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. All advertising materials mentioning features or use of this software
25 * must display the following acknowledgement:
26 * This product includes software developed by the University of
27 * California, Berkeley and its contributors.
28 * 4. Neither the name of the University nor the names of its contributors
29 * may be used to endorse or promote products derived from this software
30 * without specific prior written permission.
32 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
33 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
34 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
35 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
36 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
37 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
38 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
39 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
40 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
41 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
44 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
45 * @(#)ext2_vnops.c 8.7 (Berkeley) 2/3/94
46 * $FreeBSD: src/sys/gnu/ext2fs/ext2_vnops.c,v 1.51.2.2 2003/01/02 17:26:18 bde Exp $
47 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vnops.c,v 1.31 2006/04/28 16:45:34 swildner Exp $
50 #include "opt_quota.h"
51 #include "opt_suiddir.h"
53 #include <sys/param.h>
54 #include <sys/systm.h>
55 #include <sys/resourcevar.h>
56 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
62 #include <sys/mount.h>
64 #include <sys/vnode.h>
65 #include <sys/dirent.h>
66 #include <sys/lockf.h>
67 #include <sys/event.h>
71 #include <sys/namei.h>
72 #include <sys/signalvar.h>
73 #include <sys/unistd.h>
76 #include <vm/vm_extern.h>
77 #include <vm/vm_zone.h>
78 #include <vm/vnode_pager.h>
81 #include <sys/thread2.h>
83 #include <vfs/fifofs/fifo.h>
88 #include "ext2mount.h"
89 #include "ext2_fs_sb.h"
91 #include "ext2_extern.h"
94 static int ext2_access (struct vop_access_args *);
95 static int ext2_advlock (struct vop_advlock_args *);
96 static int ext2_chmod (struct vnode *, int, struct ucred *, struct thread *);
97 static int ext2_chown (struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
98 static int ext2_close (struct vop_close_args *);
99 static int ext2_getattr (struct vop_getattr_args *);
100 static int ext2_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
101 static int ext2_mmap (struct vop_mmap_args *);
102 static int ext2_open (struct vop_open_args *);
103 static int ext2_pathconf (struct vop_pathconf_args *);
104 static int ext2_print (struct vop_print_args *);
105 static int ext2_readlink (struct vop_readlink_args *);
106 static int ext2_setattr (struct vop_setattr_args *);
107 static int ext2_strategy (struct vop_strategy_args *);
108 static int ext2_whiteout (struct vop_old_whiteout_args *);
109 static int filt_ext2read (struct knote *kn, long hint);
110 static int filt_ext2write (struct knote *kn, long hint);
111 static int filt_ext2vnode (struct knote *kn, long hint);
112 static void filt_ext2detach (struct knote *kn);
113 static int ext2_kqfilter (struct vop_kqfilter_args *ap);
114 static int ext2spec_close (struct vop_close_args *);
115 static int ext2spec_read (struct vop_read_args *);
116 static int ext2spec_write (struct vop_write_args *);
117 static int ext2fifo_close (struct vop_close_args *);
118 static int ext2fifo_kqfilter (struct vop_kqfilter_args *);
119 static int ext2fifo_read (struct vop_read_args *);
120 static int ext2fifo_write (struct vop_write_args *);
122 static int ext2_fsync (struct vop_fsync_args *);
123 static int ext2_read (struct vop_read_args *);
124 static int ext2_write (struct vop_write_args *);
125 static int ext2_remove (struct vop_old_remove_args *);
126 static int ext2_link (struct vop_old_link_args *);
127 static int ext2_rename (struct vop_old_rename_args *);
128 static int ext2_mkdir (struct vop_old_mkdir_args *);
129 static int ext2_rmdir (struct vop_old_rmdir_args *);
130 static int ext2_create (struct vop_old_create_args *);
131 static int ext2_mknod (struct vop_old_mknod_args *);
132 static int ext2_symlink (struct vop_old_symlink_args *);
133 static int ext2_getpages (struct vop_getpages_args *);
134 static int ext2_putpages (struct vop_putpages_args *);
136 #include "ext2_readwrite.c"
142 #define SETHIGH(q, h) { \
145 tmp.val[_QUAD_HIGHWORD] = (h); \
148 #define SETLOW(q, l) { \
151 tmp.val[_QUAD_LOWWORD] = (l); \
154 #define VN_KNOTE(vp, b) \
155 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
157 #define OFSFMT(vp) ((vp)->v_mount->mnt_maxsymlinklen <= 0)
160 * A virgin directory (no blushing please).
161 * Note that the type and namlen fields are reversed relative to ufs.
162 * Also, we don't use `struct odirtemplate', since it would just cause
163 * endianness problems.
165 static struct dirtemplate ext2_mastertemplate = {
166 0, 12, 1, EXT2_FT_DIR, ".",
167 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
169 static struct dirtemplate ext2_omastertemplate = {
170 0, 12, 1, EXT2_FT_UNKNOWN, ".",
171 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
175 * Create a regular file
177 * ext2_create(struct vnode *a_dvp, struct vnode **a_vpp,
178 * struct componentname *a_cnp, struct vattr *a_vap)
181 ext2_create(struct vop_old_create_args *ap)
186 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
187 ap->a_dvp, ap->a_vpp, ap->a_cnp);
194 * Synch an open file.
196 * ext2_fsync(struct vnode *a_vp, struct ucred *a_cred, int a_waitfor,
201 static int ext2_fsync_bp(struct buf *bp, void *data);
203 struct ext2_fsync_bp_info {
209 ext2_fsync(struct vop_fsync_args *ap)
211 struct ext2_fsync_bp_info info;
212 struct vnode *vp = ap->a_vp;
216 * XXX why is all this fs specific?
220 * Flush all dirty buffers associated with a vnode.
222 ext2_discard_prealloc(VTOI(vp));
227 info.waitfor = ap->a_waitfor;
228 count = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
229 ext2_fsync_bp, &info);
233 if (ap->a_waitfor == MNT_WAIT) {
234 while (vp->v_track_write.bk_active) {
235 vp->v_track_write.bk_waitflag = 1;
236 tsleep(&vp->v_track_write, 0, "e2fsyn", 0);
239 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
240 vprint("ext2_fsync: dirty", vp);
246 return (EXT2_UPDATE(ap->a_vp, ap->a_waitfor == MNT_WAIT));
250 ext2_fsync_bp(struct buf *bp, void *data)
252 struct ext2_fsync_bp_info *info = data;
254 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
256 if ((bp->b_flags & B_DELWRI) == 0)
257 panic("ext2_fsync: not dirty");
262 * Wait for I/O associated with indirect blocks to complete,
263 * since there is no way to quickly wait for them below.
265 if (bp->b_vp == info->vp || info->waitfor == MNT_NOWAIT)
276 * ext2_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
277 * struct componentname *a_cnp, struct vattr *a_vap)
281 ext2_mknod(struct vop_old_mknod_args *ap)
283 struct vattr *vap = ap->a_vap;
284 struct vnode **vpp = ap->a_vpp;
289 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
290 ap->a_dvp, vpp, ap->a_cnp);
294 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
295 if (vap->va_rdev != VNOVAL) {
297 * Want to be able to use this to make badblock
298 * inodes, so don't truncate the dev number.
300 ip->i_rdev = vap->va_rdev;
303 * Remove inode, then reload it through VFS_VGET so it is
304 * checked to see if it is an alias of an existing entry in
307 (*vpp)->v_type = VNON;
308 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
311 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
320 * ext2_remove(struct vnode *a_dvp, struct vnode *a_vp,
321 * struct componentname *a_cnp)
324 ext2_remove(struct vop_old_remove_args *ap)
327 struct vnode *vp = ap->a_vp;
328 struct vnode *dvp = ap->a_dvp;
332 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
333 (VTOI(dvp)->i_flags & APPEND)) {
337 error = ext2_dirremove(dvp, ap->a_cnp);
340 ip->i_flag |= IN_CHANGE;
349 * ext2_link(struct vnode *a_tdvp, struct vnode *a_vp,
350 * struct componentname *a_cnp)
353 ext2_link(struct vop_old_link_args *ap)
355 struct vnode *vp = ap->a_vp;
356 struct vnode *tdvp = ap->a_tdvp;
357 struct componentname *cnp = ap->a_cnp;
358 struct thread *td = cnp->cn_td;
362 if (tdvp->v_mount != vp->v_mount) {
366 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, td))) {
370 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
374 if (ip->i_flags & (IMMUTABLE | APPEND)) {
379 ip->i_flag |= IN_CHANGE;
380 error = EXT2_UPDATE(vp, 1);
382 error = ext2_direnter(ip, tdvp, cnp);
385 ip->i_flag |= IN_CHANGE;
389 VOP_UNLOCK(vp, 0, td);
395 * Rename system call. fdvp, fvp are ref'd. tvp, tdvp are ref'd and locked.
396 * all vp's are released and must be in an unlocked state on return.
398 * ext2_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
399 * struct componentname *a_fcnp, struct vnode *a_tdvp,
400 * struct vnode *a_tvp, struct componentname *a_tcnp)
403 ext2_rename(struct vop_old_rename_args *ap)
405 struct vnode *tvp = ap->a_tvp;
406 struct vnode *tdvp = ap->a_tdvp;
407 struct vnode *fvp = ap->a_fvp;
408 struct vnode *fdvp = ap->a_fdvp;
409 struct componentname *tcnp = ap->a_tcnp;
410 struct componentname *fcnp = ap->a_fcnp;
411 struct thread *td = fcnp->cn_td;
412 struct inode *ip, *xp, *dp;
413 struct dirtemplate dirbuf;
414 int doingdirectory = 0, oldparent = 0, newparent = 0;
419 * Check for cross-device rename.
421 if ((fvp->v_mount != tdvp->v_mount) ||
422 (tvp && (fvp->v_mount != tvp->v_mount)) ||
437 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
438 (VTOI(tdvp)->i_flags & APPEND))) {
444 * Renaming a file to itself has no effect. The upper layers should
445 * not call us in that case. Temporarily just warn if they do.
452 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
456 * fvp, tvp, tdvp locked. fdvp not locked but note that fdvp may
461 if (ip->i_nlink >= LINK_MAX) {
462 VOP_UNLOCK(fvp, 0, td);
466 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
467 || (dp->i_flags & APPEND)) {
468 VOP_UNLOCK(fvp, 0, td);
472 if ((ip->i_mode & IFMT) == IFDIR) {
474 * Avoid ".", "..", and aliases of "." for obvious reasons.
476 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
477 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
478 (ip->i_flag & IN_RENAME)) {
479 VOP_UNLOCK(fvp, 0, td);
483 ip->i_flag |= IN_RENAME;
484 oldparent = dp->i_number;
489 * tvp is non-NULL if the target exists. fvp is still locked but
490 * we will unlock it soon. The 'bad' goto target requires dp and
491 * xp to be correctly assigned.
500 * 1) Bump link count while we're moving stuff
501 * around. If we crash somewhere before
502 * completing our work, the link count
503 * may be wrong, but correctable.
506 ip->i_flag |= IN_CHANGE;
507 if ((error = EXT2_UPDATE(fvp, 1)) != 0) {
508 VOP_UNLOCK(fvp, 0, td);
513 * If ".." must be changed (ie the directory gets a new
514 * parent) then the source directory must not be in the
515 * directory heirarchy above the target, as this would
516 * orphan everything below the source directory. Also
517 * the user must have write permission in the source so
518 * as to be able to change "..". We must repeat the call
519 * to namei, as the parent directory is unlocked by the
520 * call to checkpath().
522 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
523 VOP_UNLOCK(fvp, 0, td);
526 * tvp (if not NULL) and tdvp are locked. fvp and fdvp are not.
527 * dp and xp are set according to tdvp and tvp.
529 if (oldparent != dp->i_number)
530 newparent = dp->i_number;
531 if (doingdirectory && newparent) {
532 if (error) /* write access check above */
536 * Prepare for relookup, get rid of xp
544 * checkpath vput()'s tdvp (VTOI(dp)) on return no matter what,
545 * get an extra ref so we wind up with just an unlocked, ref'd
546 * tdvp. The 'out' target skips xp and tdvp cleanups. Our
547 * tdvp is now unlocked so we have to clean it up ourselves.
550 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
551 tcnp->cn_flags |= CNP_PDIRUNLOCK;
557 * relookup no longer messes with the ref count. An unlocked
558 * tdvp must be passed and if no error occurs a locked tdvp
559 * will be returned. We have to use the out target again.
561 error = relookup(tdvp, &tvp, tcnp);
563 if (tcnp->cn_flags & CNP_PDIRUNLOCK)
571 * tdvp is locked at this point. in the RENAME case tvp may
572 * be NULL without an error, assign xp accordingly. The
573 * 'bad' target can be used again after this.
580 * 2) If target doesn't exist, link the target
581 * to the source and unlink the source.
582 * Otherwise, rewrite the target directory
583 * entry to reference the source inode and
584 * expunge the original entry's existence.
586 * tdvp and tvp are cleaned up by this code. tvp is only good if
590 if (dp->i_dev != ip->i_dev)
591 panic("ext2_rename: EXDEV");
593 * Account for ".." in new directory.
594 * When source and destination have the same
595 * parent we don't fool with the link count.
597 if (doingdirectory && newparent) {
598 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
603 dp->i_flag |= IN_CHANGE;
604 error = EXT2_UPDATE(tdvp, 1);
608 error = ext2_direnter(ip, tdvp, tcnp);
610 if (doingdirectory && newparent) {
612 dp->i_flag |= IN_CHANGE;
613 EXT2_UPDATE(tdvp, 1);
619 * manual cleanup, we can't use the bad or out target after
624 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
625 panic("ext2_rename: EXDEV");
627 * Short circuit rename(foo, foo).
629 if (xp->i_number == ip->i_number)
630 panic("ext2_rename: same file");
632 * If the parent directory is "sticky", then the user must
633 * own the parent directory, or the destination of the rename,
634 * otherwise the destination may not be changed (except by
635 * root). This implements append-only directories.
637 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
638 tcnp->cn_cred->cr_uid != dp->i_uid &&
639 xp->i_uid != tcnp->cn_cred->cr_uid) {
644 * Target must be empty if a directory and have no links
645 * to it. Also, ensure source and target are compatible
646 * (both directories, or both not directories).
648 if ((xp->i_mode&IFMT) == IFDIR) {
649 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
654 if (!doingdirectory) {
658 } else if (doingdirectory) {
662 error = ext2_dirrewrite(dp, ip, tcnp);
666 * If the target directory is in the same
667 * directory as the source directory,
668 * decrement the link count on the parent
669 * of the target directory.
671 if (doingdirectory && !newparent) {
673 dp->i_flag |= IN_CHANGE;
677 * manual cleanup, we can't use the bad or out target after
683 * Adjust the link count of the target to
684 * reflect the dirrewrite above. If this is
685 * a directory it is empty and there are
686 * no links to it, so we can squash the inode and
687 * any space associated with it. We disallowed
688 * renaming over top of a directory with links to
689 * it above, as the remaining link would point to
690 * a directory without "." or ".." entries.
693 if (doingdirectory) {
694 if (--xp->i_nlink != 0)
695 panic("ext2_rename: linked directory");
696 error = EXT2_TRUNCATE(tvp, (off_t)0, IO_SYNC,
697 tcnp->cn_cred, tcnp->cn_td);
699 xp->i_flag |= IN_CHANGE;
705 * tvp and tdvp have been cleaned up. The bad and out targets may
706 * not be used. fvp and fdvp are ref'd but not locked. ip
707 * still represents the old fvp and ip->i_flag may still have IN_RENAME
708 * set (if doingdirectory).
712 * 3) Unlink the source.
714 * fdvp is locked and ref'd. ap->a_fvp holds the old lookup unlocked
715 * and ref'd, fvp will hold the new lookup locked and ref'd.
717 * After the relookup ap->a_fvp must be released as part of our
718 * cleanup, not just fdvp and fvp. And, on success, fdvp and
719 * fvp will be locked so the bad and out targets cannot be used.
721 fcnp->cn_flags &= ~CNP_MODMASK;
722 fcnp->cn_flags |= CNP_LOCKPARENT;
723 KKASSERT(fcnp->cn_flags & CNP_PDIRUNLOCK);
724 error = relookup(fdvp, &fvp, fcnp);
727 * From name has disappeared.
730 panic("ext2_rename: lost dir entry");
731 /* ip->i_flag only sets IN_RENAME if doingdirectory */
733 if (fcnp->cn_flags & CNP_PDIRUNLOCK)
739 KKASSERT((fcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
742 * This case shouldn't occur
746 * From name has disappeared.
749 panic("ext2_rename: lost dir entry");
750 /* ip->i_flag only sets IN_RENAME if doingdirectory */
758 * fvp and fdvp are both ref'd and locked.
764 * Ensure that the directory entry still exists and has not
765 * changed while the new name has been entered. If the source is
766 * a file then the entry may have been unlinked or renamed. In
767 * either case there is no further work to be done. If the source
768 * is a directory then it cannot have been rmdir'ed; its link
769 * count of three would cause a rmdir to fail with ENOTEMPTY.
770 * The IN_RENAME flag ensures that it cannot be moved by another
775 panic("ext2_rename: lost dir entry");
776 /* ip->i_flag only sets IN_RENAME if doingdirectory */
779 * If the source is a directory with a
780 * new parent, the link count of the old
781 * parent directory must be decremented
782 * and ".." set to point to the new parent.
784 if (doingdirectory && newparent) {
786 dp->i_flag |= IN_CHANGE;
787 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
788 sizeof (struct dirtemplate), (off_t)0,
789 UIO_SYSSPACE, IO_NODELOCKED,
790 tcnp->cn_cred, (int *)0, NULL);
792 /* Like ext2 little-endian: */
793 namlen = dirbuf.dotdot_type;
795 dirbuf.dotdot_name[0] != '.' ||
796 dirbuf.dotdot_name[1] != '.') {
797 ext2_dirbad(xp, (doff_t)12,
798 "rename: mangled dir");
800 dirbuf.dotdot_ino = newparent;
801 vn_rdwr(UIO_WRITE, fvp,
803 sizeof (struct dirtemplate),
804 (off_t)0, UIO_SYSSPACE,
805 IO_NODELOCKED|IO_SYNC,
806 tcnp->cn_cred, (int *)0,
811 error = ext2_dirremove(fdvp, fcnp);
814 xp->i_flag |= IN_CHANGE;
816 xp->i_flag &= ~IN_RENAME;
830 ip->i_flag &= ~IN_RENAME;
831 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
833 ip->i_flag |= IN_CHANGE;
834 ip->i_flag &= ~IN_RENAME;
845 * ext2_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
846 * struct componentname *a_cnp, struct vattr *a_vap)
849 ext2_mkdir(struct vop_old_mkdir_args *ap)
851 struct vnode *dvp = ap->a_dvp;
852 struct vattr *vap = ap->a_vap;
853 struct componentname *cnp = ap->a_cnp;
854 struct inode *ip, *dp;
856 struct dirtemplate dirtemplate, *dtp;
860 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
864 dmode = vap->va_mode & 0777;
867 * Must simulate part of ext2_makeinode here to acquire the inode,
868 * but not have it entered in the parent directory. The entry is
869 * made later after writing "." and ".." entries.
871 error = EXT2_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
875 ip->i_gid = dp->i_gid;
879 struct ucred ucred, *ucp;
883 * if we are hacking owners here, (only do this where told to)
884 * and we are not giving it TOO root, (would subvert quotas)
885 * then go ahead and give it to the other user.
886 * The new directory also inherits the SUID bit.
887 * If user's UID and dir UID are the same,
888 * 'give it away' so that the SUID is still forced on.
890 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
891 (dp->i_mode & ISUID) && dp->i_uid) {
893 ip->i_uid = dp->i_uid;
895 if (dp->i_uid != cnp->cn_cred->cr_uid) {
897 * make sure the correct user gets charged
899 * Make a dummy credential for the victim.
900 * XXX This seems to never be accessed out of
901 * our context so a stack variable is ok.
904 ucred.cr_uid = ip->i_uid;
905 ucred.cr_ngroups = 1;
906 ucred.cr_groups[0] = dp->i_gid;
911 ip->i_uid = cnp->cn_cred->cr_uid;
914 if ((error = ext2_getinoquota(ip)) ||
915 (error = ext2_chkiq(ip, 1, ucp, 0))) {
916 EXT2_VFREE(tvp, ip->i_number, dmode);
923 ip->i_uid = cnp->cn_cred->cr_uid;
925 if ((error = ext2_getinoquota(ip)) ||
926 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
927 EXT2_VFREE(tvp, ip->i_number, dmode);
933 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
935 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
937 if (cnp->cn_flags & CNP_ISWHITEOUT)
938 ip->i_flags |= UF_OPAQUE;
939 error = EXT2_UPDATE(tvp, 1);
942 * The vnode must have a VM object in order to issue buffer cache
948 * Bump link count in parent directory
949 * to reflect work done below. Should
950 * be done before reference is created
951 * so reparation is possible if we crash.
954 dp->i_flag |= IN_CHANGE;
955 error = EXT2_UPDATE(dvp, 1);
959 /* Initialize directory with "." and ".." from static template. */
960 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
961 EXT2_FEATURE_INCOMPAT_FILETYPE))
962 dtp = &ext2_mastertemplate;
964 dtp = &ext2_omastertemplate;
966 dirtemplate.dot_ino = ip->i_number;
967 dirtemplate.dotdot_ino = dp->i_number;
968 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
969 * so let's just redefine it - for this function only
972 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->s_blocksize
973 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
974 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
975 sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
976 IO_NODELOCKED|IO_SYNC, cnp->cn_cred, (int *)0, NULL);
979 dp->i_flag |= IN_CHANGE;
982 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
983 panic("ext2_mkdir: blksize"); /* XXX should grow with balloc() */
985 ip->i_size = DIRBLKSIZ;
986 ip->i_flag |= IN_CHANGE;
989 /* Directory set up, now install its entry in the parent directory. */
990 error = ext2_direnter(ip, dvp, cnp);
993 dp->i_flag |= IN_CHANGE;
997 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
998 * for us because we set the link count to 0.
1002 ip->i_flag |= IN_CHANGE;
1009 #define DIRBLKSIZ DEV_BSIZE
1013 * Rmdir system call.
1015 * ext2_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1016 * struct componentname *a_cnp)
1019 ext2_rmdir(struct vop_old_rmdir_args *ap)
1021 struct vnode *vp = ap->a_vp;
1022 struct vnode *dvp = ap->a_dvp;
1023 struct componentname *cnp = ap->a_cnp;
1024 struct thread *td = cnp->cn_td;
1025 struct inode *ip, *dp;
1032 * Verify the directory is empty (and valid).
1033 * (Rmdir ".." won't be valid since
1034 * ".." will contain a reference to
1035 * the current directory and thus be
1039 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1043 if ((dp->i_flags & APPEND)
1044 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1049 * Delete reference to directory before purging
1050 * inode. If we crash in between, the directory
1051 * will be reattached to lost+found,
1053 error = ext2_dirremove(dvp, cnp);
1057 dp->i_flag |= IN_CHANGE;
1058 VOP_UNLOCK(dvp, 0, td);
1060 * Truncate inode. The only stuff left
1061 * in the directory is "." and "..". The
1062 * "." reference is inconsequential since
1063 * we're quashing it. The ".." reference
1064 * has already been adjusted above. We've
1065 * removed the "." reference and the reference
1066 * in the parent directory, but there may be
1067 * other hard links so decrement by 2 and
1068 * worry about them later.
1071 error = EXT2_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred, td);
1072 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
1078 * symlink -- make a symbolic link
1080 * ext2_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1081 * struct componentname *a_cnp, struct vattr *a_vap,
1085 ext2_symlink(struct vop_old_symlink_args *ap)
1087 struct vnode *vp, **vpp = ap->a_vpp;
1091 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1096 len = strlen(ap->a_target);
1097 if (len < vp->v_mount->mnt_maxsymlinklen) {
1099 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1101 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1104 * Make sure we have a VM object in order to use
1107 if (vp->v_object == NULL)
1110 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1111 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred, (int *)0,
1122 * Allocate a new inode.
1125 ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1126 struct componentname *cnp)
1128 struct inode *ip, *pdir;
1134 if ((mode & IFMT) == 0)
1137 error = EXT2_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
1142 ip->i_gid = pdir->i_gid;
1146 struct ucred ucred, *ucp;
1151 * not the owner of the directory,
1152 * and we are hacking owners here, (only do this where told to)
1153 * and we are not giving it TOO root, (would subvert quotas)
1154 * then go ahead and give it to the other user.
1155 * Note that this drops off the execute bits for security.
1157 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1158 (pdir->i_mode & ISUID) &&
1159 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1160 ip->i_uid = pdir->i_uid;
1164 * make sure the correct user gets charged
1166 * Quickly knock up a dummy credential for the victim.
1167 * XXX This seems to never be accessed out of our
1168 * context so a stack variable is ok.
1171 ucred.cr_uid = ip->i_uid;
1172 ucred.cr_ngroups = 1;
1173 ucred.cr_groups[0] = pdir->i_gid;
1177 ip->i_uid = cnp->cn_cred->cr_uid;
1181 if ((error = ext2_getinoquota(ip)) ||
1182 (error = ext2_chkiq(ip, 1, ucp, 0))) {
1183 EXT2_VFREE(tvp, ip->i_number, mode);
1190 ip->i_uid = cnp->cn_cred->cr_uid;
1192 if ((error = ext2_getinoquota(ip)) ||
1193 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
1194 EXT2_VFREE(tvp, ip->i_number, mode);
1200 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1202 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1204 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
1205 suser_cred(cnp->cn_cred, PRISON_ROOT))
1206 ip->i_mode &= ~ISGID;
1208 if (cnp->cn_flags & CNP_ISWHITEOUT)
1209 ip->i_flags |= UF_OPAQUE;
1212 * Regular files and directories need VM objects. Softlinks do
1213 * not (not immediately anyway).
1215 if (tvp->v_type == VREG || tvp->v_type == VDIR)
1219 * Make sure inode goes to disk before directory entry.
1221 error = EXT2_UPDATE(tvp, 1);
1224 error = ext2_direnter(ip, dvp, cnp);
1233 * Write error occurred trying to update the inode
1234 * or the directory so must deallocate the inode.
1237 ip->i_flag |= IN_CHANGE;
1245 * XXX By default, wimp out... note that a_offset is ignored (and always
1249 ext2_getpages(struct vop_getpages_args *ap)
1251 return (vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
1258 * XXX By default, wimp out... note that a_offset is ignored (and always
1262 ext2_putpages(struct vop_putpages_args *ap)
1264 return (vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
1265 ap->a_sync, ap->a_rtvals));
1269 ext2_itimes(struct vnode *vp)
1275 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
1277 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
1278 ip->i_flag |= IN_LAZYMOD;
1280 ip->i_flag |= IN_MODIFIED;
1281 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
1283 if (ip->i_flag & IN_ACCESS) {
1284 ip->i_atime = ts.tv_sec;
1285 ip->i_atimensec = ts.tv_nsec;
1287 if (ip->i_flag & IN_UPDATE) {
1288 ip->i_mtime = ts.tv_sec;
1289 ip->i_mtimensec = ts.tv_nsec;
1292 if (ip->i_flag & IN_CHANGE) {
1293 ip->i_ctime = ts.tv_sec;
1294 ip->i_ctimensec = ts.tv_nsec;
1297 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
1305 * ext2_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1306 * struct thread *a_td)
1311 ext2_open(struct vop_open_args *ap)
1313 struct vnode *vp = ap->a_vp;
1316 * Files marked append-only must be opened for appending.
1318 if ((VTOI(vp)->i_flags & APPEND) &&
1319 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) {
1322 return (vop_stdopen(ap));
1328 * Update the times on the inode.
1330 * ext2_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1331 * struct thread *a_td)
1336 ext2_close(struct vop_close_args *ap)
1338 struct vnode *vp = ap->a_vp;
1340 if (vp->v_usecount > 1)
1342 return (vop_stdclose(ap));
1346 * ext2_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1347 * struct thread *a_td)
1351 ext2_access(struct vop_access_args *ap)
1353 struct vnode *vp = ap->a_vp;
1354 struct inode *ip = VTOI(vp);
1355 struct ucred *cred = ap->a_cred;
1356 mode_t mask, mode = ap->a_mode;
1364 * Disallow write attempts on read-only filesystems;
1365 * unless the file is a socket, fifo, or a block or
1366 * character device resident on the filesystem.
1368 if (mode & VWRITE) {
1369 switch (vp->v_type) {
1373 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1376 if ((error = ext2_getinoquota(ip)) != 0)
1385 /* If immutable bit set, nobody gets to write it. */
1386 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
1389 /* Otherwise, user id 0 always gets access. */
1390 if (cred->cr_uid == 0)
1395 /* Otherwise, check the owner. */
1396 if (cred->cr_uid == ip->i_uid) {
1403 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1406 /* Otherwise, check the groups. */
1407 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
1408 if (ip->i_gid == *gp) {
1415 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1418 /* Otherwise, check everyone else. */
1425 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1429 * ext2_getattr(struct vnode *a_vp, struct vattr *a_vap,
1430 * struct thread *a_td)
1435 ext2_getattr(struct vop_getattr_args *ap)
1437 struct vnode *vp = ap->a_vp;
1438 struct inode *ip = VTOI(vp);
1439 struct vattr *vap = ap->a_vap;
1443 * Copy from inode table
1445 vap->va_fsid = dev2udev(ip->i_dev);
1446 vap->va_fileid = ip->i_number;
1447 vap->va_mode = ip->i_mode & ~IFMT;
1448 vap->va_nlink = VFSTOEXT2(vp->v_mount)->um_i_effnlink_valid ?
1449 ip->i_effnlink : ip->i_nlink;
1450 vap->va_uid = ip->i_uid;
1451 vap->va_gid = ip->i_gid;
1452 vap->va_rdev = ip->i_rdev;
1453 vap->va_size = ip->i_din.di_size;
1454 vap->va_atime.tv_sec = ip->i_atime;
1455 vap->va_atime.tv_nsec = ip->i_atimensec;
1456 vap->va_mtime.tv_sec = ip->i_mtime;
1457 vap->va_mtime.tv_nsec = ip->i_mtimensec;
1458 vap->va_ctime.tv_sec = ip->i_ctime;
1459 vap->va_ctime.tv_nsec = ip->i_ctimensec;
1460 vap->va_flags = ip->i_flags;
1461 vap->va_gen = ip->i_gen;
1462 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1463 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
1464 vap->va_type = IFTOVT(ip->i_mode);
1465 vap->va_filerev = ip->i_modrev;
1470 * Set attribute vnode op. called from several syscalls
1472 * ext2_setattr(struct vnode *a_vp, struct vattr *a_vap,
1473 * struct ucred *a_cred, struct thread *a_td)
1477 ext2_setattr(struct vop_setattr_args *ap)
1479 struct vattr *vap = ap->a_vap;
1480 struct vnode *vp = ap->a_vp;
1481 struct inode *ip = VTOI(vp);
1482 struct ucred *cred = ap->a_cred;
1486 * Check for unsettable attributes.
1488 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
1489 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
1490 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
1491 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
1494 if (vap->va_flags != VNOVAL) {
1495 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1497 if (cred->cr_uid != ip->i_uid &&
1498 (error = suser_cred(cred, PRISON_ROOT)))
1501 * Note that a root chflags becomes a user chflags when
1502 * we are jailed, unless the jail.chflags_allowed sysctl
1505 if (cred->cr_uid == 0 &&
1506 (!jailed(cred) || jail_chflags_allowed)) {
1508 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
1511 ip->i_flags = vap->va_flags;
1514 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
1515 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
1517 ip->i_flags &= SF_SETTABLE;
1518 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
1520 ip->i_flag |= IN_CHANGE;
1521 if (vap->va_flags & (IMMUTABLE | APPEND))
1524 if (ip->i_flags & (IMMUTABLE | APPEND))
1527 * Go through the fields and update iff not VNOVAL.
1529 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
1530 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1532 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
1535 if (vap->va_size != VNOVAL) {
1537 * Disallow write attempts on read-only filesystems;
1538 * unless the file is a socket, fifo, or a block or
1539 * character device resident on the filesystem.
1541 switch (vp->v_type) {
1546 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1552 if ((error = EXT2_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
1556 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1557 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1559 if (cred->cr_uid != ip->i_uid &&
1560 (error = suser_cred(cred, PRISON_ROOT)) &&
1561 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1562 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
1564 if (vap->va_atime.tv_sec != VNOVAL)
1565 ip->i_flag |= IN_ACCESS;
1566 if (vap->va_mtime.tv_sec != VNOVAL)
1567 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1569 if (vap->va_atime.tv_sec != VNOVAL) {
1570 ip->i_atime = vap->va_atime.tv_sec;
1571 ip->i_atimensec = vap->va_atime.tv_nsec;
1573 if (vap->va_mtime.tv_sec != VNOVAL) {
1574 ip->i_mtime = vap->va_mtime.tv_sec;
1575 ip->i_mtimensec = vap->va_mtime.tv_nsec;
1577 error = EXT2_UPDATE(vp, 0);
1582 if (vap->va_mode != (mode_t)VNOVAL) {
1583 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1585 error = ext2_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
1587 VN_KNOTE(vp, NOTE_ATTRIB);
1592 * Change the mode on a file.
1593 * Inode must be locked before calling.
1596 ext2_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
1598 struct inode *ip = VTOI(vp);
1601 if (cred->cr_uid != ip->i_uid) {
1602 error = suser_cred(cred, PRISON_ROOT);
1607 if (vp->v_type != VDIR && (mode & S_ISTXT))
1609 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
1612 ip->i_mode &= ~ALLPERMS;
1613 ip->i_mode |= (mode & ALLPERMS);
1614 ip->i_flag |= IN_CHANGE;
1619 * Perform chown operation on inode ip;
1620 * inode must be locked prior to call.
1623 ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
1626 struct inode *ip = VTOI(vp);
1635 if (uid == (uid_t)VNOVAL)
1637 if (gid == (gid_t)VNOVAL)
1640 * If we don't own the file, are trying to change the owner
1641 * of the file, or are not a member of the target group,
1642 * the caller must be superuser or the call fails.
1644 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
1645 (gid != ip->i_gid && !(cred->cr_gid == gid ||
1646 groupmember((gid_t)gid, cred)))) &&
1647 (error = suser_cred(cred, PRISON_ROOT)))
1652 if ((error = ext2_getinoquota(ip)) != 0)
1655 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1656 ip->i_dquot[USRQUOTA] = NODQUOT;
1659 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1660 ip->i_dquot[GRPQUOTA] = NODQUOT;
1662 change = ip->i_blocks;
1663 (void) ext2_chkdq(ip, -change, cred, CHOWN);
1664 (void) ext2_chkiq(ip, -1, cred, CHOWN);
1665 for (i = 0; i < MAXQUOTAS; i++) {
1666 ext2_dqrele(vp, ip->i_dquot[i]);
1667 ip->i_dquot[i] = NODQUOT;
1673 if ((error = ext2_getinoquota(ip)) == 0) {
1675 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1676 ip->i_dquot[USRQUOTA] = NODQUOT;
1679 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1680 ip->i_dquot[GRPQUOTA] = NODQUOT;
1682 if ((error = ext2_chkdq(ip, change, cred, CHOWN)) == 0) {
1683 if ((error = ext2_chkiq(ip, 1, cred, CHOWN)) == 0)
1686 (void)ext2_chkdq(ip, -change, cred, CHOWN|FORCE);
1688 for (i = 0; i < MAXQUOTAS; i++) {
1689 ext2_dqrele(vp, ip->i_dquot[i]);
1690 ip->i_dquot[i] = NODQUOT;
1695 if (ext2_getinoquota(ip) == 0) {
1697 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1698 ip->i_dquot[USRQUOTA] = NODQUOT;
1701 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1702 ip->i_dquot[GRPQUOTA] = NODQUOT;
1704 (void) ext2_chkdq(ip, change, cred, FORCE|CHOWN);
1705 (void) ext2_chkiq(ip, 1, cred, FORCE|CHOWN);
1706 (void) ext2_getinoquota(ip);
1710 if (ext2_getinoquota(ip))
1711 panic("ext2_chown: lost quota");
1713 ip->i_flag |= IN_CHANGE;
1714 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
1715 ip->i_mode &= ~(ISUID | ISGID);
1722 * NB Currently unsupported.
1724 * ext2_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
1725 * struct thread *a_td)
1730 ext2_mmap(struct vop_mmap_args *ap)
1736 * whiteout vnode call
1738 * ext2_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
1742 ext2_whiteout(struct vop_old_whiteout_args *ap)
1744 return (EOPNOTSUPP);
1748 * Return target name of a symbolic link
1750 * ext2_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1754 ext2_readlink(struct vop_readlink_args *ap)
1756 struct vnode *vp = ap->a_vp;
1757 struct inode *ip = VTOI(vp);
1761 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1762 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1763 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1766 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1770 * Calculate the logical to physical mapping if not done already,
1771 * then call the device strategy routine.
1773 * In order to be able to swap to a file, the VOP_BMAP operation may not
1774 * deadlock on memory. See ext2_bmap() for details.
1776 * ext2_strategy(struct vnode *a_vp, struct bio *a_bio)
1780 ext2_strategy(struct vop_strategy_args *ap)
1782 struct bio *bio = ap->a_bio;
1784 struct buf *bp = bio->bio_buf;
1785 struct vnode *vp = ap->a_vp;
1790 if (vp->v_type == VBLK || vp->v_type == VCHR)
1791 panic("ext2_strategy: spec");
1792 nbio = push_bio(bio);
1793 if (nbio->bio_offset == NOOFFSET) {
1794 error = VOP_BMAP(vp, bio->bio_offset, NULL, &nbio->bio_offset,
1797 bp->b_error = error;
1798 bp->b_flags |= B_ERROR;
1799 /* I/O was never started on nbio, must biodone(bio) */
1803 if (nbio->bio_offset == NOOFFSET)
1806 if (nbio->bio_offset == NOOFFSET) {
1807 /* I/O was never started on nbio, must biodone(bio) */
1811 vn_strategy(ip->i_devvp, nbio);
1816 * Print out the contents of an inode.
1818 * ext2_print(struct vnode *a_vp)
1822 ext2_print(struct vop_print_args *ap)
1824 struct vnode *vp = ap->a_vp;
1825 struct inode *ip = VTOI(vp);
1827 printf("tag VT_EXT2FS, ino %lu, on dev %s (%d, %d)",
1828 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1830 if (vp->v_type == VFIFO)
1832 lockmgr_printinfo(&vp->v_lock);
1838 * Read wrapper for special devices.
1840 * ext2spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1841 * struct ucred *a_cred)
1845 ext2spec_read(struct vop_read_args *ap)
1852 resid = uio->uio_resid;
1853 error = VOCALL(spec_vnode_vops, &ap->a_head);
1855 * The inode may have been revoked during the call, so it must not
1856 * be accessed blindly here or in the other wrapper functions.
1858 ip = VTOI(ap->a_vp);
1859 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1860 ip->i_flag |= IN_ACCESS;
1865 * Write wrapper for special devices.
1867 * ext2spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1868 * struct ucred *a_cred)
1872 ext2spec_write(struct vop_write_args *ap)
1879 resid = uio->uio_resid;
1880 error = VOCALL(spec_vnode_vops, &ap->a_head);
1881 ip = VTOI(ap->a_vp);
1882 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1883 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1888 * Close wrapper for special devices.
1890 * Update the times on the inode then do device close.
1892 * ext2spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1893 * struct thread *a_td)
1897 ext2spec_close(struct vop_close_args *ap)
1899 struct vnode *vp = ap->a_vp;
1901 if (vp->v_usecount > 1)
1903 return (VOCALL(spec_vnode_vops, &ap->a_head));
1907 * Read wrapper for fifos.
1909 * ext2fifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1910 * struct ucred *a_cred)
1914 ext2fifo_read(struct vop_read_args *ap)
1921 resid = uio->uio_resid;
1922 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1923 ip = VTOI(ap->a_vp);
1924 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1925 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1926 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1931 * Write wrapper for fifos.
1933 * ext2fifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1934 * struct ucred *a_cred)
1938 ext2fifo_write(struct vop_write_args *ap)
1945 resid = uio->uio_resid;
1946 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1947 ip = VTOI(ap->a_vp);
1948 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1949 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1954 * Close wrapper for fifos.
1956 * Update the times on the inode then do device close.
1958 * ext2fifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1959 * struct thread *a_td)
1963 ext2fifo_close(struct vop_close_args *ap)
1965 struct vnode *vp = ap->a_vp;
1967 if (vp->v_usecount > 1)
1969 return (VOCALL(fifo_vnode_vops, &ap->a_head));
1973 * Kqfilter wrapper for fifos.
1975 * Fall through to ext2 kqfilter routines if needed
1979 ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1983 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1985 error = ext2_kqfilter(ap);
1990 * Return POSIX pathconf information applicable to ext2 filesystems.
1992 * ext2_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
1996 ext2_pathconf(struct vop_pathconf_args *ap)
1998 switch (ap->a_name) {
2000 *ap->a_retval = LINK_MAX;
2003 *ap->a_retval = NAME_MAX;
2006 *ap->a_retval = PATH_MAX;
2009 *ap->a_retval = PIPE_BUF;
2011 case _PC_CHOWN_RESTRICTED:
2024 * Advisory record locking support
2026 * ext2_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
2031 ext2_advlock(struct vop_advlock_args *ap)
2033 struct inode *ip = VTOI(ap->a_vp);
2035 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
2039 * Initialize the vnode associated with a new inode, handle aliased
2043 ext2_vinit(struct mount *mntp, struct vnode **vpp)
2052 switch(vp->v_type = IFTOVT(ip->i_mode)) {
2055 vp->v_ops = &mntp->mnt_vn_spec_ops;
2056 addaliasu(vp, ip->i_rdev);
2059 vp->v_ops = &mntp->mnt_vn_fifo_ops;
2063 vinitvmio(vp, ip->i_size);
2066 if ((ip->i_size >= vp->v_mount->mnt_maxsymlinklen) &&
2067 ip->i_din.di_blocks != 0
2069 vinitvmio(vp, ip->i_size);
2077 if (ip->i_number == ROOTINO)
2078 vp->v_flag |= VROOT;
2080 * Initialize modrev times
2082 getmicrouptime(&tv);
2083 SETHIGH(ip->i_modrev, tv.tv_sec);
2084 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
2089 static struct filterops ext2read_filtops =
2090 { 1, NULL, filt_ext2detach, filt_ext2read };
2091 static struct filterops ext2write_filtops =
2092 { 1, NULL, filt_ext2detach, filt_ext2write };
2093 static struct filterops ext2vnode_filtops =
2094 { 1, NULL, filt_ext2detach, filt_ext2vnode };
2097 * ext2_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2100 ext2_kqfilter(struct vop_kqfilter_args *ap)
2102 struct vnode *vp = ap->a_vp;
2103 struct knote *kn = ap->a_kn;
2106 switch (kn->kn_filter) {
2108 kn->kn_fop = &ext2read_filtops;
2111 kn->kn_fop = &ext2write_filtops;
2114 kn->kn_fop = &ext2vnode_filtops;
2120 kn->kn_hook = (caddr_t)vp;
2122 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2123 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
2124 lwkt_reltoken(&ilock);
2130 filt_ext2detach(struct knote *kn)
2132 struct vnode *vp = (struct vnode *)kn->kn_hook;
2135 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2136 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2137 kn, knote, kn_selnext);
2138 lwkt_reltoken(&ilock);
2143 filt_ext2read(struct knote *kn, long hint)
2145 struct vnode *vp = (struct vnode *)kn->kn_hook;
2146 struct inode *ip = VTOI(vp);
2149 * filesystem is gone, so set the EOF flag and schedule
2150 * the knote for deletion.
2152 if (hint == NOTE_REVOKE) {
2153 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2157 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2158 return (kn->kn_data != 0);
2163 filt_ext2write(struct knote *kn, long hint)
2166 * filesystem is gone, so set the EOF flag and schedule
2167 * the knote for deletion.
2169 if (hint == NOTE_REVOKE)
2170 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2177 filt_ext2vnode(struct knote *kn, long hint)
2179 if (kn->kn_sfflags & hint)
2180 kn->kn_fflags |= hint;
2181 if (hint == NOTE_REVOKE) {
2182 kn->kn_flags |= EV_EOF;
2185 return (kn->kn_fflags != 0);
2188 static struct vop_ops *ext2_vnode_vops;
2189 /* Global vfs data structures for ext2. */
2190 struct vnodeopv_entry_desc ext2_vnodeop_entries[] = {
2191 { &vop_default_desc, vop_defaultop },
2192 { &vop_fsync_desc, (vnodeopv_entry_t) ext2_fsync },
2193 { &vop_read_desc, (vnodeopv_entry_t) ext2_read },
2194 { &vop_reallocblks_desc, (vnodeopv_entry_t) ext2_reallocblks },
2195 { &vop_write_desc, (vnodeopv_entry_t) ext2_write },
2196 { &vop_access_desc, (vnodeopv_entry_t) ext2_access },
2197 { &vop_advlock_desc, (vnodeopv_entry_t) ext2_advlock },
2198 { &vop_bmap_desc, (vnodeopv_entry_t) ext2_bmap },
2199 { &vop_old_lookup_desc, (vnodeopv_entry_t) ext2_lookup },
2200 { &vop_close_desc, (vnodeopv_entry_t) ext2_close },
2201 { &vop_old_create_desc, (vnodeopv_entry_t) ext2_create },
2202 { &vop_getattr_desc, (vnodeopv_entry_t) ext2_getattr },
2203 { &vop_inactive_desc, (vnodeopv_entry_t) ext2_inactive },
2204 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2205 { &vop_old_link_desc, (vnodeopv_entry_t) ext2_link },
2206 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2207 { &vop_old_mkdir_desc, (vnodeopv_entry_t) ext2_mkdir },
2208 { &vop_old_mknod_desc, (vnodeopv_entry_t) ext2_mknod },
2209 { &vop_mmap_desc, (vnodeopv_entry_t) ext2_mmap },
2210 { &vop_open_desc, (vnodeopv_entry_t) ext2_open },
2211 { &vop_pathconf_desc, (vnodeopv_entry_t) ext2_pathconf },
2212 { &vop_poll_desc, (vnodeopv_entry_t) vop_stdpoll },
2213 { &vop_kqfilter_desc, (vnodeopv_entry_t) ext2_kqfilter },
2214 { &vop_print_desc, (vnodeopv_entry_t) ext2_print },
2215 { &vop_readdir_desc, (vnodeopv_entry_t) ext2_readdir },
2216 { &vop_readlink_desc, (vnodeopv_entry_t) ext2_readlink },
2217 { &vop_reclaim_desc, (vnodeopv_entry_t) ext2_reclaim },
2218 { &vop_old_remove_desc, (vnodeopv_entry_t) ext2_remove },
2219 { &vop_old_rename_desc, (vnodeopv_entry_t) ext2_rename },
2220 { &vop_old_rmdir_desc, (vnodeopv_entry_t) ext2_rmdir },
2221 { &vop_setattr_desc, (vnodeopv_entry_t) ext2_setattr },
2222 { &vop_strategy_desc, (vnodeopv_entry_t) ext2_strategy },
2223 { &vop_old_symlink_desc, (vnodeopv_entry_t) ext2_symlink },
2224 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2225 { &vop_old_whiteout_desc, (vnodeopv_entry_t) ext2_whiteout },
2226 { &vop_getpages_desc, (vnodeopv_entry_t) ext2_getpages },
2227 { &vop_putpages_desc, (vnodeopv_entry_t) ext2_putpages },
2230 static struct vnodeopv_desc ext2_vnodeop_opv_desc =
2231 { &ext2_vnode_vops, ext2_vnodeop_entries, 0 };
2233 static struct vop_ops *ext2_spec_vops;
2234 struct vnodeopv_entry_desc ext2_specop_entries[] = {
2235 { &vop_default_desc, (vnodeopv_entry_t) ext2_vnoperatespec },
2236 { &vop_fsync_desc, (vnodeopv_entry_t) ext2_fsync },
2237 { &vop_access_desc, (vnodeopv_entry_t) ext2_access },
2238 { &vop_close_desc, (vnodeopv_entry_t) ext2spec_close },
2239 { &vop_getattr_desc, (vnodeopv_entry_t) ext2_getattr },
2240 { &vop_inactive_desc, (vnodeopv_entry_t) ext2_inactive },
2241 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2242 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2243 { &vop_print_desc, (vnodeopv_entry_t) ext2_print },
2244 { &vop_read_desc, (vnodeopv_entry_t) ext2spec_read },
2245 { &vop_reclaim_desc, (vnodeopv_entry_t) ext2_reclaim },
2246 { &vop_setattr_desc, (vnodeopv_entry_t) ext2_setattr },
2247 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2248 { &vop_write_desc, (vnodeopv_entry_t) ext2spec_write },
2251 static struct vnodeopv_desc ext2_specop_opv_desc =
2252 { &ext2_spec_vops, ext2_specop_entries, 0 };
2254 static struct vop_ops *ext2_fifo_vops;
2255 struct vnodeopv_entry_desc ext2_fifoop_entries[] = {
2256 { &vop_default_desc, (vnodeopv_entry_t) ext2_vnoperatefifo },
2257 { &vop_fsync_desc, (vnodeopv_entry_t) ext2_fsync },
2258 { &vop_access_desc, (vnodeopv_entry_t) ext2_access },
2259 { &vop_close_desc, (vnodeopv_entry_t) ext2fifo_close },
2260 { &vop_getattr_desc, (vnodeopv_entry_t) ext2_getattr },
2261 { &vop_inactive_desc, (vnodeopv_entry_t) ext2_inactive },
2262 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2263 { &vop_kqfilter_desc, (vnodeopv_entry_t) ext2fifo_kqfilter },
2264 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2265 { &vop_print_desc, (vnodeopv_entry_t) ext2_print },
2266 { &vop_read_desc, (vnodeopv_entry_t) ext2fifo_read },
2267 { &vop_reclaim_desc, (vnodeopv_entry_t) ext2_reclaim },
2268 { &vop_setattr_desc, (vnodeopv_entry_t) ext2_setattr },
2269 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2270 { &vop_write_desc, (vnodeopv_entry_t) ext2fifo_write },
2273 static struct vnodeopv_desc ext2_fifoop_opv_desc =
2274 { &ext2_fifo_vops, ext2_fifoop_entries, 0 };
2276 VNODEOP_SET(ext2_vnodeop_opv_desc);
2277 VNODEOP_SET(ext2_specop_opv_desc);
2278 VNODEOP_SET(ext2_fifoop_opv_desc);
2281 * ext2_vnoperate(struct vnodeop_desc *a_desc)
2284 ext2_vnoperate(struct vop_generic_args *ap)
2286 return (VOCALL(ext2_vnode_vops, ap));
2290 * ext2_vnoperatefifo(struct vnodeop_desc *a_desc)
2293 ext2_vnoperatefifo(struct vop_generic_args *ap)
2295 return (VOCALL(ext2_fifo_vops, ap));
2299 * ext2_vnoperatespec(struct vnodeop_desc *a_desc)
2302 ext2_vnoperatespec(struct vop_generic_args *ap)
2304 return (VOCALL(ext2_spec_vops, ap));