Due to continuing issues with VOP_READ/VOP_WRITE ops being called without
[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 $
1c843a13 47 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vnops.c,v 1.29 2006/04/07 06:38:30 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
MD
61#include <sys/proc.h>
62#include <sys/mount.h>
63#include <sys/time.h>
64#include <sys/vnode.h>
1f1db49f
MD
65#include <sys/dirent.h>
66#include <sys/lockf.h>
67#include <sys/event.h>
68#include <sys/conf.h>
69#include <sys/file.h>
70#include <sys/jail.h>
984263bc 71#include <sys/namei.h>
1f1db49f
MD
72#include <sys/signalvar.h>
73#include <sys/unistd.h>
984263bc
MD
74
75#include <vm/vm.h>
76#include <vm/vm_extern.h>
77#include <vm/vm_zone.h>
78#include <vm/vnode_pager.h>
1f1db49f 79
7b95be2a 80#include <sys/buf2.h>
61670a01 81#include <sys/thread2.h>
984263bc 82
1f1db49f 83#include <vfs/fifofs/fifo.h>
984263bc 84
1f1db49f
MD
85#include "dir.h"
86#include "quota.h"
87#include "inode.h"
88#include "ext2mount.h"
1f2de5d4
MD
89#include "ext2_fs_sb.h"
90#include "fs.h"
91#include "ext2_extern.h"
92#include "ext2_fs.h"
984263bc 93
1f1db49f
MD
94static int ext2_access (struct vop_access_args *);
95static int ext2_advlock (struct vop_advlock_args *);
96static int ext2_chmod (struct vnode *, int, struct ucred *, struct thread *);
97static int ext2_chown (struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
98static int ext2_close (struct vop_close_args *);
99static int ext2_getattr (struct vop_getattr_args *);
a6ee311a 100static int ext2_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
1f1db49f
MD
101static int ext2_mmap (struct vop_mmap_args *);
102static int ext2_open (struct vop_open_args *);
103static int ext2_pathconf (struct vop_pathconf_args *);
104static int ext2_print (struct vop_print_args *);
105static int ext2_readlink (struct vop_readlink_args *);
106static int ext2_setattr (struct vop_setattr_args *);
107static int ext2_strategy (struct vop_strategy_args *);
108static int ext2_whiteout (struct vop_old_whiteout_args *);
109static int filt_ext2read (struct knote *kn, long hint);
110static int filt_ext2write (struct knote *kn, long hint);
111static int filt_ext2vnode (struct knote *kn, long hint);
112static void filt_ext2detach (struct knote *kn);
113static int ext2_kqfilter (struct vop_kqfilter_args *ap);
114static int ext2spec_close (struct vop_close_args *);
115static int ext2spec_read (struct vop_read_args *);
116static int ext2spec_write (struct vop_write_args *);
117static int ext2fifo_close (struct vop_close_args *);
118static int ext2fifo_kqfilter (struct vop_kqfilter_args *);
119static int ext2fifo_read (struct vop_read_args *);
120static int ext2fifo_write (struct vop_write_args *);
984263bc 121
a6ee311a
RG
122static int ext2_fsync (struct vop_fsync_args *);
123static int ext2_read (struct vop_read_args *);
124static int ext2_write (struct vop_write_args *);
e62afb5f
MD
125static int ext2_remove (struct vop_old_remove_args *);
126static int ext2_link (struct vop_old_link_args *);
127static int ext2_rename (struct vop_old_rename_args *);
128static int ext2_mkdir (struct vop_old_mkdir_args *);
129static int ext2_rmdir (struct vop_old_rmdir_args *);
130static int ext2_create (struct vop_old_create_args *);
131static int ext2_mknod (struct vop_old_mknod_args *);
132static int ext2_symlink (struct vop_old_symlink_args *);
a6ee311a
RG
133static int ext2_getpages (struct vop_getpages_args *);
134static int ext2_putpages (struct vop_putpages_args *);
984263bc 135
1f1db49f 136#include "ext2_readwrite.c"
984263bc 137
1f1db49f
MD
138union _qcvt {
139 int64_t qcvt;
140 int32_t val[2];
984263bc 141};
1f1db49f
MD
142#define SETHIGH(q, h) { \
143 union _qcvt tmp; \
144 tmp.qcvt = (q); \
145 tmp.val[_QUAD_HIGHWORD] = (h); \
146 (q) = tmp.qcvt; \
147}
148#define SETLOW(q, l) { \
149 union _qcvt tmp; \
150 tmp.qcvt = (q); \
151 tmp.val[_QUAD_LOWWORD] = (l); \
152 (q) = tmp.qcvt; \
153}
154#define VN_KNOTE(vp, b) \
155 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
984263bc 156
1f1db49f 157#define OFSFMT(vp) ((vp)->v_mount->mnt_maxsymlinklen <= 0)
984263bc
MD
158
159/*
160 * A virgin directory (no blushing please).
161 * Note that the type and namlen fields are reversed relative to ufs.
162 * Also, we don't use `struct odirtemplate', since it would just cause
163 * endianness problems.
164 */
1f1db49f 165static struct dirtemplate ext2_mastertemplate = {
984263bc
MD
166 0, 12, 1, EXT2_FT_DIR, ".",
167 0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
168};
1f1db49f 169static struct dirtemplate ext2_omastertemplate = {
984263bc
MD
170 0, 12, 1, EXT2_FT_UNKNOWN, ".",
171 0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
172};
173
174/*
175 * Create a regular file
0f7f7a49
CP
176 *
177 * ext2_create(struct vnode *a_dvp, struct vnode **a_vpp,
178 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
179 */
180static int
e62afb5f 181ext2_create(struct vop_old_create_args *ap)
984263bc
MD
182{
183 int error;
184
185 error =
186 ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
187 ap->a_dvp, ap->a_vpp, ap->a_cnp);
188 if (error)
189 return (error);
190 return (0);
191}
192
193/*
194 * Synch an open file.
0f7f7a49
CP
195 *
196 * ext2_fsync(struct vnode *a_vp, struct ucred *a_cred, int a_waitfor,
197 * struct proc *a_p)
984263bc
MD
198 */
199/* ARGSUSED */
6bae6177
MD
200
201static int ext2_fsync_bp(struct buf *bp, void *data);
202
203struct ext2_fsync_bp_info {
204 struct vnode *vp;
205 int waitfor;
6bae6177
MD
206};
207
984263bc 208static int
0f7f7a49 209ext2_fsync(struct vop_fsync_args *ap)
984263bc 210{
6bae6177 211 struct ext2_fsync_bp_info info;
f7aae92f 212 struct vnode *vp = ap->a_vp;
6bae6177 213 int count;
984263bc
MD
214
215 /*
216 * XXX why is all this fs specific?
217 */
218
219 /*
220 * Flush all dirty buffers associated with a vnode.
221 */
222 ext2_discard_prealloc(VTOI(vp));
223
165dba55 224 crit_enter();
6bae6177 225 info.vp = vp;
984263bc 226loop:
6bae6177
MD
227 info.waitfor = ap->a_waitfor;
228 count = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
229 ext2_fsync_bp, &info);
230 if (count)
984263bc 231 goto loop;
6bae6177 232
984263bc 233 if (ap->a_waitfor == MNT_WAIT) {
81b5c339
MD
234 while (vp->v_track_write.bk_active) {
235 vp->v_track_write.bk_waitflag = 1;
236 tsleep(&vp->v_track_write, 0, "e2fsyn", 0);
984263bc
MD
237 }
238#if DIAGNOSTIC
6bae6177 239 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
984263bc
MD
240 vprint("ext2_fsync: dirty", vp);
241 goto loop;
242 }
243#endif
244 }
165dba55 245 crit_exit();
1f1db49f 246 return (EXT2_UPDATE(ap->a_vp, ap->a_waitfor == MNT_WAIT));
984263bc
MD
247}
248
6bae6177
MD
249static int
250ext2_fsync_bp(struct buf *bp, void *data)
251{
252 struct ext2_fsync_bp_info *info = data;
253
254 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
255 return(0);
256 if ((bp->b_flags & B_DELWRI) == 0)
257 panic("ext2_fsync: not dirty");
258 bremfree(bp);
165dba55 259 crit_exit();
6bae6177
MD
260
261 /*
262 * Wait for I/O associated with indirect blocks to complete,
263 * since there is no way to quickly wait for them below.
264 */
265 if (bp->b_vp == info->vp || info->waitfor == MNT_NOWAIT)
b1ce5639 266 bawrite(bp);
6bae6177 267 else
b1ce5639 268 bwrite(bp);
165dba55 269 crit_enter();
6bae6177
MD
270 return(1);
271}
272
984263bc
MD
273/*
274 * Mknod vnode call
0f7f7a49
CP
275 *
276 * ext2_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
277 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
278 */
279/* ARGSUSED */
280static int
e62afb5f 281ext2_mknod(struct vop_old_mknod_args *ap)
984263bc
MD
282{
283 struct vattr *vap = ap->a_vap;
284 struct vnode **vpp = ap->a_vpp;
285 struct inode *ip;
286 ino_t ino;
287 int error;
288
289 error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
290 ap->a_dvp, vpp, ap->a_cnp);
291 if (error)
292 return (error);
293 ip = VTOI(*vpp);
294 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
295 if (vap->va_rdev != VNOVAL) {
296 /*
297 * Want to be able to use this to make badblock
298 * inodes, so don't truncate the dev number.
299 */
300 ip->i_rdev = vap->va_rdev;
301 }
302 /*
303 * Remove inode, then reload it through VFS_VGET so it is
304 * checked to see if it is an alias of an existing entry in
305 * the inode cache.
306 */
984263bc
MD
307 (*vpp)->v_type = VNON;
308 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
309 vgone(*vpp);
5fd012e0 310 vput(*vpp);
984263bc
MD
311 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
312 if (error) {
313 *vpp = NULL;
314 return (error);
315 }
316 return (0);
317}
318
0f7f7a49
CP
319/*
320 * ext2_remove(struct vnode *a_dvp, struct vnode *a_vp,
321 * struct componentname *a_cnp)
322 */
984263bc 323static int
e62afb5f 324ext2_remove(struct vop_old_remove_args *ap)
984263bc
MD
325{
326 struct inode *ip;
327 struct vnode *vp = ap->a_vp;
328 struct vnode *dvp = ap->a_dvp;
329 int error;
330
331 ip = VTOI(vp);
332 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
333 (VTOI(dvp)->i_flags & APPEND)) {
334 error = EPERM;
335 goto out;
336 }
337 error = ext2_dirremove(dvp, ap->a_cnp);
338 if (error == 0) {
339 ip->i_nlink--;
340 ip->i_flag |= IN_CHANGE;
341 }
342out:
343 return (error);
344}
345
346/*
347 * link vnode call
0f7f7a49
CP
348 *
349 * ext2_link(struct vnode *a_tdvp, struct vnode *a_vp,
350 * struct componentname *a_cnp)
984263bc
MD
351 */
352static int
e62afb5f 353ext2_link(struct vop_old_link_args *ap)
984263bc
MD
354{
355 struct vnode *vp = ap->a_vp;
356 struct vnode *tdvp = ap->a_tdvp;
357 struct componentname *cnp = ap->a_cnp;
7b95be2a 358 struct thread *td = cnp->cn_td;
984263bc
MD
359 struct inode *ip;
360 int error;
361
984263bc
MD
362 if (tdvp->v_mount != vp->v_mount) {
363 error = EXDEV;
364 goto out2;
365 }
5fd012e0 366 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, td))) {
984263bc
MD
367 goto out2;
368 }
369 ip = VTOI(vp);
370 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
371 error = EMLINK;
372 goto out1;
373 }
374 if (ip->i_flags & (IMMUTABLE | APPEND)) {
375 error = EPERM;
376 goto out1;
377 }
378 ip->i_nlink++;
379 ip->i_flag |= IN_CHANGE;
1f1db49f 380 error = EXT2_UPDATE(vp, 1);
984263bc
MD
381 if (!error)
382 error = ext2_direnter(ip, tdvp, cnp);
383 if (error) {
384 ip->i_nlink--;
385 ip->i_flag |= IN_CHANGE;
386 }
387out1:
388 if (tdvp != vp)
5fd012e0 389 VOP_UNLOCK(vp, 0, td);
984263bc
MD
390out2:
391 return (error);
392}
393
394/*
fad57d0e
MD
395 * Rename system call. fdvp, fvp are ref'd. tvp, tdvp are ref'd and locked.
396 * all vp's are released and must be in an unlocked state on return.
0f7f7a49
CP
397 *
398 * ext2_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
399 * struct componentname *a_fcnp, struct vnode *a_tdvp,
400 * struct vnode *a_tvp, struct componentname *a_tcnp)
984263bc
MD
401 */
402static int
e62afb5f 403ext2_rename(struct vop_old_rename_args *ap)
984263bc
MD
404{
405 struct vnode *tvp = ap->a_tvp;
f7aae92f 406 struct vnode *tdvp = ap->a_tdvp;
984263bc
MD
407 struct vnode *fvp = ap->a_fvp;
408 struct vnode *fdvp = ap->a_fdvp;
409 struct componentname *tcnp = ap->a_tcnp;
410 struct componentname *fcnp = ap->a_fcnp;
7b95be2a 411 struct thread *td = fcnp->cn_td;
984263bc
MD
412 struct inode *ip, *xp, *dp;
413 struct dirtemplate dirbuf;
414 int doingdirectory = 0, oldparent = 0, newparent = 0;
415 int error = 0;
416 u_char namlen;
417
984263bc
MD
418 /*
419 * Check for cross-device rename.
420 */
421 if ((fvp->v_mount != tdvp->v_mount) ||
fad57d0e
MD
422 (tvp && (fvp->v_mount != tvp->v_mount)) ||
423 tvp == tdvp) {
984263bc
MD
424 error = EXDEV;
425abortit:
426 if (tdvp == tvp)
427 vrele(tdvp);
428 else
429 vput(tdvp);
430 if (tvp)
431 vput(tvp);
432 vrele(fdvp);
433 vrele(fvp);
434 return (error);
435 }
436
437 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
438 (VTOI(tdvp)->i_flags & APPEND))) {
439 error = EPERM;
440 goto abortit;
441 }
442
443 /*
444 * Renaming a file to itself has no effect. The upper layers should
445 * not call us in that case. Temporarily just warn if they do.
446 */
447 if (fvp == tvp) {
984263bc
MD
448 error = 0;
449 goto abortit;
450 }
451
5fd012e0 452 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
984263bc 453 goto abortit;
fad57d0e
MD
454
455 /*
456 * fvp, tvp, tdvp locked. fdvp not locked but note that fdvp may
457 * be equal to tdvp.
458 */
984263bc
MD
459 dp = VTOI(fdvp);
460 ip = VTOI(fvp);
461 if (ip->i_nlink >= LINK_MAX) {
5fd012e0 462 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
463 error = EMLINK;
464 goto abortit;
465 }
466 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
467 || (dp->i_flags & APPEND)) {
5fd012e0 468 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
469 error = EPERM;
470 goto abortit;
471 }
472 if ((ip->i_mode & IFMT) == IFDIR) {
473 /*
474 * Avoid ".", "..", and aliases of "." for obvious reasons.
475 */
476 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
2b69e610 477 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
984263bc 478 (ip->i_flag & IN_RENAME)) {
5fd012e0 479 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
480 error = EINVAL;
481 goto abortit;
482 }
483 ip->i_flag |= IN_RENAME;
484 oldparent = dp->i_number;
485 doingdirectory++;
486 }
984263bc
MD
487
488 /*
fad57d0e
MD
489 * tvp is non-NULL if the target exists. fvp is still locked but
490 * we will unlock it soon. The 'bad' goto target requires dp and
491 * xp to be correctly assigned.
984263bc
MD
492 */
493 dp = VTOI(tdvp);
984263bc
MD
494 if (tvp)
495 xp = VTOI(tvp);
fad57d0e
MD
496 else
497 xp = NULL;
984263bc
MD
498
499 /*
500 * 1) Bump link count while we're moving stuff
501 * around. If we crash somewhere before
502 * completing our work, the link count
503 * may be wrong, but correctable.
504 */
505 ip->i_nlink++;
506 ip->i_flag |= IN_CHANGE;
1f1db49f 507 if ((error = EXT2_UPDATE(fvp, 1)) != 0) {
5fd012e0 508 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
509 goto bad;
510 }
511
512 /*
513 * If ".." must be changed (ie the directory gets a new
514 * parent) then the source directory must not be in the
515 * directory heirarchy above the target, as this would
516 * orphan everything below the source directory. Also
517 * the user must have write permission in the source so
518 * as to be able to change "..". We must repeat the call
519 * to namei, as the parent directory is unlocked by the
520 * call to checkpath().
521 */
7b95be2a 522 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
5fd012e0 523 VOP_UNLOCK(fvp, 0, td);
fad57d0e
MD
524
525 /*
526 * tvp (if not NULL) and tdvp are locked. fvp and fdvp are not.
527 * dp and xp are set according to tdvp and tvp.
528 */
984263bc
MD
529 if (oldparent != dp->i_number)
530 newparent = dp->i_number;
531 if (doingdirectory && newparent) {
532 if (error) /* write access check above */
533 goto bad;
fad57d0e
MD
534
535 /*
536 * Prepare for relookup, get rid of xp
537 */
538 if (xp != NULL) {
984263bc 539 vput(tvp);
fad57d0e
MD
540 xp = NULL;
541 }
542
543 /*
544 * checkpath vput()'s tdvp (VTOI(dp)) on return no matter what,
545 * get an extra ref so we wind up with just an unlocked, ref'd
546 * tdvp. The 'out' target skips xp and tdvp cleanups. Our
547 * tdvp is now unlocked so we have to clean it up ourselves.
548 */
549 vref(tdvp);
984263bc 550 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
6970b52a 551 tcnp->cn_flags |= CNP_PDIRUNLOCK;
fad57d0e
MD
552 if (error) {
553 vrele(tdvp);
984263bc 554 goto out;
fad57d0e
MD
555 }
556 /*
557 * relookup no longer messes with the ref count. An unlocked
558 * tdvp must be passed and if no error occurs a locked tdvp
559 * will be returned. We have to use the out target again.
560 */
984263bc 561 error = relookup(tdvp, &tvp, tcnp);
fad57d0e
MD
562 if (error) {
563 if (tcnp->cn_flags & CNP_PDIRUNLOCK)
564 vrele(tdvp);
565 else
566 vput(tdvp);
984263bc 567 goto out;
fad57d0e
MD
568 }
569
570 /*
571 * tdvp is locked at this point. in the RENAME case tvp may
572 * be NULL without an error, assign xp accordingly. The
573 * 'bad' target can be used again after this.
574 */
984263bc 575 dp = VTOI(tdvp);
984263bc
MD
576 if (tvp)
577 xp = VTOI(tvp);
578 }
579 /*
580 * 2) If target doesn't exist, link the target
581 * to the source and unlink the source.
582 * Otherwise, rewrite the target directory
583 * entry to reference the source inode and
584 * expunge the original entry's existence.
fad57d0e
MD
585 *
586 * tdvp and tvp are cleaned up by this code. tvp is only good if
587 * xp is not NULL.
984263bc
MD
588 */
589 if (xp == NULL) {
590 if (dp->i_dev != ip->i_dev)
1f1db49f 591 panic("ext2_rename: EXDEV");
984263bc
MD
592 /*
593 * Account for ".." in new directory.
594 * When source and destination have the same
595 * parent we don't fool with the link count.
596 */
597 if (doingdirectory && newparent) {
598 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
599 error = EMLINK;
600 goto bad;
601 }
602 dp->i_nlink++;
603 dp->i_flag |= IN_CHANGE;
1f1db49f 604 error = EXT2_UPDATE(tdvp, 1);
984263bc
MD
605 if (error)
606 goto bad;
607 }
608 error = ext2_direnter(ip, tdvp, tcnp);
609 if (error) {
610 if (doingdirectory && newparent) {
611 dp->i_nlink--;
612 dp->i_flag |= IN_CHANGE;
1f1db49f 613 EXT2_UPDATE(tdvp, 1);
984263bc
MD
614 }
615 goto bad;
616 }
fad57d0e
MD
617
618 /*
619 * manual cleanup, we can't use the bad or out target after
620 * this.
621 */
984263bc
MD
622 vput(tdvp);
623 } else {
624 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1f1db49f 625 panic("ext2_rename: EXDEV");
984263bc
MD
626 /*
627 * Short circuit rename(foo, foo).
628 */
629 if (xp->i_number == ip->i_number)
1f1db49f 630 panic("ext2_rename: same file");
984263bc
MD
631 /*
632 * If the parent directory is "sticky", then the user must
633 * own the parent directory, or the destination of the rename,
634 * otherwise the destination may not be changed (except by
635 * root). This implements append-only directories.
636 */
637 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
638 tcnp->cn_cred->cr_uid != dp->i_uid &&
639 xp->i_uid != tcnp->cn_cred->cr_uid) {
640 error = EPERM;
641 goto bad;
642 }
643 /*
644 * Target must be empty if a directory and have no links
645 * to it. Also, ensure source and target are compatible
646 * (both directories, or both not directories).
647 */
648 if ((xp->i_mode&IFMT) == IFDIR) {
649 if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
650 xp->i_nlink > 2) {
651 error = ENOTEMPTY;
652 goto bad;
653 }
654 if (!doingdirectory) {
655 error = ENOTDIR;
656 goto bad;
657 }
984263bc
MD
658 } else if (doingdirectory) {
659 error = EISDIR;
660 goto bad;
661 }
662 error = ext2_dirrewrite(dp, ip, tcnp);
663 if (error)
664 goto bad;
665 /*
666 * If the target directory is in the same
667 * directory as the source directory,
668 * decrement the link count on the parent
669 * of the target directory.
670 */
671 if (doingdirectory && !newparent) {
672 dp->i_nlink--;
673 dp->i_flag |= IN_CHANGE;
674 }
fad57d0e
MD
675
676 /*
677 * manual cleanup, we can't use the bad or out target after
678 * this.
679 */
984263bc 680 vput(tdvp);
fad57d0e 681
984263bc
MD
682 /*
683 * Adjust the link count of the target to
684 * reflect the dirrewrite above. If this is
685 * a directory it is empty and there are
686 * no links to it, so we can squash the inode and
687 * any space associated with it. We disallowed
688 * renaming over top of a directory with links to
689 * it above, as the remaining link would point to
690 * a directory without "." or ".." entries.
691 */
692 xp->i_nlink--;
693 if (doingdirectory) {
694 if (--xp->i_nlink != 0)
1f1db49f
MD
695 panic("ext2_rename: linked directory");
696 error = EXT2_TRUNCATE(tvp, (off_t)0, IO_SYNC,
7b95be2a 697 tcnp->cn_cred, tcnp->cn_td);
984263bc
MD
698 }
699 xp->i_flag |= IN_CHANGE;
700 vput(tvp);
701 xp = NULL;
702 }
703
fad57d0e
MD
704 /*
705 * tvp and tdvp have been cleaned up. The bad and out targets may
706 * not be used. fvp and fdvp are ref'd but not locked. ip
707 * still represents the old fvp and ip->i_flag may still have IN_RENAME
708 * set (if doingdirectory).
709 */
710
984263bc
MD
711 /*
712 * 3) Unlink the source.
fad57d0e
MD
713 *
714 * fdvp is locked and ref'd. ap->a_fvp holds the old lookup unlocked
715 * and ref'd, fvp will hold the new lookup locked and ref'd.
716 *
717 * After the relookup ap->a_fvp must be released as part of our
718 * cleanup, not just fdvp and fvp. And, on success, fdvp and
719 * fvp will be locked so the bad and out targets cannot be used.
984263bc 720 */
2b69e610 721 fcnp->cn_flags &= ~CNP_MODMASK;
fad57d0e
MD
722 fcnp->cn_flags |= CNP_LOCKPARENT;
723 KKASSERT(fcnp->cn_flags & CNP_PDIRUNLOCK);
984263bc 724 error = relookup(fdvp, &fvp, fcnp);
fad57d0e 725 if (error) {
984263bc
MD
726 /*
727 * From name has disappeared.
728 */
729 if (doingdirectory)
1f1db49f 730 panic("ext2_rename: lost dir entry");
fad57d0e 731 /* ip->i_flag only sets IN_RENAME if doingdirectory */
984263bc 732 vrele(ap->a_fvp);
fad57d0e
MD
733 if (fcnp->cn_flags & CNP_PDIRUNLOCK)
734 vrele(fdvp);
735 else
736 vput(fdvp);
984263bc
MD
737 return (0);
738 }
fad57d0e
MD
739 KKASSERT((fcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
740
741 /*
742 * This case shouldn't occur
743 */
744 if (fvp == NULL) {
745 /*
746 * From name has disappeared.
747 */
748 if (doingdirectory)
1f1db49f 749 panic("ext2_rename: lost dir entry");
fad57d0e
MD
750 /* ip->i_flag only sets IN_RENAME if doingdirectory */
751 vrele(ap->a_fvp);
752 vput(fvp);
753 vput(fdvp);
754 return (0);
755 }
756
757 /*
758 * fvp and fdvp are both ref'd and locked.
759 */
760 xp = VTOI(fvp);
761 dp = VTOI(fdvp);
762
984263bc
MD
763 /*
764 * Ensure that the directory entry still exists and has not
765 * changed while the new name has been entered. If the source is
766 * a file then the entry may have been unlinked or renamed. In
767 * either case there is no further work to be done. If the source
768 * is a directory then it cannot have been rmdir'ed; its link
769 * count of three would cause a rmdir to fail with ENOTEMPTY.
770 * The IN_RENAME flag ensures that it cannot be moved by another
771 * rename.
772 */
773 if (xp != ip) {
774 if (doingdirectory)
1f1db49f 775 panic("ext2_rename: lost dir entry");
fad57d0e 776 /* ip->i_flag only sets IN_RENAME if doingdirectory */
984263bc
MD
777 } else {
778 /*
779 * If the source is a directory with a
780 * new parent, the link count of the old
781 * parent directory must be decremented
782 * and ".." set to point to the new parent.
783 */
784 if (doingdirectory && newparent) {
785 dp->i_nlink--;
786 dp->i_flag |= IN_CHANGE;
787 error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
788 sizeof (struct dirtemplate), (off_t)0,
789 UIO_SYSSPACE, IO_NODELOCKED,
7b95be2a 790 tcnp->cn_cred, (int *)0, NULL);
984263bc 791 if (error == 0) {
1f1db49f 792 /* Like ext2 little-endian: */
984263bc
MD
793 namlen = dirbuf.dotdot_type;
794 if (namlen != 2 ||
795 dirbuf.dotdot_name[0] != '.' ||
796 dirbuf.dotdot_name[1] != '.') {
1f1db49f 797 ext2_dirbad(xp, (doff_t)12,
984263bc
MD
798 "rename: mangled dir");
799 } else {
800 dirbuf.dotdot_ino = newparent;
b1ce5639 801 vn_rdwr(UIO_WRITE, fvp,
984263bc
MD
802 (caddr_t)&dirbuf,
803 sizeof (struct dirtemplate),
804 (off_t)0, UIO_SYSSPACE,
805 IO_NODELOCKED|IO_SYNC,
806 tcnp->cn_cred, (int *)0,
7b95be2a 807 NULL);
984263bc
MD
808 }
809 }
810 }
811 error = ext2_dirremove(fdvp, fcnp);
812 if (!error) {
813 xp->i_nlink--;
814 xp->i_flag |= IN_CHANGE;
815 }
816 xp->i_flag &= ~IN_RENAME;
817 }
fad57d0e
MD
818 vput(fdvp);
819 vput(fvp);
984263bc
MD
820 vrele(ap->a_fvp);
821 return (error);
822
823bad:
824 if (xp)
825 vput(ITOV(xp));
fad57d0e
MD
826 if (dp)
827 vput(ITOV(dp));
984263bc
MD
828out:
829 if (doingdirectory)
830 ip->i_flag &= ~IN_RENAME;
5fd012e0 831 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
984263bc
MD
832 ip->i_nlink--;
833 ip->i_flag |= IN_CHANGE;
834 ip->i_flag &= ~IN_RENAME;
835 vput(fvp);
fad57d0e 836 } else {
984263bc 837 vrele(fvp);
fad57d0e 838 }
984263bc
MD
839 return (error);
840}
841
842/*
843 * Mkdir system call
0f7f7a49
CP
844 *
845 * ext2_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
846 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
847 */
848static int
e62afb5f 849ext2_mkdir(struct vop_old_mkdir_args *ap)
984263bc 850{
f7aae92f
RG
851 struct vnode *dvp = ap->a_dvp;
852 struct vattr *vap = ap->a_vap;
853 struct componentname *cnp = ap->a_cnp;
854 struct inode *ip, *dp;
984263bc
MD
855 struct vnode *tvp;
856 struct dirtemplate dirtemplate, *dtp;
857 int error, dmode;
858
984263bc
MD
859 dp = VTOI(dvp);
860 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
861 error = EMLINK;
862 goto out;
863 }
864 dmode = vap->va_mode & 0777;
865 dmode |= IFDIR;
866 /*
867 * Must simulate part of ext2_makeinode here to acquire the inode,
868 * but not have it entered in the parent directory. The entry is
869 * made later after writing "." and ".." entries.
870 */
1f1db49f 871 error = EXT2_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
984263bc
MD
872 if (error)
873 goto out;
874 ip = VTOI(tvp);
875 ip->i_gid = dp->i_gid;
876#ifdef SUIDDIR
877 {
878#ifdef QUOTA
879 struct ucred ucred, *ucp;
880 ucp = cnp->cn_cred;
1d94a9ab 881#endif
984263bc
MD
882 /*
883 * if we are hacking owners here, (only do this where told to)
884 * and we are not giving it TOO root, (would subvert quotas)
885 * then go ahead and give it to the other user.
886 * The new directory also inherits the SUID bit.
887 * If user's UID and dir UID are the same,
888 * 'give it away' so that the SUID is still forced on.
889 */
890 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
891 (dp->i_mode & ISUID) && dp->i_uid) {
892 dmode |= ISUID;
893 ip->i_uid = dp->i_uid;
894#ifdef QUOTA
895 if (dp->i_uid != cnp->cn_cred->cr_uid) {
896 /*
897 * make sure the correct user gets charged
898 * for the space.
899 * Make a dummy credential for the victim.
900 * XXX This seems to never be accessed out of
901 * our context so a stack variable is ok.
902 */
903 ucred.cr_ref = 1;
904 ucred.cr_uid = ip->i_uid;
905 ucred.cr_ngroups = 1;
906 ucred.cr_groups[0] = dp->i_gid;
907 ucp = &ucred;
908 }
1d94a9ab 909#endif
984263bc
MD
910 } else {
911 ip->i_uid = cnp->cn_cred->cr_uid;
912 }
913#ifdef QUOTA
1f1db49f
MD
914 if ((error = ext2_getinoquota(ip)) ||
915 (error = ext2_chkiq(ip, 1, ucp, 0))) {
916 EXT2_VFREE(tvp, ip->i_number, dmode);
984263bc
MD
917 vput(tvp);
918 return (error);
919 }
920#endif
921 }
922#else
923 ip->i_uid = cnp->cn_cred->cr_uid;
924#ifdef QUOTA
1f1db49f
MD
925 if ((error = ext2_getinoquota(ip)) ||
926 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
927 EXT2_VFREE(tvp, ip->i_number, dmode);
984263bc
MD
928 vput(tvp);
929 return (error);
930 }
931#endif
932#endif
933 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
934 ip->i_mode = dmode;
935 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
936 ip->i_nlink = 2;
2b69e610 937 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc 938 ip->i_flags |= UF_OPAQUE;
1f1db49f 939 error = EXT2_UPDATE(tvp, 1);
984263bc 940
93f50d60
MD
941 /*
942 * The vnode must have a VM object in order to issue buffer cache
943 * ops on it.
944 */
1c843a13 945 vinitvmio(tvp, 0);
93f50d60 946
984263bc
MD
947 /*
948 * Bump link count in parent directory
949 * to reflect work done below. Should
950 * be done before reference is created
951 * so reparation is possible if we crash.
952 */
953 dp->i_nlink++;
954 dp->i_flag |= IN_CHANGE;
1f1db49f 955 error = EXT2_UPDATE(dvp, 1);
984263bc
MD
956 if (error)
957 goto bad;
958
959 /* Initialize directory with "." and ".." from static template. */
960 if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
961 EXT2_FEATURE_INCOMPAT_FILETYPE))
1f1db49f 962 dtp = &ext2_mastertemplate;
984263bc 963 else
1f1db49f 964 dtp = &ext2_omastertemplate;
984263bc
MD
965 dirtemplate = *dtp;
966 dirtemplate.dot_ino = ip->i_number;
967 dirtemplate.dotdot_ino = dp->i_number;
968 /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
969 * so let's just redefine it - for this function only
970 */
971#undef DIRBLKSIZ
972#define DIRBLKSIZ VTOI(dvp)->i_e2fs->s_blocksize
973 dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
974 error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
975 sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
7b95be2a 976 IO_NODELOCKED|IO_SYNC, cnp->cn_cred, (int *)0, NULL);
984263bc
MD
977 if (error) {
978 dp->i_nlink--;
979 dp->i_flag |= IN_CHANGE;
980 goto bad;
981 }
1f1db49f
MD
982 if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
983 panic("ext2_mkdir: blksize"); /* XXX should grow with balloc() */
984263bc
MD
984 else {
985 ip->i_size = DIRBLKSIZ;
986 ip->i_flag |= IN_CHANGE;
987 }
988
989 /* Directory set up, now install its entry in the parent directory. */
990 error = ext2_direnter(ip, dvp, cnp);
991 if (error) {
992 dp->i_nlink--;
993 dp->i_flag |= IN_CHANGE;
994 }
995bad:
996 /*
997 * No need to do an explicit VOP_TRUNCATE here, vrele will do this
998 * for us because we set the link count to 0.
999 */
1000 if (error) {
1001 ip->i_nlink = 0;
1002 ip->i_flag |= IN_CHANGE;
1003 vput(tvp);
1004 } else
1005 *ap->a_vpp = tvp;
1006out:
1007 return (error);
1008#undef DIRBLKSIZ
1009#define DIRBLKSIZ DEV_BSIZE
1010}
1011
1012/*
1013 * Rmdir system call.
0f7f7a49
CP
1014 *
1015 * ext2_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1016 * struct componentname *a_cnp)
984263bc
MD
1017 */
1018static int
e62afb5f 1019ext2_rmdir(struct vop_old_rmdir_args *ap)
984263bc
MD
1020{
1021 struct vnode *vp = ap->a_vp;
1022 struct vnode *dvp = ap->a_dvp;
1023 struct componentname *cnp = ap->a_cnp;
7b95be2a 1024 struct thread *td = cnp->cn_td;
984263bc
MD
1025 struct inode *ip, *dp;
1026 int error;
1027
1028 ip = VTOI(vp);
1029 dp = VTOI(dvp);
1030
1031 /*
1032 * Verify the directory is empty (and valid).
1033 * (Rmdir ".." won't be valid since
1034 * ".." will contain a reference to
1035 * the current directory and thus be
1036 * non-empty.)
1037 */
1038 error = 0;
1039 if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1040 error = ENOTEMPTY;
1041 goto out;
1042 }
1043 if ((dp->i_flags & APPEND)
1044 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1045 error = EPERM;
1046 goto out;
1047 }
1048 /*
1049 * Delete reference to directory before purging
1050 * inode. If we crash in between, the directory
1051 * will be reattached to lost+found,
1052 */
1053 error = ext2_dirremove(dvp, cnp);
1054 if (error)
1055 goto out;
1056 dp->i_nlink--;
1057 dp->i_flag |= IN_CHANGE;
5fd012e0 1058 VOP_UNLOCK(dvp, 0, td);
984263bc
MD
1059 /*
1060 * Truncate inode. The only stuff left
1061 * in the directory is "." and "..". The
1062 * "." reference is inconsequential since
1063 * we're quashing it. The ".." reference
1064 * has already been adjusted above. We've
1065 * removed the "." reference and the reference
1066 * in the parent directory, but there may be
1067 * other hard links so decrement by 2 and
1068 * worry about them later.
1069 */
1070 ip->i_nlink -= 2;
1f1db49f 1071 error = EXT2_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred, td);
5fd012e0 1072 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
1073out:
1074 return (error);
1075}
1076
1077/*
1078 * symlink -- make a symbolic link
0f7f7a49
CP
1079 *
1080 * ext2_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1081 * struct componentname *a_cnp, struct vattr *a_vap,
1082 * char *a_target)
984263bc
MD
1083 */
1084static int
e62afb5f 1085ext2_symlink(struct vop_old_symlink_args *ap)
984263bc 1086{
f7aae92f
RG
1087 struct vnode *vp, **vpp = ap->a_vpp;
1088 struct inode *ip;
984263bc
MD
1089 int len, error;
1090
1091 error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1092 vpp, ap->a_cnp);
1093 if (error)
1094 return (error);
1095 vp = *vpp;
1096 len = strlen(ap->a_target);
1097 if (len < vp->v_mount->mnt_maxsymlinklen) {
1098 ip = VTOI(vp);
1099 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1100 ip->i_size = len;
1101 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1102 } else
93f50d60
MD
1103 /*
1104 * Make sure we have a VM object in order to use
1105 * the buffer cache.
1106 */
1107 if (vp->v_object == NULL)
1c843a13 1108 vinitvmio(vp, 0);
93f50d60 1109
984263bc
MD
1110 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1111 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred, (int *)0,
7b95be2a 1112 NULL);
984263bc
MD
1113 if (error)
1114 vput(vp);
1115 return (error);
1116}
1117
1118/*
1119 * Allocate a new inode.
1120 */
1121static int
0f7f7a49
CP
1122ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1123 struct componentname *cnp)
984263bc 1124{
f7aae92f 1125 struct inode *ip, *pdir;
984263bc
MD
1126 struct vnode *tvp;
1127 int error;
1128
1129 pdir = VTOI(dvp);
984263bc
MD
1130 *vpp = NULL;
1131 if ((mode & IFMT) == 0)
1132 mode |= IFREG;
1133
1f1db49f 1134 error = EXT2_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
984263bc
MD
1135 if (error) {
1136 return (error);
1137 }
1138 ip = VTOI(tvp);
1139 ip->i_gid = pdir->i_gid;
1140#ifdef SUIDDIR
1141 {
1142#ifdef QUOTA
1143 struct ucred ucred, *ucp;
1144 ucp = cnp->cn_cred;
1d94a9ab 1145#endif
984263bc
MD
1146 /*
1147 * if we are
1148 * not the owner of the directory,
1149 * and we are hacking owners here, (only do this where told to)
1150 * and we are not giving it TOO root, (would subvert quotas)
1151 * then go ahead and give it to the other user.
1152 * Note that this drops off the execute bits for security.
1153 */
1154 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1155 (pdir->i_mode & ISUID) &&
1156 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1157 ip->i_uid = pdir->i_uid;
1158 mode &= ~07111;
1159#ifdef QUOTA
1160 /*
1161 * make sure the correct user gets charged
1162 * for the space.
1163 * Quickly knock up a dummy credential for the victim.
1164 * XXX This seems to never be accessed out of our
1165 * context so a stack variable is ok.
1166 */
1167 ucred.cr_ref = 1;
1168 ucred.cr_uid = ip->i_uid;
1169 ucred.cr_ngroups = 1;
1170 ucred.cr_groups[0] = pdir->i_gid;
1171 ucp = &ucred;
1d94a9ab 1172#endif
984263bc
MD
1173 } else {
1174 ip->i_uid = cnp->cn_cred->cr_uid;
1175 }
1176
1177#ifdef QUOTA
6f1a0d44 1178 if ((error = ext2_getinoquota(ip)) ||
1f1db49f
MD
1179 (error = ext2_chkiq(ip, 1, ucp, 0))) {
1180 EXT2_VFREE(tvp, ip->i_number, mode);
984263bc
MD
1181 vput(tvp);
1182 return (error);
1183 }
1184#endif
1185 }
1186#else
1187 ip->i_uid = cnp->cn_cred->cr_uid;
1188#ifdef QUOTA
1f1db49f
MD
1189 if ((error = ext2_getinoquota(ip)) ||
1190 (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
1191 EXT2_VFREE(tvp, ip->i_number, mode);
984263bc
MD
1192 vput(tvp);
1193 return (error);
1194 }
1195#endif
1196#endif
1197 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1198 ip->i_mode = mode;
1199 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
1200 ip->i_nlink = 1;
1201 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
dadab5e9 1202 suser_cred(cnp->cn_cred, PRISON_ROOT))
984263bc
MD
1203 ip->i_mode &= ~ISGID;
1204
2b69e610 1205 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
1206 ip->i_flags |= UF_OPAQUE;
1207
1c843a13
MD
1208 /*
1209 * Regular files and directories need VM objects. Softlinks do
1210 * not (not immediately anyway).
1211 */
1212 if (tvp->v_type == VREG || tvp->v_type == VDIR)
1213 vinitvmio(tvp, 0);
1214
984263bc
MD
1215 /*
1216 * Make sure inode goes to disk before directory entry.
1217 */
1f1db49f 1218 error = EXT2_UPDATE(tvp, 1);
984263bc
MD
1219 if (error)
1220 goto bad;
1221 error = ext2_direnter(ip, dvp, cnp);
1222 if (error)
1223 goto bad;
1224
1225 *vpp = tvp;
1226 return (0);
1227
1228bad:
1229 /*
1230 * Write error occurred trying to update the inode
1231 * or the directory so must deallocate the inode.
1232 */
1233 ip->i_nlink = 0;
1234 ip->i_flag |= IN_CHANGE;
1235 vput(tvp);
1236 return (error);
1237}
1238
1239/*
1240 * get page routine
1241 *
1242 * XXX By default, wimp out... note that a_offset is ignored (and always
1243 * XXX has been).
1244 */
1245static int
0f7f7a49 1246ext2_getpages(struct vop_getpages_args *ap)
984263bc
MD
1247{
1248 return (vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
1249 ap->a_reqpage));
1250}
1251
1252/*
1253 * put page routine
1254 *
1255 * XXX By default, wimp out... note that a_offset is ignored (and always
1256 * XXX has been).
1257 */
1258static int
0f7f7a49 1259ext2_putpages(struct vop_putpages_args *ap)
984263bc
MD
1260{
1261 return (vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
1262 ap->a_sync, ap->a_rtvals));
1263}
1f1db49f
MD
1264
1265void
1266ext2_itimes(struct vnode *vp)
1267{
1268 struct inode *ip;
1269 struct timespec ts;
1270
1271 ip = VTOI(vp);
1272 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
1273 return;
1274 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
1275 ip->i_flag |= IN_LAZYMOD;
1276 else
1277 ip->i_flag |= IN_MODIFIED;
1278 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
1279 vfs_timestamp(&ts);
1280 if (ip->i_flag & IN_ACCESS) {
1281 ip->i_atime = ts.tv_sec;
1282 ip->i_atimensec = ts.tv_nsec;
1283 }
1284 if (ip->i_flag & IN_UPDATE) {
1285 ip->i_mtime = ts.tv_sec;
1286 ip->i_mtimensec = ts.tv_nsec;
1287 ip->i_modrev++;
1288 }
1289 if (ip->i_flag & IN_CHANGE) {
1290 ip->i_ctime = ts.tv_sec;
1291 ip->i_ctimensec = ts.tv_nsec;
1292 }
1293 }
1294 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
1295}
1296
1297/*
1298 * Open called.
1299 *
1300 * Nothing to do.
1301 *
1302 * ext2_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1303 * struct thread *a_td)
1304 */
1305/* ARGSUSED */
1306static
1307int
1308ext2_open(struct vop_open_args *ap)
1309{
1310 struct vnode *vp = ap->a_vp;
1311
1312 /*
1313 * Files marked append-only must be opened for appending.
1314 */
1315 if ((VTOI(vp)->i_flags & APPEND) &&
1316 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) {
1317 return (EPERM);
1318 }
1f1db49f
MD
1319 return (vop_stdopen(ap));
1320}
1321
1322/*
1323 * Close called.
1324 *
1325 * Update the times on the inode.
1326 *
1327 * ext2_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1328 * struct thread *a_td)
1329 */
1330/* ARGSUSED */
1331static
1332int
1333ext2_close(struct vop_close_args *ap)
1334{
1335 struct vnode *vp = ap->a_vp;
1336
1337 if (vp->v_usecount > 1)
1338 ext2_itimes(vp);
1339 return (vop_stdclose(ap));
1340}
1341
1342/*
1343 * ext2_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1344 * struct thread *a_td)
1345 */
1346static
1347int
1348ext2_access(struct vop_access_args *ap)
1349{
1350 struct vnode *vp = ap->a_vp;
1351 struct inode *ip = VTOI(vp);
1352 struct ucred *cred = ap->a_cred;
1353 mode_t mask, mode = ap->a_mode;
1354 gid_t *gp;
1355 int i;
1356#ifdef QUOTA
1357 int error;
1358#endif
1359
1360 /*
1361 * Disallow write attempts on read-only filesystems;
1362 * unless the file is a socket, fifo, or a block or
1363 * character device resident on the filesystem.
1364 */
1365 if (mode & VWRITE) {
1366 switch (vp->v_type) {
1367 case VDIR:
1368 case VLNK:
1369 case VREG:
1370 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1371 return (EROFS);
1372#ifdef QUOTA
1373 if ((error = ext2_getinoquota(ip)) != 0)
1374 return (error);
1375#endif
1376 break;
1377 default:
1378 break;
1379 }
1380 }
1381
1382 /* If immutable bit set, nobody gets to write it. */
1383 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
1384 return (EPERM);
1385
1386 /* Otherwise, user id 0 always gets access. */
1387 if (cred->cr_uid == 0)
1388 return (0);
1389
1390 mask = 0;
1391
1392 /* Otherwise, check the owner. */
1393 if (cred->cr_uid == ip->i_uid) {
1394 if (mode & VEXEC)
1395 mask |= S_IXUSR;
1396 if (mode & VREAD)
1397 mask |= S_IRUSR;
1398 if (mode & VWRITE)
1399 mask |= S_IWUSR;
1400 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1401 }
1402
1403 /* Otherwise, check the groups. */
1404 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
1405 if (ip->i_gid == *gp) {
1406 if (mode & VEXEC)
1407 mask |= S_IXGRP;
1408 if (mode & VREAD)
1409 mask |= S_IRGRP;
1410 if (mode & VWRITE)
1411 mask |= S_IWGRP;
1412 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1413 }
1414
1415 /* Otherwise, check everyone else. */
1416 if (mode & VEXEC)
1417 mask |= S_IXOTH;
1418 if (mode & VREAD)
1419 mask |= S_IROTH;
1420 if (mode & VWRITE)
1421 mask |= S_IWOTH;
1422 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1423}
1424
1425/*
1426 * ext2_getattr(struct vnode *a_vp, struct vattr *a_vap,
1427 * struct thread *a_td)
1428 */
1429/* ARGSUSED */
1430static
1431int
1432ext2_getattr(struct vop_getattr_args *ap)
1433{
1434 struct vnode *vp = ap->a_vp;
1435 struct inode *ip = VTOI(vp);
1436 struct vattr *vap = ap->a_vap;
1437
1f1db49f
MD
1438 ext2_itimes(vp);
1439 /*
1440 * Copy from inode table
1441 */
1442 vap->va_fsid = dev2udev(ip->i_dev);
1443 vap->va_fileid = ip->i_number;
1444 vap->va_mode = ip->i_mode & ~IFMT;
1445 vap->va_nlink = VFSTOEXT2(vp->v_mount)->um_i_effnlink_valid ?
1446 ip->i_effnlink : ip->i_nlink;
1447 vap->va_uid = ip->i_uid;
1448 vap->va_gid = ip->i_gid;
1449 vap->va_rdev = ip->i_rdev;
1450 vap->va_size = ip->i_din.di_size;
1451 vap->va_atime.tv_sec = ip->i_atime;
1452 vap->va_atime.tv_nsec = ip->i_atimensec;
1453 vap->va_mtime.tv_sec = ip->i_mtime;
1454 vap->va_mtime.tv_nsec = ip->i_mtimensec;
1455 vap->va_ctime.tv_sec = ip->i_ctime;
1456 vap->va_ctime.tv_nsec = ip->i_ctimensec;
1457 vap->va_flags = ip->i_flags;
1458 vap->va_gen = ip->i_gen;
1459 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1460 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
1461 vap->va_type = IFTOVT(ip->i_mode);
1462 vap->va_filerev = ip->i_modrev;
1f1db49f
MD
1463 return (0);
1464}
1465
1466/*
1467 * Set attribute vnode op. called from several syscalls
1468 *
1469 * ext2_setattr(struct vnode *a_vp, struct vattr *a_vap,
1470 * struct ucred *a_cred, struct thread *a_td)
1471 */
1472static
1473int
1474ext2_setattr(struct vop_setattr_args *ap)
1475{
1476 struct vattr *vap = ap->a_vap;
1477 struct vnode *vp = ap->a_vp;
1478 struct inode *ip = VTOI(vp);
1479 struct ucred *cred = ap->a_cred;
1480 int error;
1481
1482 /*
1483 * Check for unsettable attributes.
1484 */
1485 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
1486 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
1487 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
1488 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
1489 return (EINVAL);
1490 }
1491 if (vap->va_flags != VNOVAL) {
1492 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1493 return (EROFS);
1494 if (cred->cr_uid != ip->i_uid &&
1495 (error = suser_cred(cred, PRISON_ROOT)))
1496 return (error);
1497 /*
1498 * Note that a root chflags becomes a user chflags when
1499 * we are jailed, unless the jail.chflags_allowed sysctl
1500 * is set.
1501 */
1502 if (cred->cr_uid == 0 &&
1503 (!jailed(cred) || jail_chflags_allowed)) {
1504 if ((ip->i_flags
1505 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
1506 securelevel > 0)
1507 return (EPERM);
1508 ip->i_flags = vap->va_flags;
1509 } else {
1510 if (ip->i_flags
1511 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
1512 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
1513 return (EPERM);
1514 ip->i_flags &= SF_SETTABLE;
1515 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
1516 }
1517 ip->i_flag |= IN_CHANGE;
1518 if (vap->va_flags & (IMMUTABLE | APPEND))
1519 return (0);
1520 }
1521 if (ip->i_flags & (IMMUTABLE | APPEND))
1522 return (EPERM);
1523 /*
1524 * Go through the fields and update iff not VNOVAL.
1525 */
1526 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
1527 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1528 return (EROFS);
1529 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
1530 return (error);
1531 }
1532 if (vap->va_size != VNOVAL) {
1533 /*
1534 * Disallow write attempts on read-only filesystems;
1535 * unless the file is a socket, fifo, or a block or
1536 * character device resident on the filesystem.
1537 */
1538 switch (vp->v_type) {
1539 case VDIR:
1540 return (EISDIR);
1541 case VLNK:
1542 case VREG:
1543 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1544 return (EROFS);
1545 break;
1546 default:
1547 break;
1548 }
1549 if ((error = EXT2_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
1550 return (error);
1551 }
1552 ip = VTOI(vp);
1553 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1554 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1555 return (EROFS);
1556 if (cred->cr_uid != ip->i_uid &&
1557 (error = suser_cred(cred, PRISON_ROOT)) &&
1558 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1559 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
1560 return (error);
1561 if (vap->va_atime.tv_sec != VNOVAL)
1562 ip->i_flag |= IN_ACCESS;
1563 if (vap->va_mtime.tv_sec != VNOVAL)
1564 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1565 ext2_itimes(vp);
1566 if (vap->va_atime.tv_sec != VNOVAL) {
1567 ip->i_atime = vap->va_atime.tv_sec;
1568 ip->i_atimensec = vap->va_atime.tv_nsec;
1569 }
1570 if (vap->va_mtime.tv_sec != VNOVAL) {
1571 ip->i_mtime = vap->va_mtime.tv_sec;
1572 ip->i_mtimensec = vap->va_mtime.tv_nsec;
1573 }
1574 error = EXT2_UPDATE(vp, 0);
1575 if (error)
1576 return (error);
1577 }
1578 error = 0;
1579 if (vap->va_mode != (mode_t)VNOVAL) {
1580 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1581 return (EROFS);
1582 error = ext2_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
1583 }
1584 VN_KNOTE(vp, NOTE_ATTRIB);
1585 return (error);
1586}
1587
1588/*
1589 * Change the mode on a file.
1590 * Inode must be locked before calling.
1591 */
1592static int
1593ext2_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
1594{
1595 struct inode *ip = VTOI(vp);
1596 int error;
1597
1598 if (cred->cr_uid != ip->i_uid) {
1599 error = suser_cred(cred, PRISON_ROOT);
1600 if (error)
1601 return (error);
1602 }
1603 if (cred->cr_uid) {
1604 if (vp->v_type != VDIR && (mode & S_ISTXT))
1605 return (EFTYPE);
1606 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
1607 return (EPERM);
1608 }
1609 ip->i_mode &= ~ALLPERMS;
1610 ip->i_mode |= (mode & ALLPERMS);
1611 ip->i_flag |= IN_CHANGE;
1612 return (0);
1613}
1614
1615/*
1616 * Perform chown operation on inode ip;
1617 * inode must be locked prior to call.
1618 */
1619static int
1620ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
1621 struct thread *td)
1622{
1623 struct inode *ip = VTOI(vp);
1624 uid_t ouid;
1625 gid_t ogid;
1626 int error = 0;
1627#ifdef QUOTA
1628 int i;
1629 long change;
1630#endif
1631
1632 if (uid == (uid_t)VNOVAL)
1633 uid = ip->i_uid;
1634 if (gid == (gid_t)VNOVAL)
1635 gid = ip->i_gid;
1636 /*
1637 * If we don't own the file, are trying to change the owner
1638 * of the file, or are not a member of the target group,
1639 * the caller must be superuser or the call fails.
1640 */
1641 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
1642 (gid != ip->i_gid && !(cred->cr_gid == gid ||
1643 groupmember((gid_t)gid, cred)))) &&
1644 (error = suser_cred(cred, PRISON_ROOT)))
1645 return (error);
1646 ogid = ip->i_gid;
1647 ouid = ip->i_uid;
1648#ifdef QUOTA
1649 if ((error = ext2_getinoquota(ip)) != 0)
1650 return (error);
1651 if (ouid == uid) {
1652 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1653 ip->i_dquot[USRQUOTA] = NODQUOT;
1654 }
1655 if (ogid == gid) {
1656 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1657 ip->i_dquot[GRPQUOTA] = NODQUOT;
1658 }
1659 change = ip->i_blocks;
1660 (void) ext2_chkdq(ip, -change, cred, CHOWN);
1661 (void) ext2_chkiq(ip, -1, cred, CHOWN);
1662 for (i = 0; i < MAXQUOTAS; i++) {
1663 ext2_dqrele(vp, ip->i_dquot[i]);
1664 ip->i_dquot[i] = NODQUOT;
1665 }
1666#endif
1667 ip->i_gid = gid;
1668 ip->i_uid = uid;
1669#ifdef QUOTA
1670 if ((error = ext2_getinoquota(ip)) == 0) {
1671 if (ouid == uid) {
1672 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1673 ip->i_dquot[USRQUOTA] = NODQUOT;
1674 }
1675 if (ogid == gid) {
1676 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1677 ip->i_dquot[GRPQUOTA] = NODQUOT;
1678 }
1679 if ((error = ext2_chkdq(ip, change, cred, CHOWN)) == 0) {
1680 if ((error = ext2_chkiq(ip, 1, cred, CHOWN)) == 0)
1681 goto good;
1682 else
1683 (void)ext2_chkdq(ip, -change, cred, CHOWN|FORCE);
1684 }
1685 for (i = 0; i < MAXQUOTAS; i++) {
1686 ext2_dqrele(vp, ip->i_dquot[i]);
1687 ip->i_dquot[i] = NODQUOT;
1688 }
1689 }
1690 ip->i_gid = ogid;
1691 ip->i_uid = ouid;
1692 if (ext2_getinoquota(ip) == 0) {
1693 if (ouid == uid) {
1694 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1695 ip->i_dquot[USRQUOTA] = NODQUOT;
1696 }
1697 if (ogid == gid) {
1698 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1699 ip->i_dquot[GRPQUOTA] = NODQUOT;
1700 }
1701 (void) ext2_chkdq(ip, change, cred, FORCE|CHOWN);
1702 (void) ext2_chkiq(ip, 1, cred, FORCE|CHOWN);
1703 (void) ext2_getinoquota(ip);
1704 }
1705 return (error);
1706good:
1707 if (ext2_getinoquota(ip))
1708 panic("ext2_chown: lost quota");
1709#endif /* QUOTA */
1710 ip->i_flag |= IN_CHANGE;
1711 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
1712 ip->i_mode &= ~(ISUID | ISGID);
1713 return (0);
1714}
1715
1716/*
1717 * Mmap a file
1718 *
1719 * NB Currently unsupported.
1720 *
1721 * ext2_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
1722 * struct thread *a_td)
1723 */
1724/* ARGSUSED */
1725static
1726int
1727ext2_mmap(struct vop_mmap_args *ap)
1728{
1729 return (EINVAL);
1730}
1731
1732/*
1733 * whiteout vnode call
1734 *
1735 * ext2_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
1736 */
1737static
1738int
1739ext2_whiteout(struct vop_old_whiteout_args *ap)
1740{
1741 return (EOPNOTSUPP);
1742}
1743
1744/*
1745 * Return target name of a symbolic link
1746 *
1747 * ext2_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1748 */
1749static
1750int
1751ext2_readlink(struct vop_readlink_args *ap)
1752{
1753 struct vnode *vp = ap->a_vp;
1754 struct inode *ip = VTOI(vp);
1755 int isize;
1756
1757 isize = ip->i_size;
1758 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1759 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1760 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1761 return (0);
1762 }
1f1db49f
MD
1763 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1764}
1765
1766/*
1767 * Calculate the logical to physical mapping if not done already,
1768 * then call the device strategy routine.
1769 *
1770 * In order to be able to swap to a file, the VOP_BMAP operation may not
1771 * deadlock on memory. See ext2_bmap() for details.
1772 *
1773 * ext2_strategy(struct vnode *a_vp, struct bio *a_bio)
1774 */
1775static
1776int
1777ext2_strategy(struct vop_strategy_args *ap)
1778{
1779 struct bio *bio = ap->a_bio;
1780 struct bio *nbio;
1781 struct buf *bp = bio->bio_buf;
1782 struct vnode *vp = ap->a_vp;
1783 struct inode *ip;
1784 int error;
1785
1786 ip = VTOI(vp);
1787 if (vp->v_type == VBLK || vp->v_type == VCHR)
1788 panic("ext2_strategy: spec");
1789 nbio = push_bio(bio);
1790 if (nbio->bio_offset == NOOFFSET) {
1791 error = VOP_BMAP(vp, bio->bio_offset, NULL, &nbio->bio_offset,
1792 NULL, NULL);
1793 if (error) {
1794 bp->b_error = error;
1795 bp->b_flags |= B_ERROR;
1796 /* I/O was never started on nbio, must biodone(bio) */
1797 biodone(bio);
1798 return (error);
1799 }
1800 if (nbio->bio_offset == NOOFFSET)
1801 vfs_bio_clrbuf(bp);
1802 }
1803 if (nbio->bio_offset == NOOFFSET) {
1804 /* I/O was never started on nbio, must biodone(bio) */
1805 biodone(bio);
1806 return (0);
1807 }
1808 vn_strategy(ip->i_devvp, nbio);
1809 return (0);
1810}
1811
1812/*
1813 * Print out the contents of an inode.
1814 *
1815 * ext2_print(struct vnode *a_vp)
1816 */
1817static
1818int
1819ext2_print(struct vop_print_args *ap)
1820{
1821 struct vnode *vp = ap->a_vp;
1822 struct inode *ip = VTOI(vp);
1823
1824 printf("tag VT_EXT2FS, ino %lu, on dev %s (%d, %d)",
1825 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1826 minor(ip->i_dev));
1827 if (vp->v_type == VFIFO)
1828 fifo_printinfo(vp);
1829 lockmgr_printinfo(&vp->v_lock);
1830 printf("\n");
1831 return (0);
1832}
1833
1834/*
1835 * Read wrapper for special devices.
1836 *
1837 * ext2spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1838 * struct ucred *a_cred)
1839 */
1840static
1841int
1842ext2spec_read(struct vop_read_args *ap)
1843{
1844 int error, resid;
1845 struct inode *ip;
1846 struct uio *uio;
1847
1848 uio = ap->a_uio;
1849 resid = uio->uio_resid;
1850 error = VOCALL(spec_vnode_vops, &ap->a_head);
1851 /*
1852 * The inode may have been revoked during the call, so it must not
1853 * be accessed blindly here or in the other wrapper functions.
1854 */
1855 ip = VTOI(ap->a_vp);
1856 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1857 ip->i_flag |= IN_ACCESS;
1858 return (error);
1859}
1860
1861/*
1862 * Write wrapper for special devices.
1863 *
1864 * ext2spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1865 * struct ucred *a_cred)
1866 */
1867static
1868int
1869ext2spec_write(struct vop_write_args *ap)
1870{
1871 int error, resid;
1872 struct inode *ip;
1873 struct uio *uio;
1874
1875 uio = ap->a_uio;
1876 resid = uio->uio_resid;
1877 error = VOCALL(spec_vnode_vops, &ap->a_head);
1878 ip = VTOI(ap->a_vp);
1879 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1880 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1881 return (error);
1882}
1883
1884/*
1885 * Close wrapper for special devices.
1886 *
1887 * Update the times on the inode then do device close.
1888 *
1889 * ext2spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1890 * struct thread *a_td)
1891 */
1892static
1893int
1894ext2spec_close(struct vop_close_args *ap)
1895{
1896 struct vnode *vp = ap->a_vp;
1897
1898 if (vp->v_usecount > 1)
1899 ext2_itimes(vp);
1900 return (VOCALL(spec_vnode_vops, &ap->a_head));
1901}
1902
1903/*
1904 * Read wrapper for fifos.
1905 *
1906 * ext2fifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1907 * struct ucred *a_cred)
1908 */
1909static
1910int
1911ext2fifo_read(struct vop_read_args *ap)
1912{
1913 int error, resid;
1914 struct inode *ip;
1915 struct uio *uio;
1916
1917 uio = ap->a_uio;
1918 resid = uio->uio_resid;
1919 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1920 ip = VTOI(ap->a_vp);
1921 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1922 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1923 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1924 return (error);
1925}
1926
1927/*
1928 * Write wrapper for fifos.
1929 *
1930 * ext2fifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1931 * struct ucred *a_cred)
1932 */
1933static
1934int
1935ext2fifo_write(struct vop_write_args *ap)
1936{
1937 int error, resid;
1938 struct inode *ip;
1939 struct uio *uio;
1940
1941 uio = ap->a_uio;
1942 resid = uio->uio_resid;
1943 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1944 ip = VTOI(ap->a_vp);
1945 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1946 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1947 return (error);
1948}
1949
1950/*
1951 * Close wrapper for fifos.
1952 *
1953 * Update the times on the inode then do device close.
1954 *
1955 * ext2fifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1956 * struct thread *a_td)
1957 */
1958static
1959int
1960ext2fifo_close(struct vop_close_args *ap)
1961{
1962 struct vnode *vp = ap->a_vp;
1963
1964 if (vp->v_usecount > 1)
1965 ext2_itimes(vp);
1966 return (VOCALL(fifo_vnode_vops, &ap->a_head));
1967}
1968
1969/*
1970 * Kqfilter wrapper for fifos.
1971 *
1972 * Fall through to ext2 kqfilter routines if needed
1973 */
1974static
1975int
1976ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1977{
1978 int error;
1979
1980 error = VOCALL(fifo_vnode_vops, &ap->a_head);
1981 if (error)
1982 error = ext2_kqfilter(ap);
1983 return (error);
1984}
1985
1986/*
1987 * Return POSIX pathconf information applicable to ext2 filesystems.
1988 *
1989 * ext2_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
1990 */
1991static
1992int
1993ext2_pathconf(struct vop_pathconf_args *ap)
1994{
1995 switch (ap->a_name) {
1996 case _PC_LINK_MAX:
1997 *ap->a_retval = LINK_MAX;
1998 return (0);
1999 case _PC_NAME_MAX:
2000 *ap->a_retval = NAME_MAX;
2001 return (0);
2002 case _PC_PATH_MAX:
2003 *ap->a_retval = PATH_MAX;
2004 return (0);
2005 case _PC_PIPE_BUF:
2006 *ap->a_retval = PIPE_BUF;
2007 return (0);
2008 case _PC_CHOWN_RESTRICTED:
2009 *ap->a_retval = 1;
2010 return (0);
2011 case _PC_NO_TRUNC:
2012 *ap->a_retval = 1;
2013 return (0);
2014 default:
2015 return (EINVAL);
2016 }
2017 /* NOTREACHED */
2018}
2019
2020/*
2021 * Advisory record locking support
2022 *
2023 * ext2_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
2024 * int a_flags)
2025 */
2026static
2027int
2028ext2_advlock(struct vop_advlock_args *ap)
2029{
2030 struct inode *ip = VTOI(ap->a_vp);
2031
2032 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
2033}
2034
2035/*
2036 * Initialize the vnode associated with a new inode, handle aliased
2037 * vnodes.
2038 */
2039int
2040ext2_vinit(struct mount *mntp, struct vnode **vpp)
2041{
2042 struct inode *ip;
2043 struct vnode *vp;
2044 struct timeval tv;
2045
2046 vp = *vpp;
2047 ip = VTOI(vp);
2048
2049 switch(vp->v_type = IFTOVT(ip->i_mode)) {
2050 case VCHR:
2051 case VBLK:
2052 vp->v_ops = &mntp->mnt_vn_spec_ops;
2053 addaliasu(vp, ip->i_rdev);
2054 break;
2055 case VFIFO:
2056 vp->v_ops = &mntp->mnt_vn_fifo_ops;
2057 break;
93f50d60 2058 case VDIR:
1c843a13
MD
2059 case VREG:
2060 vinitvmio(vp, ip->i_size);
2061 break;
2062 case VLNK:
2063 if ((ip->i_size >= vp->v_mount->mnt_maxsymlinklen) &&
2064 ip->i_din.di_blocks != 0
2065 ) {
2066 vinitvmio(vp, ip->i_size);
2067 }
93f50d60 2068 break;
1f1db49f
MD
2069 default:
2070 break;
2071
2072 }
2073
2074 if (ip->i_number == ROOTINO)
2075 vp->v_flag |= VROOT;
2076 /*
2077 * Initialize modrev times
2078 */
2079 getmicrouptime(&tv);
2080 SETHIGH(ip->i_modrev, tv.tv_sec);
2081 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
2082 *vpp = vp;
2083 return (0);
2084}
2085
2086static struct filterops ext2read_filtops =
2087 { 1, NULL, filt_ext2detach, filt_ext2read };
2088static struct filterops ext2write_filtops =
2089 { 1, NULL, filt_ext2detach, filt_ext2write };
2090static struct filterops ext2vnode_filtops =
2091 { 1, NULL, filt_ext2detach, filt_ext2vnode };
2092
2093/*
2094 * ext2_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2095 */
2096static int
2097ext2_kqfilter(struct vop_kqfilter_args *ap)
2098{
2099 struct vnode *vp = ap->a_vp;
2100 struct knote *kn = ap->a_kn;
2101 lwkt_tokref ilock;
2102
2103 switch (kn->kn_filter) {
2104 case EVFILT_READ:
2105 kn->kn_fop = &ext2read_filtops;
2106 break;
2107 case EVFILT_WRITE:
2108 kn->kn_fop = &ext2write_filtops;
2109 break;
2110 case EVFILT_VNODE:
2111 kn->kn_fop = &ext2vnode_filtops;
2112 break;
2113 default:
2114 return (1);
2115 }
2116
2117 kn->kn_hook = (caddr_t)vp;
2118
2119 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2120 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
2121 lwkt_reltoken(&ilock);
2122
2123 return (0);
2124}
2125
2126static void
2127filt_ext2detach(struct knote *kn)
2128{
2129 struct vnode *vp = (struct vnode *)kn->kn_hook;
2130 lwkt_tokref ilock;
2131
2132 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2133 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2134 kn, knote, kn_selnext);
2135 lwkt_reltoken(&ilock);
2136}
2137
2138/*ARGSUSED*/
2139static int
2140filt_ext2read(struct knote *kn, long hint)
2141{
2142 struct vnode *vp = (struct vnode *)kn->kn_hook;
2143 struct inode *ip = VTOI(vp);
2144
2145 /*
2146 * filesystem is gone, so set the EOF flag and schedule
2147 * the knote for deletion.
2148 */
2149 if (hint == NOTE_REVOKE) {
2150 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2151 return (1);
2152 }
2153
2154 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2155 return (kn->kn_data != 0);
2156}
2157
2158/*ARGSUSED*/
2159static int
2160filt_ext2write(struct knote *kn, long hint)
2161{
2162 /*
2163 * filesystem is gone, so set the EOF flag and schedule
2164 * the knote for deletion.
2165 */
2166 if (hint == NOTE_REVOKE)
2167 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2168
2169 kn->kn_data = 0;
2170 return (1);
2171}
2172
2173static int
2174filt_ext2vnode(struct knote *kn, long hint)
2175{
2176 if (kn->kn_sfflags & hint)
2177 kn->kn_fflags |= hint;
2178 if (hint == NOTE_REVOKE) {
2179 kn->kn_flags |= EV_EOF;
2180 return (1);
2181 }
2182 return (kn->kn_fflags != 0);
2183}
2184
2185static struct vop_ops *ext2_vnode_vops;
2186/* Global vfs data structures for ext2. */
2187struct vnodeopv_entry_desc ext2_vnodeop_entries[] = {
2188 { &vop_default_desc, vop_defaultop },
2189 { &vop_fsync_desc, (vnodeopv_entry_t) ext2_fsync },
2190 { &vop_read_desc, (vnodeopv_entry_t) ext2_read },
2191 { &vop_reallocblks_desc, (vnodeopv_entry_t) ext2_reallocblks },
2192 { &vop_write_desc, (vnodeopv_entry_t) ext2_write },
2193 { &vop_access_desc, (vnodeopv_entry_t) ext2_access },
2194 { &vop_advlock_desc, (vnodeopv_entry_t) ext2_advlock },
2195 { &vop_bmap_desc, (vnodeopv_entry_t) ext2_bmap },
2196 { &vop_old_lookup_desc, (vnodeopv_entry_t) ext2_lookup },
2197 { &vop_close_desc, (vnodeopv_entry_t) ext2_close },
2198 { &vop_old_create_desc, (vnodeopv_entry_t) ext2_create },
2199 { &vop_getattr_desc, (vnodeopv_entry_t) ext2_getattr },
2200 { &vop_inactive_desc, (vnodeopv_entry_t) ext2_inactive },
2201 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2202 { &vop_old_link_desc, (vnodeopv_entry_t) ext2_link },
2203 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2204 { &vop_old_mkdir_desc, (vnodeopv_entry_t) ext2_mkdir },
2205 { &vop_old_mknod_desc, (vnodeopv_entry_t) ext2_mknod },
2206 { &vop_mmap_desc, (vnodeopv_entry_t) ext2_mmap },
2207 { &vop_open_desc, (vnodeopv_entry_t) ext2_open },
2208 { &vop_pathconf_desc, (vnodeopv_entry_t) ext2_pathconf },
2209 { &vop_poll_desc, (vnodeopv_entry_t) vop_stdpoll },
2210 { &vop_kqfilter_desc, (vnodeopv_entry_t) ext2_kqfilter },
2211 { &vop_print_desc, (vnodeopv_entry_t) ext2_print },
2212 { &vop_readdir_desc, (vnodeopv_entry_t) ext2_readdir },
2213 { &vop_readlink_desc, (vnodeopv_entry_t) ext2_readlink },
2214 { &vop_reclaim_desc, (vnodeopv_entry_t) ext2_reclaim },
2215 { &vop_old_remove_desc, (vnodeopv_entry_t) ext2_remove },
2216 { &vop_old_rename_desc, (vnodeopv_entry_t) ext2_rename },
2217 { &vop_old_rmdir_desc, (vnodeopv_entry_t) ext2_rmdir },
2218 { &vop_setattr_desc, (vnodeopv_entry_t) ext2_setattr },
2219 { &vop_strategy_desc, (vnodeopv_entry_t) ext2_strategy },
2220 { &vop_old_symlink_desc, (vnodeopv_entry_t) ext2_symlink },
2221 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2222 { &vop_old_whiteout_desc, (vnodeopv_entry_t) ext2_whiteout },
2223 { &vop_getpages_desc, (vnodeopv_entry_t) ext2_getpages },
2224 { &vop_putpages_desc, (vnodeopv_entry_t) ext2_putpages },
2225 { NULL, NULL }
2226};
2227static struct vnodeopv_desc ext2_vnodeop_opv_desc =
2228 { &ext2_vnode_vops, ext2_vnodeop_entries, 0 };
2229
2230static struct vop_ops *ext2_spec_vops;
2231struct vnodeopv_entry_desc ext2_specop_entries[] = {
2232 { &vop_default_desc, (vnodeopv_entry_t) ext2_vnoperatespec },
2233 { &vop_fsync_desc, (vnodeopv_entry_t) ext2_fsync },
2234 { &vop_access_desc, (vnodeopv_entry_t) ext2_access },
2235 { &vop_close_desc, (vnodeopv_entry_t) ext2spec_close },
2236 { &vop_getattr_desc, (vnodeopv_entry_t) ext2_getattr },
2237 { &vop_inactive_desc, (vnodeopv_entry_t) ext2_inactive },
2238 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2239 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2240 { &vop_print_desc, (vnodeopv_entry_t) ext2_print },
2241 { &vop_read_desc, (vnodeopv_entry_t) ext2spec_read },
2242 { &vop_reclaim_desc, (vnodeopv_entry_t) ext2_reclaim },
2243 { &vop_setattr_desc, (vnodeopv_entry_t) ext2_setattr },
2244 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2245 { &vop_write_desc, (vnodeopv_entry_t) ext2spec_write },
2246 { NULL, NULL }
2247};
2248static struct vnodeopv_desc ext2_specop_opv_desc =
2249 { &ext2_spec_vops, ext2_specop_entries, 0 };
2250
2251static struct vop_ops *ext2_fifo_vops;
2252struct vnodeopv_entry_desc ext2_fifoop_entries[] = {
2253 { &vop_default_desc, (vnodeopv_entry_t) ext2_vnoperatefifo },
2254 { &vop_fsync_desc, (vnodeopv_entry_t) ext2_fsync },
2255 { &vop_access_desc, (vnodeopv_entry_t) ext2_access },
2256 { &vop_close_desc, (vnodeopv_entry_t) ext2fifo_close },
2257 { &vop_getattr_desc, (vnodeopv_entry_t) ext2_getattr },
2258 { &vop_inactive_desc, (vnodeopv_entry_t) ext2_inactive },
2259 { &vop_islocked_desc, (vnodeopv_entry_t) vop_stdislocked },
2260 { &vop_kqfilter_desc, (vnodeopv_entry_t) ext2fifo_kqfilter },
2261 { &vop_lock_desc, (vnodeopv_entry_t) vop_stdlock },
2262 { &vop_print_desc, (vnodeopv_entry_t) ext2_print },
2263 { &vop_read_desc, (vnodeopv_entry_t) ext2fifo_read },
2264 { &vop_reclaim_desc, (vnodeopv_entry_t) ext2_reclaim },
2265 { &vop_setattr_desc, (vnodeopv_entry_t) ext2_setattr },
2266 { &vop_unlock_desc, (vnodeopv_entry_t) vop_stdunlock },
2267 { &vop_write_desc, (vnodeopv_entry_t) ext2fifo_write },
2268 { NULL, NULL }
2269};
2270static struct vnodeopv_desc ext2_fifoop_opv_desc =
2271 { &ext2_fifo_vops, ext2_fifoop_entries, 0 };
2272
2273VNODEOP_SET(ext2_vnodeop_opv_desc);
2274VNODEOP_SET(ext2_specop_opv_desc);
2275VNODEOP_SET(ext2_fifoop_opv_desc);
2276
2277/*
2278 * ext2_vnoperate(struct vnodeop_desc *a_desc)
2279 */
2280int
2281ext2_vnoperate(struct vop_generic_args *ap)
2282{
2283 return (VOCALL(ext2_vnode_vops, ap));
2284}
2285
2286/*
2287 * ext2_vnoperatefifo(struct vnodeop_desc *a_desc)
2288 */
2289int
2290ext2_vnoperatefifo(struct vop_generic_args *ap)
2291{
2292 return (VOCALL(ext2_fifo_vops, ap));
2293}
2294
2295/*
2296 * ext2_vnoperatespec(struct vnodeop_desc *a_desc)
2297 */
2298int
2299ext2_vnoperatespec(struct vop_generic_args *ap)
2300{
2301 return (VOCALL(ext2_spec_vops, ap));
2302}
2303