MP Implementation 1/2: Get the APIC code working again, sweetly integrate the
[dragonfly.git] / sys / vfs / ufs / ufs_vnops.c
... / ...
CommitLineData
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 $
40 * $DragonFly: src/sys/vfs/ufs/ufs_vnops.c,v 1.6 2003/07/06 21:23:55 dillon Exp $
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>
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>
54#include <sys/namei.h>
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
69#include <miscfs/fifofs/fifo.h>
70
71#include <ufs/ufs/quota.h>
72#include <ufs/ufs/inode.h>
73#include <ufs/ufs/dir.h>
74#include <ufs/ufs/ufsmount.h>
75#include <ufs/ufs/ufs_extern.h>
76#ifdef UFS_DIRHASH
77#include <ufs/ufs/dirhash.h>
78#endif
79
80static int ufs_access __P((struct vop_access_args *));
81static int ufs_advlock __P((struct vop_advlock_args *));
82static int ufs_chmod __P((struct vnode *, int, struct ucred *, struct thread *));
83static int ufs_chown __P((struct vnode *, uid_t, gid_t, struct ucred *, struct thread *));
84static int ufs_close __P((struct vop_close_args *));
85static int ufs_create __P((struct vop_create_args *));
86static int ufs_getattr __P((struct vop_getattr_args *));
87static int ufs_link __P((struct vop_link_args *));
88static int ufs_makeinode __P((int mode, struct vnode *, struct vnode **, struct componentname *));
89static int ufs_missingop __P((struct vop_generic_args *ap));
90static int ufs_mkdir __P((struct vop_mkdir_args *));
91static int ufs_mknod __P((struct vop_mknod_args *));
92static int ufs_mmap __P((struct vop_mmap_args *));
93static int ufs_open __P((struct vop_open_args *));
94static int ufs_pathconf __P((struct vop_pathconf_args *));
95static int ufs_print __P((struct vop_print_args *));
96static int ufs_readdir __P((struct vop_readdir_args *));
97static int ufs_readlink __P((struct vop_readlink_args *));
98static int ufs_remove __P((struct vop_remove_args *));
99static int ufs_rename __P((struct vop_rename_args *));
100static int ufs_rmdir __P((struct vop_rmdir_args *));
101static int ufs_setattr __P((struct vop_setattr_args *));
102static int ufs_strategy __P((struct vop_strategy_args *));
103static int ufs_symlink __P((struct vop_symlink_args *));
104static int ufs_whiteout __P((struct vop_whiteout_args *));
105static int ufsfifo_close __P((struct vop_close_args *));
106static int ufsfifo_kqfilter __P((struct vop_kqfilter_args *));
107static int ufsfifo_read __P((struct vop_read_args *));
108static int ufsfifo_write __P((struct vop_write_args *));
109static int ufsspec_close __P((struct vop_close_args *));
110static int ufsspec_read __P((struct vop_read_args *));
111static int ufsspec_write __P((struct vop_write_args *));
112static int filt_ufsread __P((struct knote *kn, long hint));
113static int filt_ufswrite __P((struct knote *kn, long hint));
114static int filt_ufsvnode __P((struct knote *kn, long hint));
115static void filt_ufsdetach __P((struct knote *kn));
116static int ufs_kqfilter __P((struct vop_kqfilter_args *ap));
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
150ufs_itimes(vp)
151 struct vnode *vp;
152{
153 struct inode *ip;
154 struct timespec ts;
155
156 ip = VTOI(vp);
157 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
158 return;
159 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
160 ip->i_flag |= IN_LAZYMOD;
161 else
162 ip->i_flag |= IN_MODIFIED;
163 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
164 vfs_timestamp(&ts);
165 if (ip->i_flag & IN_ACCESS) {
166 ip->i_atime = ts.tv_sec;
167 ip->i_atimensec = ts.tv_nsec;
168 }
169 if (ip->i_flag & IN_UPDATE) {
170 ip->i_mtime = ts.tv_sec;
171 ip->i_mtimensec = ts.tv_nsec;
172 ip->i_modrev++;
173 }
174 if (ip->i_flag & IN_CHANGE) {
175 ip->i_ctime = ts.tv_sec;
176 ip->i_ctimensec = ts.tv_nsec;
177 }
178 }
179 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
180}
181
182/*
183 * Create a regular file
184 */
185int
186ufs_create(ap)
187 struct vop_create_args /* {
188 struct vnode *a_dvp;
189 struct vnode **a_vpp;
190 struct componentname *a_cnp;
191 struct vattr *a_vap;
192 } */ *ap;
193{
194 int error;
195
196 error =
197 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
198 ap->a_dvp, ap->a_vpp, ap->a_cnp);
199 if (error)
200 return (error);
201 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
202 return (0);
203}
204
205/*
206 * Mknod vnode call
207 */
208/* ARGSUSED */
209int
210ufs_mknod(ap)
211 struct vop_mknod_args /* {
212 struct vnode *a_dvp;
213 struct vnode **a_vpp;
214 struct componentname *a_cnp;
215 struct vattr *a_vap;
216 } */ *ap;
217{
218 struct vattr *vap = ap->a_vap;
219 struct vnode **vpp = ap->a_vpp;
220 struct inode *ip;
221 ino_t ino;
222 int error;
223
224 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
225 ap->a_dvp, vpp, ap->a_cnp);
226 if (error)
227 return (error);
228 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
229 ip = VTOI(*vpp);
230 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
231 if (vap->va_rdev != VNOVAL) {
232 /*
233 * Want to be able to use this to make badblock
234 * inodes, so don't truncate the dev number.
235 */
236 ip->i_rdev = vap->va_rdev;
237 }
238 /*
239 * Remove inode, then reload it through VFS_VGET so it is
240 * checked to see if it is an alias of an existing entry in
241 * the inode cache.
242 */
243 vput(*vpp);
244 (*vpp)->v_type = VNON;
245 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
246 vgone(*vpp);
247 error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
248 if (error) {
249 *vpp = NULL;
250 return (error);
251 }
252 return (0);
253}
254
255/*
256 * Open called.
257 *
258 * Nothing to do.
259 */
260/* ARGSUSED */
261int
262ufs_open(ap)
263 struct vop_open_args /* {
264 struct vnode *a_vp;
265 int a_mode;
266 struct ucred *a_cred;
267 struct thread *a_td;
268 } */ *ap;
269{
270
271 /*
272 * Files marked append-only must be opened for appending.
273 */
274 if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
275 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
276 return (EPERM);
277 return (0);
278}
279
280/*
281 * Close called.
282 *
283 * Update the times on the inode.
284 */
285/* ARGSUSED */
286int
287ufs_close(ap)
288 struct vop_close_args /* {
289 struct vnode *a_vp;
290 int a_fflag;
291 struct ucred *a_cred;
292 struct thread *a_td;
293 } */ *ap;
294{
295 register struct vnode *vp = ap->a_vp;
296
297 lwkt_gettoken(&vp->v_interlock);
298 if (vp->v_usecount > 1)
299 ufs_itimes(vp);
300 lwkt_reltoken(&vp->v_interlock);
301 return (0);
302}
303
304int
305ufs_access(ap)
306 struct vop_access_args /* {
307 struct vnode *a_vp;
308 int a_mode;
309 struct ucred *a_cred;
310 struct thread *a_td;
311 } */ *ap;
312{
313 struct vnode *vp = ap->a_vp;
314 struct inode *ip = VTOI(vp);
315 struct ucred *cred = ap->a_cred;
316 mode_t mask, mode = ap->a_mode;
317 register gid_t *gp;
318 int i;
319#ifdef QUOTA
320 int error;
321#endif
322
323 /*
324 * Disallow write attempts on read-only file systems;
325 * unless the file is a socket, fifo, or a block or
326 * character device resident on the file system.
327 */
328 if (mode & VWRITE) {
329 switch (vp->v_type) {
330 case VDIR:
331 case VLNK:
332 case VREG:
333 if (vp->v_mount->mnt_flag & MNT_RDONLY)
334 return (EROFS);
335#ifdef QUOTA
336 if ((error = getinoquota(ip)) != 0)
337 return (error);
338#endif
339 break;
340 default:
341 break;
342 }
343 }
344
345 /* If immutable bit set, nobody gets to write it. */
346 if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
347 return (EPERM);
348
349 /* Otherwise, user id 0 always gets access. */
350 if (cred->cr_uid == 0)
351 return (0);
352
353 mask = 0;
354
355 /* Otherwise, check the owner. */
356 if (cred->cr_uid == ip->i_uid) {
357 if (mode & VEXEC)
358 mask |= S_IXUSR;
359 if (mode & VREAD)
360 mask |= S_IRUSR;
361 if (mode & VWRITE)
362 mask |= S_IWUSR;
363 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
364 }
365
366 /* Otherwise, check the groups. */
367 for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
368 if (ip->i_gid == *gp) {
369 if (mode & VEXEC)
370 mask |= S_IXGRP;
371 if (mode & VREAD)
372 mask |= S_IRGRP;
373 if (mode & VWRITE)
374 mask |= S_IWGRP;
375 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
376 }
377
378 /* Otherwise, check everyone else. */
379 if (mode & VEXEC)
380 mask |= S_IXOTH;
381 if (mode & VREAD)
382 mask |= S_IROTH;
383 if (mode & VWRITE)
384 mask |= S_IWOTH;
385 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
386}
387
388/* ARGSUSED */
389int
390ufs_getattr(ap)
391 struct vop_getattr_args /* {
392 struct vnode *a_vp;
393 struct vattr *a_vap;
394 struct thread *a_td;
395 } */ *ap;
396{
397 register struct vnode *vp = ap->a_vp;
398 register struct inode *ip = VTOI(vp);
399 register struct vattr *vap = ap->a_vap;
400
401 ufs_itimes(vp);
402 /*
403 * Copy from inode table
404 */
405 vap->va_fsid = dev2udev(ip->i_dev);
406 vap->va_fileid = ip->i_number;
407 vap->va_mode = ip->i_mode & ~IFMT;
408 vap->va_nlink = VFSTOUFS(vp->v_mount)->um_i_effnlink_valid ?
409 ip->i_effnlink : ip->i_nlink;
410 vap->va_uid = ip->i_uid;
411 vap->va_gid = ip->i_gid;
412 vap->va_rdev = ip->i_rdev;
413 vap->va_size = ip->i_din.di_size;
414 vap->va_atime.tv_sec = ip->i_atime;
415 vap->va_atime.tv_nsec = ip->i_atimensec;
416 vap->va_mtime.tv_sec = ip->i_mtime;
417 vap->va_mtime.tv_nsec = ip->i_mtimensec;
418 vap->va_ctime.tv_sec = ip->i_ctime;
419 vap->va_ctime.tv_nsec = ip->i_ctimensec;
420 vap->va_flags = ip->i_flags;
421 vap->va_gen = ip->i_gen;
422 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
423 vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
424 vap->va_type = IFTOVT(ip->i_mode);
425 vap->va_filerev = ip->i_modrev;
426 return (0);
427}
428
429/*
430 * Set attribute vnode op. called from several syscalls
431 */
432int
433ufs_setattr(ap)
434 struct vop_setattr_args /* {
435 struct vnode *a_vp;
436 struct vattr *a_vap;
437 struct ucred *a_cred;
438 struct thread *a_td;
439 } */ *ap;
440{
441 struct vattr *vap = ap->a_vap;
442 struct vnode *vp = ap->a_vp;
443 struct inode *ip = VTOI(vp);
444 struct ucred *cred = ap->a_cred;
445 int error;
446
447 /*
448 * Check for unsettable attributes.
449 */
450 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
451 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
452 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
453 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
454 return (EINVAL);
455 }
456 if (vap->va_flags != VNOVAL) {
457 if (vp->v_mount->mnt_flag & MNT_RDONLY)
458 return (EROFS);
459 if (cred->cr_uid != ip->i_uid &&
460 (error = suser_cred(cred, PRISON_ROOT)))
461 return (error);
462 if ((cred->cr_uid == 0) && (cred->cr_prison == NULL)) {
463 if ((ip->i_flags
464 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
465 securelevel > 0)
466 return (EPERM);
467 ip->i_flags = vap->va_flags;
468 } else {
469 if (ip->i_flags
470 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
471 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
472 return (EPERM);
473 ip->i_flags &= SF_SETTABLE;
474 ip->i_flags |= (vap->va_flags & UF_SETTABLE);
475 }
476 ip->i_flag |= IN_CHANGE;
477 if (vap->va_flags & (IMMUTABLE | APPEND))
478 return (0);
479 }
480 if (ip->i_flags & (IMMUTABLE | APPEND))
481 return (EPERM);
482 /*
483 * Go through the fields and update iff not VNOVAL.
484 */
485 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
486 if (vp->v_mount->mnt_flag & MNT_RDONLY)
487 return (EROFS);
488 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, ap->a_td)) != 0)
489 return (error);
490 }
491 if (vap->va_size != VNOVAL) {
492 /*
493 * Disallow write attempts on read-only file systems;
494 * unless the file is a socket, fifo, or a block or
495 * character device resident on the file system.
496 */
497 switch (vp->v_type) {
498 case VDIR:
499 return (EISDIR);
500 case VLNK:
501 case VREG:
502 if (vp->v_mount->mnt_flag & MNT_RDONLY)
503 return (EROFS);
504 break;
505 default:
506 break;
507 }
508 if ((error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, ap->a_td)) != 0)
509 return (error);
510 }
511 ip = VTOI(vp);
512 if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
513 if (vp->v_mount->mnt_flag & MNT_RDONLY)
514 return (EROFS);
515 if (cred->cr_uid != ip->i_uid &&
516 (error = suser_cred(cred, PRISON_ROOT)) &&
517 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
518 (error = VOP_ACCESS(vp, VWRITE, cred, ap->a_td))))
519 return (error);
520 if (vap->va_atime.tv_sec != VNOVAL)
521 ip->i_flag |= IN_ACCESS;
522 if (vap->va_mtime.tv_sec != VNOVAL)
523 ip->i_flag |= IN_CHANGE | IN_UPDATE;
524 ufs_itimes(vp);
525 if (vap->va_atime.tv_sec != VNOVAL) {
526 ip->i_atime = vap->va_atime.tv_sec;
527 ip->i_atimensec = vap->va_atime.tv_nsec;
528 }
529 if (vap->va_mtime.tv_sec != VNOVAL) {
530 ip->i_mtime = vap->va_mtime.tv_sec;
531 ip->i_mtimensec = vap->va_mtime.tv_nsec;
532 }
533 error = UFS_UPDATE(vp, 0);
534 if (error)
535 return (error);
536 }
537 error = 0;
538 if (vap->va_mode != (mode_t)VNOVAL) {
539 if (vp->v_mount->mnt_flag & MNT_RDONLY)
540 return (EROFS);
541 error = ufs_chmod(vp, (int)vap->va_mode, cred, ap->a_td);
542 }
543 VN_KNOTE(vp, NOTE_ATTRIB);
544 return (error);
545}
546
547/*
548 * Change the mode on a file.
549 * Inode must be locked before calling.
550 */
551static int
552ufs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
553{
554 register struct inode *ip = VTOI(vp);
555 int error;
556
557 if (cred->cr_uid != ip->i_uid) {
558 error = suser_cred(cred, PRISON_ROOT);
559 if (error)
560 return (error);
561 }
562 if (cred->cr_uid) {
563 if (vp->v_type != VDIR && (mode & S_ISTXT))
564 return (EFTYPE);
565 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
566 return (EPERM);
567 }
568 ip->i_mode &= ~ALLPERMS;
569 ip->i_mode |= (mode & ALLPERMS);
570 ip->i_flag |= IN_CHANGE;
571 return (0);
572}
573
574/*
575 * Perform chown operation on inode ip;
576 * inode must be locked prior to call.
577 */
578static int
579ufs_chown(vp, uid, gid, cred, td)
580 register struct vnode *vp;
581 uid_t uid;
582 gid_t gid;
583 struct ucred *cred;
584 struct thread *td;
585{
586 register struct inode *ip = VTOI(vp);
587 uid_t ouid;
588 gid_t ogid;
589 int error = 0;
590#ifdef QUOTA
591 register int i;
592 long change;
593#endif
594
595 if (uid == (uid_t)VNOVAL)
596 uid = ip->i_uid;
597 if (gid == (gid_t)VNOVAL)
598 gid = ip->i_gid;
599 /*
600 * If we don't own the file, are trying to change the owner
601 * of the file, or are not a member of the target group,
602 * the caller must be superuser or the call fails.
603 */
604 if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
605 (gid != ip->i_gid && !groupmember((gid_t)gid, cred))) &&
606 (error = suser_cred(cred, PRISON_ROOT)))
607 return (error);
608 ogid = ip->i_gid;
609 ouid = ip->i_uid;
610#ifdef QUOTA
611 if ((error = getinoquota(ip)) != 0)
612 return (error);
613 if (ouid == uid) {
614 dqrele(vp, ip->i_dquot[USRQUOTA]);
615 ip->i_dquot[USRQUOTA] = NODQUOT;
616 }
617 if (ogid == gid) {
618 dqrele(vp, ip->i_dquot[GRPQUOTA]);
619 ip->i_dquot[GRPQUOTA] = NODQUOT;
620 }
621 change = ip->i_blocks;
622 (void) chkdq(ip, -change, cred, CHOWN);
623 (void) chkiq(ip, -1, cred, CHOWN);
624 for (i = 0; i < MAXQUOTAS; i++) {
625 dqrele(vp, ip->i_dquot[i]);
626 ip->i_dquot[i] = NODQUOT;
627 }
628#endif
629 ip->i_gid = gid;
630 ip->i_uid = uid;
631#ifdef QUOTA
632 if ((error = 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 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
642 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
643 goto good;
644 else
645 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
646 }
647 for (i = 0; i < MAXQUOTAS; i++) {
648 dqrele(vp, ip->i_dquot[i]);
649 ip->i_dquot[i] = NODQUOT;
650 }
651 }
652 ip->i_gid = ogid;
653 ip->i_uid = ouid;
654 if (getinoquota(ip) == 0) {
655 if (ouid == uid) {
656 dqrele(vp, ip->i_dquot[USRQUOTA]);
657 ip->i_dquot[USRQUOTA] = NODQUOT;
658 }
659 if (ogid == gid) {
660 dqrele(vp, ip->i_dquot[GRPQUOTA]);
661 ip->i_dquot[GRPQUOTA] = NODQUOT;
662 }
663 (void) chkdq(ip, change, cred, FORCE|CHOWN);
664 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
665 (void) getinoquota(ip);
666 }
667 return (error);
668good:
669 if (getinoquota(ip))
670 panic("ufs_chown: lost quota");
671#endif /* QUOTA */
672 ip->i_flag |= IN_CHANGE;
673 if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
674 ip->i_mode &= ~(ISUID | ISGID);
675 return (0);
676}
677
678/*
679 * Mmap a file
680 *
681 * NB Currently unsupported.
682 */
683/* ARGSUSED */
684int
685ufs_mmap(ap)
686 struct vop_mmap_args /* {
687 struct vnode *a_vp;
688 int a_fflags;
689 struct ucred *a_cred;
690 struct thread *a_td;
691 } */ *ap;
692{
693
694 return (EINVAL);
695}
696
697int
698ufs_remove(ap)
699 struct vop_remove_args /* {
700 struct vnode *a_dvp;
701 struct vnode *a_vp;
702 struct componentname *a_cnp;
703 } */ *ap;
704{
705 struct inode *ip;
706 struct vnode *vp = ap->a_vp;
707 struct vnode *dvp = ap->a_dvp;
708 int error;
709
710 ip = VTOI(vp);
711 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
712 (VTOI(dvp)->i_flags & APPEND)) {
713 error = EPERM;
714 goto out;
715 }
716 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
717 VN_KNOTE(vp, NOTE_DELETE);
718 VN_KNOTE(dvp, NOTE_WRITE);
719out:
720 return (error);
721}
722
723/*
724 * link vnode call
725 */
726int
727ufs_link(ap)
728 struct vop_link_args /* {
729 struct vnode *a_tdvp;
730 struct vnode *a_vp;
731 struct componentname *a_cnp;
732 } */ *ap;
733{
734 struct vnode *vp = ap->a_vp;
735 struct vnode *tdvp = ap->a_tdvp;
736 struct componentname *cnp = ap->a_cnp;
737 struct thread *td = cnp->cn_td;
738 struct inode *ip;
739 struct direct newdir;
740 int error;
741
742#ifdef DIAGNOSTIC
743 if ((cnp->cn_flags & HASBUF) == 0)
744 panic("ufs_link: no name");
745#endif
746 if (tdvp->v_mount != vp->v_mount) {
747 error = EXDEV;
748 goto out2;
749 }
750 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, td))) {
751 goto out2;
752 }
753 ip = VTOI(vp);
754 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
755 error = EMLINK;
756 goto out1;
757 }
758 if (ip->i_flags & (IMMUTABLE | APPEND)) {
759 error = EPERM;
760 goto out1;
761 }
762 ip->i_effnlink++;
763 ip->i_nlink++;
764 ip->i_flag |= IN_CHANGE;
765 if (DOINGSOFTDEP(vp))
766 softdep_change_linkcnt(ip);
767 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
768 if (!error) {
769 ufs_makedirentry(ip, cnp, &newdir);
770 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
771 }
772
773 if (error) {
774 ip->i_effnlink--;
775 ip->i_nlink--;
776 ip->i_flag |= IN_CHANGE;
777 if (DOINGSOFTDEP(vp))
778 softdep_change_linkcnt(ip);
779 }
780out1:
781 if (tdvp != vp)
782 VOP_UNLOCK(vp, 0, td);
783out2:
784 VN_KNOTE(vp, NOTE_LINK);
785 VN_KNOTE(tdvp, NOTE_WRITE);
786 return (error);
787}
788
789/*
790 * whiteout vnode call
791 */
792int
793ufs_whiteout(ap)
794 struct vop_whiteout_args /* {
795 struct vnode *a_dvp;
796 struct componentname *a_cnp;
797 int a_flags;
798 } */ *ap;
799{
800 struct vnode *dvp = ap->a_dvp;
801 struct componentname *cnp = ap->a_cnp;
802 struct direct newdir;
803 int error = 0;
804
805 switch (ap->a_flags) {
806 case LOOKUP:
807 /* 4.4 format directories support whiteout operations */
808 if (dvp->v_mount->mnt_maxsymlinklen > 0)
809 return (0);
810 return (EOPNOTSUPP);
811
812 case CREATE:
813 /* create a new directory whiteout */
814#ifdef DIAGNOSTIC
815 if ((cnp->cn_flags & SAVENAME) == 0)
816 panic("ufs_whiteout: missing name");
817 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
818 panic("ufs_whiteout: old format filesystem");
819#endif
820
821 newdir.d_ino = WINO;
822 newdir.d_namlen = cnp->cn_namelen;
823 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
824 newdir.d_type = DT_WHT;
825 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
826 break;
827
828 case DELETE:
829 /* remove an existing directory whiteout */
830#ifdef DIAGNOSTIC
831 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
832 panic("ufs_whiteout: old format filesystem");
833#endif
834
835 cnp->cn_flags &= ~DOWHITEOUT;
836 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
837 break;
838 default:
839 panic("ufs_whiteout: unknown op");
840 }
841 return (error);
842}
843
844/*
845 * Rename system call.
846 * rename("foo", "bar");
847 * is essentially
848 * unlink("bar");
849 * link("foo", "bar");
850 * unlink("foo");
851 * but ``atomically''. Can't do full commit without saving state in the
852 * inode on disk which isn't feasible at this time. Best we can do is
853 * always guarantee the target exists.
854 *
855 * Basic algorithm is:
856 *
857 * 1) Bump link count on source while we're linking it to the
858 * target. This also ensure the inode won't be deleted out
859 * from underneath us while we work (it may be truncated by
860 * a concurrent `trunc' or `open' for creation).
861 * 2) Link source to destination. If destination already exists,
862 * delete it first.
863 * 3) Unlink source reference to inode if still around. If a
864 * directory was moved and the parent of the destination
865 * is different from the source, patch the ".." entry in the
866 * directory.
867 */
868int
869ufs_rename(ap)
870 struct vop_rename_args /* {
871 struct vnode *a_fdvp;
872 struct vnode *a_fvp;
873 struct componentname *a_fcnp;
874 struct vnode *a_tdvp;
875 struct vnode *a_tvp;
876 struct componentname *a_tcnp;
877 } */ *ap;
878{
879 struct vnode *tvp = ap->a_tvp;
880 register struct vnode *tdvp = ap->a_tdvp;
881 struct vnode *fvp = ap->a_fvp;
882 struct vnode *fdvp = ap->a_fdvp;
883 struct componentname *tcnp = ap->a_tcnp;
884 struct componentname *fcnp = ap->a_fcnp;
885 struct thread *td = fcnp->cn_td;
886 struct inode *ip, *xp, *dp;
887 struct direct newdir;
888 int doingdirectory = 0, oldparent = 0, newparent = 0;
889 int error = 0, ioflag;
890
891#ifdef DIAGNOSTIC
892 if ((tcnp->cn_flags & HASBUF) == 0 ||
893 (fcnp->cn_flags & HASBUF) == 0)
894 panic("ufs_rename: no name");
895#endif
896 /*
897 * Check for cross-device rename.
898 */
899 if ((fvp->v_mount != tdvp->v_mount) ||
900 (tvp && (fvp->v_mount != tvp->v_mount))) {
901 error = EXDEV;
902abortit:
903 if (tdvp == tvp)
904 vrele(tdvp);
905 else
906 vput(tdvp);
907 if (tvp)
908 vput(tvp);
909 vrele(fdvp);
910 vrele(fvp);
911 return (error);
912 }
913
914 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
915 (VTOI(tdvp)->i_flags & APPEND))) {
916 error = EPERM;
917 goto abortit;
918 }
919
920 /*
921 * Renaming a file to itself has no effect. The upper layers should
922 * not call us in that case. Temporarily just warn if they do.
923 */
924 if (fvp == tvp) {
925 printf("ufs_rename: fvp == tvp (can't happen)\n");
926 error = 0;
927 goto abortit;
928 }
929
930 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
931 goto abortit;
932 dp = VTOI(fdvp);
933 ip = VTOI(fvp);
934 if (ip->i_nlink >= LINK_MAX) {
935 VOP_UNLOCK(fvp, 0, td);
936 error = EMLINK;
937 goto abortit;
938 }
939 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
940 || (dp->i_flags & APPEND)) {
941 VOP_UNLOCK(fvp, 0, td);
942 error = EPERM;
943 goto abortit;
944 }
945 if ((ip->i_mode & IFMT) == IFDIR) {
946 /*
947 * Avoid ".", "..", and aliases of "." for obvious reasons.
948 */
949 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
950 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
951 (ip->i_flag & IN_RENAME)) {
952 VOP_UNLOCK(fvp, 0, td);
953 error = EINVAL;
954 goto abortit;
955 }
956 ip->i_flag |= IN_RENAME;
957 oldparent = dp->i_number;
958 doingdirectory = 1;
959 }
960 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
961 vrele(fdvp);
962
963 /*
964 * When the target exists, both the directory
965 * and target vnodes are returned locked.
966 */
967 dp = VTOI(tdvp);
968 xp = NULL;
969 if (tvp)
970 xp = VTOI(tvp);
971
972 /*
973 * 1) Bump link count while we're moving stuff
974 * around. If we crash somewhere before
975 * completing our work, the link count
976 * may be wrong, but correctable.
977 */
978 ip->i_effnlink++;
979 ip->i_nlink++;
980 ip->i_flag |= IN_CHANGE;
981 if (DOINGSOFTDEP(fvp))
982 softdep_change_linkcnt(ip);
983 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
984 DOINGASYNC(fvp)))) != 0) {
985 VOP_UNLOCK(fvp, 0, td);
986 goto bad;
987 }
988
989 /*
990 * If ".." must be changed (ie the directory gets a new
991 * parent) then the source directory must not be in the
992 * directory heirarchy above the target, as this would
993 * orphan everything below the source directory. Also
994 * the user must have write permission in the source so
995 * as to be able to change "..". We must repeat the call
996 * to namei, as the parent directory is unlocked by the
997 * call to checkpath().
998 */
999 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_td);
1000 VOP_UNLOCK(fvp, 0, td);
1001 if (oldparent != dp->i_number)
1002 newparent = dp->i_number;
1003 if (doingdirectory && newparent) {
1004 if (error) /* write access check above */
1005 goto bad;
1006 if (xp != NULL)
1007 vput(tvp);
1008 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
1009 if (error)
1010 goto out;
1011 if ((tcnp->cn_flags & SAVESTART) == 0)
1012 panic("ufs_rename: lost to startdir");
1013 VREF(tdvp);
1014 error = relookup(tdvp, &tvp, tcnp);
1015 if (error)
1016 goto out;
1017 vrele(tdvp);
1018 dp = VTOI(tdvp);
1019 xp = NULL;
1020 if (tvp)
1021 xp = VTOI(tvp);
1022 }
1023 /*
1024 * 2) If target doesn't exist, link the target
1025 * to the source and unlink the source.
1026 * Otherwise, rewrite the target directory
1027 * entry to reference the source inode and
1028 * expunge the original entry's existence.
1029 */
1030 if (xp == NULL) {
1031 if (dp->i_dev != ip->i_dev)
1032 panic("ufs_rename: EXDEV");
1033 /*
1034 * Account for ".." in new directory.
1035 * When source and destination have the same
1036 * parent we don't fool with the link count.
1037 */
1038 if (doingdirectory && newparent) {
1039 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1040 error = EMLINK;
1041 goto bad;
1042 }
1043 dp->i_effnlink++;
1044 dp->i_nlink++;
1045 dp->i_flag |= IN_CHANGE;
1046 if (DOINGSOFTDEP(tdvp))
1047 softdep_change_linkcnt(dp);
1048 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1049 DOINGASYNC(tdvp)));
1050 if (error)
1051 goto bad;
1052 }
1053 ufs_makedirentry(ip, tcnp, &newdir);
1054 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1055 if (error) {
1056 if (doingdirectory && newparent) {
1057 dp->i_effnlink--;
1058 dp->i_nlink--;
1059 dp->i_flag |= IN_CHANGE;
1060 if (DOINGSOFTDEP(tdvp))
1061 softdep_change_linkcnt(dp);
1062 (void)UFS_UPDATE(tdvp, 1);
1063 }
1064 goto bad;
1065 }
1066 VN_KNOTE(tdvp, NOTE_WRITE);
1067 vput(tdvp);
1068 } else {
1069 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1070 panic("ufs_rename: EXDEV");
1071 /*
1072 * Short circuit rename(foo, foo).
1073 */
1074 if (xp->i_number == ip->i_number)
1075 panic("ufs_rename: same file");
1076 /*
1077 * If the parent directory is "sticky", then the user must
1078 * own the parent directory, or the destination of the rename,
1079 * otherwise the destination may not be changed (except by
1080 * root). This implements append-only directories.
1081 */
1082 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1083 tcnp->cn_cred->cr_uid != dp->i_uid &&
1084 xp->i_uid != tcnp->cn_cred->cr_uid) {
1085 error = EPERM;
1086 goto bad;
1087 }
1088 /*
1089 * Target must be empty if a directory and have no links
1090 * to it. Also, ensure source and target are compatible
1091 * (both directories, or both not directories).
1092 */
1093 if ((xp->i_mode&IFMT) == IFDIR) {
1094 if ((xp->i_effnlink > 2) ||
1095 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1096 error = ENOTEMPTY;
1097 goto bad;
1098 }
1099 if (!doingdirectory) {
1100 error = ENOTDIR;
1101 goto bad;
1102 }
1103 cache_purge(tdvp);
1104 } else if (doingdirectory) {
1105 error = EISDIR;
1106 goto bad;
1107 }
1108 error = ufs_dirrewrite(dp, xp, ip->i_number,
1109 IFTODT(ip->i_mode),
1110 (doingdirectory && newparent) ? newparent : doingdirectory);
1111 if (error)
1112 goto bad;
1113 if (doingdirectory) {
1114 if (!newparent) {
1115 dp->i_effnlink--;
1116 if (DOINGSOFTDEP(tdvp))
1117 softdep_change_linkcnt(dp);
1118 }
1119 xp->i_effnlink--;
1120 if (DOINGSOFTDEP(tvp))
1121 softdep_change_linkcnt(xp);
1122 }
1123 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1124 /*
1125 * Truncate inode. The only stuff left in the directory
1126 * is "." and "..". The "." reference is inconsequential
1127 * since we are quashing it. We have removed the "."
1128 * reference and the reference in the parent directory,
1129 * but there may be other hard links. The soft
1130 * dependency code will arrange to do these operations
1131 * after the parent directory entry has been deleted on
1132 * disk, so when running with that code we avoid doing
1133 * them now.
1134 */
1135 if (!newparent) {
1136 dp->i_nlink--;
1137 dp->i_flag |= IN_CHANGE;
1138 }
1139 xp->i_nlink--;
1140 xp->i_flag |= IN_CHANGE;
1141 ioflag = DOINGASYNC(tvp) ? 0 : IO_SYNC;
1142 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
1143 tcnp->cn_cred, tcnp->cn_td)) != 0)
1144 goto bad;
1145 }
1146 VN_KNOTE(tdvp, NOTE_WRITE);
1147 vput(tdvp);
1148 VN_KNOTE(tvp, NOTE_DELETE);
1149 vput(tvp);
1150 xp = NULL;
1151 }
1152
1153 /*
1154 * 3) Unlink the source.
1155 */
1156 fcnp->cn_flags &= ~MODMASK;
1157 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1158 if ((fcnp->cn_flags & SAVESTART) == 0)
1159 panic("ufs_rename: lost from startdir");
1160 VREF(fdvp);
1161 error = relookup(fdvp, &fvp, fcnp);
1162 if (error == 0)
1163 vrele(fdvp);
1164 if (fvp != NULL) {
1165 xp = VTOI(fvp);
1166 dp = VTOI(fdvp);
1167 } else {
1168 /*
1169 * From name has disappeared.
1170 */
1171 if (doingdirectory)
1172 panic("ufs_rename: lost dir entry");
1173 vrele(ap->a_fvp);
1174 return (0);
1175 }
1176 /*
1177 * Ensure that the directory entry still exists and has not
1178 * changed while the new name has been entered. If the source is
1179 * a file then the entry may have been unlinked or renamed. In
1180 * either case there is no further work to be done. If the source
1181 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1182 * flag ensures that it cannot be moved by another rename or removed
1183 * by a rmdir.
1184 */
1185 if (xp != ip) {
1186 if (doingdirectory)
1187 panic("ufs_rename: lost dir entry");
1188 } else {
1189 /*
1190 * If the source is a directory with a
1191 * new parent, the link count of the old
1192 * parent directory must be decremented
1193 * and ".." set to point to the new parent.
1194 */
1195 if (doingdirectory && newparent) {
1196 xp->i_offset = mastertemplate.dot_reclen;
1197 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
1198 cache_purge(fdvp);
1199 }
1200 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1201 xp->i_flag &= ~IN_RENAME;
1202 }
1203 VN_KNOTE(fvp, NOTE_RENAME);
1204 if (dp)
1205 vput(fdvp);
1206 if (xp)
1207 vput(fvp);
1208 vrele(ap->a_fvp);
1209 return (error);
1210
1211bad:
1212 if (xp)
1213 vput(ITOV(xp));
1214 vput(ITOV(dp));
1215out:
1216 if (doingdirectory)
1217 ip->i_flag &= ~IN_RENAME;
1218 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
1219 ip->i_effnlink--;
1220 ip->i_nlink--;
1221 ip->i_flag |= IN_CHANGE;
1222 ip->i_flag &= ~IN_RENAME;
1223 if (DOINGSOFTDEP(fvp))
1224 softdep_change_linkcnt(ip);
1225 vput(fvp);
1226 } else
1227 vrele(fvp);
1228 return (error);
1229}
1230
1231/*
1232 * Mkdir system call
1233 */
1234int
1235ufs_mkdir(ap)
1236 struct vop_mkdir_args /* {
1237 struct vnode *a_dvp;
1238 struct vnode **a_vpp;
1239 struct componentname *a_cnp;
1240 struct vattr *a_vap;
1241 } */ *ap;
1242{
1243 register struct vnode *dvp = ap->a_dvp;
1244 register struct vattr *vap = ap->a_vap;
1245 register struct componentname *cnp = ap->a_cnp;
1246 register struct inode *ip, *dp;
1247 struct vnode *tvp;
1248 struct buf *bp;
1249 struct dirtemplate dirtemplate, *dtp;
1250 struct direct newdir;
1251 int error, dmode;
1252 long blkoff;
1253
1254#ifdef DIAGNOSTIC
1255 if ((cnp->cn_flags & HASBUF) == 0)
1256 panic("ufs_mkdir: no name");
1257#endif
1258 dp = VTOI(dvp);
1259 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1260 error = EMLINK;
1261 goto out;
1262 }
1263 dmode = vap->va_mode & 0777;
1264 dmode |= IFDIR;
1265 /*
1266 * Must simulate part of ufs_makeinode here to acquire the inode,
1267 * but not have it entered in the parent directory. The entry is
1268 * made later after writing "." and ".." entries.
1269 */
1270 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1271 if (error)
1272 goto out;
1273 ip = VTOI(tvp);
1274 ip->i_gid = dp->i_gid;
1275#ifdef SUIDDIR
1276 {
1277#ifdef QUOTA
1278 struct ucred ucred, *ucp;
1279 ucp = cnp->cn_cred;
1280#endif
1281 /*
1282 * If we are hacking owners here, (only do this where told to)
1283 * and we are not giving it TO root, (would subvert quotas)
1284 * then go ahead and give it to the other user.
1285 * The new directory also inherits the SUID bit.
1286 * If user's UID and dir UID are the same,
1287 * 'give it away' so that the SUID is still forced on.
1288 */
1289 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1290 (dp->i_mode & ISUID) && dp->i_uid) {
1291 dmode |= ISUID;
1292 ip->i_uid = dp->i_uid;
1293#ifdef QUOTA
1294 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1295 /*
1296 * Make sure the correct user gets charged
1297 * for the space.
1298 * Make a dummy credential for the victim.
1299 * XXX This seems to never be accessed out of
1300 * our context so a stack variable is ok.
1301 */
1302 ucred.cr_ref = 1;
1303 ucred.cr_uid = ip->i_uid;
1304 ucred.cr_ngroups = 1;
1305 ucred.cr_groups[0] = dp->i_gid;
1306 ucp = &ucred;
1307 }
1308#endif
1309 } else
1310 ip->i_uid = cnp->cn_cred->cr_uid;
1311#ifdef QUOTA
1312 if ((error = getinoquota(ip)) ||
1313 (error = chkiq(ip, 1, ucp, 0))) {
1314 UFS_VFREE(tvp, ip->i_number, dmode);
1315 vput(tvp);
1316 return (error);
1317 }
1318#endif
1319 }
1320#else /* !SUIDDIR */
1321 ip->i_uid = cnp->cn_cred->cr_uid;
1322#ifdef QUOTA
1323 if ((error = getinoquota(ip)) ||
1324 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1325 UFS_VFREE(tvp, ip->i_number, dmode);
1326 vput(tvp);
1327 return (error);
1328 }
1329#endif
1330#endif /* !SUIDDIR */
1331 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1332 ip->i_mode = dmode;
1333 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1334 ip->i_effnlink = 2;
1335 ip->i_nlink = 2;
1336 if (DOINGSOFTDEP(tvp))
1337 softdep_change_linkcnt(ip);
1338 if (cnp->cn_flags & ISWHITEOUT)
1339 ip->i_flags |= UF_OPAQUE;
1340
1341 /*
1342 * Bump link count in parent directory to reflect work done below.
1343 * Should be done before reference is created so cleanup is
1344 * possible if we crash.
1345 */
1346 dp->i_effnlink++;
1347 dp->i_nlink++;
1348 dp->i_flag |= IN_CHANGE;
1349 if (DOINGSOFTDEP(dvp))
1350 softdep_change_linkcnt(dp);
1351 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1352 if (error)
1353 goto bad;
1354
1355 /*
1356 * Initialize directory with "." and ".." from static template.
1357 */
1358 if (dvp->v_mount->mnt_maxsymlinklen > 0
1359 )
1360 dtp = &mastertemplate;
1361 else
1362 dtp = (struct dirtemplate *)&omastertemplate;
1363 dirtemplate = *dtp;
1364 dirtemplate.dot_ino = ip->i_number;
1365 dirtemplate.dotdot_ino = dp->i_number;
1366 if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1367 B_CLRBUF, &bp)) != 0)
1368 goto bad;
1369 ip->i_size = DIRBLKSIZ;
1370 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1371 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1372 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1373 if (DOINGSOFTDEP(tvp)) {
1374 /*
1375 * Ensure that the entire newly allocated block is a
1376 * valid directory so that future growth within the
1377 * block does not have to ensure that the block is
1378 * written before the inode.
1379 */
1380 blkoff = DIRBLKSIZ;
1381 while (blkoff < bp->b_bcount) {
1382 ((struct direct *)
1383 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1384 blkoff += DIRBLKSIZ;
1385 }
1386 }
1387 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1388 DOINGASYNC(tvp)))) != 0) {
1389 (void)VOP_BWRITE(bp->b_vp, bp);
1390 goto bad;
1391 }
1392 /*
1393 * Directory set up, now install its entry in the parent directory.
1394 *
1395 * If we are not doing soft dependencies, then we must write out the
1396 * buffer containing the new directory body before entering the new
1397 * name in the parent. If we are doing soft dependencies, then the
1398 * buffer containing the new directory body will be passed to and
1399 * released in the soft dependency code after the code has attached
1400 * an appropriate ordering dependency to the buffer which ensures that
1401 * the buffer is written before the new name is written in the parent.
1402 */
1403 if (DOINGASYNC(dvp))
1404 bdwrite(bp);
1405 else if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp->b_vp, bp))))
1406 goto bad;
1407 ufs_makedirentry(ip, cnp, &newdir);
1408 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1409
1410bad:
1411 if (error == 0) {
1412 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1413 *ap->a_vpp = tvp;
1414 } else {
1415 dp->i_effnlink--;
1416 dp->i_nlink--;
1417 dp->i_flag |= IN_CHANGE;
1418 if (DOINGSOFTDEP(dvp))
1419 softdep_change_linkcnt(dp);
1420 /*
1421 * No need to do an explicit VOP_TRUNCATE here, vrele will
1422 * do this for us because we set the link count to 0.
1423 */
1424 ip->i_effnlink = 0;
1425 ip->i_nlink = 0;
1426 ip->i_flag |= IN_CHANGE;
1427 if (DOINGSOFTDEP(tvp))
1428 softdep_change_linkcnt(ip);
1429 vput(tvp);
1430 }
1431out:
1432 return (error);
1433}
1434
1435/*
1436 * Rmdir system call.
1437 */
1438int
1439ufs_rmdir(ap)
1440 struct vop_rmdir_args /* {
1441 struct vnode *a_dvp;
1442 struct vnode *a_vp;
1443 struct componentname *a_cnp;
1444 } */ *ap;
1445{
1446 struct vnode *vp = ap->a_vp;
1447 struct vnode *dvp = ap->a_dvp;
1448 struct componentname *cnp = ap->a_cnp;
1449 struct inode *ip, *dp;
1450 int error, ioflag;
1451
1452 ip = VTOI(vp);
1453 dp = VTOI(dvp);
1454
1455 /*
1456 * Do not remove a directory that is in the process of being renamed.
1457 * Verify the directory is empty (and valid). Rmdir ".." will not be
1458 * valid since ".." will contain a reference to the current directory
1459 * and thus be non-empty. Do not allow the removal of mounted on
1460 * directories (this can happen when an NFS exported filesystem
1461 * tries to remove a locally mounted on directory).
1462 */
1463 error = 0;
1464 if (ip->i_flag & IN_RENAME) {
1465 error = EINVAL;
1466 goto out;
1467 }
1468 if (ip->i_effnlink != 2 ||
1469 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1470 error = ENOTEMPTY;
1471 goto out;
1472 }
1473 if ((dp->i_flags & APPEND)
1474 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1475 error = EPERM;
1476 goto out;
1477 }
1478 if (vp->v_mountedhere != 0) {
1479 error = EINVAL;
1480 goto out;
1481 }
1482 /*
1483 * Delete reference to directory before purging
1484 * inode. If we crash in between, the directory
1485 * will be reattached to lost+found,
1486 */
1487 dp->i_effnlink--;
1488 ip->i_effnlink--;
1489 if (DOINGSOFTDEP(vp)) {
1490 softdep_change_linkcnt(dp);
1491 softdep_change_linkcnt(ip);
1492 }
1493 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1494 if (error) {
1495 dp->i_effnlink++;
1496 ip->i_effnlink++;
1497 if (DOINGSOFTDEP(vp)) {
1498 softdep_change_linkcnt(dp);
1499 softdep_change_linkcnt(ip);
1500 }
1501 goto out;
1502 }
1503 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1504 cache_purge(dvp);
1505 /*
1506 * Truncate inode. The only stuff left in the directory is "." and
1507 * "..". The "." reference is inconsequential since we are quashing
1508 * it. The soft dependency code will arrange to do these operations
1509 * after the parent directory entry has been deleted on disk, so
1510 * when running with that code we avoid doing them now.
1511 */
1512 if (!DOINGSOFTDEP(vp)) {
1513 dp->i_nlink--;
1514 dp->i_flag |= IN_CHANGE;
1515 ip->i_nlink--;
1516 ip->i_flag |= IN_CHANGE;
1517 ioflag = DOINGASYNC(vp) ? 0 : IO_SYNC;
1518 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
1519 cnp->cn_td);
1520 }
1521 cache_purge(vp);
1522#ifdef UFS_DIRHASH
1523 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
1524 if (ip->i_dirhash != NULL)
1525 ufsdirhash_free(ip);
1526#endif
1527out:
1528 VN_KNOTE(vp, NOTE_DELETE);
1529 return (error);
1530}
1531
1532/*
1533 * symlink -- make a symbolic link
1534 */
1535int
1536ufs_symlink(ap)
1537 struct vop_symlink_args /* {
1538 struct vnode *a_dvp;
1539 struct vnode **a_vpp;
1540 struct componentname *a_cnp;
1541 struct vattr *a_vap;
1542 char *a_target;
1543 } */ *ap;
1544{
1545 register struct vnode *vp, **vpp = ap->a_vpp;
1546 register struct inode *ip;
1547 int len, error;
1548
1549 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1550 vpp, ap->a_cnp);
1551 if (error)
1552 return (error);
1553 VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1554 vp = *vpp;
1555 len = strlen(ap->a_target);
1556 if (len < vp->v_mount->mnt_maxsymlinklen) {
1557 ip = VTOI(vp);
1558 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1559 ip->i_size = len;
1560 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1561 } else
1562 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1563 UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred,
1564 (int *)0, NULL);
1565 if (error)
1566 vput(vp);
1567 return (error);
1568}
1569
1570/*
1571 * Vnode op for reading directories.
1572 *
1573 * The routine below assumes that the on-disk format of a directory
1574 * is the same as that defined by <sys/dirent.h>. If the on-disk
1575 * format changes, then it will be necessary to do a conversion
1576 * from the on-disk format that read returns to the format defined
1577 * by <sys/dirent.h>.
1578 */
1579int
1580ufs_readdir(ap)
1581 struct vop_readdir_args /* {
1582 struct vnode *a_vp;
1583 struct uio *a_uio;
1584 struct ucred *a_cred;
1585 int *a_eofflag;
1586 int *ncookies;
1587 u_long **a_cookies;
1588 } */ *ap;
1589{
1590 register struct uio *uio = ap->a_uio;
1591 int error;
1592 size_t count, lost;
1593 off_t off;
1594
1595 if (ap->a_ncookies != NULL)
1596 /*
1597 * Ensure that the block is aligned. The caller can use
1598 * the cookies to determine where in the block to start.
1599 */
1600 uio->uio_offset &= ~(DIRBLKSIZ - 1);
1601 off = uio->uio_offset;
1602 count = uio->uio_resid;
1603 /* Make sure we don't return partial entries. */
1604 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
1605 return (EINVAL);
1606 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1607 lost = uio->uio_resid - count;
1608 uio->uio_resid = count;
1609 uio->uio_iov->iov_len = count;
1610# if (BYTE_ORDER == LITTLE_ENDIAN)
1611 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
1612 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1613 } else {
1614 struct dirent *dp, *edp;
1615 struct uio auio;
1616 struct iovec aiov;
1617 caddr_t dirbuf;
1618 int readcnt;
1619 u_char tmp;
1620
1621 auio = *uio;
1622 auio.uio_iov = &aiov;
1623 auio.uio_iovcnt = 1;
1624 auio.uio_segflg = UIO_SYSSPACE;
1625 aiov.iov_len = count;
1626 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1627 aiov.iov_base = dirbuf;
1628 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
1629 if (error == 0) {
1630 readcnt = count - auio.uio_resid;
1631 edp = (struct dirent *)&dirbuf[readcnt];
1632 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
1633 tmp = dp->d_namlen;
1634 dp->d_namlen = dp->d_type;
1635 dp->d_type = tmp;
1636 if (dp->d_reclen > 0) {
1637 dp = (struct dirent *)
1638 ((char *)dp + dp->d_reclen);
1639 } else {
1640 error = EIO;
1641 break;
1642 }
1643 }
1644 if (dp >= edp)
1645 error = uiomove(dirbuf, readcnt, uio);
1646 }
1647 FREE(dirbuf, M_TEMP);
1648 }
1649# else
1650 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1651# endif
1652 if (!error && ap->a_ncookies != NULL) {
1653 struct dirent* dpStart;
1654 struct dirent* dpEnd;
1655 struct dirent* dp;
1656 int ncookies;
1657 u_long *cookies;
1658 u_long *cookiep;
1659
1660 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1661 panic("ufs_readdir: unexpected uio from NFS server");
1662 dpStart = (struct dirent *)
1663 (uio->uio_iov->iov_base - (uio->uio_offset - off));
1664 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
1665 for (dp = dpStart, ncookies = 0;
1666 dp < dpEnd;
1667 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
1668 ncookies++;
1669 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1670 M_WAITOK);
1671 for (dp = dpStart, cookiep = cookies;
1672 dp < dpEnd;
1673 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
1674 off += dp->d_reclen;
1675 *cookiep++ = (u_long) off;
1676 }
1677 *ap->a_ncookies = ncookies;
1678 *ap->a_cookies = cookies;
1679 }
1680 uio->uio_resid += lost;
1681 if (ap->a_eofflag)
1682 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
1683 return (error);
1684}
1685
1686/*
1687 * Return target name of a symbolic link
1688 */
1689int
1690ufs_readlink(ap)
1691 struct vop_readlink_args /* {
1692 struct vnode *a_vp;
1693 struct uio *a_uio;
1694 struct ucred *a_cred;
1695 } */ *ap;
1696{
1697 register struct vnode *vp = ap->a_vp;
1698 register struct inode *ip = VTOI(vp);
1699 int isize;
1700
1701 isize = ip->i_size;
1702 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1703 (ip->i_din.di_blocks == 0)) { /* XXX - for old fastlink support */
1704 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1705 return (0);
1706 }
1707 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1708}
1709
1710/*
1711 * Calculate the logical to physical mapping if not done already,
1712 * then call the device strategy routine.
1713 *
1714 * In order to be able to swap to a file, the VOP_BMAP operation may not
1715 * deadlock on memory. See ufs_bmap() for details.
1716 */
1717int
1718ufs_strategy(ap)
1719 struct vop_strategy_args /* {
1720 struct vnode *a_vp;
1721 struct buf *a_bp;
1722 } */ *ap;
1723{
1724 register struct buf *bp = ap->a_bp;
1725 register struct vnode *vp = ap->a_vp;
1726 register struct inode *ip;
1727 int error;
1728
1729 ip = VTOI(vp);
1730 if (vp->v_type == VBLK || vp->v_type == VCHR)
1731 panic("ufs_strategy: spec");
1732 if (bp->b_blkno == bp->b_lblkno) {
1733 error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
1734 if (error) {
1735 bp->b_error = error;
1736 bp->b_flags |= B_ERROR;
1737 biodone(bp);
1738 return (error);
1739 }
1740 if ((long)bp->b_blkno == -1)
1741 vfs_bio_clrbuf(bp);
1742 }
1743 if ((long)bp->b_blkno == -1) {
1744 biodone(bp);
1745 return (0);
1746 }
1747 vp = ip->i_devvp;
1748 bp->b_dev = vp->v_rdev;
1749 VOP_STRATEGY(vp, bp);
1750 return (0);
1751}
1752
1753/*
1754 * Print out the contents of an inode.
1755 */
1756int
1757ufs_print(ap)
1758 struct vop_print_args /* {
1759 struct vnode *a_vp;
1760 } */ *ap;
1761{
1762 register struct vnode *vp = ap->a_vp;
1763 register struct inode *ip = VTOI(vp);
1764
1765 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)",
1766 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1767 minor(ip->i_dev));
1768 if (vp->v_type == VFIFO)
1769 fifo_printinfo(vp);
1770 lockmgr_printinfo(&ip->i_lock);
1771 printf("\n");
1772 return (0);
1773}
1774
1775/*
1776 * Read wrapper for special devices.
1777 */
1778int
1779ufsspec_read(ap)
1780 struct vop_read_args /* {
1781 struct vnode *a_vp;
1782 struct uio *a_uio;
1783 int a_ioflag;
1784 struct ucred *a_cred;
1785 } */ *ap;
1786{
1787 int error, resid;
1788 struct inode *ip;
1789 struct uio *uio;
1790
1791 uio = ap->a_uio;
1792 resid = uio->uio_resid;
1793 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap);
1794 /*
1795 * The inode may have been revoked during the call, so it must not
1796 * be accessed blindly here or in the other wrapper functions.
1797 */
1798 ip = VTOI(ap->a_vp);
1799 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1800 ip->i_flag |= IN_ACCESS;
1801 return (error);
1802}
1803
1804/*
1805 * Write wrapper for special devices.
1806 */
1807int
1808ufsspec_write(ap)
1809 struct vop_write_args /* {
1810 struct vnode *a_vp;
1811 struct uio *a_uio;
1812 int a_ioflag;
1813 struct ucred *a_cred;
1814 } */ *ap;
1815{
1816 int error, resid;
1817 struct inode *ip;
1818 struct uio *uio;
1819
1820 uio = ap->a_uio;
1821 resid = uio->uio_resid;
1822 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap);
1823 ip = VTOI(ap->a_vp);
1824 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1825 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1826 return (error);
1827}
1828
1829/*
1830 * Close wrapper for special devices.
1831 *
1832 * Update the times on the inode then do device close.
1833 */
1834int
1835ufsspec_close(ap)
1836 struct vop_close_args /* {
1837 struct vnode *a_vp;
1838 int a_fflag;
1839 struct ucred *a_cred;
1840 struct thread *a_td;
1841 } */ *ap;
1842{
1843 struct vnode *vp = ap->a_vp;
1844
1845 lwkt_gettoken(&vp->v_interlock);
1846 if (vp->v_usecount > 1)
1847 ufs_itimes(vp);
1848 lwkt_reltoken(&vp->v_interlock);
1849 return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
1850}
1851
1852/*
1853 * Read wrapper for fifos.
1854 */
1855int
1856ufsfifo_read(ap)
1857 struct vop_read_args /* {
1858 struct vnode *a_vp;
1859 struct uio *a_uio;
1860 int a_ioflag;
1861 struct ucred *a_cred;
1862 } */ *ap;
1863{
1864 int error, resid;
1865 struct inode *ip;
1866 struct uio *uio;
1867
1868 uio = ap->a_uio;
1869 resid = uio->uio_resid;
1870 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap);
1871 ip = VTOI(ap->a_vp);
1872 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1873 (uio->uio_resid != resid || (error == 0 && resid != 0)))
1874 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1875 return (error);
1876}
1877
1878/*
1879 * Write wrapper for fifos.
1880 */
1881int
1882ufsfifo_write(ap)
1883 struct vop_write_args /* {
1884 struct vnode *a_vp;
1885 struct uio *a_uio;
1886 int a_ioflag;
1887 struct ucred *a_cred;
1888 } */ *ap;
1889{
1890 int error, resid;
1891 struct inode *ip;
1892 struct uio *uio;
1893
1894 uio = ap->a_uio;
1895 resid = uio->uio_resid;
1896 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap);
1897 ip = VTOI(ap->a_vp);
1898 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1899 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1900 return (error);
1901}
1902
1903/*
1904 * Close wrapper for fifos.
1905 *
1906 * Update the times on the inode then do device close.
1907 */
1908int
1909ufsfifo_close(ap)
1910 struct vop_close_args /* {
1911 struct vnode *a_vp;
1912 int a_fflag;
1913 struct ucred *a_cred;
1914 struct thread *a_td;
1915 } */ *ap;
1916{
1917 struct vnode *vp = ap->a_vp;
1918
1919 lwkt_gettoken(&vp->v_interlock);
1920 if (vp->v_usecount > 1)
1921 ufs_itimes(vp);
1922 lwkt_reltoken(&vp->v_interlock);
1923 return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
1924}
1925
1926/*
1927 * Kqfilter wrapper for fifos.
1928 *
1929 * Fall through to ufs kqfilter routines if needed
1930 */
1931int
1932ufsfifo_kqfilter(ap)
1933 struct vop_kqfilter_args *ap;
1934{
1935 int error;
1936
1937 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_kqfilter), ap);
1938 if (error)
1939 error = ufs_kqfilter(ap);
1940 return (error);
1941}
1942
1943/*
1944 * Return POSIX pathconf information applicable to ufs filesystems.
1945 */
1946int
1947ufs_pathconf(ap)
1948 struct vop_pathconf_args /* {
1949 struct vnode *a_vp;
1950 int a_name;
1951 int *a_retval;
1952 } */ *ap;
1953{
1954
1955 switch (ap->a_name) {
1956 case _PC_LINK_MAX:
1957 *ap->a_retval = LINK_MAX;
1958 return (0);
1959 case _PC_NAME_MAX:
1960 *ap->a_retval = NAME_MAX;
1961 return (0);
1962 case _PC_PATH_MAX:
1963 *ap->a_retval = PATH_MAX;
1964 return (0);
1965 case _PC_PIPE_BUF:
1966 *ap->a_retval = PIPE_BUF;
1967 return (0);
1968 case _PC_CHOWN_RESTRICTED:
1969 *ap->a_retval = 1;
1970 return (0);
1971 case _PC_NO_TRUNC:
1972 *ap->a_retval = 1;
1973 return (0);
1974 default:
1975 return (EINVAL);
1976 }
1977 /* NOTREACHED */
1978}
1979
1980/*
1981 * Advisory record locking support
1982 */
1983int
1984ufs_advlock(ap)
1985 struct vop_advlock_args /* {
1986 struct vnode *a_vp;
1987 caddr_t a_id;
1988 int a_op;
1989 struct flock *a_fl;
1990 int a_flags;
1991 } */ *ap;
1992{
1993 register struct inode *ip = VTOI(ap->a_vp);
1994
1995 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1996}
1997
1998/*
1999 * Initialize the vnode associated with a new inode, handle aliased
2000 * vnodes.
2001 */
2002int
2003ufs_vinit(mntp, specops, fifoops, vpp)
2004 struct mount *mntp;
2005 vop_t **specops;
2006 vop_t **fifoops;
2007 struct vnode **vpp;
2008{
2009 struct inode *ip;
2010 struct vnode *vp;
2011 struct timeval tv;
2012
2013 vp = *vpp;
2014 ip = VTOI(vp);
2015 switch(vp->v_type = IFTOVT(ip->i_mode)) {
2016 case VCHR:
2017 case VBLK:
2018 vp->v_op = specops;
2019 addaliasu(vp, ip->i_rdev);
2020 break;
2021 case VFIFO:
2022 vp->v_op = fifoops;
2023 break;
2024 default:
2025 break;
2026
2027 }
2028 if (ip->i_number == ROOTINO)
2029 vp->v_flag |= VROOT;
2030 /*
2031 * Initialize modrev times
2032 */
2033 getmicrouptime(&tv);
2034 SETHIGH(ip->i_modrev, tv.tv_sec);
2035 SETLOW(ip->i_modrev, tv.tv_usec * 4294);
2036 *vpp = vp;
2037 return (0);
2038}
2039
2040/*
2041 * Allocate a new inode.
2042 */
2043int
2044ufs_makeinode(mode, dvp, vpp, cnp)
2045 int mode;
2046 struct vnode *dvp;
2047 struct vnode **vpp;
2048 struct componentname *cnp;
2049{
2050 register struct inode *ip, *pdir;
2051 struct direct newdir;
2052 struct vnode *tvp;
2053 int error;
2054
2055 pdir = VTOI(dvp);
2056#ifdef DIAGNOSTIC
2057 if ((cnp->cn_flags & HASBUF) == 0)
2058 panic("ufs_makeinode: no name");
2059#endif
2060 *vpp = NULL;
2061 if ((mode & IFMT) == 0)
2062 mode |= IFREG;
2063
2064 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
2065 if (error)
2066 return (error);
2067 ip = VTOI(tvp);
2068 ip->i_gid = pdir->i_gid;
2069#ifdef SUIDDIR
2070 {
2071#ifdef QUOTA
2072 struct ucred ucred, *ucp;
2073 ucp = cnp->cn_cred;
2074#endif
2075 /*
2076 * If we are not the owner of the directory,
2077 * and we are hacking owners here, (only do this where told to)
2078 * and we are not giving it TO root, (would subvert quotas)
2079 * then go ahead and give it to the other user.
2080 * Note that this drops off the execute bits for security.
2081 */
2082 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
2083 (pdir->i_mode & ISUID) &&
2084 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
2085 ip->i_uid = pdir->i_uid;
2086 mode &= ~07111;
2087#ifdef QUOTA
2088 /*
2089 * Make sure the correct user gets charged
2090 * for the space.
2091 * Quickly knock up a dummy credential for the victim.
2092 * XXX This seems to never be accessed out of our
2093 * context so a stack variable is ok.
2094 */
2095 ucred.cr_ref = 1;
2096 ucred.cr_uid = ip->i_uid;
2097 ucred.cr_ngroups = 1;
2098 ucred.cr_groups[0] = pdir->i_gid;
2099 ucp = &ucred;
2100#endif
2101 } else
2102 ip->i_uid = cnp->cn_cred->cr_uid;
2103
2104#ifdef QUOTA
2105 if ((error = getinoquota(ip)) ||
2106 (error = chkiq(ip, 1, ucp, 0))) {
2107 UFS_VFREE(tvp, ip->i_number, mode);
2108 vput(tvp);
2109 return (error);
2110 }
2111#endif
2112 }
2113#else /* !SUIDDIR */
2114 ip->i_uid = cnp->cn_cred->cr_uid;
2115#ifdef QUOTA
2116 if ((error = getinoquota(ip)) ||
2117 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2118 UFS_VFREE(tvp, ip->i_number, mode);
2119 vput(tvp);
2120 return (error);
2121 }
2122#endif
2123#endif /* !SUIDDIR */
2124 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2125 ip->i_mode = mode;
2126 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2127 ip->i_effnlink = 1;
2128 ip->i_nlink = 1;
2129 if (DOINGSOFTDEP(tvp))
2130 softdep_change_linkcnt(ip);
2131 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
2132 suser_cred(cnp->cn_cred, 0)) {
2133 ip->i_mode &= ~ISGID;
2134 }
2135
2136 if (cnp->cn_flags & ISWHITEOUT)
2137 ip->i_flags |= UF_OPAQUE;
2138
2139 /*
2140 * Make sure inode goes to disk before directory entry.
2141 */
2142 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2143 if (error)
2144 goto bad;
2145 ufs_makedirentry(ip, cnp, &newdir);
2146 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2147 if (error)
2148 goto bad;
2149 *vpp = tvp;
2150 return (0);
2151
2152bad:
2153 /*
2154 * Write error occurred trying to update the inode
2155 * or the directory so must deallocate the inode.
2156 */
2157 ip->i_effnlink = 0;
2158 ip->i_nlink = 0;
2159 ip->i_flag |= IN_CHANGE;
2160 if (DOINGSOFTDEP(tvp))
2161 softdep_change_linkcnt(ip);
2162 vput(tvp);
2163 return (error);
2164}
2165
2166static int
2167ufs_missingop(ap)
2168 struct vop_generic_args *ap;
2169{
2170
2171 panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2172 return (EOPNOTSUPP);
2173}
2174
2175static struct filterops ufsread_filtops =
2176 { 1, NULL, filt_ufsdetach, filt_ufsread };
2177static struct filterops ufswrite_filtops =
2178 { 1, NULL, filt_ufsdetach, filt_ufswrite };
2179static struct filterops ufsvnode_filtops =
2180 { 1, NULL, filt_ufsdetach, filt_ufsvnode };
2181
2182static int
2183ufs_kqfilter(ap)
2184 struct vop_kqfilter_args /* {
2185 struct vnode *a_vp;
2186 struct knote *a_kn;
2187 } */ *ap;
2188{
2189 struct vnode *vp = ap->a_vp;
2190 struct knote *kn = ap->a_kn;
2191
2192 switch (kn->kn_filter) {
2193 case EVFILT_READ:
2194 kn->kn_fop = &ufsread_filtops;
2195 break;
2196 case EVFILT_WRITE:
2197 kn->kn_fop = &ufswrite_filtops;
2198 break;
2199 case EVFILT_VNODE:
2200 kn->kn_fop = &ufsvnode_filtops;
2201 break;
2202 default:
2203 return (1);
2204 }
2205
2206 kn->kn_hook = (caddr_t)vp;
2207
2208 lwkt_gettoken(&vp->v_pollinfo.vpi_token);
2209 SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
2210 lwkt_reltoken(&vp->v_pollinfo.vpi_token);
2211
2212 return (0);
2213}
2214
2215static void
2216filt_ufsdetach(struct knote *kn)
2217{
2218 struct vnode *vp = (struct vnode *)kn->kn_hook;
2219
2220 lwkt_gettoken(&vp->v_pollinfo.vpi_token);
2221 SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2222 kn, knote, kn_selnext);
2223 lwkt_reltoken(&vp->v_pollinfo.vpi_token);
2224}
2225
2226/*ARGSUSED*/
2227static int
2228filt_ufsread(struct knote *kn, long hint)
2229{
2230 struct vnode *vp = (struct vnode *)kn->kn_hook;
2231 struct inode *ip = VTOI(vp);
2232
2233 /*
2234 * filesystem is gone, so set the EOF flag and schedule
2235 * the knote for deletion.
2236 */
2237 if (hint == NOTE_REVOKE) {
2238 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2239 return (1);
2240 }
2241
2242 kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2243 return (kn->kn_data != 0);
2244}
2245
2246/*ARGSUSED*/
2247static int
2248filt_ufswrite(struct knote *kn, long hint)
2249{
2250
2251 /*
2252 * filesystem is gone, so set the EOF flag and schedule
2253 * the knote for deletion.
2254 */
2255 if (hint == NOTE_REVOKE)
2256 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2257
2258 kn->kn_data = 0;
2259 return (1);
2260}
2261
2262static int
2263filt_ufsvnode(struct knote *kn, long hint)
2264{
2265
2266 if (kn->kn_sfflags & hint)
2267 kn->kn_fflags |= hint;
2268 if (hint == NOTE_REVOKE) {
2269 kn->kn_flags |= EV_EOF;
2270 return (1);
2271 }
2272 return (kn->kn_fflags != 0);
2273}
2274
2275/* Global vfs data structures for ufs. */
2276static vop_t **ufs_vnodeop_p;
2277static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2278 { &vop_default_desc, (vop_t *) vop_defaultop },
2279 { &vop_fsync_desc, (vop_t *) ufs_missingop },
2280 { &vop_read_desc, (vop_t *) ufs_missingop },
2281 { &vop_reallocblks_desc, (vop_t *) ufs_missingop },
2282 { &vop_write_desc, (vop_t *) ufs_missingop },
2283 { &vop_access_desc, (vop_t *) ufs_access },
2284 { &vop_advlock_desc, (vop_t *) ufs_advlock },
2285 { &vop_bmap_desc, (vop_t *) ufs_bmap },
2286 { &vop_cachedlookup_desc, (vop_t *) ufs_lookup },
2287 { &vop_close_desc, (vop_t *) ufs_close },
2288 { &vop_create_desc, (vop_t *) ufs_create },
2289 { &vop_getattr_desc, (vop_t *) ufs_getattr },
2290 { &vop_inactive_desc, (vop_t *) ufs_inactive },
2291 { &vop_islocked_desc, (vop_t *) vop_stdislocked },
2292 { &vop_link_desc, (vop_t *) ufs_link },
2293 { &vop_lock_desc, (vop_t *) vop_stdlock },
2294 { &vop_lookup_desc, (vop_t *) vfs_cache_lookup },
2295 { &vop_mkdir_desc, (vop_t *) ufs_mkdir },
2296 { &vop_mknod_desc, (vop_t *) ufs_mknod },
2297 { &vop_mmap_desc, (vop_t *) ufs_mmap },
2298 { &vop_open_desc, (vop_t *) ufs_open },
2299 { &vop_pathconf_desc, (vop_t *) ufs_pathconf },
2300 { &vop_poll_desc, (vop_t *) vop_stdpoll },
2301 { &vop_kqfilter_desc, (vop_t *) ufs_kqfilter },
2302 { &vop_print_desc, (vop_t *) ufs_print },
2303 { &vop_readdir_desc, (vop_t *) ufs_readdir },
2304 { &vop_readlink_desc, (vop_t *) ufs_readlink },
2305 { &vop_reclaim_desc, (vop_t *) ufs_reclaim },
2306 { &vop_remove_desc, (vop_t *) ufs_remove },
2307 { &vop_rename_desc, (vop_t *) ufs_rename },
2308 { &vop_rmdir_desc, (vop_t *) ufs_rmdir },
2309 { &vop_setattr_desc, (vop_t *) ufs_setattr },
2310 { &vop_strategy_desc, (vop_t *) ufs_strategy },
2311 { &vop_symlink_desc, (vop_t *) ufs_symlink },
2312 { &vop_unlock_desc, (vop_t *) vop_stdunlock },
2313 { &vop_whiteout_desc, (vop_t *) ufs_whiteout },
2314 { NULL, NULL }
2315};
2316static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2317 { &ufs_vnodeop_p, ufs_vnodeop_entries };
2318
2319static vop_t **ufs_specop_p;
2320static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2321 { &vop_default_desc, (vop_t *) spec_vnoperate },
2322 { &vop_fsync_desc, (vop_t *) ufs_missingop },
2323 { &vop_access_desc, (vop_t *) ufs_access },
2324 { &vop_close_desc, (vop_t *) ufsspec_close },
2325 { &vop_getattr_desc, (vop_t *) ufs_getattr },
2326 { &vop_inactive_desc, (vop_t *) ufs_inactive },
2327 { &vop_islocked_desc, (vop_t *) vop_stdislocked },
2328 { &vop_lock_desc, (vop_t *) vop_stdlock },
2329 { &vop_print_desc, (vop_t *) ufs_print },
2330 { &vop_read_desc, (vop_t *) ufsspec_read },
2331 { &vop_reclaim_desc, (vop_t *) ufs_reclaim },
2332 { &vop_setattr_desc, (vop_t *) ufs_setattr },
2333 { &vop_unlock_desc, (vop_t *) vop_stdunlock },
2334 { &vop_write_desc, (vop_t *) ufsspec_write },
2335 { NULL, NULL }
2336};
2337static struct vnodeopv_desc ufs_specop_opv_desc =
2338 { &ufs_specop_p, ufs_specop_entries };
2339
2340static vop_t **ufs_fifoop_p;
2341static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2342 { &vop_default_desc, (vop_t *) fifo_vnoperate },
2343 { &vop_fsync_desc, (vop_t *) ufs_missingop },
2344 { &vop_access_desc, (vop_t *) ufs_access },
2345 { &vop_close_desc, (vop_t *) ufsfifo_close },
2346 { &vop_getattr_desc, (vop_t *) ufs_getattr },
2347 { &vop_inactive_desc, (vop_t *) ufs_inactive },
2348 { &vop_islocked_desc, (vop_t *) vop_stdislocked },
2349 { &vop_kqfilter_desc, (vop_t *) ufsfifo_kqfilter },
2350 { &vop_lock_desc, (vop_t *) vop_stdlock },
2351 { &vop_print_desc, (vop_t *) ufs_print },
2352 { &vop_read_desc, (vop_t *) ufsfifo_read },
2353 { &vop_reclaim_desc, (vop_t *) ufs_reclaim },
2354 { &vop_setattr_desc, (vop_t *) ufs_setattr },
2355 { &vop_unlock_desc, (vop_t *) vop_stdunlock },
2356 { &vop_write_desc, (vop_t *) ufsfifo_write },
2357 { NULL, NULL }
2358};
2359static struct vnodeopv_desc ufs_fifoop_opv_desc =
2360 { &ufs_fifoop_p, ufs_fifoop_entries };
2361
2362VNODEOP_SET(ufs_vnodeop_opv_desc);
2363VNODEOP_SET(ufs_specop_opv_desc);
2364VNODEOP_SET(ufs_fifoop_opv_desc);
2365
2366int
2367ufs_vnoperate(ap)
2368 struct vop_generic_args /* {
2369 struct vnodeop_desc *a_desc;
2370 } */ *ap;
2371{
2372 return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap));
2373}
2374
2375int
2376ufs_vnoperatefifo(ap)
2377 struct vop_generic_args /* {
2378 struct vnodeop_desc *a_desc;
2379 } */ *ap;
2380{
2381 return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap));
2382}
2383
2384int
2385ufs_vnoperatespec(ap)
2386 struct vop_generic_args /* {
2387 struct vnodeop_desc *a_desc;
2388 } */ *ap;
2389{
2390 return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap));
2391}