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 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
35 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD: releng/10.2/sys/ufs/ufs/ufs_vnops.c 276648 2015-01-04 00:46:06Z kib $");
39
40 #include "opt_quota.h"
41 #include "opt_suiddir.h"
42 #include "opt_ufs.h"
43 #include "opt_ffs.h"
44
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/malloc.h>
48 #include <sys/namei.h>
49 #include <sys/kernel.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
52 #include <sys/stat.h>
53 #include <sys/bio.h>
54 #include <sys/buf.h>
55 #include <sys/mount.h>
56 #include <sys/priv.h>
57 #include <sys/refcount.h>
58 #include <sys/unistd.h>
59 #include <sys/vnode.h>
60 #include <sys/dirent.h>
61 #include <sys/lockf.h>
62 #include <sys/conf.h>
63 #include <sys/acl.h>
64
65 #include <security/mac/mac_framework.h>
66
67 #include <sys/file.h> /* XXX */
68
69 #include <vm/vm.h>
70 #include <vm/vm_extern.h>
71
72 #include <ufs/ufs/acl.h>
73 #include <ufs/ufs/extattr.h>
74 #include <ufs/ufs/quota.h>
75 #include <ufs/ufs/inode.h>
76 #include <ufs/ufs/dir.h>
77 #include <ufs/ufs/ufsmount.h>
78 #include <ufs/ufs/ufs_extern.h>
79 #ifdef UFS_DIRHASH
80 #include <ufs/ufs/dirhash.h>
81 #endif
82 #ifdef UFS_GJOURNAL
83 #include <ufs/ufs/gjournal.h>
84 FEATURE(ufs_gjournal, "Journaling support through GEOM for UFS");
85 #endif
86
87 #ifdef QUOTA
88 FEATURE(ufs_quota, "UFS disk quotas support");
89 FEATURE(ufs_quota64, "64bit UFS disk quotas support");
90 #endif
91
92 #ifdef SUIDDIR
93 FEATURE(suiddir,
94 "Give all new files in directory the same ownership as the directory");
95 #endif
96
97
98 #include <ufs/ffs/ffs_extern.h>
99
100 static vop_accessx_t ufs_accessx;
101 static int ufs_chmod(struct vnode *, int, struct ucred *, struct thread *);
102 static int ufs_chown(struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
103 static vop_close_t ufs_close;
104 static vop_create_t ufs_create;
105 static vop_getattr_t ufs_getattr;
106 static vop_ioctl_t ufs_ioctl;
107 static vop_link_t ufs_link;
108 static int ufs_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
109 static vop_markatime_t ufs_markatime;
110 static vop_mkdir_t ufs_mkdir;
111 static vop_mknod_t ufs_mknod;
112 static vop_open_t ufs_open;
113 static vop_pathconf_t ufs_pathconf;
114 static vop_print_t ufs_print;
115 static vop_readlink_t ufs_readlink;
116 static vop_remove_t ufs_remove;
117 static vop_rename_t ufs_rename;
118 static vop_rmdir_t ufs_rmdir;
119 static vop_setattr_t ufs_setattr;
120 static vop_strategy_t ufs_strategy;
121 static vop_symlink_t ufs_symlink;
122 static vop_whiteout_t ufs_whiteout;
123 static vop_close_t ufsfifo_close;
124 static vop_kqfilter_t ufsfifo_kqfilter;
125 static vop_pathconf_t ufsfifo_pathconf;
126
127 SYSCTL_NODE(_vfs, OID_AUTO, ufs, CTLFLAG_RD, 0, "UFS filesystem");
128
129 /*
130 * A virgin directory (no blushing please).
131 */
132 static struct dirtemplate mastertemplate = {
133 0, 12, DT_DIR, 1, ".",
134 0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
135 };
136 static struct odirtemplate omastertemplate = {
137 0, 12, 1, ".",
138 0, DIRBLKSIZ - 12, 2, ".."
139 };
140
141 static void
142 ufs_itimes_locked(struct vnode *vp)
143 {
144 struct inode *ip;
145 struct timespec ts;
146
147 ASSERT_VI_LOCKED(vp, __func__);
148
149 ip = VTOI(vp);
150 if (UFS_RDONLY(ip))
151 goto out;
152 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
153 return;
154
155 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
156 ip->i_flag |= IN_LAZYMOD;
157 else if (((vp->v_mount->mnt_kern_flag &
158 (MNTK_SUSPENDED | MNTK_SUSPEND)) == 0) ||
159 (ip->i_flag & (IN_CHANGE | IN_UPDATE)))
160 ip->i_flag |= IN_MODIFIED;
161 else if (ip->i_flag & IN_ACCESS)
162 ip->i_flag |= IN_LAZYACCESS;
163 vfs_timestamp(&ts);
164 if (ip->i_flag & IN_ACCESS) {
165 DIP_SET(ip, i_atime, ts.tv_sec);
166 DIP_SET(ip, i_atimensec, ts.tv_nsec);
167 }
168 if (ip->i_flag & IN_UPDATE) {
169 DIP_SET(ip, i_mtime, ts.tv_sec);
170 DIP_SET(ip, i_mtimensec, ts.tv_nsec);
171 }
172 if (ip->i_flag & IN_CHANGE) {
173 DIP_SET(ip, i_ctime, ts.tv_sec);
174 DIP_SET(ip, i_ctimensec, ts.tv_nsec);
175 DIP_SET(ip, i_modrev, DIP(ip, i_modrev) + 1);
176 }
177
178 out:
179 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
180 }
181
182 void
183 ufs_itimes(struct vnode *vp)
184 {
185
186 VI_LOCK(vp);
187 ufs_itimes_locked(vp);
188 VI_UNLOCK(vp);
189 }
190
191 /*
192 * Create a regular file
193 */
194 static int
195 ufs_create(ap)
196 struct vop_create_args /* {
197 struct vnode *a_dvp;
198 struct vnode **a_vpp;
199 struct componentname *a_cnp;
200 struct vattr *a_vap;
201 } */ *ap;
202 {
203 int error;
204
205 error =
206 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
207 ap->a_dvp, ap->a_vpp, ap->a_cnp);
208 if (error != 0)
209 return (error);
210 if ((ap->a_cnp->cn_flags & MAKEENTRY) != 0)
211 cache_enter(ap->a_dvp, *ap->a_vpp, ap->a_cnp);
212 return (0);
213 }
214
215 /*
216 * Mknod vnode call
217 */
218 /* ARGSUSED */
219 static int
220 ufs_mknod(ap)
221 struct vop_mknod_args /* {
222 struct vnode *a_dvp;
223 struct vnode **a_vpp;
224 struct componentname *a_cnp;
225 struct vattr *a_vap;
226 } */ *ap;
227 {
228 struct vattr *vap = ap->a_vap;
229 struct vnode **vpp = ap->a_vpp;
230 struct inode *ip;
231 ino_t ino;
232 int error;
233
234 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
235 ap->a_dvp, vpp, ap->a_cnp);
236 if (error)
237 return (error);
238 ip = VTOI(*vpp);
239 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
240 if (vap->va_rdev != VNOVAL) {
241 /*
242 * Want to be able to use this to make badblock
243 * inodes, so don't truncate the dev number.
244 */
245 DIP_SET(ip, i_rdev, vap->va_rdev);
246 }
247 /*
248 * Remove inode, then reload it through VFS_VGET so it is
249 * checked to see if it is an alias of an existing entry in
250 * the inode cache. XXX I don't believe this is necessary now.
251 */
252 (*vpp)->v_type = VNON;
253 ino = ip->i_number; /* Save this before vgone() invalidates ip. */
254 vgone(*vpp);
255 vput(*vpp);
256 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
257 if (error) {
258 *vpp = NULL;
259 return (error);
260 }
261 return (0);
262 }
263
264 /*
265 * Open called.
266 */
267 /* ARGSUSED */
268 static int
269 ufs_open(struct vop_open_args *ap)
270 {
271 struct vnode *vp = ap->a_vp;
272 struct inode *ip;
273
274 if (vp->v_type == VCHR || vp->v_type == VBLK)
275 return (EOPNOTSUPP);
276
277 ip = VTOI(vp);
278 /*
279 * Files marked append-only must be opened for appending.
280 */
281 if ((ip->i_flags & APPEND) &&
282 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
283 return (EPERM);
284 vnode_create_vobject(vp, DIP(ip, i_size), ap->a_td);
285 return (0);
286 }
287
288 /*
289 * Close called.
290 *
291 * Update the times on the inode.
292 */
293 /* ARGSUSED */
294 static int
295 ufs_close(ap)
296 struct vop_close_args /* {
297 struct vnode *a_vp;
298 int a_fflag;
299 struct ucred *a_cred;
300 struct thread *a_td;
301 } */ *ap;
302 {
303 struct vnode *vp = ap->a_vp;
304 int usecount;
305
306 VI_LOCK(vp);
307 usecount = vp->v_usecount;
308 if (usecount > 1)
309 ufs_itimes_locked(vp);
310 VI_UNLOCK(vp);
311 return (0);
312 }
313
314 static int
315 ufs_accessx(ap)
316 struct vop_accessx_args /* {
317 struct vnode *a_vp;
318 accmode_t a_accmode;
319 struct ucred *a_cred;
320 struct thread *a_td;
321 } */ *ap;
322 {
323 struct vnode *vp = ap->a_vp;
324 struct inode *ip = VTOI(vp);
325 accmode_t accmode = ap->a_accmode;
326 int error;
327 #ifdef QUOTA
328 int relocked;
329 #endif
330 #ifdef UFS_ACL
331 struct acl *acl;
332 acl_type_t type;
333 #endif
334
335 /*
336 * Disallow write attempts on read-only filesystems;
337 * unless the file is a socket, fifo, or a block or
338 * character device resident on the filesystem.
339 */
340 if (accmode & VMODIFY_PERMS) {
341 switch (vp->v_type) {
342 case VDIR:
343 case VLNK:
344 case VREG:
345 if (vp->v_mount->mnt_flag & MNT_RDONLY)
346 return (EROFS);
347 #ifdef QUOTA
348 /*
349 * Inode is accounted in the quotas only if struct
350 * dquot is attached to it. VOP_ACCESS() is called
351 * from vn_open_cred() and provides a convenient
352 * point to call getinoquota().
353 */
354 if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
355
356 /*
357 * Upgrade vnode lock, since getinoquota()
358 * requires exclusive lock to modify inode.
359 */
360 relocked = 1;
361 vhold(vp);
362 vn_lock(vp, LK_UPGRADE | LK_RETRY);
363 VI_LOCK(vp);
364 if (vp->v_iflag & VI_DOOMED) {
365 vdropl(vp);
366 error = ENOENT;
367 goto relock;
368 }
369 vdropl(vp);
370 } else
371 relocked = 0;
372 error = getinoquota(ip);
373 relock:
374 if (relocked)
375 vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
376 if (error != 0)
377 return (error);
378 #endif
379 break;
380 default:
381 break;
382 }
383 }
384
385 /*
386 * If immutable bit set, nobody gets to write it. "& ~VADMIN_PERMS"
387 * is here, because without it, * it would be impossible for the owner
388 * to remove the IMMUTABLE flag.
389 */
390 if ((accmode & (VMODIFY_PERMS & ~VADMIN_PERMS)) &&
391 (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
392 return (EPERM);
393
394 #ifdef UFS_ACL
395 if ((vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS)) != 0) {
396 if (vp->v_mount->mnt_flag & MNT_NFS4ACLS)
397 type = ACL_TYPE_NFS4;
398 else
399 type = ACL_TYPE_ACCESS;
400
401 acl = acl_alloc(M_WAITOK);
402 if (type == ACL_TYPE_NFS4)
403 error = ufs_getacl_nfs4_internal(vp, acl, ap->a_td);
404 else
405 error = VOP_GETACL(vp, type, acl, ap->a_cred, ap->a_td);
406 switch (error) {
407 case 0:
408 if (type == ACL_TYPE_NFS4) {
409 error = vaccess_acl_nfs4(vp->v_type, ip->i_uid,
410 ip->i_gid, acl, accmode, ap->a_cred, NULL);
411 } else {
412 error = vfs_unixify_accmode(&accmode);
413 if (error == 0)
414 error = vaccess_acl_posix1e(vp->v_type, ip->i_uid,
415 ip->i_gid, acl, accmode, ap->a_cred, NULL);
416 }
417 break;
418 default:
419 if (error != EOPNOTSUPP)
420 printf(
421 "ufs_accessx(): Error retrieving ACL on object (%d).\n",
422 error);
423 /*
424 * XXX: Fall back until debugged. Should
425 * eventually possibly log an error, and return
426 * EPERM for safety.
427 */
428 error = vfs_unixify_accmode(&accmode);
429 if (error == 0)
430 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
431 ip->i_gid, accmode, ap->a_cred, NULL);
432 }
433 acl_free(acl);
434
435 return (error);
436 }
437 #endif /* !UFS_ACL */
438 error = vfs_unixify_accmode(&accmode);
439 if (error == 0)
440 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
441 accmode, ap->a_cred, NULL);
442 return (error);
443 }
444
445 /* ARGSUSED */
446 static int
447 ufs_getattr(ap)
448 struct vop_getattr_args /* {
449 struct vnode *a_vp;
450 struct vattr *a_vap;
451 struct ucred *a_cred;
452 } */ *ap;
453 {
454 struct vnode *vp = ap->a_vp;
455 struct inode *ip = VTOI(vp);
456 struct vattr *vap = ap->a_vap;
457
458 VI_LOCK(vp);
459 ufs_itimes_locked(vp);
460 if (ip->i_ump->um_fstype == UFS1) {
461 vap->va_atime.tv_sec = ip->i_din1->di_atime;
462 vap->va_atime.tv_nsec = ip->i_din1->di_atimensec;
463 } else {
464 vap->va_atime.tv_sec = ip->i_din2->di_atime;
465 vap->va_atime.tv_nsec = ip->i_din2->di_atimensec;
466 }
467 VI_UNLOCK(vp);
468 /*
469 * Copy from inode table
470 */
471 vap->va_fsid = dev2udev(ip->i_dev);
472 vap->va_fileid = ip->i_number;
473 vap->va_mode = ip->i_mode & ~IFMT;
474 vap->va_nlink = ip->i_effnlink;
475 vap->va_uid = ip->i_uid;
476 vap->va_gid = ip->i_gid;
477 if (ip->i_ump->um_fstype == UFS1) {
478 vap->va_rdev = ip->i_din1->di_rdev;
479 vap->va_size = ip->i_din1->di_size;
480 vap->va_mtime.tv_sec = ip->i_din1->di_mtime;
481 vap->va_mtime.tv_nsec = ip->i_din1->di_mtimensec;
482 vap->va_ctime.tv_sec = ip->i_din1->di_ctime;
483 vap->va_ctime.tv_nsec = ip->i_din1->di_ctimensec;
484 vap->va_bytes = dbtob((u_quad_t)ip->i_din1->di_blocks);
485 vap->va_filerev = ip->i_din1->di_modrev;
486 } else {
487 vap->va_rdev = ip->i_din2->di_rdev;
488 vap->va_size = ip->i_din2->di_size;
489 vap->va_mtime.tv_sec = ip->i_din2->di_mtime;
490 vap->va_mtime.tv_nsec = ip->i_din2->di_mtimensec;
491 vap->va_ctime.tv_sec = ip->i_din2->di_ctime;
492 vap->va_ctime.tv_nsec = ip->i_din2->di_ctimensec;
493 vap->va_birthtime.tv_sec = ip->i_din2->di_birthtime;
494 vap->va_birthtime.tv_nsec = ip->i_din2->di_birthnsec;
495 vap->va_bytes = dbtob((u_quad_t)ip->i_din2->di_blocks);
496 vap->va_filerev = ip->i_din2->di_modrev;
497 }
498 vap->va_flags = ip->i_flags;
499 vap->va_gen = ip->i_gen;
500 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
501 vap->va_type = IFTOVT(ip->i_mode);
502 return (0);
503 }
504
505 /*
506 * Set attribute vnode op. called from several syscalls
507 */
508 static int
509 ufs_setattr(ap)
510 struct vop_setattr_args /* {
511 struct vnode *a_vp;
512 struct vattr *a_vap;
513 struct ucred *a_cred;
514 } */ *ap;
515 {
516 struct vattr *vap = ap->a_vap;
517 struct vnode *vp = ap->a_vp;
518 struct inode *ip = VTOI(vp);
519 struct ucred *cred = ap->a_cred;
520 struct thread *td = curthread;
521 int error;
522
523 /*
524 * Check for unsettable attributes.
525 */
526 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
527 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
528 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
529 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
530 return (EINVAL);
531 }
532 if (vap->va_flags != VNOVAL) {
533 if ((vap->va_flags & ~(SF_APPEND | SF_ARCHIVED | SF_IMMUTABLE |
534 SF_NOUNLINK | SF_SNAPSHOT | UF_APPEND | UF_ARCHIVE |
535 UF_HIDDEN | UF_IMMUTABLE | UF_NODUMP | UF_NOUNLINK |
536 UF_OFFLINE | UF_OPAQUE | UF_READONLY | UF_REPARSE |
537 UF_SPARSE | UF_SYSTEM)) != 0)
538 return (EOPNOTSUPP);
539 if (vp->v_mount->mnt_flag & MNT_RDONLY)
540 return (EROFS);
541 /*
542 * Callers may only modify the file flags on objects they
543 * have VADMIN rights for.
544 */
545 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
546 return (error);
547 /*
548 * Unprivileged processes are not permitted to unset system
549 * flags, or modify flags if any system flags are set.
550 * Privileged non-jail processes may not modify system flags
551 * if securelevel > 0 and any existing system flags are set.
552 * Privileged jail processes behave like privileged non-jail
553 * processes if the security.jail.chflags_allowed sysctl is
554 * is non-zero; otherwise, they behave like unprivileged
555 * processes.
556 */
557 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
558 if (ip->i_flags &
559 (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
560 error = securelevel_gt(cred, 0);
561 if (error)
562 return (error);
563 }
564 /* The snapshot flag cannot be toggled. */
565 if ((vap->va_flags ^ ip->i_flags) & SF_SNAPSHOT)
566 return (EPERM);
567 } else {
568 if (ip->i_flags &
569 (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
570 ((vap->va_flags ^ ip->i_flags) & SF_SETTABLE))
571 return (EPERM);
572 }
573 ip->i_flags = vap->va_flags;
574 DIP_SET(ip, i_flags, vap->va_flags);
575 ip->i_flag |= IN_CHANGE;
576 error = UFS_UPDATE(vp, 0);
577 if (ip->i_flags & (IMMUTABLE | APPEND))
578 return (error);
579 }
580 /*
581 * If immutable or append, no one can change any of its attributes
582 * except the ones already handled (in some cases, file flags
583 * including the immutability flags themselves for the superuser).
584 */
585 if (ip->i_flags & (IMMUTABLE | APPEND))
586 return (EPERM);
587 /*
588 * Go through the fields and update iff not VNOVAL.
589 */
590 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
591 if (vp->v_mount->mnt_flag & MNT_RDONLY)
592 return (EROFS);
593 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred,
594 td)) != 0)
595 return (error);
596 }
597 if (vap->va_size != VNOVAL) {
598 /*
599 * XXX most of the following special cases should be in
600 * callers instead of in N filesystems. The VDIR check
601 * mostly already is.
602 */
603 switch (vp->v_type) {
604 case VDIR:
605 return (EISDIR);
606 case VLNK:
607 case VREG:
608 /*
609 * Truncation should have an effect in these cases.
610 * Disallow it if the filesystem is read-only or
611 * the file is being snapshotted.
612 */
613 if (vp->v_mount->mnt_flag & MNT_RDONLY)
614 return (EROFS);
615 if ((ip->i_flags & SF_SNAPSHOT) != 0)
616 return (EPERM);
617 break;
618 default:
619 /*
620 * According to POSIX, the result is unspecified
621 * for file types other than regular files,
622 * directories and shared memory objects. We
623 * don't support shared memory objects in the file
624 * system, and have dubious support for truncating
625 * symlinks. Just ignore the request in other cases.
626 */
627 return (0);
628 }
629 if ((error = UFS_TRUNCATE(vp, vap->va_size, IO_NORMAL,
630 cred)) != 0)
631 return (error);
632 }
633 if (vap->va_atime.tv_sec != VNOVAL ||
634 vap->va_mtime.tv_sec != VNOVAL ||
635 vap->va_birthtime.tv_sec != VNOVAL) {
636 if (vp->v_mount->mnt_flag & MNT_RDONLY)
637 return (EROFS);
638 if ((ip->i_flags & SF_SNAPSHOT) != 0)
639 return (EPERM);
640 error = vn_utimes_perm(vp, vap, cred, td);
641 if (error != 0)
642 return (error);
643 if (vap->va_atime.tv_sec != VNOVAL)
644 ip->i_flag |= IN_ACCESS;
645 if (vap->va_mtime.tv_sec != VNOVAL)
646 ip->i_flag |= IN_CHANGE | IN_UPDATE;
647 if (vap->va_birthtime.tv_sec != VNOVAL &&
648 ip->i_ump->um_fstype == UFS2)
649 ip->i_flag |= IN_MODIFIED;
650 ufs_itimes(vp);
651 if (vap->va_atime.tv_sec != VNOVAL) {
652 DIP_SET(ip, i_atime, vap->va_atime.tv_sec);
653 DIP_SET(ip, i_atimensec, vap->va_atime.tv_nsec);
654 }
655 if (vap->va_mtime.tv_sec != VNOVAL) {
656 DIP_SET(ip, i_mtime, vap->va_mtime.tv_sec);
657 DIP_SET(ip, i_mtimensec, vap->va_mtime.tv_nsec);
658 }
659 if (vap->va_birthtime.tv_sec != VNOVAL &&
660 ip->i_ump->um_fstype == UFS2) {
661 ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec;
662 ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec;
663 }
664 error = UFS_UPDATE(vp, 0);
665 if (error)
666 return (error);
667 }
668 error = 0;
669 if (vap->va_mode != (mode_t)VNOVAL) {
670 if (vp->v_mount->mnt_flag & MNT_RDONLY)
671 return (EROFS);
672 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode &
673 (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH)))
674 return (EPERM);
675 error = ufs_chmod(vp, (int)vap->va_mode, cred, td);
676 }
677 return (error);
678 }
679
680 #ifdef UFS_ACL
681 static int
682 ufs_update_nfs4_acl_after_mode_change(struct vnode *vp, int mode,
683 int file_owner_id, struct ucred *cred, struct thread *td)
684 {
685 int error;
686 struct acl *aclp;
687
688 aclp = acl_alloc(M_WAITOK);
689 error = ufs_getacl_nfs4_internal(vp, aclp, td);
690 /*
691 * We don't have to handle EOPNOTSUPP here, as the filesystem claims
692 * it supports ACLs.
693 */
694 if (error)
695 goto out;
696
697 acl_nfs4_sync_acl_from_mode(aclp, mode, file_owner_id);
698 error = ufs_setacl_nfs4_internal(vp, aclp, td);
699
700 out:
701 acl_free(aclp);
702 return (error);
703 }
704 #endif /* UFS_ACL */
705
706 /*
707 * Mark this file's access time for update for vfs_mark_atime(). This
708 * is called from execve() and mmap().
709 */
710 static int
711 ufs_markatime(ap)
712 struct vop_markatime_args /* {
713 struct vnode *a_vp;
714 } */ *ap;
715 {
716 struct vnode *vp = ap->a_vp;
717 struct inode *ip = VTOI(vp);
718
719 VI_LOCK(vp);
720 ip->i_flag |= IN_ACCESS;
721 VI_UNLOCK(vp);
722 /*
723 * XXXKIB No UFS_UPDATE(ap->a_vp, 0) there.
724 */
725 return (0);
726 }
727
728 /*
729 * Change the mode on a file.
730 * Inode must be locked before calling.
731 */
732 static int
733 ufs_chmod(vp, mode, cred, td)
734 struct vnode *vp;
735 int mode;
736 struct ucred *cred;
737 struct thread *td;
738 {
739 struct inode *ip = VTOI(vp);
740 int error;
741
742 /*
743 * To modify the permissions on a file, must possess VADMIN
744 * for that file.
745 */
746 if ((error = VOP_ACCESSX(vp, VWRITE_ACL, cred, td)))
747 return (error);
748 /*
749 * Privileged processes may set the sticky bit on non-directories,
750 * as well as set the setgid bit on a file with a group that the
751 * process is not a member of. Both of these are allowed in
752 * jail(8).
753 */
754 if (vp->v_type != VDIR && (mode & S_ISTXT)) {
755 if (priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0))
756 return (EFTYPE);
757 }
758 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
759 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
760 if (error)
761 return (error);
762 }
763
764 /*
765 * Deny setting setuid if we are not the file owner.
766 */
767 if ((mode & ISUID) && ip->i_uid != cred->cr_uid) {
768 error = priv_check_cred(cred, PRIV_VFS_ADMIN, 0);
769 if (error)
770 return (error);
771 }
772
773 ip->i_mode &= ~ALLPERMS;
774 ip->i_mode |= (mode & ALLPERMS);
775 DIP_SET(ip, i_mode, ip->i_mode);
776 ip->i_flag |= IN_CHANGE;
777 #ifdef UFS_ACL
778 if ((vp->v_mount->mnt_flag & MNT_NFS4ACLS) != 0)
779 error = ufs_update_nfs4_acl_after_mode_change(vp, mode, ip->i_uid, cred, td);
780 #endif
781 if (error == 0 && (ip->i_flag & IN_CHANGE) != 0)
782 error = UFS_UPDATE(vp, 0);
783
784 return (error);
785 }
786
787 /*
788 * Perform chown operation on inode ip;
789 * inode must be locked prior to call.
790 */
791 static int
792 ufs_chown(vp, uid, gid, cred, td)
793 struct vnode *vp;
794 uid_t uid;
795 gid_t gid;
796 struct ucred *cred;
797 struct thread *td;
798 {
799 struct inode *ip = VTOI(vp);
800 uid_t ouid;
801 gid_t ogid;
802 int error = 0;
803 #ifdef QUOTA
804 int i;
805 ufs2_daddr_t change;
806 #endif
807
808 if (uid == (uid_t)VNOVAL)
809 uid = ip->i_uid;
810 if (gid == (gid_t)VNOVAL)
811 gid = ip->i_gid;
812 /*
813 * To modify the ownership of a file, must possess VADMIN for that
814 * file.
815 */
816 if ((error = VOP_ACCESSX(vp, VWRITE_OWNER, cred, td)))
817 return (error);
818 /*
819 * To change the owner of a file, or change the group of a file to a
820 * group of which we are not a member, the caller must have
821 * privilege.
822 */
823 if (((uid != ip->i_uid && uid != cred->cr_uid) ||
824 (gid != ip->i_gid && !groupmember(gid, cred))) &&
825 (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
826 return (error);
827 ogid = ip->i_gid;
828 ouid = ip->i_uid;
829 #ifdef QUOTA
830 if ((error = getinoquota(ip)) != 0)
831 return (error);
832 if (ouid == uid) {
833 dqrele(vp, ip->i_dquot[USRQUOTA]);
834 ip->i_dquot[USRQUOTA] = NODQUOT;
835 }
836 if (ogid == gid) {
837 dqrele(vp, ip->i_dquot[GRPQUOTA]);
838 ip->i_dquot[GRPQUOTA] = NODQUOT;
839 }
840 change = DIP(ip, i_blocks);
841 (void) chkdq(ip, -change, cred, CHOWN);
842 (void) chkiq(ip, -1, cred, CHOWN);
843 for (i = 0; i < MAXQUOTAS; i++) {
844 dqrele(vp, ip->i_dquot[i]);
845 ip->i_dquot[i] = NODQUOT;
846 }
847 #endif
848 ip->i_gid = gid;
849 DIP_SET(ip, i_gid, gid);
850 ip->i_uid = uid;
851 DIP_SET(ip, i_uid, uid);
852 #ifdef QUOTA
853 if ((error = getinoquota(ip)) == 0) {
854 if (ouid == uid) {
855 dqrele(vp, ip->i_dquot[USRQUOTA]);
856 ip->i_dquot[USRQUOTA] = NODQUOT;
857 }
858 if (ogid == gid) {
859 dqrele(vp, ip->i_dquot[GRPQUOTA]);
860 ip->i_dquot[GRPQUOTA] = NODQUOT;
861 }
862 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
863 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
864 goto good;
865 else
866 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
867 }
868 for (i = 0; i < MAXQUOTAS; i++) {
869 dqrele(vp, ip->i_dquot[i]);
870 ip->i_dquot[i] = NODQUOT;
871 }
872 }
873 ip->i_gid = ogid;
874 DIP_SET(ip, i_gid, ogid);
875 ip->i_uid = ouid;
876 DIP_SET(ip, i_uid, ouid);
877 if (getinoquota(ip) == 0) {
878 if (ouid == uid) {
879 dqrele(vp, ip->i_dquot[USRQUOTA]);
880 ip->i_dquot[USRQUOTA] = NODQUOT;
881 }
882 if (ogid == gid) {
883 dqrele(vp, ip->i_dquot[GRPQUOTA]);
884 ip->i_dquot[GRPQUOTA] = NODQUOT;
885 }
886 (void) chkdq(ip, change, cred, FORCE|CHOWN);
887 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
888 (void) getinoquota(ip);
889 }
890 return (error);
891 good:
892 if (getinoquota(ip))
893 panic("ufs_chown: lost quota");
894 #endif /* QUOTA */
895 ip->i_flag |= IN_CHANGE;
896 if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
897 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0)) {
898 ip->i_mode &= ~(ISUID | ISGID);
899 DIP_SET(ip, i_mode, ip->i_mode);
900 }
901 }
902 error = UFS_UPDATE(vp, 0);
903 return (error);
904 }
905
906 static int
907 ufs_remove(ap)
908 struct vop_remove_args /* {
909 struct vnode *a_dvp;
910 struct vnode *a_vp;
911 struct componentname *a_cnp;
912 } */ *ap;
913 {
914 struct inode *ip;
915 struct vnode *vp = ap->a_vp;
916 struct vnode *dvp = ap->a_dvp;
917 int error;
918 struct thread *td;
919
920 td = curthread;
921 ip = VTOI(vp);
922 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
923 (VTOI(dvp)->i_flags & APPEND)) {
924 error = EPERM;
925 goto out;
926 }
927 #ifdef UFS_GJOURNAL
928 ufs_gjournal_orphan(vp);
929 #endif
930 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
931 if (ip->i_nlink <= 0)
932 vp->v_vflag |= VV_NOSYNC;
933 if ((ip->i_flags & SF_SNAPSHOT) != 0) {
934 /*
935 * Avoid deadlock where another thread is trying to
936 * update the inodeblock for dvp and is waiting on
937 * snaplk. Temporary unlock the vnode lock for the
938 * unlinked file and sync the directory. This should
939 * allow vput() of the directory to not block later on
940 * while holding the snapshot vnode locked, assuming
941 * that the directory hasn't been unlinked too.
942 */
943 VOP_UNLOCK(vp, 0);
944 (void) VOP_FSYNC(dvp, MNT_WAIT, td);
945 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
946 }
947 out:
948 return (error);
949 }
950
951 /*
952 * link vnode call
953 */
954 static int
955 ufs_link(ap)
956 struct vop_link_args /* {
957 struct vnode *a_tdvp;
958 struct vnode *a_vp;
959 struct componentname *a_cnp;
960 } */ *ap;
961 {
962 struct vnode *vp = ap->a_vp;
963 struct vnode *tdvp = ap->a_tdvp;
964 struct componentname *cnp = ap->a_cnp;
965 struct inode *ip;
966 struct direct newdir;
967 int error;
968
969 #ifdef INVARIANTS
970 if ((cnp->cn_flags & HASBUF) == 0)
971 panic("ufs_link: no name");
972 #endif
973 if (VTOI(tdvp)->i_effnlink < 2)
974 panic("ufs_link: Bad link count %d on parent",
975 VTOI(tdvp)->i_effnlink);
976 ip = VTOI(vp);
977 if ((nlink_t)ip->i_nlink >= LINK_MAX) {
978 error = EMLINK;
979 goto out;
980 }
981 /*
982 * The file may have been removed after namei droped the original
983 * lock.
984 */
985 if (ip->i_effnlink == 0) {
986 error = ENOENT;
987 goto out;
988 }
989 if (ip->i_flags & (IMMUTABLE | APPEND)) {
990 error = EPERM;
991 goto out;
992 }
993 ip->i_effnlink++;
994 ip->i_nlink++;
995 DIP_SET(ip, i_nlink, ip->i_nlink);
996 ip->i_flag |= IN_CHANGE;
997 if (DOINGSOFTDEP(vp))
998 softdep_setup_link(VTOI(tdvp), ip);
999 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
1000 if (!error) {
1001 ufs_makedirentry(ip, cnp, &newdir);
1002 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL, 0);
1003 }
1004
1005 if (error) {
1006 ip->i_effnlink--;
1007 ip->i_nlink--;
1008 DIP_SET(ip, i_nlink, ip->i_nlink);
1009 ip->i_flag |= IN_CHANGE;
1010 if (DOINGSOFTDEP(vp))
1011 softdep_revert_link(VTOI(tdvp), ip);
1012 }
1013 out:
1014 return (error);
1015 }
1016
1017 /*
1018 * whiteout vnode call
1019 */
1020 static int
1021 ufs_whiteout(ap)
1022 struct vop_whiteout_args /* {
1023 struct vnode *a_dvp;
1024 struct componentname *a_cnp;
1025 int a_flags;
1026 } */ *ap;
1027 {
1028 struct vnode *dvp = ap->a_dvp;
1029 struct componentname *cnp = ap->a_cnp;
1030 struct direct newdir;
1031 int error = 0;
1032
1033 switch (ap->a_flags) {
1034 case LOOKUP:
1035 /* 4.4 format directories support whiteout operations */
1036 if (dvp->v_mount->mnt_maxsymlinklen > 0)
1037 return (0);
1038 return (EOPNOTSUPP);
1039
1040 case CREATE:
1041 /* create a new directory whiteout */
1042 #ifdef INVARIANTS
1043 if ((cnp->cn_flags & SAVENAME) == 0)
1044 panic("ufs_whiteout: missing name");
1045 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
1046 panic("ufs_whiteout: old format filesystem");
1047 #endif
1048
1049 newdir.d_ino = WINO;
1050 newdir.d_namlen = cnp->cn_namelen;
1051 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
1052 newdir.d_type = DT_WHT;
1053 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL, 0);
1054 break;
1055
1056 case DELETE:
1057 /* remove an existing directory whiteout */
1058 #ifdef INVARIANTS
1059 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
1060 panic("ufs_whiteout: old format filesystem");
1061 #endif
1062
1063 cnp->cn_flags &= ~DOWHITEOUT;
1064 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
1065 break;
1066 default:
1067 panic("ufs_whiteout: unknown op");
1068 }
1069 return (error);
1070 }
1071
1072 static volatile int rename_restarts;
1073 SYSCTL_INT(_vfs_ufs, OID_AUTO, rename_restarts, CTLFLAG_RD,
1074 __DEVOLATILE(int *, &rename_restarts), 0,
1075 "Times rename had to restart due to lock contention");
1076
1077 /*
1078 * Rename system call.
1079 * rename("foo", "bar");
1080 * is essentially
1081 * unlink("bar");
1082 * link("foo", "bar");
1083 * unlink("foo");
1084 * but ``atomically''. Can't do full commit without saving state in the
1085 * inode on disk which isn't feasible at this time. Best we can do is
1086 * always guarantee the target exists.
1087 *
1088 * Basic algorithm is:
1089 *
1090 * 1) Bump link count on source while we're linking it to the
1091 * target. This also ensure the inode won't be deleted out
1092 * from underneath us while we work (it may be truncated by
1093 * a concurrent `trunc' or `open' for creation).
1094 * 2) Link source to destination. If destination already exists,
1095 * delete it first.
1096 * 3) Unlink source reference to inode if still around. If a
1097 * directory was moved and the parent of the destination
1098 * is different from the source, patch the ".." entry in the
1099 * directory.
1100 */
1101 static int
1102 ufs_rename(ap)
1103 struct vop_rename_args /* {
1104 struct vnode *a_fdvp;
1105 struct vnode *a_fvp;
1106 struct componentname *a_fcnp;
1107 struct vnode *a_tdvp;
1108 struct vnode *a_tvp;
1109 struct componentname *a_tcnp;
1110 } */ *ap;
1111 {
1112 struct vnode *tvp = ap->a_tvp;
1113 struct vnode *tdvp = ap->a_tdvp;
1114 struct vnode *fvp = ap->a_fvp;
1115 struct vnode *fdvp = ap->a_fdvp;
1116 struct vnode *nvp;
1117 struct componentname *tcnp = ap->a_tcnp;
1118 struct componentname *fcnp = ap->a_fcnp;
1119 struct thread *td = fcnp->cn_thread;
1120 struct inode *fip, *tip, *tdp, *fdp;
1121 struct direct newdir;
1122 off_t endoff;
1123 int doingdirectory, newparent;
1124 int error = 0;
1125 struct mount *mp;
1126 ino_t ino;
1127
1128 #ifdef INVARIANTS
1129 if ((tcnp->cn_flags & HASBUF) == 0 ||
1130 (fcnp->cn_flags & HASBUF) == 0)
1131 panic("ufs_rename: no name");
1132 #endif
1133 endoff = 0;
1134 mp = tdvp->v_mount;
1135 VOP_UNLOCK(tdvp, 0);
1136 if (tvp && tvp != tdvp)
1137 VOP_UNLOCK(tvp, 0);
1138 /*
1139 * Check for cross-device rename.
1140 */
1141 if ((fvp->v_mount != tdvp->v_mount) ||
1142 (tvp && (fvp->v_mount != tvp->v_mount))) {
1143 error = EXDEV;
1144 mp = NULL;
1145 goto releout;
1146 }
1147 relock:
1148 /*
1149 * We need to acquire 2 to 4 locks depending on whether tvp is NULL
1150 * and fdvp and tdvp are the same directory. Subsequently we need
1151 * to double-check all paths and in the directory rename case we
1152 * need to verify that we are not creating a directory loop. To
1153 * handle this we acquire all but fdvp using non-blocking
1154 * acquisitions. If we fail to acquire any lock in the path we will
1155 * drop all held locks, acquire the new lock in a blocking fashion,
1156 * and then release it and restart the rename. This acquire/release
1157 * step ensures that we do not spin on a lock waiting for release.
1158 */
1159 error = vn_lock(fdvp, LK_EXCLUSIVE);
1160 if (error)
1161 goto releout;
1162 if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
1163 VOP_UNLOCK(fdvp, 0);
1164 error = vn_lock(tdvp, LK_EXCLUSIVE);
1165 if (error)
1166 goto releout;
1167 VOP_UNLOCK(tdvp, 0);
1168 atomic_add_int(&rename_restarts, 1);
1169 goto relock;
1170 }
1171 /*
1172 * Re-resolve fvp to be certain it still exists and fetch the
1173 * correct vnode.
1174 */
1175 error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
1176 if (error) {
1177 VOP_UNLOCK(fdvp, 0);
1178 VOP_UNLOCK(tdvp, 0);
1179 goto releout;
1180 }
1181 error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
1182 if (error) {
1183 VOP_UNLOCK(fdvp, 0);
1184 VOP_UNLOCK(tdvp, 0);
1185 if (error != EBUSY)
1186 goto releout;
1187 error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
1188 if (error != 0)
1189 goto releout;
1190 VOP_UNLOCK(nvp, 0);
1191 vrele(fvp);
1192 fvp = nvp;
1193 atomic_add_int(&rename_restarts, 1);
1194 goto relock;
1195 }
1196 vrele(fvp);
1197 fvp = nvp;
1198 /*
1199 * Re-resolve tvp and acquire the vnode lock if present.
1200 */
1201 error = ufs_lookup_ino(tdvp, NULL, tcnp, &ino);
1202 if (error != 0 && error != EJUSTRETURN) {
1203 VOP_UNLOCK(fdvp, 0);
1204 VOP_UNLOCK(tdvp, 0);
1205 VOP_UNLOCK(fvp, 0);
1206 goto releout;
1207 }
1208 /*
1209 * If tvp disappeared we just carry on.
1210 */
1211 if (error == EJUSTRETURN && tvp != NULL) {
1212 vrele(tvp);
1213 tvp = NULL;
1214 }
1215 /*
1216 * Get the tvp ino if the lookup succeeded. We may have to restart
1217 * if the non-blocking acquire fails.
1218 */
1219 if (error == 0) {
1220 nvp = NULL;
1221 error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
1222 if (tvp)
1223 vrele(tvp);
1224 tvp = nvp;
1225 if (error) {
1226 VOP_UNLOCK(fdvp, 0);
1227 VOP_UNLOCK(tdvp, 0);
1228 VOP_UNLOCK(fvp, 0);
1229 if (error != EBUSY)
1230 goto releout;
1231 error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
1232 if (error != 0)
1233 goto releout;
1234 vput(nvp);
1235 atomic_add_int(&rename_restarts, 1);
1236 goto relock;
1237 }
1238 }
1239 fdp = VTOI(fdvp);
1240 fip = VTOI(fvp);
1241 tdp = VTOI(tdvp);
1242 tip = NULL;
1243 if (tvp)
1244 tip = VTOI(tvp);
1245 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
1246 (VTOI(tdvp)->i_flags & APPEND))) {
1247 error = EPERM;
1248 goto unlockout;
1249 }
1250 /*
1251 * Renaming a file to itself has no effect. The upper layers should
1252 * not call us in that case. However, things could change after
1253 * we drop the locks above.
1254 */
1255 if (fvp == tvp) {
1256 error = 0;
1257 goto unlockout;
1258 }
1259 doingdirectory = 0;
1260 newparent = 0;
1261 ino = fip->i_number;
1262 if (fip->i_nlink >= LINK_MAX) {
1263 error = EMLINK;
1264 goto unlockout;
1265 }
1266 if ((fip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
1267 || (fdp->i_flags & APPEND)) {
1268 error = EPERM;
1269 goto unlockout;
1270 }
1271 if ((fip->i_mode & IFMT) == IFDIR) {
1272 /*
1273 * Avoid ".", "..", and aliases of "." for obvious reasons.
1274 */
1275 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
1276 fdp == fip ||
1277 (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) {
1278 error = EINVAL;
1279 goto unlockout;
1280 }
1281 if (fdp->i_number != tdp->i_number)
1282 newparent = tdp->i_number;
1283 doingdirectory = 1;
1284 }
1285 if ((fvp->v_type == VDIR && fvp->v_mountedhere != NULL) ||
1286 (tvp != NULL && tvp->v_type == VDIR &&
1287 tvp->v_mountedhere != NULL)) {
1288 error = EXDEV;
1289 goto unlockout;
1290 }
1291
1292 /*
1293 * If ".." must be changed (ie the directory gets a new
1294 * parent) then the source directory must not be in the
1295 * directory hierarchy above the target, as this would
1296 * orphan everything below the source directory. Also
1297 * the user must have write permission in the source so
1298 * as to be able to change "..".
1299 */
1300 if (doingdirectory && newparent) {
1301 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
1302 if (error)
1303 goto unlockout;
1304 error = ufs_checkpath(ino, fdp->i_number, tdp, tcnp->cn_cred,
1305 &ino);
1306 /*
1307 * We encountered a lock that we have to wait for. Unlock
1308 * everything else and VGET before restarting.
1309 */
1310 if (ino) {
1311 VOP_UNLOCK(fdvp, 0);
1312 VOP_UNLOCK(fvp, 0);
1313 VOP_UNLOCK(tdvp, 0);
1314 if (tvp)
1315 VOP_UNLOCK(tvp, 0);
1316 error = VFS_VGET(mp, ino, LK_SHARED, &nvp);
1317 if (error == 0)
1318 vput(nvp);
1319 atomic_add_int(&rename_restarts, 1);
1320 goto relock;
1321 }
1322 if (error)
1323 goto unlockout;
1324 if ((tcnp->cn_flags & SAVESTART) == 0)
1325 panic("ufs_rename: lost to startdir");
1326 }
1327 if (fip->i_effnlink == 0 || fdp->i_effnlink == 0 ||
1328 tdp->i_effnlink == 0)
1329 panic("Bad effnlink fip %p, fdp %p, tdp %p", fip, fdp, tdp);
1330
1331 /*
1332 * 1) Bump link count while we're moving stuff
1333 * around. If we crash somewhere before
1334 * completing our work, the link count
1335 * may be wrong, but correctable.
1336 */
1337 fip->i_effnlink++;
1338 fip->i_nlink++;
1339 DIP_SET(fip, i_nlink, fip->i_nlink);
1340 fip->i_flag |= IN_CHANGE;
1341 if (DOINGSOFTDEP(fvp))
1342 softdep_setup_link(tdp, fip);
1343 error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) | DOINGASYNC(fvp)));
1344 if (error)
1345 goto bad;
1346
1347 /*
1348 * 2) If target doesn't exist, link the target
1349 * to the source and unlink the source.
1350 * Otherwise, rewrite the target directory
1351 * entry to reference the source inode and
1352 * expunge the original entry's existence.
1353 */
1354 if (tip == NULL) {
1355 if (tdp->i_dev != fip->i_dev)
1356 panic("ufs_rename: EXDEV");
1357 if (doingdirectory && newparent) {
1358 /*
1359 * Account for ".." in new directory.
1360 * When source and destination have the same
1361 * parent we don't adjust the link count. The
1362 * actual link modification is completed when
1363 * .. is rewritten below.
1364 */
1365 if ((nlink_t)tdp->i_nlink >= LINK_MAX) {
1366 error = EMLINK;
1367 goto bad;
1368 }
1369 }
1370 ufs_makedirentry(fip, tcnp, &newdir);
1371 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL, 1);
1372 if (error)
1373 goto bad;
1374 /* Setup tdvp for directory compaction if needed. */
1375 if (tdp->i_count && tdp->i_endoff &&
1376 tdp->i_endoff < tdp->i_size)
1377 endoff = tdp->i_endoff;
1378 } else {
1379 if (tip->i_dev != tdp->i_dev || tip->i_dev != fip->i_dev)
1380 panic("ufs_rename: EXDEV");
1381 /*
1382 * Short circuit rename(foo, foo).
1383 */
1384 if (tip->i_number == fip->i_number)
1385 panic("ufs_rename: same file");
1386 /*
1387 * If the parent directory is "sticky", then the caller
1388 * must possess VADMIN for the parent directory, or the
1389 * destination of the rename. This implements append-only
1390 * directories.
1391 */
1392 if ((tdp->i_mode & S_ISTXT) &&
1393 VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) &&
1394 VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) {
1395 error = EPERM;
1396 goto bad;
1397 }
1398 /*
1399 * Target must be empty if a directory and have no links
1400 * to it. Also, ensure source and target are compatible
1401 * (both directories, or both not directories).
1402 */
1403 if ((tip->i_mode & IFMT) == IFDIR) {
1404 if ((tip->i_effnlink > 2) ||
1405 !ufs_dirempty(tip, tdp->i_number, tcnp->cn_cred)) {
1406 error = ENOTEMPTY;
1407 goto bad;
1408 }
1409 if (!doingdirectory) {
1410 error = ENOTDIR;
1411 goto bad;
1412 }
1413 cache_purge(tdvp);
1414 } else if (doingdirectory) {
1415 error = EISDIR;
1416 goto bad;
1417 }
1418 if (doingdirectory) {
1419 if (!newparent) {
1420 tdp->i_effnlink--;
1421 if (DOINGSOFTDEP(tdvp))
1422 softdep_change_linkcnt(tdp);
1423 }
1424 tip->i_effnlink--;
1425 if (DOINGSOFTDEP(tvp))
1426 softdep_change_linkcnt(tip);
1427 }
1428 error = ufs_dirrewrite(tdp, tip, fip->i_number,
1429 IFTODT(fip->i_mode),
1430 (doingdirectory && newparent) ? newparent : doingdirectory);
1431 if (error) {
1432 if (doingdirectory) {
1433 if (!newparent) {
1434 tdp->i_effnlink++;
1435 if (DOINGSOFTDEP(tdvp))
1436 softdep_change_linkcnt(tdp);
1437 }
1438 tip->i_effnlink++;
1439 if (DOINGSOFTDEP(tvp))
1440 softdep_change_linkcnt(tip);
1441 }
1442 }
1443 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1444 /*
1445 * The only stuff left in the directory is "."
1446 * and "..". The "." reference is inconsequential
1447 * since we are quashing it. We have removed the "."
1448 * reference and the reference in the parent directory,
1449 * but there may be other hard links. The soft
1450 * dependency code will arrange to do these operations
1451 * after the parent directory entry has been deleted on
1452 * disk, so when running with that code we avoid doing
1453 * them now.
1454 */
1455 if (!newparent) {
1456 tdp->i_nlink--;
1457 DIP_SET(tdp, i_nlink, tdp->i_nlink);
1458 tdp->i_flag |= IN_CHANGE;
1459 }
1460 tip->i_nlink--;
1461 DIP_SET(tip, i_nlink, tip->i_nlink);
1462 tip->i_flag |= IN_CHANGE;
1463 }
1464 }
1465
1466 /*
1467 * 3) Unlink the source. We have to resolve the path again to
1468 * fixup the directory offset and count for ufs_dirremove.
1469 */
1470 if (fdvp == tdvp) {
1471 error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
1472 if (error)
1473 panic("ufs_rename: from entry went away!");
1474 if (ino != fip->i_number)
1475 panic("ufs_rename: ino mismatch %ju != %ju\n",
1476 (uintmax_t)ino, (uintmax_t)fip->i_number);
1477 }
1478 /*
1479 * If the source is a directory with a
1480 * new parent, the link count of the old
1481 * parent directory must be decremented
1482 * and ".." set to point to the new parent.
1483 */
1484 if (doingdirectory && newparent) {
1485 /*
1486 * If tip exists we simply use its link, otherwise we must
1487 * add a new one.
1488 */
1489 if (tip == NULL) {
1490 tdp->i_effnlink++;
1491 tdp->i_nlink++;
1492 DIP_SET(tdp, i_nlink, tdp->i_nlink);
1493 tdp->i_flag |= IN_CHANGE;
1494 if (DOINGSOFTDEP(tdvp))
1495 softdep_setup_dotdot_link(tdp, fip);
1496 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
1497 DOINGASYNC(tdvp)));
1498 /* Don't go to bad here as the new link exists. */
1499 if (error)
1500 goto unlockout;
1501 } else if (DOINGSUJ(tdvp))
1502 /* Journal must account for each new link. */
1503 softdep_setup_dotdot_link(tdp, fip);
1504 fip->i_offset = mastertemplate.dot_reclen;
1505 ufs_dirrewrite(fip, fdp, newparent, DT_DIR, 0);
1506 cache_purge(fdvp);
1507 }
1508 error = ufs_dirremove(fdvp, fip, fcnp->cn_flags, 0);
1509 /*
1510 * The kern_renameat() looks up the fvp using the DELETE flag, which
1511 * causes the removal of the name cache entry for fvp.
1512 * As the relookup of the fvp is done in two steps:
1513 * ufs_lookup_ino() and then VFS_VGET(), another thread might do a
1514 * normal lookup of the from name just before the VFS_VGET() call,
1515 * causing the cache entry to be re-instantiated.
1516 *
1517 * The same issue also applies to tvp if it exists as
1518 * otherwise we may have a stale name cache entry for the new
1519 * name that references the old i-node if it has other links
1520 * or open file descriptors.
1521 */
1522 cache_purge(fvp);
1523 if (tvp)
1524 cache_purge(tvp);
1525 cache_purge_negative(tdvp);
1526
1527 unlockout:
1528 vput(fdvp);
1529 vput(fvp);
1530 if (tvp)
1531 vput(tvp);
1532 /*
1533 * If compaction or fsync was requested do it now that other locks
1534 * are no longer needed.
1535 */
1536 if (error == 0 && endoff != 0) {
1537 #ifdef UFS_DIRHASH
1538 if (tdp->i_dirhash != NULL)
1539 ufsdirhash_dirtrunc(tdp, endoff);
1540 #endif
1541 UFS_TRUNCATE(tdvp, endoff, IO_NORMAL | IO_SYNC, tcnp->cn_cred);
1542 }
1543 if (error == 0 && tdp->i_flag & IN_NEEDSYNC)
1544 error = VOP_FSYNC(tdvp, MNT_WAIT, td);
1545 vput(tdvp);
1546 return (error);
1547
1548 bad:
1549 fip->i_effnlink--;
1550 fip->i_nlink--;
1551 DIP_SET(fip, i_nlink, fip->i_nlink);
1552 fip->i_flag |= IN_CHANGE;
1553 if (DOINGSOFTDEP(fvp))
1554 softdep_revert_link(tdp, fip);
1555 goto unlockout;
1556
1557 releout:
1558 vrele(fdvp);
1559 vrele(fvp);
1560 vrele(tdvp);
1561 if (tvp)
1562 vrele(tvp);
1563
1564 return (error);
1565 }
1566
1567 #ifdef UFS_ACL
1568 static int
1569 ufs_do_posix1e_acl_inheritance_dir(struct vnode *dvp, struct vnode *tvp,
1570 mode_t dmode, struct ucred *cred, struct thread *td)
1571 {
1572 int error;
1573 struct inode *ip = VTOI(tvp);
1574 struct acl *dacl, *acl;
1575
1576 acl = acl_alloc(M_WAITOK);
1577 dacl = acl_alloc(M_WAITOK);
1578
1579 /*
1580 * Retrieve default ACL from parent, if any.
1581 */
1582 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
1583 switch (error) {
1584 case 0:
1585 /*
1586 * Retrieved a default ACL, so merge mode and ACL if
1587 * necessary. If the ACL is empty, fall through to
1588 * the "not defined or available" case.
1589 */
1590 if (acl->acl_cnt != 0) {
1591 dmode = acl_posix1e_newfilemode(dmode, acl);
1592 ip->i_mode = dmode;
1593 DIP_SET(ip, i_mode, dmode);
1594 *dacl = *acl;
1595 ufs_sync_acl_from_inode(ip, acl);
1596 break;
1597 }
1598 /* FALLTHROUGH */
1599
1600 case EOPNOTSUPP:
1601 /*
1602 * Just use the mode as-is.
1603 */
1604 ip->i_mode = dmode;
1605 DIP_SET(ip, i_mode, dmode);
1606 error = 0;
1607 goto out;
1608
1609 default:
1610 goto out;
1611 }
1612
1613 /*
1614 * XXX: If we abort now, will Soft Updates notify the extattr
1615 * code that the EAs for the file need to be released?
1616 */
1617 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
1618 if (error == 0)
1619 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, cred, td);
1620 switch (error) {
1621 case 0:
1622 break;
1623
1624 case EOPNOTSUPP:
1625 /*
1626 * XXX: This should not happen, as EOPNOTSUPP above
1627 * was supposed to free acl.
1628 */
1629 printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
1630 /*
1631 panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()");
1632 */
1633 break;
1634
1635 default:
1636 goto out;
1637 }
1638
1639 out:
1640 acl_free(acl);
1641 acl_free(dacl);
1642
1643 return (error);
1644 }
1645
1646 static int
1647 ufs_do_posix1e_acl_inheritance_file(struct vnode *dvp, struct vnode *tvp,
1648 mode_t mode, struct ucred *cred, struct thread *td)
1649 {
1650 int error;
1651 struct inode *ip = VTOI(tvp);
1652 struct acl *acl;
1653
1654 acl = acl_alloc(M_WAITOK);
1655
1656 /*
1657 * Retrieve default ACL for parent, if any.
1658 */
1659 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
1660 switch (error) {
1661 case 0:
1662 /*
1663 * Retrieved a default ACL, so merge mode and ACL if
1664 * necessary.
1665 */
1666 if (acl->acl_cnt != 0) {
1667 /*
1668 * Two possible ways for default ACL to not
1669 * be present. First, the EA can be
1670 * undefined, or second, the default ACL can
1671 * be blank. If it's blank, fall through to
1672 * the it's not defined case.
1673 */
1674 mode = acl_posix1e_newfilemode(mode, acl);
1675 ip->i_mode = mode;
1676 DIP_SET(ip, i_mode, mode);
1677 ufs_sync_acl_from_inode(ip, acl);
1678 break;
1679 }
1680 /* FALLTHROUGH */
1681
1682 case EOPNOTSUPP:
1683 /*
1684 * Just use the mode as-is.
1685 */
1686 ip->i_mode = mode;
1687 DIP_SET(ip, i_mode, mode);
1688 error = 0;
1689 goto out;
1690
1691 default:
1692 goto out;
1693 }
1694
1695 /*
1696 * XXX: If we abort now, will Soft Updates notify the extattr
1697 * code that the EAs for the file need to be released?
1698 */
1699 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
1700 switch (error) {
1701 case 0:
1702 break;
1703
1704 case EOPNOTSUPP:
1705 /*
1706 * XXX: This should not happen, as EOPNOTSUPP above was
1707 * supposed to free acl.
1708 */
1709 printf("ufs_makeinode: VOP_GETACL() but no "
1710 "VOP_SETACL()\n");
1711 /* panic("ufs_makeinode: VOP_GETACL() but no "
1712 "VOP_SETACL()"); */
1713 break;
1714
1715 default:
1716 goto out;
1717 }
1718
1719 out:
1720 acl_free(acl);
1721
1722 return (error);
1723 }
1724
1725 static int
1726 ufs_do_nfs4_acl_inheritance(struct vnode *dvp, struct vnode *tvp,
1727 mode_t child_mode, struct ucred *cred, struct thread *td)
1728 {
1729 int error;
1730 struct acl *parent_aclp, *child_aclp;
1731
1732 parent_aclp = acl_alloc(M_WAITOK);
1733 child_aclp = acl_alloc(M_WAITOK | M_ZERO);
1734
1735 error = ufs_getacl_nfs4_internal(dvp, parent_aclp, td);
1736 if (error)
1737 goto out;
1738 acl_nfs4_compute_inherited_acl(parent_aclp, child_aclp,
1739 child_mode, VTOI(tvp)->i_uid, tvp->v_type == VDIR);
1740 error = ufs_setacl_nfs4_internal(tvp, child_aclp, td);
1741 if (error)
1742 goto out;
1743 out:
1744 acl_free(parent_aclp);
1745 acl_free(child_aclp);
1746
1747 return (error);
1748 }
1749 #endif
1750
1751 /*
1752 * Mkdir system call
1753 */
1754 static int
1755 ufs_mkdir(ap)
1756 struct vop_mkdir_args /* {
1757 struct vnode *a_dvp;
1758 struct vnode **a_vpp;
1759 struct componentname *a_cnp;
1760 struct vattr *a_vap;
1761 } */ *ap;
1762 {
1763 struct vnode *dvp = ap->a_dvp;
1764 struct vattr *vap = ap->a_vap;
1765 struct componentname *cnp = ap->a_cnp;
1766 struct inode *ip, *dp;
1767 struct vnode *tvp;
1768 struct buf *bp;
1769 struct dirtemplate dirtemplate, *dtp;
1770 struct direct newdir;
1771 int error, dmode;
1772 long blkoff;
1773
1774 #ifdef INVARIANTS
1775 if ((cnp->cn_flags & HASBUF) == 0)
1776 panic("ufs_mkdir: no name");
1777 #endif
1778 dp = VTOI(dvp);
1779 if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1780 error = EMLINK;
1781 goto out;
1782 }
1783 dmode = vap->va_mode & 0777;
1784 dmode |= IFDIR;
1785 /*
1786 * Must simulate part of ufs_makeinode here to acquire the inode,
1787 * but not have it entered in the parent directory. The entry is
1788 * made later after writing "." and ".." entries.
1789 */
1790 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1791 if (error)
1792 goto out;
1793 ip = VTOI(tvp);
1794 ip->i_gid = dp->i_gid;
1795 DIP_SET(ip, i_gid, dp->i_gid);
1796 #ifdef SUIDDIR
1797 {
1798 #ifdef QUOTA
1799 struct ucred ucred, *ucp;
1800 gid_t ucred_group;
1801 ucp = cnp->cn_cred;
1802 #endif
1803 /*
1804 * If we are hacking owners here, (only do this where told to)
1805 * and we are not giving it TO root, (would subvert quotas)
1806 * then go ahead and give it to the other user.
1807 * The new directory also inherits the SUID bit.
1808 * If user's UID and dir UID are the same,
1809 * 'give it away' so that the SUID is still forced on.
1810 */
1811 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1812 (dp->i_mode & ISUID) && dp->i_uid) {
1813 dmode |= ISUID;
1814 ip->i_uid = dp->i_uid;
1815 DIP_SET(ip, i_uid, dp->i_uid);
1816 #ifdef QUOTA
1817 if (dp->i_uid != cnp->cn_cred->cr_uid) {
1818 /*
1819 * Make sure the correct user gets charged
1820 * for the space.
1821 * Make a dummy credential for the victim.
1822 * XXX This seems to never be accessed out of
1823 * our context so a stack variable is ok.
1824 */
1825 refcount_init(&ucred.cr_ref, 1);
1826 ucred.cr_uid = ip->i_uid;
1827 ucred.cr_ngroups = 1;
1828 ucred.cr_groups = &ucred_group;
1829 ucred.cr_groups[0] = dp->i_gid;
1830 ucp = &ucred;
1831 }
1832 #endif
1833 } else {
1834 ip->i_uid = cnp->cn_cred->cr_uid;
1835 DIP_SET(ip, i_uid, ip->i_uid);
1836 }
1837 #ifdef QUOTA
1838 if ((error = getinoquota(ip)) ||
1839 (error = chkiq(ip, 1, ucp, 0))) {
1840 if (DOINGSOFTDEP(tvp))
1841 softdep_revert_link(dp, ip);
1842 UFS_VFREE(tvp, ip->i_number, dmode);
1843 vput(tvp);
1844 return (error);
1845 }
1846 #endif
1847 }
1848 #else /* !SUIDDIR */
1849 ip->i_uid = cnp->cn_cred->cr_uid;
1850 DIP_SET(ip, i_uid, ip->i_uid);
1851 #ifdef QUOTA
1852 if ((error = getinoquota(ip)) ||
1853 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1854 if (DOINGSOFTDEP(tvp))
1855 softdep_revert_link(dp, ip);
1856 UFS_VFREE(tvp, ip->i_number, dmode);
1857 vput(tvp);
1858 return (error);
1859 }
1860 #endif
1861 #endif /* !SUIDDIR */
1862 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1863 ip->i_mode = dmode;
1864 DIP_SET(ip, i_mode, dmode);
1865 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
1866 ip->i_effnlink = 2;
1867 ip->i_nlink = 2;
1868 DIP_SET(ip, i_nlink, 2);
1869
1870 if (cnp->cn_flags & ISWHITEOUT) {
1871 ip->i_flags |= UF_OPAQUE;
1872 DIP_SET(ip, i_flags, ip->i_flags);
1873 }
1874
1875 /*
1876 * Bump link count in parent directory to reflect work done below.
1877 * Should be done before reference is created so cleanup is
1878 * possible if we crash.
1879 */
1880 dp->i_effnlink++;
1881 dp->i_nlink++;
1882 DIP_SET(dp, i_nlink, dp->i_nlink);
1883 dp->i_flag |= IN_CHANGE;
1884 if (DOINGSOFTDEP(dvp))
1885 softdep_setup_mkdir(dp, ip);
1886 error = UFS_UPDATE(dvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
1887 if (error)
1888 goto bad;
1889 #ifdef MAC
1890 if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
1891 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
1892 dvp, tvp, cnp);
1893 if (error)
1894 goto bad;
1895 }
1896 #endif
1897 #ifdef UFS_ACL
1898 if (dvp->v_mount->mnt_flag & MNT_ACLS) {
1899 error = ufs_do_posix1e_acl_inheritance_dir(dvp, tvp, dmode,
1900 cnp->cn_cred, cnp->cn_thread);
1901 if (error)
1902 goto bad;
1903 } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
1904 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, dmode,
1905 cnp->cn_cred, cnp->cn_thread);
1906 if (error)
1907 goto bad;
1908 }
1909 #endif /* !UFS_ACL */
1910
1911 /*
1912 * Initialize directory with "." and ".." from static template.
1913 */
1914 if (dvp->v_mount->mnt_maxsymlinklen > 0)
1915 dtp = &mastertemplate;
1916 else
1917 dtp = (struct dirtemplate *)&omastertemplate;
1918 dirtemplate = *dtp;
1919 dirtemplate.dot_ino = ip->i_number;
1920 dirtemplate.dotdot_ino = dp->i_number;
1921 if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1922 BA_CLRBUF, &bp)) != 0)
1923 goto bad;
1924 ip->i_size = DIRBLKSIZ;
1925 DIP_SET(ip, i_size, DIRBLKSIZ);
1926 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1927 vnode_pager_setsize(tvp, (u_long)ip->i_size);
1928 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1929 if (DOINGSOFTDEP(tvp)) {
1930 /*
1931 * Ensure that the entire newly allocated block is a
1932 * valid directory so that future growth within the
1933 * block does not have to ensure that the block is
1934 * written before the inode.
1935 */
1936 blkoff = DIRBLKSIZ;
1937 while (blkoff < bp->b_bcount) {
1938 ((struct direct *)
1939 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1940 blkoff += DIRBLKSIZ;
1941 }
1942 }
1943 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
1944 DOINGASYNC(tvp)))) != 0) {
1945 (void)bwrite(bp);
1946 goto bad;
1947 }
1948 /*
1949 * Directory set up, now install its entry in the parent directory.
1950 *
1951 * If we are not doing soft dependencies, then we must write out the
1952 * buffer containing the new directory body before entering the new
1953 * name in the parent. If we are doing soft dependencies, then the
1954 * buffer containing the new directory body will be passed to and
1955 * released in the soft dependency code after the code has attached
1956 * an appropriate ordering dependency to the buffer which ensures that
1957 * the buffer is written before the new name is written in the parent.
1958 */
1959 if (DOINGASYNC(dvp))
1960 bdwrite(bp);
1961 else if (!DOINGSOFTDEP(dvp) && ((error = bwrite(bp))))
1962 goto bad;
1963 ufs_makedirentry(ip, cnp, &newdir);
1964 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp, 0);
1965
1966 bad:
1967 if (error == 0) {
1968 *ap->a_vpp = tvp;
1969 } else {
1970 dp->i_effnlink--;
1971 dp->i_nlink--;
1972 DIP_SET(dp, i_nlink, dp->i_nlink);
1973 dp->i_flag |= IN_CHANGE;
1974 /*
1975 * No need to do an explicit VOP_TRUNCATE here, vrele will
1976 * do this for us because we set the link count to 0.
1977 */
1978 ip->i_effnlink = 0;
1979 ip->i_nlink = 0;
1980 DIP_SET(ip, i_nlink, 0);
1981 ip->i_flag |= IN_CHANGE;
1982 if (DOINGSOFTDEP(tvp))
1983 softdep_revert_mkdir(dp, ip);
1984
1985 vput(tvp);
1986 }
1987 out:
1988 return (error);
1989 }
1990
1991 /*
1992 * Rmdir system call.
1993 */
1994 static int
1995 ufs_rmdir(ap)
1996 struct vop_rmdir_args /* {
1997 struct vnode *a_dvp;
1998 struct vnode *a_vp;
1999 struct componentname *a_cnp;
2000 } */ *ap;
2001 {
2002 struct vnode *vp = ap->a_vp;
2003 struct vnode *dvp = ap->a_dvp;
2004 struct componentname *cnp = ap->a_cnp;
2005 struct inode *ip, *dp;
2006 int error;
2007
2008 ip = VTOI(vp);
2009 dp = VTOI(dvp);
2010
2011 /*
2012 * Do not remove a directory that is in the process of being renamed.
2013 * Verify the directory is empty (and valid). Rmdir ".." will not be
2014 * valid since ".." will contain a reference to the current directory
2015 * and thus be non-empty. Do not allow the removal of mounted on
2016 * directories (this can happen when an NFS exported filesystem
2017 * tries to remove a locally mounted on directory).
2018 */
2019 error = 0;
2020 if (ip->i_effnlink < 2) {
2021 error = EINVAL;
2022 goto out;
2023 }
2024 if (dp->i_effnlink < 3)
2025 panic("ufs_dirrem: Bad link count %d on parent",
2026 dp->i_effnlink);
2027 if (!ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
2028 error = ENOTEMPTY;
2029 goto out;
2030 }
2031 if ((dp->i_flags & APPEND)
2032 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
2033 error = EPERM;
2034 goto out;
2035 }
2036 if (vp->v_mountedhere != 0) {
2037 error = EINVAL;
2038 goto out;
2039 }
2040 #ifdef UFS_GJOURNAL
2041 ufs_gjournal_orphan(vp);
2042 #endif
2043 /*
2044 * Delete reference to directory before purging
2045 * inode. If we crash in between, the directory
2046 * will be reattached to lost+found,
2047 */
2048 dp->i_effnlink--;
2049 ip->i_effnlink--;
2050 if (DOINGSOFTDEP(vp))
2051 softdep_setup_rmdir(dp, ip);
2052 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
2053 if (error) {
2054 dp->i_effnlink++;
2055 ip->i_effnlink++;
2056 if (DOINGSOFTDEP(vp))
2057 softdep_revert_rmdir(dp, ip);
2058 goto out;
2059 }
2060 cache_purge(dvp);
2061 /*
2062 * The only stuff left in the directory is "." and "..". The "."
2063 * reference is inconsequential since we are quashing it. The soft
2064 * dependency code will arrange to do these operations after
2065 * the parent directory entry has been deleted on disk, so
2066 * when running with that code we avoid doing them now.
2067 */
2068 if (!DOINGSOFTDEP(vp)) {
2069 dp->i_nlink--;
2070 DIP_SET(dp, i_nlink, dp->i_nlink);
2071 dp->i_flag |= IN_CHANGE;
2072 error = UFS_UPDATE(dvp, 0);
2073 ip->i_nlink--;
2074 DIP_SET(ip, i_nlink, ip->i_nlink);
2075 ip->i_flag |= IN_CHANGE;
2076 }
2077 cache_purge(vp);
2078 #ifdef UFS_DIRHASH
2079 /* Kill any active hash; i_effnlink == 0, so it will not come back. */
2080 if (ip->i_dirhash != NULL)
2081 ufsdirhash_free(ip);
2082 #endif
2083 out:
2084 return (error);
2085 }
2086
2087 /*
2088 * symlink -- make a symbolic link
2089 */
2090 static int
2091 ufs_symlink(ap)
2092 struct vop_symlink_args /* {
2093 struct vnode *a_dvp;
2094 struct vnode **a_vpp;
2095 struct componentname *a_cnp;
2096 struct vattr *a_vap;
2097 char *a_target;
2098 } */ *ap;
2099 {
2100 struct vnode *vp, **vpp = ap->a_vpp;
2101 struct inode *ip;
2102 int len, error;
2103
2104 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
2105 vpp, ap->a_cnp);
2106 if (error)
2107 return (error);
2108 vp = *vpp;
2109 len = strlen(ap->a_target);
2110 if (len < vp->v_mount->mnt_maxsymlinklen) {
2111 ip = VTOI(vp);
2112 bcopy(ap->a_target, SHORTLINK(ip), len);
2113 ip->i_size = len;
2114 DIP_SET(ip, i_size, len);
2115 ip->i_flag |= IN_CHANGE | IN_UPDATE;
2116 error = UFS_UPDATE(vp, 0);
2117 } else
2118 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
2119 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
2120 ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
2121 if (error)
2122 vput(vp);
2123 return (error);
2124 }
2125
2126 /*
2127 * Vnode op for reading directories.
2128 */
2129 int
2130 ufs_readdir(ap)
2131 struct vop_readdir_args /* {
2132 struct vnode *a_vp;
2133 struct uio *a_uio;
2134 struct ucred *a_cred;
2135 int *a_eofflag;
2136 int *a_ncookies;
2137 u_long **a_cookies;
2138 } */ *ap;
2139 {
2140 struct vnode *vp = ap->a_vp;
2141 struct uio *uio = ap->a_uio;
2142 struct buf *bp;
2143 struct inode *ip;
2144 struct direct *dp, *edp;
2145 u_long *cookies;
2146 struct dirent dstdp;
2147 off_t offset, startoffset;
2148 size_t readcnt, skipcnt;
2149 ssize_t startresid;
2150 int ncookies;
2151 int error;
2152
2153 if (uio->uio_offset < 0)
2154 return (EINVAL);
2155 ip = VTOI(vp);
2156 if (ip->i_effnlink == 0)
2157 return (0);
2158 if (ap->a_ncookies != NULL) {
2159 ncookies = uio->uio_resid;
2160 if (uio->uio_offset >= ip->i_size)
2161 ncookies = 0;
2162 else if (ip->i_size - uio->uio_offset < ncookies)
2163 ncookies = ip->i_size - uio->uio_offset;
2164 ncookies = ncookies / (offsetof(struct direct, d_name) + 4) + 1;
2165 cookies = malloc(ncookies * sizeof(*cookies), M_TEMP, M_WAITOK);
2166 *ap->a_ncookies = ncookies;
2167 *ap->a_cookies = cookies;
2168 } else {
2169 ncookies = 0;
2170 cookies = NULL;
2171 }
2172 offset = startoffset = uio->uio_offset;
2173 startresid = uio->uio_resid;
2174 error = 0;
2175 while (error == 0 && uio->uio_resid > 0 &&
2176 uio->uio_offset < ip->i_size) {
2177 error = ffs_blkatoff(vp, uio->uio_offset, NULL, &bp);
2178 if (error)
2179 break;
2180 if (bp->b_offset + bp->b_bcount > ip->i_size)
2181 readcnt = ip->i_size - bp->b_offset;
2182 else
2183 readcnt = bp->b_bcount;
2184 skipcnt = (size_t)(uio->uio_offset - bp->b_offset) &
2185 ~(size_t)(DIRBLKSIZ - 1);
2186 offset = bp->b_offset + skipcnt;
2187 dp = (struct direct *)&bp->b_data[skipcnt];
2188 edp = (struct direct *)&bp->b_data[readcnt];
2189 while (error == 0 && uio->uio_resid > 0 && dp < edp) {
2190 if (dp->d_reclen <= offsetof(struct direct, d_name) ||
2191 (caddr_t)dp + dp->d_reclen > (caddr_t)edp) {
2192 error = EIO;
2193 break;
2194 }
2195 #if BYTE_ORDER == LITTLE_ENDIAN
2196 /* Old filesystem format. */
2197 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
2198 dstdp.d_namlen = dp->d_type;
2199 dstdp.d_type = dp->d_namlen;
2200 } else
2201 #endif
2202 {
2203 dstdp.d_namlen = dp->d_namlen;
2204 dstdp.d_type = dp->d_type;
2205 }
2206 if (offsetof(struct direct, d_name) + dstdp.d_namlen >
2207 dp->d_reclen) {
2208 error = EIO;
2209 break;
2210 }
2211 if (offset < startoffset || dp->d_ino == 0)
2212 goto nextentry;
2213 dstdp.d_fileno = dp->d_ino;
2214 dstdp.d_reclen = GENERIC_DIRSIZ(&dstdp);
2215 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
2216 dstdp.d_name[dstdp.d_namlen] = '\0';
2217 if (dstdp.d_reclen > uio->uio_resid) {
2218 if (uio->uio_resid == startresid)
2219 error = EINVAL;
2220 else
2221 error = EJUSTRETURN;
2222 break;
2223 }
2224 /* Advance dp. */
2225 error = uiomove((caddr_t)&dstdp, dstdp.d_reclen, uio);
2226 if (error)
2227 break;
2228 if (cookies != NULL) {
2229 KASSERT(ncookies > 0,
2230 ("ufs_readdir: cookies buffer too small"));
2231 *cookies = offset + dp->d_reclen;
2232 cookies++;
2233 ncookies--;
2234 }
2235 nextentry:
2236 offset += dp->d_reclen;
2237 dp = (struct direct *)((caddr_t)dp + dp->d_reclen);
2238 }
2239 bqrelse(bp);
2240 uio->uio_offset = offset;
2241 }
2242 /* We need to correct uio_offset. */
2243 uio->uio_offset = offset;
2244 if (error == EJUSTRETURN)
2245 error = 0;
2246 if (ap->a_ncookies != NULL) {
2247 if (error == 0) {
2248 ap->a_ncookies -= ncookies;
2249 } else {
2250 free(*ap->a_cookies, M_TEMP);
2251 *ap->a_ncookies = 0;
2252 *ap->a_cookies = NULL;
2253 }
2254 }
2255 if (error == 0 && ap->a_eofflag)
2256 *ap->a_eofflag = ip->i_size <= uio->uio_offset;
2257 return (error);
2258 }
2259
2260 /*
2261 * Return target name of a symbolic link
2262 */
2263 static int
2264 ufs_readlink(ap)
2265 struct vop_readlink_args /* {
2266 struct vnode *a_vp;
2267 struct uio *a_uio;
2268 struct ucred *a_cred;
2269 } */ *ap;
2270 {
2271 struct vnode *vp = ap->a_vp;
2272 struct inode *ip = VTOI(vp);
2273 doff_t isize;
2274
2275 isize = ip->i_size;
2276 if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
2277 DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */
2278 return (uiomove(SHORTLINK(ip), isize, ap->a_uio));
2279 }
2280 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
2281 }
2282
2283 /*
2284 * Calculate the logical to physical mapping if not done already,
2285 * then call the device strategy routine.
2286 *
2287 * In order to be able to swap to a file, the ufs_bmaparray() operation may not
2288 * deadlock on memory. See ufs_bmap() for details.
2289 */
2290 static int
2291 ufs_strategy(ap)
2292 struct vop_strategy_args /* {
2293 struct vnode *a_vp;
2294 struct buf *a_bp;
2295 } */ *ap;
2296 {
2297 struct buf *bp = ap->a_bp;
2298 struct vnode *vp = ap->a_vp;
2299 struct bufobj *bo;
2300 struct inode *ip;
2301 ufs2_daddr_t blkno;
2302 int error;
2303
2304 ip = VTOI(vp);
2305 if (bp->b_blkno == bp->b_lblkno) {
2306 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL);
2307 bp->b_blkno = blkno;
2308 if (error) {
2309 bp->b_error = error;
2310 bp->b_ioflags |= BIO_ERROR;
2311 bufdone(bp);
2312 return (0);
2313 }
2314 if ((long)bp->b_blkno == -1)
2315 vfs_bio_clrbuf(bp);
2316 }
2317 if ((long)bp->b_blkno == -1) {
2318 bufdone(bp);
2319 return (0);
2320 }
2321 bp->b_iooffset = dbtob(bp->b_blkno);
2322 bo = ip->i_umbufobj;
2323 BO_STRATEGY(bo, bp);
2324 return (0);
2325 }
2326
2327 /*
2328 * Print out the contents of an inode.
2329 */
2330 static int
2331 ufs_print(ap)
2332 struct vop_print_args /* {
2333 struct vnode *a_vp;
2334 } */ *ap;
2335 {
2336 struct vnode *vp = ap->a_vp;
2337 struct inode *ip = VTOI(vp);
2338
2339 printf("\tino %lu, on dev %s", (u_long)ip->i_number,
2340 devtoname(ip->i_dev));
2341 if (vp->v_type == VFIFO)
2342 fifo_printinfo(vp);
2343 printf("\n");
2344 return (0);
2345 }
2346
2347 /*
2348 * Close wrapper for fifos.
2349 *
2350 * Update the times on the inode then do device close.
2351 */
2352 static int
2353 ufsfifo_close(ap)
2354 struct vop_close_args /* {
2355 struct vnode *a_vp;
2356 int a_fflag;
2357 struct ucred *a_cred;
2358 struct thread *a_td;
2359 } */ *ap;
2360 {
2361 struct vnode *vp = ap->a_vp;
2362 int usecount;
2363
2364 VI_LOCK(vp);
2365 usecount = vp->v_usecount;
2366 if (usecount > 1)
2367 ufs_itimes_locked(vp);
2368 VI_UNLOCK(vp);
2369 return (fifo_specops.vop_close(ap));
2370 }
2371
2372 /*
2373 * Kqfilter wrapper for fifos.
2374 *
2375 * Fall through to ufs kqfilter routines if needed
2376 */
2377 static int
2378 ufsfifo_kqfilter(ap)
2379 struct vop_kqfilter_args *ap;
2380 {
2381 int error;
2382
2383 error = fifo_specops.vop_kqfilter(ap);
2384 if (error)
2385 error = vfs_kqfilter(ap);
2386 return (error);
2387 }
2388
2389 /*
2390 * Return POSIX pathconf information applicable to fifos.
2391 */
2392 static int
2393 ufsfifo_pathconf(ap)
2394 struct vop_pathconf_args /* {
2395 struct vnode *a_vp;
2396 int a_name;
2397 int *a_retval;
2398 } */ *ap;
2399 {
2400
2401 switch (ap->a_name) {
2402 case _PC_ACL_EXTENDED:
2403 case _PC_ACL_NFS4:
2404 case _PC_ACL_PATH_MAX:
2405 case _PC_MAC_PRESENT:
2406 return (ufs_pathconf(ap));
2407 default:
2408 return (fifo_specops.vop_pathconf(ap));
2409 }
2410 /* NOTREACHED */
2411 }
2412
2413 /*
2414 * Return POSIX pathconf information applicable to ufs filesystems.
2415 */
2416 static int
2417 ufs_pathconf(ap)
2418 struct vop_pathconf_args /* {
2419 struct vnode *a_vp;
2420 int a_name;
2421 int *a_retval;
2422 } */ *ap;
2423 {
2424 int error;
2425
2426 error = 0;
2427 switch (ap->a_name) {
2428 case _PC_LINK_MAX:
2429 *ap->a_retval = LINK_MAX;
2430 break;
2431 case _PC_NAME_MAX:
2432 *ap->a_retval = NAME_MAX;
2433 break;
2434 case _PC_PATH_MAX:
2435 *ap->a_retval = PATH_MAX;
2436 break;
2437 case _PC_PIPE_BUF:
2438 *ap->a_retval = PIPE_BUF;
2439 break;
2440 case _PC_CHOWN_RESTRICTED:
2441 *ap->a_retval = 1;
2442 break;
2443 case _PC_NO_TRUNC:
2444 *ap->a_retval = 1;
2445 break;
2446 case _PC_ACL_EXTENDED:
2447 #ifdef UFS_ACL
2448 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
2449 *ap->a_retval = 1;
2450 else
2451 *ap->a_retval = 0;
2452 #else
2453 *ap->a_retval = 0;
2454 #endif
2455 break;
2456
2457 case _PC_ACL_NFS4:
2458 #ifdef UFS_ACL
2459 if (ap->a_vp->v_mount->mnt_flag & MNT_NFS4ACLS)
2460 *ap->a_retval = 1;
2461 else
2462 *ap->a_retval = 0;
2463 #else
2464 *ap->a_retval = 0;
2465 #endif
2466 break;
2467
2468 case _PC_ACL_PATH_MAX:
2469 #ifdef UFS_ACL
2470 if (ap->a_vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS))
2471 *ap->a_retval = ACL_MAX_ENTRIES;
2472 else
2473 *ap->a_retval = 3;
2474 #else
2475 *ap->a_retval = 3;
2476 #endif
2477 break;
2478 case _PC_MAC_PRESENT:
2479 #ifdef MAC
2480 if (ap->a_vp->v_mount->mnt_flag & MNT_MULTILABEL)
2481 *ap->a_retval = 1;
2482 else
2483 *ap->a_retval = 0;
2484 #else
2485 *ap->a_retval = 0;
2486 #endif
2487 break;
2488 case _PC_MIN_HOLE_SIZE:
2489 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
2490 break;
2491 case _PC_ASYNC_IO:
2492 /* _PC_ASYNC_IO should have been handled by upper layers. */
2493 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
2494 error = EINVAL;
2495 break;
2496 case _PC_PRIO_IO:
2497 *ap->a_retval = 0;
2498 break;
2499 case _PC_SYNC_IO:
2500 *ap->a_retval = 0;
2501 break;
2502 case _PC_ALLOC_SIZE_MIN:
2503 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
2504 break;
2505 case _PC_FILESIZEBITS:
2506 *ap->a_retval = 64;
2507 break;
2508 case _PC_REC_INCR_XFER_SIZE:
2509 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
2510 break;
2511 case _PC_REC_MAX_XFER_SIZE:
2512 *ap->a_retval = -1; /* means ``unlimited'' */
2513 break;
2514 case _PC_REC_MIN_XFER_SIZE:
2515 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
2516 break;
2517 case _PC_REC_XFER_ALIGN:
2518 *ap->a_retval = PAGE_SIZE;
2519 break;
2520 case _PC_SYMLINK_MAX:
2521 *ap->a_retval = MAXPATHLEN;
2522 break;
2523
2524 default:
2525 error = EINVAL;
2526 break;
2527 }
2528 return (error);
2529 }
2530
2531 /*
2532 * Initialize the vnode associated with a new inode, handle aliased
2533 * vnodes.
2534 */
2535 int
2536 ufs_vinit(mntp, fifoops, vpp)
2537 struct mount *mntp;
2538 struct vop_vector *fifoops;
2539 struct vnode **vpp;
2540 {
2541 struct inode *ip;
2542 struct vnode *vp;
2543
2544 vp = *vpp;
2545 ip = VTOI(vp);
2546 vp->v_type = IFTOVT(ip->i_mode);
2547 if (vp->v_type == VFIFO)
2548 vp->v_op = fifoops;
2549 ASSERT_VOP_LOCKED(vp, "ufs_vinit");
2550 if (ip->i_number == ROOTINO)
2551 vp->v_vflag |= VV_ROOT;
2552 *vpp = vp;
2553 return (0);
2554 }
2555
2556 /*
2557 * Allocate a new inode.
2558 * Vnode dvp must be locked.
2559 */
2560 static int
2561 ufs_makeinode(mode, dvp, vpp, cnp)
2562 int mode;
2563 struct vnode *dvp;
2564 struct vnode **vpp;
2565 struct componentname *cnp;
2566 {
2567 struct inode *ip, *pdir;
2568 struct direct newdir;
2569 struct vnode *tvp;
2570 int error;
2571
2572 pdir = VTOI(dvp);
2573 #ifdef INVARIANTS
2574 if ((cnp->cn_flags & HASBUF) == 0)
2575 panic("ufs_makeinode: no name");
2576 #endif
2577 *vpp = NULL;
2578 if ((mode & IFMT) == 0)
2579 mode |= IFREG;
2580
2581 if (VTOI(dvp)->i_effnlink < 2)
2582 panic("ufs_makeinode: Bad link count %d on parent",
2583 VTOI(dvp)->i_effnlink);
2584 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
2585 if (error)
2586 return (error);
2587 ip = VTOI(tvp);
2588 ip->i_gid = pdir->i_gid;
2589 DIP_SET(ip, i_gid, pdir->i_gid);
2590 #ifdef SUIDDIR
2591 {
2592 #ifdef QUOTA
2593 struct ucred ucred, *ucp;
2594 gid_t ucred_group;
2595 ucp = cnp->cn_cred;
2596 #endif
2597 /*
2598 * If we are not the owner of the directory,
2599 * and we are hacking owners here, (only do this where told to)
2600 * and we are not giving it TO root, (would subvert quotas)
2601 * then go ahead and give it to the other user.
2602 * Note that this drops off the execute bits for security.
2603 */
2604 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
2605 (pdir->i_mode & ISUID) &&
2606 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
2607 ip->i_uid = pdir->i_uid;
2608 DIP_SET(ip, i_uid, ip->i_uid);
2609 mode &= ~07111;
2610 #ifdef QUOTA
2611 /*
2612 * Make sure the correct user gets charged
2613 * for the space.
2614 * Quickly knock up a dummy credential for the victim.
2615 * XXX This seems to never be accessed out of our
2616 * context so a stack variable is ok.
2617 */
2618 refcount_init(&ucred.cr_ref, 1);
2619 ucred.cr_uid = ip->i_uid;
2620 ucred.cr_ngroups = 1;
2621 ucred.cr_groups = &ucred_group;
2622 ucred.cr_groups[0] = pdir->i_gid;
2623 ucp = &ucred;
2624 #endif
2625 } else {
2626 ip->i_uid = cnp->cn_cred->cr_uid;
2627 DIP_SET(ip, i_uid, ip->i_uid);
2628 }
2629
2630 #ifdef QUOTA
2631 if ((error = getinoquota(ip)) ||
2632 (error = chkiq(ip, 1, ucp, 0))) {
2633 if (DOINGSOFTDEP(tvp))
2634 softdep_revert_link(pdir, ip);
2635 UFS_VFREE(tvp, ip->i_number, mode);
2636 vput(tvp);
2637 return (error);
2638 }
2639 #endif
2640 }
2641 #else /* !SUIDDIR */
2642 ip->i_uid = cnp->cn_cred->cr_uid;
2643 DIP_SET(ip, i_uid, ip->i_uid);
2644 #ifdef QUOTA
2645 if ((error = getinoquota(ip)) ||
2646 (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2647 if (DOINGSOFTDEP(tvp))
2648 softdep_revert_link(pdir, ip);
2649 UFS_VFREE(tvp, ip->i_number, mode);
2650 vput(tvp);
2651 return (error);
2652 }
2653 #endif
2654 #endif /* !SUIDDIR */
2655 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2656 ip->i_mode = mode;
2657 DIP_SET(ip, i_mode, mode);
2658 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
2659 ip->i_effnlink = 1;
2660 ip->i_nlink = 1;
2661 DIP_SET(ip, i_nlink, 1);
2662 if (DOINGSOFTDEP(tvp))
2663 softdep_setup_create(VTOI(dvp), ip);
2664 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
2665 priv_check_cred(cnp->cn_cred, PRIV_VFS_SETGID, 0)) {
2666 ip->i_mode &= ~ISGID;
2667 DIP_SET(ip, i_mode, ip->i_mode);
2668 }
2669
2670 if (cnp->cn_flags & ISWHITEOUT) {
2671 ip->i_flags |= UF_OPAQUE;
2672 DIP_SET(ip, i_flags, ip->i_flags);
2673 }
2674
2675 /*
2676 * Make sure inode goes to disk before directory entry.
2677 */
2678 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
2679 if (error)
2680 goto bad;
2681 #ifdef MAC
2682 if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
2683 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
2684 dvp, tvp, cnp);
2685 if (error)
2686 goto bad;
2687 }
2688 #endif
2689 #ifdef UFS_ACL
2690 if (dvp->v_mount->mnt_flag & MNT_ACLS) {
2691 error = ufs_do_posix1e_acl_inheritance_file(dvp, tvp, mode,
2692 cnp->cn_cred, cnp->cn_thread);
2693 if (error)
2694 goto bad;
2695 } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
2696 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, mode,
2697 cnp->cn_cred, cnp->cn_thread);
2698 if (error)
2699 goto bad;
2700 }
2701 #endif /* !UFS_ACL */
2702 ufs_makedirentry(ip, cnp, &newdir);
2703 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL, 0);
2704 if (error)
2705 goto bad;
2706 *vpp = tvp;
2707 return (0);
2708
2709 bad:
2710 /*
2711 * Write error occurred trying to update the inode
2712 * or the directory so must deallocate the inode.
2713 */
2714 ip->i_effnlink = 0;
2715 ip->i_nlink = 0;
2716 DIP_SET(ip, i_nlink, 0);
2717 ip->i_flag |= IN_CHANGE;
2718 if (DOINGSOFTDEP(tvp))
2719 softdep_revert_create(VTOI(dvp), ip);
2720 vput(tvp);
2721 return (error);
2722 }
2723
2724 static int
2725 ufs_ioctl(struct vop_ioctl_args *ap)
2726 {
2727
2728 switch (ap->a_command) {
2729 case FIOSEEKDATA:
2730 case FIOSEEKHOLE:
2731 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
2732 (off_t *)ap->a_data, ap->a_cred));
2733 default:
2734 return (ENOTTY);
2735 }
2736 }
2737
2738 /* Global vfs data structures for ufs. */
2739 struct vop_vector ufs_vnodeops = {
2740 .vop_default = &default_vnodeops,
2741 .vop_fsync = VOP_PANIC,
2742 .vop_read = VOP_PANIC,
2743 .vop_reallocblks = VOP_PANIC,
2744 .vop_write = VOP_PANIC,
2745 .vop_accessx = ufs_accessx,
2746 .vop_bmap = ufs_bmap,
2747 .vop_cachedlookup = ufs_lookup,
2748 .vop_close = ufs_close,
2749 .vop_create = ufs_create,
2750 .vop_getattr = ufs_getattr,
2751 .vop_inactive = ufs_inactive,
2752 .vop_ioctl = ufs_ioctl,
2753 .vop_link = ufs_link,
2754 .vop_lookup = vfs_cache_lookup,
2755 .vop_markatime = ufs_markatime,
2756 .vop_mkdir = ufs_mkdir,
2757 .vop_mknod = ufs_mknod,
2758 .vop_open = ufs_open,
2759 .vop_pathconf = ufs_pathconf,
2760 .vop_poll = vop_stdpoll,
2761 .vop_print = ufs_print,
2762 .vop_readdir = ufs_readdir,
2763 .vop_readlink = ufs_readlink,
2764 .vop_reclaim = ufs_reclaim,
2765 .vop_remove = ufs_remove,
2766 .vop_rename = ufs_rename,
2767 .vop_rmdir = ufs_rmdir,
2768 .vop_setattr = ufs_setattr,
2769 #ifdef MAC
2770 .vop_setlabel = vop_stdsetlabel_ea,
2771 #endif
2772 .vop_strategy = ufs_strategy,
2773 .vop_symlink = ufs_symlink,
2774 .vop_whiteout = ufs_whiteout,
2775 #ifdef UFS_EXTATTR
2776 .vop_getextattr = ufs_getextattr,
2777 .vop_deleteextattr = ufs_deleteextattr,
2778 .vop_setextattr = ufs_setextattr,
2779 #endif
2780 #ifdef UFS_ACL
2781 .vop_getacl = ufs_getacl,
2782 .vop_setacl = ufs_setacl,
2783 .vop_aclcheck = ufs_aclcheck,
2784 #endif
2785 };
2786
2787 struct vop_vector ufs_fifoops = {
2788 .vop_default = &fifo_specops,
2789 .vop_fsync = VOP_PANIC,
2790 .vop_accessx = ufs_accessx,
2791 .vop_close = ufsfifo_close,
2792 .vop_getattr = ufs_getattr,
2793 .vop_inactive = ufs_inactive,
2794 .vop_kqfilter = ufsfifo_kqfilter,
2795 .vop_markatime = ufs_markatime,
2796 .vop_pathconf = ufsfifo_pathconf,
2797 .vop_print = ufs_print,
2798 .vop_read = VOP_PANIC,
2799 .vop_reclaim = ufs_reclaim,
2800 .vop_setattr = ufs_setattr,
2801 #ifdef MAC
2802 .vop_setlabel = vop_stdsetlabel_ea,
2803 #endif
2804 .vop_write = VOP_PANIC,
2805 #ifdef UFS_EXTATTR
2806 .vop_getextattr = ufs_getextattr,
2807 .vop_deleteextattr = ufs_deleteextattr,
2808 .vop_setextattr = ufs_setextattr,
2809 #endif
2810 #ifdef UFS_ACL
2811 .vop_getacl = ufs_getacl,
2812 .vop_setacl = ufs_setacl,
2813 .vop_aclcheck = ufs_aclcheck,
2814 #endif
2815 };
Cache object: 7bb419942cf238aae21f4705f433b02b
|