VFS messaging/interfacing work stage 1/99. This stage replaces the old
[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 $
2d3e977e 40 * $DragonFly: src/sys/vfs/ufs/ufs_vnops.c,v 1.16 2004/08/13 17:51:13 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
MD
186 */
187int
0973c589 188ufs_create(struct vop_create_args *ap)
984263bc
MD
189{
190 int error;
191
192 error =
193 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
194 ap->a_dvp, ap->a_vpp, ap->a_cnp);
195 if (error)
196 return (error);
197 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
198 return (0);
199}
200
201/*
202 * Mknod vnode call
0973c589
CP
203 *
204 * ufs_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
205 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
206 */
207/* ARGSUSED */
208int
0973c589 209ufs_mknod(struct vop_mknod_args *ap)
984263bc
MD
210{
211 struct vattr *vap = ap->a_vap;
212 struct vnode **vpp = ap->a_vpp;
213 struct inode *ip;
214 ino_t ino;
215 int error;
216
217 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
218 ap->a_dvp, vpp, ap->a_cnp);
219 if (error)
220 return (error);
221 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
222 ip = VTOI(*vpp);
223 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
224 if (vap->va_rdev != VNOVAL) {
225 /*
226 * Want to be able to use this to make badblock
227 * inodes, so don't truncate the dev number.
228 */
229 ip->i_rdev = vap->va_rdev;
230 }
231 /*
232 * Remove inode, then reload it through VFS_VGET so it is
233 * checked to see if it is an alias of an existing entry in
234 * the inode cache.
235 */
236 vput(*vpp);
237 (*vpp)->v_type = VNON;
238 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
239 vgone(*vpp);
240 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
241 if (error) {
242 *vpp = NULL;
243 return (error);
244 }
245 return (0);
246}
247
248/*
249 * Open called.
250 *
251 * Nothing to do.
0973c589
CP
252 *
253 * ufs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
254 * struct thread *a_td)
984263bc
MD
255 */
256/* ARGSUSED */
257int
0973c589 258ufs_open(struct vop_open_args *ap)
984263bc 259{
984263bc
MD
260 /*
261 * Files marked append-only must be opened for appending.
262 */
263 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
264 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
265 return (EPERM);
266 return (0);
267}
268
269/*
270 * Close called.
271 *
272 * Update the times on the inode.
0973c589
CP
273 *
274 * ufs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
275 * struct thread *a_td)
984263bc
MD
276 */
277/* ARGSUSED */
278int
0973c589 279ufs_close(struct vop_close_args *ap)
984263bc 280{
3ff2135f 281 struct vnode *vp = ap->a_vp;
41a01a4d 282 lwkt_tokref vlock;
984263bc 283
41a01a4d 284 lwkt_gettoken(&vlock, vp->v_interlock);
984263bc
MD
285 if (vp->v_usecount > 1)
286 ufs_itimes(vp);
41a01a4d 287 lwkt_reltoken(&vlock);
984263bc
MD
288 return (0);
289}
290
0973c589
CP
291/*
292 * ufs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
293 * struct thread *a_td)
294 */
984263bc 295int
0973c589 296ufs_access(struct vop_access_args *ap)
984263bc
MD
297{
298 struct vnode *vp = ap->a_vp;
299 struct inode *ip = VTOI(vp);
300 struct ucred *cred = ap->a_cred;
301 mode_t mask, mode = ap->a_mode;
3ff2135f 302 gid_t *gp;
984263bc
MD
303 int i;
304#ifdef QUOTA
305 int error;
306#endif
307
308 /*
f719c866 309 * Disallow write attempts on read-only filesystems;
984263bc 310 * unless the file is a socket, fifo, or a block or
f719c866 311 * character device resident on the filesystem.
984263bc
MD
312 */
313 if (mode & VWRITE) {
314 switch (vp->v_type) {
315 case VDIR:
316 case VLNK:
317 case VREG:
318 if (vp->v_mount->mnt_flag & MNT_RDONLY)
319 return (EROFS);
320#ifdef QUOTA
321 if ((error = getinoquota(ip)) != 0)
322 return (error);
323#endif
324 break;
325 default:
326 break;
327 }
328 }
329
330 /* If immutable bit set, nobody gets to write it. */
331 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
332 return (EPERM);
333
334 /* Otherwise, user id 0 always gets access. */
335 if (cred->cr_uid == 0)
336 return (0);
337
338 mask = 0;
339
340 /* Otherwise, check the owner. */
341 if (cred->cr_uid == ip->i_uid) {
342 if (mode & VEXEC)
343 mask |= S_IXUSR;
344 if (mode & VREAD)
345 mask |= S_IRUSR;
346 if (mode & VWRITE)
347 mask |= S_IWUSR;
348 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
349 }
350
351 /* Otherwise, check the groups. */
352 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
353 if (ip->i_gid == *gp) {
354 if (mode & VEXEC)
355 mask |= S_IXGRP;
356 if (mode & VREAD)
357 mask |= S_IRGRP;
358 if (mode & VWRITE)
359 mask |= S_IWGRP;
360 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
361 }
362
363 /* Otherwise, check everyone else. */
364 if (mode & VEXEC)
365 mask |= S_IXOTH;
366 if (mode & VREAD)
367 mask |= S_IROTH;
368 if (mode & VWRITE)
369 mask |= S_IWOTH;
370 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
371}
372
0973c589
CP
373/*
374 * ufs_getattr(struct vnode *a_vp, struct vattr *a_vap,
375 * struct thread *a_td)
376 */
984263bc
MD
377/* ARGSUSED */
378int
0973c589 379ufs_getattr(struct vop_getattr_args *ap)
984263bc 380{
3ff2135f
RG
381 struct vnode *vp = ap->a_vp;
382 struct inode *ip = VTOI(vp);
383 struct vattr *vap = ap->a_vap;
984263bc
MD
384
385 ufs_itimes(vp);
386 /*
387 * Copy from inode table
388 */
389 vap->va_fsid = dev2udev(ip->i_dev);
390 vap->va_fileid = ip->i_number;
391 vap->va_mode = ip->i_mode & ~IFMT;
392 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ?
393 ip->i_effnlink : ip->i_nlink;
394 vap->va_uid = ip->i_uid;
395 vap->va_gid = ip->i_gid;
396 vap->va_rdev = ip->i_rdev;
397 vap->va_size = ip->i_din.di_size;
398 vap->va_atime.tv_sec = ip->i_atime;
399 vap->va_atime.tv_nsec = ip->i_atimensec;
400 vap->va_mtime.tv_sec = ip->i_mtime;
401 vap->va_mtime.tv_nsec = ip->i_mtimensec;
402 vap->va_ctime.tv_sec = ip->i_ctime;
403 vap->va_ctime.tv_nsec = ip->i_ctimensec;
404 vap->va_flags = ip->i_flags;
405 vap->va_gen = ip->i_gen;
406 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
407 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
408 vap->va_type = IFTOVT(ip->i_mode);
409 vap->va_filerev = ip->i_modrev;
410 return (0);
411}
412
413/*
414 * Set attribute vnode op. called from several syscalls
0973c589
CP
415 *
416 * ufs_setattr(struct vnode *a_vp, struct vattr *a_vap,
417 * struct ucred *a_cred, struct thread *a_td)
984263bc
MD
418 */
419int
0973c589 420ufs_setattr(struct vop_setattr_args *ap)
984263bc
MD
421{
422 struct vattr *vap = ap->a_vap;
423 struct vnode *vp = ap->a_vp;
424 struct inode *ip = VTOI(vp);
425 struct ucred *cred = ap->a_cred;
984263bc
MD
426 int error;
427
428 /*
429 * Check for unsettable attributes.
430 */
431 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
432 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
433 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
434 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
435 return (EINVAL);
436 }
437 if (vap->va_flags != VNOVAL) {
438 if (vp->v_mount->mnt_flag & MNT_RDONLY)
439 return (EROFS);
440 if (cred->cr_uid != ip->i_uid &&
dadab5e9 441 (error = suser_cred(cred, PRISON_ROOT)))
984263bc 442 return (error);
41c20dac 443 if ((cred->cr_uid == 0) && (cred->cr_prison == NULL)) {
984263bc
MD
444 if ((ip->i_flags
445 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
446 securelevel > 0)
447 return (EPERM);
448 ip->i_flags = vap->va_flags;
449 } else {
450 if (ip->i_flags
451 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
452 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
453 return (EPERM);
454 ip->i_flags &= SF_SETTABLE;
455 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
456 }
457 ip->i_flag |= IN_CHANGE;
458 if (vap->va_flags & (IMMUTABLE | APPEND))
459 return (0);
460 }
461 if (ip->i_flags & (IMMUTABLE | APPEND))
462 return (EPERM);
463 /*
464 * Go through the fields and update iff not VNOVAL.
465 */
466 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
467 if (vp->v_mount->mnt_flag & MNT_RDONLY)
468 return (EROFS);
dadab5e9 469 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
984263bc
MD
470 return (error);
471 }
472 if (vap->va_size != VNOVAL) {
473 /*
f719c866 474 * Disallow write attempts on read-only filesystems;
984263bc 475 * unless the file is a socket, fifo, or a block or
f719c866 476 * character device resident on the filesystem.
984263bc
MD
477 */
478 switch (vp->v_type) {
479 case VDIR:
480 return (EISDIR);
481 case VLNK:
482 case VREG:
483 if (vp->v_mount->mnt_flag & MNT_RDONLY)
484 return (EROFS);
485 break;
486 default:
487 break;
488 }
dadab5e9 489 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
984263bc
MD
490 return (error);
491 }
492 ip = VTOI(vp);
493 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
494 if (vp->v_mount->mnt_flag & MNT_RDONLY)
495 return (EROFS);
496 if (cred->cr_uid != ip->i_uid &&
dadab5e9 497 (error = suser_cred(cred, PRISON_ROOT)) &&
984263bc 498 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
dadab5e9 499 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
984263bc
MD
500 return (error);
501 if (vap->va_atime.tv_sec != VNOVAL)
502 ip->i_flag |= IN_ACCESS;
503 if (vap->va_mtime.tv_sec != VNOVAL)
504 ip->i_flag |= IN_CHANGE | IN_UPDATE;
505 ufs_itimes(vp);
506 if (vap->va_atime.tv_sec != VNOVAL) {
507 ip->i_atime = vap->va_atime.tv_sec;
508 ip->i_atimensec = vap->va_atime.tv_nsec;
509 }
510 if (vap->va_mtime.tv_sec != VNOVAL) {
511 ip->i_mtime = vap->va_mtime.tv_sec;
512 ip->i_mtimensec = vap->va_mtime.tv_nsec;
513 }
514 error = UFS_UPDATE(vp, 0);
515 if (error)
516 return (error);
517 }
518 error = 0;
519 if (vap->va_mode != (mode_t)VNOVAL) {
520 if (vp->v_mount->mnt_flag & MNT_RDONLY)
521 return (EROFS);
dadab5e9 522 error = ufs_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
984263bc
MD
523 }
524 VN_KNOTE(vp, NOTE_ATTRIB);
525 return (error);
526}
527
528/*
529 * Change the mode on a file.
530 * Inode must be locked before calling.
531 */
532static int
dadab5e9 533ufs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
984263bc 534{
3ff2135f 535 struct inode *ip = VTOI(vp);
984263bc
MD
536 int error;
537
538 if (cred->cr_uid != ip->i_uid) {
dadab5e9 539 error = suser_cred(cred, PRISON_ROOT);
984263bc
MD
540 if (error)
541 return (error);
542 }
543 if (cred->cr_uid) {
544 if (vp->v_type != VDIR && (mode & S_ISTXT))
545 return (EFTYPE);
546 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
547 return (EPERM);
548 }
549 ip->i_mode &= ~ALLPERMS;
550 ip->i_mode |= (mode & ALLPERMS);
551 ip->i_flag |= IN_CHANGE;
552 return (0);
553}
554
555/*
556 * Perform chown operation on inode ip;
557 * inode must be locked prior to call.
558 */
559static int
0973c589
CP
560ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
561 struct thread *td)
984263bc 562{
3ff2135f 563 struct inode *ip = VTOI(vp);
984263bc
MD
564 uid_t ouid;
565 gid_t ogid;
566 int error = 0;
567#ifdef QUOTA
3ff2135f 568 int i;
984263bc
MD
569 long change;
570#endif
571
572 if (uid == (uid_t)VNOVAL)
573 uid = ip->i_uid;
574 if (gid == (gid_t)VNOVAL)
575 gid = ip->i_gid;
576 /*
577 * If we don't own the file, are trying to change the owner
578 * of the file, or are not a member of the target group,
579 * the caller must be superuser or the call fails.
580 */
581 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
a10e3626
DR
582 (gid != ip->i_gid && !(cred->cr_gid == gid ||
583 groupmember((gid_t)gid, cred)))) &&
dadab5e9 584 (error = suser_cred(cred, PRISON_ROOT)))
984263bc
MD
585 return (error);
586 ogid = ip->i_gid;
587 ouid = ip->i_uid;
588#ifdef QUOTA
589 if ((error = getinoquota(ip)) != 0)
590 return (error);
591 if (ouid == uid) {
592 dqrele(vp, ip->i_dquot[USRQUOTA]);
593 ip->i_dquot[USRQUOTA] = NODQUOT;
594 }
595 if (ogid == gid) {
596 dqrele(vp, ip->i_dquot[GRPQUOTA]);
597 ip->i_dquot[GRPQUOTA] = NODQUOT;
598 }
599 change = ip->i_blocks;
600 (void) chkdq(ip, -change, cred, CHOWN);
601 (void) chkiq(ip, -1, cred, CHOWN);
602 for (i = 0; i < MAXQUOTAS; i++) {
603 dqrele(vp, ip->i_dquot[i]);
604 ip->i_dquot[i] = NODQUOT;
605 }
606#endif
607 ip->i_gid = gid;
608 ip->i_uid = uid;
609#ifdef QUOTA
610 if ((error = getinoquota(ip)) == 0) {
611 if (ouid == uid) {
612 dqrele(vp, ip->i_dquot[USRQUOTA]);
613 ip->i_dquot[USRQUOTA] = NODQUOT;
614 }
615 if (ogid == gid) {
616 dqrele(vp, ip->i_dquot[GRPQUOTA]);
617 ip->i_dquot[GRPQUOTA] = NODQUOT;
618 }
619 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
620 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
621 goto good;
622 else
623 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
624 }
625 for (i = 0; i < MAXQUOTAS; i++) {
626 dqrele(vp, ip->i_dquot[i]);
627 ip->i_dquot[i] = NODQUOT;
628 }
629 }
630 ip->i_gid = ogid;
631 ip->i_uid = ouid;
632 if (getinoquota(ip) == 0) {
633 if (ouid == uid) {
634 dqrele(vp, ip->i_dquot[USRQUOTA]);
635 ip->i_dquot[USRQUOTA] = NODQUOT;
636 }
637 if (ogid == gid) {
638 dqrele(vp, ip->i_dquot[GRPQUOTA]);
639 ip->i_dquot[GRPQUOTA] = NODQUOT;
640 }
641 (void) chkdq(ip, change, cred, FORCE|CHOWN);
642 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
643 (void) getinoquota(ip);
644 }
645 return (error);
646good:
647 if (getinoquota(ip))
648 panic("ufs_chown: lost quota");
649#endif /* QUOTA */
650 ip->i_flag |= IN_CHANGE;
651 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
652 ip->i_mode &= ~(ISUID | ISGID);
653 return (0);
654}
655
656/*
657 * Mmap a file
658 *
659 * NB Currently unsupported.
0973c589
CP
660 *
661 * ufs_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred,
662 * struct thread *a_td)
984263bc
MD
663 */
664/* ARGSUSED */
665int
0973c589 666ufs_mmap(struct vop_mmap_args *ap)
984263bc 667{
984263bc
MD
668 return (EINVAL);
669}
670
0973c589
CP
671/*
672 * ufs_remove(struct vnode *a_dvp, struct vnode *a_vp,
673 * struct componentname *a_cnp)
674 */
984263bc 675int
0973c589 676ufs_remove(struct vop_remove_args *ap)
984263bc
MD
677{
678 struct inode *ip;
679 struct vnode *vp = ap->a_vp;
680 struct vnode *dvp = ap->a_dvp;
681 int error;
682
683 ip = VTOI(vp);
684 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
685 (VTOI(dvp)->i_flags & APPEND)) {
686 error = EPERM;
687 goto out;
688 }
689 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
690 VN_KNOTE(vp, NOTE_DELETE);
691 VN_KNOTE(dvp, NOTE_WRITE);
692out:
693 return (error);
694}
695
696/*
697 * link vnode call
0973c589
CP
698 *
699 * ufs_link(struct vnode *a_tdvp, struct vnode *a_vp,
700 * struct componentname *a_cnp)
984263bc
MD
701 */
702int
0973c589 703ufs_link(struct vop_link_args *ap)
984263bc
MD
704{
705 struct vnode *vp = ap->a_vp;
706 struct vnode *tdvp = ap->a_tdvp;
707 struct componentname *cnp = ap->a_cnp;
dadab5e9 708 struct thread *td = cnp->cn_td;
984263bc
MD
709 struct inode *ip;
710 struct direct newdir;
711 int error;
712
713#ifdef DIAGNOSTIC
2b69e610 714 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
715 panic("ufs_link: no name");
716#endif
717 if (tdvp->v_mount != vp->v_mount) {
718 error = EXDEV;
719 goto out2;
720 }
41a01a4d 721 if (tdvp != vp && (error = vn_lock(vp, NULL, LK_EXCLUSIVE, td))) {
984263bc
MD
722 goto out2;
723 }
724 ip = VTOI(vp);
725 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
726 error = EMLINK;
727 goto out1;
728 }
729 if (ip->i_flags & (IMMUTABLE | APPEND)) {
730 error = EPERM;
731 goto out1;
732 }
733 ip->i_effnlink++;
734 ip->i_nlink++;
735 ip->i_flag |= IN_CHANGE;
736 if (DOINGSOFTDEP(vp))
737 softdep_change_linkcnt(ip);
738 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
739 if (!error) {
740 ufs_makedirentry(ip, cnp, &newdir);
741 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
742 }
743
744 if (error) {
745 ip->i_effnlink--;
746 ip->i_nlink--;
747 ip->i_flag |= IN_CHANGE;
748 if (DOINGSOFTDEP(vp))
749 softdep_change_linkcnt(ip);
750 }
751out1:
752 if (tdvp != vp)
41a01a4d 753 VOP_UNLOCK(vp, NULL, 0, td);
984263bc
MD
754out2:
755 VN_KNOTE(vp, NOTE_LINK);
756 VN_KNOTE(tdvp, NOTE_WRITE);
757 return (error);
758}
759
760/*
761 * whiteout vnode call
0973c589
CP
762 *
763 * ufs_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
984263bc
MD
764 */
765int
0973c589 766ufs_whiteout(struct vop_whiteout_args *ap)
984263bc
MD
767{
768 struct vnode *dvp = ap->a_dvp;
769 struct componentname *cnp = ap->a_cnp;
770 struct direct newdir;
771 int error = 0;
772
773 switch (ap->a_flags) {
2b69e610 774 case NAMEI_LOOKUP:
984263bc
MD
775 /* 4.4 format directories support whiteout operations */
776 if (dvp->v_mount->mnt_maxsymlinklen > 0)
777 return (0);
778 return (EOPNOTSUPP);
779
2b69e610 780 case NAMEI_CREATE:
984263bc
MD
781 /* create a new directory whiteout */
782#ifdef DIAGNOSTIC
2b69e610 783 if ((cnp->cn_flags & CNP_SAVENAME) == 0)
984263bc
MD
784 panic("ufs_whiteout: missing name");
785 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
786 panic("ufs_whiteout: old format filesystem");
787#endif
788
789 newdir.d_ino = WINO;
790 newdir.d_namlen = cnp->cn_namelen;
791 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
792 newdir.d_type = DT_WHT;
793 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
794 break;
795
2b69e610 796 case NAMEI_DELETE:
984263bc
MD
797 /* remove an existing directory whiteout */
798#ifdef DIAGNOSTIC
799 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
800 panic("ufs_whiteout: old format filesystem");
801#endif
802
2b69e610 803 cnp->cn_flags &= ~CNP_DOWHITEOUT;
984263bc
MD
804 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
805 break;
806 default:
807 panic("ufs_whiteout: unknown op");
808 }
809 return (error);
810}
811
812/*
813 * Rename system call.
814 * rename("foo", "bar");
815 * is essentially
816 * unlink("bar");
817 * link("foo", "bar");
818 * unlink("foo");
819 * but ``atomically''. Can't do full commit without saving state in the
820 * inode on disk which isn't feasible at this time. Best we can do is
821 * always guarantee the target exists.
822 *
823 * Basic algorithm is:
824 *
825 * 1) Bump link count on source while we're linking it to the
826 * target. This also ensure the inode won't be deleted out
827 * from underneath us while we work (it may be truncated by
828 * a concurrent `trunc' or `open' for creation).
829 * 2) Link source to destination. If destination already exists,
830 * delete it first.
831 * 3) Unlink source reference to inode if still around. If a
832 * directory was moved and the parent of the destination
833 * is different from the source, patch the ".." entry in the
834 * directory.
0973c589
CP
835 *
836 * ufs_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
837 * struct componentname *a_fcnp, struct vnode *a_tdvp,
838 * struct vnode *a_tvp, struct componentname *a_tcnp)
984263bc
MD
839 */
840int
0973c589 841ufs_rename(struct vop_rename_args *ap)
984263bc
MD
842{
843 struct vnode *tvp = ap->a_tvp;
3ff2135f 844 struct vnode *tdvp = ap->a_tdvp;
984263bc
MD
845 struct vnode *fvp = ap->a_fvp;
846 struct vnode *fdvp = ap->a_fdvp;
847 struct componentname *tcnp = ap->a_tcnp;
848 struct componentname *fcnp = ap->a_fcnp;
dadab5e9 849 struct thread *td = fcnp->cn_td;
984263bc
MD
850 struct inode *ip, *xp, *dp;
851 struct direct newdir;
852 int doingdirectory = 0, oldparent = 0, newparent = 0;
853 int error = 0, ioflag;
854
855#ifdef DIAGNOSTIC
2b69e610
MD
856 if ((tcnp->cn_flags & CNP_HASBUF) == 0 ||
857 (fcnp->cn_flags & CNP_HASBUF) == 0)
984263bc 858 panic("ufs_rename: no name");
f719c866 859#endif /* DIAGNOSTIC */
984263bc
MD
860 /*
861 * Check for cross-device rename.
862 */
863 if ((fvp->v_mount != tdvp->v_mount) ||
864 (tvp && (fvp->v_mount != tvp->v_mount))) {
865 error = EXDEV;
866abortit:
867 if (tdvp == tvp)
868 vrele(tdvp);
869 else
870 vput(tdvp);
871 if (tvp)
872 vput(tvp);
873 vrele(fdvp);
874 vrele(fvp);
875 return (error);
876 }
877
878 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
879 (VTOI(tdvp)->i_flags & APPEND))) {
880 error = EPERM;
881 goto abortit;
882 }
883
884 /*
885 * Renaming a file to itself has no effect. The upper layers should
886 * not call us in that case. Temporarily just warn if they do.
887 */
888 if (fvp == tvp) {
889 printf("ufs_rename: fvp == tvp (can't happen)\n");
890 error = 0;
891 goto abortit;
892 }
893
41a01a4d 894 if ((error = vn_lock(fvp, NULL, LK_EXCLUSIVE, td)) != 0)
984263bc
MD
895 goto abortit;
896 dp = VTOI(fdvp);
897 ip = VTOI(fvp);
898 if (ip->i_nlink >= LINK_MAX) {
41a01a4d 899 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
900 error = EMLINK;
901 goto abortit;
902 }
903 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
904 || (dp->i_flags & APPEND)) {
41a01a4d 905 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
906 error = EPERM;
907 goto abortit;
908 }
909 if ((ip->i_mode & IFMT) == IFDIR) {
910 /*
911 * Avoid ".", "..", and aliases of "." for obvious reasons.
912 */
913 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
2b69e610 914 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
984263bc 915 (ip->i_flag & IN_RENAME)) {
41a01a4d 916 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
917 error = EINVAL;
918 goto abortit;
919 }
920 ip->i_flag |= IN_RENAME;
921 oldparent = dp->i_number;
922 doingdirectory = 1;
923 }
924 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
925 vrele(fdvp);
926
927 /*
928 * When the target exists, both the directory
929 * and target vnodes are returned locked.
930 */
931 dp = VTOI(tdvp);
932 xp = NULL;
933 if (tvp)
934 xp = VTOI(tvp);
935
936 /*
937 * 1) Bump link count while we're moving stuff
938 * around. If we crash somewhere before
939 * completing our work, the link count
940 * may be wrong, but correctable.
941 */
942 ip->i_effnlink++;
943 ip->i_nlink++;
944 ip->i_flag |= IN_CHANGE;
945 if (DOINGSOFTDEP(fvp))
946 softdep_change_linkcnt(ip);
947 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
948 DOINGASYNC(fvp)))) != 0) {
41a01a4d 949 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
950 goto bad;
951 }
952
953 /*
954 * If ".." must be changed (ie the directory gets a new
955 * parent) then the source directory must not be in the
956 * directory heirarchy above the target, as this would
957 * orphan everything below the source directory. Also
958 * the user must have write permission in the source so
959 * as to be able to change "..". We must repeat the call
960 * to namei, as the parent directory is unlocked by the
961 * call to checkpath().
962 */
dadab5e9 963 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
41a01a4d 964 VOP_UNLOCK(fvp, NULL, 0, td);
984263bc
MD
965 if (oldparent != dp->i_number)
966 newparent = dp->i_number;
967 if (doingdirectory && newparent) {
968 if (error) /* write access check above */
969 goto bad;
970 if (xp != NULL)
971 vput(tvp);
972 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
973 if (error)
974 goto out;
2b69e610 975 if ((tcnp->cn_flags & CNP_SAVESTART) == 0)
984263bc 976 panic("ufs_rename: lost to startdir");
597aea93 977 vref(tdvp);
984263bc
MD
978 error = relookup(tdvp, &tvp, tcnp);
979 if (error)
980 goto out;
981 vrele(tdvp);
982 dp = VTOI(tdvp);
983 xp = NULL;
984 if (tvp)
985 xp = VTOI(tvp);
986 }
987 /*
988 * 2) If target doesn't exist, link the target
989 * to the source and unlink the source.
990 * Otherwise, rewrite the target directory
991 * entry to reference the source inode and
992 * expunge the original entry's existence.
993 */
994 if (xp == NULL) {
995 if (dp->i_dev != ip->i_dev)
996 panic("ufs_rename: EXDEV");
997 /*
998 * Account for ".." in new directory.
999 * When source and destination have the same
1000 * parent we don't fool with the link count.
1001 */
1002 if (doingdirectory && newparent) {
1003 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1004 error = EMLINK;
1005 goto bad;
1006 }
1007 dp->i_effnlink++;
1008 dp->i_nlink++;
1009 dp->i_flag |= IN_CHANGE;
1010 if (DOINGSOFTDEP(tdvp))
1011 softdep_change_linkcnt(dp);
1012 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1013 DOINGASYNC(tdvp)));
1014 if (error)
1015 goto bad;
1016 }
1017 ufs_makedirentry(ip, tcnp, &newdir);
1018 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1019 if (error) {
1020 if (doingdirectory && newparent) {
1021 dp->i_effnlink--;
1022 dp->i_nlink--;
1023 dp->i_flag |= IN_CHANGE;
1024 if (DOINGSOFTDEP(tdvp))
1025 softdep_change_linkcnt(dp);
1026 (void)UFS_UPDATE(tdvp, 1);
1027 }
1028 goto bad;
1029 }
1030 VN_KNOTE(tdvp, NOTE_WRITE);
1031 vput(tdvp);
1032 } else {
1033 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1034 panic("ufs_rename: EXDEV");
1035 /*
1036 * Short circuit rename(foo, foo).
1037 */
1038 if (xp->i_number == ip->i_number)
1039 panic("ufs_rename: same file");
1040 /*
1041 * If the parent directory is "sticky", then the user must
1042 * own the parent directory, or the destination of the rename,
1043 * otherwise the destination may not be changed (except by
1044 * root). This implements append-only directories.
1045 */
1046 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1047 tcnp->cn_cred->cr_uid != dp->i_uid &&
1048 xp->i_uid != tcnp->cn_cred->cr_uid) {
1049 error = EPERM;
1050 goto bad;
1051 }
1052 /*
1053 * Target must be empty if a directory and have no links
1054 * to it. Also, ensure source and target are compatible
1055 * (both directories, or both not directories).
1056 */
1057 if ((xp->i_mode&IFMT) == IFDIR) {
1058 if ((xp->i_effnlink > 2) ||
1059 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1060 error = ENOTEMPTY;
1061 goto bad;
1062 }
1063 if (!doingdirectory) {
1064 error = ENOTDIR;
1065 goto bad;
1066 }
1067 cache_purge(tdvp);
1068 } else if (doingdirectory) {
1069 error = EISDIR;
1070 goto bad;
1071 }
1072 error = ufs_dirrewrite(dp, xp, ip->i_number,
1073 IFTODT(ip->i_mode),
1074 (doingdirectory && newparent) ? newparent : doingdirectory);
1075 if (error)
1076 goto bad;
1077 if (doingdirectory) {
1078 if (!newparent) {
1079 dp->i_effnlink--;
1080 if (DOINGSOFTDEP(tdvp))
1081 softdep_change_linkcnt(dp);
1082 }
1083 xp->i_effnlink--;
1084 if (DOINGSOFTDEP(tvp))
1085 softdep_change_linkcnt(xp);
1086 }
1087 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1088 /*
1089 * Truncate inode. The only stuff left in the directory
1090 * is "." and "..". The "." reference is inconsequential
1091 * since we are quashing it. We have removed the "."
1092 * reference and the reference in the parent directory,
1093 * but there may be other hard links. The soft
1094 * dependency code will arrange to do these operations
1095 * after the parent directory entry has been deleted on
1096 * disk, so when running with that code we avoid doing
1097 * them now.
1098 */
1099 if (!newparent) {
1100 dp->i_nlink--;
1101 dp->i_flag |= IN_CHANGE;
1102 }
1103 xp->i_nlink--;
1104 xp->i_flag |= IN_CHANGE;
1105 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC;
1106 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
dadab5e9 1107 tcnp->cn_cred, tcnp->cn_td)) != 0)
984263bc
MD
1108 goto bad;
1109 }
1110 VN_KNOTE(tdvp, NOTE_WRITE);
1111 vput(tdvp);
1112 VN_KNOTE(tvp, NOTE_DELETE);
1113 vput(tvp);
1114 xp = NULL;
1115 }
1116
1117 /*
1118 * 3) Unlink the source.
1119 */
2b69e610
MD
1120 fcnp->cn_flags &= ~CNP_MODMASK;
1121 fcnp->cn_flags |= CNP_LOCKPARENT | CNP_LOCKLEAF;
1122 if ((fcnp->cn_flags & CNP_SAVESTART) == 0)
984263bc 1123 panic("ufs_rename: lost from startdir");
597aea93 1124 vref(fdvp);
984263bc
MD
1125 error = relookup(fdvp, &fvp, fcnp);
1126 if (error == 0)
1127 vrele(fdvp);
1128 if (fvp != NULL) {
1129 xp = VTOI(fvp);
1130 dp = VTOI(fdvp);
1131 } else {
1132 /*
1133 * From name has disappeared.
1134 */
1135 if (doingdirectory)
1136 panic("ufs_rename: lost dir entry");
1137 vrele(ap->a_fvp);
1138 return (0);
1139 }
1140 /*
1141 * Ensure that the directory entry still exists and has not
1142 * changed while the new name has been entered. If the source is
1143 * a file then the entry may have been unlinked or renamed. In
1144 * either case there is no further work to be done. If the source
1145 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1146 * flag ensures that it cannot be moved by another rename or removed
1147 * by a rmdir.
1148 */
1149 if (xp != ip) {
1150 if (doingdirectory)
1151 panic("ufs_rename: lost dir entry");
1152 } else {
1153 /*
1154 * If the source is a directory with a
1155 * new parent, the link count of the old
1156 * parent directory must be decremented
1157 * and ".." set to point to the new parent.
1158 */
1159 if (doingdirectory && newparent) {
1160 xp->i_offset = mastertemplate.dot_reclen;
1161 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
1162 cache_purge(fdvp);
1163 }
1164 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1165 xp->i_flag &= ~IN_RENAME;
1166 }
1167 VN_KNOTE(fvp, NOTE_RENAME);
1168 if (dp)
1169 vput(fdvp);
1170 if (xp)
1171 vput(fvp);
1172 vrele(ap->a_fvp);
1173 return (error);
1174
1175bad:
1176 if (xp)
1177 vput(ITOV(xp));
1178 vput(ITOV(dp));
1179out:
1180 if (doingdirectory)
1181 ip->i_flag &= ~IN_RENAME;
41a01a4d 1182 if (vn_lock(fvp, NULL, LK_EXCLUSIVE, td) == 0) {
984263bc
MD
1183 ip->i_effnlink--;
1184 ip->i_nlink--;
1185 ip->i_flag |= IN_CHANGE;
1186 ip->i_flag &= ~IN_RENAME;
1187 if (DOINGSOFTDEP(fvp))
1188 softdep_change_linkcnt(ip);
1189 vput(fvp);
1190 } else
1191 vrele(fvp);
1192 return (error);
1193}
1194
1195/*
1196 * Mkdir system call
0973c589
CP
1197 *
1198 * ufs_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
1199 * struct componentname *a_cnp, struct vattr *a_vap)
984263bc
MD
1200 */
1201int
0973c589 1202ufs_mkdir(struct vop_mkdir_args *ap)
984263bc 1203{
3ff2135f
RG
1204 struct vnode *dvp = ap->a_dvp;
1205 struct vattr *vap = ap->a_vap;
1206 struct componentname *cnp = ap->a_cnp;
1207 struct inode *ip, *dp;
984263bc
MD
1208 struct vnode *tvp;
1209 struct buf *bp;
1210 struct dirtemplate dirtemplate, *dtp;
1211 struct direct newdir;
1212 int error, dmode;
1213 long blkoff;
1214
1215#ifdef DIAGNOSTIC
2b69e610 1216 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
1217 panic("ufs_mkdir: no name");
1218#endif
1219 dp = VTOI(dvp);
1220 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1221 error = EMLINK;
1222 goto out;
1223 }
1224 dmode = vap->va_mode & 0777;
1225 dmode |= IFDIR;
1226 /*
1227 * Must simulate part of ufs_makeinode here to acquire the inode,
1228 * but not have it entered in the parent directory. The entry is
1229 * made later after writing "." and ".." entries.
1230 */
1231 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1232 if (error)
1233 goto out;
1234 ip = VTOI(tvp);
1235 ip->i_gid = dp->i_gid;
1236#ifdef SUIDDIR
1237 {
1238#ifdef QUOTA
1239 struct ucred ucred, *ucp;
1240 ucp = cnp->cn_cred;
1241#endif
1242 /*
1243 * If we are hacking owners here, (only do this where told to)
1244 * and we are not giving it TO root, (would subvert quotas)
1245 * then go ahead and give it to the other user.
1246 * The new directory also inherits the SUID bit.
1247 * If user's UID and dir UID are the same,
1248 * 'give it away' so that the SUID is still forced on.
1249 */
1250 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1251 (dp->i_mode & ISUID) && dp->i_uid) {
1252 dmode |= ISUID;
1253 ip->i_uid = dp->i_uid;
1254#ifdef QUOTA
1255 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1256 /*
1257 * Make sure the correct user gets charged
1258 * for the space.
1259 * Make a dummy credential for the victim.
1260 * XXX This seems to never be accessed out of
1261 * our context so a stack variable is ok.
1262 */
1263 ucred.cr_ref = 1;
1264 ucred.cr_uid = ip->i_uid;
1265 ucred.cr_ngroups = 1;
1266 ucred.cr_groups[0] = dp->i_gid;
1267 ucp = &ucred;
1268 }
1269#endif
1270 } else
1271 ip->i_uid = cnp->cn_cred->cr_uid;
1272#ifdef QUOTA
1273 if ((error = getinoquota(ip)) ||
1274 (error = chkiq(ip, 1, ucp, 0))) {
1275 UFS_VFREE(tvp, ip->i_number, dmode);
1276 vput(tvp);
1277 return (error);
1278 }
1279#endif
1280 }
1281#else /* !SUIDDIR */
1282 ip->i_uid = cnp->cn_cred->cr_uid;
1283#ifdef QUOTA
1284 if ((error = getinoquota(ip)) ||
1285 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1286 UFS_VFREE(tvp, ip->i_number, dmode);
1287 vput(tvp);
1288 return (error);
1289 }
1290#endif
1291#endif /* !SUIDDIR */
1292 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1293 ip->i_mode = dmode;
1294 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1295 ip->i_effnlink = 2;
1296 ip->i_nlink = 2;
1297 if (DOINGSOFTDEP(tvp))
1298 softdep_change_linkcnt(ip);
2b69e610 1299 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
1300 ip->i_flags |= UF_OPAQUE;
1301
1302 /*
1303 * Bump link count in parent directory to reflect work done below.
1304 * Should be done before reference is created so cleanup is
1305 * possible if we crash.
1306 */
1307 dp->i_effnlink++;
1308 dp->i_nlink++;
1309 dp->i_flag |= IN_CHANGE;
1310 if (DOINGSOFTDEP(dvp))
1311 softdep_change_linkcnt(dp);
1312 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1313 if (error)
1314 goto bad;
1315
1316 /*
1317 * Initialize directory with "." and ".." from static template.
1318 */
1319 if (dvp->v_mount->mnt_maxsymlinklen > 0
1320 )
1321 dtp = &mastertemplate;
1322 else
1323 dtp = (struct dirtemplate *)&omastertemplate;
1324 dirtemplate = *dtp;
1325 dirtemplate.dot_ino = ip->i_number;
1326 dirtemplate.dotdot_ino = dp->i_number;
1327 if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1328 B_CLRBUF, &bp)) != 0)
1329 goto bad;
1330 ip->i_size = DIRBLKSIZ;
1331 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1332 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1333 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1334 if (DOINGSOFTDEP(tvp)) {
1335 /*
1336 * Ensure that the entire newly allocated block is a
1337 * valid directory so that future growth within the
1338 * block does not have to ensure that the block is
1339 * written before the inode.
1340 */
1341 blkoff = DIRBLKSIZ;
1342 while (blkoff < bp->b_bcount) {
1343 ((struct direct *)
1344 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1345 blkoff += DIRBLKSIZ;
1346 }
1347 }
1348 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1349 DOINGASYNC(tvp)))) != 0) {
1350 (void)VOP_BWRITE(bp->b_vp, bp);
1351 goto bad;
1352 }
1353 /*
1354 * Directory set up, now install its entry in the parent directory.
1355 *
1356 * If we are not doing soft dependencies, then we must write out the
1357 * buffer containing the new directory body before entering the new
1358 * name in the parent. If we are doing soft dependencies, then the
1359 * buffer containing the new directory body will be passed to and
1360 * released in the soft dependency code after the code has attached
1361 * an appropriate ordering dependency to the buffer which ensures that
1362 * the buffer is written before the new name is written in the parent.
1363 */
1364 if (DOINGASYNC(dvp))
1365 bdwrite(bp);
1366 else if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp->b_vp, bp))))
1367 goto bad;
1368 ufs_makedirentry(ip, cnp, &newdir);
1369 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1370
1371bad:
1372 if (error == 0) {
1373 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1374 *ap->a_vpp = tvp;
1375 } else {
1376 dp->i_effnlink--;
1377 dp->i_nlink--;
1378 dp->i_flag |= IN_CHANGE;
1379 if (DOINGSOFTDEP(dvp))
1380 softdep_change_linkcnt(dp);
1381 /*
1382 * No need to do an explicit VOP_TRUNCATE here, vrele will
1383 * do this for us because we set the link count to 0.
1384 */
1385 ip->i_effnlink = 0;
1386 ip->i_nlink = 0;
1387 ip->i_flag |= IN_CHANGE;
1388 if (DOINGSOFTDEP(tvp))
1389 softdep_change_linkcnt(ip);
1390 vput(tvp);
1391 }
1392out:
1393 return (error);
1394}
1395
1396/*
1397 * Rmdir system call.
0973c589
CP
1398 *
1399 * ufs_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1400 * struct componentname *a_cnp)
984263bc
MD
1401 */
1402int
0973c589 1403ufs_rmdir(struct vop_rmdir_args *ap)
984263bc
MD
1404{
1405 struct vnode *vp = ap->a_vp;
1406 struct vnode *dvp = ap->a_dvp;
1407 struct componentname *cnp = ap->a_cnp;
1408 struct inode *ip, *dp;
1409 int error, ioflag;
1410
1411 ip = VTOI(vp);
1412 dp = VTOI(dvp);
1413
1414 /*
1415 * Do not remove a directory that is in the process of being renamed.
1416 * Verify the directory is empty (and valid). Rmdir ".." will not be
1417 * valid since ".." will contain a reference to the current directory
1418 * and thus be non-empty. Do not allow the removal of mounted on
1419 * directories (this can happen when an NFS exported filesystem
1420 * tries to remove a locally mounted on directory).
1421 */
1422 error = 0;
1423 if (ip->i_flag & IN_RENAME) {
1424 error = EINVAL;
1425 goto out;
1426 }
1427 if (ip->i_effnlink != 2 ||
1428 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1429 error = ENOTEMPTY;
1430 goto out;
1431 }
1432 if ((dp->i_flags & APPEND)
1433 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1434 error = EPERM;
1435 goto out;
1436 }
1437 if (vp->v_mountedhere != 0) {
1438 error = EINVAL;
1439 goto out;
1440 }
1441 /*
1442 * Delete reference to directory before purging
1443 * inode. If we crash in between, the directory
1444 * will be reattached to lost+found,
1445 */
1446 dp->i_effnlink--;
1447 ip->i_effnlink--;
1448 if (DOINGSOFTDEP(vp)) {
1449 softdep_change_linkcnt(dp);
1450 softdep_change_linkcnt(ip);
1451 }
1452 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1453 if (error) {
1454 dp->i_effnlink++;
1455 ip->i_effnlink++;
1456 if (DOINGSOFTDEP(vp)) {
1457 softdep_change_linkcnt(dp);
1458 softdep_change_linkcnt(ip);
1459 }
1460 goto out;
1461 }
1462 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1463 cache_purge(dvp);
1464 /*
1465 * Truncate inode. The only stuff left in the directory is "." and
1466 * "..". The "." reference is inconsequential since we are quashing
1467 * it. The soft dependency code will arrange to do these operations
1468 * after the parent directory entry has been deleted on disk, so
1469 * when running with that code we avoid doing them now.
1470 */
1471 if (!DOINGSOFTDEP(vp)) {
1472 dp->i_nlink--;
1473 dp->i_flag |= IN_CHANGE;
1474 ip->i_nlink--;
1475 ip->i_flag |= IN_CHANGE;
1476 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC;
1477 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
dadab5e9 1478 cnp->cn_td);
984263bc
MD
1479 }
1480 cache_purge(vp);
1481#ifdef UFS_DIRHASH
1482 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
1483 if (ip->i_dirhash != NULL)
1484 ufsdirhash_free(ip);
1485#endif
1486out:
1487 VN_KNOTE(vp, NOTE_DELETE);
1488 return (error);
1489}
1490
1491/*
1492 * symlink -- make a symbolic link
0973c589
CP
1493 *
1494 * ufs_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1495 * struct componentname *a_cnp, struct vattr *a_vap,
1496 * char *a_target)
984263bc
MD
1497 */
1498int
0973c589 1499ufs_symlink(struct vop_symlink_args *ap)
984263bc 1500{
3ff2135f
RG
1501 struct vnode *vp, **vpp = ap->a_vpp;
1502 struct inode *ip;
984263bc
MD
1503 int len, error;
1504
1505 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1506 vpp, ap->a_cnp);
1507 if (error)
1508 return (error);
1509 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1510 vp = *vpp;
1511 len = strlen(ap->a_target);
1512 if (len < vp->v_mount->mnt_maxsymlinklen) {
1513 ip = VTOI(vp);
1514 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1515 ip->i_size = len;
1516 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1517 } else
1518 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
dadab5e9
MD
1519 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred,
1520 (int *)0, NULL);
984263bc
MD
1521 if (error)
1522 vput(vp);
1523 return (error);
1524}
1525
1526/*
1527 * Vnode op for reading directories.
1528 *
1529 * The routine below assumes that the on-disk format of a directory
1530 * is the same as that defined by <sys/dirent.h>. If the on-disk
1531 * format changes, then it will be necessary to do a conversion
1532 * from the on-disk format that read returns to the format defined
1533 * by <sys/dirent.h>.
0973c589
CP
1534 *
1535 * ufs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
1536 * int *a_eofflag, int *ncookies, u_long **a_cookies)
984263bc
MD
1537 */
1538int
0973c589 1539ufs_readdir(struct vop_readdir_args *ap)
984263bc 1540{
3ff2135f 1541 struct uio *uio = ap->a_uio;
984263bc
MD
1542 int error;
1543 size_t count, lost;
1544 off_t off;
1545
1546 if (ap->a_ncookies != NULL)
1547 /*
1548 * Ensure that the block is aligned. The caller can use
1549 * the cookies to determine where in the block to start.
1550 */
1551 uio->uio_offset &= ~(DIRBLKSIZ - 1);
1552 off = uio->uio_offset;
1553 count = uio->uio_resid;
1554 /* Make sure we don't return partial entries. */
1555 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
1556 return (EINVAL);
1557 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1558 lost = uio->uio_resid - count;
1559 uio->uio_resid = count;
1560 uio->uio_iov->iov_len = count;
1561# if (BYTE_ORDER == LITTLE_ENDIAN)
1562 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
1563 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1564 } else {
1565 struct dirent *dp, *edp;
1566 struct uio auio;
1567 struct iovec aiov;
1568 caddr_t dirbuf;
1569 int readcnt;
1570 u_char tmp;
1571
1572 auio = *uio;
1573 auio.uio_iov = &aiov;
1574 auio.uio_iovcnt = 1;
1575 auio.uio_segflg = UIO_SYSSPACE;
1576 aiov.iov_len = count;
1577 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1578 aiov.iov_base = dirbuf;
1579 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
1580 if (error == 0) {
1581 readcnt = count - auio.uio_resid;
1582 edp = (struct dirent *)&dirbuf[readcnt];
1583 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
1584 tmp = dp->d_namlen;
1585 dp->d_namlen = dp->d_type;
1586 dp->d_type = tmp;
1587 if (dp->d_reclen > 0) {
1588 dp = (struct dirent *)
1589 ((char *)dp + dp->d_reclen);
1590 } else {
1591 error = EIO;
1592 break;
1593 }
1594 }
1595 if (dp >= edp)
1596 error = uiomove(dirbuf, readcnt, uio);
1597 }
1598 FREE(dirbuf, M_TEMP);
1599 }
1600# else
1601 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1602# endif
1603 if (!error && ap->a_ncookies != NULL) {
1604 struct dirent* dpStart;
1605 struct dirent* dpEnd;
1606 struct dirent* dp;
1607 int ncookies;
1608 u_long *cookies;
1609 u_long *cookiep;
1610
1611 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1612 panic("ufs_readdir: unexpected uio from NFS server");
1613 dpStart = (struct dirent *)
1614 (uio->uio_iov->iov_base - (uio->uio_offset - off));
1615 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
1616 for (dp = dpStart, ncookies = 0;
1617 dp < dpEnd;
1618 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
1619 ncookies++;
1620 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1621 M_WAITOK);
1622 for (dp = dpStart, cookiep = cookies;
1623 dp < dpEnd;
1624 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
1625 off += dp->d_reclen;
1626 *cookiep++ = (u_long) off;
1627 }
1628 *ap->a_ncookies = ncookies;
1629 *ap->a_cookies = cookies;
1630 }
1631 uio->uio_resid += lost;
1632 if (ap->a_eofflag)
1633 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
1634 return (error);
1635}
1636
1637/*
1638 * Return target name of a symbolic link
0973c589
CP
1639 *
1640 * ufs_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
984263bc
MD
1641 */
1642int
0973c589 1643ufs_readlink(struct vop_readlink_args *ap)
984263bc 1644{
3ff2135f
RG
1645 struct vnode *vp = ap->a_vp;
1646 struct inode *ip = VTOI(vp);
984263bc
MD
1647 int isize;
1648
1649 isize = ip->i_size;
1650 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
f719c866 1651 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
984263bc
MD
1652 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1653 return (0);
1654 }
1655 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1656}
1657
1658/*
1659 * Calculate the logical to physical mapping if not done already,
1660 * then call the device strategy routine.
1661 *
1662 * In order to be able to swap to a file, the VOP_BMAP operation may not
1663 * deadlock on memory. See ufs_bmap() for details.
0973c589
CP
1664 *
1665 * ufs_strategy(struct vnode *a_vp, struct buf *a_bp)
984263bc
MD
1666 */
1667int
0973c589 1668ufs_strategy(struct vop_strategy_args *ap)
984263bc 1669{
3ff2135f
RG
1670 struct buf *bp = ap->a_bp;
1671 struct vnode *vp = ap->a_vp;
1672 struct inode *ip;
984263bc
MD
1673 int error;
1674
1675 ip = VTOI(vp);
1676 if (vp->v_type == VBLK || vp->v_type == VCHR)
1677 panic("ufs_strategy: spec");
1678 if (bp->b_blkno == bp->b_lblkno) {
1679 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
1680 if (error) {
1681 bp->b_error = error;
1682 bp->b_flags |= B_ERROR;
1683 biodone(bp);
1684 return (error);
1685 }
1686 if ((long)bp->b_blkno == -1)
1687 vfs_bio_clrbuf(bp);
1688 }
1689 if ((long)bp->b_blkno == -1) {
1690 biodone(bp);
1691 return (0);
1692 }
1693 vp = ip->i_devvp;
1694 bp->b_dev = vp->v_rdev;
1695 VOP_STRATEGY(vp, bp);
1696 return (0);
1697}
1698
1699/*
1700 * Print out the contents of an inode.
0973c589
CP
1701 *
1702 * ufs_print(struct vnode *a_vp)
984263bc
MD
1703 */
1704int
0973c589 1705ufs_print(struct vop_print_args *ap)
984263bc 1706{
3ff2135f
RG
1707 struct vnode *vp = ap->a_vp;
1708 struct inode *ip = VTOI(vp);
984263bc
MD
1709
1710 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)",
1711 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1712 minor(ip->i_dev));
1713 if (vp->v_type == VFIFO)
1714 fifo_printinfo(vp);
1715 lockmgr_printinfo(&ip->i_lock);
1716 printf("\n");
1717 return (0);
1718}
1719
1720/*
1721 * Read wrapper for special devices.
0973c589
CP
1722 *
1723 * ufsspec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1724 * struct ucred *a_cred)
984263bc
MD
1725 */
1726int
0973c589 1727ufsspec_read(struct vop_read_args *ap)
984263bc
MD
1728{
1729 int error, resid;
1730 struct inode *ip;
1731 struct uio *uio;
1732
1733 uio = ap->a_uio;
1734 resid = uio->uio_resid;
2d3e977e 1735 error = VOCALL(spec_vnode_vops, VOFFSET(vop_read), &ap->a_head);
984263bc
MD
1736 /*
1737 * The inode may have been revoked during the call, so it must not
1738 * be accessed blindly here or in the other wrapper functions.
1739 */
1740 ip = VTOI(ap->a_vp);
1741 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1742 ip->i_flag |= IN_ACCESS;
1743 return (error);
1744}
1745
1746/*
1747 * Write wrapper for special devices.
0973c589
CP
1748 *
1749 * ufsspec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1750 * struct ucred *a_cred)
984263bc
MD
1751 */
1752int
0973c589 1753ufsspec_write(struct vop_write_args *ap)
984263bc
MD
1754{
1755 int error, resid;
1756 struct inode *ip;
1757 struct uio *uio;
1758
1759 uio = ap->a_uio;
1760 resid = uio->uio_resid;
2d3e977e 1761 error = VOCALL(spec_vnode_vops, VOFFSET(vop_write), &ap->a_head);
984263bc
MD
1762 ip = VTOI(ap->a_vp);
1763 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1764 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1765 return (error);
1766}
1767
1768/*
1769 * Close wrapper for special devices.
1770 *
1771 * Update the times on the inode then do device close.
0973c589
CP
1772 *
1773 * ufsspec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1774 * struct thread *a_td)
984263bc
MD
1775 */
1776int
0973c589 1777ufsspec_close(struct vop_close_args *ap)
984263bc
MD
1778{
1779 struct vnode *vp = ap->a_vp;
41a01a4d 1780 lwkt_tokref vlock;
984263bc 1781
41a01a4d 1782 lwkt_gettoken(&vlock, vp->v_interlock);
984263bc
MD
1783 if (vp->v_usecount > 1)
1784 ufs_itimes(vp);
41a01a4d 1785 lwkt_reltoken(&vlock);
2d3e977e 1786 return (VOCALL(spec_vnode_vops, VOFFSET(vop_close), &ap->a_head));
984263bc
MD
1787}
1788
1789/*
1790 * Read wrapper for fifos.
0973c589
CP
1791 *
1792 * ufsfifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1793 * struct ucred *a_cred)
984263bc
MD
1794 */
1795int
0973c589 1796ufsfifo_read(struct vop_read_args *ap)
984263bc
MD
1797{
1798 int error, resid;
1799 struct inode *ip;
1800 struct uio *uio;
1801
1802 uio = ap->a_uio;
1803 resid = uio->uio_resid;
2d3e977e 1804 error = VOCALL(fifo_vnode_vops, VOFFSET(vop_read), &ap->a_head);
984263bc
MD
1805 ip = VTOI(ap->a_vp);
1806 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1807 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1808 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1809 return (error);
1810}
1811
1812/*
1813 * Write wrapper for fifos.
0973c589
CP
1814 *
1815 * ufsfifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1816 * struct ucred *a_cred)
984263bc
MD
1817 */
1818int
0973c589 1819ufsfifo_write(struct vop_write_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;
2d3e977e 1827 error = VOCALL(fifo_vnode_vops, VOFFSET(vop_write), &ap->a_head);
984263bc
MD
1828 ip = VTOI(ap->a_vp);
1829 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1830 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1831 return (error);
1832}
1833
1834/*
1835 * Close wrapper for fifos.
1836 *
1837 * Update the times on the inode then do device close.
0973c589
CP
1838 *
1839 * ufsfifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred,
1840 * struct thread *a_td)
984263bc
MD
1841 */
1842int
0973c589 1843ufsfifo_close(struct vop_close_args *ap)
984263bc
MD
1844{
1845 struct vnode *vp = ap->a_vp;
41a01a4d 1846 lwkt_tokref vlock;
984263bc 1847
41a01a4d 1848 lwkt_gettoken(&vlock, vp->v_interlock);
984263bc
MD
1849 if (vp->v_usecount > 1)
1850 ufs_itimes(vp);
41a01a4d 1851 lwkt_reltoken(&vlock);
2d3e977e 1852 return (VOCALL(fifo_vnode_vops, VOFFSET(vop_close), &ap->a_head));
984263bc
MD
1853}
1854
1855/*
1856 * Kqfilter wrapper for fifos.
1857 *
1858 * Fall through to ufs kqfilter routines if needed
1859 */
1860int
0973c589 1861ufsfifo_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
1862{
1863 int error;
1864
2d3e977e 1865 error = VOCALL(fifo_vnode_vops, VOFFSET(vop_kqfilter), &ap->a_head);
984263bc
MD
1866 if (error)
1867 error = ufs_kqfilter(ap);
1868 return (error);
1869}
1870
1871/*
1872 * Return POSIX pathconf information applicable to ufs filesystems.
0973c589
CP
1873 *
1874 * ufs_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
984263bc
MD
1875 */
1876int
0973c589 1877ufs_pathconf(struct vop_pathconf_args *ap)
984263bc 1878{
984263bc
MD
1879 switch (ap->a_name) {
1880 case _PC_LINK_MAX:
1881 *ap->a_retval = LINK_MAX;
1882 return (0);
1883 case _PC_NAME_MAX:
1884 *ap->a_retval = NAME_MAX;
1885 return (0);
1886 case _PC_PATH_MAX:
1887 *ap->a_retval = PATH_MAX;
1888 return (0);
1889 case _PC_PIPE_BUF:
1890 *ap->a_retval = PIPE_BUF;
1891 return (0);
1892 case _PC_CHOWN_RESTRICTED:
1893 *ap->a_retval = 1;
1894 return (0);
1895 case _PC_NO_TRUNC:
1896 *ap->a_retval = 1;
1897 return (0);
1898 default:
1899 return (EINVAL);
1900 }
1901 /* NOTREACHED */
1902}
1903
1904/*
1905 * Advisory record locking support
0973c589
CP
1906 *
1907 * ufs_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
1908 * int a_flags)
984263bc
MD
1909 */
1910int
0973c589 1911ufs_advlock(struct vop_advlock_args *ap)
984263bc 1912{
3ff2135f 1913 struct inode *ip = VTOI(ap->a_vp);
984263bc
MD
1914
1915 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1916}
1917
1918/*
1919 * Initialize the vnode associated with a new inode, handle aliased
1920 * vnodes.
1921 */
1922int
2d3e977e 1923ufs_vinit(struct mount *mntp, struct vop_ops *specops, struct vop_ops *fifoops,
0973c589 1924 struct vnode **vpp)
984263bc
MD
1925{
1926 struct inode *ip;
1927 struct vnode *vp;
1928 struct timeval tv;
1929
1930 vp = *vpp;
1931 ip = VTOI(vp);
1932 switch(vp->v_type = IFTOVT(ip->i_mode)) {
1933 case VCHR:
1934 case VBLK:
2d3e977e 1935 vp->v_vops = specops;
984263bc
MD
1936 addaliasu(vp, ip->i_rdev);
1937 break;
1938 case VFIFO:
2d3e977e 1939 vp->v_vops = fifoops;
984263bc
MD
1940 break;
1941 default:
1942 break;
1943
1944 }
1945 if (ip->i_number == ROOTINO)
1946 vp->v_flag |= VROOT;
1947 /*
1948 * Initialize modrev times
1949 */
1950 getmicrouptime(&tv);
1951 SETHIGH(ip->i_modrev, tv.tv_sec);
1952 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
1953 *vpp = vp;
1954 return (0);
1955}
1956
1957/*
1958 * Allocate a new inode.
1959 */
1960int
0973c589
CP
1961ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1962 struct componentname *cnp)
984263bc 1963{
3ff2135f 1964 struct inode *ip, *pdir;
984263bc
MD
1965 struct direct newdir;
1966 struct vnode *tvp;
1967 int error;
1968
1969 pdir = VTOI(dvp);
1970#ifdef DIAGNOSTIC
2b69e610 1971 if ((cnp->cn_flags & CNP_HASBUF) == 0)
984263bc
MD
1972 panic("ufs_makeinode: no name");
1973#endif
1974 *vpp = NULL;
1975 if ((mode & IFMT) == 0)
1976 mode |= IFREG;
1977
1978 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
1979 if (error)
1980 return (error);
1981 ip = VTOI(tvp);
1982 ip->i_gid = pdir->i_gid;
1983#ifdef SUIDDIR
1984 {
1985#ifdef QUOTA
1986 struct ucred ucred, *ucp;
1987 ucp = cnp->cn_cred;
1988#endif
1989 /*
1990 * If we are not the owner of the directory,
1991 * and we are hacking owners here, (only do this where told to)
1992 * and we are not giving it TO root, (would subvert quotas)
1993 * then go ahead and give it to the other user.
1994 * Note that this drops off the execute bits for security.
1995 */
1996 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1997 (pdir->i_mode & ISUID) &&
1998 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1999 ip->i_uid = pdir->i_uid;
2000 mode &= ~07111;
2001#ifdef QUOTA
2002 /*
2003 * Make sure the correct user gets charged
2004 * for the space.
2005 * Quickly knock up a dummy credential for the victim.
2006 * XXX This seems to never be accessed out of our
2007 * context so a stack variable is ok.
2008 */
2009 ucred.cr_ref = 1;
2010 ucred.cr_uid = ip->i_uid;
2011 ucred.cr_ngroups = 1;
2012 ucred.cr_groups[0] = pdir->i_gid;
2013 ucp = &ucred;
2014#endif
2015 } else
2016 ip->i_uid = cnp->cn_cred->cr_uid;
2017
2018#ifdef QUOTA
2019 if ((error = getinoquota(ip)) ||
2020 (error = chkiq(ip, 1, ucp, 0))) {
2021 UFS_VFREE(tvp, ip->i_number, mode);
2022 vput(tvp);
2023 return (error);
2024 }
2025#endif
2026 }
2027#else /* !SUIDDIR */
2028 ip->i_uid = cnp->cn_cred->cr_uid;
2029#ifdef QUOTA
2030 if ((error = getinoquota(ip)) ||
2031 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2032 UFS_VFREE(tvp, ip->i_number, mode);
2033 vput(tvp);
2034 return (error);
2035 }
2036#endif
2037#endif /* !SUIDDIR */
2038 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2039 ip->i_mode = mode;
2040 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2041 ip->i_effnlink = 1;
2042 ip->i_nlink = 1;
2043 if (DOINGSOFTDEP(tvp))
2044 softdep_change_linkcnt(ip);
2045 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
dadab5e9 2046 suser_cred(cnp->cn_cred, 0)) {
984263bc 2047 ip->i_mode &= ~ISGID;
41c20dac 2048 }
984263bc 2049
2b69e610 2050 if (cnp->cn_flags & CNP_ISWHITEOUT)
984263bc
MD
2051 ip->i_flags |= UF_OPAQUE;
2052
2053 /*
2054 * Make sure inode goes to disk before directory entry.
2055 */
2056 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2057 if (error)
2058 goto bad;
2059 ufs_makedirentry(ip, cnp, &newdir);
2060 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2061 if (error)
2062 goto bad;
2063 *vpp = tvp;
2064 return (0);
2065
2066bad:
2067 /*
2068 * Write error occurred trying to update the inode
2069 * or the directory so must deallocate the inode.
2070 */
2071 ip->i_effnlink = 0;
2072 ip->i_nlink = 0;
2073 ip->i_flag |= IN_CHANGE;
2074 if (DOINGSOFTDEP(tvp))
2075 softdep_change_linkcnt(ip);
2076 vput(tvp);
2077 return (error);
2078}
2079
2080static int
0973c589 2081ufs_missingop(struct vop_generic_args *ap)
984263bc 2082{
984263bc
MD
2083 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2084 return (EOPNOTSUPP);
2085}
2086
2087static struct filterops ufsread_filtops =
2088 { 1, NULL, filt_ufsdetach, filt_ufsread };
2089static struct filterops ufswrite_filtops =
2090 { 1, NULL, filt_ufsdetach, filt_ufswrite };
2091static struct filterops ufsvnode_filtops =
2092 { 1, NULL, filt_ufsdetach, filt_ufsvnode };
2093
0973c589
CP
2094/*
2095 * ufs_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2096 */
984263bc 2097static int
0973c589 2098ufs_kqfilter(struct vop_kqfilter_args *ap)
984263bc
MD
2099{
2100 struct vnode *vp = ap->a_vp;
2101 struct knote *kn = ap->a_kn;
41a01a4d 2102 lwkt_tokref ilock;
984263bc
MD
2103
2104 switch (kn->kn_filter) {
2105 case EVFILT_READ:
2106 kn->kn_fop = &ufsread_filtops;
2107 break;
2108 case EVFILT_WRITE:
2109 kn->kn_fop = &ufswrite_filtops;
2110 break;
2111 case EVFILT_VNODE:
2112 kn->kn_fop = &ufsvnode_filtops;
2113 break;
2114 default:
2115 return (1);
2116 }
2117
2118 kn->kn_hook = (caddr_t)vp;
2119
41a01a4d 2120 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc 2121 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
41a01a4d 2122 lwkt_reltoken(&ilock);
984263bc
MD
2123
2124 return (0);
2125}
2126
2127static void
2128filt_ufsdetach(struct knote *kn)
2129{
2130 struct vnode *vp = (struct vnode *)kn->kn_hook;
41a01a4d 2131 lwkt_tokref ilock;
984263bc 2132
41a01a4d 2133 lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
984263bc
MD
2134 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2135 kn, knote, kn_selnext);
41a01a4d 2136 lwkt_reltoken(&ilock);
984263bc
MD
2137}
2138
2139/*ARGSUSED*/
2140static int
2141filt_ufsread(struct knote *kn, long hint)
2142{
2143 struct vnode *vp = (struct vnode *)kn->kn_hook;
2144 struct inode *ip = VTOI(vp);
2145
2146 /*
2147 * filesystem is gone, so set the EOF flag and schedule
2148 * the knote for deletion.
2149 */
2150 if (hint == NOTE_REVOKE) {
2151 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2152 return (1);
2153 }
2154
2155 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2156 return (kn->kn_data != 0);
2157}
2158
2159/*ARGSUSED*/
2160static int
2161filt_ufswrite(struct knote *kn, long hint)
2162{
984263bc
MD
2163 /*
2164 * filesystem is gone, so set the EOF flag and schedule
2165 * the knote for deletion.
2166 */
2167 if (hint == NOTE_REVOKE)
2168 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2169
2170 kn->kn_data = 0;
2171 return (1);
2172}
2173
2174static int
2175filt_ufsvnode(struct knote *kn, long hint)
2176{
984263bc
MD
2177 if (kn->kn_sfflags & hint)
2178 kn->kn_fflags |= hint;
2179 if (hint == NOTE_REVOKE) {
2180 kn->kn_flags |= EV_EOF;
2181 return (1);
2182 }
2183 return (kn->kn_fflags != 0);
2184}
2185
2186/* Global vfs data structures for ufs. */
2d3e977e 2187static struct vop_ops *ufs_vnode_vops;
984263bc 2188static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2d3e977e
MD
2189 { &vop_default_desc, vop_defaultop },
2190 { &vop_fsync_desc, (void *) ufs_missingop },
2191 { &vop_read_desc, (void *) ufs_missingop },
2192 { &vop_reallocblks_desc, (void *) ufs_missingop },
2193 { &vop_write_desc, (void *) ufs_missingop },
2194 { &vop_access_desc, (void *) ufs_access },
2195 { &vop_advlock_desc, (void *) ufs_advlock },
2196 { &vop_bmap_desc, (void *) ufs_bmap },
2197 { &vop_cachedlookup_desc, (void *) ufs_lookup },
2198 { &vop_close_desc, (void *) ufs_close },
2199 { &vop_create_desc, (void *) ufs_create },
2200 { &vop_getattr_desc, (void *) ufs_getattr },
2201 { &vop_inactive_desc, (void *) ufs_inactive },
2202 { &vop_islocked_desc, (void *) vop_stdislocked },
2203 { &vop_link_desc, (void *) ufs_link },
2204 { &vop_lock_desc, (void *) vop_stdlock },
2205 { &vop_lookup_desc, (void *) vfs_cache_lookup },
2206 { &vop_mkdir_desc, (void *) ufs_mkdir },
2207 { &vop_mknod_desc, (void *) ufs_mknod },
2208 { &vop_mmap_desc, (void *) ufs_mmap },
2209 { &vop_open_desc, (void *) ufs_open },
2210 { &vop_pathconf_desc, (void *) ufs_pathconf },
2211 { &vop_poll_desc, (void *) vop_stdpoll },
2212 { &vop_kqfilter_desc, (void *) ufs_kqfilter },
2213 { &vop_print_desc, (void *) ufs_print },
2214 { &vop_readdir_desc, (void *) ufs_readdir },
2215 { &vop_readlink_desc, (void *) ufs_readlink },
2216 { &vop_reclaim_desc, (void *) ufs_reclaim },
2217 { &vop_remove_desc, (void *) ufs_remove },
2218 { &vop_rename_desc, (void *) ufs_rename },
2219 { &vop_rmdir_desc, (void *) ufs_rmdir },
2220 { &vop_setattr_desc, (void *) ufs_setattr },
2221 { &vop_strategy_desc, (void *) ufs_strategy },
2222 { &vop_symlink_desc, (void *) ufs_symlink },
2223 { &vop_unlock_desc, (void *) vop_stdunlock },
2224 { &vop_whiteout_desc, (void *) ufs_whiteout },
984263bc
MD
2225 { NULL, NULL }
2226};
2227static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2d3e977e 2228 { &ufs_vnode_vops, ufs_vnodeop_entries };
984263bc 2229
2d3e977e 2230static struct vop_ops *ufs_spec_vops;
984263bc 2231static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2d3e977e
MD
2232 { &vop_default_desc, (void *) spec_vnoperate },
2233 { &vop_fsync_desc, (void *) ufs_missingop },
2234 { &vop_access_desc, (void *) ufs_access },
2235 { &vop_close_desc, (void *) ufsspec_close },
2236 { &vop_getattr_desc, (void *) ufs_getattr },
2237 { &vop_inactive_desc, (void *) ufs_inactive },
2238 { &vop_islocked_desc, (void *) vop_stdislocked },
2239 { &vop_lock_desc, (void *) vop_stdlock },
2240 { &vop_print_desc, (void *) ufs_print },
2241 { &vop_read_desc, (void *) ufsspec_read },
2242 { &vop_reclaim_desc, (void *) ufs_reclaim },
2243 { &vop_setattr_desc, (void *) ufs_setattr },
2244 { &vop_unlock_desc, (void *) vop_stdunlock },
2245 { &vop_write_desc, (void *) ufsspec_write },
984263bc
MD
2246 { NULL, NULL }
2247};
2248static struct vnodeopv_desc ufs_specop_opv_desc =
2d3e977e 2249 { &ufs_spec_vops, ufs_specop_entries };
984263bc 2250
2d3e977e 2251static struct vop_ops *ufs_fifo_vops;
984263bc 2252static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2d3e977e
MD
2253 { &vop_default_desc, (void *) fifo_vnoperate },
2254 { &vop_fsync_desc, (void *) ufs_missingop },
2255 { &vop_access_desc, (void *) ufs_access },
2256 { &vop_close_desc, (void *) ufsfifo_close },
2257 { &vop_getattr_desc, (void *) ufs_getattr },
2258 { &vop_inactive_desc, (void *) ufs_inactive },
2259 { &vop_islocked_desc, (void *) vop_stdislocked },
2260 { &vop_kqfilter_desc, (void *) ufsfifo_kqfilter },
2261 { &vop_lock_desc, (void *) vop_stdlock },
2262 { &vop_print_desc, (void *) ufs_print },
2263 { &vop_read_desc, (void *) ufsfifo_read },
2264 { &vop_reclaim_desc, (void *) ufs_reclaim },
2265 { &vop_setattr_desc, (void *) ufs_setattr },
2266 { &vop_unlock_desc, (void *) vop_stdunlock },
2267 { &vop_write_desc, (void *) ufsfifo_write },
984263bc
MD
2268 { NULL, NULL }
2269};
2270static struct vnodeopv_desc ufs_fifoop_opv_desc =
2d3e977e 2271 { &ufs_fifo_vops, ufs_fifoop_entries };
984263bc
MD
2272
2273VNODEOP_SET(ufs_vnodeop_opv_desc);
2274VNODEOP_SET(ufs_specop_opv_desc);
2275VNODEOP_SET(ufs_fifoop_opv_desc);
2276
0973c589
CP
2277/*
2278 * ufs_vnoperate(struct vnodeop_desc *a_desc)
2279 */
984263bc 2280int
0973c589 2281ufs_vnoperate(struct vop_generic_args *ap)
984263bc 2282{
2d3e977e 2283 return (VOCALL(ufs_vnode_vops, ap->a_desc->vdesc_offset, ap));
984263bc
MD
2284}
2285
0973c589
CP
2286/*
2287 * ufs_vnoperatefifo(struct vnodeop_desc *a_desc)
2288 */
984263bc 2289int
0973c589 2290ufs_vnoperatefifo(struct vop_generic_args *ap)
984263bc 2291{
2d3e977e 2292 return (VOCALL(ufs_fifo_vops, ap->a_desc->vdesc_offset, ap));
984263bc
MD
2293}
2294
0973c589
CP
2295/*
2296 * ufs_vnoperatespec(struct vnodeop_desc *a_desc)
2297 */
984263bc 2298int
0973c589 2299ufs_vnoperatespec(struct vop_generic_args *ap)
984263bc 2300{
2d3e977e 2301 return (VOCALL(ufs_spec_vops, ap->a_desc->vdesc_offset, ap));
984263bc 2302}