kernel - VM PAGER part 2/2 - addendum, fix compile time issue w/ext2
[dragonfly.git] / sys / vfs / gnu / ext2fs / ext2_vnops.c
blobc91d70a6fd5c64e8ab7ad120510ae1d39ed06709
1 /*
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
6 */
7 /*
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
18 * are met:
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
42 * SUCH DAMAGE.
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 1.43 2008/06/19 23:27:39 dillon 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>
57 #include <sys/stat.h>
58 #include <sys/fcntl.h>
59 #include <sys/buf.h>
60 #include <sys/stat.h>
61 #include <sys/proc.h>
62 #include <sys/priv.h>
63 #include <sys/mount.h>
64 #include <sys/time.h>
65 #include <sys/vnode.h>
66 #include <sys/dirent.h>
67 #include <sys/lockf.h>
68 #include <sys/event.h>
69 #include <sys/conf.h>
70 #include <sys/file.h>
71 #include <sys/jail.h>
72 #include <sys/namei.h>
73 #include <sys/signalvar.h>
74 #include <sys/unistd.h>
76 #include <vm/vm.h>
77 #include <vm/vm_extern.h>
78 #include <vm/vm_zone.h>
79 #include <vm/vnode_pager.h>
81 #include <sys/buf2.h>
82 #include <sys/thread2.h>
84 #include <vfs/fifofs/fifo.h>
86 #include "dir.h"
87 #include "quota.h"
88 #include "inode.h"
89 #include "ext2mount.h"
90 #include "ext2_fs_sb.h"
91 #include "fs.h"
92 #include "ext2_extern.h"
93 #include "ext2_fs.h"
95 static int ext2_access (struct vop_access_args *);
96 static int ext2_advlock (struct vop_advlock_args *);
97 static int ext2_chmod (struct vnode *, int, struct ucred *);
98 static int ext2_chown (struct vnode *, uid_t, gid_t, struct ucred *);
99 static int ext2_close (struct vop_close_args *);
100 static int ext2_getattr (struct vop_getattr_args *);
101 static int ext2_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
102 static int ext2_mmap (struct vop_mmap_args *);
103 static int ext2_open (struct vop_open_args *);
104 static int ext2_pathconf (struct vop_pathconf_args *);
105 static int ext2_print (struct vop_print_args *);
106 static int ext2_readlink (struct vop_readlink_args *);
107 static int ext2_setattr (struct vop_setattr_args *);
108 static int ext2_strategy (struct vop_strategy_args *);
109 static int ext2_whiteout (struct vop_old_whiteout_args *);
110 static int filt_ext2read (struct knote *kn, long hint);
111 static int filt_ext2write (struct knote *kn, long hint);
112 static int filt_ext2vnode (struct knote *kn, long hint);
113 static void filt_ext2detach (struct knote *kn);
114 static int ext2_kqfilter (struct vop_kqfilter_args *ap);
115 static int ext2fifo_close (struct vop_close_args *);
116 static int ext2fifo_kqfilter (struct vop_kqfilter_args *);
117 static int ext2fifo_read (struct vop_read_args *);
118 static int ext2fifo_write (struct vop_write_args *);
120 static int ext2_fsync (struct vop_fsync_args *);
121 static int ext2_read (struct vop_read_args *);
122 static int ext2_write (struct vop_write_args *);
123 static int ext2_remove (struct vop_old_remove_args *);
124 static int ext2_link (struct vop_old_link_args *);
125 static int ext2_rename (struct vop_old_rename_args *);
126 static int ext2_mkdir (struct vop_old_mkdir_args *);
127 static int ext2_rmdir (struct vop_old_rmdir_args *);
128 static int ext2_create (struct vop_old_create_args *);
129 static int ext2_mknod (struct vop_old_mknod_args *);
130 static int ext2_symlink (struct vop_old_symlink_args *);
132 #include "ext2_readwrite.c"
134 union _qcvt {
135 int64_t qcvt;
136 int32_t val[2];
138 #define SETHIGH(q, h) { \
139 union _qcvt tmp; \
140 tmp.qcvt = (q); \
141 tmp.val[_QUAD_HIGHWORD] = (h); \
142 (q) = tmp.qcvt; \
144 #define SETLOW(q, l) { \
145 union _qcvt tmp; \
146 tmp.qcvt = (q); \
147 tmp.val[_QUAD_LOWWORD] = (l); \
148 (q) = tmp.qcvt; \
150 #define VN_KNOTE(vp, b) \
151 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
153 #define OFSFMT(vp) ((vp)->v_mount->mnt_maxsymlinklen <= 0)
156 * A virgin directory (no blushing please).
157 * Note that the type and namlen fields are reversed relative to ufs.
158 * Also, we don't use `struct odirtemplate', since it would just cause
159 * endianness problems.
161 static struct dirtemplate ext2_mastertemplate = {
162 0, 12, 1, EXT2_FT_DIR, ".",
163 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
165 static struct dirtemplate ext2_omastertemplate = {
166 0, 12, 1, EXT2_FT_UNKNOWN, ".",
167 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
171 * Create a regular file
173 * ext2_create(struct vnode *a_dvp, struct vnode **a_vpp,
174 * struct componentname *a_cnp, struct vattr *a_vap)
176 static int
177 ext2_create(struct vop_old_create_args *ap)
179 int error;
181 error =
182 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
183 ap->a_dvp, ap->a_vpp, ap->a_cnp);
184 if (error)
185 return (error);
186 return (0);
190 * Synch an open file.
192 * ext2_fsync(struct vnode *a_vp, struct ucred *a_cred, int a_waitfor,
193 * struct proc *a_p)
195 /* ARGSUSED */
197 static int ext2_fsync_bp(struct buf *bp, void *data);
199 struct ext2_fsync_bp_info {
200 struct vnode *vp;
201 int waitfor;
204 static int
205 ext2_fsync(struct vop_fsync_args *ap)
207 struct ext2_fsync_bp_info info;
208 struct vnode *vp = ap->a_vp;
209 lwkt_tokref vlock;
210 int count;
213 * XXX why is all this fs specific?
217 * Flush all dirty buffers associated with a vnode.
219 ext2_discard_prealloc(VTOI(vp));
221 lwkt_gettoken(&vlock, &vp->v_token);
222 info.vp = vp;
223 loop:
224 info.waitfor = ap->a_waitfor;
225 count = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
226 ext2_fsync_bp, &info);
227 if (count)
228 goto loop;
230 if (ap->a_waitfor == MNT_WAIT) {
231 bio_track_wait(&vp->v_track_write, 0, 0);
232 #if DIAGNOSTIC
233 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
234 vprint("ext2_fsync: dirty", vp);
235 goto loop;
237 #endif
239 lwkt_reltoken(&vlock);
240 return (EXT2_UPDATE(ap->a_vp, ap->a_waitfor == MNT_WAIT));
243 static int
244 ext2_fsync_bp(struct buf *bp, void *data)
246 struct ext2_fsync_bp_info *info = data;
248 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
249 return(0);
250 if ((bp->b_flags & B_DELWRI) == 0)
251 panic("ext2_fsync: not dirty");
252 bremfree(bp);
255 * Wait for I/O associated with indirect blocks to complete,
256 * since there is no way to quickly wait for them below.
258 if (bp->b_vp == info->vp || info->waitfor == MNT_NOWAIT)
259 bawrite(bp);
260 else
261 bwrite(bp);
262 return(1);
266 * Mknod vnode call
268 * ext2_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
269 * struct componentname *a_cnp, struct vattr *a_vap)
271 /* ARGSUSED */
272 static int
273 ext2_mknod(struct vop_old_mknod_args *ap)
275 struct vattr *vap = ap->a_vap;
276 struct vnode **vpp = ap->a_vpp;
277 struct inode *ip;
278 ino_t ino;
279 int error;
281 if (vap->va_rmajor != VNOVAL &&
282 makeudev(vap->va_rmajor, vap->va_rminor) == NOUDEV) {
283 return (EINVAL);
286 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
287 ap->a_dvp, vpp, ap->a_cnp);
288 if (error)
289 return (error);
290 ip = VTOI(*vpp);
291 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
292 if (vap->va_rmajor != VNOVAL) {
294 * Want to be able to use this to make badblock
295 * inodes, so don't truncate the dev number.
297 ip->i_rdev = makeudev(vap->va_rmajor, vap->va_rminor);
300 * Remove inode, then reload it through VFS_VGET so it is
301 * checked to see if it is an alias of an existing entry in
302 * the inode cache.
304 (*vpp)->v_type = VNON;
305 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
306 vgone_vxlocked(*vpp);
307 vput(*vpp);
308 error = VFS_VGET(ap->a_dvp->v_mount, NULL, ino, vpp);
309 if (error) {
310 *vpp = NULL;
311 return (error);
313 return (0);
317 * ext2_remove(struct vnode *a_dvp, struct vnode *a_vp,
318 * struct componentname *a_cnp)
320 static int
321 ext2_remove(struct vop_old_remove_args *ap)
323 struct inode *ip;
324 struct vnode *vp = ap->a_vp;
325 struct vnode *dvp = ap->a_dvp;
326 int error;
328 ip = VTOI(vp);
329 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
330 (VTOI(dvp)->i_flags & APPEND)) {
331 error = EPERM;
332 goto out;
334 error = ext2_dirremove(dvp, ap->a_cnp);
335 if (error == 0) {
336 ip->i_nlink--;
337 ip->i_flag |= IN_CHANGE;
339 out:
340 return (error);
344 * link vnode call
346 * ext2_link(struct vnode *a_tdvp, struct vnode *a_vp,
347 * struct componentname *a_cnp)
349 static int
350 ext2_link(struct vop_old_link_args *ap)
352 struct vnode *vp = ap->a_vp;
353 struct vnode *tdvp = ap->a_tdvp;
354 struct componentname *cnp = ap->a_cnp;
355 struct inode *ip;
356 int error;
358 if (tdvp->v_mount != vp->v_mount) {
359 error = EXDEV;
360 goto out2;
362 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE))) {
363 goto out2;
365 ip = VTOI(vp);
366 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
367 error = EMLINK;
368 goto out1;
370 if (ip->i_flags & (IMMUTABLE | APPEND)) {
371 error = EPERM;
372 goto out1;
374 ip->i_nlink++;
375 ip->i_flag |= IN_CHANGE;
376 error = EXT2_UPDATE(vp, 1);
377 if (!error)
378 error = ext2_direnter(ip, tdvp, cnp);
379 if (error) {
380 ip->i_nlink--;
381 ip->i_flag |= IN_CHANGE;
383 out1:
384 if (tdvp != vp)
385 vn_unlock(vp);
386 out2:
387 return (error);
391 * Rename system call. fdvp, fvp are ref'd. tvp, tdvp are ref'd and locked.
392 * all vp's are released and must be in an unlocked state on return.
394 * ext2_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
395 * struct componentname *a_fcnp, struct vnode *a_tdvp,
396 * struct vnode *a_tvp, struct componentname *a_tcnp)
398 static int
399 ext2_rename(struct vop_old_rename_args *ap)
401 struct vnode *tvp = ap->a_tvp;
402 struct vnode *tdvp = ap->a_tdvp;
403 struct vnode *fvp = ap->a_fvp;
404 struct vnode *fdvp = ap->a_fdvp;
405 struct componentname *tcnp = ap->a_tcnp;
406 struct componentname *fcnp = ap->a_fcnp;
407 struct inode *ip, *xp, *dp;
408 struct dirtemplate dirbuf;
409 int doingdirectory = 0, oldparent = 0, newparent = 0;
410 int error = 0;
411 u_char namlen;
414 * Check for cross-device rename.
416 if ((fvp->v_mount != tdvp->v_mount) ||
417 (tvp && (fvp->v_mount != tvp->v_mount)) ||
418 tvp == tdvp) {
419 error = EXDEV;
420 abortit:
421 if (tdvp == tvp)
422 vrele(tdvp);
423 else
424 vput(tdvp);
425 if (tvp)
426 vput(tvp);
427 vrele(fdvp);
428 vrele(fvp);
429 return (error);
432 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
433 (VTOI(tdvp)->i_flags & APPEND))) {
434 error = EPERM;
435 goto abortit;
439 * Renaming a file to itself has no effect. The upper layers should
440 * not call us in that case. Temporarily just warn if they do.
442 if (fvp == tvp) {
443 error = 0;
444 goto abortit;
447 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
448 goto abortit;
451 * fvp, tvp, tdvp locked. fdvp not locked but note that fdvp may
452 * be equal to tdvp.
454 dp = VTOI(fdvp);
455 ip = VTOI(fvp);
456 if (ip->i_nlink >= LINK_MAX) {
457 vn_unlock(fvp);
458 error = EMLINK;
459 goto abortit;
461 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
462 || (dp->i_flags & APPEND)) {
463 vn_unlock(fvp);
464 error = EPERM;
465 goto abortit;
467 if ((ip->i_mode & IFMT) == IFDIR) {
469 * Avoid ".", "..", and aliases of "." for obvious reasons.
471 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
472 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
473 (ip->i_flag & IN_RENAME)) {
474 vn_unlock(fvp);
475 error = EINVAL;
476 goto abortit;
478 ip->i_flag |= IN_RENAME;
479 oldparent = dp->i_number;
480 doingdirectory++;
484 * tvp is non-NULL if the target exists. fvp is still locked but
485 * we will unlock it soon. The 'bad' goto target requires dp and
486 * xp to be correctly assigned.
488 dp = VTOI(tdvp);
489 if (tvp)
490 xp = VTOI(tvp);
491 else
492 xp = NULL;
495 * 1) Bump link count while we're moving stuff
496 * around. If we crash somewhere before
497 * completing our work, the link count
498 * may be wrong, but correctable.
500 ip->i_nlink++;
501 ip->i_flag |= IN_CHANGE;
502 if ((error = EXT2_UPDATE(fvp, 1)) != 0) {
503 vn_unlock(fvp);
504 goto bad;
508 * If ".." must be changed (ie the directory gets a new
509 * parent) then the source directory must not be in the
510 * directory heirarchy above the target, as this would
511 * orphan everything below the source directory. Also
512 * the user must have write permission in the source so
513 * as to be able to change "..". We must repeat the call
514 * to namei, as the parent directory is unlocked by the
515 * call to checkpath().
517 error = VOP_EACCESS(fvp, VWRITE, tcnp->cn_cred);
518 vn_unlock(fvp);
521 * tvp (if not NULL) and tdvp are locked. fvp and fdvp are not.
522 * dp and xp are set according to tdvp and tvp.
524 if (oldparent != dp->i_number)
525 newparent = dp->i_number;
526 if (doingdirectory && newparent) {
527 if (error) /* write access check above */
528 goto bad;
531 * Prepare for relookup, get rid of xp
533 if (xp != NULL) {
534 vput(tvp);
535 xp = NULL;
539 * checkpath vput()'s tdvp (VTOI(dp)) on return no matter what,
540 * get an extra ref so we wind up with just an unlocked, ref'd
541 * tdvp. The 'out' target skips xp and tdvp cleanups. Our
542 * tdvp is now unlocked so we have to clean it up ourselves.
544 vref(tdvp);
545 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
546 tcnp->cn_flags |= CNP_PDIRUNLOCK;
547 if (error) {
548 vrele(tdvp);
549 goto out;
552 * relookup no longer messes with the ref count. An unlocked
553 * tdvp must be passed and if no error occurs a locked tdvp
554 * will be returned. We have to use the out target again.
556 error = relookup(tdvp, &tvp, tcnp);
557 if (error) {
558 if (tcnp->cn_flags & CNP_PDIRUNLOCK)
559 vrele(tdvp);
560 else
561 vput(tdvp);
562 goto out;
566 * tdvp is locked at this point. in the RENAME case tvp may
567 * be NULL without an error, assign xp accordingly. The
568 * 'bad' target can be used again after this.
570 dp = VTOI(tdvp);
571 if (tvp)
572 xp = VTOI(tvp);
575 * 2) If target doesn't exist, link the target
576 * to the source and unlink the source.
577 * Otherwise, rewrite the target directory
578 * entry to reference the source inode and
579 * expunge the original entry's existence.
581 * tdvp and tvp are cleaned up by this code. tvp is only good if
582 * xp is not NULL.
584 if (xp == NULL) {
585 if (dp->i_dev != ip->i_dev)
586 panic("ext2_rename: EXDEV");
588 * Account for ".." in new directory.
589 * When source and destination have the same
590 * parent we don't fool with the link count.
592 if (doingdirectory && newparent) {
593 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
594 error = EMLINK;
595 goto bad;
597 dp->i_nlink++;
598 dp->i_flag |= IN_CHANGE;
599 error = EXT2_UPDATE(tdvp, 1);
600 if (error)
601 goto bad;
603 error = ext2_direnter(ip, tdvp, tcnp);
604 if (error) {
605 if (doingdirectory && newparent) {
606 dp->i_nlink--;
607 dp->i_flag |= IN_CHANGE;
608 EXT2_UPDATE(tdvp, 1);
610 goto bad;
614 * manual cleanup, we can't use the bad or out target after
615 * this.
617 vput(tdvp);
618 } else {
619 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
620 panic("ext2_rename: EXDEV");
622 * Short circuit rename(foo, foo).
624 if (xp->i_number == ip->i_number)
625 panic("ext2_rename: same file");
627 * If the parent directory is "sticky", then the user must
628 * own the parent directory, or the destination of the rename,
629 * otherwise the destination may not be changed (except by
630 * root). This implements append-only directories.
632 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
633 tcnp->cn_cred->cr_uid != dp->i_uid &&
634 xp->i_uid != tcnp->cn_cred->cr_uid) {
635 error = EPERM;
636 goto bad;
639 * Target must be empty if a directory and have no links
640 * to it. Also, ensure source and target are compatible
641 * (both directories, or both not directories).
643 if ((xp->i_mode&IFMT) == IFDIR) {
644 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
645 xp->i_nlink > 2) {
646 error = ENOTEMPTY;
647 goto bad;
649 if (!doingdirectory) {
650 error = ENOTDIR;
651 goto bad;
653 } else if (doingdirectory) {
654 error = EISDIR;
655 goto bad;
657 error = ext2_dirrewrite(dp, ip, tcnp);
658 if (error)
659 goto bad;
661 * If the target directory is in the same
662 * directory as the source directory,
663 * decrement the link count on the parent
664 * of the target directory.
666 if (doingdirectory && !newparent) {
667 dp->i_nlink--;
668 dp->i_flag |= IN_CHANGE;
672 * manual cleanup, we can't use the bad or out target after
673 * this.
675 vput(tdvp);
678 * Adjust the link count of the target to
679 * reflect the dirrewrite above. If this is
680 * a directory it is empty and there are
681 * no links to it, so we can squash the inode and
682 * any space associated with it. We disallowed
683 * renaming over top of a directory with links to
684 * it above, as the remaining link would point to
685 * a directory without "." or ".." entries.
687 xp->i_nlink--;
688 if (doingdirectory) {
689 if (--xp->i_nlink != 0)
690 panic("ext2_rename: linked directory");
691 error = EXT2_TRUNCATE(tvp, (off_t)0, IO_SYNC,
692 tcnp->cn_cred);
694 xp->i_flag |= IN_CHANGE;
695 vput(tvp);
696 xp = NULL;
700 * tvp and tdvp have been cleaned up. The bad and out targets may
701 * not be used. fvp and fdvp are ref'd but not locked. ip
702 * still represents the old fvp and ip->i_flag may still have IN_RENAME
703 * set (if doingdirectory).
707 * 3) Unlink the source.
709 * fdvp is locked and ref'd. ap->a_fvp holds the old lookup unlocked
710 * and ref'd, fvp will hold the new lookup locked and ref'd.
712 * After the relookup ap->a_fvp must be released as part of our
713 * cleanup, not just fdvp and fvp. And, on success, fdvp and
714 * fvp will be locked so the bad and out targets cannot be used.
716 fcnp->cn_flags &= ~CNP_MODMASK;
717 fcnp->cn_flags |= CNP_LOCKPARENT;
718 KKASSERT(fcnp->cn_flags & CNP_PDIRUNLOCK);
719 error = relookup(fdvp, &fvp, fcnp);
720 if (error) {
722 * From name has disappeared.
724 if (doingdirectory)
725 panic("ext2_rename: lost dir entry");
726 /* ip->i_flag only sets IN_RENAME if doingdirectory */
727 vrele(ap->a_fvp);
728 if (fcnp->cn_flags & CNP_PDIRUNLOCK)
729 vrele(fdvp);
730 else
731 vput(fdvp);
732 return (0);
734 KKASSERT((fcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
737 * This case shouldn't occur
739 if (fvp == NULL) {
741 * From name has disappeared.
743 if (doingdirectory)
744 panic("ext2_rename: lost dir entry");
745 /* ip->i_flag only sets IN_RENAME if doingdirectory */
746 vrele(ap->a_fvp);
747 vput(fvp);
748 vput(fdvp);
749 return (0);
753 * fvp and fdvp are both ref'd and locked.
755 xp = VTOI(fvp);
756 dp = VTOI(fdvp);
759 * Ensure that the directory entry still exists and has not
760 * changed while the new name has been entered. If the source is
761 * a file then the entry may have been unlinked or renamed. In
762 * either case there is no further work to be done. If the source
763 * is a directory then it cannot have been rmdir'ed; its link
764 * count of three would cause a rmdir to fail with ENOTEMPTY.
765 * The IN_RENAME flag ensures that it cannot be moved by another
766 * rename.
768 if (xp != ip) {
769 if (doingdirectory)
770 panic("ext2_rename: lost dir entry");
771 /* ip->i_flag only sets IN_RENAME if doingdirectory */
772 } else {
774 * If the source is a directory with a
775 * new parent, the link count of the old
776 * parent directory must be decremented
777 * and ".." set to point to the new parent.
779 if (doingdirectory && newparent) {
780 dp->i_nlink--;
781 dp->i_flag |= IN_CHANGE;
782 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
783 sizeof (struct dirtemplate), (off_t)0,
784 UIO_SYSSPACE, IO_NODELOCKED,
785 tcnp->cn_cred, NULL);
786 if (error == 0) {
787 /* Like ext2 little-endian: */
788 namlen = dirbuf.dotdot_type;
789 if (namlen != 2 ||
790 dirbuf.dotdot_name[0] != '.' ||
791 dirbuf.dotdot_name[1] != '.') {
792 ext2_dirbad(xp, (doff_t)12,
793 "rename: mangled dir");
794 } else {
795 dirbuf.dotdot_ino = newparent;
796 vn_rdwr(UIO_WRITE, fvp,
797 (caddr_t)&dirbuf,
798 sizeof (struct dirtemplate),
799 (off_t)0, UIO_SYSSPACE,
800 IO_NODELOCKED|IO_SYNC,
801 tcnp->cn_cred, NULL);
805 error = ext2_dirremove(fdvp, fcnp);
806 if (!error) {
807 xp->i_nlink--;
808 xp->i_flag |= IN_CHANGE;
810 xp->i_flag &= ~IN_RENAME;
812 vput(fdvp);
813 vput(fvp);
814 vrele(ap->a_fvp);
815 return (error);
817 bad:
818 if (xp)
819 vput(ITOV(xp));
820 if (dp)
821 vput(ITOV(dp));
822 out:
823 if (doingdirectory)
824 ip->i_flag &= ~IN_RENAME;
825 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
826 ip->i_nlink--;
827 ip->i_flag |= IN_CHANGE;
828 ip->i_flag &= ~IN_RENAME;
829 vput(fvp);
830 } else {
831 vrele(fvp);
833 return (error);
837 * Mkdir system call
839 * ext2_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
840 * struct componentname *a_cnp, struct vattr *a_vap)
842 static int
843 ext2_mkdir(struct vop_old_mkdir_args *ap)
845 struct vnode *dvp = ap->a_dvp;
846 struct vattr *vap = ap->a_vap;
847 struct componentname *cnp = ap->a_cnp;
848 struct inode *ip, *dp;
849 struct vnode *tvp;
850 struct dirtemplate dirtemplate, *dtp;
851 int error, dmode;
853 dp = VTOI(dvp);
854 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
855 error = EMLINK;
856 goto out;
858 dmode = vap->va_mode & 0777;
859 dmode |= IFDIR;
861 * Must simulate part of ext2_makeinode here to acquire the inode,
862 * but not have it entered in the parent directory. The entry is
863 * made later after writing "." and ".." entries.
865 error = EXT2_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
866 if (error)
867 goto out;
868 ip = VTOI(tvp);
869 ip->i_gid = dp->i_gid;
870 #ifdef SUIDDIR
872 #ifdef QUOTA
873 struct ucred ucred, *ucp;
874 ucp = cnp->cn_cred;
875 #endif
877 * if we are hacking owners here, (only do this where told to)
878 * and we are not giving it TOO root, (would subvert quotas)
879 * then go ahead and give it to the other user.
880 * The new directory also inherits the SUID bit.
881 * If user's UID and dir UID are the same,
882 * 'give it away' so that the SUID is still forced on.
884 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
885 (dp->i_mode & ISUID) && dp->i_uid) {
886 dmode |= ISUID;
887 ip->i_uid = dp->i_uid;
888 #ifdef QUOTA
889 if (dp->i_uid != cnp->cn_cred->cr_uid) {
891 * make sure the correct user gets charged
892 * for the space.
893 * Make a dummy credential for the victim.
894 * XXX This seems to never be accessed out of
895 * our context so a stack variable is ok.
897 ucred.cr_ref = 1;
898 ucred.cr_uid = ip->i_uid;
899 ucred.cr_ngroups = 1;
900 ucred.cr_groups[0] = dp->i_gid;
901 ucp = &ucred;
903 #endif
904 } else {
905 ip->i_uid = cnp->cn_cred->cr_uid;
907 #ifdef QUOTA
908 if ((error = ext2_getinoquota(ip)) ||
909 (error = ext2_chkiq(ip, 1, ucp, 0))) {
910 EXT2_VFREE(tvp, ip->i_number, dmode);
911 vput(tvp);
912 return (error);
914 #endif
916 #else
917 ip->i_uid = cnp->cn_cred->cr_uid;
918 #ifdef QUOTA
919 if ((error = ext2_getinoquota(ip)) ||
920 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
921 EXT2_VFREE(tvp, ip->i_number, dmode);
922 vput(tvp);
923 return (error);
925 #endif
926 #endif
927 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
928 ip->i_mode = dmode;
929 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
930 ip->i_nlink = 2;
931 if (cnp->cn_flags & CNP_ISWHITEOUT)
932 ip->i_flags |= UF_OPAQUE;
933 error = EXT2_UPDATE(tvp, 1);
936 * The vnode must have a VM object in order to issue buffer cache
937 * ops on it.
939 vinitvmio(tvp, 0, PAGE_SIZE, -1);
942 * Bump link count in parent directory
943 * to reflect work done below. Should
944 * be done before reference is created
945 * so reparation is possible if we crash.
947 dp->i_nlink++;
948 dp->i_flag |= IN_CHANGE;
949 error = EXT2_UPDATE(dvp, 1);
950 if (error)
951 goto bad;
953 /* Initialize directory with "." and ".." from static template. */
954 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
955 EXT2_FEATURE_INCOMPAT_FILETYPE))
956 dtp = &ext2_mastertemplate;
957 else
958 dtp = &ext2_omastertemplate;
959 dirtemplate = *dtp;
960 dirtemplate.dot_ino = ip->i_number;
961 dirtemplate.dotdot_ino = dp->i_number;
962 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
963 * so let's just redefine it - for this function only
965 #undef DIRBLKSIZ
966 #define DIRBLKSIZ VTOI(dvp)->i_e2fs->s_blocksize
967 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
968 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
969 sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
970 IO_NODELOCKED|IO_SYNC, cnp->cn_cred, NULL);
971 if (error) {
972 dp->i_nlink--;
973 dp->i_flag |= IN_CHANGE;
974 goto bad;
976 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
977 panic("ext2_mkdir: blksize"); /* XXX should grow with balloc() */
978 else {
979 ip->i_size = DIRBLKSIZ;
980 ip->i_flag |= IN_CHANGE;
983 /* Directory set up, now install its entry in the parent directory. */
984 error = ext2_direnter(ip, dvp, cnp);
985 if (error) {
986 dp->i_nlink--;
987 dp->i_flag |= IN_CHANGE;
989 bad:
991 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
992 * for us because we set the link count to 0.
994 if (error) {
995 ip->i_nlink = 0;
996 ip->i_flag |= IN_CHANGE;
997 vput(tvp);
998 } else
999 *ap->a_vpp = tvp;
1000 out:
1001 return (error);
1002 #undef DIRBLKSIZ
1003 #define DIRBLKSIZ DEV_BSIZE
1007 * Rmdir system call.
1009 * ext2_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1010 * struct componentname *a_cnp)
1012 static int
1013 ext2_rmdir(struct vop_old_rmdir_args *ap)
1015 struct vnode *vp = ap->a_vp;
1016 struct vnode *dvp = ap->a_dvp;
1017 struct componentname *cnp = ap->a_cnp;
1018 struct inode *ip, *dp;
1019 int error;
1021 ip = VTOI(vp);
1022 dp = VTOI(dvp);
1025 * Verify the directory is empty (and valid).
1026 * (Rmdir ".." won't be valid since
1027 * ".." will contain a reference to
1028 * the current directory and thus be
1029 * non-empty.)
1031 error = 0;
1032 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1033 error = ENOTEMPTY;
1034 goto out;
1036 if ((dp->i_flags & APPEND)
1037 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1038 error = EPERM;
1039 goto out;
1042 * Delete reference to directory before purging
1043 * inode. If we crash in between, the directory
1044 * will be reattached to lost+found,
1046 error = ext2_dirremove(dvp, cnp);
1047 if (error)
1048 goto out;
1049 dp->i_nlink--;
1050 dp->i_flag |= IN_CHANGE;
1051 vn_unlock(dvp);
1053 * Truncate inode. The only stuff left
1054 * in the directory is "." and "..". The
1055 * "." reference is inconsequential since
1056 * we're quashing it. The ".." reference
1057 * has already been adjusted above. We've
1058 * removed the "." reference and the reference
1059 * in the parent directory, but there may be
1060 * other hard links so decrement by 2 and
1061 * worry about them later.
1063 ip->i_nlink -= 2;
1064 error = EXT2_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
1065 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
1066 out:
1067 return (error);
1071 * symlink -- make a symbolic link
1073 * ext2_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1074 * struct componentname *a_cnp, struct vattr *a_vap,
1075 * char *a_target)
1077 static int
1078 ext2_symlink(struct vop_old_symlink_args *ap)
1080 struct vnode *vp, **vpp = ap->a_vpp;
1081 struct inode *ip;
1082 int len, error;
1084 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1085 vpp, ap->a_cnp);
1086 if (error)
1087 return (error);
1088 vp = *vpp;
1089 len = strlen(ap->a_target);
1090 if (len < vp->v_mount->mnt_maxsymlinklen) {
1091 ip = VTOI(vp);
1092 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1093 ip->i_size = len;
1094 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1095 } else {
1097 * Make sure we have a VM object in order to use
1098 * the buffer cache.
1100 if (vp->v_object == NULL)
1101 vinitvmio(vp, 0, PAGE_SIZE, -1);
1103 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1104 UIO_SYSSPACE, IO_NODELOCKED,
1105 ap->a_cnp->cn_cred, NULL);
1107 if (error)
1108 vput(vp);
1111 return (error);
1115 * Allocate a new inode.
1117 static int
1118 ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1119 struct componentname *cnp)
1121 struct inode *ip, *pdir;
1122 struct vnode *tvp;
1123 int error;
1125 pdir = VTOI(dvp);
1126 *vpp = NULL;
1127 if ((mode & IFMT) == 0)
1128 mode |= IFREG;
1130 error = EXT2_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
1131 if (error) {
1132 return (error);
1134 ip = VTOI(tvp);
1135 ip->i_gid = pdir->i_gid;
1136 #ifdef SUIDDIR
1138 #ifdef QUOTA
1139 struct ucred ucred, *ucp;
1140 ucp = cnp->cn_cred;
1141 #endif
1143 * if we are
1144 * not the owner of the directory,
1145 * and we are hacking owners here, (only do this where told to)
1146 * and we are not giving it TOO root, (would subvert quotas)
1147 * then go ahead and give it to the other user.
1148 * Note that this drops off the execute bits for security.
1150 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1151 (pdir->i_mode & ISUID) &&
1152 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1153 ip->i_uid = pdir->i_uid;
1154 mode &= ~07111;
1155 #ifdef QUOTA
1157 * make sure the correct user gets charged
1158 * for the space.
1159 * Quickly knock up a dummy credential for the victim.
1160 * XXX This seems to never be accessed out of our
1161 * context so a stack variable is ok.
1163 ucred.cr_ref = 1;
1164 ucred.cr_uid = ip->i_uid;
1165 ucred.cr_ngroups = 1;
1166 ucred.cr_groups[0] = pdir->i_gid;
1167 ucp = &ucred;
1168 #endif
1169 } else {
1170 ip->i_uid = cnp->cn_cred->cr_uid;
1173 #ifdef QUOTA
1174 if ((error = ext2_getinoquota(ip)) ||
1175 (error = ext2_chkiq(ip, 1, ucp, 0))) {
1176 EXT2_VFREE(tvp, ip->i_number, mode);
1177 vput(tvp);
1178 return (error);
1180 #endif
1182 #else
1183 ip->i_uid = cnp->cn_cred->cr_uid;
1184 #ifdef QUOTA
1185 if ((error = ext2_getinoquota(ip)) ||
1186 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
1187 EXT2_VFREE(tvp, ip->i_number, mode);
1188 vput(tvp);
1189 return (error);
1191 #endif
1192 #endif
1193 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1194 ip->i_mode = mode;
1195 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1196 ip->i_nlink = 1;
1197 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
1198 priv_check_cred(cnp->cn_cred, PRIV_VFS_SETGID, 0))
1199 ip->i_mode &= ~ISGID;
1201 if (cnp->cn_flags & CNP_ISWHITEOUT)
1202 ip->i_flags |= UF_OPAQUE;
1205 * Regular files and directories need VM objects. Softlinks do
1206 * not (not immediately anyway).
1208 if (tvp->v_type == VREG || tvp->v_type == VDIR)
1209 vinitvmio(tvp, 0, PAGE_SIZE, -1);
1212 * Make sure inode goes to disk before directory entry.
1214 error = EXT2_UPDATE(tvp, 1);
1215 if (error)
1216 goto bad;
1217 error = ext2_direnter(ip, dvp, cnp);
1218 if (error)
1219 goto bad;
1221 *vpp = tvp;
1222 return (0);
1224 bad:
1226 * Write error occurred trying to update the inode
1227 * or the directory so must deallocate the inode.
1229 ip->i_nlink = 0;
1230 ip->i_flag |= IN_CHANGE;
1231 vput(tvp);
1232 return (error);
1235 void
1236 ext2_itimes(struct vnode *vp)
1238 struct inode *ip;
1239 struct timespec ts;
1241 ip = VTOI(vp);
1242 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
1243 return;
1244 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
1245 ip->i_flag |= IN_LAZYMOD;
1246 else
1247 ip->i_flag |= IN_MODIFIED;
1248 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
1249 vfs_timestamp(&ts);
1250 if (ip->i_flag & IN_ACCESS) {
1251 ip->i_atime = ts.tv_sec;
1252 ip->i_atimensec = ts.tv_nsec;
1254 if (ip->i_flag & IN_UPDATE) {
1255 ip->i_mtime = ts.tv_sec;
1256 ip->i_mtimensec = ts.tv_nsec;
1257 ip->i_modrev++;
1259 if (ip->i_flag & IN_CHANGE) {
1260 ip->i_ctime = ts.tv_sec;
1261 ip->i_ctimensec = ts.tv_nsec;
1264 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
1268 * Open called.
1270 * Nothing to do.
1272 * ext2_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1273 * struct file *a_fp)
1275 /* ARGSUSED */
1276 static
1278 ext2_open(struct vop_open_args *ap)
1280 struct vnode *vp = ap->a_vp;
1283 * Files marked append-only must be opened for appending.
1285 if ((VTOI(vp)->i_flags & APPEND) &&
1286 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) {
1287 return (EPERM);
1289 return (vop_stdopen(ap));
1293 * Close called.
1295 * Update the times on the inode.
1297 * ext2_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1299 /* ARGSUSED */
1300 static
1302 ext2_close(struct vop_close_args *ap)
1304 struct vnode *vp = ap->a_vp;
1306 if (vp->v_sysref.refcnt > 1)
1307 ext2_itimes(vp);
1308 return (vop_stdclose(ap));
1312 * ext2_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred)
1314 static
1316 ext2_access(struct vop_access_args *ap)
1318 struct vnode *vp = ap->a_vp;
1319 struct inode *ip = VTOI(vp);
1320 int error;
1322 #ifdef QUOTA
1323 if (ap->a_mode & VWRITE) {
1324 switch (vp->v_type) {
1325 case VDIR:
1326 case VLNK:
1327 case VREG:
1328 if ((error = ext2_getinoquota(ip)) != 0)
1329 return (error);
1330 break;
1331 default:
1332 break;
1335 #endif
1337 error = vop_helper_access(ap, ip->i_uid, ip->i_gid, ip->i_mode,
1338 ip->i_flags);
1339 return (error);
1343 * ext2_getattr(struct vnode *a_vp, struct vattr *a_vap)
1345 /* ARGSUSED */
1346 static
1348 ext2_getattr(struct vop_getattr_args *ap)
1350 struct vnode *vp = ap->a_vp;
1351 struct inode *ip = VTOI(vp);
1352 struct vattr *vap = ap->a_vap;
1354 ext2_itimes(vp);
1356 * Copy from inode table
1358 vap->va_fsid = dev2udev(ip->i_dev);
1359 vap->va_fileid = ip->i_number;
1360 vap->va_mode = ip->i_mode & ~IFMT;
1361 vap->va_nlink = VFSTOEXT2(vp->v_mount)->um_i_effnlink_valid ?
1362 ip->i_effnlink : ip->i_nlink;
1363 vap->va_uid = ip->i_uid;
1364 vap->va_gid = ip->i_gid;
1365 vap->va_rmajor = umajor(ip->i_rdev);
1366 vap->va_rminor = uminor(ip->i_rdev);
1367 vap->va_size = ip->i_din.di_size;
1368 vap->va_atime.tv_sec = ip->i_atime;
1369 vap->va_atime.tv_nsec = ip->i_atimensec;
1370 vap->va_mtime.tv_sec = ip->i_mtime;
1371 vap->va_mtime.tv_nsec = ip->i_mtimensec;
1372 vap->va_ctime.tv_sec = ip->i_ctime;
1373 vap->va_ctime.tv_nsec = ip->i_ctimensec;
1374 vap->va_flags = ip->i_flags;
1375 vap->va_gen = ip->i_gen;
1376 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1377 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
1378 vap->va_type = IFTOVT(ip->i_mode);
1379 vap->va_filerev = ip->i_modrev;
1380 return (0);
1384 * Set attribute vnode op. called from several syscalls
1386 * ext2_setattr(struct vnode *a_vp, struct vattr *a_vap, struct ucred *a_cred)
1388 static
1390 ext2_setattr(struct vop_setattr_args *ap)
1392 struct vattr *vap = ap->a_vap;
1393 struct vnode *vp = ap->a_vp;
1394 struct inode *ip = VTOI(vp);
1395 struct ucred *cred = ap->a_cred;
1396 int error;
1399 * Check for unsettable attributes.
1401 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
1402 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
1403 (vap->va_blocksize != VNOVAL) || (vap->va_rmajor != VNOVAL) ||
1404 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
1405 return (EINVAL);
1407 if (vap->va_flags != VNOVAL) {
1408 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1409 return (EROFS);
1410 if (cred->cr_uid != ip->i_uid &&
1411 (error = priv_check_cred(cred, PRIV_VFS_SETATTR, 0)))
1412 return (error);
1414 * Note that a root chflags becomes a user chflags when
1415 * we are jailed, unless the jail.chflags_allowed sysctl
1416 * is set.
1418 if (cred->cr_uid == 0 &&
1419 (!jailed(cred) || jail_chflags_allowed)) {
1420 if ((ip->i_flags
1421 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
1422 securelevel > 0)
1423 return (EPERM);
1424 ip->i_flags = vap->va_flags;
1425 } else {
1426 if (ip->i_flags
1427 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
1428 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
1429 return (EPERM);
1430 ip->i_flags &= SF_SETTABLE;
1431 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
1433 ip->i_flag |= IN_CHANGE;
1434 if (vap->va_flags & (IMMUTABLE | APPEND))
1435 return (0);
1437 if (ip->i_flags & (IMMUTABLE | APPEND))
1438 return (EPERM);
1440 * Go through the fields and update iff not VNOVAL.
1442 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
1443 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1444 return (EROFS);
1445 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred)) != 0)
1446 return (error);
1448 if (vap->va_size != VNOVAL) {
1450 * Disallow write attempts on read-only filesystems;
1451 * unless the file is a socket, fifo, or a block or
1452 * character device resident on the filesystem.
1454 switch (vp->v_type) {
1455 case VDIR:
1456 return (EISDIR);
1457 case VLNK:
1458 case VREG:
1459 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1460 return (EROFS);
1461 break;
1462 default:
1463 break;
1465 if ((error = EXT2_TRUNCATE(vp, vap->va_size, 0, cred)) != 0)
1466 return (error);
1468 ip = VTOI(vp);
1469 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1470 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1471 return (EROFS);
1472 if (cred->cr_uid != ip->i_uid &&
1473 (error = priv_check_cred(cred, PRIV_VFS_SETATTR, 0)) &&
1474 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1475 (error = VOP_EACCESS(vp, VWRITE, cred))))
1476 return (error);
1477 if (vap->va_atime.tv_sec != VNOVAL)
1478 ip->i_flag |= IN_ACCESS;
1479 if (vap->va_mtime.tv_sec != VNOVAL)
1480 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1481 ext2_itimes(vp);
1482 if (vap->va_atime.tv_sec != VNOVAL) {
1483 ip->i_atime = vap->va_atime.tv_sec;
1484 ip->i_atimensec = vap->va_atime.tv_nsec;
1486 if (vap->va_mtime.tv_sec != VNOVAL) {
1487 ip->i_mtime = vap->va_mtime.tv_sec;
1488 ip->i_mtimensec = vap->va_mtime.tv_nsec;
1490 error = EXT2_UPDATE(vp, 0);
1491 if (error)
1492 return (error);
1494 error = 0;
1495 if (vap->va_mode != (mode_t)VNOVAL) {
1496 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1497 return (EROFS);
1498 error = ext2_chmod(vp, (int)vap->va_mode, cred);
1500 VN_KNOTE(vp, NOTE_ATTRIB);
1501 return (error);
1505 * Change the mode on a file.
1506 * Inode must be locked before calling.
1508 static int
1509 ext2_chmod(struct vnode *vp, int mode, struct ucred *cred)
1511 struct inode *ip = VTOI(vp);
1512 int error;
1514 if (cred->cr_uid != ip->i_uid) {
1515 error = priv_check_cred(cred, PRIV_VFS_CHMOD, 0);
1516 if (error)
1517 return (error);
1519 if (cred->cr_uid) {
1520 if (vp->v_type != VDIR && (mode & S_ISTXT))
1521 return (EFTYPE);
1522 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
1523 return (EPERM);
1525 ip->i_mode &= ~ALLPERMS;
1526 ip->i_mode |= (mode & ALLPERMS);
1527 ip->i_flag |= IN_CHANGE;
1528 return (0);
1532 * Perform chown operation on inode ip;
1533 * inode must be locked prior to call.
1535 static int
1536 ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred)
1538 struct inode *ip = VTOI(vp);
1539 uid_t ouid;
1540 gid_t ogid;
1541 int error = 0;
1542 #ifdef QUOTA
1543 int i;
1544 long change;
1545 #endif
1547 if (uid == (uid_t)VNOVAL)
1548 uid = ip->i_uid;
1549 if (gid == (gid_t)VNOVAL)
1550 gid = ip->i_gid;
1552 * If we don't own the file, are trying to change the owner
1553 * of the file, or are not a member of the target group,
1554 * the caller must be superuser or the call fails.
1556 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
1557 (gid != ip->i_gid && !(cred->cr_gid == gid ||
1558 groupmember((gid_t)gid, cred)))) &&
1559 (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
1560 return (error);
1561 ogid = ip->i_gid;
1562 ouid = ip->i_uid;
1563 #ifdef QUOTA
1564 if ((error = ext2_getinoquota(ip)) != 0)
1565 return (error);
1566 if (ouid == uid) {
1567 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1568 ip->i_dquot[USRQUOTA] = NODQUOT;
1570 if (ogid == gid) {
1571 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1572 ip->i_dquot[GRPQUOTA] = NODQUOT;
1574 change = ip->i_blocks;
1575 (void) ext2_chkdq(ip, -change, cred, CHOWN);
1576 (void) ext2_chkiq(ip, -1, cred, CHOWN);
1577 for (i = 0; i < MAXQUOTAS; i++) {
1578 ext2_dqrele(vp, ip->i_dquot[i]);
1579 ip->i_dquot[i] = NODQUOT;
1581 #endif
1582 ip->i_gid = gid;
1583 ip->i_uid = uid;
1584 #ifdef QUOTA
1585 if ((error = ext2_getinoquota(ip)) == 0) {
1586 if (ouid == uid) {
1587 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1588 ip->i_dquot[USRQUOTA] = NODQUOT;
1590 if (ogid == gid) {
1591 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1592 ip->i_dquot[GRPQUOTA] = NODQUOT;
1594 if ((error = ext2_chkdq(ip, change, cred, CHOWN)) == 0) {
1595 if ((error = ext2_chkiq(ip, 1, cred, CHOWN)) == 0)
1596 goto good;
1597 else
1598 (void)ext2_chkdq(ip, -change, cred, CHOWN|FORCE);
1600 for (i = 0; i < MAXQUOTAS; i++) {
1601 ext2_dqrele(vp, ip->i_dquot[i]);
1602 ip->i_dquot[i] = NODQUOT;
1605 ip->i_gid = ogid;
1606 ip->i_uid = ouid;
1607 if (ext2_getinoquota(ip) == 0) {
1608 if (ouid == uid) {
1609 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1610 ip->i_dquot[USRQUOTA] = NODQUOT;
1612 if (ogid == gid) {
1613 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1614 ip->i_dquot[GRPQUOTA] = NODQUOT;
1616 (void) ext2_chkdq(ip, change, cred, FORCE|CHOWN);
1617 (void) ext2_chkiq(ip, 1, cred, FORCE|CHOWN);
1618 (void) ext2_getinoquota(ip);
1620 return (error);
1621 good:
1622 if (ext2_getinoquota(ip))
1623 panic("ext2_chown: lost quota");
1624 #endif /* QUOTA */
1625 ip->i_flag |= IN_CHANGE;
1626 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
1627 ip->i_mode &= ~(ISUID | ISGID);
1628 return (0);
1632 * Mmap a file
1634 * NB Currently unsupported.
1636 * ext2_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred)
1638 /* ARGSUSED */
1639 static
1641 ext2_mmap(struct vop_mmap_args *ap)
1643 return (EINVAL);
1647 * whiteout vnode call
1649 * ext2_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
1651 static
1653 ext2_whiteout(struct vop_old_whiteout_args *ap)
1655 return (EOPNOTSUPP);
1659 * Return target name of a symbolic link
1661 * ext2_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1663 static
1665 ext2_readlink(struct vop_readlink_args *ap)
1667 struct vnode *vp = ap->a_vp;
1668 struct inode *ip = VTOI(vp);
1669 int isize;
1671 isize = ip->i_size;
1672 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1673 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1674 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1675 return (0);
1677 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1681 * Calculate the logical to physical mapping if not done already,
1682 * then call the device strategy routine.
1684 * In order to be able to swap to a file, the VOP_BMAP operation may not
1685 * deadlock on memory. See ext2_bmap() for details.
1687 * ext2_strategy(struct vnode *a_vp, struct bio *a_bio)
1689 static
1691 ext2_strategy(struct vop_strategy_args *ap)
1693 struct bio *bio = ap->a_bio;
1694 struct bio *nbio;
1695 struct buf *bp = bio->bio_buf;
1696 struct vnode *vp = ap->a_vp;
1697 struct inode *ip;
1698 int error;
1700 ip = VTOI(vp);
1701 if (vp->v_type == VBLK || vp->v_type == VCHR)
1702 panic("ext2_strategy: spec");
1703 nbio = push_bio(bio);
1704 if (nbio->bio_offset == NOOFFSET) {
1705 error = VOP_BMAP(vp, bio->bio_offset, &nbio->bio_offset,
1706 NULL, NULL, bp->b_cmd);
1707 if (error) {
1708 bp->b_error = error;
1709 bp->b_flags |= B_ERROR;
1710 /* I/O was never started on nbio, must biodone(bio) */
1711 biodone(bio);
1712 return (error);
1714 if (nbio->bio_offset == NOOFFSET)
1715 vfs_bio_clrbuf(bp);
1717 if (nbio->bio_offset == NOOFFSET) {
1718 /* I/O was never started on nbio, must biodone(bio) */
1719 biodone(bio);
1720 return (0);
1722 vn_strategy(ip->i_devvp, nbio);
1723 return (0);
1727 * Print out the contents of an inode.
1729 * ext2_print(struct vnode *a_vp)
1731 static
1733 ext2_print(struct vop_print_args *ap)
1735 struct vnode *vp = ap->a_vp;
1736 struct inode *ip = VTOI(vp);
1738 kprintf("tag VT_EXT2FS, ino %lu, on dev %s (%d, %d)",
1739 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1740 minor(ip->i_dev));
1741 if (vp->v_type == VFIFO)
1742 fifo_printinfo(vp);
1743 lockmgr_printinfo(&vp->v_lock);
1744 kprintf("\n");
1745 return (0);
1749 * Read wrapper for fifos.
1751 * ext2fifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1752 * struct ucred *a_cred)
1754 static
1756 ext2fifo_read(struct vop_read_args *ap)
1758 int error, resid;
1759 struct inode *ip;
1760 struct uio *uio;
1762 uio = ap->a_uio;
1763 resid = uio->uio_resid;
1764 error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1765 ip = VTOI(ap->a_vp);
1766 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1767 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1768 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1769 return (error);
1773 * Write wrapper for fifos.
1775 * ext2fifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1776 * struct ucred *a_cred)
1778 static
1780 ext2fifo_write(struct vop_write_args *ap)
1782 int error, resid;
1783 struct inode *ip;
1784 struct uio *uio;
1786 uio = ap->a_uio;
1787 resid = uio->uio_resid;
1788 error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1789 ip = VTOI(ap->a_vp);
1790 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1791 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1792 return (error);
1796 * Close wrapper for fifos.
1798 * Update the times on the inode then do device close.
1800 * ext2fifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1802 static
1804 ext2fifo_close(struct vop_close_args *ap)
1806 struct vnode *vp = ap->a_vp;
1808 if (vp->v_sysref.refcnt > 1)
1809 ext2_itimes(vp);
1810 return (VOCALL(&fifo_vnode_vops, &ap->a_head));
1814 * Kqfilter wrapper for fifos.
1816 * Fall through to ext2 kqfilter routines if needed
1818 static
1820 ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1822 int error;
1824 error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1825 if (error)
1826 error = ext2_kqfilter(ap);
1827 return (error);
1831 * Return POSIX pathconf information applicable to ext2 filesystems.
1833 * ext2_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
1835 static
1837 ext2_pathconf(struct vop_pathconf_args *ap)
1839 switch (ap->a_name) {
1840 case _PC_LINK_MAX:
1841 *ap->a_retval = LINK_MAX;
1842 return (0);
1843 case _PC_NAME_MAX:
1844 *ap->a_retval = NAME_MAX;
1845 return (0);
1846 case _PC_PATH_MAX:
1847 *ap->a_retval = PATH_MAX;
1848 return (0);
1849 case _PC_PIPE_BUF:
1850 *ap->a_retval = PIPE_BUF;
1851 return (0);
1852 case _PC_CHOWN_RESTRICTED:
1853 *ap->a_retval = 1;
1854 return (0);
1855 case _PC_NO_TRUNC:
1856 *ap->a_retval = 1;
1857 return (0);
1858 default:
1859 return (EINVAL);
1861 /* NOTREACHED */
1865 * Advisory record locking support
1867 * ext2_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
1868 * int a_flags)
1870 static
1872 ext2_advlock(struct vop_advlock_args *ap)
1874 struct inode *ip = VTOI(ap->a_vp);
1876 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1880 * Initialize the vnode associated with a new inode, handle aliased
1881 * vnodes.
1884 ext2_vinit(struct mount *mntp, struct vnode **vpp)
1886 struct inode *ip;
1887 struct vnode *vp;
1888 struct timeval tv;
1890 vp = *vpp;
1891 ip = VTOI(vp);
1893 switch(vp->v_type = IFTOVT(ip->i_mode)) {
1894 case VCHR:
1895 case VBLK:
1896 vp->v_ops = &mntp->mnt_vn_spec_ops;
1897 addaliasu(vp, umajor(ip->i_rdev), uminor(ip->i_rdev));
1898 break;
1899 case VFIFO:
1900 vp->v_ops = &mntp->mnt_vn_fifo_ops;
1901 break;
1902 case VDIR:
1903 case VREG:
1904 vinitvmio(vp, ip->i_size, PAGE_SIZE, -1); /* XXX */
1905 break;
1906 case VLNK:
1907 if ((ip->i_size >= vp->v_mount->mnt_maxsymlinklen) &&
1908 ip->i_din.di_blocks != 0
1910 vinitvmio(vp, ip->i_size, PAGE_SIZE, -1);
1912 break;
1913 default:
1914 break;
1918 if (ip->i_number == ROOTINO)
1919 vp->v_flag |= VROOT;
1921 * Initialize modrev times
1923 getmicrouptime(&tv);
1924 SETHIGH(ip->i_modrev, tv.tv_sec);
1925 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
1926 *vpp = vp;
1927 return (0);
1930 static struct filterops ext2read_filtops =
1931 { 1, NULL, filt_ext2detach, filt_ext2read };
1932 static struct filterops ext2write_filtops =
1933 { 1, NULL, filt_ext2detach, filt_ext2write };
1934 static struct filterops ext2vnode_filtops =
1935 { 1, NULL, filt_ext2detach, filt_ext2vnode };
1938 * ext2_kqfilter(struct vnode *a_vp, struct knote *a_kn)
1940 static int
1941 ext2_kqfilter(struct vop_kqfilter_args *ap)
1943 struct vnode *vp = ap->a_vp;
1944 struct knote *kn = ap->a_kn;
1945 lwkt_tokref vlock;
1947 switch (kn->kn_filter) {
1948 case EVFILT_READ:
1949 kn->kn_fop = &ext2read_filtops;
1950 break;
1951 case EVFILT_WRITE:
1952 kn->kn_fop = &ext2write_filtops;
1953 break;
1954 case EVFILT_VNODE:
1955 kn->kn_fop = &ext2vnode_filtops;
1956 break;
1957 default:
1958 return (1);
1961 kn->kn_hook = (caddr_t)vp;
1963 lwkt_gettoken(&vlock, &vp->v_token);
1964 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
1965 lwkt_reltoken(&vlock);
1967 return (0);
1970 static void
1971 filt_ext2detach(struct knote *kn)
1973 struct vnode *vp = (struct vnode *)kn->kn_hook;
1974 lwkt_tokref vlock;
1976 lwkt_gettoken(&vlock, &vp->v_token);
1977 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
1978 kn, knote, kn_selnext);
1979 lwkt_reltoken(&vlock);
1982 /*ARGSUSED*/
1983 static int
1984 filt_ext2read(struct knote *kn, long hint)
1986 struct vnode *vp = (struct vnode *)kn->kn_hook;
1987 struct inode *ip = VTOI(vp);
1990 * filesystem is gone, so set the EOF flag and schedule
1991 * the knote for deletion.
1993 if (hint == NOTE_REVOKE) {
1994 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1995 return (1);
1998 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
1999 return (kn->kn_data != 0);
2002 /*ARGSUSED*/
2003 static int
2004 filt_ext2write(struct knote *kn, long hint)
2007 * filesystem is gone, so set the EOF flag and schedule
2008 * the knote for deletion.
2010 if (hint == NOTE_REVOKE)
2011 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2013 kn->kn_data = 0;
2014 return (1);
2017 static int
2018 filt_ext2vnode(struct knote *kn, long hint)
2020 if (kn->kn_sfflags & hint)
2021 kn->kn_fflags |= hint;
2022 if (hint == NOTE_REVOKE) {
2023 kn->kn_flags |= EV_EOF;
2024 return (1);
2026 return (kn->kn_fflags != 0);
2029 struct vop_ops ext2_vnode_vops = {
2030 .vop_default = vop_defaultop,
2031 .vop_fsync = ext2_fsync,
2032 .vop_read = ext2_read,
2033 .vop_reallocblks = ext2_reallocblks,
2034 .vop_write = ext2_write,
2035 .vop_access = ext2_access,
2036 .vop_advlock = ext2_advlock,
2037 .vop_bmap = ext2_bmap,
2038 .vop_old_lookup = ext2_lookup,
2039 .vop_close = ext2_close,
2040 .vop_old_create = ext2_create,
2041 .vop_getattr = ext2_getattr,
2042 .vop_inactive = ext2_inactive,
2043 .vop_old_link = ext2_link,
2044 .vop_old_mkdir = ext2_mkdir,
2045 .vop_old_mknod = ext2_mknod,
2046 .vop_mmap = ext2_mmap,
2047 .vop_open = ext2_open,
2048 .vop_pathconf = ext2_pathconf,
2049 .vop_poll = vop_stdpoll,
2050 .vop_kqfilter = ext2_kqfilter,
2051 .vop_print = ext2_print,
2052 .vop_readdir = ext2_readdir,
2053 .vop_readlink = ext2_readlink,
2054 .vop_reclaim = ext2_reclaim,
2055 .vop_old_remove = ext2_remove,
2056 .vop_old_rename = ext2_rename,
2057 .vop_old_rmdir = ext2_rmdir,
2058 .vop_setattr = ext2_setattr,
2059 .vop_strategy = ext2_strategy,
2060 .vop_old_symlink = ext2_symlink,
2061 .vop_old_whiteout = ext2_whiteout,
2062 .vop_getpages = vop_stdgetpages,
2063 .vop_putpages = vop_stdputpages
2066 struct vop_ops ext2_spec_vops = {
2067 .vop_default = vop_defaultop,
2068 .vop_fsync = ext2_fsync,
2069 .vop_access = ext2_access,
2070 .vop_close = ext2_close,
2071 .vop_getattr = ext2_getattr,
2072 .vop_inactive = ext2_inactive,
2073 .vop_print = ext2_print,
2074 .vop_read = vop_stdnoread,
2075 .vop_reclaim = ext2_reclaim,
2076 .vop_setattr = ext2_setattr,
2077 .vop_write = vop_stdnowrite
2080 struct vop_ops ext2_fifo_vops = {
2081 .vop_default = ext2_vnoperatefifo,
2082 .vop_fsync = ext2_fsync,
2083 .vop_access = ext2_access,
2084 .vop_close = ext2fifo_close,
2085 .vop_getattr = ext2_getattr,
2086 .vop_inactive = ext2_inactive,
2087 .vop_kqfilter = ext2fifo_kqfilter,
2088 .vop_print = ext2_print,
2089 .vop_read = ext2fifo_read,
2090 .vop_reclaim = ext2_reclaim,
2091 .vop_setattr = ext2_setattr,
2092 .vop_write = ext2fifo_write
2095 VNODEOP_SET(ext2_vnode_vops);
2096 VNODEOP_SET(ext2_spec_vops);
2097 VNODEOP_SET(ext2_fifo_vops);
2100 * ext2_vnoperate()
2103 ext2_vnoperate(struct vop_generic_args *ap)
2105 return (VOCALL(&ext2_vnode_vops, ap));
2109 * ext2_vnoperatefifo()
2112 ext2_vnoperatefifo(struct vop_generic_args *ap)
2114 return (VOCALL(&ext2_fifo_vops, ap));