M_EXT_CLUSTER was not being properly inherited in m_copym(), m_copypacket(),
[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 $
8bb5da37 40 * $DragonFly: src/sys/vfs/ufs/ufs_vnops.c,v 1.18 2004/08/24 13:52:59 drhodus 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 */
238 vput(*vpp);
239 (*vpp)->v_type = VNON;
240 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
241 vgone(*vpp);
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;
41a01a4d 286 lwkt_tokref vlock;
984263bc 287
41a01a4d 288 lwkt_gettoken(&vlock, vp->v_interlock);
984263bc
MD
289 if (vp->v_usecount > 1)
290 ufs_itimes(vp);
41a01a4d 291 lwkt_reltoken(&vlock);
984263bc
MD
292 return (0);
293}
294
0973c589
CP
295/*
296 * ufs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
297 * struct thread *a_td)
298 */
0961aa92 299static
984263bc 300int
0973c589 301ufs_access(struct vop_access_args *ap)
984263bc
MD
302{
303 struct vnode *vp = ap->a_vp;
304 struct inode *ip = VTOI(vp);
305 struct ucred *cred = ap->a_cred;
306 mode_t mask, mode = ap->a_mode;
3ff2135f 307 gid_t *gp;
984263bc
MD
308 int i;
309#ifdef QUOTA
310 int error;
311#endif
312
313 /*
f719c866 314 * Disallow write attempts on read-only filesystems;
984263bc 315 * unless the file is a socket, fifo, or a block or
f719c866 316 * character device resident on the filesystem.
984263bc
MD
317 */
318 if (mode & VWRITE) {
319 switch (vp->v_type) {
320 case VDIR:
321 case VLNK:
322 case VREG:
323 if (vp->v_mount->mnt_flag & MNT_RDONLY)
324 return (EROFS);
325#ifdef QUOTA
326 if ((error = getinoquota(ip)) != 0)
327 return (error);
328#endif
329 break;
330 default:
331 break;
332 }
333 }
334
335 /* If immutable bit set, nobody gets to write it. */
336 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
337 return (EPERM);
338
339 /* Otherwise, user id 0 always gets access. */
340 if (cred->cr_uid == 0)
341 return (0);
342
343 mask = 0;
344
345 /* Otherwise, check the owner. */
346 if (cred->cr_uid == ip->i_uid) {
347 if (mode & VEXEC)
348 mask |= S_IXUSR;
349 if (mode & VREAD)
350 mask |= S_IRUSR;
351 if (mode & VWRITE)
352 mask |= S_IWUSR;
353 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
354 }
355
356 /* Otherwise, check the groups. */
357 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
358 if (ip->i_gid == *gp) {
359 if (mode & VEXEC)
360 mask |= S_IXGRP;
361 if (mode & VREAD)
362 mask |= S_IRGRP;
363 if (mode & VWRITE)
364 mask |= S_IWGRP;
365 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
366 }
367
368 /* Otherwise, check everyone else. */
369 if (mode & VEXEC)
370 mask |= S_IXOTH;
371 if (mode & VREAD)
372 mask |= S_IROTH;
373 if (mode & VWRITE)
374 mask |= S_IWOTH;
375 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
376}
377
0973c589
CP
378/*
379 * ufs_getattr(struct vnode *a_vp, struct vattr *a_vap,
380 * struct thread *a_td)
381 */
984263bc 382/* ARGSUSED */
0961aa92 383static
984263bc 384int
0973c589 385ufs_getattr(struct vop_getattr_args *ap)
984263bc 386{
3ff2135f
RG
387 struct vnode *vp = ap->a_vp;
388 struct inode *ip = VTOI(vp);
389 struct vattr *vap = ap->a_vap;
984263bc
MD
390
391 ufs_itimes(vp);
392 /*
393 * Copy from inode table
394 */
395 vap->va_fsid = dev2udev(ip->i_dev);
396 vap->va_fileid = ip->i_number;
397 vap->va_mode = ip->i_mode & ~IFMT;
398 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ?
399 ip->i_effnlink : ip->i_nlink;
400 vap->va_uid = ip->i_uid;
401 vap->va_gid = ip->i_gid;
402 vap->va_rdev = ip->i_rdev;
403 vap->va_size = ip->i_din.di_size;
404 vap->va_atime.tv_sec = ip->i_atime;
405 vap->va_atime.tv_nsec = ip->i_atimensec;
406 vap->va_mtime.tv_sec = ip->i_mtime;
407 vap->va_mtime.tv_nsec = ip->i_mtimensec;
408 vap->va_ctime.tv_sec = ip->i_ctime;
409 vap->va_ctime.tv_nsec = ip->i_ctimensec;
410 vap->va_flags = ip->i_flags;
411 vap->va_gen = ip->i_gen;
412 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
413 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
414 vap->va_type = IFTOVT(ip->i_mode);
415 vap->va_filerev = ip->i_modrev;
416 return (0);
417}
418
419/*
420 * Set attribute vnode op. called from several syscalls
0973c589
CP
421 *
422 * ufs_setattr(struct vnode *a_vp, struct vattr *a_vap,
423 * struct ucred *a_cred, struct thread *a_td)
984263bc 424 */
0961aa92 425static
984263bc 426int
0973c589 427ufs_setattr(struct vop_setattr_args *ap)
984263bc
MD
428{
429 struct vattr *vap = ap->a_vap;
430 struct vnode *vp = ap->a_vp;
431 struct inode *ip = VTOI(vp);
432 struct ucred *cred = ap->a_cred;
984263bc
MD
433 int error;
434
435 /*
436 * Check for unsettable attributes.
437 */
438 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
439 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
440 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
441 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
442 return (EINVAL);
443 }
444 if (vap->va_flags != VNOVAL) {
445 if (vp->v_mount->mnt_flag & MNT_RDONLY)
446 return (EROFS);
447 if (cred->cr_uid != ip->i_uid &&
dadab5e9 448 (error = suser_cred(cred, PRISON_ROOT)))
984263bc 449 return (error);
41c20dac 450 if ((cred->cr_uid == 0) && (cred->cr_prison == NULL)) {
984263bc
MD
451 if ((ip->i_flags
452 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
453 securelevel > 0)
454 return (EPERM);
455 ip->i_flags = vap->va_flags;
456 } else {
457 if (ip->i_flags
458 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
459 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
460 return (EPERM);
461 ip->i_flags &= SF_SETTABLE;
462 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
463 }
464 ip->i_flag |= IN_CHANGE;
465 if (vap->va_flags & (IMMUTABLE | APPEND))
466 return (0);
467 }
468 if (ip->i_flags & (IMMUTABLE | APPEND))
469 return (EPERM);
470 /*
471 * Go through the fields and update iff not VNOVAL.
472 */
473 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
474 if (vp->v_mount->mnt_flag & MNT_RDONLY)
475 return (EROFS);
dadab5e9 476 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
984263bc
MD
477 return (error);
478 }
479 if (vap->va_size != VNOVAL) {
480 /*
f719c866 481 * Disallow write attempts on read-only filesystems;
984263bc 482 * unless the file is a socket, fifo, or a block or
f719c866 483 * character device resident on the filesystem.
984263bc
MD
484 */
485 switch (vp->v_type) {
486 case VDIR:
487 return (EISDIR);
488 case VLNK:
489 case VREG:
490 if (vp->v_mount->mnt_flag & MNT_RDONLY)
491 return (EROFS);
492 break;
493 default:
494 break;
495 }
dadab5e9 496 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
984263bc
MD
497 return (error);
498 }
499 ip = VTOI(vp);
500 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
501 if (vp->v_mount->mnt_flag & MNT_RDONLY)
502 return (EROFS);
503 if (cred->cr_uid != ip->i_uid &&
dadab5e9 504 (error = suser_cred(cred, PRISON_ROOT)) &&
984263bc 505 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
dadab5e9 506 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
984263bc
MD
507 return (error);
508 if (vap->va_atime.tv_sec != VNOVAL)
509 ip->i_flag |= IN_ACCESS;
510 if (vap->va_mtime.tv_sec != VNOVAL)
511 ip->i_flag |= IN_CHANGE | IN_UPDATE;
512 ufs_itimes(vp);
513 if (vap->va_atime.tv_sec != VNOVAL) {
514 ip->i_atime = vap->va_atime.tv_sec;
515 ip->i_atimensec = vap->va_atime.tv_nsec;
516 }
517 if (vap->va_mtime.tv_sec != VNOVAL) {
518 ip->i_mtime = vap->va_mtime.tv_sec;
519 ip->i_mtimensec = vap->va_mtime.tv_nsec;
520 }
521 error = UFS_UPDATE(vp, 0);
522 if (error)
523 return (error);
524 }
525 error = 0;
526 if (vap->va_mode != (mode_t)VNOVAL) {
527 if (vp->v_mount->mnt_flag & MNT_RDONLY)
528 return (EROFS);
dadab5e9 529 error = ufs_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
984263bc
MD
530 }
531 VN_KNOTE(vp, NOTE_ATTRIB);
532 return (error);
533}
534
535/*
536 * Change the mode on a file.
537 * Inode must be locked before calling.
538 */
539static int
dadab5e9 540ufs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
984263bc 541{
3ff2135f 542 struct inode *ip = VTOI(vp);
984263bc
MD
543 int error;
544
545 if (cred->cr_uid != ip->i_uid) {
dadab5e9 546 error = suser_cred(cred, PRISON_ROOT);
984263bc
MD
547 if (error)
548 return (error);
549 }
550 if (cred->cr_uid) {
551 if (vp->v_type != VDIR && (mode & S_ISTXT))
552 return (EFTYPE);
553 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
554 return (EPERM);
555 }
556 ip->i_mode &= ~ALLPERMS;
557 ip->i_mode |= (mode & ALLPERMS);
558 ip->i_flag |= IN_CHANGE;
559 return (0);
560}
561
562/*
563 * Perform chown operation on inode ip;
564 * inode must be locked prior to call.
565 */
566static int
0973c589
CP
567ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
568 struct thread *td)
984263bc 569{
3ff2135f 570 struct inode *ip = VTOI(vp);
984263bc
MD
571 uid_t ouid;
572 gid_t ogid;
573 int error = 0;
574#ifdef QUOTA
3ff2135f 575 int i;
984263bc
MD
576 long change;
577#endif
578
579 if (uid == (uid_t)VNOVAL)
580 uid = ip->i_uid;
581 if (gid == (gid_t)VNOVAL)
582 gid = ip->i_gid;
583 /*
584 * If we don't own the file, are trying to change the owner
585 * of the file, or are not a member of the target group,
586 * the caller must be superuser or the call fails.
587 */
588 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
a10e3626
DR
589 (gid != ip->i_gid && !(cred->cr_gid == gid ||
590 groupmember((gid_t)gid, cred)))) &&
dadab5e9 591 (error = suser_cred(cred, PRISON_ROOT)))
984263bc
MD
592 return (error);
593 ogid = ip->i_gid;
594 ouid = ip->i_uid;
595#ifdef QUOTA
596 if ((error = getinoquota(ip)) != 0)
597 return (error);
598 if (ouid == uid) {
599 dqrele(vp, ip->i_dquot[USRQUOTA]);
600 ip->i_dquot[USRQUOTA] = NODQUOT;
601 }
602 if (ogid == gid) {
603 dqrele(vp, ip->i_dquot[GRPQUOTA]);
604 ip->i_dquot[GRPQUOTA] = NODQUOT;
605 }
606 change = ip->i_blocks;
607 (void) chkdq(ip, -change, cred, CHOWN);
608 (void) chkiq(ip, -1, cred, CHOWN);
609 for (i = 0; i < MAXQUOTAS; i++) {
610 dqrele(vp, ip->i_dquot[i]);
611 ip->i_dquot[i] = NODQUOT;
612 }
613#endif
614 ip->i_gid = gid;
615 ip->i_uid = uid;
616#ifdef QUOTA
617 if ((error = getinoquota(ip)) == 0) {
618 if (ouid == uid) {
619 dqrele(vp, ip->i_dquot[USRQUOTA]);
620 ip->i_dquot[USRQUOTA] = NODQUOT;
621 }
622 if (ogid == gid) {
623 dqrele(vp, ip->i_dquot[GRPQUOTA]);
624 ip->i_dquot[GRPQUOTA] = NODQUOT;
625 }
626 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
627 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
628 goto good;
629 else
630 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
631 }
632 for (i = 0; i < MAXQUOTAS; i++) {
633 dqrele(vp, ip->i_dquot[i]);
634 ip->i_dquot[i] = NODQUOT;
635 }
636 }
637 ip->i_gid = ogid;
638 ip->i_uid = ouid;
639 if (getinoquota(ip) == 0) {
640 if (ouid == uid) {
641 dqrele(vp, ip->i_dquot[USRQUOTA]);
642 ip->i_dquot[USRQUOTA] = NODQUOT;
643 }
644 if (ogid == gid) {
645 dqrele(vp, ip->i_dquot[GRPQUOTA]);
646 ip->i_dquot[GRPQUOTA] = NODQUOT;
647 }
648 (void) chkdq(ip, change, cred, FORCE|CHOWN);
649 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
650 (void) getinoquota(ip);
651 }
652 return (error);
653good:
654 if (getinoquota(ip))
655 panic("ufs_chown: lost quota");
656#endif /* QUOTA */
657 ip->i_flag |= IN_CHANGE;
658 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
659 ip->i_mode &= ~(ISUID | ISGID);
660 return (0);
661}
662
663/*
664 * Mmap a file
665 *
666 * NB Currently unsupported.
0973c589
CP
667 *
668 * ufs_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
669 * struct thread *a_td)
984263bc
MD
670 */
671/* ARGSUSED */
0961aa92 672static
984263bc 673int
0973c589 674ufs_mmap(struct vop_mmap_args *ap)
984263bc 675{
984263bc
MD
676 return (EINVAL);
677}
678
0973c589
CP
679/*
680 * ufs_remove(struct vnode *a_dvp, struct vnode *a_vp,
681 * struct componentname *a_cnp)
682 */
0961aa92 683static
984263bc 684int
0973c589 685ufs_remove(struct vop_remove_args *ap)
984263bc
MD
686{
687 struct inode *ip;
688 struct vnode *vp = ap->a_vp;
689 struct vnode *dvp = ap->a_dvp;
690 int error;
691
692 ip = VTOI(vp);
693 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
694 (VTOI(dvp)->i_flags & APPEND)) {
695 error = EPERM;
696 goto out;
697 }
698 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
699 VN_KNOTE(vp, NOTE_DELETE);
700 VN_KNOTE(dvp, NOTE_WRITE);
701out:
702 return (error);
703}
704
705/*
706 * link vnode call
0973c589
CP
707 *
708 * ufs_link(struct vnode *a_tdvp, struct vnode *a_vp,
709 * struct componentname *a_cnp)
984263bc 710 */
0961aa92 711static
984263bc 712int
0973c589 713ufs_link(struct vop_link_args *ap)
984263bc
MD
714{
715 struct vnode *vp = ap->a_vp;
716 struct vnode *tdvp = ap->a_tdvp;
717 struct componentname *cnp = ap->a_cnp;
dadab5e9 718 struct thread *td = cnp->cn_td;
984263bc
MD
719 struct inode *ip;
720 struct direct newdir;
721 int error;
722
723#ifdef DIAGNOSTIC
2b69e610 724 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
725 panic("ufs_link: no name");
726#endif
727 if (tdvp->v_mount != vp->v_mount) {
728 error = EXDEV;
729 goto out2;
730 }
41a01a4d 731 if (tdvp != vp && (error = vn_lock(vp, NULL, LK_EXCLUSIVE, td))) {
984263bc
MD
732 goto out2;
733 }
734 ip = VTOI(vp);
735 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
736 error = EMLINK;
737 goto out1;
738 }
739 if (ip->i_flags & (IMMUTABLE | APPEND)) {
740 error = EPERM;
741 goto out1;
742 }
743 ip->i_effnlink++;
744 ip->i_nlink++;
745 ip->i_flag |= IN_CHANGE;
746 if (DOINGSOFTDEP(vp))
747 softdep_change_linkcnt(ip);
748 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
749 if (!error) {
750 ufs_makedirentry(ip, cnp, &newdir);
751 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
752 }
753
754 if (error) {
755 ip->i_effnlink--;
756 ip->i_nlink--;
757 ip->i_flag |= IN_CHANGE;
758 if (DOINGSOFTDEP(vp))
759 softdep_change_linkcnt(ip);
760 }
761out1:
762 if (tdvp != vp)
41a01a4d 763 VOP_UNLOCK(vp, NULL, 0, td);
984263bc
MD
764out2:
765 VN_KNOTE(vp, NOTE_LINK);
766 VN_KNOTE(tdvp, NOTE_WRITE);
767 return (error);
768}
769
770/*
771 * whiteout vnode call
0973c589
CP
772 *
773 * ufs_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
984263bc 774 */
0961aa92 775static
984263bc 776int
0973c589 777ufs_whiteout(struct vop_whiteout_args *ap)
984263bc
MD
778{
779 struct vnode *dvp = ap->a_dvp;
780 struct componentname *cnp = ap->a_cnp;
781 struct direct newdir;
782 int error = 0;
783
784 switch (ap->a_flags) {
2b69e610 785 case NAMEI_LOOKUP:
984263bc
MD
786 /* 4.4 format directories support whiteout operations */
787 if (dvp->v_mount->mnt_maxsymlinklen > 0)
788 return (0);
789 return (EOPNOTSUPP);
790
2b69e610 791 case NAMEI_CREATE:
984263bc
MD
792 /* create a new directory whiteout */
793#ifdef DIAGNOSTIC
2b69e610 794 if ((cnp->cn_flags & CNP_SAVENAME) == 0)
984263bc
MD
795 panic("ufs_whiteout: missing name");
796 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
797 panic("ufs_whiteout: old format filesystem");
798#endif
799
800 newdir.d_ino = WINO;
801 newdir.d_namlen = cnp->cn_namelen;
802 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
803 newdir.d_type = DT_WHT;
804 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
805 break;
806
2b69e610 807 case NAMEI_DELETE:
984263bc
MD
808 /* remove an existing directory whiteout */
809#ifdef DIAGNOSTIC
810 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
811 panic("ufs_whiteout: old format filesystem");
812#endif
813
2b69e610 814 cnp->cn_flags &= ~CNP_DOWHITEOUT;
984263bc
MD
815 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
816 break;
817 default:
818 panic("ufs_whiteout: unknown op");
819 }
820 return (error);
821}
822
823/*
824 * Rename system call.
825 * rename("foo", "bar");
826 * is essentially
827 * unlink("bar");
828 * link("foo", "bar");
829 * unlink("foo");
830 * but ``atomically''. Can't do full commit without saving state in the
831 * inode on disk which isn't feasible at this time. Best we can do is
832 * always guarantee the target exists.
833 *
834 * Basic algorithm is:
835 *
836 * 1) Bump link count on source while we're linking it to the
837 * target. This also ensure the inode won't be deleted out
838 * from underneath us while we work (it may be truncated by
839 * a concurrent `trunc' or `open' for creation).
840 * 2) Link source to destination. If destination already exists,
841 * delete it first.
842 * 3) Unlink source reference to inode if still around. If a
843 * directory was moved and the parent of the destination
844 * is different from the source, patch the ".." entry in the
845 * directory.
0973c589
CP
846 *
847 * ufs_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
848 * struct componentname *a_fcnp, struct vnode *a_tdvp,
849 * struct vnode *a_tvp, struct componentname *a_tcnp)
984263bc 850 */
0961aa92 851static
984263bc 852int
0973c589 853ufs_rename(struct vop_rename_args *ap)
984263bc
MD
854{
855 struct vnode *tvp = ap->a_tvp;
3ff2135f 856 struct vnode *tdvp = ap->a_tdvp;
984263bc
MD
857 struct vnode *fvp = ap->a_fvp;
858 struct vnode *fdvp = ap->a_fdvp;
859 struct componentname *tcnp = ap->a_tcnp;
860 struct componentname *fcnp = ap->a_fcnp;
dadab5e9 861 struct thread *td = fcnp->cn_td;
984263bc
MD
862 struct inode *ip, *xp, *dp;
863 struct direct newdir;
8bb5da37 864 ino_t doingdirectory = 0, oldparent = 0, newparent = 0;
984263bc
MD
865 int error = 0, ioflag;
866
867#ifdef DIAGNOSTIC
2b69e610
MD
868 if ((tcnp->cn_flags & CNP_HASBUF) == 0 ||
869 (fcnp->cn_flags & CNP_HASBUF) == 0)
984263bc 870 panic("ufs_rename: no name");
f719c866 871#endif /* DIAGNOSTIC */
984263bc
MD
872 /*
873 * Check for cross-device rename.
874 */
875 if ((fvp->v_mount != tdvp->v_mount) ||
876 (tvp && (fvp->v_mount != tvp->v_mount))) {
877 error = EXDEV;
878abortit:
879 if (tdvp == tvp)
880 vrele(tdvp);
881 else
882 vput(tdvp);
883 if (tvp)
884 vput(tvp);
885 vrele(fdvp);
886 vrele(fvp);
887 return (error);
888 }
889
890 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
891 (VTOI(tdvp)->i_flags & APPEND))) {
892 error = EPERM;
893 goto abortit;
894 }
895
896 /*
897 * Renaming a file to itself has no effect. The upper layers should
898 * not call us in that case. Temporarily just warn if they do.
899 */
900 if (fvp == tvp) {
901 printf("ufs_rename: fvp == tvp (can't happen)\n");
902 error = 0;
903 goto abortit;
904 }
905
41a01a4d 906 if ((error = vn_lock(fvp, NULL, LK_EXCLUSIVE, td)) != 0)
984263bc
MD
907 goto abortit;
908 dp = VTOI(fdvp);
909 ip = VTOI(fvp);
910 if (ip->i_nlink >= LINK_MAX) {
41a01a4d 911 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
912 error = EMLINK;
913 goto abortit;
914 }
915 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
916 || (dp->i_flags & APPEND)) {
41a01a4d 917 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
918 error = EPERM;
919 goto abortit;
920 }
921 if ((ip->i_mode & IFMT) == IFDIR) {
922 /*
923 * Avoid ".", "..", and aliases of "." for obvious reasons.
924 */
925 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
2b69e610 926 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
984263bc 927 (ip->i_flag & IN_RENAME)) {
41a01a4d 928 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
929 error = EINVAL;
930 goto abortit;
931 }
932 ip->i_flag |= IN_RENAME;
933 oldparent = dp->i_number;
934 doingdirectory = 1;
935 }
936 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
937 vrele(fdvp);
938
939 /*
940 * When the target exists, both the directory
941 * and target vnodes are returned locked.
942 */
943 dp = VTOI(tdvp);
944 xp = NULL;
945 if (tvp)
946 xp = VTOI(tvp);
947
948 /*
949 * 1) Bump link count while we're moving stuff
950 * around. If we crash somewhere before
951 * completing our work, the link count
952 * may be wrong, but correctable.
953 */
954 ip->i_effnlink++;
955 ip->i_nlink++;
956 ip->i_flag |= IN_CHANGE;
957 if (DOINGSOFTDEP(fvp))
958 softdep_change_linkcnt(ip);
959 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
960 DOINGASYNC(fvp)))) != 0) {
41a01a4d 961 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
962 goto bad;
963 }
964
965 /*
966 * If ".." must be changed (ie the directory gets a new
967 * parent) then the source directory must not be in the
968 * directory heirarchy above the target, as this would
969 * orphan everything below the source directory. Also
970 * the user must have write permission in the source so
971 * as to be able to change "..". We must repeat the call
972 * to namei, as the parent directory is unlocked by the
973 * call to checkpath().
974 */
dadab5e9 975 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
41a01a4d 976 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
977 if (oldparent != dp->i_number)
978 newparent = dp->i_number;
979 if (doingdirectory && newparent) {
980 if (error) /* write access check above */
981 goto bad;
982 if (xp != NULL)
983 vput(tvp);
984 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
985 if (error)
986 goto out;
2b69e610 987 if ((tcnp->cn_flags & CNP_SAVESTART) == 0)
984263bc 988 panic("ufs_rename: lost to startdir");
597aea93 989 vref(tdvp);
984263bc
MD
990 error = relookup(tdvp, &tvp, tcnp);
991 if (error)
992 goto out;
993 vrele(tdvp);
994 dp = VTOI(tdvp);
995 xp = NULL;
996 if (tvp)
997 xp = VTOI(tvp);
998 }
999 /*
1000 * 2) If target doesn't exist, link the target
1001 * to the source and unlink the source.
1002 * Otherwise, rewrite the target directory
1003 * entry to reference the source inode and
1004 * expunge the original entry's existence.
1005 */
1006 if (xp == NULL) {
1007 if (dp->i_dev != ip->i_dev)
1008 panic("ufs_rename: EXDEV");
1009 /*
1010 * Account for ".." in new directory.
1011 * When source and destination have the same
1012 * parent we don't fool with the link count.
1013 */
1014 if (doingdirectory && newparent) {
1015 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1016 error = EMLINK;
1017 goto bad;
1018 }
1019 dp->i_effnlink++;
1020 dp->i_nlink++;
1021 dp->i_flag |= IN_CHANGE;
1022 if (DOINGSOFTDEP(tdvp))
1023 softdep_change_linkcnt(dp);
1024 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1025 DOINGASYNC(tdvp)));
1026 if (error)
1027 goto bad;
1028 }
1029 ufs_makedirentry(ip, tcnp, &newdir);
1030 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1031 if (error) {
1032 if (doingdirectory && newparent) {
1033 dp->i_effnlink--;
1034 dp->i_nlink--;
1035 dp->i_flag |= IN_CHANGE;
1036 if (DOINGSOFTDEP(tdvp))
1037 softdep_change_linkcnt(dp);
1038 (void)UFS_UPDATE(tdvp, 1);
1039 }
1040 goto bad;
1041 }
1042 VN_KNOTE(tdvp, NOTE_WRITE);
1043 vput(tdvp);
1044 } else {
1045 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1046 panic("ufs_rename: EXDEV");
1047 /*
1048 * Short circuit rename(foo, foo).
1049 */
1050 if (xp->i_number == ip->i_number)
1051 panic("ufs_rename: same file");
1052 /*
1053 * If the parent directory is "sticky", then the user must
1054 * own the parent directory, or the destination of the rename,
1055 * otherwise the destination may not be changed (except by
1056 * root). This implements append-only directories.
1057 */
1058 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1059 tcnp->cn_cred->cr_uid != dp->i_uid &&
1060 xp->i_uid != tcnp->cn_cred->cr_uid) {
1061 error = EPERM;
1062 goto bad;
1063 }
1064 /*
1065 * Target must be empty if a directory and have no links
1066 * to it. Also, ensure source and target are compatible
1067 * (both directories, or both not directories).
1068 */
1069 if ((xp->i_mode&IFMT) == IFDIR) {
1070 if ((xp->i_effnlink > 2) ||
1071 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1072 error = ENOTEMPTY;
1073 goto bad;
1074 }
1075 if (!doingdirectory) {
1076 error = ENOTDIR;
1077 goto bad;
1078 }
1079 cache_purge(tdvp);
1080 } else if (doingdirectory) {
1081 error = EISDIR;
1082 goto bad;
1083 }
1084 error = ufs_dirrewrite(dp, xp, ip->i_number,
1085 IFTODT(ip->i_mode),
1086 (doingdirectory && newparent) ? newparent : doingdirectory);
1087 if (error)
1088 goto bad;
1089 if (doingdirectory) {
1090 if (!newparent) {
1091 dp->i_effnlink--;
1092 if (DOINGSOFTDEP(tdvp))
1093 softdep_change_linkcnt(dp);
1094 }
1095 xp->i_effnlink--;
1096 if (DOINGSOFTDEP(tvp))
1097 softdep_change_linkcnt(xp);
1098 }
1099 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1100 /*
1101 * Truncate inode. The only stuff left in the directory
1102 * is "." and "..". The "." reference is inconsequential
1103 * since we are quashing it. We have removed the "."
1104 * reference and the reference in the parent directory,
1105 * but there may be other hard links. The soft
1106 * dependency code will arrange to do these operations
1107 * after the parent directory entry has been deleted on
1108 * disk, so when running with that code we avoid doing
1109 * them now.
1110 */
1111 if (!newparent) {
1112 dp->i_nlink--;
1113 dp->i_flag |= IN_CHANGE;
1114 }
1115 xp->i_nlink--;
1116 xp->i_flag |= IN_CHANGE;
1117 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC;
1118 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
dadab5e9 1119 tcnp->cn_cred, tcnp->cn_td)) != 0)
984263bc
MD
1120 goto bad;
1121 }
1122 VN_KNOTE(tdvp, NOTE_WRITE);
1123 vput(tdvp);
1124 VN_KNOTE(tvp, NOTE_DELETE);
1125 vput(tvp);
1126 xp = NULL;
1127 }
1128
1129 /*
1130 * 3) Unlink the source.
1131 */
2b69e610
MD
1132 fcnp->cn_flags &= ~CNP_MODMASK;
1133 fcnp->cn_flags |= CNP_LOCKPARENT | CNP_LOCKLEAF;
1134 if ((fcnp->cn_flags & CNP_SAVESTART) == 0)
984263bc 1135 panic("ufs_rename: lost from startdir");
597aea93 1136 vref(fdvp);
984263bc
MD
1137 error = relookup(fdvp, &fvp, fcnp);
1138 if (error == 0)
1139 vrele(fdvp);
1140 if (fvp != NULL) {
1141 xp = VTOI(fvp);
1142 dp = VTOI(fdvp);
1143 } else {
1144 /*
1145 * From name has disappeared.
1146 */
1147 if (doingdirectory)
1148 panic("ufs_rename: lost dir entry");
1149 vrele(ap->a_fvp);
1150 return (0);
1151 }
1152 /*
1153 * Ensure that the directory entry still exists and has not
1154 * changed while the new name has been entered. If the source is
1155 * a file then the entry may have been unlinked or renamed. In
1156 * either case there is no further work to be done. If the source
1157 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1158 * flag ensures that it cannot be moved by another rename or removed
1159 * by a rmdir.
1160 */
1161 if (xp != ip) {
1162 if (doingdirectory)
1163 panic("ufs_rename: lost dir entry");
1164 } else {
1165 /*
1166 * If the source is a directory with a
1167 * new parent, the link count of the old
1168 * parent directory must be decremented
1169 * and ".." set to point to the new parent.
1170 */
1171 if (doingdirectory && newparent) {
1172 xp->i_offset = mastertemplate.dot_reclen;
1173 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
1174 cache_purge(fdvp);
1175 }
1176 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1177 xp->i_flag &= ~IN_RENAME;
1178 }
1179 VN_KNOTE(fvp, NOTE_RENAME);
1180 if (dp)
1181 vput(fdvp);
1182 if (xp)
1183 vput(fvp);
1184 vrele(ap->a_fvp);
1185 return (error);
1186
1187bad:
1188 if (xp)
1189 vput(ITOV(xp));
1190 vput(ITOV(dp));
1191out:
1192 if (doingdirectory)
1193 ip->i_flag &= ~IN_RENAME;
41a01a4d 1194 if (vn_lock(fvp, NULL, LK_EXCLUSIVE, td) == 0) {
984263bc
MD
1195 ip->i_effnlink--;
1196 ip->i_nlink--;
1197 ip->i_flag |= IN_CHANGE;
1198 ip->i_flag &= ~IN_RENAME;
1199 if (DOINGSOFTDEP(fvp))
1200 softdep_change_linkcnt(ip);
1201 vput(fvp);
1202 } else
1203 vrele(fvp);
1204 return (error);
1205}
1206
1207/*
1208 * Mkdir system call
0973c589
CP
1209 *
1210 * ufs_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
1211 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc 1212 */
0961aa92 1213static
984263bc 1214int
0973c589 1215ufs_mkdir(struct vop_mkdir_args *ap)
984263bc 1216{
3ff2135f
RG
1217 struct vnode *dvp = ap->a_dvp;
1218 struct vattr *vap = ap->a_vap;
1219 struct componentname *cnp = ap->a_cnp;
1220 struct inode *ip, *dp;
984263bc
MD
1221 struct vnode *tvp;
1222 struct buf *bp;
1223 struct dirtemplate dirtemplate, *dtp;
1224 struct direct newdir;
1225 int error, dmode;
1226 long blkoff;
1227
1228#ifdef DIAGNOSTIC
2b69e610 1229 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
1230 panic("ufs_mkdir: no name");
1231#endif
1232 dp = VTOI(dvp);
1233 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1234 error = EMLINK;
1235 goto out;
1236 }
1237 dmode = vap->va_mode & 0777;
1238 dmode |= IFDIR;
1239 /*
1240 * Must simulate part of ufs_makeinode here to acquire the inode,
1241 * but not have it entered in the parent directory. The entry is
1242 * made later after writing "." and ".." entries.
1243 */
1244 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1245 if (error)
1246 goto out;
1247 ip = VTOI(tvp);
1248 ip->i_gid = dp->i_gid;
1249#ifdef SUIDDIR
1250 {
1251#ifdef QUOTA
1252 struct ucred ucred, *ucp;
1253 ucp = cnp->cn_cred;
1254#endif
1255 /*
1256 * If we are hacking owners here, (only do this where told to)
1257 * and we are not giving it TO root, (would subvert quotas)
1258 * then go ahead and give it to the other user.
1259 * The new directory also inherits the SUID bit.
1260 * If user's UID and dir UID are the same,
1261 * 'give it away' so that the SUID is still forced on.
1262 */
1263 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1264 (dp->i_mode & ISUID) && dp->i_uid) {
1265 dmode |= ISUID;
1266 ip->i_uid = dp->i_uid;
1267#ifdef QUOTA
1268 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1269 /*
1270 * Make sure the correct user gets charged
1271 * for the space.
1272 * Make a dummy credential for the victim.
1273 * XXX This seems to never be accessed out of
1274 * our context so a stack variable is ok.
1275 */
1276 ucred.cr_ref = 1;
1277 ucred.cr_uid = ip->i_uid;
1278 ucred.cr_ngroups = 1;
1279 ucred.cr_groups[0] = dp->i_gid;
1280 ucp = &ucred;
1281 }
1282#endif
1283 } else
1284 ip->i_uid = cnp->cn_cred->cr_uid;
1285#ifdef QUOTA
1286 if ((error = getinoquota(ip)) ||
1287 (error = chkiq(ip, 1, ucp, 0))) {
1288 UFS_VFREE(tvp, ip->i_number, dmode);
1289 vput(tvp);
1290 return (error);
1291 }
1292#endif
1293 }
1294#else /* !SUIDDIR */
1295 ip->i_uid = cnp->cn_cred->cr_uid;
1296#ifdef QUOTA
1297 if ((error = getinoquota(ip)) ||
1298 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1299 UFS_VFREE(tvp, ip->i_number, dmode);
1300 vput(tvp);
1301 return (error);
1302 }
1303#endif
1304#endif /* !SUIDDIR */
1305 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1306 ip->i_mode = dmode;
1307 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1308 ip->i_effnlink = 2;
1309 ip->i_nlink = 2;
1310 if (DOINGSOFTDEP(tvp))
1311 softdep_change_linkcnt(ip);
2b69e610 1312 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
1313 ip->i_flags |= UF_OPAQUE;
1314
1315 /*
1316 * Bump link count in parent directory to reflect work done below.
1317 * Should be done before reference is created so cleanup is
1318 * possible if we crash.
1319 */
1320 dp->i_effnlink++;
1321 dp->i_nlink++;
1322 dp->i_flag |= IN_CHANGE;
1323 if (DOINGSOFTDEP(dvp))
1324 softdep_change_linkcnt(dp);
1325 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1326 if (error)
1327 goto bad;
1328
1329 /*
1330 * Initialize directory with "." and ".." from static template.
1331 */
1332 if (dvp->v_mount->mnt_maxsymlinklen > 0
1333 )
1334 dtp = &mastertemplate;
1335 else
1336 dtp = (struct dirtemplate *)&omastertemplate;
1337 dirtemplate = *dtp;
1338 dirtemplate.dot_ino = ip->i_number;
1339 dirtemplate.dotdot_ino = dp->i_number;
1340 if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1341 B_CLRBUF, &bp)) != 0)
1342 goto bad;
1343 ip->i_size = DIRBLKSIZ;
1344 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1345 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1346 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1347 if (DOINGSOFTDEP(tvp)) {
1348 /*
1349 * Ensure that the entire newly allocated block is a
1350 * valid directory so that future growth within the
1351 * block does not have to ensure that the block is
1352 * written before the inode.
1353 */
1354 blkoff = DIRBLKSIZ;
1355 while (blkoff < bp->b_bcount) {
1356 ((struct direct *)
1357 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1358 blkoff += DIRBLKSIZ;
1359 }
1360 }
1361 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1362 DOINGASYNC(tvp)))) != 0) {
1363 (void)VOP_BWRITE(bp->b_vp, bp);
1364 goto bad;
1365 }
1366 /*
1367 * Directory set up, now install its entry in the parent directory.
1368 *
1369 * If we are not doing soft dependencies, then we must write out the
1370 * buffer containing the new directory body before entering the new
1371 * name in the parent. If we are doing soft dependencies, then the
1372 * buffer containing the new directory body will be passed to and
1373 * released in the soft dependency code after the code has attached
1374 * an appropriate ordering dependency to the buffer which ensures that
1375 * the buffer is written before the new name is written in the parent.
1376 */
1377 if (DOINGASYNC(dvp))
1378 bdwrite(bp);
1379 else if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp->b_vp, bp))))
1380 goto bad;
1381 ufs_makedirentry(ip, cnp, &newdir);
1382 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1383
1384bad:
1385 if (error == 0) {
1386 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1387 *ap->a_vpp = tvp;
1388 } else {
1389 dp->i_effnlink--;
1390 dp->i_nlink--;
1391 dp->i_flag |= IN_CHANGE;
1392 if (DOINGSOFTDEP(dvp))
1393 softdep_change_linkcnt(dp);
1394 /*
1395 * No need to do an explicit VOP_TRUNCATE here, vrele will
1396 * do this for us because we set the link count to 0.
1397 */
1398 ip->i_effnlink = 0;
1399 ip->i_nlink = 0;
1400 ip->i_flag |= IN_CHANGE;
1401 if (DOINGSOFTDEP(tvp))
1402 softdep_change_linkcnt(ip);
1403 vput(tvp);
1404 }
1405out:
1406 return (error);
1407}
1408
1409/*
1410 * Rmdir system call.
0973c589
CP
1411 *
1412 * ufs_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1413 * struct componentname *a_cnp)
984263bc 1414 */
0961aa92 1415static
984263bc 1416int
0973c589 1417ufs_rmdir(struct vop_rmdir_args *ap)
984263bc
MD
1418{
1419 struct vnode *vp = ap->a_vp;
1420 struct vnode *dvp = ap->a_dvp;
1421 struct componentname *cnp = ap->a_cnp;
1422 struct inode *ip, *dp;
1423 int error, ioflag;
1424
1425 ip = VTOI(vp);
1426 dp = VTOI(dvp);
1427
1428 /*
1429 * Do not remove a directory that is in the process of being renamed.
1430 * Verify the directory is empty (and valid). Rmdir ".." will not be
1431 * valid since ".." will contain a reference to the current directory
1432 * and thus be non-empty. Do not allow the removal of mounted on
1433 * directories (this can happen when an NFS exported filesystem
1434 * tries to remove a locally mounted on directory).
1435 */
1436 error = 0;
1437 if (ip->i_flag & IN_RENAME) {
1438 error = EINVAL;
1439 goto out;
1440 }
1441 if (ip->i_effnlink != 2 ||
1442 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1443 error = ENOTEMPTY;
1444 goto out;
1445 }
1446 if ((dp->i_flags & APPEND)
1447 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1448 error = EPERM;
1449 goto out;
1450 }
1451 if (vp->v_mountedhere != 0) {
1452 error = EINVAL;
1453 goto out;
1454 }
1455 /*
1456 * Delete reference to directory before purging
1457 * inode. If we crash in between, the directory
1458 * will be reattached to lost+found,
1459 */
1460 dp->i_effnlink--;
1461 ip->i_effnlink--;
1462 if (DOINGSOFTDEP(vp)) {
1463 softdep_change_linkcnt(dp);
1464 softdep_change_linkcnt(ip);
1465 }
1466 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1467 if (error) {
1468 dp->i_effnlink++;
1469 ip->i_effnlink++;
1470 if (DOINGSOFTDEP(vp)) {
1471 softdep_change_linkcnt(dp);
1472 softdep_change_linkcnt(ip);
1473 }
1474 goto out;
1475 }
1476 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1477 cache_purge(dvp);
1478 /*
1479 * Truncate inode. The only stuff left in the directory is "." and
1480 * "..". The "." reference is inconsequential since we are quashing
1481 * it. The soft dependency code will arrange to do these operations
1482 * after the parent directory entry has been deleted on disk, so
1483 * when running with that code we avoid doing them now.
1484 */
1485 if (!DOINGSOFTDEP(vp)) {
1486 dp->i_nlink--;
1487 dp->i_flag |= IN_CHANGE;
1488 ip->i_nlink--;
1489 ip->i_flag |= IN_CHANGE;
1490 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC;
1491 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
dadab5e9 1492 cnp->cn_td);
984263bc
MD
1493 }
1494 cache_purge(vp);
1495#ifdef UFS_DIRHASH
1496 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
1497 if (ip->i_dirhash != NULL)
1498 ufsdirhash_free(ip);
1499#endif
1500out:
1501 VN_KNOTE(vp, NOTE_DELETE);
1502 return (error);
1503}
1504
1505/*
1506 * symlink -- make a symbolic link
0973c589
CP
1507 *
1508 * ufs_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1509 * struct componentname *a_cnp, struct vattr *a_vap,
1510 * char *a_target)
984263bc 1511 */
0961aa92 1512static
984263bc 1513int
0973c589 1514ufs_symlink(struct vop_symlink_args *ap)
984263bc 1515{
3ff2135f
RG
1516 struct vnode *vp, **vpp = ap->a_vpp;
1517 struct inode *ip;
984263bc
MD
1518 int len, error;
1519
1520 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1521 vpp, ap->a_cnp);
1522 if (error)
1523 return (error);
1524 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1525 vp = *vpp;
1526 len = strlen(ap->a_target);
1527 if (len < vp->v_mount->mnt_maxsymlinklen) {
1528 ip = VTOI(vp);
1529 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1530 ip->i_size = len;
1531 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1532 } else
1533 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
dadab5e9
MD
1534 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred,
1535 (int *)0, NULL);
984263bc
MD
1536 if (error)
1537 vput(vp);
1538 return (error);
1539}
1540
1541/*
1542 * Vnode op for reading directories.
1543 *
1544 * The routine below assumes that the on-disk format of a directory
1545 * is the same as that defined by <sys/dirent.h>. If the on-disk
1546 * format changes, then it will be necessary to do a conversion
1547 * from the on-disk format that read returns to the format defined
1548 * by <sys/dirent.h>.
0973c589
CP
1549 *
1550 * ufs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
1551 * int *a_eofflag, int *ncookies, u_long **a_cookies)
984263bc 1552 */
0961aa92 1553static
984263bc 1554int
0973c589 1555ufs_readdir(struct vop_readdir_args *ap)
984263bc 1556{
3ff2135f 1557 struct uio *uio = ap->a_uio;
984263bc
MD
1558 int error;
1559 size_t count, lost;
1560 off_t off;
1561
1562 if (ap->a_ncookies != NULL)
1563 /*
1564 * Ensure that the block is aligned. The caller can use
1565 * the cookies to determine where in the block to start.
1566 */
1567 uio->uio_offset &= ~(DIRBLKSIZ - 1);
1568 off = uio->uio_offset;
1569 count = uio->uio_resid;
1570 /* Make sure we don't return partial entries. */
1571 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
1572 return (EINVAL);
1573 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1574 lost = uio->uio_resid - count;
1575 uio->uio_resid = count;
1576 uio->uio_iov->iov_len = count;
1577# if (BYTE_ORDER == LITTLE_ENDIAN)
1578 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
1579 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1580 } else {
1581 struct dirent *dp, *edp;
1582 struct uio auio;
1583 struct iovec aiov;
1584 caddr_t dirbuf;
1585 int readcnt;
1586 u_char tmp;
1587
1588 auio = *uio;
1589 auio.uio_iov = &aiov;
1590 auio.uio_iovcnt = 1;
1591 auio.uio_segflg = UIO_SYSSPACE;
1592 aiov.iov_len = count;
1593 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1594 aiov.iov_base = dirbuf;
1595 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
1596 if (error == 0) {
1597 readcnt = count - auio.uio_resid;
1598 edp = (struct dirent *)&dirbuf[readcnt];
1599 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
1600 tmp = dp->d_namlen;
1601 dp->d_namlen = dp->d_type;
1602 dp->d_type = tmp;
1603 if (dp->d_reclen > 0) {
1604 dp = (struct dirent *)
1605 ((char *)dp + dp->d_reclen);
1606 } else {
1607 error = EIO;
1608 break;
1609 }
1610 }
1611 if (dp >= edp)
1612 error = uiomove(dirbuf, readcnt, uio);
1613 }
1614 FREE(dirbuf, M_TEMP);
1615 }
1616# else
1617 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1618# endif
1619 if (!error && ap->a_ncookies != NULL) {
1620 struct dirent* dpStart;
1621 struct dirent* dpEnd;
1622 struct dirent* dp;
1623 int ncookies;
1624 u_long *cookies;
1625 u_long *cookiep;
1626
1627 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1628 panic("ufs_readdir: unexpected uio from NFS server");
1629 dpStart = (struct dirent *)
1630 (uio->uio_iov->iov_base - (uio->uio_offset - off));
1631 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
1632 for (dp = dpStart, ncookies = 0;
1633 dp < dpEnd;
1634 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
1635 ncookies++;
1636 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1637 M_WAITOK);
1638 for (dp = dpStart, cookiep = cookies;
1639 dp < dpEnd;
1640 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
1641 off += dp->d_reclen;
1642 *cookiep++ = (u_long) off;
1643 }
1644 *ap->a_ncookies = ncookies;
1645 *ap->a_cookies = cookies;
1646 }
1647 uio->uio_resid += lost;
1648 if (ap->a_eofflag)
1649 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
1650 return (error);
1651}
1652
1653/*
1654 * Return target name of a symbolic link
0973c589
CP
1655 *
1656 * ufs_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
984263bc 1657 */
0961aa92 1658static
984263bc 1659int
0973c589 1660ufs_readlink(struct vop_readlink_args *ap)
984263bc 1661{
3ff2135f
RG
1662 struct vnode *vp = ap->a_vp;
1663 struct inode *ip = VTOI(vp);
984263bc
MD
1664 int isize;
1665
1666 isize = ip->i_size;
1667 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
f719c866 1668 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
984263bc
MD
1669 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1670 return (0);
1671 }
1672 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1673}
1674
1675/*
1676 * Calculate the logical to physical mapping if not done already,
1677 * then call the device strategy routine.
1678 *
1679 * In order to be able to swap to a file, the VOP_BMAP operation may not
1680 * deadlock on memory. See ufs_bmap() for details.
0973c589
CP
1681 *
1682 * ufs_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc 1683 */
0961aa92 1684static
984263bc 1685int
0973c589 1686ufs_strategy(struct vop_strategy_args *ap)
984263bc 1687{
3ff2135f
RG
1688 struct buf *bp = ap->a_bp;
1689 struct vnode *vp = ap->a_vp;
1690 struct inode *ip;
984263bc
MD
1691 int error;
1692
1693 ip = VTOI(vp);
1694 if (vp->v_type == VBLK || vp->v_type == VCHR)
1695 panic("ufs_strategy: spec");
1696 if (bp->b_blkno == bp->b_lblkno) {
1697 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
1698 if (error) {
1699 bp->b_error = error;
1700 bp->b_flags |= B_ERROR;
1701 biodone(bp);
1702 return (error);
1703 }
8bb5da37 1704 if (bp->b_blkno == -1)
984263bc
MD
1705 vfs_bio_clrbuf(bp);
1706 }
8bb5da37 1707 if (bp->b_blkno == -1) {
984263bc
MD
1708 biodone(bp);
1709 return (0);
1710 }
1711 vp = ip->i_devvp;
1712 bp->b_dev = vp->v_rdev;
1713 VOP_STRATEGY(vp, bp);
1714 return (0);
1715}
1716
1717/*
1718 * Print out the contents of an inode.
0973c589
CP
1719 *
1720 * ufs_print(struct vnode *a_vp)
984263bc 1721 */
0961aa92 1722static
984263bc 1723int
0973c589 1724ufs_print(struct vop_print_args *ap)
984263bc 1725{
3ff2135f
RG
1726 struct vnode *vp = ap->a_vp;
1727 struct inode *ip = VTOI(vp);
984263bc
MD
1728
1729 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)",
1730 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1731 minor(ip->i_dev));
1732 if (vp->v_type == VFIFO)
1733 fifo_printinfo(vp);
1734 lockmgr_printinfo(&ip->i_lock);
1735 printf("\n");
1736 return (0);
1737}
1738
1739/*
1740 * Read wrapper for special devices.
0973c589
CP
1741 *
1742 * ufsspec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1743 * struct ucred *a_cred)
984263bc 1744 */
0961aa92 1745static
984263bc 1746int
0973c589 1747ufsspec_read(struct vop_read_args *ap)
984263bc
MD
1748{
1749 int error, resid;
1750 struct inode *ip;
1751 struct uio *uio;
1752
1753 uio = ap->a_uio;
1754 resid = uio->uio_resid;
0961aa92 1755 error = VOCALL(spec_vnode_vops, &ap->a_head);
984263bc
MD
1756 /*
1757 * The inode may have been revoked during the call, so it must not
1758 * be accessed blindly here or in the other wrapper functions.
1759 */
1760 ip = VTOI(ap->a_vp);
1761 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1762 ip->i_flag |= IN_ACCESS;
1763 return (error);
1764}
1765
1766/*
1767 * Write wrapper for special devices.
0973c589
CP
1768 *
1769 * ufsspec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1770 * struct ucred *a_cred)
984263bc 1771 */
0961aa92 1772static
984263bc 1773int
0973c589 1774ufsspec_write(struct vop_write_args *ap)
984263bc
MD
1775{
1776 int error, resid;
1777 struct inode *ip;
1778 struct uio *uio;
1779
1780 uio = ap->a_uio;
1781 resid = uio->uio_resid;
0961aa92 1782 error = VOCALL(spec_vnode_vops, &ap->a_head);
984263bc
MD
1783 ip = VTOI(ap->a_vp);
1784 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1785 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1786 return (error);
1787}
1788
1789/*
1790 * Close wrapper for special devices.
1791 *
1792 * Update the times on the inode then do device close.
0973c589
CP
1793 *
1794 * ufsspec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1795 * struct thread *a_td)
984263bc 1796 */
0961aa92 1797static
984263bc 1798int
0973c589 1799ufsspec_close(struct vop_close_args *ap)
984263bc
MD
1800{
1801 struct vnode *vp = ap->a_vp;
41a01a4d 1802 lwkt_tokref vlock;
984263bc 1803
41a01a4d 1804 lwkt_gettoken(&vlock, vp->v_interlock);
984263bc
MD
1805 if (vp->v_usecount > 1)
1806 ufs_itimes(vp);
41a01a4d 1807 lwkt_reltoken(&vlock);
0961aa92 1808 return (VOCALL(spec_vnode_vops, &ap->a_head));
984263bc
MD
1809}
1810
1811/*
1812 * Read wrapper for fifos.
0973c589
CP
1813 *
1814 * ufsfifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1815 * struct ucred *a_cred)
984263bc 1816 */
0961aa92 1817static
984263bc 1818int
0973c589 1819ufsfifo_read(struct vop_read_args *ap)
984263bc
MD
1820{
1821 int error, resid;
1822 struct inode *ip;
1823 struct uio *uio;
1824
1825 uio = ap->a_uio;
1826 resid = uio->uio_resid;
0961aa92 1827 error = VOCALL(fifo_vnode_vops, &ap->a_head);
984263bc
MD
1828 ip = VTOI(ap->a_vp);
1829 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1830 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1831 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1832 return (error);
1833}
1834
1835/*
1836 * Write wrapper for fifos.
0973c589
CP
1837 *
1838 * ufsfifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1839 * struct ucred *a_cred)
984263bc 1840 */
0961aa92 1841static
984263bc 1842int
0973c589 1843ufsfifo_write(struct vop_write_args *ap)
984263bc
MD
1844{
1845 int error, resid;
1846 struct inode *ip;
1847 struct uio *uio;
1848
1849 uio = ap->a_uio;
1850 resid = uio->uio_resid;
0961aa92 1851 error = VOCALL(fifo_vnode_vops, &ap->a_head);
984263bc
MD
1852 ip = VTOI(ap->a_vp);
1853 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1854 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1855 return (error);
1856}
1857
1858/*
1859 * Close wrapper for fifos.
1860 *
1861 * Update the times on the inode then do device close.
0973c589
CP
1862 *
1863 * ufsfifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1864 * struct thread *a_td)
984263bc 1865 */
0961aa92 1866static
984263bc 1867int
0973c589 1868ufsfifo_close(struct vop_close_args *ap)
984263bc
MD
1869{
1870 struct vnode *vp = ap->a_vp;
41a01a4d 1871 lwkt_tokref vlock;
984263bc 1872
41a01a4d 1873 lwkt_gettoken(&vlock, vp->v_interlock);
984263bc
MD
1874 if (vp->v_usecount > 1)
1875 ufs_itimes(vp);
41a01a4d 1876 lwkt_reltoken(&vlock);
0961aa92 1877 return (VOCALL(fifo_vnode_vops, &ap->a_head));
984263bc
MD
1878}
1879
1880/*
1881 * Kqfilter wrapper for fifos.
1882 *
1883 * Fall through to ufs kqfilter routines if needed
1884 */
0961aa92 1885static
984263bc 1886int
0973c589 1887ufsfifo_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
1888{
1889 int error;
1890
0961aa92 1891 error = VOCALL(fifo_vnode_vops, &ap->a_head);
984263bc
MD
1892 if (error)
1893 error = ufs_kqfilter(ap);
1894 return (error);
1895}
1896
1897/*
1898 * Return POSIX pathconf information applicable to ufs filesystems.
0973c589
CP
1899 *
1900 * ufs_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
984263bc 1901 */
0961aa92 1902static
984263bc 1903int
0973c589 1904ufs_pathconf(struct vop_pathconf_args *ap)
984263bc 1905{
984263bc
MD
1906 switch (ap->a_name) {
1907 case _PC_LINK_MAX:
1908 *ap->a_retval = LINK_MAX;
1909 return (0);
1910 case _PC_NAME_MAX:
1911 *ap->a_retval = NAME_MAX;
1912 return (0);
1913 case _PC_PATH_MAX:
1914 *ap->a_retval = PATH_MAX;
1915 return (0);
1916 case _PC_PIPE_BUF:
1917 *ap->a_retval = PIPE_BUF;
1918 return (0);
1919 case _PC_CHOWN_RESTRICTED:
1920 *ap->a_retval = 1;
1921 return (0);
1922 case _PC_NO_TRUNC:
1923 *ap->a_retval = 1;
1924 return (0);
1925 default:
1926 return (EINVAL);
1927 }
1928 /* NOTREACHED */
1929}
1930
1931/*
1932 * Advisory record locking support
0973c589
CP
1933 *
1934 * ufs_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
1935 * int a_flags)
984263bc 1936 */
0961aa92 1937static
984263bc 1938int
0973c589 1939ufs_advlock(struct vop_advlock_args *ap)
984263bc 1940{
3ff2135f 1941 struct inode *ip = VTOI(ap->a_vp);
984263bc
MD
1942
1943 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1944}
1945
1946/*
1947 * Initialize the vnode associated with a new inode, handle aliased
1948 * vnodes.
1949 */
1950int
0961aa92 1951ufs_vinit(struct mount *mntp, struct vnode **vpp)
984263bc
MD
1952{
1953 struct inode *ip;
1954 struct vnode *vp;
1955 struct timeval tv;
1956
1957 vp = *vpp;
1958 ip = VTOI(vp);
1959 switch(vp->v_type = IFTOVT(ip->i_mode)) {
1960 case VCHR:
1961 case VBLK:
0961aa92 1962 vp->v_ops = mntp->mnt_vn_spec_ops;
984263bc
MD
1963 addaliasu(vp, ip->i_rdev);
1964 break;
1965 case VFIFO:
0961aa92 1966 vp->v_ops = mntp->mnt_vn_fifo_ops;
984263bc
MD
1967 break;
1968 default:
1969 break;
1970
1971 }
1972 if (ip->i_number == ROOTINO)
1973 vp->v_flag |= VROOT;
1974 /*
1975 * Initialize modrev times
1976 */
1977 getmicrouptime(&tv);
1978 SETHIGH(ip->i_modrev, tv.tv_sec);
1979 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
1980 *vpp = vp;
1981 return (0);
1982}
1983
1984/*
1985 * Allocate a new inode.
1986 */
0961aa92 1987static
984263bc 1988int
0973c589
CP
1989ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1990 struct componentname *cnp)
984263bc 1991{
3ff2135f 1992 struct inode *ip, *pdir;
984263bc
MD
1993 struct direct newdir;
1994 struct vnode *tvp;
1995 int error;
1996
1997 pdir = VTOI(dvp);
1998#ifdef DIAGNOSTIC
2b69e610 1999 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
2000 panic("ufs_makeinode: no name");
2001#endif
2002 *vpp = NULL;
2003 if ((mode & IFMT) == 0)
2004 mode |= IFREG;
2005
2006 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
2007 if (error)
2008 return (error);
2009 ip = VTOI(tvp);
2010 ip->i_gid = pdir->i_gid;
2011#ifdef SUIDDIR
2012 {
2013#ifdef QUOTA
2014 struct ucred ucred, *ucp;
2015 ucp = cnp->cn_cred;
2016#endif
2017 /*
2018 * If we are not the owner of the directory,
2019 * and we are hacking owners here, (only do this where told to)
2020 * and we are not giving it TO root, (would subvert quotas)
2021 * then go ahead and give it to the other user.
2022 * Note that this drops off the execute bits for security.
2023 */
2024 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
2025 (pdir->i_mode & ISUID) &&
2026 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
2027 ip->i_uid = pdir->i_uid;
2028 mode &= ~07111;
2029#ifdef QUOTA
2030 /*
2031 * Make sure the correct user gets charged
2032 * for the space.
2033 * Quickly knock up a dummy credential for the victim.
2034 * XXX This seems to never be accessed out of our
2035 * context so a stack variable is ok.
2036 */
2037 ucred.cr_ref = 1;
2038 ucred.cr_uid = ip->i_uid;
2039 ucred.cr_ngroups = 1;
2040 ucred.cr_groups[0] = pdir->i_gid;
2041 ucp = &ucred;
2042#endif
2043 } else
2044 ip->i_uid = cnp->cn_cred->cr_uid;
2045
2046#ifdef QUOTA
2047 if ((error = getinoquota(ip)) ||
2048 (error = chkiq(ip, 1, ucp, 0))) {
2049 UFS_VFREE(tvp, ip->i_number, mode);
2050 vput(tvp);
2051 return (error);
2052 }
2053#endif
2054 }
2055#else /* !SUIDDIR */
2056 ip->i_uid = cnp->cn_cred->cr_uid;
2057#ifdef QUOTA
2058 if ((error = getinoquota(ip)) ||
2059 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2060 UFS_VFREE(tvp, ip->i_number, mode);
2061 vput(tvp);
2062 return (error);
2063 }
2064#endif
2065#endif /* !SUIDDIR */
2066 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2067 ip->i_mode = mode;
2068 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2069 ip->i_effnlink = 1;
2070 ip->i_nlink = 1;
2071 if (DOINGSOFTDEP(tvp))
2072 softdep_change_linkcnt(ip);
2073 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
dadab5e9 2074 suser_cred(cnp->cn_cred, 0)) {
984263bc 2075 ip->i_mode &= ~ISGID;
41c20dac 2076 }
984263bc 2077
2b69e610 2078 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
2079 ip->i_flags |= UF_OPAQUE;
2080
2081 /*
2082 * Make sure inode goes to disk before directory entry.
2083 */
2084 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2085 if (error)
2086 goto bad;
2087 ufs_makedirentry(ip, cnp, &newdir);
2088 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2089 if (error)
2090 goto bad;
2091 *vpp = tvp;
2092 return (0);
2093
2094bad:
2095 /*
2096 * Write error occurred trying to update the inode
2097 * or the directory so must deallocate the inode.
2098 */
2099 ip->i_effnlink = 0;
2100 ip->i_nlink = 0;
2101 ip->i_flag |= IN_CHANGE;
2102 if (DOINGSOFTDEP(tvp))
2103 softdep_change_linkcnt(ip);
2104 vput(tvp);
2105 return (error);
2106}
2107
2108static int
0973c589 2109ufs_missingop(struct vop_generic_args *ap)
984263bc 2110{
984263bc
MD
2111 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2112 return (EOPNOTSUPP);
2113}
2114
2115static struct filterops ufsread_filtops =
2116 { 1, NULL, filt_ufsdetach, filt_ufsread };
2117static struct filterops ufswrite_filtops =
2118 { 1, NULL, filt_ufsdetach, filt_ufswrite };
2119static struct filterops ufsvnode_filtops =
2120 { 1, NULL, filt_ufsdetach, filt_ufsvnode };
2121
0973c589
CP
2122/*
2123 * ufs_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2124 */
984263bc 2125static int
0973c589 2126ufs_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
2127{
2128 struct vnode *vp = ap->a_vp;
2129 struct knote *kn = ap->a_kn;
41a01a4d 2130 lwkt_tokref ilock;
984263bc
MD
2131
2132 switch (kn->kn_filter) {
2133 case EVFILT_READ:
2134 kn->kn_fop = &ufsread_filtops;
2135 break;
2136 case EVFILT_WRITE:
2137 kn->kn_fop = &ufswrite_filtops;
2138 break;
2139 case EVFILT_VNODE:
2140 kn->kn_fop = &ufsvnode_filtops;
2141 break;
2142 default:
2143 return (1);
2144 }
2145
2146 kn->kn_hook = (caddr_t)vp;
2147
41a01a4d 2148 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc 2149 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
41a01a4d 2150 lwkt_reltoken(&ilock);
984263bc
MD
2151
2152 return (0);
2153}
2154
2155static void
2156filt_ufsdetach(struct knote *kn)
2157{
2158 struct vnode *vp = (struct vnode *)kn->kn_hook;
41a01a4d 2159 lwkt_tokref ilock;
984263bc 2160
41a01a4d 2161 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc
MD
2162 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2163 kn, knote, kn_selnext);
41a01a4d 2164 lwkt_reltoken(&ilock);
984263bc
MD
2165}
2166
2167/*ARGSUSED*/
2168static int
2169filt_ufsread(struct knote *kn, long hint)
2170{
2171 struct vnode *vp = (struct vnode *)kn->kn_hook;
2172 struct inode *ip = VTOI(vp);
2173
2174 /*
2175 * filesystem is gone, so set the EOF flag and schedule
2176 * the knote for deletion.
2177 */
2178 if (hint == NOTE_REVOKE) {
2179 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2180 return (1);
2181 }
2182
2183 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2184 return (kn->kn_data != 0);
2185}
2186
2187/*ARGSUSED*/
2188static int
2189filt_ufswrite(struct knote *kn, long hint)
2190{
984263bc
MD
2191 /*
2192 * filesystem is gone, so set the EOF flag and schedule
2193 * the knote for deletion.
2194 */
2195 if (hint == NOTE_REVOKE)
2196 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2197
2198 kn->kn_data = 0;
2199 return (1);
2200}
2201
2202static int
2203filt_ufsvnode(struct knote *kn, long hint)
2204{
984263bc
MD
2205 if (kn->kn_sfflags & hint)
2206 kn->kn_fflags |= hint;
2207 if (hint == NOTE_REVOKE) {
2208 kn->kn_flags |= EV_EOF;
2209 return (1);
2210 }
2211 return (kn->kn_fflags != 0);
2212}
2213
2214/* Global vfs data structures for ufs. */
2d3e977e 2215static struct vop_ops *ufs_vnode_vops;
984263bc 2216static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2d3e977e
MD
2217 { &vop_default_desc, vop_defaultop },
2218 { &vop_fsync_desc, (void *) ufs_missingop },
2219 { &vop_read_desc, (void *) ufs_missingop },
2220 { &vop_reallocblks_desc, (void *) ufs_missingop },
2221 { &vop_write_desc, (void *) ufs_missingop },
2222 { &vop_access_desc, (void *) ufs_access },
2223 { &vop_advlock_desc, (void *) ufs_advlock },
2224 { &vop_bmap_desc, (void *) ufs_bmap },
2225 { &vop_cachedlookup_desc, (void *) ufs_lookup },
2226 { &vop_close_desc, (void *) ufs_close },
2227 { &vop_create_desc, (void *) ufs_create },
2228 { &vop_getattr_desc, (void *) ufs_getattr },
2229 { &vop_inactive_desc, (void *) ufs_inactive },
2230 { &vop_islocked_desc, (void *) vop_stdislocked },
2231 { &vop_link_desc, (void *) ufs_link },
2232 { &vop_lock_desc, (void *) vop_stdlock },
2233 { &vop_lookup_desc, (void *) vfs_cache_lookup },
2234 { &vop_mkdir_desc, (void *) ufs_mkdir },
2235 { &vop_mknod_desc, (void *) ufs_mknod },
2236 { &vop_mmap_desc, (void *) ufs_mmap },
2237 { &vop_open_desc, (void *) ufs_open },
2238 { &vop_pathconf_desc, (void *) ufs_pathconf },
2239 { &vop_poll_desc, (void *) vop_stdpoll },
2240 { &vop_kqfilter_desc, (void *) ufs_kqfilter },
2241 { &vop_print_desc, (void *) ufs_print },
2242 { &vop_readdir_desc, (void *) ufs_readdir },
2243 { &vop_readlink_desc, (void *) ufs_readlink },
2244 { &vop_reclaim_desc, (void *) ufs_reclaim },
2245 { &vop_remove_desc, (void *) ufs_remove },
2246 { &vop_rename_desc, (void *) ufs_rename },
2247 { &vop_rmdir_desc, (void *) ufs_rmdir },
2248 { &vop_setattr_desc, (void *) ufs_setattr },
2249 { &vop_strategy_desc, (void *) ufs_strategy },
2250 { &vop_symlink_desc, (void *) ufs_symlink },
2251 { &vop_unlock_desc, (void *) vop_stdunlock },
2252 { &vop_whiteout_desc, (void *) ufs_whiteout },
984263bc
MD
2253 { NULL, NULL }
2254};
2255static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2d3e977e 2256 { &ufs_vnode_vops, ufs_vnodeop_entries };
984263bc 2257
2d3e977e 2258static struct vop_ops *ufs_spec_vops;
984263bc 2259static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2d3e977e
MD
2260 { &vop_default_desc, (void *) spec_vnoperate },
2261 { &vop_fsync_desc, (void *) ufs_missingop },
2262 { &vop_access_desc, (void *) ufs_access },
2263 { &vop_close_desc, (void *) ufsspec_close },
2264 { &vop_getattr_desc, (void *) ufs_getattr },
2265 { &vop_inactive_desc, (void *) ufs_inactive },
2266 { &vop_islocked_desc, (void *) vop_stdislocked },
2267 { &vop_lock_desc, (void *) vop_stdlock },
2268 { &vop_print_desc, (void *) ufs_print },
2269 { &vop_read_desc, (void *) ufsspec_read },
2270 { &vop_reclaim_desc, (void *) ufs_reclaim },
2271 { &vop_setattr_desc, (void *) ufs_setattr },
2272 { &vop_unlock_desc, (void *) vop_stdunlock },
2273 { &vop_write_desc, (void *) ufsspec_write },
984263bc
MD
2274 { NULL, NULL }
2275};
2276static struct vnodeopv_desc ufs_specop_opv_desc =
2d3e977e 2277 { &ufs_spec_vops, ufs_specop_entries };
984263bc 2278
2d3e977e 2279static struct vop_ops *ufs_fifo_vops;
984263bc 2280static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2d3e977e
MD
2281 { &vop_default_desc, (void *) fifo_vnoperate },
2282 { &vop_fsync_desc, (void *) ufs_missingop },
2283 { &vop_access_desc, (void *) ufs_access },
2284 { &vop_close_desc, (void *) ufsfifo_close },
2285 { &vop_getattr_desc, (void *) ufs_getattr },
2286 { &vop_inactive_desc, (void *) ufs_inactive },
2287 { &vop_islocked_desc, (void *) vop_stdislocked },
2288 { &vop_kqfilter_desc, (void *) ufsfifo_kqfilter },
2289 { &vop_lock_desc, (void *) vop_stdlock },
2290 { &vop_print_desc, (void *) ufs_print },
2291 { &vop_read_desc, (void *) ufsfifo_read },
2292 { &vop_reclaim_desc, (void *) ufs_reclaim },
2293 { &vop_setattr_desc, (void *) ufs_setattr },
2294 { &vop_unlock_desc, (void *) vop_stdunlock },
2295 { &vop_write_desc, (void *) ufsfifo_write },
984263bc
MD
2296 { NULL, NULL }
2297};
2298static struct vnodeopv_desc ufs_fifoop_opv_desc =
2d3e977e 2299 { &ufs_fifo_vops, ufs_fifoop_entries };
984263bc
MD
2300
2301VNODEOP_SET(ufs_vnodeop_opv_desc);
2302VNODEOP_SET(ufs_specop_opv_desc);
2303VNODEOP_SET(ufs_fifoop_opv_desc);
2304
0973c589
CP
2305/*
2306 * ufs_vnoperate(struct vnodeop_desc *a_desc)
2307 */
984263bc 2308int
0973c589 2309ufs_vnoperate(struct vop_generic_args *ap)
984263bc 2310{
0961aa92 2311 return (VOCALL(ufs_vnode_vops, ap));
984263bc
MD
2312}
2313
0973c589
CP
2314/*
2315 * ufs_vnoperatefifo(struct vnodeop_desc *a_desc)
2316 */
984263bc 2317int
0973c589 2318ufs_vnoperatefifo(struct vop_generic_args *ap)
984263bc 2319{
0961aa92 2320 return (VOCALL(ufs_fifo_vops, ap));
984263bc
MD
2321}
2322
0973c589
CP
2323/*
2324 * ufs_vnoperatespec(struct vnodeop_desc *a_desc)
2325 */
984263bc 2326int
0973c589 2327ufs_vnoperatespec(struct vop_generic_args *ap)
984263bc 2328{
0961aa92 2329 return (VOCALL(ufs_spec_vops, ap));
984263bc 2330}