kernel - Do a better job with the filesystem background sync
[dragonfly.git] / sys / vfs / gnu / ext2fs / ext2_vnops.c
CommitLineData
984263bc
MD
1/*
2 * modified for EXT2FS support in Lites 1.1
3 *
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.
15 *
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.
31 *
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.
43 *
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 $
e92ca23a 47 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vnops.c,v 1.43 2008/06/19 23:27:39 dillon Exp $
984263bc
MD
48 */
49
50#include "opt_quota.h"
51#include "opt_suiddir.h"
52
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>
1f1db49f 58#include <sys/fcntl.h>
984263bc 59#include <sys/buf.h>
1f1db49f 60#include <sys/stat.h>
984263bc 61#include <sys/proc.h>
895c1f85 62#include <sys/priv.h>
984263bc
MD
63#include <sys/mount.h>
64#include <sys/time.h>
65#include <sys/vnode.h>
1f1db49f
MD
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>
984263bc 72#include <sys/namei.h>
1f1db49f
MD
73#include <sys/signalvar.h>
74#include <sys/unistd.h>
984263bc
MD
75
76#include <vm/vm.h>
77#include <vm/vm_extern.h>
78#include <vm/vm_zone.h>
79#include <vm/vnode_pager.h>
1f1db49f 80
7b95be2a 81#include <sys/buf2.h>
61670a01 82#include <sys/thread2.h>
984263bc 83
1f1db49f 84#include <vfs/fifofs/fifo.h>
984263bc 85
1f1db49f
MD
86#include "dir.h"
87#include "quota.h"
88#include "inode.h"
89#include "ext2mount.h"
1f2de5d4
MD
90#include "ext2_fs_sb.h"
91#include "fs.h"
92#include "ext2_extern.h"
93#include "ext2_fs.h"
984263bc 94
1f1db49f
MD
95static int ext2_access (struct vop_access_args *);
96static int ext2_advlock (struct vop_advlock_args *);
87de5057
MD
97static int ext2_chmod (struct vnode *, int, struct ucred *);
98static int ext2_chown (struct vnode *, uid_t, gid_t, struct ucred *);
1f1db49f
MD
99static int ext2_close (struct vop_close_args *);
100static int ext2_getattr (struct vop_getattr_args *);
a6ee311a 101static int ext2_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
1f1db49f
MD
102static int ext2_mmap (struct vop_mmap_args *);
103static int ext2_open (struct vop_open_args *);
104static int ext2_pathconf (struct vop_pathconf_args *);
105static int ext2_print (struct vop_print_args *);
106static int ext2_readlink (struct vop_readlink_args *);
107static int ext2_setattr (struct vop_setattr_args *);
108static int ext2_strategy (struct vop_strategy_args *);
109static int ext2_whiteout (struct vop_old_whiteout_args *);
110static int filt_ext2read (struct knote *kn, long hint);
111static int filt_ext2write (struct knote *kn, long hint);
112static int filt_ext2vnode (struct knote *kn, long hint);
113static void filt_ext2detach (struct knote *kn);
114static int ext2_kqfilter (struct vop_kqfilter_args *ap);
1f1db49f
MD
115static int ext2fifo_close (struct vop_close_args *);
116static int ext2fifo_kqfilter (struct vop_kqfilter_args *);
117static int ext2fifo_read (struct vop_read_args *);
118static int ext2fifo_write (struct vop_write_args *);
984263bc 119
a6ee311a
RG
120static int ext2_fsync (struct vop_fsync_args *);
121static int ext2_read (struct vop_read_args *);
122static int ext2_write (struct vop_write_args *);
e62afb5f
MD
123static int ext2_remove (struct vop_old_remove_args *);
124static int ext2_link (struct vop_old_link_args *);
125static int ext2_rename (struct vop_old_rename_args *);
126static int ext2_mkdir (struct vop_old_mkdir_args *);
127static int ext2_rmdir (struct vop_old_rmdir_args *);
128static int ext2_create (struct vop_old_create_args *);
129static int ext2_mknod (struct vop_old_mknod_args *);
130static int ext2_symlink (struct vop_old_symlink_args *);
984263bc 131
1f1db49f 132#include "ext2_readwrite.c"
984263bc 133
1f1db49f
MD
134union _qcvt {
135 int64_t qcvt;
136 int32_t val[2];
984263bc 137};
1f1db49f
MD
138#define SETHIGH(q, h) { \
139 union _qcvt tmp; \
140 tmp.qcvt = (q); \
141 tmp.val[_QUAD_HIGHWORD] = (h); \
142 (q) = tmp.qcvt; \
143}
144#define SETLOW(q, l) { \
145 union _qcvt tmp; \
146 tmp.qcvt = (q); \
147 tmp.val[_QUAD_LOWWORD] = (l); \
148 (q) = tmp.qcvt; \
149}
150#define VN_KNOTE(vp, b) \
5b22f1a7 151 KNOTE(&vp->v_pollinfo.vpi_kqinfo.ki_note, (b))
984263bc 152
1f1db49f 153#define OFSFMT(vp) ((vp)->v_mount->mnt_maxsymlinklen <= 0)
984263bc
MD
154
155/*
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.
160 */
1f1db49f 161static struct dirtemplate ext2_mastertemplate = {
984263bc
MD
162 0, 12, 1, EXT2_FT_DIR, ".",
163 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
164};
1f1db49f 165static struct dirtemplate ext2_omastertemplate = {
984263bc
MD
166 0, 12, 1, EXT2_FT_UNKNOWN, ".",
167 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
168};
169
170/*
171 * Create a regular file
0f7f7a49
CP
172 *
173 * ext2_create(struct vnode *a_dvp, struct vnode **a_vpp,
174 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
175 */
176static int
e62afb5f 177ext2_create(struct vop_old_create_args *ap)
984263bc
MD
178{
179 int error;
180
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);
187}
188
189/*
190 * Synch an open file.
0f7f7a49
CP
191 *
192 * ext2_fsync(struct vnode *a_vp, struct ucred *a_cred, int a_waitfor,
193 * struct proc *a_p)
984263bc
MD
194 */
195/* ARGSUSED */
6bae6177
MD
196
197static int ext2_fsync_bp(struct buf *bp, void *data);
198
199struct ext2_fsync_bp_info {
200 struct vnode *vp;
201 int waitfor;
6bae6177
MD
202};
203
984263bc 204static int
0f7f7a49 205ext2_fsync(struct vop_fsync_args *ap)
984263bc 206{
6bae6177 207 struct ext2_fsync_bp_info info;
f7aae92f 208 struct vnode *vp = ap->a_vp;
6bae6177 209 int count;
984263bc
MD
210
211 /*
212 * XXX why is all this fs specific?
213 */
214
215 /*
216 * Flush all dirty buffers associated with a vnode.
217 */
218 ext2_discard_prealloc(VTOI(vp));
219
3b998fa9 220 lwkt_gettoken(&vp->v_token);
6bae6177 221 info.vp = vp;
984263bc 222loop:
6bae6177
MD
223 info.waitfor = ap->a_waitfor;
224 count = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
225 ext2_fsync_bp, &info);
226 if (count)
984263bc 227 goto loop;
6bae6177 228
984263bc 229 if (ap->a_waitfor == MNT_WAIT) {
a9a20f98 230 bio_track_wait(&vp->v_track_write, 0, 0);
984263bc 231#if DIAGNOSTIC
6bae6177 232 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
984263bc
MD
233 vprint("ext2_fsync: dirty", vp);
234 goto loop;
235 }
236#endif
237 }
3b998fa9 238 lwkt_reltoken(&vp->v_token);
1f1db49f 239 return (EXT2_UPDATE(ap->a_vp, ap->a_waitfor == MNT_WAIT));
984263bc
MD
240}
241
6bae6177
MD
242static int
243ext2_fsync_bp(struct buf *bp, void *data)
244{
245 struct ext2_fsync_bp_info *info = data;
246
247 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
248 return(0);
249 if ((bp->b_flags & B_DELWRI) == 0)
250 panic("ext2_fsync: not dirty");
251 bremfree(bp);
6bae6177
MD
252
253 /*
254 * Wait for I/O associated with indirect blocks to complete,
255 * since there is no way to quickly wait for them below.
256 */
28271622 257 if (bp->b_vp == info->vp || (info->waitfor & MNT_NOWAIT))
b1ce5639 258 bawrite(bp);
6bae6177 259 else
b1ce5639 260 bwrite(bp);
6bae6177
MD
261 return(1);
262}
263
984263bc
MD
264/*
265 * Mknod vnode call
0f7f7a49
CP
266 *
267 * ext2_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
268 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
269 */
270/* ARGSUSED */
271static int
e62afb5f 272ext2_mknod(struct vop_old_mknod_args *ap)
984263bc
MD
273{
274 struct vattr *vap = ap->a_vap;
275 struct vnode **vpp = ap->a_vpp;
276 struct inode *ip;
277 ino_t ino;
278 int error;
279
0e9b9130
MD
280 if (vap->va_rmajor != VNOVAL &&
281 makeudev(vap->va_rmajor, vap->va_rminor) == NOUDEV) {
282 return (EINVAL);
283 }
284
984263bc
MD
285 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
286 ap->a_dvp, vpp, ap->a_cnp);
287 if (error)
288 return (error);
289 ip = VTOI(*vpp);
290 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
0e9b9130 291 if (vap->va_rmajor != VNOVAL) {
984263bc
MD
292 /*
293 * Want to be able to use this to make badblock
294 * inodes, so don't truncate the dev number.
295 */
0e9b9130 296 ip->i_rdev = makeudev(vap->va_rmajor, vap->va_rminor);
984263bc
MD
297 }
298 /*
299 * Remove inode, then reload it through VFS_VGET so it is
300 * checked to see if it is an alias of an existing entry in
301 * the inode cache.
302 */
984263bc
MD
303 (*vpp)->v_type = VNON;
304 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
3c37c940 305 vgone_vxlocked(*vpp);
5fd012e0 306 vput(*vpp);
b9b0a6d0 307 error = VFS_VGET(ap->a_dvp->v_mount, NULL, ino, vpp);
984263bc
MD
308 if (error) {
309 *vpp = NULL;
310 return (error);
311 }
312 return (0);
313}
314
0f7f7a49
CP
315/*
316 * ext2_remove(struct vnode *a_dvp, struct vnode *a_vp,
317 * struct componentname *a_cnp)
318 */
984263bc 319static int
e62afb5f 320ext2_remove(struct vop_old_remove_args *ap)
984263bc
MD
321{
322 struct inode *ip;
323 struct vnode *vp = ap->a_vp;
324 struct vnode *dvp = ap->a_dvp;
325 int error;
326
327 ip = VTOI(vp);
328 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
329 (VTOI(dvp)->i_flags & APPEND)) {
330 error = EPERM;
331 goto out;
332 }
333 error = ext2_dirremove(dvp, ap->a_cnp);
334 if (error == 0) {
335 ip->i_nlink--;
336 ip->i_flag |= IN_CHANGE;
337 }
338out:
339 return (error);
340}
341
342/*
343 * link vnode call
0f7f7a49
CP
344 *
345 * ext2_link(struct vnode *a_tdvp, struct vnode *a_vp,
346 * struct componentname *a_cnp)
984263bc
MD
347 */
348static int
e62afb5f 349ext2_link(struct vop_old_link_args *ap)
984263bc
MD
350{
351 struct vnode *vp = ap->a_vp;
352 struct vnode *tdvp = ap->a_tdvp;
353 struct componentname *cnp = ap->a_cnp;
984263bc
MD
354 struct inode *ip;
355 int error;
356
984263bc
MD
357 if (tdvp->v_mount != vp->v_mount) {
358 error = EXDEV;
359 goto out2;
360 }
ca466bae 361 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE))) {
984263bc
MD
362 goto out2;
363 }
364 ip = VTOI(vp);
365 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
366 error = EMLINK;
367 goto out1;
368 }
369 if (ip->i_flags & (IMMUTABLE | APPEND)) {
370 error = EPERM;
371 goto out1;
372 }
373 ip->i_nlink++;
374 ip->i_flag |= IN_CHANGE;
1f1db49f 375 error = EXT2_UPDATE(vp, 1);
984263bc
MD
376 if (!error)
377 error = ext2_direnter(ip, tdvp, cnp);
378 if (error) {
379 ip->i_nlink--;
380 ip->i_flag |= IN_CHANGE;
381 }
382out1:
383 if (tdvp != vp)
a11aaa81 384 vn_unlock(vp);
984263bc
MD
385out2:
386 return (error);
387}
388
389/*
fad57d0e
MD
390 * Rename system call. fdvp, fvp are ref'd. tvp, tdvp are ref'd and locked.
391 * all vp's are released and must be in an unlocked state on return.
0f7f7a49
CP
392 *
393 * ext2_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
394 * struct componentname *a_fcnp, struct vnode *a_tdvp,
395 * struct vnode *a_tvp, struct componentname *a_tcnp)
984263bc
MD
396 */
397static int
e62afb5f 398ext2_rename(struct vop_old_rename_args *ap)
984263bc
MD
399{
400 struct vnode *tvp = ap->a_tvp;
f7aae92f 401 struct vnode *tdvp = ap->a_tdvp;
984263bc
MD
402 struct vnode *fvp = ap->a_fvp;
403 struct vnode *fdvp = ap->a_fdvp;
404 struct componentname *tcnp = ap->a_tcnp;
405 struct componentname *fcnp = ap->a_fcnp;
984263bc
MD
406 struct inode *ip, *xp, *dp;
407 struct dirtemplate dirbuf;
408 int doingdirectory = 0, oldparent = 0, newparent = 0;
409 int error = 0;
410 u_char namlen;
411
984263bc
MD
412 /*
413 * Check for cross-device rename.
414 */
415 if ((fvp->v_mount != tdvp->v_mount) ||
fad57d0e
MD
416 (tvp && (fvp->v_mount != tvp->v_mount)) ||
417 tvp == tdvp) {
984263bc
MD
418 error = EXDEV;
419abortit:
420 if (tdvp == tvp)
421 vrele(tdvp);
422 else
423 vput(tdvp);
424 if (tvp)
425 vput(tvp);
426 vrele(fdvp);
427 vrele(fvp);
428 return (error);
429 }
430
431 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
432 (VTOI(tdvp)->i_flags & APPEND))) {
433 error = EPERM;
434 goto abortit;
435 }
436
437 /*
438 * Renaming a file to itself has no effect. The upper layers should
439 * not call us in that case. Temporarily just warn if they do.
440 */
441 if (fvp == tvp) {
984263bc
MD
442 error = 0;
443 goto abortit;
444 }
445
ca466bae 446 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
984263bc 447 goto abortit;
fad57d0e
MD
448
449 /*
450 * fvp, tvp, tdvp locked. fdvp not locked but note that fdvp may
451 * be equal to tdvp.
452 */
984263bc
MD
453 dp = VTOI(fdvp);
454 ip = VTOI(fvp);
455 if (ip->i_nlink >= LINK_MAX) {
a11aaa81 456 vn_unlock(fvp);
984263bc
MD
457 error = EMLINK;
458 goto abortit;
459 }
460 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
461 || (dp->i_flags & APPEND)) {
a11aaa81 462 vn_unlock(fvp);
984263bc
MD
463 error = EPERM;
464 goto abortit;
465 }
466 if ((ip->i_mode & IFMT) == IFDIR) {
467 /*
468 * Avoid ".", "..", and aliases of "." for obvious reasons.
469 */
470 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
2b69e610 471 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
984263bc 472 (ip->i_flag & IN_RENAME)) {
a11aaa81 473 vn_unlock(fvp);
984263bc
MD
474 error = EINVAL;
475 goto abortit;
476 }
477 ip->i_flag |= IN_RENAME;
478 oldparent = dp->i_number;
479 doingdirectory++;
480 }
984263bc
MD
481
482 /*
fad57d0e
MD
483 * tvp is non-NULL if the target exists. fvp is still locked but
484 * we will unlock it soon. The 'bad' goto target requires dp and
485 * xp to be correctly assigned.
984263bc
MD
486 */
487 dp = VTOI(tdvp);
984263bc
MD
488 if (tvp)
489 xp = VTOI(tvp);
fad57d0e
MD
490 else
491 xp = NULL;
984263bc
MD
492
493 /*
494 * 1) Bump link count while we're moving stuff
495 * around. If we crash somewhere before
496 * completing our work, the link count
497 * may be wrong, but correctable.
498 */
499 ip->i_nlink++;
500 ip->i_flag |= IN_CHANGE;
1f1db49f 501 if ((error = EXT2_UPDATE(fvp, 1)) != 0) {
a11aaa81 502 vn_unlock(fvp);
984263bc
MD
503 goto bad;
504 }
505
506 /*
507 * If ".." must be changed (ie the directory gets a new
508 * parent) then the source directory must not be in the
509 * directory heirarchy above the target, as this would
510 * orphan everything below the source directory. Also
511 * the user must have write permission in the source so
512 * as to be able to change "..". We must repeat the call
513 * to namei, as the parent directory is unlocked by the
514 * call to checkpath().
515 */
cb66845a 516 error = VOP_EACCESS(fvp, VWRITE, tcnp->cn_cred);
a11aaa81 517 vn_unlock(fvp);
fad57d0e
MD
518
519 /*
520 * tvp (if not NULL) and tdvp are locked. fvp and fdvp are not.
521 * dp and xp are set according to tdvp and tvp.
522 */
984263bc
MD
523 if (oldparent != dp->i_number)
524 newparent = dp->i_number;
525 if (doingdirectory && newparent) {
526 if (error) /* write access check above */
527 goto bad;
fad57d0e
MD
528
529 /*
530 * Prepare for relookup, get rid of xp
531 */
532 if (xp != NULL) {
984263bc 533 vput(tvp);
fad57d0e
MD
534 xp = NULL;
535 }
536
537 /*
538 * checkpath vput()'s tdvp (VTOI(dp)) on return no matter what,
539 * get an extra ref so we wind up with just an unlocked, ref'd
540 * tdvp. The 'out' target skips xp and tdvp cleanups. Our
541 * tdvp is now unlocked so we have to clean it up ourselves.
542 */
543 vref(tdvp);
984263bc 544 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
6970b52a 545 tcnp->cn_flags |= CNP_PDIRUNLOCK;
fad57d0e
MD
546 if (error) {
547 vrele(tdvp);
984263bc 548 goto out;
fad57d0e
MD
549 }
550 /*
551 * relookup no longer messes with the ref count. An unlocked
552 * tdvp must be passed and if no error occurs a locked tdvp
553 * will be returned. We have to use the out target again.
554 */
984263bc 555 error = relookup(tdvp, &tvp, tcnp);
fad57d0e
MD
556 if (error) {
557 if (tcnp->cn_flags & CNP_PDIRUNLOCK)
558 vrele(tdvp);
559 else
560 vput(tdvp);
984263bc 561 goto out;
fad57d0e
MD
562 }
563
564 /*
565 * tdvp is locked at this point. in the RENAME case tvp may
566 * be NULL without an error, assign xp accordingly. The
567 * 'bad' target can be used again after this.
568 */
984263bc 569 dp = VTOI(tdvp);
984263bc
MD
570 if (tvp)
571 xp = VTOI(tvp);
572 }
573 /*
574 * 2) If target doesn't exist, link the target
575 * to the source and unlink the source.
576 * Otherwise, rewrite the target directory
577 * entry to reference the source inode and
578 * expunge the original entry's existence.
fad57d0e
MD
579 *
580 * tdvp and tvp are cleaned up by this code. tvp is only good if
581 * xp is not NULL.
984263bc
MD
582 */
583 if (xp == NULL) {
584 if (dp->i_dev != ip->i_dev)
1f1db49f 585 panic("ext2_rename: EXDEV");
984263bc
MD
586 /*
587 * Account for ".." in new directory.
588 * When source and destination have the same
589 * parent we don't fool with the link count.
590 */
591 if (doingdirectory && newparent) {
592 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
593 error = EMLINK;
594 goto bad;
595 }
596 dp->i_nlink++;
597 dp->i_flag |= IN_CHANGE;
1f1db49f 598 error = EXT2_UPDATE(tdvp, 1);
984263bc
MD
599 if (error)
600 goto bad;
601 }
602 error = ext2_direnter(ip, tdvp, tcnp);
603 if (error) {
604 if (doingdirectory && newparent) {
605 dp->i_nlink--;
606 dp->i_flag |= IN_CHANGE;
1f1db49f 607 EXT2_UPDATE(tdvp, 1);
984263bc
MD
608 }
609 goto bad;
610 }
fad57d0e
MD
611
612 /*
613 * manual cleanup, we can't use the bad or out target after
614 * this.
615 */
984263bc
MD
616 vput(tdvp);
617 } else {
618 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1f1db49f 619 panic("ext2_rename: EXDEV");
984263bc
MD
620 /*
621 * Short circuit rename(foo, foo).
622 */
623 if (xp->i_number == ip->i_number)
1f1db49f 624 panic("ext2_rename: same file");
984263bc
MD
625 /*
626 * If the parent directory is "sticky", then the user must
627 * own the parent directory, or the destination of the rename,
628 * otherwise the destination may not be changed (except by
629 * root). This implements append-only directories.
630 */
631 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
632 tcnp->cn_cred->cr_uid != dp->i_uid &&
633 xp->i_uid != tcnp->cn_cred->cr_uid) {
634 error = EPERM;
635 goto bad;
636 }
637 /*
638 * Target must be empty if a directory and have no links
639 * to it. Also, ensure source and target are compatible
640 * (both directories, or both not directories).
641 */
642 if ((xp->i_mode&IFMT) == IFDIR) {
643 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
644 xp->i_nlink > 2) {
645 error = ENOTEMPTY;
646 goto bad;
647 }
648 if (!doingdirectory) {
649 error = ENOTDIR;
650 goto bad;
651 }
984263bc
MD
652 } else if (doingdirectory) {
653 error = EISDIR;
654 goto bad;
655 }
656 error = ext2_dirrewrite(dp, ip, tcnp);
657 if (error)
658 goto bad;
659 /*
660 * If the target directory is in the same
661 * directory as the source directory,
662 * decrement the link count on the parent
663 * of the target directory.
664 */
665 if (doingdirectory && !newparent) {
666 dp->i_nlink--;
667 dp->i_flag |= IN_CHANGE;
668 }
fad57d0e
MD
669
670 /*
671 * manual cleanup, we can't use the bad or out target after
672 * this.
673 */
984263bc 674 vput(tdvp);
fad57d0e 675
984263bc
MD
676 /*
677 * Adjust the link count of the target to
678 * reflect the dirrewrite above. If this is
679 * a directory it is empty and there are
680 * no links to it, so we can squash the inode and
681 * any space associated with it. We disallowed
682 * renaming over top of a directory with links to
683 * it above, as the remaining link would point to
684 * a directory without "." or ".." entries.
685 */
686 xp->i_nlink--;
687 if (doingdirectory) {
688 if (--xp->i_nlink != 0)
1f1db49f
MD
689 panic("ext2_rename: linked directory");
690 error = EXT2_TRUNCATE(tvp, (off_t)0, IO_SYNC,
87de5057 691 tcnp->cn_cred);
984263bc
MD
692 }
693 xp->i_flag |= IN_CHANGE;
694 vput(tvp);
695 xp = NULL;
696 }
697
698 /*
fad57d0e
MD
699 * tvp and tdvp have been cleaned up. The bad and out targets may
700 * not be used. fvp and fdvp are ref'd but not locked. ip
701 * still represents the old fvp and ip->i_flag may still have IN_RENAME
702 * set (if doingdirectory).
703 */
704
705 /*
984263bc 706 * 3) Unlink the source.
fad57d0e
MD
707 *
708 * fdvp is locked and ref'd. ap->a_fvp holds the old lookup unlocked
709 * and ref'd, fvp will hold the new lookup locked and ref'd.
710 *
711 * After the relookup ap->a_fvp must be released as part of our
712 * cleanup, not just fdvp and fvp. And, on success, fdvp and
713 * fvp will be locked so the bad and out targets cannot be used.
984263bc 714 */
2b69e610 715 fcnp->cn_flags &= ~CNP_MODMASK;
fad57d0e
MD
716 fcnp->cn_flags |= CNP_LOCKPARENT;
717 KKASSERT(fcnp->cn_flags & CNP_PDIRUNLOCK);
984263bc 718 error = relookup(fdvp, &fvp, fcnp);
fad57d0e 719 if (error) {
984263bc
MD
720 /*
721 * From name has disappeared.
722 */
723 if (doingdirectory)
1f1db49f 724 panic("ext2_rename: lost dir entry");
fad57d0e 725 /* ip->i_flag only sets IN_RENAME if doingdirectory */
984263bc 726 vrele(ap->a_fvp);
fad57d0e
MD
727 if (fcnp->cn_flags & CNP_PDIRUNLOCK)
728 vrele(fdvp);
729 else
730 vput(fdvp);
984263bc
MD
731 return (0);
732 }
fad57d0e
MD
733 KKASSERT((fcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
734
735 /*
736 * This case shouldn't occur
737 */
738 if (fvp == NULL) {
739 /*
740 * From name has disappeared.
741 */
742 if (doingdirectory)
1f1db49f 743 panic("ext2_rename: lost dir entry");
fad57d0e
MD
744 /* ip->i_flag only sets IN_RENAME if doingdirectory */
745 vrele(ap->a_fvp);
746 vput(fvp);
747 vput(fdvp);
748 return (0);
749 }
750
751 /*
752 * fvp and fdvp are both ref'd and locked.
753 */
754 xp = VTOI(fvp);
755 dp = VTOI(fdvp);
756
984263bc
MD
757 /*
758 * Ensure that the directory entry still exists and has not
759 * changed while the new name has been entered. If the source is
760 * a file then the entry may have been unlinked or renamed. In
761 * either case there is no further work to be done. If the source
762 * is a directory then it cannot have been rmdir'ed; its link
763 * count of three would cause a rmdir to fail with ENOTEMPTY.
764 * The IN_RENAME flag ensures that it cannot be moved by another
765 * rename.
766 */
767 if (xp != ip) {
768 if (doingdirectory)
1f1db49f 769 panic("ext2_rename: lost dir entry");
fad57d0e 770 /* ip->i_flag only sets IN_RENAME if doingdirectory */
984263bc
MD
771 } else {
772 /*
773 * If the source is a directory with a
774 * new parent, the link count of the old
775 * parent directory must be decremented
776 * and ".." set to point to the new parent.
777 */
778 if (doingdirectory && newparent) {
779 dp->i_nlink--;
780 dp->i_flag |= IN_CHANGE;
781 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
87de5057
MD
782 sizeof (struct dirtemplate), (off_t)0,
783 UIO_SYSSPACE, IO_NODELOCKED,
60233e58 784 tcnp->cn_cred, NULL);
984263bc 785 if (error == 0) {
1f1db49f 786 /* Like ext2 little-endian: */
984263bc
MD
787 namlen = dirbuf.dotdot_type;
788 if (namlen != 2 ||
789 dirbuf.dotdot_name[0] != '.' ||
790 dirbuf.dotdot_name[1] != '.') {
1f1db49f 791 ext2_dirbad(xp, (doff_t)12,
984263bc
MD
792 "rename: mangled dir");
793 } else {
794 dirbuf.dotdot_ino = newparent;
b1ce5639 795 vn_rdwr(UIO_WRITE, fvp,
87de5057
MD
796 (caddr_t)&dirbuf,
797 sizeof (struct dirtemplate),
798 (off_t)0, UIO_SYSSPACE,
799 IO_NODELOCKED|IO_SYNC,
60233e58 800 tcnp->cn_cred, NULL);
984263bc
MD
801 }
802 }
803 }
804 error = ext2_dirremove(fdvp, fcnp);
805 if (!error) {
806 xp->i_nlink--;
807 xp->i_flag |= IN_CHANGE;
808 }
809 xp->i_flag &= ~IN_RENAME;
810 }
fad57d0e
MD
811 vput(fdvp);
812 vput(fvp);
984263bc
MD
813 vrele(ap->a_fvp);
814 return (error);
815
816bad:
817 if (xp)
818 vput(ITOV(xp));
fad57d0e
MD
819 if (dp)
820 vput(ITOV(dp));
984263bc
MD
821out:
822 if (doingdirectory)
823 ip->i_flag &= ~IN_RENAME;
ca466bae 824 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
984263bc
MD
825 ip->i_nlink--;
826 ip->i_flag |= IN_CHANGE;
827 ip->i_flag &= ~IN_RENAME;
828 vput(fvp);
fad57d0e 829 } else {
984263bc 830 vrele(fvp);
fad57d0e 831 }
984263bc
MD
832 return (error);
833}
834
835/*
836 * Mkdir system call
0f7f7a49
CP
837 *
838 * ext2_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
839 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
840 */
841static int
e62afb5f 842ext2_mkdir(struct vop_old_mkdir_args *ap)
984263bc 843{
f7aae92f
RG
844 struct vnode *dvp = ap->a_dvp;
845 struct vattr *vap = ap->a_vap;
846 struct componentname *cnp = ap->a_cnp;
847 struct inode *ip, *dp;
984263bc
MD
848 struct vnode *tvp;
849 struct dirtemplate dirtemplate, *dtp;
850 int error, dmode;
851
984263bc
MD
852 dp = VTOI(dvp);
853 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
854 error = EMLINK;
855 goto out;
856 }
857 dmode = vap->va_mode & 0777;
858 dmode |= IFDIR;
859 /*
860 * Must simulate part of ext2_makeinode here to acquire the inode,
861 * but not have it entered in the parent directory. The entry is
862 * made later after writing "." and ".." entries.
863 */
1f1db49f 864 error = EXT2_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
984263bc
MD
865 if (error)
866 goto out;
867 ip = VTOI(tvp);
868 ip->i_gid = dp->i_gid;
869#ifdef SUIDDIR
870 {
871#ifdef QUOTA
872 struct ucred ucred, *ucp;
873 ucp = cnp->cn_cred;
1d94a9ab 874#endif
984263bc
MD
875 /*
876 * if we are hacking owners here, (only do this where told to)
877 * and we are not giving it TOO root, (would subvert quotas)
878 * then go ahead and give it to the other user.
879 * The new directory also inherits the SUID bit.
880 * If user's UID and dir UID are the same,
881 * 'give it away' so that the SUID is still forced on.
882 */
883 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
884 (dp->i_mode & ISUID) && dp->i_uid) {
885 dmode |= ISUID;
886 ip->i_uid = dp->i_uid;
887#ifdef QUOTA
888 if (dp->i_uid != cnp->cn_cred->cr_uid) {
889 /*
890 * make sure the correct user gets charged
891 * for the space.
892 * Make a dummy credential for the victim.
893 * XXX This seems to never be accessed out of
894 * our context so a stack variable is ok.
895 */
896 ucred.cr_ref = 1;
897 ucred.cr_uid = ip->i_uid;
898 ucred.cr_ngroups = 1;
899 ucred.cr_groups[0] = dp->i_gid;
900 ucp = &ucred;
901 }
1d94a9ab 902#endif
984263bc
MD
903 } else {
904 ip->i_uid = cnp->cn_cred->cr_uid;
905 }
906#ifdef QUOTA
1f1db49f
MD
907 if ((error = ext2_getinoquota(ip)) ||
908 (error = ext2_chkiq(ip, 1, ucp, 0))) {
909 EXT2_VFREE(tvp, ip->i_number, dmode);
984263bc
MD
910 vput(tvp);
911 return (error);
912 }
913#endif
914 }
915#else
916 ip->i_uid = cnp->cn_cred->cr_uid;
917#ifdef QUOTA
1f1db49f
MD
918 if ((error = ext2_getinoquota(ip)) ||
919 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
920 EXT2_VFREE(tvp, ip->i_number, dmode);
984263bc
MD
921 vput(tvp);
922 return (error);
923 }
924#endif
925#endif
926 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
927 ip->i_mode = dmode;
928 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
929 ip->i_nlink = 2;
2b69e610 930 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc 931 ip->i_flags |= UF_OPAQUE;
1f1db49f 932 error = EXT2_UPDATE(tvp, 1);
984263bc
MD
933
934 /*
93f50d60
MD
935 * The vnode must have a VM object in order to issue buffer cache
936 * ops on it.
937 */
b0d18f7d 938 vinitvmio(tvp, 0, PAGE_SIZE, -1);
93f50d60
MD
939
940 /*
984263bc
MD
941 * Bump link count in parent directory
942 * to reflect work done below. Should
943 * be done before reference is created
944 * so reparation is possible if we crash.
945 */
946 dp->i_nlink++;
947 dp->i_flag |= IN_CHANGE;
1f1db49f 948 error = EXT2_UPDATE(dvp, 1);
984263bc
MD
949 if (error)
950 goto bad;
951
952 /* Initialize directory with "." and ".." from static template. */
953 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
954 EXT2_FEATURE_INCOMPAT_FILETYPE))
1f1db49f 955 dtp = &ext2_mastertemplate;
984263bc 956 else
1f1db49f 957 dtp = &ext2_omastertemplate;
984263bc
MD
958 dirtemplate = *dtp;
959 dirtemplate.dot_ino = ip->i_number;
960 dirtemplate.dotdot_ino = dp->i_number;
961 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
962 * so let's just redefine it - for this function only
963 */
964#undef DIRBLKSIZ
965#define DIRBLKSIZ VTOI(dvp)->i_e2fs->s_blocksize
966 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
967 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
87de5057 968 sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
60233e58 969 IO_NODELOCKED|IO_SYNC, cnp->cn_cred, NULL);
984263bc
MD
970 if (error) {
971 dp->i_nlink--;
972 dp->i_flag |= IN_CHANGE;
973 goto bad;
974 }
1f1db49f
MD
975 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
976 panic("ext2_mkdir: blksize"); /* XXX should grow with balloc() */
984263bc
MD
977 else {
978 ip->i_size = DIRBLKSIZ;
979 ip->i_flag |= IN_CHANGE;
980 }
981
982 /* Directory set up, now install its entry in the parent directory. */
983 error = ext2_direnter(ip, dvp, cnp);
984 if (error) {
985 dp->i_nlink--;
986 dp->i_flag |= IN_CHANGE;
987 }
988bad:
989 /*
990 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
991 * for us because we set the link count to 0.
992 */
993 if (error) {
994 ip->i_nlink = 0;
995 ip->i_flag |= IN_CHANGE;
996 vput(tvp);
997 } else
998 *ap->a_vpp = tvp;
999out:
1000 return (error);
1001#undef DIRBLKSIZ
1002#define DIRBLKSIZ DEV_BSIZE
1003}
1004
1005/*
1006 * Rmdir system call.
0f7f7a49
CP
1007 *
1008 * ext2_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1009 * struct componentname *a_cnp)
984263bc
MD
1010 */
1011static int
e62afb5f 1012ext2_rmdir(struct vop_old_rmdir_args *ap)
984263bc
MD
1013{
1014 struct vnode *vp = ap->a_vp;
1015 struct vnode *dvp = ap->a_dvp;
1016 struct componentname *cnp = ap->a_cnp;
984263bc
MD
1017 struct inode *ip, *dp;
1018 int error;
1019
1020 ip = VTOI(vp);
1021 dp = VTOI(dvp);
1022
1023 /*
1024 * Verify the directory is empty (and valid).
1025 * (Rmdir ".." won't be valid since
1026 * ".." will contain a reference to
1027 * the current directory and thus be
1028 * non-empty.)
1029 */
1030 error = 0;
1031 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1032 error = ENOTEMPTY;
1033 goto out;
1034 }
1035 if ((dp->i_flags & APPEND)
1036 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1037 error = EPERM;
1038 goto out;
1039 }
1040 /*
1041 * Delete reference to directory before purging
1042 * inode. If we crash in between, the directory
1043 * will be reattached to lost+found,
1044 */
1045 error = ext2_dirremove(dvp, cnp);
1046 if (error)
1047 goto out;
1048 dp->i_nlink--;
1049 dp->i_flag |= IN_CHANGE;
a11aaa81 1050 vn_unlock(dvp);
984263bc
MD
1051 /*
1052 * Truncate inode. The only stuff left
1053 * in the directory is "." and "..". The
1054 * "." reference is inconsequential since
1055 * we're quashing it. The ".." reference
1056 * has already been adjusted above. We've
1057 * removed the "." reference and the reference
1058 * in the parent directory, but there may be
1059 * other hard links so decrement by 2 and
1060 * worry about them later.
1061 */
1062 ip->i_nlink -= 2;
87de5057 1063 error = EXT2_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
ca466bae 1064 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
1065out:
1066 return (error);
1067}
1068
1069/*
1070 * symlink -- make a symbolic link
0f7f7a49
CP
1071 *
1072 * ext2_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1073 * struct componentname *a_cnp, struct vattr *a_vap,
1074 * char *a_target)
984263bc
MD
1075 */
1076static int
e62afb5f 1077ext2_symlink(struct vop_old_symlink_args *ap)
984263bc 1078{
f7aae92f
RG
1079 struct vnode *vp, **vpp = ap->a_vpp;
1080 struct inode *ip;
984263bc
MD
1081 int len, error;
1082
1083 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1084 vpp, ap->a_cnp);
1085 if (error)
1086 return (error);
1087 vp = *vpp;
1088 len = strlen(ap->a_target);
1089 if (len < vp->v_mount->mnt_maxsymlinklen) {
1090 ip = VTOI(vp);
1091 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1092 ip->i_size = len;
1093 ip->i_flag |= IN_CHANGE | IN_UPDATE;
6c23cda9 1094 } else {
93f50d60
MD
1095 /*
1096 * Make sure we have a VM object in order to use
1097 * the buffer cache.
1098 */
1099 if (vp->v_object == NULL)
b0d18f7d 1100 vinitvmio(vp, 0, PAGE_SIZE, -1);
93f50d60 1101
984263bc 1102 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
87de5057 1103 UIO_SYSSPACE, IO_NODELOCKED,
60233e58 1104 ap->a_cnp->cn_cred, NULL);
dd41ecb0
SW
1105
1106 if (error)
1107 vput(vp);
6c23cda9
SW
1108 }
1109
984263bc
MD
1110 return (error);
1111}
1112
1113/*
1114 * Allocate a new inode.
1115 */
1116static int
0f7f7a49
CP
1117ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1118 struct componentname *cnp)
984263bc 1119{
f7aae92f 1120 struct inode *ip, *pdir;
984263bc
MD
1121 struct vnode *tvp;
1122 int error;
1123
1124 pdir = VTOI(dvp);
984263bc
MD
1125 *vpp = NULL;
1126 if ((mode & IFMT) == 0)
1127 mode |= IFREG;
1128
1f1db49f 1129 error = EXT2_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
984263bc
MD
1130 if (error) {
1131 return (error);
1132 }
1133 ip = VTOI(tvp);
1134 ip->i_gid = pdir->i_gid;
1135#ifdef SUIDDIR
1136 {
1137#ifdef QUOTA
1138 struct ucred ucred, *ucp;
1139 ucp = cnp->cn_cred;
1d94a9ab 1140#endif
984263bc
MD
1141 /*
1142 * if we are
1143 * not the owner of the directory,
1144 * and we are hacking owners here, (only do this where told to)
1145 * and we are not giving it TOO root, (would subvert quotas)
1146 * then go ahead and give it to the other user.
1147 * Note that this drops off the execute bits for security.
1148 */
1149 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1150 (pdir->i_mode & ISUID) &&
1151 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1152 ip->i_uid = pdir->i_uid;
1153 mode &= ~07111;
1154#ifdef QUOTA
1155 /*
1156 * make sure the correct user gets charged
1157 * for the space.
1158 * Quickly knock up a dummy credential for the victim.
1159 * XXX This seems to never be accessed out of our
1160 * context so a stack variable is ok.
1161 */
1162 ucred.cr_ref = 1;
1163 ucred.cr_uid = ip->i_uid;
1164 ucred.cr_ngroups = 1;
1165 ucred.cr_groups[0] = pdir->i_gid;
1166 ucp = &ucred;
1d94a9ab 1167#endif
984263bc
MD
1168 } else {
1169 ip->i_uid = cnp->cn_cred->cr_uid;
1170 }
1171
1172#ifdef QUOTA
6f1a0d44 1173 if ((error = ext2_getinoquota(ip)) ||
1f1db49f
MD
1174 (error = ext2_chkiq(ip, 1, ucp, 0))) {
1175 EXT2_VFREE(tvp, ip->i_number, mode);
984263bc
MD
1176 vput(tvp);
1177 return (error);
1178 }
1179#endif
1180 }
1181#else
1182 ip->i_uid = cnp->cn_cred->cr_uid;
1183#ifdef QUOTA
1f1db49f
MD
1184 if ((error = ext2_getinoquota(ip)) ||
1185 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
1186 EXT2_VFREE(tvp, ip->i_number, mode);
984263bc
MD
1187 vput(tvp);
1188 return (error);
1189 }
1190#endif
1191#endif
1192 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1193 ip->i_mode = mode;
1194 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1195 ip->i_nlink = 1;
1196 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
61b7a688 1197 priv_check_cred(cnp->cn_cred, PRIV_VFS_SETGID, 0))
984263bc
MD
1198 ip->i_mode &= ~ISGID;
1199
2b69e610 1200 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
1201 ip->i_flags |= UF_OPAQUE;
1202
1203 /*
1c843a13
MD
1204 * Regular files and directories need VM objects. Softlinks do
1205 * not (not immediately anyway).
1206 */
1207 if (tvp->v_type == VREG || tvp->v_type == VDIR)
b0d18f7d 1208 vinitvmio(tvp, 0, PAGE_SIZE, -1);
1c843a13
MD
1209
1210 /*
984263bc
MD
1211 * Make sure inode goes to disk before directory entry.
1212 */
1f1db49f 1213 error = EXT2_UPDATE(tvp, 1);
984263bc
MD
1214 if (error)
1215 goto bad;
1216 error = ext2_direnter(ip, dvp, cnp);
1217 if (error)
1218 goto bad;
1219
1220 *vpp = tvp;
1221 return (0);
1222
1223bad:
1224 /*
1225 * Write error occurred trying to update the inode
1226 * or the directory so must deallocate the inode.
1227 */
1228 ip->i_nlink = 0;
1229 ip->i_flag |= IN_CHANGE;
1230 vput(tvp);
1231 return (error);
1232}
1233
1f1db49f
MD
1234void
1235ext2_itimes(struct vnode *vp)
1236{
1237 struct inode *ip;
1238 struct timespec ts;
1239
1240 ip = VTOI(vp);
1241 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
1242 return;
1243 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
1244 ip->i_flag |= IN_LAZYMOD;
1245 else
1246 ip->i_flag |= IN_MODIFIED;
1247 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
1248 vfs_timestamp(&ts);
1249 if (ip->i_flag & IN_ACCESS) {
1250 ip->i_atime = ts.tv_sec;
1251 ip->i_atimensec = ts.tv_nsec;
1252 }
1253 if (ip->i_flag & IN_UPDATE) {
1254 ip->i_mtime = ts.tv_sec;
1255 ip->i_mtimensec = ts.tv_nsec;
1256 ip->i_modrev++;
1257 }
1258 if (ip->i_flag & IN_CHANGE) {
1259 ip->i_ctime = ts.tv_sec;
1260 ip->i_ctimensec = ts.tv_nsec;
1261 }
1262 }
1263 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
1264}
1265
1266/*
1267 * Open called.
1268 *
1269 * Nothing to do.
1270 *
b478fdce
SW
1271 * ext2_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1272 * struct file *a_fp)
1f1db49f
MD
1273 */
1274/* ARGSUSED */
1275static
1276int
1277ext2_open(struct vop_open_args *ap)
1278{
1279 struct vnode *vp = ap->a_vp;
1280
1281 /*
1282 * Files marked append-only must be opened for appending.
1283 */
1284 if ((VTOI(vp)->i_flags & APPEND) &&
1285 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) {
1286 return (EPERM);
1287 }
1f1db49f
MD
1288 return (vop_stdopen(ap));
1289}
1290
1291/*
1292 * Close called.
1293 *
1294 * Update the times on the inode.
1295 *
87de5057 1296 * ext2_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1f1db49f
MD
1297 */
1298/* ARGSUSED */
1299static
1300int
1301ext2_close(struct vop_close_args *ap)
1302{
1303 struct vnode *vp = ap->a_vp;
1304
3c37c940 1305 if (vp->v_sysref.refcnt > 1)
1f1db49f
MD
1306 ext2_itimes(vp);
1307 return (vop_stdclose(ap));
1308}
1309
1310/*
87de5057 1311 * ext2_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred)
1f1db49f
MD
1312 */
1313static
1314int
1315ext2_access(struct vop_access_args *ap)
1316{
1317 struct vnode *vp = ap->a_vp;
1318 struct inode *ip = VTOI(vp);
1f1db49f 1319 int error;
1f1db49f 1320
1dbf38bb
NT
1321#ifdef QUOTA
1322 if (ap->a_mode & VWRITE) {
1f1db49f
MD
1323 switch (vp->v_type) {
1324 case VDIR:
1325 case VLNK:
1326 case VREG:
1f1db49f
MD
1327 if ((error = ext2_getinoquota(ip)) != 0)
1328 return (error);
1f1db49f
MD
1329 break;
1330 default:
1331 break;
1332 }
1333 }
1dbf38bb 1334#endif
1f1db49f 1335
1dbf38bb
NT
1336 error = vop_helper_access(ap, ip->i_uid, ip->i_gid, ip->i_mode,
1337 ip->i_flags);
1338 return (error);
1f1db49f
MD
1339}
1340
1341/*
87de5057 1342 * ext2_getattr(struct vnode *a_vp, struct vattr *a_vap)
1f1db49f
MD
1343 */
1344/* ARGSUSED */
1345static
1346int
1347ext2_getattr(struct vop_getattr_args *ap)
1348{
1349 struct vnode *vp = ap->a_vp;
1350 struct inode *ip = VTOI(vp);
1351 struct vattr *vap = ap->a_vap;
1352
1f1db49f
MD
1353 ext2_itimes(vp);
1354 /*
1355 * Copy from inode table
1356 */
1357 vap->va_fsid = dev2udev(ip->i_dev);
1358 vap->va_fileid = ip->i_number;
1359 vap->va_mode = ip->i_mode & ~IFMT;
1360 vap->va_nlink = VFSTOEXT2(vp->v_mount)->um_i_effnlink_valid ?
1361 ip->i_effnlink : ip->i_nlink;
1362 vap->va_uid = ip->i_uid;
1363 vap->va_gid = ip->i_gid;
0e9b9130
MD
1364 vap->va_rmajor = umajor(ip->i_rdev);
1365 vap->va_rminor = uminor(ip->i_rdev);
1f1db49f
MD
1366 vap->va_size = ip->i_din.di_size;
1367 vap->va_atime.tv_sec = ip->i_atime;
1368 vap->va_atime.tv_nsec = ip->i_atimensec;
1369 vap->va_mtime.tv_sec = ip->i_mtime;
1370 vap->va_mtime.tv_nsec = ip->i_mtimensec;
1371 vap->va_ctime.tv_sec = ip->i_ctime;
1372 vap->va_ctime.tv_nsec = ip->i_ctimensec;
1373 vap->va_flags = ip->i_flags;
1374 vap->va_gen = ip->i_gen;
1375 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1376 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
1377 vap->va_type = IFTOVT(ip->i_mode);
1378 vap->va_filerev = ip->i_modrev;
1f1db49f
MD
1379 return (0);
1380}
1381
1382/*
1383 * Set attribute vnode op. called from several syscalls
1384 *
87de5057 1385 * ext2_setattr(struct vnode *a_vp, struct vattr *a_vap, struct ucred *a_cred)
1f1db49f
MD
1386 */
1387static
1388int
1389ext2_setattr(struct vop_setattr_args *ap)
1390{
1391 struct vattr *vap = ap->a_vap;
1392 struct vnode *vp = ap->a_vp;
1393 struct inode *ip = VTOI(vp);
1394 struct ucred *cred = ap->a_cred;
1395 int error;
1396
1397 /*
1398 * Check for unsettable attributes.
1399 */
1400 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
1401 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
0e9b9130 1402 (vap->va_blocksize != VNOVAL) || (vap->va_rmajor != VNOVAL) ||
1f1db49f
MD
1403 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
1404 return (EINVAL);
1405 }
1406 if (vap->va_flags != VNOVAL) {
1407 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1408 return (EROFS);
1409 if (cred->cr_uid != ip->i_uid &&
db813a0c 1410 (error = priv_check_cred(cred, PRIV_VFS_SETATTR, 0)))
1f1db49f
MD
1411 return (error);
1412 /*
1413 * Note that a root chflags becomes a user chflags when
1414 * we are jailed, unless the jail.chflags_allowed sysctl
1415 * is set.
1416 */
1417 if (cred->cr_uid == 0 &&
1418 (!jailed(cred) || jail_chflags_allowed)) {
1419 if ((ip->i_flags
1420 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
1421 securelevel > 0)
1422 return (EPERM);
1423 ip->i_flags = vap->va_flags;
1424 } else {
1425 if (ip->i_flags
1426 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
1427 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
1428 return (EPERM);
1429 ip->i_flags &= SF_SETTABLE;
1430 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
1431 }
1432 ip->i_flag |= IN_CHANGE;
1433 if (vap->va_flags & (IMMUTABLE | APPEND))
1434 return (0);
1435 }
1436 if (ip->i_flags & (IMMUTABLE | APPEND))
1437 return (EPERM);
1438 /*
1439 * Go through the fields and update iff not VNOVAL.
1440 */
1441 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
1442 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1443 return (EROFS);
87de5057 1444 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred)) != 0)
1f1db49f
MD
1445 return (error);
1446 }
1447 if (vap->va_size != VNOVAL) {
1448 /*
1449 * Disallow write attempts on read-only filesystems;
1450 * unless the file is a socket, fifo, or a block or
1451 * character device resident on the filesystem.
1452 */
1453 switch (vp->v_type) {
1454 case VDIR:
1455 return (EISDIR);
1456 case VLNK:
1457 case VREG:
1458 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1459 return (EROFS);
1460 break;
1461 default:
1462 break;
1463 }
87de5057 1464 if ((error = EXT2_TRUNCATE(vp, vap->va_size, 0, cred)) != 0)
1f1db49f
MD
1465 return (error);
1466 }
1467 ip = VTOI(vp);
1468 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1469 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1470 return (EROFS);
1471 if (cred->cr_uid != ip->i_uid &&
db813a0c 1472 (error = priv_check_cred(cred, PRIV_VFS_SETATTR, 0)) &&
1f1db49f 1473 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
cb66845a 1474 (error = VOP_EACCESS(vp, VWRITE, cred))))
1f1db49f
MD
1475 return (error);
1476 if (vap->va_atime.tv_sec != VNOVAL)
1477 ip->i_flag |= IN_ACCESS;
1478 if (vap->va_mtime.tv_sec != VNOVAL)
1479 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1480 ext2_itimes(vp);
1481 if (vap->va_atime.tv_sec != VNOVAL) {
1482 ip->i_atime = vap->va_atime.tv_sec;
1483 ip->i_atimensec = vap->va_atime.tv_nsec;
1484 }
1485 if (vap->va_mtime.tv_sec != VNOVAL) {
1486 ip->i_mtime = vap->va_mtime.tv_sec;
1487 ip->i_mtimensec = vap->va_mtime.tv_nsec;
1488 }
1489 error = EXT2_UPDATE(vp, 0);
1490 if (error)
1491 return (error);
1492 }
1493 error = 0;
1494 if (vap->va_mode != (mode_t)VNOVAL) {
1495 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1496 return (EROFS);
87de5057 1497 error = ext2_chmod(vp, (int)vap->va_mode, cred);
1f1db49f
MD
1498 }
1499 VN_KNOTE(vp, NOTE_ATTRIB);
1500 return (error);
1501}
1502
1503/*
1504 * Change the mode on a file.
1505 * Inode must be locked before calling.
1506 */
1507static int
87de5057 1508ext2_chmod(struct vnode *vp, int mode, struct ucred *cred)
1f1db49f
MD
1509{
1510 struct inode *ip = VTOI(vp);
1511 int error;
1512
1513 if (cred->cr_uid != ip->i_uid) {
a79e2ed8 1514 error = priv_check_cred(cred, PRIV_VFS_CHMOD, 0);
1f1db49f
MD
1515 if (error)
1516 return (error);
1517 }
1518 if (cred->cr_uid) {
1519 if (vp->v_type != VDIR && (mode & S_ISTXT))
1520 return (EFTYPE);
1521 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
1522 return (EPERM);
1523 }
1524 ip->i_mode &= ~ALLPERMS;
1525 ip->i_mode |= (mode & ALLPERMS);
1526 ip->i_flag |= IN_CHANGE;
1527 return (0);
1528}
1529
1530/*
1531 * Perform chown operation on inode ip;
1532 * inode must be locked prior to call.
1533 */
1534static int
87de5057 1535ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred)
1f1db49f
MD
1536{
1537 struct inode *ip = VTOI(vp);
1538 uid_t ouid;
1539 gid_t ogid;
1540 int error = 0;
1541#ifdef QUOTA
1542 int i;
1543 long change;
1544#endif
1545
1546 if (uid == (uid_t)VNOVAL)
1547 uid = ip->i_uid;
1548 if (gid == (gid_t)VNOVAL)
1549 gid = ip->i_gid;
1550 /*
1551 * If we don't own the file, are trying to change the owner
1552 * of the file, or are not a member of the target group,
1553 * the caller must be superuser or the call fails.
1554 */
1555 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
1556 (gid != ip->i_gid && !(cred->cr_gid == gid ||
1557 groupmember((gid_t)gid, cred)))) &&
a79e2ed8 1558 (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
1f1db49f
MD
1559 return (error);
1560 ogid = ip->i_gid;
1561 ouid = ip->i_uid;
1562#ifdef QUOTA
1563 if ((error = ext2_getinoquota(ip)) != 0)
1564 return (error);
1565 if (ouid == uid) {
1566 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1567 ip->i_dquot[USRQUOTA] = NODQUOT;
1568 }
1569 if (ogid == gid) {
1570 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1571 ip->i_dquot[GRPQUOTA] = NODQUOT;
1572 }
1573 change = ip->i_blocks;
1574 (void) ext2_chkdq(ip, -change, cred, CHOWN);
1575 (void) ext2_chkiq(ip, -1, cred, CHOWN);
1576 for (i = 0; i < MAXQUOTAS; i++) {
1577 ext2_dqrele(vp, ip->i_dquot[i]);
1578 ip->i_dquot[i] = NODQUOT;
1579 }
1580#endif
1581 ip->i_gid = gid;
1582 ip->i_uid = uid;
1583#ifdef QUOTA
1584 if ((error = ext2_getinoquota(ip)) == 0) {
1585 if (ouid == uid) {
1586 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1587 ip->i_dquot[USRQUOTA] = NODQUOT;
1588 }
1589 if (ogid == gid) {
1590 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1591 ip->i_dquot[GRPQUOTA] = NODQUOT;
1592 }
1593 if ((error = ext2_chkdq(ip, change, cred, CHOWN)) == 0) {
1594 if ((error = ext2_chkiq(ip, 1, cred, CHOWN)) == 0)
1595 goto good;
1596 else
1597 (void)ext2_chkdq(ip, -change, cred, CHOWN|FORCE);
1598 }
1599 for (i = 0; i < MAXQUOTAS; i++) {
1600 ext2_dqrele(vp, ip->i_dquot[i]);
1601 ip->i_dquot[i] = NODQUOT;
1602 }
1603 }
1604 ip->i_gid = ogid;
1605 ip->i_uid = ouid;
1606 if (ext2_getinoquota(ip) == 0) {
1607 if (ouid == uid) {
1608 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1609 ip->i_dquot[USRQUOTA] = NODQUOT;
1610 }
1611 if (ogid == gid) {
1612 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1613 ip->i_dquot[GRPQUOTA] = NODQUOT;
1614 }
1615 (void) ext2_chkdq(ip, change, cred, FORCE|CHOWN);
1616 (void) ext2_chkiq(ip, 1, cred, FORCE|CHOWN);
1617 (void) ext2_getinoquota(ip);
1618 }
1619 return (error);
1620good:
1621 if (ext2_getinoquota(ip))
1622 panic("ext2_chown: lost quota");
1623#endif /* QUOTA */
1624 ip->i_flag |= IN_CHANGE;
1625 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
1626 ip->i_mode &= ~(ISUID | ISGID);
1627 return (0);
1628}
1629
1630/*
1631 * Mmap a file
1632 *
1633 * NB Currently unsupported.
1634 *
87de5057 1635 * ext2_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred)
1f1db49f
MD
1636 */
1637/* ARGSUSED */
1638static
1639int
1640ext2_mmap(struct vop_mmap_args *ap)
1641{
1642 return (EINVAL);
1643}
1644
1645/*
1646 * whiteout vnode call
1647 *
1648 * ext2_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
1649 */
1650static
1651int
1652ext2_whiteout(struct vop_old_whiteout_args *ap)
1653{
1654 return (EOPNOTSUPP);
1655}
1656
1657/*
1658 * Return target name of a symbolic link
1659 *
1660 * ext2_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1661 */
1662static
1663int
1664ext2_readlink(struct vop_readlink_args *ap)
1665{
1666 struct vnode *vp = ap->a_vp;
1667 struct inode *ip = VTOI(vp);
1668 int isize;
1669
1670 isize = ip->i_size;
1671 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1672 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1673 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1674 return (0);
1675 }
1f1db49f
MD
1676 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1677}
1678
1679/*
1680 * Calculate the logical to physical mapping if not done already,
1681 * then call the device strategy routine.
1682 *
1683 * In order to be able to swap to a file, the VOP_BMAP operation may not
1684 * deadlock on memory. See ext2_bmap() for details.
1685 *
1686 * ext2_strategy(struct vnode *a_vp, struct bio *a_bio)
1687 */
1688static
1689int
1690ext2_strategy(struct vop_strategy_args *ap)
1691{
1692 struct bio *bio = ap->a_bio;
1693 struct bio *nbio;
1694 struct buf *bp = bio->bio_buf;
1695 struct vnode *vp = ap->a_vp;
1696 struct inode *ip;
1697 int error;
1698
1699 ip = VTOI(vp);
1700 if (vp->v_type == VBLK || vp->v_type == VCHR)
1701 panic("ext2_strategy: spec");
1702 nbio = push_bio(bio);
1703 if (nbio->bio_offset == NOOFFSET) {
08daea96 1704 error = VOP_BMAP(vp, bio->bio_offset, &nbio->bio_offset,
e92ca23a 1705 NULL, NULL, bp->b_cmd);
1f1db49f
MD
1706 if (error) {
1707 bp->b_error = error;
1708 bp->b_flags |= B_ERROR;
1709 /* I/O was never started on nbio, must biodone(bio) */
1710 biodone(bio);
1711 return (error);
1712 }
1713 if (nbio->bio_offset == NOOFFSET)
1714 vfs_bio_clrbuf(bp);
1715 }
1716 if (nbio->bio_offset == NOOFFSET) {
1717 /* I/O was never started on nbio, must biodone(bio) */
1718 biodone(bio);
1719 return (0);
1720 }
1721 vn_strategy(ip->i_devvp, nbio);
1722 return (0);
1723}
1724
1725/*
1726 * Print out the contents of an inode.
1727 *
1728 * ext2_print(struct vnode *a_vp)
1729 */
1730static
1731int
1732ext2_print(struct vop_print_args *ap)
1733{
1734 struct vnode *vp = ap->a_vp;
1735 struct inode *ip = VTOI(vp);
1736
086c1d7e 1737 kprintf("tag VT_EXT2FS, ino %lu, on dev %s (%d, %d)",
1f1db49f
MD
1738 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1739 minor(ip->i_dev));
1740 if (vp->v_type == VFIFO)
1741 fifo_printinfo(vp);
1742 lockmgr_printinfo(&vp->v_lock);
086c1d7e 1743 kprintf("\n");
1f1db49f
MD
1744 return (0);
1745}
1746
1747/*
1f1db49f
MD
1748 * Read wrapper for fifos.
1749 *
1750 * ext2fifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1751 * struct ucred *a_cred)
1752 */
1753static
1754int
1755ext2fifo_read(struct vop_read_args *ap)
1756{
1757 int error, resid;
1758 struct inode *ip;
1759 struct uio *uio;
1760
1761 uio = ap->a_uio;
1762 resid = uio->uio_resid;
66a1ddf5 1763 error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1f1db49f
MD
1764 ip = VTOI(ap->a_vp);
1765 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1766 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1767 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1768 return (error);
1769}
1770
1771/*
1772 * Write wrapper for fifos.
1773 *
1774 * ext2fifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1775 * struct ucred *a_cred)
1776 */
1777static
1778int
1779ext2fifo_write(struct vop_write_args *ap)
1780{
1781 int error, resid;
1782 struct inode *ip;
1783 struct uio *uio;
1784
1785 uio = ap->a_uio;
1786 resid = uio->uio_resid;
66a1ddf5 1787 error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1f1db49f
MD
1788 ip = VTOI(ap->a_vp);
1789 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1790 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1791 return (error);
1792}
1793
1794/*
1795 * Close wrapper for fifos.
1796 *
1797 * Update the times on the inode then do device close.
1798 *
87de5057 1799 * ext2fifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1f1db49f
MD
1800 */
1801static
1802int
1803ext2fifo_close(struct vop_close_args *ap)
1804{
1805 struct vnode *vp = ap->a_vp;
1806
3c37c940 1807 if (vp->v_sysref.refcnt > 1)
1f1db49f 1808 ext2_itimes(vp);
66a1ddf5 1809 return (VOCALL(&fifo_vnode_vops, &ap->a_head));
1f1db49f
MD
1810}
1811
1812/*
1813 * Kqfilter wrapper for fifos.
1814 *
1815 * Fall through to ext2 kqfilter routines if needed
1816 */
1817static
1818int
1819ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1820{
1821 int error;
1822
66a1ddf5 1823 error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1f1db49f
MD
1824 if (error)
1825 error = ext2_kqfilter(ap);
1826 return (error);
1827}
1828
1829/*
1830 * Return POSIX pathconf information applicable to ext2 filesystems.
1831 *
1832 * ext2_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
1833 */
1834static
1835int
1836ext2_pathconf(struct vop_pathconf_args *ap)
1837{
1838 switch (ap->a_name) {
1839 case _PC_LINK_MAX:
1840 *ap->a_retval = LINK_MAX;
1841 return (0);
1842 case _PC_NAME_MAX:
1843 *ap->a_retval = NAME_MAX;
1844 return (0);
1845 case _PC_PATH_MAX:
1846 *ap->a_retval = PATH_MAX;
1847 return (0);
1848 case _PC_PIPE_BUF:
1849 *ap->a_retval = PIPE_BUF;
1850 return (0);
1851 case _PC_CHOWN_RESTRICTED:
1852 *ap->a_retval = 1;
1853 return (0);
1854 case _PC_NO_TRUNC:
1855 *ap->a_retval = 1;
1856 return (0);
1857 default:
1858 return (EINVAL);
1859 }
1860 /* NOTREACHED */
1861}
1862
1863/*
1864 * Advisory record locking support
1865 *
1866 * ext2_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
1867 * int a_flags)
1868 */
1869static
1870int
1871ext2_advlock(struct vop_advlock_args *ap)
1872{
1873 struct inode *ip = VTOI(ap->a_vp);
1874
1875 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1876}
1877
1878/*
1879 * Initialize the vnode associated with a new inode, handle aliased
1880 * vnodes.
1881 */
1882int
1883ext2_vinit(struct mount *mntp, struct vnode **vpp)
1884{
1885 struct inode *ip;
1886 struct vnode *vp;
1887 struct timeval tv;
1888
1889 vp = *vpp;
1890 ip = VTOI(vp);
1891
1892 switch(vp->v_type = IFTOVT(ip->i_mode)) {
1893 case VCHR:
1894 case VBLK:
1895 vp->v_ops = &mntp->mnt_vn_spec_ops;
0e9b9130 1896 addaliasu(vp, umajor(ip->i_rdev), uminor(ip->i_rdev));
1f1db49f
MD
1897 break;
1898 case VFIFO:
1899 vp->v_ops = &mntp->mnt_vn_fifo_ops;
1900 break;
93f50d60 1901 case VDIR:
1c843a13 1902 case VREG:
b0d18f7d 1903 vinitvmio(vp, ip->i_size, PAGE_SIZE, -1); /* XXX */
1c843a13
MD
1904 break;
1905 case VLNK:
1906 if ((ip->i_size >= vp->v_mount->mnt_maxsymlinklen) &&
1907 ip->i_din.di_blocks != 0
1908 ) {
b0d18f7d 1909 vinitvmio(vp, ip->i_size, PAGE_SIZE, -1);
1c843a13 1910 }
93f50d60 1911 break;
1f1db49f
MD
1912 default:
1913 break;
1914
1915 }
1916
1917 if (ip->i_number == ROOTINO)
1918 vp->v_flag |= VROOT;
1919 /*
1920 * Initialize modrev times
1921 */
1922 getmicrouptime(&tv);
1923 SETHIGH(ip->i_modrev, tv.tv_sec);
1924 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
1925 *vpp = vp;
1926 return (0);
1927}
1928
1929static struct filterops ext2read_filtops =
4c91dbc9 1930 { FILTEROP_ISFD, NULL, filt_ext2detach, filt_ext2read };
1f1db49f 1931static struct filterops ext2write_filtops =
4c91dbc9 1932 { FILTEROP_ISFD, NULL, filt_ext2detach, filt_ext2write };
1f1db49f 1933static struct filterops ext2vnode_filtops =
4c91dbc9 1934 { FILTEROP_ISFD, NULL, filt_ext2detach, filt_ext2vnode };
1f1db49f
MD
1935
1936/*
1937 * ext2_kqfilter(struct vnode *a_vp, struct knote *a_kn)
1938 */
1939static int
1940ext2_kqfilter(struct vop_kqfilter_args *ap)
1941{
1942 struct vnode *vp = ap->a_vp;
1943 struct knote *kn = ap->a_kn;
1f1db49f
MD
1944
1945 switch (kn->kn_filter) {
1946 case EVFILT_READ:
1947 kn->kn_fop = &ext2read_filtops;
1948 break;
1949 case EVFILT_WRITE:
1950 kn->kn_fop = &ext2write_filtops;
1951 break;
1952 case EVFILT_VNODE:
1953 kn->kn_fop = &ext2vnode_filtops;
1954 break;
1955 default:
b287d649 1956 return (EOPNOTSUPP);
1f1db49f
MD
1957 }
1958
1959 kn->kn_hook = (caddr_t)vp;
1960
5b22f1a7 1961 /* XXX: kq token actually protects the list */
3b998fa9 1962 lwkt_gettoken(&vp->v_token);
5b22f1a7 1963 knote_insert(&vp->v_pollinfo.vpi_kqinfo.ki_note, kn);
3b998fa9 1964 lwkt_reltoken(&vp->v_token);
1f1db49f
MD
1965
1966 return (0);
1967}
1968
1969static void
1970filt_ext2detach(struct knote *kn)
1971{
1972 struct vnode *vp = (struct vnode *)kn->kn_hook;
1f1db49f 1973
3b998fa9 1974 lwkt_gettoken(&vp->v_token);
5b22f1a7 1975 knote_remove(&vp->v_pollinfo.vpi_kqinfo.ki_note, kn);
3b998fa9 1976 lwkt_reltoken(&vp->v_token);
1f1db49f
MD
1977}
1978
1979/*ARGSUSED*/
1980static int
1981filt_ext2read(struct knote *kn, long hint)
1982{
1983 struct vnode *vp = (struct vnode *)kn->kn_hook;
1984 struct inode *ip = VTOI(vp);
57b24f4e 1985 off_t off;
1f1db49f
MD
1986
1987 /*
1988 * filesystem is gone, so set the EOF flag and schedule
1989 * the knote for deletion.
1990 */
1991 if (hint == NOTE_REVOKE) {
1992 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1993 return (1);
1994 }
57b24f4e
MD
1995 off = ip->i_size - kn->kn_fp->f_offset;
1996 kn->kn_data = (off < INTPTR_MAX) ? off : INTPTR_MAX;
1997 if (kn->kn_sfflags & NOTE_OLDAPI)
1998 return(1);
1f1db49f
MD
1999 return (kn->kn_data != 0);
2000}
2001
2002/*ARGSUSED*/
2003static int
2004filt_ext2write(struct knote *kn, long hint)
2005{
2006 /*
2007 * filesystem is gone, so set the EOF flag and schedule
2008 * the knote for deletion.
2009 */
2010 if (hint == NOTE_REVOKE)
2011 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2012
2013 kn->kn_data = 0;
2014 return (1);
2015}
2016
2017static int
2018filt_ext2vnode(struct knote *kn, long hint)
2019{
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);
2025 }
2026 return (kn->kn_fflags != 0);
2027}
2028
66a1ddf5
MD
2029struct 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,
66a1ddf5 2043 .vop_old_link = ext2_link,
66a1ddf5
MD
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,
66a1ddf5
MD
2049 .vop_kqfilter = ext2_kqfilter,
2050 .vop_print = ext2_print,
2051 .vop_readdir = ext2_readdir,
2052 .vop_readlink = ext2_readlink,
2053 .vop_reclaim = ext2_reclaim,
2054 .vop_old_remove = ext2_remove,
2055 .vop_old_rename = ext2_rename,
2056 .vop_old_rmdir = ext2_rmdir,
2057 .vop_setattr = ext2_setattr,
2058 .vop_strategy = ext2_strategy,
2059 .vop_old_symlink = ext2_symlink,
66a1ddf5 2060 .vop_old_whiteout = ext2_whiteout,
a9de949a 2061 .vop_getpages = vop_stdgetpages,
1787385d 2062 .vop_putpages = vop_stdputpages
1f1db49f 2063};
66a1ddf5
MD
2064
2065struct vop_ops ext2_spec_vops = {
8be7edad 2066 .vop_default = vop_defaultop,
66a1ddf5
MD
2067 .vop_fsync = ext2_fsync,
2068 .vop_access = ext2_access,
8be7edad 2069 .vop_close = ext2_close,
66a1ddf5
MD
2070 .vop_getattr = ext2_getattr,
2071 .vop_inactive = ext2_inactive,
66a1ddf5 2072 .vop_print = ext2_print,
8be7edad 2073 .vop_read = vop_stdnoread,
66a1ddf5
MD
2074 .vop_reclaim = ext2_reclaim,
2075 .vop_setattr = ext2_setattr,
8be7edad 2076 .vop_write = vop_stdnowrite
1f1db49f 2077};
66a1ddf5
MD
2078
2079struct vop_ops ext2_fifo_vops = {
2080 .vop_default = ext2_vnoperatefifo,
2081 .vop_fsync = ext2_fsync,
2082 .vop_access = ext2_access,
2083 .vop_close = ext2fifo_close,
2084 .vop_getattr = ext2_getattr,
2085 .vop_inactive = ext2_inactive,
66a1ddf5 2086 .vop_kqfilter = ext2fifo_kqfilter,
66a1ddf5
MD
2087 .vop_print = ext2_print,
2088 .vop_read = ext2fifo_read,
2089 .vop_reclaim = ext2_reclaim,
2090 .vop_setattr = ext2_setattr,
66a1ddf5 2091 .vop_write = ext2fifo_write
1f1db49f 2092};
1f1db49f 2093
66a1ddf5
MD
2094VNODEOP_SET(ext2_vnode_vops);
2095VNODEOP_SET(ext2_spec_vops);
2096VNODEOP_SET(ext2_fifo_vops);
1f1db49f
MD
2097
2098/*
31bd717a 2099 * ext2_vnoperate()
1f1db49f
MD
2100 */
2101int
2102ext2_vnoperate(struct vop_generic_args *ap)
2103{
66a1ddf5 2104 return (VOCALL(&ext2_vnode_vops, ap));
1f1db49f
MD
2105}
2106
2107/*
31bd717a 2108 * ext2_vnoperatefifo()
1f1db49f
MD
2109 */
2110int
2111ext2_vnoperatefifo(struct vop_generic_args *ap)
2112{
66a1ddf5 2113 return (VOCALL(&ext2_fifo_vops, ap));
1f1db49f
MD
2114}
2115