Style(9):
[dragonfly.git] / sys / vfs / ufs / ufs_vnops.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1982, 1986, 1989, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
39 * $FreeBSD: src/sys/ufs/ufs/ufs_vnops.c,v 1.131.2.8 2003/01/02 17:26:19 bde Exp $
5fd012e0 40 * $DragonFly: src/sys/vfs/ufs/ufs_vnops.c,v 1.23 2004/10/12 19:21:12 dillon Exp $
984263bc
MD
41 */
42
43#include "opt_quota.h"
44#include "opt_suiddir.h"
45#include "opt_ufs.h"
46
47#include <sys/param.h>
48#include <sys/systm.h>
984263bc
MD
49#include <sys/kernel.h>
50#include <sys/fcntl.h>
51#include <sys/stat.h>
52#include <sys/buf.h>
53#include <sys/proc.h>
dadab5e9 54#include <sys/namei.h>
984263bc
MD
55#include <sys/mount.h>
56#include <sys/unistd.h>
57#include <sys/vnode.h>
58#include <sys/malloc.h>
59#include <sys/dirent.h>
60#include <sys/lockf.h>
61#include <sys/event.h>
62#include <sys/conf.h>
63
64#include <sys/file.h> /* XXX */
65
66#include <vm/vm.h>
67#include <vm/vm_extern.h>
68
1f2de5d4 69#include <vfs/fifofs/fifo.h>
984263bc 70
1f2de5d4
MD
71#include "quota.h"
72#include "inode.h"
73#include "dir.h"
74#include "ufsmount.h"
75#include "ufs_extern.h"
984263bc 76#ifdef UFS_DIRHASH
1f2de5d4 77#include "dirhash.h"
984263bc
MD
78#endif
79
a6ee311a
RG
80static int ufs_access (struct vop_access_args *);
81static int ufs_advlock (struct vop_advlock_args *);
82static int ufs_chmod (struct vnode *, int, struct ucred *, struct thread *);
83static int ufs_chown (struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
84static int ufs_close (struct vop_close_args *);
85static int ufs_create (struct vop_create_args *);
86static int ufs_getattr (struct vop_getattr_args *);
87static int ufs_link (struct vop_link_args *);
88static int ufs_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
89static int ufs_missingop (struct vop_generic_args *ap);
90static int ufs_mkdir (struct vop_mkdir_args *);
91static int ufs_mknod (struct vop_mknod_args *);
92static int ufs_mmap (struct vop_mmap_args *);
93static int ufs_open (struct vop_open_args *);
94static int ufs_pathconf (struct vop_pathconf_args *);
95static int ufs_print (struct vop_print_args *);
96static int ufs_readdir (struct vop_readdir_args *);
97static int ufs_readlink (struct vop_readlink_args *);
98static int ufs_remove (struct vop_remove_args *);
99static int ufs_rename (struct vop_rename_args *);
100static int ufs_rmdir (struct vop_rmdir_args *);
101static int ufs_setattr (struct vop_setattr_args *);
102static int ufs_strategy (struct vop_strategy_args *);
103static int ufs_symlink (struct vop_symlink_args *);
104static int ufs_whiteout (struct vop_whiteout_args *);
105static int ufsfifo_close (struct vop_close_args *);
106static int ufsfifo_kqfilter (struct vop_kqfilter_args *);
107static int ufsfifo_read (struct vop_read_args *);
108static int ufsfifo_write (struct vop_write_args *);
109static int ufsspec_close (struct vop_close_args *);
110static int ufsspec_read (struct vop_read_args *);
111static int ufsspec_write (struct vop_write_args *);
112static int filt_ufsread (struct knote *kn, long hint);
113static int filt_ufswrite (struct knote *kn, long hint);
114static int filt_ufsvnode (struct knote *kn, long hint);
115static void filt_ufsdetach (struct knote *kn);
116static int ufs_kqfilter (struct vop_kqfilter_args *ap);
984263bc
MD
117
118union _qcvt {
119 int64_t qcvt;
120 int32_t val[2];
121};
122#define SETHIGH(q, h) { \
123 union _qcvt tmp; \
124 tmp.qcvt = (q); \
125 tmp.val[_QUAD_HIGHWORD] = (h); \
126 (q) = tmp.qcvt; \
127}
128#define SETLOW(q, l) { \
129 union _qcvt tmp; \
130 tmp.qcvt = (q); \
131 tmp.val[_QUAD_LOWWORD] = (l); \
132 (q) = tmp.qcvt; \
133}
134#define VN_KNOTE(vp, b) \
135 KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
136
137/*
138 * A virgin directory (no blushing please).
139 */
140static struct dirtemplate mastertemplate = {
141 0, 12, DT_DIR, 1, ".",
142 0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
143};
144static struct odirtemplate omastertemplate = {
145 0, 12, 1, ".",
146 0, DIRBLKSIZ - 12, 2, ".."
147};
148
149void
0973c589 150ufs_itimes(struct vnode *vp)
984263bc
MD
151{
152 struct inode *ip;
153 struct timespec ts;
154
155 ip = VTOI(vp);
156 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
157 return;
158 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
159 ip->i_flag |= IN_LAZYMOD;
160 else
161 ip->i_flag |= IN_MODIFIED;
162 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
163 vfs_timestamp(&ts);
164 if (ip->i_flag & IN_ACCESS) {
165 ip->i_atime = ts.tv_sec;
166 ip->i_atimensec = ts.tv_nsec;
167 }
168 if (ip->i_flag & IN_UPDATE) {
169 ip->i_mtime = ts.tv_sec;
170 ip->i_mtimensec = ts.tv_nsec;
171 ip->i_modrev++;
172 }
173 if (ip->i_flag & IN_CHANGE) {
174 ip->i_ctime = ts.tv_sec;
175 ip->i_ctimensec = ts.tv_nsec;
176 }
177 }
178 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
179}
180
181/*
182 * Create a regular file
0973c589
CP
183 *
184 * ufs_create(struct vnode *a_dvp, struct vnode **a_vpp,
185 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc 186 */
0961aa92 187static
984263bc 188int
0973c589 189ufs_create(struct vop_create_args *ap)
984263bc
MD
190{
191 int error;
192
193 error =
194 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
195 ap->a_dvp, ap->a_vpp, ap->a_cnp);
196 if (error)
197 return (error);
198 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
199 return (0);
200}
201
202/*
203 * Mknod vnode call
0973c589
CP
204 *
205 * ufs_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
206 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
207 */
208/* ARGSUSED */
0961aa92 209static
984263bc 210int
0973c589 211ufs_mknod(struct vop_mknod_args *ap)
984263bc
MD
212{
213 struct vattr *vap = ap->a_vap;
214 struct vnode **vpp = ap->a_vpp;
215 struct inode *ip;
216 ino_t ino;
217 int error;
218
219 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
220 ap->a_dvp, vpp, ap->a_cnp);
221 if (error)
222 return (error);
223 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
224 ip = VTOI(*vpp);
225 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
226 if (vap->va_rdev != VNOVAL) {
227 /*
228 * Want to be able to use this to make badblock
229 * inodes, so don't truncate the dev number.
230 */
231 ip->i_rdev = vap->va_rdev;
232 }
233 /*
234 * Remove inode, then reload it through VFS_VGET so it is
235 * checked to see if it is an alias of an existing entry in
236 * the inode cache.
237 */
984263bc
MD
238 (*vpp)->v_type = VNON;
239 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
240 vgone(*vpp);
5fd012e0 241 vput(*vpp);
984263bc
MD
242 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
243 if (error) {
244 *vpp = NULL;
245 return (error);
246 }
247 return (0);
248}
249
250/*
251 * Open called.
252 *
253 * Nothing to do.
0973c589
CP
254 *
255 * ufs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
256 * struct thread *a_td)
984263bc
MD
257 */
258/* ARGSUSED */
0961aa92 259static
984263bc 260int
0973c589 261ufs_open(struct vop_open_args *ap)
984263bc 262{
984263bc
MD
263 /*
264 * Files marked append-only must be opened for appending.
265 */
266 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
267 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
268 return (EPERM);
269 return (0);
270}
271
272/*
273 * Close called.
274 *
275 * Update the times on the inode.
0973c589
CP
276 *
277 * ufs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
278 * struct thread *a_td)
984263bc
MD
279 */
280/* ARGSUSED */
0961aa92 281static
984263bc 282int
0973c589 283ufs_close(struct vop_close_args *ap)
984263bc 284{
3ff2135f 285 struct vnode *vp = ap->a_vp;
984263bc 286
984263bc
MD
287 if (vp->v_usecount > 1)
288 ufs_itimes(vp);
984263bc
MD
289 return (0);
290}
291
0973c589
CP
292/*
293 * ufs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
294 * struct thread *a_td)
295 */
0961aa92 296static
984263bc 297int
0973c589 298ufs_access(struct vop_access_args *ap)
984263bc
MD
299{
300 struct vnode *vp = ap->a_vp;
301 struct inode *ip = VTOI(vp);
302 struct ucred *cred = ap->a_cred;
303 mode_t mask, mode = ap->a_mode;
3ff2135f 304 gid_t *gp;
984263bc
MD
305 int i;
306#ifdef QUOTA
307 int error;
308#endif
309
310 /*
f719c866 311 * Disallow write attempts on read-only filesystems;
984263bc 312 * unless the file is a socket, fifo, or a block or
f719c866 313 * character device resident on the filesystem.
984263bc
MD
314 */
315 if (mode & VWRITE) {
316 switch (vp->v_type) {
317 case VDIR:
318 case VLNK:
319 case VREG:
320 if (vp->v_mount->mnt_flag & MNT_RDONLY)
321 return (EROFS);
322#ifdef QUOTA
323 if ((error = getinoquota(ip)) != 0)
324 return (error);
325#endif
326 break;
327 default:
328 break;
329 }
330 }
331
332 /* If immutable bit set, nobody gets to write it. */
333 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
334 return (EPERM);
335
336 /* Otherwise, user id 0 always gets access. */
337 if (cred->cr_uid == 0)
338 return (0);
339
340 mask = 0;
341
342 /* Otherwise, check the owner. */
343 if (cred->cr_uid == ip->i_uid) {
344 if (mode & VEXEC)
345 mask |= S_IXUSR;
346 if (mode & VREAD)
347 mask |= S_IRUSR;
348 if (mode & VWRITE)
349 mask |= S_IWUSR;
350 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
351 }
352
353 /* Otherwise, check the groups. */
354 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
355 if (ip->i_gid == *gp) {
356 if (mode & VEXEC)
357 mask |= S_IXGRP;
358 if (mode & VREAD)
359 mask |= S_IRGRP;
360 if (mode & VWRITE)
361 mask |= S_IWGRP;
362 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
363 }
364
365 /* Otherwise, check everyone else. */
366 if (mode & VEXEC)
367 mask |= S_IXOTH;
368 if (mode & VREAD)
369 mask |= S_IROTH;
370 if (mode & VWRITE)
371 mask |= S_IWOTH;
372 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
373}
374
0973c589
CP
375/*
376 * ufs_getattr(struct vnode *a_vp, struct vattr *a_vap,
377 * struct thread *a_td)
378 */
984263bc 379/* ARGSUSED */
0961aa92 380static
984263bc 381int
0973c589 382ufs_getattr(struct vop_getattr_args *ap)
984263bc 383{
3ff2135f
RG
384 struct vnode *vp = ap->a_vp;
385 struct inode *ip = VTOI(vp);
386 struct vattr *vap = ap->a_vap;
984263bc
MD
387
388 ufs_itimes(vp);
389 /*
390 * Copy from inode table
391 */
392 vap->va_fsid = dev2udev(ip->i_dev);
393 vap->va_fileid = ip->i_number;
394 vap->va_mode = ip->i_mode & ~IFMT;
395 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ?
396 ip->i_effnlink : ip->i_nlink;
397 vap->va_uid = ip->i_uid;
398 vap->va_gid = ip->i_gid;
399 vap->va_rdev = ip->i_rdev;
400 vap->va_size = ip->i_din.di_size;
401 vap->va_atime.tv_sec = ip->i_atime;
402 vap->va_atime.tv_nsec = ip->i_atimensec;
403 vap->va_mtime.tv_sec = ip->i_mtime;
404 vap->va_mtime.tv_nsec = ip->i_mtimensec;
405 vap->va_ctime.tv_sec = ip->i_ctime;
406 vap->va_ctime.tv_nsec = ip->i_ctimensec;
407 vap->va_flags = ip->i_flags;
408 vap->va_gen = ip->i_gen;
409 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
410 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
411 vap->va_type = IFTOVT(ip->i_mode);
412 vap->va_filerev = ip->i_modrev;
413 return (0);
414}
415
416/*
417 * Set attribute vnode op. called from several syscalls
0973c589
CP
418 *
419 * ufs_setattr(struct vnode *a_vp, struct vattr *a_vap,
420 * struct ucred *a_cred, struct thread *a_td)
984263bc 421 */
0961aa92 422static
984263bc 423int
0973c589 424ufs_setattr(struct vop_setattr_args *ap)
984263bc
MD
425{
426 struct vattr *vap = ap->a_vap;
427 struct vnode *vp = ap->a_vp;
428 struct inode *ip = VTOI(vp);
429 struct ucred *cred = ap->a_cred;
984263bc
MD
430 int error;
431
432 /*
433 * Check for unsettable attributes.
434 */
435 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
436 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
437 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
438 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
439 return (EINVAL);
440 }
441 if (vap->va_flags != VNOVAL) {
442 if (vp->v_mount->mnt_flag & MNT_RDONLY)
443 return (EROFS);
444 if (cred->cr_uid != ip->i_uid &&
dadab5e9 445 (error = suser_cred(cred, PRISON_ROOT)))
984263bc 446 return (error);
41c20dac 447 if ((cred->cr_uid == 0) && (cred->cr_prison == NULL)) {
984263bc
MD
448 if ((ip->i_flags
449 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
450 securelevel > 0)
451 return (EPERM);
452 ip->i_flags = vap->va_flags;
453 } else {
454 if (ip->i_flags
455 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
456 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
457 return (EPERM);
458 ip->i_flags &= SF_SETTABLE;
459 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
460 }
461 ip->i_flag |= IN_CHANGE;
462 if (vap->va_flags & (IMMUTABLE | APPEND))
463 return (0);
464 }
465 if (ip->i_flags & (IMMUTABLE | APPEND))
466 return (EPERM);
467 /*
468 * Go through the fields and update iff not VNOVAL.
469 */
470 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
471 if (vp->v_mount->mnt_flag & MNT_RDONLY)
472 return (EROFS);
dadab5e9 473 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
984263bc
MD
474 return (error);
475 }
476 if (vap->va_size != VNOVAL) {
477 /*
f719c866 478 * Disallow write attempts on read-only filesystems;
984263bc 479 * unless the file is a socket, fifo, or a block or
f719c866 480 * character device resident on the filesystem.
984263bc
MD
481 */
482 switch (vp->v_type) {
483 case VDIR:
484 return (EISDIR);
485 case VLNK:
486 case VREG:
487 if (vp->v_mount->mnt_flag & MNT_RDONLY)
488 return (EROFS);
489 break;
490 default:
491 break;
492 }
dadab5e9 493 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
984263bc
MD
494 return (error);
495 }
496 ip = VTOI(vp);
497 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
498 if (vp->v_mount->mnt_flag & MNT_RDONLY)
499 return (EROFS);
500 if (cred->cr_uid != ip->i_uid &&
dadab5e9 501 (error = suser_cred(cred, PRISON_ROOT)) &&
984263bc 502 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
dadab5e9 503 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
984263bc
MD
504 return (error);
505 if (vap->va_atime.tv_sec != VNOVAL)
506 ip->i_flag |= IN_ACCESS;
507 if (vap->va_mtime.tv_sec != VNOVAL)
508 ip->i_flag |= IN_CHANGE | IN_UPDATE;
509 ufs_itimes(vp);
510 if (vap->va_atime.tv_sec != VNOVAL) {
511 ip->i_atime = vap->va_atime.tv_sec;
512 ip->i_atimensec = vap->va_atime.tv_nsec;
513 }
514 if (vap->va_mtime.tv_sec != VNOVAL) {
515 ip->i_mtime = vap->va_mtime.tv_sec;
516 ip->i_mtimensec = vap->va_mtime.tv_nsec;
517 }
518 error = UFS_UPDATE(vp, 0);
519 if (error)
520 return (error);
521 }
522 error = 0;
523 if (vap->va_mode != (mode_t)VNOVAL) {
524 if (vp->v_mount->mnt_flag & MNT_RDONLY)
525 return (EROFS);
dadab5e9 526 error = ufs_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
984263bc
MD
527 }
528 VN_KNOTE(vp, NOTE_ATTRIB);
529 return (error);
530}
531
532/*
533 * Change the mode on a file.
534 * Inode must be locked before calling.
535 */
536static int
dadab5e9 537ufs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
984263bc 538{
3ff2135f 539 struct inode *ip = VTOI(vp);
984263bc
MD
540 int error;
541
542 if (cred->cr_uid != ip->i_uid) {
dadab5e9 543 error = suser_cred(cred, PRISON_ROOT);
984263bc
MD
544 if (error)
545 return (error);
546 }
547 if (cred->cr_uid) {
548 if (vp->v_type != VDIR && (mode & S_ISTXT))
549 return (EFTYPE);
550 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
551 return (EPERM);
552 }
553 ip->i_mode &= ~ALLPERMS;
554 ip->i_mode |= (mode & ALLPERMS);
555 ip->i_flag |= IN_CHANGE;
556 return (0);
557}
558
559/*
560 * Perform chown operation on inode ip;
561 * inode must be locked prior to call.
562 */
563static int
0973c589
CP
564ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
565 struct thread *td)
984263bc 566{
3ff2135f 567 struct inode *ip = VTOI(vp);
984263bc
MD
568 uid_t ouid;
569 gid_t ogid;
570 int error = 0;
571#ifdef QUOTA
3ff2135f 572 int i;
984263bc
MD
573 long change;
574#endif
575
576 if (uid == (uid_t)VNOVAL)
577 uid = ip->i_uid;
578 if (gid == (gid_t)VNOVAL)
579 gid = ip->i_gid;
580 /*
581 * If we don't own the file, are trying to change the owner
582 * of the file, or are not a member of the target group,
583 * the caller must be superuser or the call fails.
584 */
585 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
a10e3626
DR
586 (gid != ip->i_gid && !(cred->cr_gid == gid ||
587 groupmember((gid_t)gid, cred)))) &&
dadab5e9 588 (error = suser_cred(cred, PRISON_ROOT)))
984263bc
MD
589 return (error);
590 ogid = ip->i_gid;
591 ouid = ip->i_uid;
592#ifdef QUOTA
593 if ((error = getinoquota(ip)) != 0)
594 return (error);
595 if (ouid == uid) {
596 dqrele(vp, ip->i_dquot[USRQUOTA]);
597 ip->i_dquot[USRQUOTA] = NODQUOT;
598 }
599 if (ogid == gid) {
600 dqrele(vp, ip->i_dquot[GRPQUOTA]);
601 ip->i_dquot[GRPQUOTA] = NODQUOT;
602 }
603 change = ip->i_blocks;
604 (void) chkdq(ip, -change, cred, CHOWN);
605 (void) chkiq(ip, -1, cred, CHOWN);
606 for (i = 0; i < MAXQUOTAS; i++) {
607 dqrele(vp, ip->i_dquot[i]);
608 ip->i_dquot[i] = NODQUOT;
609 }
610#endif
611 ip->i_gid = gid;
612 ip->i_uid = uid;
613#ifdef QUOTA
614 if ((error = getinoquota(ip)) == 0) {
615 if (ouid == uid) {
616 dqrele(vp, ip->i_dquot[USRQUOTA]);
617 ip->i_dquot[USRQUOTA] = NODQUOT;
618 }
619 if (ogid == gid) {
620 dqrele(vp, ip->i_dquot[GRPQUOTA]);
621 ip->i_dquot[GRPQUOTA] = NODQUOT;
622 }
623 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
624 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
625 goto good;
626 else
627 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
628 }
629 for (i = 0; i < MAXQUOTAS; i++) {
630 dqrele(vp, ip->i_dquot[i]);
631 ip->i_dquot[i] = NODQUOT;
632 }
633 }
634 ip->i_gid = ogid;
635 ip->i_uid = ouid;
636 if (getinoquota(ip) == 0) {
637 if (ouid == uid) {
638 dqrele(vp, ip->i_dquot[USRQUOTA]);
639 ip->i_dquot[USRQUOTA] = NODQUOT;
640 }
641 if (ogid == gid) {
642 dqrele(vp, ip->i_dquot[GRPQUOTA]);
643 ip->i_dquot[GRPQUOTA] = NODQUOT;
644 }
645 (void) chkdq(ip, change, cred, FORCE|CHOWN);
646 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
647 (void) getinoquota(ip);
648 }
649 return (error);
650good:
651 if (getinoquota(ip))
652 panic("ufs_chown: lost quota");
653#endif /* QUOTA */
654 ip->i_flag |= IN_CHANGE;
655 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
656 ip->i_mode &= ~(ISUID | ISGID);
657 return (0);
658}
659
660/*
661 * Mmap a file
662 *
663 * NB Currently unsupported.
0973c589
CP
664 *
665 * ufs_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
666 * struct thread *a_td)
984263bc
MD
667 */
668/* ARGSUSED */
0961aa92 669static
984263bc 670int
0973c589 671ufs_mmap(struct vop_mmap_args *ap)
984263bc 672{
984263bc
MD
673 return (EINVAL);
674}
675
0973c589
CP
676/*
677 * ufs_remove(struct vnode *a_dvp, struct vnode *a_vp,
678 * struct componentname *a_cnp)
679 */
0961aa92 680static
984263bc 681int
0973c589 682ufs_remove(struct vop_remove_args *ap)
984263bc
MD
683{
684 struct inode *ip;
685 struct vnode *vp = ap->a_vp;
686 struct vnode *dvp = ap->a_dvp;
687 int error;
688
689 ip = VTOI(vp);
690 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
691 (VTOI(dvp)->i_flags & APPEND)) {
692 error = EPERM;
693 goto out;
694 }
695 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
696 VN_KNOTE(vp, NOTE_DELETE);
697 VN_KNOTE(dvp, NOTE_WRITE);
698out:
699 return (error);
700}
701
702/*
703 * link vnode call
0973c589
CP
704 *
705 * ufs_link(struct vnode *a_tdvp, struct vnode *a_vp,
706 * struct componentname *a_cnp)
984263bc 707 */
0961aa92 708static
984263bc 709int
0973c589 710ufs_link(struct vop_link_args *ap)
984263bc
MD
711{
712 struct vnode *vp = ap->a_vp;
713 struct vnode *tdvp = ap->a_tdvp;
714 struct componentname *cnp = ap->a_cnp;
dadab5e9 715 struct thread *td = cnp->cn_td;
984263bc
MD
716 struct inode *ip;
717 struct direct newdir;
718 int error;
719
720#ifdef DIAGNOSTIC
2b69e610 721 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
722 panic("ufs_link: no name");
723#endif
724 if (tdvp->v_mount != vp->v_mount) {
725 error = EXDEV;
726 goto out2;
727 }
5fd012e0 728 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, td))) {
984263bc
MD
729 goto out2;
730 }
731 ip = VTOI(vp);
732 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
733 error = EMLINK;
734 goto out1;
735 }
736 if (ip->i_flags & (IMMUTABLE | APPEND)) {
737 error = EPERM;
738 goto out1;
739 }
740 ip->i_effnlink++;
741 ip->i_nlink++;
742 ip->i_flag |= IN_CHANGE;
743 if (DOINGSOFTDEP(vp))
744 softdep_change_linkcnt(ip);
745 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
746 if (!error) {
747 ufs_makedirentry(ip, cnp, &newdir);
748 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
749 }
750
751 if (error) {
752 ip->i_effnlink--;
753 ip->i_nlink--;
754 ip->i_flag |= IN_CHANGE;
755 if (DOINGSOFTDEP(vp))
756 softdep_change_linkcnt(ip);
757 }
758out1:
759 if (tdvp != vp)
5fd012e0 760 VOP_UNLOCK(vp, 0, td);
984263bc
MD
761out2:
762 VN_KNOTE(vp, NOTE_LINK);
763 VN_KNOTE(tdvp, NOTE_WRITE);
764 return (error);
765}
766
767/*
768 * whiteout vnode call
0973c589
CP
769 *
770 * ufs_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
984263bc 771 */
0961aa92 772static
984263bc 773int
0973c589 774ufs_whiteout(struct vop_whiteout_args *ap)
984263bc
MD
775{
776 struct vnode *dvp = ap->a_dvp;
777 struct componentname *cnp = ap->a_cnp;
778 struct direct newdir;
779 int error = 0;
780
781 switch (ap->a_flags) {
2b69e610 782 case NAMEI_LOOKUP:
984263bc
MD
783 /* 4.4 format directories support whiteout operations */
784 if (dvp->v_mount->mnt_maxsymlinklen > 0)
785 return (0);
786 return (EOPNOTSUPP);
787
2b69e610 788 case NAMEI_CREATE:
984263bc
MD
789 /* create a new directory whiteout */
790#ifdef DIAGNOSTIC
2b69e610 791 if ((cnp->cn_flags & CNP_SAVENAME) == 0)
984263bc
MD
792 panic("ufs_whiteout: missing name");
793 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
794 panic("ufs_whiteout: old format filesystem");
795#endif
796
797 newdir.d_ino = WINO;
798 newdir.d_namlen = cnp->cn_namelen;
799 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
800 newdir.d_type = DT_WHT;
801 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
802 break;
803
2b69e610 804 case NAMEI_DELETE:
984263bc
MD
805 /* remove an existing directory whiteout */
806#ifdef DIAGNOSTIC
807 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
808 panic("ufs_whiteout: old format filesystem");
809#endif
810
2b69e610 811 cnp->cn_flags &= ~CNP_DOWHITEOUT;
984263bc
MD
812 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
813 break;
814 default:
815 panic("ufs_whiteout: unknown op");
816 }
817 return (error);
818}
819
820/*
821 * Rename system call.
822 * rename("foo", "bar");
823 * is essentially
824 * unlink("bar");
825 * link("foo", "bar");
826 * unlink("foo");
827 * but ``atomically''. Can't do full commit without saving state in the
828 * inode on disk which isn't feasible at this time. Best we can do is
829 * always guarantee the target exists.
830 *
831 * Basic algorithm is:
832 *
833 * 1) Bump link count on source while we're linking it to the
834 * target. This also ensure the inode won't be deleted out
835 * from underneath us while we work (it may be truncated by
836 * a concurrent `trunc' or `open' for creation).
837 * 2) Link source to destination. If destination already exists,
838 * delete it first.
839 * 3) Unlink source reference to inode if still around. If a
840 * directory was moved and the parent of the destination
841 * is different from the source, patch the ".." entry in the
842 * directory.
0973c589
CP
843 *
844 * ufs_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
845 * struct componentname *a_fcnp, struct vnode *a_tdvp,
846 * struct vnode *a_tvp, struct componentname *a_tcnp)
984263bc 847 */
0961aa92 848static
984263bc 849int
0973c589 850ufs_rename(struct vop_rename_args *ap)
984263bc
MD
851{
852 struct vnode *tvp = ap->a_tvp;
3ff2135f 853 struct vnode *tdvp = ap->a_tdvp;
984263bc
MD
854 struct vnode *fvp = ap->a_fvp;
855 struct vnode *fdvp = ap->a_fdvp;
856 struct componentname *tcnp = ap->a_tcnp;
857 struct componentname *fcnp = ap->a_fcnp;
dadab5e9 858 struct thread *td = fcnp->cn_td;
984263bc
MD
859 struct inode *ip, *xp, *dp;
860 struct direct newdir;
9ab06300
MD
861 ino_t oldparent = 0, newparent = 0;
862 int doingdirectory = 0;
984263bc
MD
863 int error = 0, ioflag;
864
865#ifdef DIAGNOSTIC
2b69e610
MD
866 if ((tcnp->cn_flags & CNP_HASBUF) == 0 ||
867 (fcnp->cn_flags & CNP_HASBUF) == 0)
984263bc 868 panic("ufs_rename: no name");
f719c866 869#endif /* DIAGNOSTIC */
984263bc
MD
870 /*
871 * Check for cross-device rename.
872 */
873 if ((fvp->v_mount != tdvp->v_mount) ||
874 (tvp && (fvp->v_mount != tvp->v_mount))) {
875 error = EXDEV;
876abortit:
877 if (tdvp == tvp)
878 vrele(tdvp);
879 else
880 vput(tdvp);
881 if (tvp)
882 vput(tvp);
883 vrele(fdvp);
884 vrele(fvp);
885 return (error);
886 }
887
888 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
889 (VTOI(tdvp)->i_flags & APPEND))) {
890 error = EPERM;
891 goto abortit;
892 }
893
894 /*
895 * Renaming a file to itself has no effect. The upper layers should
896 * not call us in that case. Temporarily just warn if they do.
897 */
898 if (fvp == tvp) {
899 printf("ufs_rename: fvp == tvp (can't happen)\n");
900 error = 0;
901 goto abortit;
902 }
903
5fd012e0 904 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
984263bc
MD
905 goto abortit;
906 dp = VTOI(fdvp);
907 ip = VTOI(fvp);
908 if (ip->i_nlink >= LINK_MAX) {
5fd012e0 909 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
910 error = EMLINK;
911 goto abortit;
912 }
913 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
914 || (dp->i_flags & APPEND)) {
5fd012e0 915 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
916 error = EPERM;
917 goto abortit;
918 }
919 if ((ip->i_mode & IFMT) == IFDIR) {
920 /*
921 * Avoid ".", "..", and aliases of "." for obvious reasons.
922 */
923 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
2b69e610 924 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
984263bc 925 (ip->i_flag & IN_RENAME)) {
5fd012e0 926 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
927 error = EINVAL;
928 goto abortit;
929 }
930 ip->i_flag |= IN_RENAME;
931 oldparent = dp->i_number;
932 doingdirectory = 1;
933 }
934 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
935 vrele(fdvp);
936
937 /*
938 * When the target exists, both the directory
939 * and target vnodes are returned locked.
940 */
941 dp = VTOI(tdvp);
942 xp = NULL;
943 if (tvp)
944 xp = VTOI(tvp);
945
946 /*
947 * 1) Bump link count while we're moving stuff
948 * around. If we crash somewhere before
949 * completing our work, the link count
950 * may be wrong, but correctable.
951 */
952 ip->i_effnlink++;
953 ip->i_nlink++;
954 ip->i_flag |= IN_CHANGE;
955 if (DOINGSOFTDEP(fvp))
956 softdep_change_linkcnt(ip);
957 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
958 DOINGASYNC(fvp)))) != 0) {
5fd012e0 959 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
960 goto bad;
961 }
962
963 /*
964 * If ".." must be changed (ie the directory gets a new
965 * parent) then the source directory must not be in the
966 * directory heirarchy above the target, as this would
967 * orphan everything below the source directory. Also
968 * the user must have write permission in the source so
969 * as to be able to change "..". We must repeat the call
970 * to namei, as the parent directory is unlocked by the
971 * call to checkpath().
972 */
dadab5e9 973 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
5fd012e0 974 VOP_UNLOCK(fvp, 0, td);
984263bc
MD
975 if (oldparent != dp->i_number)
976 newparent = dp->i_number;
977 if (doingdirectory && newparent) {
978 if (error) /* write access check above */
979 goto bad;
980 if (xp != NULL)
981 vput(tvp);
982 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
983 if (error)
984 goto out;
2b69e610 985 if ((tcnp->cn_flags & CNP_SAVESTART) == 0)
984263bc 986 panic("ufs_rename: lost to startdir");
597aea93 987 vref(tdvp);
984263bc
MD
988 error = relookup(tdvp, &tvp, tcnp);
989 if (error)
990 goto out;
991 vrele(tdvp);
992 dp = VTOI(tdvp);
993 xp = NULL;
994 if (tvp)
995 xp = VTOI(tvp);
996 }
997 /*
998 * 2) If target doesn't exist, link the target
999 * to the source and unlink the source.
1000 * Otherwise, rewrite the target directory
1001 * entry to reference the source inode and
1002 * expunge the original entry's existence.
1003 */
1004 if (xp == NULL) {
1005 if (dp->i_dev != ip->i_dev)
1006 panic("ufs_rename: EXDEV");
1007 /*
1008 * Account for ".." in new directory.
1009 * When source and destination have the same
1010 * parent we don't fool with the link count.
1011 */
1012 if (doingdirectory && newparent) {
1013 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1014 error = EMLINK;
1015 goto bad;
1016 }
1017 dp->i_effnlink++;
1018 dp->i_nlink++;
1019 dp->i_flag |= IN_CHANGE;
1020 if (DOINGSOFTDEP(tdvp))
1021 softdep_change_linkcnt(dp);
1022 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1023 DOINGASYNC(tdvp)));
1024 if (error)
1025 goto bad;
1026 }
1027 ufs_makedirentry(ip, tcnp, &newdir);
1028 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1029 if (error) {
1030 if (doingdirectory && newparent) {
1031 dp->i_effnlink--;
1032 dp->i_nlink--;
1033 dp->i_flag |= IN_CHANGE;
1034 if (DOINGSOFTDEP(tdvp))
1035 softdep_change_linkcnt(dp);
1036 (void)UFS_UPDATE(tdvp, 1);
1037 }
1038 goto bad;
1039 }
1040 VN_KNOTE(tdvp, NOTE_WRITE);
1041 vput(tdvp);
1042 } else {
1043 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1044 panic("ufs_rename: EXDEV");
1045 /*
1046 * Short circuit rename(foo, foo).
1047 */
1048 if (xp->i_number == ip->i_number)
1049 panic("ufs_rename: same file");
1050 /*
1051 * If the parent directory is "sticky", then the user must
1052 * own the parent directory, or the destination of the rename,
1053 * otherwise the destination may not be changed (except by
1054 * root). This implements append-only directories.
1055 */
1056 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1057 tcnp->cn_cred->cr_uid != dp->i_uid &&
1058 xp->i_uid != tcnp->cn_cred->cr_uid) {
1059 error = EPERM;
1060 goto bad;
1061 }
1062 /*
1063 * Target must be empty if a directory and have no links
1064 * to it. Also, ensure source and target are compatible
1065 * (both directories, or both not directories).
8e005a45
MD
1066 *
1067 * Purge the file or directory being replaced from the
1068 * nameccache.
984263bc
MD
1069 */
1070 if ((xp->i_mode&IFMT) == IFDIR) {
1071 if ((xp->i_effnlink > 2) ||
1072 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1073 error = ENOTEMPTY;
1074 goto bad;
1075 }
1076 if (!doingdirectory) {
1077 error = ENOTDIR;
1078 goto bad;
1079 }
8e005a45
MD
1080 cache_purge(tvp);
1081 } else if (doingdirectory == 0) {
1082 cache_purge(tvp);
1083 } else {
984263bc
MD
1084 error = EISDIR;
1085 goto bad;
1086 }
9ab06300
MD
1087 /*
1088 * note: inode passed to ufs_dirrewrite() is 0 for a
1089 * non-directory file rename, 1 for a directory rename
1090 * in the same directory, and > 1 for an inode representing
1091 * the new directory.
1092 */
984263bc
MD
1093 error = ufs_dirrewrite(dp, xp, ip->i_number,
1094 IFTODT(ip->i_mode),
9ab06300
MD
1095 (doingdirectory && newparent) ?
1096 newparent : (ino_t)doingdirectory);
984263bc
MD
1097 if (error)
1098 goto bad;
1099 if (doingdirectory) {
1100 if (!newparent) {
1101 dp->i_effnlink--;
1102 if (DOINGSOFTDEP(tdvp))
1103 softdep_change_linkcnt(dp);
1104 }
1105 xp->i_effnlink--;
1106 if (DOINGSOFTDEP(tvp))
1107 softdep_change_linkcnt(xp);
1108 }
1109 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1110 /*
1111 * Truncate inode. The only stuff left in the directory
1112 * is "." and "..". The "." reference is inconsequential
1113 * since we are quashing it. We have removed the "."
1114 * reference and the reference in the parent directory,
1115 * but there may be other hard links. The soft
1116 * dependency code will arrange to do these operations
1117 * after the parent directory entry has been deleted on
1118 * disk, so when running with that code we avoid doing
1119 * them now.
1120 */
1121 if (!newparent) {
1122 dp->i_nlink--;
1123 dp->i_flag |= IN_CHANGE;
1124 }
1125 xp->i_nlink--;
1126 xp->i_flag |= IN_CHANGE;
1127 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC;
1128 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
dadab5e9 1129 tcnp->cn_cred, tcnp->cn_td)) != 0)
984263bc
MD
1130 goto bad;
1131 }
1132 VN_KNOTE(tdvp, NOTE_WRITE);
1133 vput(tdvp);
1134 VN_KNOTE(tvp, NOTE_DELETE);
1135 vput(tvp);
1136 xp = NULL;
1137 }
1138
1139 /*
1140 * 3) Unlink the source.
1141 */
2b69e610
MD
1142 fcnp->cn_flags &= ~CNP_MODMASK;
1143 fcnp->cn_flags |= CNP_LOCKPARENT | CNP_LOCKLEAF;
1144 if ((fcnp->cn_flags & CNP_SAVESTART) == 0)
984263bc 1145 panic("ufs_rename: lost from startdir");
597aea93 1146 vref(fdvp);
984263bc
MD
1147 error = relookup(fdvp, &fvp, fcnp);
1148 if (error == 0)
1149 vrele(fdvp);
1150 if (fvp != NULL) {
1151 xp = VTOI(fvp);
1152 dp = VTOI(fdvp);
1153 } else {
1154 /*
1155 * From name has disappeared.
1156 */
1157 if (doingdirectory)
1158 panic("ufs_rename: lost dir entry");
1159 vrele(ap->a_fvp);
1160 return (0);
1161 }
1162 /*
1163 * Ensure that the directory entry still exists and has not
1164 * changed while the new name has been entered. If the source is
1165 * a file then the entry may have been unlinked or renamed. In
1166 * either case there is no further work to be done. If the source
1167 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1168 * flag ensures that it cannot be moved by another rename or removed
1169 * by a rmdir.
1170 */
1171 if (xp != ip) {
1172 if (doingdirectory)
1173 panic("ufs_rename: lost dir entry");
1174 } else {
1175 /*
1176 * If the source is a directory with a
1177 * new parent, the link count of the old
1178 * parent directory must be decremented
1179 * and ".." set to point to the new parent.
1180 */
1181 if (doingdirectory && newparent) {
1182 xp->i_offset = mastertemplate.dot_reclen;
1183 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
8e005a45 1184 /*cache_purge(fdvp);*/
984263bc
MD
1185 }
1186 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1187 xp->i_flag &= ~IN_RENAME;
1188 }
8e005a45
MD
1189
1190 /*
1191 * The old API can only do a kitchen sink invalidation. It will
1192 * not be possible to '..' through the disconnected fvp (if a
1193 * directory) until it is reconnected by a forward lookup.
1194 */
1195 cache_purge(fvp);
984263bc
MD
1196 VN_KNOTE(fvp, NOTE_RENAME);
1197 if (dp)
1198 vput(fdvp);
1199 if (xp)
1200 vput(fvp);
1201 vrele(ap->a_fvp);
1202 return (error);
1203
1204bad:
1205 if (xp)
1206 vput(ITOV(xp));
1207 vput(ITOV(dp));
1208out:
1209 if (doingdirectory)
1210 ip->i_flag &= ~IN_RENAME;
5fd012e0 1211 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
984263bc
MD
1212 ip->i_effnlink--;
1213 ip->i_nlink--;
1214 ip->i_flag |= IN_CHANGE;
1215 ip->i_flag &= ~IN_RENAME;
1216 if (DOINGSOFTDEP(fvp))
1217 softdep_change_linkcnt(ip);
1218 vput(fvp);
1219 } else
1220 vrele(fvp);
1221 return (error);
1222}
1223
1224/*
1225 * Mkdir system call
0973c589
CP
1226 *
1227 * ufs_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
1228 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc 1229 */
0961aa92 1230static
984263bc 1231int
0973c589 1232ufs_mkdir(struct vop_mkdir_args *ap)
984263bc 1233{
3ff2135f
RG
1234 struct vnode *dvp = ap->a_dvp;
1235 struct vattr *vap = ap->a_vap;
1236 struct componentname *cnp = ap->a_cnp;
1237 struct inode *ip, *dp;
984263bc
MD
1238 struct vnode *tvp;
1239 struct buf *bp;
1240 struct dirtemplate dirtemplate, *dtp;
1241 struct direct newdir;
1242 int error, dmode;
1243 long blkoff;
1244
1245#ifdef DIAGNOSTIC
2b69e610 1246 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
1247 panic("ufs_mkdir: no name");
1248#endif
1249 dp = VTOI(dvp);
1250 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1251 error = EMLINK;
1252 goto out;
1253 }
1254 dmode = vap->va_mode & 0777;
1255 dmode |= IFDIR;
1256 /*
1257 * Must simulate part of ufs_makeinode here to acquire the inode,
1258 * but not have it entered in the parent directory. The entry is
1259 * made later after writing "." and ".." entries.
1260 */
1261 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1262 if (error)
1263 goto out;
1264 ip = VTOI(tvp);
1265 ip->i_gid = dp->i_gid;
1266#ifdef SUIDDIR
1267 {
1268#ifdef QUOTA
1269 struct ucred ucred, *ucp;
1270 ucp = cnp->cn_cred;
1271#endif
1272 /*
1273 * If we are hacking owners here, (only do this where told to)
1274 * and we are not giving it TO root, (would subvert quotas)
1275 * then go ahead and give it to the other user.
1276 * The new directory also inherits the SUID bit.
1277 * If user's UID and dir UID are the same,
1278 * 'give it away' so that the SUID is still forced on.
1279 */
1280 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1281 (dp->i_mode & ISUID) && dp->i_uid) {
1282 dmode |= ISUID;
1283 ip->i_uid = dp->i_uid;
1284#ifdef QUOTA
1285 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1286 /*
1287 * Make sure the correct user gets charged
1288 * for the space.
1289 * Make a dummy credential for the victim.
1290 * XXX This seems to never be accessed out of
1291 * our context so a stack variable is ok.
1292 */
1293 ucred.cr_ref = 1;
1294 ucred.cr_uid = ip->i_uid;
1295 ucred.cr_ngroups = 1;
1296 ucred.cr_groups[0] = dp->i_gid;
1297 ucp = &ucred;
1298 }
1299#endif
1300 } else
1301 ip->i_uid = cnp->cn_cred->cr_uid;
1302#ifdef QUOTA
1303 if ((error = getinoquota(ip)) ||
1304 (error = chkiq(ip, 1, ucp, 0))) {
1305 UFS_VFREE(tvp, ip->i_number, dmode);
1306 vput(tvp);
1307 return (error);
1308 }
1309#endif
1310 }
1311#else /* !SUIDDIR */
1312 ip->i_uid = cnp->cn_cred->cr_uid;
1313#ifdef QUOTA
1314 if ((error = getinoquota(ip)) ||
1315 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1316 UFS_VFREE(tvp, ip->i_number, dmode);
1317 vput(tvp);
1318 return (error);
1319 }
1320#endif
1321#endif /* !SUIDDIR */
1322 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1323 ip->i_mode = dmode;
1324 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1325 ip->i_effnlink = 2;
1326 ip->i_nlink = 2;
1327 if (DOINGSOFTDEP(tvp))
1328 softdep_change_linkcnt(ip);
2b69e610 1329 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
1330 ip->i_flags |= UF_OPAQUE;
1331
1332 /*
1333 * Bump link count in parent directory to reflect work done below.
1334 * Should be done before reference is created so cleanup is
1335 * possible if we crash.
1336 */
1337 dp->i_effnlink++;
1338 dp->i_nlink++;
1339 dp->i_flag |= IN_CHANGE;
1340 if (DOINGSOFTDEP(dvp))
1341 softdep_change_linkcnt(dp);
1342 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1343 if (error)
1344 goto bad;
1345
1346 /*
1347 * Initialize directory with "." and ".." from static template.
1348 */
1349 if (dvp->v_mount->mnt_maxsymlinklen > 0
1350 )
1351 dtp = &mastertemplate;
1352 else
1353 dtp = (struct dirtemplate *)&omastertemplate;
1354 dirtemplate = *dtp;
1355 dirtemplate.dot_ino = ip->i_number;
1356 dirtemplate.dotdot_ino = dp->i_number;
1357 if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1358 B_CLRBUF, &bp)) != 0)
1359 goto bad;
1360 ip->i_size = DIRBLKSIZ;
1361 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1362 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1363 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1364 if (DOINGSOFTDEP(tvp)) {
1365 /*
1366 * Ensure that the entire newly allocated block is a
1367 * valid directory so that future growth within the
1368 * block does not have to ensure that the block is
1369 * written before the inode.
1370 */
1371 blkoff = DIRBLKSIZ;
1372 while (blkoff < bp->b_bcount) {
1373 ((struct direct *)
1374 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1375 blkoff += DIRBLKSIZ;
1376 }
1377 }
1378 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1379 DOINGASYNC(tvp)))) != 0) {
1380 (void)VOP_BWRITE(bp->b_vp, bp);
1381 goto bad;
1382 }
1383 /*
1384 * Directory set up, now install its entry in the parent directory.
1385 *
1386 * If we are not doing soft dependencies, then we must write out the
1387 * buffer containing the new directory body before entering the new
1388 * name in the parent. If we are doing soft dependencies, then the
1389 * buffer containing the new directory body will be passed to and
1390 * released in the soft dependency code after the code has attached
1391 * an appropriate ordering dependency to the buffer which ensures that
1392 * the buffer is written before the new name is written in the parent.
1393 */
1394 if (DOINGASYNC(dvp))
1395 bdwrite(bp);
1396 else if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp->b_vp, bp))))
1397 goto bad;
1398 ufs_makedirentry(ip, cnp, &newdir);
1399 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1400
1401bad:
1402 if (error == 0) {
1403 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1404 *ap->a_vpp = tvp;
1405 } else {
1406 dp->i_effnlink--;
1407 dp->i_nlink--;
1408 dp->i_flag |= IN_CHANGE;
1409 if (DOINGSOFTDEP(dvp))
1410 softdep_change_linkcnt(dp);
1411 /*
1412 * No need to do an explicit VOP_TRUNCATE here, vrele will
1413 * do this for us because we set the link count to 0.
1414 */
1415 ip->i_effnlink = 0;
1416 ip->i_nlink = 0;
1417 ip->i_flag |= IN_CHANGE;
1418 if (DOINGSOFTDEP(tvp))
1419 softdep_change_linkcnt(ip);
1420 vput(tvp);
1421 }
1422out:
1423 return (error);
1424}
1425
1426/*
1427 * Rmdir system call.
0973c589
CP
1428 *
1429 * ufs_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1430 * struct componentname *a_cnp)
984263bc 1431 */
0961aa92 1432static
984263bc 1433int
0973c589 1434ufs_rmdir(struct vop_rmdir_args *ap)
984263bc
MD
1435{
1436 struct vnode *vp = ap->a_vp;
1437 struct vnode *dvp = ap->a_dvp;
1438 struct componentname *cnp = ap->a_cnp;
1439 struct inode *ip, *dp;
1440 int error, ioflag;
1441
1442 ip = VTOI(vp);
1443 dp = VTOI(dvp);
1444
1445 /*
1446 * Do not remove a directory that is in the process of being renamed.
1447 * Verify the directory is empty (and valid). Rmdir ".." will not be
1448 * valid since ".." will contain a reference to the current directory
1449 * and thus be non-empty. Do not allow the removal of mounted on
1450 * directories (this can happen when an NFS exported filesystem
1451 * tries to remove a locally mounted on directory).
1452 */
1453 error = 0;
1454 if (ip->i_flag & IN_RENAME) {
1455 error = EINVAL;
1456 goto out;
1457 }
1458 if (ip->i_effnlink != 2 ||
1459 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1460 error = ENOTEMPTY;
1461 goto out;
1462 }
1463 if ((dp->i_flags & APPEND)
1464 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1465 error = EPERM;
1466 goto out;
1467 }
1468 if (vp->v_mountedhere != 0) {
1469 error = EINVAL;
1470 goto out;
1471 }
1472 /*
1473 * Delete reference to directory before purging
1474 * inode. If we crash in between, the directory
1475 * will be reattached to lost+found,
1476 */
1477 dp->i_effnlink--;
1478 ip->i_effnlink--;
1479 if (DOINGSOFTDEP(vp)) {
1480 softdep_change_linkcnt(dp);
1481 softdep_change_linkcnt(ip);
1482 }
1483 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1484 if (error) {
1485 dp->i_effnlink++;
1486 ip->i_effnlink++;
1487 if (DOINGSOFTDEP(vp)) {
1488 softdep_change_linkcnt(dp);
1489 softdep_change_linkcnt(ip);
1490 }
1491 goto out;
1492 }
1493 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
984263bc
MD
1494 /*
1495 * Truncate inode. The only stuff left in the directory is "." and
1496 * "..". The "." reference is inconsequential since we are quashing
1497 * it. The soft dependency code will arrange to do these operations
1498 * after the parent directory entry has been deleted on disk, so
1499 * when running with that code we avoid doing them now.
1500 */
1501 if (!DOINGSOFTDEP(vp)) {
1502 dp->i_nlink--;
1503 dp->i_flag |= IN_CHANGE;
1504 ip->i_nlink--;
1505 ip->i_flag |= IN_CHANGE;
1506 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC;
1507 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
dadab5e9 1508 cnp->cn_td);
984263bc
MD
1509 }
1510 cache_purge(vp);
1511#ifdef UFS_DIRHASH
1512 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
1513 if (ip->i_dirhash != NULL)
1514 ufsdirhash_free(ip);
1515#endif
1516out:
1517 VN_KNOTE(vp, NOTE_DELETE);
1518 return (error);
1519}
1520
1521/*
1522 * symlink -- make a symbolic link
0973c589
CP
1523 *
1524 * ufs_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1525 * struct componentname *a_cnp, struct vattr *a_vap,
1526 * char *a_target)
984263bc 1527 */
0961aa92 1528static
984263bc 1529int
0973c589 1530ufs_symlink(struct vop_symlink_args *ap)
984263bc 1531{
3ff2135f
RG
1532 struct vnode *vp, **vpp = ap->a_vpp;
1533 struct inode *ip;
984263bc
MD
1534 int len, error;
1535
1536 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1537 vpp, ap->a_cnp);
1538 if (error)
1539 return (error);
1540 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1541 vp = *vpp;
1542 len = strlen(ap->a_target);
1543 if (len < vp->v_mount->mnt_maxsymlinklen) {
1544 ip = VTOI(vp);
1545 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1546 ip->i_size = len;
1547 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1548 } else
1549 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
dadab5e9
MD
1550 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred,
1551 (int *)0, NULL);
984263bc
MD
1552 if (error)
1553 vput(vp);
1554 return (error);
1555}
1556
1557/*
1558 * Vnode op for reading directories.
1559 *
1560 * The routine below assumes that the on-disk format of a directory
1561 * is the same as that defined by <sys/dirent.h>. If the on-disk
1562 * format changes, then it will be necessary to do a conversion
1563 * from the on-disk format that read returns to the format defined
1564 * by <sys/dirent.h>.
0973c589
CP
1565 *
1566 * ufs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
1567 * int *a_eofflag, int *ncookies, u_long **a_cookies)
984263bc 1568 */
0961aa92 1569static
984263bc 1570int
0973c589 1571ufs_readdir(struct vop_readdir_args *ap)
984263bc 1572{
3ff2135f 1573 struct uio *uio = ap->a_uio;
984263bc
MD
1574 int error;
1575 size_t count, lost;
1576 off_t off;
1577
1578 if (ap->a_ncookies != NULL)
1579 /*
1580 * Ensure that the block is aligned. The caller can use
1581 * the cookies to determine where in the block to start.
1582 */
1583 uio->uio_offset &= ~(DIRBLKSIZ - 1);
1584 off = uio->uio_offset;
1585 count = uio->uio_resid;
1586 /* Make sure we don't return partial entries. */
1587 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
1588 return (EINVAL);
1589 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1590 lost = uio->uio_resid - count;
1591 uio->uio_resid = count;
1592 uio->uio_iov->iov_len = count;
1593# if (BYTE_ORDER == LITTLE_ENDIAN)
1594 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
1595 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1596 } else {
1597 struct dirent *dp, *edp;
1598 struct uio auio;
1599 struct iovec aiov;
1600 caddr_t dirbuf;
1601 int readcnt;
1602 u_char tmp;
1603
1604 auio = *uio;
1605 auio.uio_iov = &aiov;
1606 auio.uio_iovcnt = 1;
1607 auio.uio_segflg = UIO_SYSSPACE;
1608 aiov.iov_len = count;
1609 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1610 aiov.iov_base = dirbuf;
1611 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
1612 if (error == 0) {
1613 readcnt = count - auio.uio_resid;
1614 edp = (struct dirent *)&dirbuf[readcnt];
1615 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
1616 tmp = dp->d_namlen;
1617 dp->d_namlen = dp->d_type;
1618 dp->d_type = tmp;
1619 if (dp->d_reclen > 0) {
1620 dp = (struct dirent *)
1621 ((char *)dp + dp->d_reclen);
1622 } else {
1623 error = EIO;
1624 break;
1625 }
1626 }
1627 if (dp >= edp)
1628 error = uiomove(dirbuf, readcnt, uio);
1629 }
1630 FREE(dirbuf, M_TEMP);
1631 }
1632# else
1633 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1634# endif
1635 if (!error && ap->a_ncookies != NULL) {
1636 struct dirent* dpStart;
1637 struct dirent* dpEnd;
1638 struct dirent* dp;
1639 int ncookies;
1640 u_long *cookies;
1641 u_long *cookiep;
1642
1643 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1644 panic("ufs_readdir: unexpected uio from NFS server");
1645 dpStart = (struct dirent *)
1646 (uio->uio_iov->iov_base - (uio->uio_offset - off));
1647 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
1648 for (dp = dpStart, ncookies = 0;
1649 dp < dpEnd;
1650 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
1651 ncookies++;
1652 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1653 M_WAITOK);
1654 for (dp = dpStart, cookiep = cookies;
1655 dp < dpEnd;
1656 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
1657 off += dp->d_reclen;
1658 *cookiep++ = (u_long) off;
1659 }
1660 *ap->a_ncookies = ncookies;
1661 *ap->a_cookies = cookies;
1662 }
1663 uio->uio_resid += lost;
1664 if (ap->a_eofflag)
1665 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
1666 return (error);
1667}
1668
1669/*
1670 * Return target name of a symbolic link
0973c589
CP
1671 *
1672 * ufs_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
984263bc 1673 */
0961aa92 1674static
984263bc 1675int
0973c589 1676ufs_readlink(struct vop_readlink_args *ap)
984263bc 1677{
3ff2135f
RG
1678 struct vnode *vp = ap->a_vp;
1679 struct inode *ip = VTOI(vp);
984263bc
MD
1680 int isize;
1681
1682 isize = ip->i_size;
1683 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
f719c866 1684 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
984263bc
MD
1685 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1686 return (0);
1687 }
1688 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1689}
1690
1691/*
1692 * Calculate the logical to physical mapping if not done already,
1693 * then call the device strategy routine.
1694 *
1695 * In order to be able to swap to a file, the VOP_BMAP operation may not
1696 * deadlock on memory. See ufs_bmap() for details.
0973c589
CP
1697 *
1698 * ufs_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc 1699 */
0961aa92 1700static
984263bc 1701int
0973c589 1702ufs_strategy(struct vop_strategy_args *ap)
984263bc 1703{
3ff2135f
RG
1704 struct buf *bp = ap->a_bp;
1705 struct vnode *vp = ap->a_vp;
1706 struct inode *ip;
984263bc
MD
1707 int error;
1708
1709 ip = VTOI(vp);
1710 if (vp->v_type == VBLK || vp->v_type == VCHR)
1711 panic("ufs_strategy: spec");
1712 if (bp->b_blkno == bp->b_lblkno) {
1713 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
1714 if (error) {
1715 bp->b_error = error;
1716 bp->b_flags |= B_ERROR;
1717 biodone(bp);
1718 return (error);
1719 }
9ab06300 1720 if (bp->b_blkno == (daddr_t)-1)
984263bc
MD
1721 vfs_bio_clrbuf(bp);
1722 }
9ab06300 1723 if (bp->b_blkno == (daddr_t)-1) {
984263bc
MD
1724 biodone(bp);
1725 return (0);
1726 }
1727 vp = ip->i_devvp;
1728 bp->b_dev = vp->v_rdev;
1729 VOP_STRATEGY(vp, bp);
1730 return (0);
1731}
1732
1733/*
1734 * Print out the contents of an inode.
0973c589
CP
1735 *
1736 * ufs_print(struct vnode *a_vp)
984263bc 1737 */
0961aa92 1738static
984263bc 1739int
0973c589 1740ufs_print(struct vop_print_args *ap)
984263bc 1741{
3ff2135f
RG
1742 struct vnode *vp = ap->a_vp;
1743 struct inode *ip = VTOI(vp);
984263bc
MD
1744
1745 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)",
1746 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1747 minor(ip->i_dev));
1748 if (vp->v_type == VFIFO)
1749 fifo_printinfo(vp);
3446c007 1750 lockmgr_printinfo(&vp->v_lock);
984263bc
MD
1751 printf("\n");
1752 return (0);
1753}
1754
1755/*
1756 * Read wrapper for special devices.
0973c589
CP
1757 *
1758 * ufsspec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1759 * struct ucred *a_cred)
984263bc 1760 */
0961aa92 1761static
984263bc 1762int
0973c589 1763ufsspec_read(struct vop_read_args *ap)
984263bc
MD
1764{
1765 int error, resid;
1766 struct inode *ip;
1767 struct uio *uio;
1768
1769 uio = ap->a_uio;
1770 resid = uio->uio_resid;
0961aa92 1771 error = VOCALL(spec_vnode_vops, &ap->a_head);
984263bc
MD
1772 /*
1773 * The inode may have been revoked during the call, so it must not
1774 * be accessed blindly here or in the other wrapper functions.
1775 */
1776 ip = VTOI(ap->a_vp);
1777 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1778 ip->i_flag |= IN_ACCESS;
1779 return (error);
1780}
1781
1782/*
1783 * Write wrapper for special devices.
0973c589
CP
1784 *
1785 * ufsspec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1786 * struct ucred *a_cred)
984263bc 1787 */
0961aa92 1788static
984263bc 1789int
0973c589 1790ufsspec_write(struct vop_write_args *ap)
984263bc
MD
1791{
1792 int error, resid;
1793 struct inode *ip;
1794 struct uio *uio;
1795
1796 uio = ap->a_uio;
1797 resid = uio->uio_resid;
0961aa92 1798 error = VOCALL(spec_vnode_vops, &ap->a_head);
984263bc
MD
1799 ip = VTOI(ap->a_vp);
1800 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1801 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1802 return (error);
1803}
1804
1805/*
1806 * Close wrapper for special devices.
1807 *
1808 * Update the times on the inode then do device close.
0973c589
CP
1809 *
1810 * ufsspec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1811 * struct thread *a_td)
984263bc 1812 */
0961aa92 1813static
984263bc 1814int
0973c589 1815ufsspec_close(struct vop_close_args *ap)
984263bc
MD
1816{
1817 struct vnode *vp = ap->a_vp;
1818
984263bc
MD
1819 if (vp->v_usecount > 1)
1820 ufs_itimes(vp);
0961aa92 1821 return (VOCALL(spec_vnode_vops, &ap->a_head));
984263bc
MD
1822}
1823
1824/*
1825 * Read wrapper for fifos.
0973c589
CP
1826 *
1827 * ufsfifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1828 * struct ucred *a_cred)
984263bc 1829 */
0961aa92 1830static
984263bc 1831int
0973c589 1832ufsfifo_read(struct vop_read_args *ap)
984263bc
MD
1833{
1834 int error, resid;
1835 struct inode *ip;
1836 struct uio *uio;
1837
1838 uio = ap->a_uio;
1839 resid = uio->uio_resid;
0961aa92 1840 error = VOCALL(fifo_vnode_vops, &ap->a_head);
984263bc
MD
1841 ip = VTOI(ap->a_vp);
1842 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1843 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1844 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1845 return (error);
1846}
1847
1848/*
1849 * Write wrapper for fifos.
0973c589
CP
1850 *
1851 * ufsfifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1852 * struct ucred *a_cred)
984263bc 1853 */
0961aa92 1854static
984263bc 1855int
0973c589 1856ufsfifo_write(struct vop_write_args *ap)
984263bc
MD
1857{
1858 int error, resid;
1859 struct inode *ip;
1860 struct uio *uio;
1861
1862 uio = ap->a_uio;
1863 resid = uio->uio_resid;
0961aa92 1864 error = VOCALL(fifo_vnode_vops, &ap->a_head);
984263bc
MD
1865 ip = VTOI(ap->a_vp);
1866 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1867 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1868 return (error);
1869}
1870
1871/*
1872 * Close wrapper for fifos.
1873 *
1874 * Update the times on the inode then do device close.
0973c589
CP
1875 *
1876 * ufsfifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1877 * struct thread *a_td)
984263bc 1878 */
0961aa92 1879static
984263bc 1880int
0973c589 1881ufsfifo_close(struct vop_close_args *ap)
984263bc
MD
1882{
1883 struct vnode *vp = ap->a_vp;
1884
984263bc
MD
1885 if (vp->v_usecount > 1)
1886 ufs_itimes(vp);
0961aa92 1887 return (VOCALL(fifo_vnode_vops, &ap->a_head));
984263bc
MD
1888}
1889
1890/*
1891 * Kqfilter wrapper for fifos.
1892 *
1893 * Fall through to ufs kqfilter routines if needed
1894 */
0961aa92 1895static
984263bc 1896int
0973c589 1897ufsfifo_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
1898{
1899 int error;
1900
0961aa92 1901 error = VOCALL(fifo_vnode_vops, &ap->a_head);
984263bc
MD
1902 if (error)
1903 error = ufs_kqfilter(ap);
1904 return (error);
1905}
1906
1907/*
1908 * Return POSIX pathconf information applicable to ufs filesystems.
0973c589
CP
1909 *
1910 * ufs_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
984263bc 1911 */
0961aa92 1912static
984263bc 1913int
0973c589 1914ufs_pathconf(struct vop_pathconf_args *ap)
984263bc 1915{
984263bc
MD
1916 switch (ap->a_name) {
1917 case _PC_LINK_MAX:
1918 *ap->a_retval = LINK_MAX;
1919 return (0);
1920 case _PC_NAME_MAX:
1921 *ap->a_retval = NAME_MAX;
1922 return (0);
1923 case _PC_PATH_MAX:
1924 *ap->a_retval = PATH_MAX;
1925 return (0);
1926 case _PC_PIPE_BUF:
1927 *ap->a_retval = PIPE_BUF;
1928 return (0);
1929 case _PC_CHOWN_RESTRICTED:
1930 *ap->a_retval = 1;
1931 return (0);
1932 case _PC_NO_TRUNC:
1933 *ap->a_retval = 1;
1934 return (0);
1935 default:
1936 return (EINVAL);
1937 }
1938 /* NOTREACHED */
1939}
1940
1941/*
1942 * Advisory record locking support
0973c589
CP
1943 *
1944 * ufs_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
1945 * int a_flags)
984263bc 1946 */
0961aa92 1947static
984263bc 1948int
0973c589 1949ufs_advlock(struct vop_advlock_args *ap)
984263bc 1950{
3ff2135f 1951 struct inode *ip = VTOI(ap->a_vp);
984263bc
MD
1952
1953 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1954}
1955
1956/*
1957 * Initialize the vnode associated with a new inode, handle aliased
1958 * vnodes.
1959 */
1960int
0961aa92 1961ufs_vinit(struct mount *mntp, struct vnode **vpp)
984263bc
MD
1962{
1963 struct inode *ip;
1964 struct vnode *vp;
1965 struct timeval tv;
1966
1967 vp = *vpp;
1968 ip = VTOI(vp);
1969 switch(vp->v_type = IFTOVT(ip->i_mode)) {
1970 case VCHR:
1971 case VBLK:
0961aa92 1972 vp->v_ops = mntp->mnt_vn_spec_ops;
984263bc
MD
1973 addaliasu(vp, ip->i_rdev);
1974 break;
1975 case VFIFO:
0961aa92 1976 vp->v_ops = mntp->mnt_vn_fifo_ops;
984263bc
MD
1977 break;
1978 default:
1979 break;
1980
1981 }
1982 if (ip->i_number == ROOTINO)
1983 vp->v_flag |= VROOT;
1984 /*
1985 * Initialize modrev times
1986 */
1987 getmicrouptime(&tv);
1988 SETHIGH(ip->i_modrev, tv.tv_sec);
1989 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
1990 *vpp = vp;
1991 return (0);
1992}
1993
1994/*
1995 * Allocate a new inode.
1996 */
0961aa92 1997static
984263bc 1998int
0973c589
CP
1999ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
2000 struct componentname *cnp)
984263bc 2001{
3ff2135f 2002 struct inode *ip, *pdir;
984263bc
MD
2003 struct direct newdir;
2004 struct vnode *tvp;
2005 int error;
2006
2007 pdir = VTOI(dvp);
2008#ifdef DIAGNOSTIC
2b69e610 2009 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
2010 panic("ufs_makeinode: no name");
2011#endif
2012 *vpp = NULL;
2013 if ((mode & IFMT) == 0)
2014 mode |= IFREG;
2015
2016 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
2017 if (error)
2018 return (error);
2019 ip = VTOI(tvp);
2020 ip->i_gid = pdir->i_gid;
2021#ifdef SUIDDIR
2022 {
2023#ifdef QUOTA
2024 struct ucred ucred, *ucp;
2025 ucp = cnp->cn_cred;
2026#endif
2027 /*
2028 * If we are not the owner of the directory,
2029 * and we are hacking owners here, (only do this where told to)
2030 * and we are not giving it TO root, (would subvert quotas)
2031 * then go ahead and give it to the other user.
2032 * Note that this drops off the execute bits for security.
2033 */
2034 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
2035 (pdir->i_mode & ISUID) &&
2036 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
2037 ip->i_uid = pdir->i_uid;
2038 mode &= ~07111;
2039#ifdef QUOTA
2040 /*
2041 * Make sure the correct user gets charged
2042 * for the space.
2043 * Quickly knock up a dummy credential for the victim.
2044 * XXX This seems to never be accessed out of our
2045 * context so a stack variable is ok.
2046 */
2047 ucred.cr_ref = 1;
2048 ucred.cr_uid = ip->i_uid;
2049 ucred.cr_ngroups = 1;
2050 ucred.cr_groups[0] = pdir->i_gid;
2051 ucp = &ucred;
2052#endif
2053 } else
2054 ip->i_uid = cnp->cn_cred->cr_uid;
2055
2056#ifdef QUOTA
2057 if ((error = getinoquota(ip)) ||
2058 (error = chkiq(ip, 1, ucp, 0))) {
2059 UFS_VFREE(tvp, ip->i_number, mode);
2060 vput(tvp);
2061 return (error);
2062 }
2063#endif
2064 }
2065#else /* !SUIDDIR */
2066 ip->i_uid = cnp->cn_cred->cr_uid;
2067#ifdef QUOTA
2068 if ((error = getinoquota(ip)) ||
2069 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2070 UFS_VFREE(tvp, ip->i_number, mode);
2071 vput(tvp);
2072 return (error);
2073 }
2074#endif
2075#endif /* !SUIDDIR */
2076 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2077 ip->i_mode = mode;
2078 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2079 ip->i_effnlink = 1;
2080 ip->i_nlink = 1;
2081 if (DOINGSOFTDEP(tvp))
2082 softdep_change_linkcnt(ip);
2083 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
dadab5e9 2084 suser_cred(cnp->cn_cred, 0)) {
984263bc 2085 ip->i_mode &= ~ISGID;
41c20dac 2086 }
984263bc 2087
2b69e610 2088 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
2089 ip->i_flags |= UF_OPAQUE;
2090
2091 /*
2092 * Make sure inode goes to disk before directory entry.
2093 */
2094 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2095 if (error)
2096 goto bad;
2097 ufs_makedirentry(ip, cnp, &newdir);
2098 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2099 if (error)
2100 goto bad;
2101 *vpp = tvp;
2102 return (0);
2103
2104bad:
2105 /*
2106 * Write error occurred trying to update the inode
2107 * or the directory so must deallocate the inode.
2108 */
2109 ip->i_effnlink = 0;
2110 ip->i_nlink = 0;
2111 ip->i_flag |= IN_CHANGE;
2112 if (DOINGSOFTDEP(tvp))
2113 softdep_change_linkcnt(ip);
2114 vput(tvp);
2115 return (error);
2116}
2117
2118static int
0973c589 2119ufs_missingop(struct vop_generic_args *ap)
984263bc 2120{
984263bc
MD
2121 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2122 return (EOPNOTSUPP);
2123}
2124
2125static struct filterops ufsread_filtops =
2126 { 1, NULL, filt_ufsdetach, filt_ufsread };
2127static struct filterops ufswrite_filtops =
2128 { 1, NULL, filt_ufsdetach, filt_ufswrite };
2129static struct filterops ufsvnode_filtops =
2130 { 1, NULL, filt_ufsdetach, filt_ufsvnode };
2131
0973c589
CP
2132/*
2133 * ufs_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2134 */
984263bc 2135static int
0973c589 2136ufs_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
2137{
2138 struct vnode *vp = ap->a_vp;
2139 struct knote *kn = ap->a_kn;
41a01a4d 2140 lwkt_tokref ilock;
984263bc
MD
2141
2142 switch (kn->kn_filter) {
2143 case EVFILT_READ:
2144 kn->kn_fop = &ufsread_filtops;
2145 break;
2146 case EVFILT_WRITE:
2147 kn->kn_fop = &ufswrite_filtops;
2148 break;
2149 case EVFILT_VNODE:
2150 kn->kn_fop = &ufsvnode_filtops;
2151 break;
2152 default:
2153 return (1);
2154 }
2155
2156 kn->kn_hook = (caddr_t)vp;
2157
41a01a4d 2158 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc 2159 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
41a01a4d 2160 lwkt_reltoken(&ilock);
984263bc
MD
2161
2162 return (0);
2163}
2164
2165static void
2166filt_ufsdetach(struct knote *kn)
2167{
2168 struct vnode *vp = (struct vnode *)kn->kn_hook;
41a01a4d 2169 lwkt_tokref ilock;
984263bc 2170
41a01a4d 2171 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc
MD
2172 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2173 kn, knote, kn_selnext);
41a01a4d 2174 lwkt_reltoken(&ilock);
984263bc
MD
2175}
2176
2177/*ARGSUSED*/
2178static int
2179filt_ufsread(struct knote *kn, long hint)
2180{
2181 struct vnode *vp = (struct vnode *)kn->kn_hook;
2182 struct inode *ip = VTOI(vp);
2183
2184 /*
2185 * filesystem is gone, so set the EOF flag and schedule
2186 * the knote for deletion.
2187 */
2188 if (hint == NOTE_REVOKE) {
2189 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2190 return (1);
2191 }
2192
2193 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2194 return (kn->kn_data != 0);
2195}
2196
2197/*ARGSUSED*/
2198static int
2199filt_ufswrite(struct knote *kn, long hint)
2200{
984263bc
MD
2201 /*
2202 * filesystem is gone, so set the EOF flag and schedule
2203 * the knote for deletion.
2204 */
2205 if (hint == NOTE_REVOKE)
2206 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2207
2208 kn->kn_data = 0;
2209 return (1);
2210}
2211
2212static int
2213filt_ufsvnode(struct knote *kn, long hint)
2214{
984263bc
MD
2215 if (kn->kn_sfflags & hint)
2216 kn->kn_fflags |= hint;
2217 if (hint == NOTE_REVOKE) {
2218 kn->kn_flags |= EV_EOF;
2219 return (1);
2220 }
2221 return (kn->kn_fflags != 0);
2222}
2223
2224/* Global vfs data structures for ufs. */
2d3e977e 2225static struct vop_ops *ufs_vnode_vops;
984263bc 2226static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2d3e977e
MD
2227 { &vop_default_desc, vop_defaultop },
2228 { &vop_fsync_desc, (void *) ufs_missingop },
2229 { &vop_read_desc, (void *) ufs_missingop },
2230 { &vop_reallocblks_desc, (void *) ufs_missingop },
2231 { &vop_write_desc, (void *) ufs_missingop },
2232 { &vop_access_desc, (void *) ufs_access },
2233 { &vop_advlock_desc, (void *) ufs_advlock },
2234 { &vop_bmap_desc, (void *) ufs_bmap },
2235 { &vop_cachedlookup_desc, (void *) ufs_lookup },
2236 { &vop_close_desc, (void *) ufs_close },
2237 { &vop_create_desc, (void *) ufs_create },
2238 { &vop_getattr_desc, (void *) ufs_getattr },
2239 { &vop_inactive_desc, (void *) ufs_inactive },
2240 { &vop_islocked_desc, (void *) vop_stdislocked },
2241 { &vop_link_desc, (void *) ufs_link },
2242 { &vop_lock_desc, (void *) vop_stdlock },
2243 { &vop_lookup_desc, (void *) vfs_cache_lookup },
2244 { &vop_mkdir_desc, (void *) ufs_mkdir },
2245 { &vop_mknod_desc, (void *) ufs_mknod },
2246 { &vop_mmap_desc, (void *) ufs_mmap },
2247 { &vop_open_desc, (void *) ufs_open },
2248 { &vop_pathconf_desc, (void *) ufs_pathconf },
2249 { &vop_poll_desc, (void *) vop_stdpoll },
2250 { &vop_kqfilter_desc, (void *) ufs_kqfilter },
2251 { &vop_print_desc, (void *) ufs_print },
2252 { &vop_readdir_desc, (void *) ufs_readdir },
2253 { &vop_readlink_desc, (void *) ufs_readlink },
2254 { &vop_reclaim_desc, (void *) ufs_reclaim },
2255 { &vop_remove_desc, (void *) ufs_remove },
2256 { &vop_rename_desc, (void *) ufs_rename },
2257 { &vop_rmdir_desc, (void *) ufs_rmdir },
2258 { &vop_setattr_desc, (void *) ufs_setattr },
2259 { &vop_strategy_desc, (void *) ufs_strategy },
2260 { &vop_symlink_desc, (void *) ufs_symlink },
2261 { &vop_unlock_desc, (void *) vop_stdunlock },
2262 { &vop_whiteout_desc, (void *) ufs_whiteout },
984263bc
MD
2263 { NULL, NULL }
2264};
2265static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2d3e977e 2266 { &ufs_vnode_vops, ufs_vnodeop_entries };
984263bc 2267
2d3e977e 2268static struct vop_ops *ufs_spec_vops;
984263bc 2269static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2d3e977e
MD
2270 { &vop_default_desc, (void *) spec_vnoperate },
2271 { &vop_fsync_desc, (void *) ufs_missingop },
2272 { &vop_access_desc, (void *) ufs_access },
2273 { &vop_close_desc, (void *) ufsspec_close },
2274 { &vop_getattr_desc, (void *) ufs_getattr },
2275 { &vop_inactive_desc, (void *) ufs_inactive },
2276 { &vop_islocked_desc, (void *) vop_stdislocked },
2277 { &vop_lock_desc, (void *) vop_stdlock },
2278 { &vop_print_desc, (void *) ufs_print },
2279 { &vop_read_desc, (void *) ufsspec_read },
2280 { &vop_reclaim_desc, (void *) ufs_reclaim },
2281 { &vop_setattr_desc, (void *) ufs_setattr },
2282 { &vop_unlock_desc, (void *) vop_stdunlock },
2283 { &vop_write_desc, (void *) ufsspec_write },
984263bc
MD
2284 { NULL, NULL }
2285};
2286static struct vnodeopv_desc ufs_specop_opv_desc =
2d3e977e 2287 { &ufs_spec_vops, ufs_specop_entries };
984263bc 2288
2d3e977e 2289static struct vop_ops *ufs_fifo_vops;
984263bc 2290static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2d3e977e
MD
2291 { &vop_default_desc, (void *) fifo_vnoperate },
2292 { &vop_fsync_desc, (void *) ufs_missingop },
2293 { &vop_access_desc, (void *) ufs_access },
2294 { &vop_close_desc, (void *) ufsfifo_close },
2295 { &vop_getattr_desc, (void *) ufs_getattr },
2296 { &vop_inactive_desc, (void *) ufs_inactive },
2297 { &vop_islocked_desc, (void *) vop_stdislocked },
2298 { &vop_kqfilter_desc, (void *) ufsfifo_kqfilter },
2299 { &vop_lock_desc, (void *) vop_stdlock },
2300 { &vop_print_desc, (void *) ufs_print },
2301 { &vop_read_desc, (void *) ufsfifo_read },
2302 { &vop_reclaim_desc, (void *) ufs_reclaim },
2303 { &vop_setattr_desc, (void *) ufs_setattr },
2304 { &vop_unlock_desc, (void *) vop_stdunlock },
2305 { &vop_write_desc, (void *) ufsfifo_write },
984263bc
MD
2306 { NULL, NULL }
2307};
2308static struct vnodeopv_desc ufs_fifoop_opv_desc =
2d3e977e 2309 { &ufs_fifo_vops, ufs_fifoop_entries };
984263bc
MD
2310
2311VNODEOP_SET(ufs_vnodeop_opv_desc);
2312VNODEOP_SET(ufs_specop_opv_desc);
2313VNODEOP_SET(ufs_fifoop_opv_desc);
2314
0973c589
CP
2315/*
2316 * ufs_vnoperate(struct vnodeop_desc *a_desc)
2317 */
984263bc 2318int
0973c589 2319ufs_vnoperate(struct vop_generic_args *ap)
984263bc 2320{
0961aa92 2321 return (VOCALL(ufs_vnode_vops, ap));
984263bc
MD
2322}
2323
0973c589
CP
2324/*
2325 * ufs_vnoperatefifo(struct vnodeop_desc *a_desc)
2326 */
984263bc 2327int
0973c589 2328ufs_vnoperatefifo(struct vop_generic_args *ap)
984263bc 2329{
0961aa92 2330 return (VOCALL(ufs_fifo_vops, ap));
984263bc
MD
2331}
2332
0973c589
CP
2333/*
2334 * ufs_vnoperatespec(struct vnodeop_desc *a_desc)
2335 */
984263bc 2336int
0973c589 2337ufs_vnoperatespec(struct vop_generic_args *ap)
984263bc 2338{
0961aa92 2339 return (VOCALL(ufs_spec_vops, ap));
984263bc 2340}