
521 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
525 ip->i_mode &= ~ALLPERMS;
526 ip->i_mode |= (mode & ALLPERMS);
527 ip->i_flag |= IN_CHANGE;
532 * Perform chown operation on inode ip;
533 * inode must be locked prior to call.
536 ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
539 struct inode *ip = VTOI(vp);
544 if (uid == (uid_t)VNOVAL)
546 if (gid == (gid_t)VNOVAL)
549 * To modify the ownership of a file, must possess VADMIN
552 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
555 * To change the owner of a file, or change the group of a file
556 * to a group of which we are not a member, the caller must
559 if (uid != ip->i_uid || (gid != ip->i_gid &&
560 !groupmember(gid, cred))) {
561 error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0);
569 ip->i_flag |= IN_CHANGE;
570 if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
571 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0) != 0)
572 ip->i_mode &= ~(ISUID | ISGID);
578 * Synch an open file.
582 ext2_fsync(struct vop_fsync_args *ap)
585 * Flush all dirty buffers associated with a vnode.
590 return (ext2_update(ap->a_vp, ap->a_waitfor == MNT_WAIT));
598 ext2_mknod(struct vop_mknod_args *ap)
600 struct vattr *vap = ap->a_vap;
601 struct vnode **vpp = ap->a_vpp;
606 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
607 ap->a_dvp, vpp, ap->a_cnp);
611 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
612 if (vap->va_rdev != VNOVAL) {
614 * Want to be able to use this to make badblock
615 * inodes, so don't truncate the dev number.
617 ip->i_rdev = vap->va_rdev;
620 * Remove inode, then reload it through VFS_VGET so it is
621 * checked to see if it is an alias of an existing entry in
622 * the inode cache. XXX I don't believe this is necessary now.
624 (*vpp)->v_type = VNON;
625 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
628 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
637 ext2_remove(struct vop_remove_args *ap)
640 struct vnode *vp = ap->a_vp;
641 struct vnode *dvp = ap->a_dvp;
645 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
646 (VTOI(dvp)->i_flags & APPEND)) {
650 error = ext2_dirremove(dvp, ap->a_cnp);
653 ip->i_flag |= IN_CHANGE;
663 ext2_link(struct vop_link_args *ap)
665 struct vnode *vp = ap->a_vp;
666 struct vnode *tdvp = ap->a_tdvp;
667 struct componentname *cnp = ap->a_cnp;
672 if ((cnp->cn_flags & HASBUF) == 0)
673 panic("ext2_link: no name");
675 if (tdvp->v_mount != vp->v_mount) {
680 if ((nlink_t)ip->i_nlink >= EXT2_LINK_MAX) {
684 if (ip->i_flags & (IMMUTABLE | APPEND)) {
689 ip->i_flag |= IN_CHANGE;
690 error = ext2_update(vp, !DOINGASYNC(vp));
692 error = ext2_direnter(ip, tdvp, cnp);
695 ip->i_flag |= IN_CHANGE;
702 * Rename system call.
703 * rename("foo", "bar");
706 * link("foo", "bar");
708 * but ``atomically''. Can't do full commit without saving state in the
709 * inode on disk which isn't feasible at this time. Best we can do is
710 * always guarantee the target exists.
712 * Basic algorithm is:
714 * 1) Bump link count on source while we're linking it to the
715 * target. This also ensure the inode won't be deleted out
716 * from underneath us while we work (it may be truncated by
717 * a concurrent `trunc' or `open' for creation).
718 * 2) Link source to destination. If destination already exists,
720 * 3) Unlink source reference to inode if still around. If a
721 * directory was moved and the parent of the destination
722 * is different from the source, patch the ".." entry in the
726 ext2_rename(struct vop_rename_args *ap)
728 struct vnode *tvp = ap->a_tvp;
729 struct vnode *tdvp = ap->a_tdvp;
730 struct vnode *fvp = ap->a_fvp;
731 struct vnode *fdvp = ap->a_fdvp;
732 struct componentname *tcnp = ap->a_tcnp;
733 struct componentname *fcnp = ap->a_fcnp;
734 struct inode *ip, *xp, *dp;
735 struct dirtemplate dirbuf;
736 int doingdirectory = 0, oldparent = 0, newparent = 0;
741 if ((tcnp->cn_flags & HASBUF) == 0 ||
742 (fcnp->cn_flags & HASBUF) == 0)
743 panic("ext2_rename: no name");
746 * Check for cross-device rename.
748 if ((fvp->v_mount != tdvp->v_mount) ||
749 (tvp && (fvp->v_mount != tvp->v_mount))) {
763 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
764 (VTOI(tdvp)->i_flags & APPEND))) {
770 * Renaming a file to itself has no effect. The upper layers should
771 * not call us in that case. Temporarily just warn if they do.
774 printf("ext2_rename: fvp == tvp (can't happen)\n");
779 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
783 if (ip->i_nlink >= EXT2_LINK_MAX) {
788 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
789 || (dp->i_flags & APPEND)) {
794 if ((ip->i_mode & IFMT) == IFDIR) {
796 * Avoid ".", "..", and aliases of "." for obvious reasons.
798 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
799 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
800 (ip->i_flag & IN_RENAME)) {
805 ip->i_flag |= IN_RENAME;
806 oldparent = dp->i_number;
812 * When the target exists, both the directory
813 * and target vnodes are returned locked.
821 * 1) Bump link count while we're moving stuff
822 * around. If we crash somewhere before
823 * completing our work, the link count
824 * may be wrong, but correctable.
827 ip->i_flag |= IN_CHANGE;
828 if ((error = ext2_update(fvp, !DOINGASYNC(fvp))) != 0) {
834 * If ".." must be changed (ie the directory gets a new
835 * parent) then the source directory must not be in the
836 * directory hierarchy above the target, as this would
837 * orphan everything below the source directory. Also
838 * the user must have write permission in the source so
839 * as to be able to change "..". We must repeat the call
840 * to namei, as the parent directory is unlocked by the
841 * call to checkpath().
843 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
845 if (oldparent != dp->i_number)
846 newparent = dp->i_number;
847 if (doingdirectory && newparent) {
848 if (error) /* write access check above */
852 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
856 error = relookup(tdvp, &tvp, tcnp);
866 * 2) If target doesn't exist, link the target
867 * to the source and unlink the source.
868 * Otherwise, rewrite the target directory
869 * entry to reference the source inode and
870 * expunge the original entry's existence.
873 if (dp->i_devvp != ip->i_devvp)
874 panic("ext2_rename: EXDEV");
876 * Account for ".." in new directory.
877 * When source and destination have the same
878 * parent we don't fool with the link count.
880 if (doingdirectory && newparent) {
881 if ((nlink_t)dp->i_nlink >= EXT2_LINK_MAX) {
886 dp->i_flag |= IN_CHANGE;
887 error = ext2_update(tdvp, !DOINGASYNC(tdvp));
891 error = ext2_direnter(ip, tdvp, tcnp);
893 if (doingdirectory && newparent) {
895 dp->i_flag |= IN_CHANGE;
896 (void)ext2_update(tdvp, 1);
902 if (xp->i_devvp != dp->i_devvp || xp->i_devvp != ip->i_devvp)
903 panic("ext2_rename: EXDEV");
905 * Short circuit rename(foo, foo).
907 if (xp->i_number == ip->i_number)
908 panic("ext2_rename: same file");
910 * If the parent directory is "sticky", then the user must
911 * own the parent directory, or the destination of the rename,
912 * otherwise the destination may not be changed (except by
913 * root). This implements append-only directories.
915 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
916 tcnp->cn_cred->cr_uid != dp->i_uid &&
917 xp->i_uid != tcnp->cn_cred->cr_uid) {
922 * Target must be empty if a directory and have no links
923 * to it. Also, ensure source and target are compatible
924 * (both directories, or both not directories).
926 if ((xp->i_mode&IFMT) == IFDIR) {
927 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
932 if (!doingdirectory) {
937 } else if (doingdirectory) {
941 error = ext2_dirrewrite(dp, ip, tcnp);
945 * If the target directory is in the same
946 * directory as the source directory,
947 * decrement the link count on the parent
948 * of the target directory.
950 if (doingdirectory && !newparent) {
952 dp->i_flag |= IN_CHANGE;
956 * Adjust the link count of the target to
957 * reflect the dirrewrite above. If this is
958 * a directory it is empty and there are
959 * no links to it, so we can squash the inode and
960 * any space associated with it. We disallowed
961 * renaming over top of a directory with links to
962 * it above, as the remaining link would point to
963 * a directory without "." or ".." entries.
966 if (doingdirectory) {
967 if (--xp->i_nlink != 0)
968 panic("ext2_rename: linked directory");
969 error = ext2_truncate(tvp, (off_t)0, IO_SYNC,
970 tcnp->cn_cred, tcnp->cn_thread);
972 xp->i_flag |= IN_CHANGE;
978 * 3) Unlink the source.
980 fcnp->cn_flags &= ~MODMASK;
981 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
983 error = relookup(fdvp, &fvp, fcnp);
991 * From name has disappeared.
994 panic("ext2_rename: lost dir entry");
999 * Ensure that the directory entry still exists and has not
1000 * changed while the new name has been entered. If the source is
1001 * a file then the entry may have been unlinked or renamed. In
1002 * either case there is no further work to be done. If the source
1003 * is a directory then it cannot have been rmdir'ed; its link
1004 * count of three would cause a rmdir to fail with ENOTEMPTY.
1005 * The IN_RENAME flag ensures that it cannot be moved by another
1010 panic("ext2_rename: lost dir entry");
1013 * If the source is a directory with a
1014 * new parent, the link count of the old
1015 * parent directory must be decremented
1016 * and ".." set to point to the new parent.
1018 if (doingdirectory && newparent) {
1020 dp->i_flag |= IN_CHANGE;
1021 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
1022 sizeof(struct dirtemplate), (off_t)0,
1023 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1024 tcnp->cn_cred, NOCRED, NULL, NULL);
1026 /* Like ufs little-endian: */
1027 namlen = dirbuf.dotdot_type;
1029 dirbuf.dotdot_name[0] != '.' ||
1030 dirbuf.dotdot_name[1] != '.') {
1031 ext2_dirbad(xp, (doff_t)12,
1032 "rename: mangled dir");
1034 dirbuf.dotdot_ino = newparent;
1035 (void) vn_rdwr(UIO_WRITE, fvp,
1037 sizeof(struct dirtemplate),
1038 (off_t)0, UIO_SYSSPACE,
1039 IO_NODELOCKED | IO_SYNC |
1040 IO_NOMACCHECK, tcnp->cn_cred,
1041 NOCRED, NULL, NULL);
1046 error = ext2_dirremove(fdvp, fcnp);
1049 xp->i_flag |= IN_CHANGE;
1051 xp->i_flag &= ~IN_RENAME;
1066 ip->i_flag &= ~IN_RENAME;
1067 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
1069 ip->i_flag |= IN_CHANGE;
1070 ip->i_flag &= ~IN_RENAME;
1081 ext2_mkdir(struct vop_mkdir_args *ap)
1083 struct vnode *dvp = ap->a_dvp;
1084 struct vattr *vap = ap->a_vap;
1085 struct componentname *cnp = ap->a_cnp;
1086 struct inode *ip, *dp;
1088 struct dirtemplate dirtemplate, *dtp;
1092 if ((cnp->cn_flags & HASBUF) == 0)
1093 panic("ext2_mkdir: no name");
1096 if ((nlink_t)dp->i_nlink >= EXT2_LINK_MAX) {
1100 dmode = vap->va_mode & 0777;
1103 * Must simulate part of ext2_makeinode here to acquire the inode,
1104 * but not have it entered in the parent directory. The entry is
1105 * made later after writing "." and ".." entries.
1107 error = ext2_valloc(dvp, dmode, cnp->cn_cred, &tvp);
1111 ip->i_gid = dp->i_gid;
1115 * if we are hacking owners here, (only do this where told to)
1116 * and we are not giving it TOO root, (would subvert quotas)
1117 * then go ahead and give it to the other user.
1118 * The new directory also inherits the SUID bit.
1119 * If user's UID and dir UID are the same,
1120 * 'give it away' so that the SUID is still forced on.
1122 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1123 (dp->i_mode & ISUID) && dp->i_uid) {
1125 ip->i_uid = dp->i_uid;
1127 ip->i_uid = cnp->cn_cred->cr_uid;
1131 ip->i_uid = cnp->cn_cred->cr_uid;
1133 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1135 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1137 if (cnp->cn_flags & ISWHITEOUT)
1138 ip->i_flags |= UF_OPAQUE;
1139 error = ext2_update(tvp, 1);
1142 * Bump link count in parent directory
1143 * to reflect work done below. Should
1144 * be done before reference is created
1145 * so reparation is possible if we crash.
1148 dp->i_flag |= IN_CHANGE;
1149 error = ext2_update(dvp, !DOINGASYNC(dvp));
1153 /* Initialize directory with "." and ".." from static template. */
1154 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs,
1155 EXT2F_INCOMPAT_FTYPE))
1156 dtp = &mastertemplate;
1158 dtp = &omastertemplate;
1160 dirtemplate.dot_ino = ip->i_number;
1161 dirtemplate.dotdot_ino = dp->i_number;
1162 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
1163 * so let's just redefine it - for this function only
1166 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->e2fs_bsize
1167 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
1168 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
1169 sizeof(dirtemplate), (off_t)0, UIO_SYSSPACE,
1170 IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
1174 dp->i_flag |= IN_CHANGE;
1177 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
1178 /* XXX should grow with balloc() */
1179 panic("ext2_mkdir: blksize");
1181 ip->i_size = DIRBLKSIZ;
1182 ip->i_flag |= IN_CHANGE;
1185 /* Directory set up, now install its entry in the parent directory. */
1186 error = ext2_direnter(ip, dvp, cnp);
1189 dp->i_flag |= IN_CHANGE;
1193 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
1194 * for us because we set the link count to 0.
1198 ip->i_flag |= IN_CHANGE;
1205 #define DIRBLKSIZ DEV_BSIZE
1209 * Rmdir system call.
1212 ext2_rmdir(struct vop_rmdir_args *ap)
1214 struct vnode *vp = ap->a_vp;
1215 struct vnode *dvp = ap->a_dvp;
1216 struct componentname *cnp = ap->a_cnp;
1217 struct inode *ip, *dp;
1224 * Verify the directory is empty (and valid).
1225 * (Rmdir ".." won't be valid since
1226 * ".." will contain a reference to
1227 * the current directory and thus be
1231 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1235 if ((dp->i_flags & APPEND)
1236 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1241 * Delete reference to directory before purging
1242 * inode. If we crash in between, the directory
1243 * will be reattached to lost+found,
1245 error = ext2_dirremove(dvp, cnp);
1249 dp->i_flag |= IN_CHANGE;
1253 * Truncate inode. The only stuff left
1254 * in the directory is "." and "..". The
1255 * "." reference is inconsequential since
1256 * we're quashing it. The ".." reference
1257 * has already been adjusted above. We've
1258 * removed the "." reference and the reference
1259 * in the parent directory, but there may be
1260 * other hard links so decrement by 2 and
1261 * worry about them later.
1264 error = ext2_truncate(vp, (off_t)0, IO_SYNC, cnp->cn_cred,
1266 cache_purge(ITOV(ip));
1267 if (vn_lock(dvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
1269 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
1270 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1277 * symlink -- make a symbolic link
1280 ext2_symlink(struct vop_symlink_args *ap)
1282 struct vnode *vp, **vpp = ap->a_vpp;
1286 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1291 len = strlen(ap->a_target);
1292 if (len < vp->v_mount->mnt_maxsymlinklen) {
1294 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1296 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1298 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1299 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1300 ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
1307 * Return target name of a symbolic link
1310 ext2_readlink(struct vop_readlink_args *ap)
1312 struct vnode *vp = ap->a_vp;
1313 struct inode *ip = VTOI(vp);
1317 if (isize < vp->v_mount->mnt_maxsymlinklen) {
1318 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1321 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1325 * Calculate the logical to physical mapping if not done already,
1326 * then call the device strategy routine.
1328 * In order to be able to swap to a file, the ext2_bmaparray() operation may not
1329 * deadlock on memory. See ext2_bmap() for details.
1332 ext2_strategy(struct vop_strategy_args *ap)
1334 struct buf *bp = ap->a_bp;
1335 struct vnode *vp = ap->a_vp;
1342 if (vp->v_type == VBLK || vp->v_type == VCHR)
1343 panic("ext2_strategy: spec");
1344 if (bp->b_blkno == bp->b_lblkno) {
1345 error = ext2_bmaparray(vp, bp->b_lblkno, &blkno, NULL, NULL);
1346 bp->b_blkno = blkno;
1348 bp->b_error = error;
1349 bp->b_ioflags |= BIO_ERROR;
1353 if ((long)bp->b_blkno == -1)
1356 if ((long)bp->b_blkno == -1) {
1360 bp->b_iooffset = dbtob(bp->b_blkno);
1361 bo = VFSTOEXT2(vp->v_mount)->um_bo;
1362 BO_STRATEGY(bo, bp);
1367 * Print out the contents of an inode.
1370 ext2_print(struct vop_print_args *ap)
1372 struct vnode *vp = ap->a_vp;
1373 struct inode *ip = VTOI(vp);
1375 vn_printf(ip->i_devvp, "\tino %lu", (u_long)ip->i_number);
1376 if (vp->v_type == VFIFO)
1383 * Close wrapper for fifos.
1385 * Update the times on the inode then do device close.
1388 ext2fifo_close(struct vop_close_args *ap)
1390 struct vnode *vp = ap->a_vp;
1393 if (vp->v_usecount > 1)
1394 ext2_itimes_locked(vp);
1396 return (fifo_specops.vop_close(ap));
1400 * Kqfilter wrapper for fifos.
1402 * Fall through to ext2 kqfilter routines if needed
1405 ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1409 error = fifo_specops.vop_kqfilter(ap);
1411 error = vfs_kqfilter(ap);
1416 * Return POSIX pathconf information applicable to ext2 filesystems.
1419 ext2_pathconf(struct vop_pathconf_args *ap)
1423 switch (ap->a_name) {
1425 *ap->a_retval = EXT2_LINK_MAX;
1428 *ap->a_retval = NAME_MAX;
1431 *ap->a_retval = PATH_MAX;
1434 *ap->a_retval = PIPE_BUF;
1436 case _PC_CHOWN_RESTRICTED:
1442 case _PC_MIN_HOLE_SIZE:
1443 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1446 /* _PC_ASYNC_IO should have been handled by upper layers. */
1447 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
1456 case _PC_ALLOC_SIZE_MIN:
1457 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
1459 case _PC_FILESIZEBITS:
1462 case _PC_REC_INCR_XFER_SIZE:
1463 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1465 case _PC_REC_MAX_XFER_SIZE:
1466 *ap->a_retval = -1; /* means ``unlimited'' */
1468 case _PC_REC_MIN_XFER_SIZE:
1469 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
1471 case _PC_REC_XFER_ALIGN:
1472 *ap->a_retval = PAGE_SIZE;
1474 case _PC_SYMLINK_MAX:
1475 *ap->a_retval = MAXPATHLEN;
1486 * Vnode pointer to File handle
1490 ext2_vptofh(struct vop_vptofh_args *ap)
1495 ip = VTOI(ap->a_vp);
1496 ufhp = (struct ufid *)ap->a_fhp;
1497 ufhp->ufid_len = sizeof(struct ufid);
1498 ufhp->ufid_ino = ip->i_number;
1499 ufhp->ufid_gen = ip->i_gen;
1504 * Initialize the vnode associated with a new inode, handle aliased
1508 ext2_vinit(struct mount *mntp, struct vop_vector *fifoops, struct vnode **vpp)
1515 vp->v_type = IFTOVT(ip->i_mode);
1516 if (vp->v_type == VFIFO)
1519 if (ip->i_number == EXT2_ROOTINO)
1520 vp->v_vflag |= VV_ROOT;
1521 ip->i_modrev = init_va_filerev();
1527 * Allocate a new inode.
1530 ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1531 struct componentname *cnp)
1533 struct inode *ip, *pdir;
1539 if ((cnp->cn_flags & HASBUF) == 0)
1540 panic("ext2_makeinode: no name");
1543 if ((mode & IFMT) == 0)
1546 error = ext2_valloc(dvp, mode, cnp->cn_cred, &tvp);
1551 ip->i_gid = pdir->i_gid;
1556 * not the owner of the directory,
1557 * and we are hacking owners here, (only do this where told to)
1558 * and we are not giving it TOO root, (would subvert quotas)
1559 * then go ahead and give it to the other user.
1560 * Note that this drops off the execute bits for security.
1562 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1563 (pdir->i_mode & ISUID) &&
1564 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1565 ip->i_uid = pdir->i_uid;
1568 ip->i_uid = cnp->cn_cred->cr_uid;
1572 ip->i_uid = cnp->cn_cred->cr_uid;
1574 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1576 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1578 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred)) {
1579 if (priv_check_cred(cnp->cn_cred, PRIV_VFS_RETAINSUGID, 0))
1580 ip->i_mode &= ~ISGID;
1583 if (cnp->cn_flags & ISWHITEOUT)
1584 ip->i_flags |= UF_OPAQUE;
1587 * Make sure inode goes to disk before directory entry.
1589 error = ext2_update(tvp, !DOINGASYNC(tvp));
1592 error = ext2_direnter(ip, dvp, cnp);
1601 * Write error occurred trying to update the inode
1602 * or the directory so must deallocate the inode.
1605 ip->i_flag |= IN_CHANGE;
1611 * Vnode op for reading.
1614 ext2_read(struct vop_read_args *ap)
1623 /*EXT4_EXT_LOCK(ip);*/
1624 if (ip->i_flags & EXT4_EXTENTS)
1625 error = ext4_ext_read(ap);
1627 error = ext2_ind_read(ap);
1628 /*EXT4_EXT_UNLOCK(ip);*/
1634 * Vnode op for reading.
1637 ext2_ind_read(struct vop_read_args *ap)
1642 struct m_ext2fs *fs;
1644 daddr_t lbn, nextlbn;
1646 long size, xfersize, blkoffset;
1647 int error, orig_resid, seqcount;
1652 ioflag = ap->a_ioflag;
1654 seqcount = ap->a_ioflag >> IO_SEQSHIFT;
1658 if (uio->uio_rw != UIO_READ)
1659 panic("%s: mode", "ext2_read");
1661 if (vp->v_type == VLNK) {
1662 if ((int)ip->i_size < vp->v_mount->mnt_maxsymlinklen)
1663 panic("%s: short symlink", "ext2_read");
1664 } else if (vp->v_type != VREG && vp->v_type != VDIR)
1665 panic("%s: type %d", "ext2_read", vp->v_type);
1667 orig_resid = uio->uio_resid;
1668 KASSERT(orig_resid >= 0, ("ext2_read: uio->uio_resid < 0"));
1669 if (orig_resid == 0)
1671 KASSERT(uio->uio_offset >= 0, ("ext2_read: uio->uio_offset < 0"));
1673 if (uio->uio_offset < ip->i_size &&
1674 uio->uio_offset >= fs->e2fs_maxfilesize)
1677 for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
1678 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
1680 lbn = lblkno(fs, uio->uio_offset);
1682 size = blksize(fs, ip, lbn);
1683 blkoffset = blkoff(fs, uio->uio_offset);
1685 xfersize = fs->e2fs_fsize - blkoffset;
1686 if (uio->uio_resid < xfersize)
1687 xfersize = uio->uio_resid;
1688 if (bytesinfile < xfersize)
1689 xfersize = bytesinfile;
1691 if (lblktosize(fs, nextlbn) >= ip->i_size)
1692 error = bread(vp, lbn, size, NOCRED, &bp);
1693 else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0)
1694 error = cluster_read(vp, ip->i_size, lbn, size,
1695 NOCRED, blkoffset + uio->uio_resid, seqcount, &bp);
1696 else if (seqcount > 1) {
1697 u_int nextsize = blksize(fs, ip, nextlbn);
1698 error = breadn(vp, lbn,
1699 size, &nextlbn, &nextsize, 1, NOCRED, &bp);
1701 error = bread(vp, lbn, size, NOCRED, &bp);
1709 * If IO_DIRECT then set B_DIRECT for the buffer. This
1710 * will cause us to attempt to release the buffer later on
1711 * and will cause the buffer cache to attempt to free the
1714 if (ioflag & IO_DIRECT)
1715 bp->b_flags |= B_DIRECT;
1718 * We should only get non-zero b_resid when an I/O error
1719 * has occurred, which should cause us to break above.
1720 * However, if the short read did not cause an error,
1721 * then we want to ensure that we do not uiomove bad
1722 * or uninitialized data.
1724 size -= bp->b_resid;
1725 if (size < xfersize) {
1730 error = uiomove((char *)bp->b_data + blkoffset,
1731 (int)xfersize, uio);
1735 if (ioflag & (IO_VMIO|IO_DIRECT)) {
1737 * If it's VMIO or direct I/O, then we don't
1738 * need the buf, mark it available for
1739 * freeing. If it's non-direct VMIO, the VM has
1742 bp->b_flags |= B_RELBUF;
1746 * Otherwise let whoever
1747 * made the request take care of
1748 * freeing it. We just queue
1749 * it onto another list.
1756 * This can only happen in the case of an error
1757 * because the loop above resets bp to NULL on each iteration
1758 * and on normal completion has not set a new value into it.
1759 * so it must have come from a 'break' statement
1762 if (ioflag & (IO_VMIO|IO_DIRECT)) {
1763 bp->b_flags |= B_RELBUF;
1770 if ((error == 0 || uio->uio_resid != orig_resid) &&
1771 (vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
1772 ip->i_flag |= IN_ACCESS;
1777 ext2_ioctl(struct vop_ioctl_args *ap)
1780 switch (ap->a_command) {
1783 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
1784 (off_t *)ap->a_data, ap->a_cred));
1791 * this function handles ext4 extents block mapping
1794 ext4_ext_read(struct vop_read_args *ap)
1799 struct m_ext2fs *fs;
1801 struct ext4_extent nex, *ep;
1802 struct ext4_extent_path path;
1803 daddr_t lbn, newblk;
1808 long size, xfersize, blkoffset;
1813 memset(&path, 0, sizeof(path));
1815 orig_resid = uio->uio_resid;
1816 KASSERT(orig_resid >= 0, ("%s: uio->uio_resid < 0", __func__));
1817 if (orig_resid == 0)
1819 KASSERT(uio->uio_offset >= 0, ("%s: uio->uio_offset < 0", __func__));
1821 if (uio->uio_offset < ip->i_size && uio->uio_offset >= fs->e2fs_maxfilesize)
1824 while (uio->uio_resid > 0) {
1825 if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
1827 lbn = lblkno(fs, uio->uio_offset);
1828 size = blksize(fs, ip, lbn);
1829 blkoffset = blkoff(fs, uio->uio_offset);
1831 xfersize = fs->e2fs_fsize - blkoffset;
1832 xfersize = MIN(xfersize, uio->uio_resid);
1833 xfersize = MIN(xfersize, bytesinfile);
1835 /* get block from ext4 extent cache */
1836 cache_type = ext4_ext_in_cache(ip, lbn, &nex);
1837 switch (cache_type) {
1838 case EXT4_EXT_CACHE_NO:
1839 ext4_ext_find_extent(fs, ip, lbn, &path);
1844 ext4_ext_put_cache(ip, ep, EXT4_EXT_CACHE_IN);
1846 newblk = lbn - ep->e_blk + (ep->e_start_lo |
1847 (daddr_t)ep->e_start_hi << 32);
1849 if (path.ep_bp != NULL) {
1855 case EXT4_EXT_CACHE_GAP:
1856 /* block has not been allocated yet */
1859 case EXT4_EXT_CACHE_IN:
1860 newblk = lbn - nex.e_blk + (nex.e_start_lo |
1861 (daddr_t)nex.e_start_hi << 32);
1865 panic("%s: invalid cache type", __func__);
1868 error = bread(ip->i_devvp, fsbtodb(fs, newblk), size, NOCRED, &bp);
1874 size -= bp->b_resid;
1875 if (size < xfersize) {
1882 error = uiomove(bp->b_data + blkoffset, (int)xfersize, uio);
1892 * Vnode op for writing.
1895 ext2_write(struct vop_write_args *ap)
1900 struct m_ext2fs *fs;
1904 int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
1906 ioflag = ap->a_ioflag;
1910 seqcount = ioflag >> IO_SEQSHIFT;
1914 if (uio->uio_rw != UIO_WRITE)
1915 panic("%s: mode", "ext2_write");
1918 switch (vp->v_type) {
1920 if (ioflag & IO_APPEND)
1921 uio->uio_offset = ip->i_size;
1922 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
1928 /* XXX differs from ffs -- this is called from ext2_mkdir(). */
1929 if ((ioflag & IO_SYNC) == 0)
1930 panic("ext2_write: nonsync dir write");
1933 panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
1934 vp->v_type, (intmax_t)uio->uio_offset,
1935 (intmax_t)uio->uio_resid);
1938 KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
1939 KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
1941 if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
1944 * Maybe this should be above the vnode op call, but so long as
1945 * file servers have no limits, I don't think it matters.
1947 if (vn_rlimit_fsize(vp, uio, uio->uio_td))
1950 resid = uio->uio_resid;
1952 if (seqcount > BA_SEQMAX)
1953 flags = BA_SEQMAX << BA_SEQSHIFT;
1955 flags = seqcount << BA_SEQSHIFT;
1956 if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
1959 for (error = 0; uio->uio_resid > 0;) {
1960 lbn = lblkno(fs, uio->uio_offset);
1961 blkoffset = blkoff(fs, uio->uio_offset);
1962 xfersize = fs->e2fs_fsize - blkoffset;
1963 if (uio->uio_resid < xfersize)
1964 xfersize = uio->uio_resid;
1965 if (uio->uio_offset + xfersize > ip->i_size)
1966 vnode_pager_setsize(vp, uio->uio_offset + xfersize);
1969 * We must perform a read-before-write if the transfer size
1970 * does not cover the entire buffer.
1972 if (fs->e2fs_bsize > xfersize)
1975 flags &= ~BA_CLRBUF;
1976 error = ext2_balloc(ip, lbn, blkoffset + xfersize,
1977 ap->a_cred, &bp, flags);
1981 if ((ioflag & (IO_SYNC|IO_INVAL)) == (IO_SYNC|IO_INVAL))
1982 bp->b_flags |= B_NOCACHE;
1983 if (uio->uio_offset + xfersize > ip->i_size)
1984 ip->i_size = uio->uio_offset + xfersize;
1985 size = blksize(fs, ip, lbn) - bp->b_resid;
1986 if (size < xfersize)
1990 uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
1992 * If the buffer is not already filled and we encounter an
1993 * error while trying to fill it, we have to clear out any
1994 * garbage data from the pages instantiated for the buffer.
1995 * If we do not, a failed uiomove() during a write can leave
1996 * the prior contents of the pages exposed to a userland mmap.
1998 * Note that we need only clear buffers with a transfer size
1999 * equal to the block size because buffers with a shorter
2000 * transfer size were cleared above by the call to ext2_balloc()
2001 * with the BA_CLRBUF flag set.
2003 * If the source region for uiomove identically mmaps the
2004 * buffer, uiomove() performed the NOP copy, and the buffer
2005 * content remains valid because the page fault handler
2006 * validated the pages.
2008 if (error != 0 && (bp->b_flags & B_CACHE) == 0 &&
2009 fs->e2fs_bsize == xfersize)
2011 if (ioflag & (IO_VMIO|IO_DIRECT)) {
2012 bp->b_flags |= B_RELBUF;
2016 * If IO_SYNC each buffer is written synchronously. Otherwise
2017 * if we have a severe page deficiency write the buffer
2018 * asynchronously. Otherwise try to cluster, and if that
2019 * doesn't do it then either do an async write (if O_DIRECT),
2020 * or a delayed write (if not).
2022 if (ioflag & IO_SYNC) {
2024 } else if (vm_page_count_severe() ||
2025 buf_dirty_count_severe() ||
2026 (ioflag & IO_ASYNC)) {
2027 bp->b_flags |= B_CLUSTEROK;
2029 } else if (xfersize + blkoffset == fs->e2fs_fsize) {
2030 if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
2031 bp->b_flags |= B_CLUSTEROK;
2032 cluster_write(vp, bp, ip->i_size, seqcount);
2036 } else if (ioflag & IO_DIRECT) {
2037 bp->b_flags |= B_CLUSTEROK;
2040 bp->b_flags |= B_CLUSTEROK;
2043 if (error || xfersize == 0)
2047 * If we successfully wrote any data, and we are not the superuser
2048 * we clear the setuid and setgid bits as a precaution against
2051 if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid &&
2053 if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
2054 ip->i_mode &= ~(ISUID | ISGID);
2057 if (ioflag & IO_UNIT) {
2058 (void)ext2_truncate(vp, osize,
2059 ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
2060 uio->uio_offset -= resid - uio->uio_resid;
2061 uio->uio_resid = resid;
2064 if (uio->uio_resid != resid) {
2065 ip->i_flag |= IN_CHANGE | IN_UPDATE;
2066 if (ioflag & IO_SYNC)
2067 error = ext2_update(vp, 1);