xref: /netbsd-src/sys/ufs/ufs/ufs_vnops.c (revision d25ffa98a4bfca1fe272f3c182496ec9934faac7)
1 /*	$NetBSD: ufs_vnops.c,v 1.189 2011/04/30 14:24:27 hannken Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Wasabi Systems, 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  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1986, 1989, 1993, 1995
34  *	The Regents of the University of California.  All rights reserved.
35  * (c) UNIX System Laboratories, Inc.
36  * All or some portions of this file are derived from material licensed
37  * to the University of California by American Telephone and Telegraph
38  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
39  * the permission of UNIX System Laboratories, Inc.
40  *
41  * Redistribution and use in source and binary forms, with or without
42  * modification, are permitted provided that the following conditions
43  * are met:
44  * 1. Redistributions of source code must retain the above copyright
45  *    notice, this list of conditions and the following disclaimer.
46  * 2. Redistributions in binary form must reproduce the above copyright
47  *    notice, this list of conditions and the following disclaimer in the
48  *    documentation and/or other materials provided with the distribution.
49  * 3. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ufs_vnops.c	8.28 (Berkeley) 7/31/95
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: ufs_vnops.c,v 1.189 2011/04/30 14:24:27 hannken Exp $");
70 
71 #if defined(_KERNEL_OPT)
72 #include "opt_ffs.h"
73 #include "opt_quota.h"
74 #endif
75 
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/namei.h>
79 #include <sys/resourcevar.h>
80 #include <sys/kernel.h>
81 #include <sys/file.h>
82 #include <sys/stat.h>
83 #include <sys/buf.h>
84 #include <sys/proc.h>
85 #include <sys/mount.h>
86 #include <sys/vnode.h>
87 #include <sys/malloc.h>
88 #include <sys/dirent.h>
89 #include <sys/lockf.h>
90 #include <sys/kauth.h>
91 #include <sys/wapbl.h>
92 #include <sys/fstrans.h>
93 
94 #include <miscfs/specfs/specdev.h>
95 #include <miscfs/fifofs/fifo.h>
96 #include <miscfs/genfs/genfs.h>
97 
98 #include <ufs/ufs/inode.h>
99 #include <ufs/ufs/dir.h>
100 #include <ufs/ufs/ufsmount.h>
101 #include <ufs/ufs/ufs_bswap.h>
102 #include <ufs/ufs/ufs_extern.h>
103 #include <ufs/ufs/ufs_wapbl.h>
104 #ifdef UFS_DIRHASH
105 #include <ufs/ufs/dirhash.h>
106 #endif
107 #include <ufs/ext2fs/ext2fs_extern.h>
108 #include <ufs/ffs/ffs_extern.h>
109 #include <ufs/lfs/lfs_extern.h>
110 
111 #include <uvm/uvm.h>
112 
113 static int ufs_chmod(struct vnode *, int, kauth_cred_t, struct lwp *);
114 static int ufs_chown(struct vnode *, uid_t, gid_t, kauth_cred_t,
115     struct lwp *);
116 
117 /*
118  * A virgin directory (no blushing please).
119  */
120 static const struct dirtemplate mastertemplate = {
121 	0,	12,		DT_DIR,	1,	".",
122 	0,	DIRBLKSIZ - 12,	DT_DIR,	2,	".."
123 };
124 
125 /*
126  * Create a regular file
127  */
128 int
129 ufs_create(void *v)
130 {
131 	struct vop_create_args /* {
132 		struct vnode		*a_dvp;
133 		struct vnode		**a_vpp;
134 		struct componentname	*a_cnp;
135 		struct vattr		*a_vap;
136 	} */ *ap = v;
137 	int	error;
138 
139 	/*
140 	 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
141 	 * ufs_makeinode
142 	 */
143 	fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
144 	error =
145 	    ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
146 			  ap->a_dvp, ap->a_vpp, ap->a_cnp);
147 	if (error) {
148 		fstrans_done(ap->a_dvp->v_mount);
149 		return (error);
150 	}
151 	UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
152 	fstrans_done(ap->a_dvp->v_mount);
153 	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
154 	return (0);
155 }
156 
157 /*
158  * Mknod vnode call
159  */
160 /* ARGSUSED */
161 int
162 ufs_mknod(void *v)
163 {
164 	struct vop_mknod_args /* {
165 		struct vnode		*a_dvp;
166 		struct vnode		**a_vpp;
167 		struct componentname	*a_cnp;
168 		struct vattr		*a_vap;
169 	} */ *ap = v;
170 	struct vattr	*vap;
171 	struct vnode	**vpp;
172 	struct inode	*ip;
173 	int		error;
174 	struct mount	*mp;
175 	ino_t		ino;
176 
177 	vap = ap->a_vap;
178 	vpp = ap->a_vpp;
179 
180 	/*
181 	 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
182 	 * ufs_makeinode
183 	 */
184 	fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
185 	if ((error =
186 	    ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
187 	    ap->a_dvp, vpp, ap->a_cnp)) != 0)
188 		goto out;
189 	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
190 	ip = VTOI(*vpp);
191 	mp  = (*vpp)->v_mount;
192 	ino = ip->i_number;
193 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
194 	if (vap->va_rdev != VNOVAL) {
195 		struct ufsmount *ump = ip->i_ump;
196 		/*
197 		 * Want to be able to use this to make badblock
198 		 * inodes, so don't truncate the dev number.
199 		 */
200 		if (ump->um_fstype == UFS1)
201 			ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev,
202 			    UFS_MPNEEDSWAP(ump));
203 		else
204 			ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev,
205 			    UFS_MPNEEDSWAP(ump));
206 	}
207 	UFS_WAPBL_UPDATE(*vpp, NULL, NULL, 0);
208 	UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
209 	/*
210 	 * Remove inode so that it will be reloaded by VFS_VGET and
211 	 * checked to see if it is an alias of an existing entry in
212 	 * the inode cache.
213 	 */
214 	(*vpp)->v_type = VNON;
215 	VOP_UNLOCK(*vpp);
216 	vgone(*vpp);
217 	error = VFS_VGET(mp, ino, vpp);
218 out:
219 	fstrans_done(ap->a_dvp->v_mount);
220 	if (error != 0) {
221 		*vpp = NULL;
222 		return (error);
223 	}
224 	return (0);
225 }
226 
227 /*
228  * Open called.
229  *
230  * Nothing to do.
231  */
232 /* ARGSUSED */
233 int
234 ufs_open(void *v)
235 {
236 	struct vop_open_args /* {
237 		struct vnode	*a_vp;
238 		int		a_mode;
239 		kauth_cred_t	a_cred;
240 	} */ *ap = v;
241 
242 	/*
243 	 * Files marked append-only must be opened for appending.
244 	 */
245 	if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
246 	    (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
247 		return (EPERM);
248 	return (0);
249 }
250 
251 /*
252  * Close called.
253  *
254  * Update the times on the inode.
255  */
256 /* ARGSUSED */
257 int
258 ufs_close(void *v)
259 {
260 	struct vop_close_args /* {
261 		struct vnode	*a_vp;
262 		int		a_fflag;
263 		kauth_cred_t	a_cred;
264 	} */ *ap = v;
265 	struct vnode	*vp;
266 	struct inode	*ip;
267 
268 	vp = ap->a_vp;
269 	ip = VTOI(vp);
270 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
271 	if (vp->v_usecount > 1)
272 		UFS_ITIMES(vp, NULL, NULL, NULL);
273 	fstrans_done(vp->v_mount);
274 	return (0);
275 }
276 
277 static int
278 ufs_check_possible(struct vnode *vp, struct inode *ip, mode_t mode,
279     kauth_cred_t cred)
280 {
281 #if defined(QUOTA) || defined(QUOTA2)
282 	int error;
283 #endif
284 
285 	/*
286 	 * Disallow write attempts on read-only file systems;
287 	 * unless the file is a socket, fifo, or a block or
288 	 * character device resident on the file system.
289 	 */
290 	if (mode & VWRITE) {
291 		switch (vp->v_type) {
292 		case VDIR:
293 		case VLNK:
294 		case VREG:
295 			if (vp->v_mount->mnt_flag & MNT_RDONLY)
296 				return (EROFS);
297 #if defined(QUOTA) || defined(QUOTA2)
298 			fstrans_start(vp->v_mount, FSTRANS_SHARED);
299 			error = chkdq(ip, 0, cred, 0);
300 			fstrans_done(vp->v_mount);
301 			if (error != 0)
302 				return error;
303 #endif
304 			break;
305 		case VBAD:
306 		case VBLK:
307 		case VCHR:
308 		case VSOCK:
309 		case VFIFO:
310 		case VNON:
311 		default:
312 			break;
313 		}
314 	}
315 
316 	/* If it is a snapshot, nobody gets access to it. */
317 	if ((ip->i_flags & SF_SNAPSHOT))
318 		return (EPERM);
319 	/* If immutable bit set, nobody gets to write it. */
320 	if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
321 		return (EPERM);
322 
323 	return 0;
324 }
325 
326 static int
327 ufs_check_permitted(struct vnode *vp, struct inode *ip, mode_t mode,
328     kauth_cred_t cred)
329 {
330 
331 	return genfs_can_access(vp->v_type, ip->i_mode & ALLPERMS, ip->i_uid,
332 	    ip->i_gid, mode, cred);
333 }
334 
335 int
336 ufs_access(void *v)
337 {
338 	struct vop_access_args /* {
339 		struct vnode	*a_vp;
340 		int		a_mode;
341 		kauth_cred_t	a_cred;
342 	} */ *ap = v;
343 	struct vnode	*vp;
344 	struct inode	*ip;
345 	mode_t		mode;
346 	int		error;
347 
348 	vp = ap->a_vp;
349 	ip = VTOI(vp);
350 	mode = ap->a_mode;
351 
352 	error = ufs_check_possible(vp, ip, mode, ap->a_cred);
353 	if (error)
354 		return error;
355 
356 	error = ufs_check_permitted(vp, ip, mode, ap->a_cred);
357 
358 	return error;
359 }
360 
361 /* ARGSUSED */
362 int
363 ufs_getattr(void *v)
364 {
365 	struct vop_getattr_args /* {
366 		struct vnode	*a_vp;
367 		struct vattr	*a_vap;
368 		kauth_cred_t	a_cred;
369 	} */ *ap = v;
370 	struct vnode	*vp;
371 	struct inode	*ip;
372 	struct vattr	*vap;
373 
374 	vp = ap->a_vp;
375 	ip = VTOI(vp);
376 	vap = ap->a_vap;
377 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
378 	UFS_ITIMES(vp, NULL, NULL, NULL);
379 
380 	/*
381 	 * Copy from inode table
382 	 */
383 	vap->va_fsid = ip->i_dev;
384 	vap->va_fileid = ip->i_number;
385 	vap->va_mode = ip->i_mode & ALLPERMS;
386 	vap->va_nlink = ip->i_nlink;
387 	vap->va_uid = ip->i_uid;
388 	vap->va_gid = ip->i_gid;
389 	vap->va_size = vp->v_size;
390 	if (ip->i_ump->um_fstype == UFS1) {
391 		vap->va_rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev,
392 		    UFS_MPNEEDSWAP(ip->i_ump));
393 		vap->va_atime.tv_sec = ip->i_ffs1_atime;
394 		vap->va_atime.tv_nsec = ip->i_ffs1_atimensec;
395 		vap->va_mtime.tv_sec = ip->i_ffs1_mtime;
396 		vap->va_mtime.tv_nsec = ip->i_ffs1_mtimensec;
397 		vap->va_ctime.tv_sec = ip->i_ffs1_ctime;
398 		vap->va_ctime.tv_nsec = ip->i_ffs1_ctimensec;
399 		vap->va_birthtime.tv_sec = 0;
400 		vap->va_birthtime.tv_nsec = 0;
401 		vap->va_bytes = dbtob((u_quad_t)ip->i_ffs1_blocks);
402 	} else {
403 		vap->va_rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev,
404 		    UFS_MPNEEDSWAP(ip->i_ump));
405 		vap->va_atime.tv_sec = ip->i_ffs2_atime;
406 		vap->va_atime.tv_nsec = ip->i_ffs2_atimensec;
407 		vap->va_mtime.tv_sec = ip->i_ffs2_mtime;
408 		vap->va_mtime.tv_nsec = ip->i_ffs2_mtimensec;
409 		vap->va_ctime.tv_sec = ip->i_ffs2_ctime;
410 		vap->va_ctime.tv_nsec = ip->i_ffs2_ctimensec;
411 		vap->va_birthtime.tv_sec = ip->i_ffs2_birthtime;
412 		vap->va_birthtime.tv_nsec = ip->i_ffs2_birthnsec;
413 		vap->va_bytes = dbtob(ip->i_ffs2_blocks);
414 	}
415 	vap->va_gen = ip->i_gen;
416 	vap->va_flags = ip->i_flags;
417 
418 	/* this doesn't belong here */
419 	if (vp->v_type == VBLK)
420 		vap->va_blocksize = BLKDEV_IOSIZE;
421 	else if (vp->v_type == VCHR)
422 		vap->va_blocksize = MAXBSIZE;
423 	else
424 		vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
425 	vap->va_type = vp->v_type;
426 	vap->va_filerev = ip->i_modrev;
427 	fstrans_done(vp->v_mount);
428 	return (0);
429 }
430 
431 /*
432  * Set attribute vnode op. called from several syscalls
433  */
434 int
435 ufs_setattr(void *v)
436 {
437 	struct vop_setattr_args /* {
438 		struct vnode	*a_vp;
439 		struct vattr	*a_vap;
440 		kauth_cred_t	a_cred;
441 	} */ *ap = v;
442 	struct vattr	*vap;
443 	struct vnode	*vp;
444 	struct inode	*ip;
445 	kauth_cred_t	cred;
446 	struct lwp	*l;
447 	int		error;
448 
449 	vap = ap->a_vap;
450 	vp = ap->a_vp;
451 	ip = VTOI(vp);
452 	cred = ap->a_cred;
453 	l = curlwp;
454 
455 	/*
456 	 * Check for unsettable attributes.
457 	 */
458 	if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
459 	    (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
460 	    (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
461 	    ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
462 		return (EINVAL);
463 	}
464 
465 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
466 
467 	if (vap->va_flags != VNOVAL) {
468 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
469 			error = EROFS;
470 			goto out;
471 		}
472 		if (kauth_cred_geteuid(cred) != ip->i_uid &&
473 		    (error = kauth_authorize_generic(cred,
474 		    KAUTH_GENERIC_ISSUSER, NULL)))
475 			goto out;
476 		if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
477 		    NULL) == 0) {
478 			if ((ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) &&
479 			    kauth_authorize_system(l->l_cred,
480 			     KAUTH_SYSTEM_CHSYSFLAGS, 0, NULL, NULL, NULL)) {
481 				error = EPERM;
482 				goto out;
483 			}
484 			/* Snapshot flag cannot be set or cleared */
485 			if ((vap->va_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) !=
486 			    (ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))) {
487 				error = EPERM;
488 				goto out;
489 			}
490 			error = UFS_WAPBL_BEGIN(vp->v_mount);
491 			if (error)
492 				goto out;
493 			ip->i_flags = vap->va_flags;
494 			DIP_ASSIGN(ip, flags, ip->i_flags);
495 		} else {
496 			if ((ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) ||
497 			    (vap->va_flags & UF_SETTABLE) != vap->va_flags) {
498 				error = EPERM;
499 				goto out;
500 			}
501 			if ((ip->i_flags & SF_SETTABLE) !=
502 			    (vap->va_flags & SF_SETTABLE)) {
503 				error = EPERM;
504 				goto out;
505 			}
506 			error = UFS_WAPBL_BEGIN(vp->v_mount);
507 			if (error)
508 				goto out;
509 			ip->i_flags &= SF_SETTABLE;
510 			ip->i_flags |= (vap->va_flags & UF_SETTABLE);
511 			DIP_ASSIGN(ip, flags, ip->i_flags);
512 		}
513 		ip->i_flag |= IN_CHANGE;
514 		UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
515 		UFS_WAPBL_END(vp->v_mount);
516 		if (vap->va_flags & (IMMUTABLE | APPEND)) {
517 			error = 0;
518 			goto out;
519 		}
520 	}
521 	if (ip->i_flags & (IMMUTABLE | APPEND)) {
522 		error = EPERM;
523 		goto out;
524 	}
525 	/*
526 	 * Go through the fields and update iff not VNOVAL.
527 	 */
528 	if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
529 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
530 			error = EROFS;
531 			goto out;
532 		}
533 		error = UFS_WAPBL_BEGIN(vp->v_mount);
534 		if (error)
535 			goto out;
536 		error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, l);
537 		UFS_WAPBL_END(vp->v_mount);
538 		if (error)
539 			goto out;
540 	}
541 	if (vap->va_size != VNOVAL) {
542 		/*
543 		 * Disallow write attempts on read-only file systems;
544 		 * unless the file is a socket, fifo, or a block or
545 		 * character device resident on the file system.
546 		 */
547 		switch (vp->v_type) {
548 		case VDIR:
549 			error = EISDIR;
550 			goto out;
551 		case VCHR:
552 		case VBLK:
553 		case VFIFO:
554 			break;
555 		case VREG:
556 			if (vp->v_mount->mnt_flag & MNT_RDONLY) {
557 				error = EROFS;
558 				goto out;
559 			}
560 			if ((ip->i_flags & SF_SNAPSHOT) != 0) {
561 				error = EPERM;
562 				goto out;
563 			}
564 			error = UFS_WAPBL_BEGIN(vp->v_mount);
565 			if (error)
566 				goto out;
567 			/*
568 			 * When journaling, only truncate one indirect block
569 			 * at a time.
570 			 */
571 			if (vp->v_mount->mnt_wapbl) {
572 				uint64_t incr = MNINDIR(ip->i_ump) <<
573 				    vp->v_mount->mnt_fs_bshift; /* Power of 2 */
574 				uint64_t base = NDADDR <<
575 				    vp->v_mount->mnt_fs_bshift;
576 				while (!error && ip->i_size > base + incr &&
577 				    ip->i_size > vap->va_size + incr) {
578 					/*
579 					 * round down to next full indirect
580 					 * block boundary.
581 					 */
582 					uint64_t nsize = base +
583 					    ((ip->i_size - base - 1) &
584 					    ~(incr - 1));
585 					error = UFS_TRUNCATE(vp, nsize, 0,
586 					    cred);
587 					if (error == 0) {
588 						UFS_WAPBL_END(vp->v_mount);
589 						error =
590 						   UFS_WAPBL_BEGIN(vp->v_mount);
591 					}
592 				}
593 			}
594 			if (!error)
595 				error = UFS_TRUNCATE(vp, vap->va_size, 0, cred);
596 			UFS_WAPBL_END(vp->v_mount);
597 			if (error)
598 				goto out;
599 			break;
600 		default:
601 			error = EOPNOTSUPP;
602 			goto out;
603 		}
604 	}
605 	ip = VTOI(vp);
606 	if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL ||
607 	    vap->va_birthtime.tv_sec != VNOVAL) {
608 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
609 			error = EROFS;
610 			goto out;
611 		}
612 		if ((ip->i_flags & SF_SNAPSHOT) != 0) {
613 			error = EPERM;
614 			goto out;
615 		}
616 		error = genfs_can_chtimes(vp, vap->va_vaflags, ip->i_uid, cred);
617 		if (error)
618 			goto out;
619 		error = UFS_WAPBL_BEGIN(vp->v_mount);
620 		if (error)
621 			goto out;
622 		if (vap->va_atime.tv_sec != VNOVAL)
623 			if (!(vp->v_mount->mnt_flag & MNT_NOATIME))
624 				ip->i_flag |= IN_ACCESS;
625 		if (vap->va_mtime.tv_sec != VNOVAL)
626 			ip->i_flag |= IN_CHANGE | IN_UPDATE;
627 		if (vap->va_birthtime.tv_sec != VNOVAL &&
628 		    ip->i_ump->um_fstype == UFS2) {
629 			ip->i_ffs2_birthtime = vap->va_birthtime.tv_sec;
630 			ip->i_ffs2_birthnsec = vap->va_birthtime.tv_nsec;
631 		}
632 		error = UFS_UPDATE(vp, &vap->va_atime, &vap->va_mtime, 0);
633 		UFS_WAPBL_END(vp->v_mount);
634 		if (error)
635 			goto out;
636 	}
637 	error = 0;
638 	if (vap->va_mode != (mode_t)VNOVAL) {
639 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
640 			error = EROFS;
641 			goto out;
642 		}
643 		if ((ip->i_flags & SF_SNAPSHOT) != 0 &&
644 		    (vap->va_mode & (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP |
645 		     S_IXOTH | S_IWOTH))) {
646 			error = EPERM;
647 			goto out;
648 		}
649 		error = UFS_WAPBL_BEGIN(vp->v_mount);
650 		if (error)
651 			goto out;
652 		error = ufs_chmod(vp, (int)vap->va_mode, cred, l);
653 		UFS_WAPBL_END(vp->v_mount);
654 	}
655 	VN_KNOTE(vp, NOTE_ATTRIB);
656 out:
657 	fstrans_done(vp->v_mount);
658 	return (error);
659 }
660 
661 /*
662  * Change the mode on a file.
663  * Inode must be locked before calling.
664  */
665 static int
666 ufs_chmod(struct vnode *vp, int mode, kauth_cred_t cred, struct lwp *l)
667 {
668 	struct inode	*ip;
669 	int		error;
670 
671 	UFS_WAPBL_JLOCK_ASSERT(vp->v_mount);
672 
673 	ip = VTOI(vp);
674 
675 	error = genfs_can_chmod(vp, cred, ip->i_uid, ip->i_gid, mode);
676 	if (error)
677 		return (error);
678 
679 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
680 	ip->i_mode &= ~ALLPERMS;
681 	ip->i_mode |= (mode & ALLPERMS);
682 	ip->i_flag |= IN_CHANGE;
683 	DIP_ASSIGN(ip, mode, ip->i_mode);
684 	UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
685 	fstrans_done(vp->v_mount);
686 	return (0);
687 }
688 
689 /*
690  * Perform chown operation on inode ip;
691  * inode must be locked prior to call.
692  */
693 static int
694 ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, kauth_cred_t cred,
695     	struct lwp *l)
696 {
697 	struct inode	*ip;
698 	int		error = 0;
699 #if defined(QUOTA) || defined(QUOTA2)
700 	uid_t		ouid;
701 	gid_t		ogid;
702 	int64_t		change;
703 #endif
704 	ip = VTOI(vp);
705 	error = 0;
706 
707 	if (uid == (uid_t)VNOVAL)
708 		uid = ip->i_uid;
709 	if (gid == (gid_t)VNOVAL)
710 		gid = ip->i_gid;
711 
712 	error = genfs_can_chown(vp, cred, ip->i_uid, ip->i_gid, uid, gid);
713 	if (error)
714 		return (error);
715 
716 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
717 #if defined(QUOTA) || defined(QUOTA2)
718 	ogid = ip->i_gid;
719 	ouid = ip->i_uid;
720 	change = DIP(ip, blocks);
721 	(void) chkdq(ip, -change, cred, 0);
722 	(void) chkiq(ip, -1, cred, 0);
723 #endif
724 	ip->i_gid = gid;
725 	DIP_ASSIGN(ip, gid, gid);
726 	ip->i_uid = uid;
727 	DIP_ASSIGN(ip, uid, uid);
728 #if defined(QUOTA) || defined(QUOTA2)
729 	if ((error = chkdq(ip, change, cred, 0)) == 0) {
730 		if ((error = chkiq(ip, 1, cred, 0)) == 0)
731 			goto good;
732 		else
733 			(void) chkdq(ip, -change, cred, FORCE);
734 	}
735 	ip->i_gid = ogid;
736 	DIP_ASSIGN(ip, gid, ogid);
737 	ip->i_uid = ouid;
738 	DIP_ASSIGN(ip, uid, ouid);
739 	(void) chkdq(ip, change, cred, FORCE);
740 	(void) chkiq(ip, 1, cred, FORCE);
741 	fstrans_done(vp->v_mount);
742 	return (error);
743  good:
744 #endif /* QUOTA || QUOTA2 */
745 	ip->i_flag |= IN_CHANGE;
746 	UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
747 	fstrans_done(vp->v_mount);
748 	return (0);
749 }
750 
751 int
752 ufs_remove(void *v)
753 {
754 	struct vop_remove_args /* {
755 		struct vnode		*a_dvp;
756 		struct vnode		*a_vp;
757 		struct componentname	*a_cnp;
758 	} */ *ap = v;
759 	struct vnode	*vp, *dvp;
760 	struct inode	*ip;
761 	int		error;
762 
763 	vp = ap->a_vp;
764 	dvp = ap->a_dvp;
765 	ip = VTOI(vp);
766 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
767 	if (vp->v_type == VDIR || (ip->i_flags & (IMMUTABLE | APPEND)) ||
768 	    (VTOI(dvp)->i_flags & APPEND))
769 		error = EPERM;
770 	else {
771 		error = UFS_WAPBL_BEGIN(dvp->v_mount);
772 		if (error == 0) {
773 			error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
774 			UFS_WAPBL_END(dvp->v_mount);
775 		}
776 	}
777 	VN_KNOTE(vp, NOTE_DELETE);
778 	VN_KNOTE(dvp, NOTE_WRITE);
779 	if (dvp == vp)
780 		vrele(vp);
781 	else
782 		vput(vp);
783 	vput(dvp);
784 	fstrans_done(dvp->v_mount);
785 	return (error);
786 }
787 
788 /*
789  * ufs_link: create hard link.
790  */
791 int
792 ufs_link(void *v)
793 {
794 	struct vop_link_args /* {
795 		struct vnode *a_dvp;
796 		struct vnode *a_vp;
797 		struct componentname *a_cnp;
798 	} */ *ap = v;
799 	struct vnode *dvp = ap->a_dvp;
800 	struct vnode *vp = ap->a_vp;
801 	struct componentname *cnp = ap->a_cnp;
802 	struct inode *ip;
803 	struct direct *newdir;
804 	int error;
805 
806 	KASSERT(dvp != vp);
807 	KASSERT(vp->v_type != VDIR);
808 	KASSERT(dvp->v_mount == vp->v_mount);
809 
810 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
811 	error = vn_lock(vp, LK_EXCLUSIVE);
812 	if (error) {
813 		VOP_ABORTOP(dvp, cnp);
814 		goto out2;
815 	}
816 	ip = VTOI(vp);
817 	if ((nlink_t)ip->i_nlink >= LINK_MAX) {
818 		VOP_ABORTOP(dvp, cnp);
819 		error = EMLINK;
820 		goto out1;
821 	}
822 	if (ip->i_flags & (IMMUTABLE | APPEND)) {
823 		VOP_ABORTOP(dvp, cnp);
824 		error = EPERM;
825 		goto out1;
826 	}
827 	error = UFS_WAPBL_BEGIN(vp->v_mount);
828 	if (error) {
829 		VOP_ABORTOP(dvp, cnp);
830 		goto out1;
831 	}
832 	ip->i_nlink++;
833 	DIP_ASSIGN(ip, nlink, ip->i_nlink);
834 	ip->i_flag |= IN_CHANGE;
835 	error = UFS_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
836 	if (!error) {
837 		newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
838 		ufs_makedirentry(ip, cnp, newdir);
839 		error = ufs_direnter(dvp, vp, newdir, cnp, NULL);
840 		pool_cache_put(ufs_direct_cache, newdir);
841 	}
842 	if (error) {
843 		ip->i_nlink--;
844 		DIP_ASSIGN(ip, nlink, ip->i_nlink);
845 		ip->i_flag |= IN_CHANGE;
846 		UFS_WAPBL_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
847 	}
848 	UFS_WAPBL_END(vp->v_mount);
849  out1:
850 	VOP_UNLOCK(vp);
851  out2:
852 	VN_KNOTE(vp, NOTE_LINK);
853 	VN_KNOTE(dvp, NOTE_WRITE);
854 	vput(dvp);
855 	fstrans_done(dvp->v_mount);
856 	return (error);
857 }
858 
859 /*
860  * whiteout vnode call
861  */
862 int
863 ufs_whiteout(void *v)
864 {
865 	struct vop_whiteout_args /* {
866 		struct vnode		*a_dvp;
867 		struct componentname	*a_cnp;
868 		int			a_flags;
869 	} */ *ap = v;
870 	struct vnode		*dvp = ap->a_dvp;
871 	struct componentname	*cnp = ap->a_cnp;
872 	struct direct		*newdir;
873 	int			error;
874 	struct ufsmount		*ump = VFSTOUFS(dvp->v_mount);
875 
876 	error = 0;
877 	switch (ap->a_flags) {
878 	case LOOKUP:
879 		/* 4.4 format directories support whiteout operations */
880 		if (ump->um_maxsymlinklen > 0)
881 			return (0);
882 		return (EOPNOTSUPP);
883 
884 	case CREATE:
885 		/* create a new directory whiteout */
886 		fstrans_start(dvp->v_mount, FSTRANS_SHARED);
887 #ifdef DIAGNOSTIC
888 		if (ump->um_maxsymlinklen <= 0)
889 			panic("ufs_whiteout: old format filesystem");
890 #endif
891 
892 		newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
893 		newdir->d_ino = WINO;
894 		newdir->d_namlen = cnp->cn_namelen;
895 		memcpy(newdir->d_name, cnp->cn_nameptr,
896 		    (size_t)cnp->cn_namelen);
897 		newdir->d_name[cnp->cn_namelen] = '\0';
898 		newdir->d_type = DT_WHT;
899 		error = ufs_direnter(dvp, NULL, newdir, cnp, NULL);
900 		pool_cache_put(ufs_direct_cache, newdir);
901 		break;
902 
903 	case DELETE:
904 		/* remove an existing directory whiteout */
905 		fstrans_start(dvp->v_mount, FSTRANS_SHARED);
906 #ifdef DIAGNOSTIC
907 		if (ump->um_maxsymlinklen <= 0)
908 			panic("ufs_whiteout: old format filesystem");
909 #endif
910 
911 		cnp->cn_flags &= ~DOWHITEOUT;
912 		error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
913 		break;
914 	default:
915 		panic("ufs_whiteout: unknown op");
916 		/* NOTREACHED */
917 	}
918 	fstrans_done(dvp->v_mount);
919 	return (error);
920 }
921 
922 
923 /*
924  * Rename vnode operation
925  * 	rename("foo", "bar");
926  * is essentially
927  *	unlink("bar");
928  *	link("foo", "bar");
929  *	unlink("foo");
930  * but ``atomically''.  Can't do full commit without saving state in the
931  * inode on disk which isn't feasible at this time.  Best we can do is
932  * always guarantee the target exists.
933  *
934  * Basic algorithm is:
935  *
936  * 1) Bump link count on source while we're linking it to the
937  *    target.  This also ensure the inode won't be deleted out
938  *    from underneath us while we work (it may be truncated by
939  *    a concurrent `trunc' or `open' for creation).
940  * 2) Link source to destination.  If destination already exists,
941  *    delete it first.
942  * 3) Unlink source reference to inode if still around. If a
943  *    directory was moved and the parent of the destination
944  *    is different from the source, patch the ".." entry in the
945  *    directory.
946  */
947 int
948 ufs_rename(void *v)
949 {
950 	struct vop_rename_args  /* {
951 		struct vnode		*a_fdvp;
952 		struct vnode		*a_fvp;
953 		struct componentname	*a_fcnp;
954 		struct vnode		*a_tdvp;
955 		struct vnode		*a_tvp;
956 		struct componentname	*a_tcnp;
957 	} */ *ap = v;
958 	struct vnode		*tvp, *tdvp, *fvp, *fdvp;
959 	struct componentname	*tcnp, *fcnp;
960 	struct inode		*ip, *xp, *dp;
961 	struct mount		*mp;
962 	struct direct		*newdir;
963 	int			doingdirectory, oldparent, newparent, error;
964 
965 #ifdef WAPBL
966 	if (ap->a_tdvp->v_mount->mnt_wapbl)
967 		return wapbl_ufs_rename(v);
968 #endif
969 
970 	tvp = ap->a_tvp;
971 	tdvp = ap->a_tdvp;
972 	fvp = ap->a_fvp;
973 	fdvp = ap->a_fdvp;
974 	tcnp = ap->a_tcnp;
975 	fcnp = ap->a_fcnp;
976 	doingdirectory = oldparent = newparent = error = 0;
977 
978 	/*
979 	 * Check for cross-device rename.
980 	 */
981 	if ((fvp->v_mount != tdvp->v_mount) ||
982 	    (tvp && (fvp->v_mount != tvp->v_mount))) {
983 		error = EXDEV;
984  abortit:
985 		VOP_ABORTOP(tdvp, tcnp); /* XXX, why not in NFS? */
986 		if (tdvp == tvp)
987 			vrele(tdvp);
988 		else
989 			vput(tdvp);
990 		if (tvp)
991 			vput(tvp);
992 		VOP_ABORTOP(fdvp, fcnp); /* XXX, why not in NFS? */
993 		vrele(fdvp);
994 		vrele(fvp);
995 		return (error);
996 	}
997 
998 	/*
999 	 * Check if just deleting a link name.
1000 	 */
1001 	if (tvp && ((VTOI(tvp)->i_flags & (IMMUTABLE | APPEND)) ||
1002 	    (VTOI(tdvp)->i_flags & APPEND))) {
1003 		error = EPERM;
1004 		goto abortit;
1005 	}
1006 	if (fvp == tvp) {
1007 		if (fvp->v_type == VDIR) {
1008 			error = EINVAL;
1009 			goto abortit;
1010 		}
1011 
1012 		/* Release destination completely. */
1013 		VOP_ABORTOP(tdvp, tcnp);
1014 		vput(tdvp);
1015 		vput(tvp);
1016 
1017 		/* Delete source. */
1018 		vrele(fvp);
1019 		fcnp->cn_flags &= ~(MODMASK);
1020 		fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1021 		fcnp->cn_nameiop = DELETE;
1022 		vn_lock(fdvp, LK_EXCLUSIVE | LK_RETRY);
1023 		if ((error = relookup(fdvp, &fvp, fcnp, 0))) {
1024 			vput(fdvp);
1025 			return (error);
1026 		}
1027 		return (VOP_REMOVE(fdvp, fvp, fcnp));
1028 	}
1029 	if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
1030 		goto abortit;
1031 	dp = VTOI(fdvp);
1032 	ip = VTOI(fvp);
1033 	if ((nlink_t) ip->i_nlink >= LINK_MAX) {
1034 		VOP_UNLOCK(fvp);
1035 		error = EMLINK;
1036 		goto abortit;
1037 	}
1038 	if ((ip->i_flags & (IMMUTABLE | APPEND)) ||
1039 		(dp->i_flags & APPEND)) {
1040 		VOP_UNLOCK(fvp);
1041 		error = EPERM;
1042 		goto abortit;
1043 	}
1044 	if ((ip->i_mode & IFMT) == IFDIR) {
1045 		/*
1046 		 * Avoid ".", "..", and aliases of "." for obvious reasons.
1047 		 */
1048 		if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
1049 		    dp == ip ||
1050 		    (fcnp->cn_flags & ISDOTDOT) ||
1051 		    (tcnp->cn_flags & ISDOTDOT) ||
1052 		    (ip->i_flag & IN_RENAME)) {
1053 			VOP_UNLOCK(fvp);
1054 			error = EINVAL;
1055 			goto abortit;
1056 		}
1057 		ip->i_flag |= IN_RENAME;
1058 		oldparent = dp->i_number;
1059 		doingdirectory = 1;
1060 	}
1061 	VN_KNOTE(fdvp, NOTE_WRITE);		/* XXXLUKEM/XXX: right place? */
1062 
1063 	/*
1064 	 * When the target exists, both the directory
1065 	 * and target vnodes are returned locked.
1066 	 */
1067 	dp = VTOI(tdvp);
1068 	xp = NULL;
1069 	if (tvp)
1070 		xp = VTOI(tvp);
1071 
1072 	mp = fdvp->v_mount;
1073 	fstrans_start(mp, FSTRANS_SHARED);
1074 
1075 	/*
1076 	 * 1) Bump link count while we're moving stuff
1077 	 *    around.  If we crash somewhere before
1078 	 *    completing our work, the link count
1079 	 *    may be wrong, but correctable.
1080 	 */
1081 	ip->i_nlink++;
1082 	DIP_ASSIGN(ip, nlink, ip->i_nlink);
1083 	ip->i_flag |= IN_CHANGE;
1084 	if ((error = UFS_UPDATE(fvp, NULL, NULL, UPDATE_DIROP)) != 0) {
1085 		VOP_UNLOCK(fvp);
1086 		goto bad;
1087 	}
1088 
1089 	/*
1090 	 * If ".." must be changed (ie the directory gets a new
1091 	 * parent) then the source directory must not be in the
1092 	 * directory hierarchy above the target, as this would
1093 	 * orphan everything below the source directory. Also
1094 	 * the user must have write permission in the source so
1095 	 * as to be able to change "..". We must repeat the call
1096 	 * to namei, as the parent directory is unlocked by the
1097 	 * call to checkpath().
1098 	 */
1099 	error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
1100 	VOP_UNLOCK(fvp);
1101 	if (oldparent != dp->i_number)
1102 		newparent = dp->i_number;
1103 	if (doingdirectory && newparent) {
1104 		if (error)	/* write access check above */
1105 			goto bad;
1106 		if (xp != NULL)
1107 			vput(tvp);
1108 		vref(tdvp);	/* compensate for the ref checkpath loses */
1109 		if ((error = ufs_checkpath(ip, dp, tcnp->cn_cred)) != 0) {
1110 			vrele(tdvp);
1111 			goto out;
1112 		}
1113 		vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY);
1114 		error = relookup(tdvp, &tvp, tcnp, 0);
1115 		if (error != 0) {
1116 			vput(tdvp);
1117 			goto out;
1118 		}
1119 		dp = VTOI(tdvp);
1120 		xp = NULL;
1121 		if (tvp)
1122 			xp = VTOI(tvp);
1123 	}
1124 	/*
1125 	 * 2) If target doesn't exist, link the target
1126 	 *    to the source and unlink the source.
1127 	 *    Otherwise, rewrite the target directory
1128 	 *    entry to reference the source inode and
1129 	 *    expunge the original entry's existence.
1130 	 */
1131 	if (xp == NULL) {
1132 		if (dp->i_dev != ip->i_dev)
1133 			panic("rename: EXDEV");
1134 		/*
1135 		 * Account for ".." in new directory.
1136 		 * When source and destination have the same
1137 		 * parent we don't fool with the link count.
1138 		 */
1139 		if (doingdirectory && newparent) {
1140 			if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1141 				error = EMLINK;
1142 				goto bad;
1143 			}
1144 			dp->i_nlink++;
1145 			DIP_ASSIGN(dp, nlink, dp->i_nlink);
1146 			dp->i_flag |= IN_CHANGE;
1147 			if ((error = UFS_UPDATE(tdvp, NULL, NULL,
1148 			    UPDATE_DIROP)) != 0) {
1149 				dp->i_nlink--;
1150 				DIP_ASSIGN(dp, nlink, dp->i_nlink);
1151 				dp->i_flag |= IN_CHANGE;
1152 				goto bad;
1153 			}
1154 		}
1155 		newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
1156 		ufs_makedirentry(ip, tcnp, newdir);
1157 		error = ufs_direnter(tdvp, NULL, newdir, tcnp, NULL);
1158 		pool_cache_put(ufs_direct_cache, newdir);
1159 		if (error != 0) {
1160 			if (doingdirectory && newparent) {
1161 				dp->i_nlink--;
1162 				DIP_ASSIGN(dp, nlink, dp->i_nlink);
1163 				dp->i_flag |= IN_CHANGE;
1164 				(void)UFS_UPDATE(tdvp, NULL, NULL,
1165 						 UPDATE_WAIT|UPDATE_DIROP);
1166 			}
1167 			goto bad;
1168 		}
1169 		VN_KNOTE(tdvp, NOTE_WRITE);
1170 		vput(tdvp);
1171 	} else {
1172 		if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1173 			panic("rename: EXDEV");
1174 		/*
1175 		 * Short circuit rename(foo, foo).
1176 		 */
1177 		if (xp->i_number == ip->i_number)
1178 			panic("rename: same file");
1179 		/*
1180 		 * If the parent directory is "sticky", then the user must
1181 		 * own the parent directory, or the destination of the rename,
1182 		 * otherwise the destination may not be changed (except by
1183 		 * root). This implements append-only directories.
1184 		 */
1185 		if ((dp->i_mode & S_ISTXT) &&
1186 		    kauth_authorize_generic(tcnp->cn_cred,
1187 		     KAUTH_GENERIC_ISSUSER, NULL) != 0 &&
1188 		    kauth_cred_geteuid(tcnp->cn_cred) != dp->i_uid &&
1189 		    xp->i_uid != kauth_cred_geteuid(tcnp->cn_cred)) {
1190 			error = EPERM;
1191 			goto bad;
1192 		}
1193 		/*
1194 		 * Target must be empty if a directory and have no links
1195 		 * to it. Also, ensure source and target are compatible
1196 		 * (both directories, or both not directories).
1197 		 */
1198 		if ((xp->i_mode & IFMT) == IFDIR) {
1199 			if (xp->i_nlink > 2 ||
1200 			    !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1201 				error = ENOTEMPTY;
1202 				goto bad;
1203 			}
1204 			if (!doingdirectory) {
1205 				error = ENOTDIR;
1206 				goto bad;
1207 			}
1208 			cache_purge(tdvp);
1209 		} else if (doingdirectory) {
1210 			error = EISDIR;
1211 			goto bad;
1212 		}
1213 		if ((error = ufs_dirrewrite(dp, xp, ip->i_number,
1214 		    IFTODT(ip->i_mode), doingdirectory && newparent ?
1215 		    newparent : doingdirectory, IN_CHANGE | IN_UPDATE)) != 0)
1216 			goto bad;
1217 		if (doingdirectory) {
1218 			/*
1219 			 * Truncate inode. The only stuff left in the directory
1220 			 * is "." and "..". The "." reference is inconsequential
1221 			 * since we are quashing it. We have removed the "."
1222 			 * reference and the reference in the parent directory,
1223 			 * but there may be other hard links.
1224 			 */
1225 			if (!newparent) {
1226 				dp->i_nlink--;
1227 				DIP_ASSIGN(dp, nlink, dp->i_nlink);
1228 				dp->i_flag |= IN_CHANGE;
1229 			}
1230 			xp->i_nlink--;
1231 			DIP_ASSIGN(xp, nlink, xp->i_nlink);
1232 			xp->i_flag |= IN_CHANGE;
1233 			if ((error = UFS_TRUNCATE(tvp, (off_t)0, IO_SYNC,
1234 			    tcnp->cn_cred)))
1235 				goto bad;
1236 		}
1237 		VN_KNOTE(tdvp, NOTE_WRITE);
1238 		vput(tdvp);
1239 		VN_KNOTE(tvp, NOTE_DELETE);
1240 		vput(tvp);
1241 		xp = NULL;
1242 	}
1243 
1244 	/*
1245 	 * 3) Unlink the source.
1246 	 */
1247 	fcnp->cn_flags &= ~(MODMASK);
1248 	fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1249 	vn_lock(fdvp, LK_EXCLUSIVE | LK_RETRY);
1250 	if ((error = relookup(fdvp, &fvp, fcnp, 0))) {
1251 		vput(fdvp);
1252 		vrele(ap->a_fvp);
1253 		goto out2;
1254 	}
1255 	if (fvp != NULL) {
1256 		xp = VTOI(fvp);
1257 		dp = VTOI(fdvp);
1258 	} else {
1259 		/*
1260 		 * From name has disappeared.
1261 		 */
1262 		if (doingdirectory)
1263 			panic("rename: lost dir entry");
1264 		vrele(ap->a_fvp);
1265 		error = 0;
1266 		goto out2;
1267 	}
1268 	/*
1269 	 * Ensure that the directory entry still exists and has not
1270 	 * changed while the new name has been entered. If the source is
1271 	 * a file then the entry may have been unlinked or renamed. In
1272 	 * either case there is no further work to be done. If the source
1273 	 * is a directory then it cannot have been rmdir'ed; The IRENAME
1274 	 * flag ensures that it cannot be moved by another rename or removed
1275 	 * by a rmdir.
1276 	 */
1277 	if (xp != ip) {
1278 		if (doingdirectory)
1279 			panic("rename: lost dir entry");
1280 	} else {
1281 		/*
1282 		 * If the source is a directory with a
1283 		 * new parent, the link count of the old
1284 		 * parent directory must be decremented
1285 		 * and ".." set to point to the new parent.
1286 		 */
1287 		if (doingdirectory && newparent) {
1288 			KASSERT(dp != NULL);
1289 			xp->i_offset = mastertemplate.dot_reclen;
1290 			ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0, IN_CHANGE);
1291 			cache_purge(fdvp);
1292 		}
1293 		error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1294 		xp->i_flag &= ~IN_RENAME;
1295 	}
1296 	VN_KNOTE(fvp, NOTE_RENAME);
1297 	if (dp)
1298 		vput(fdvp);
1299 	if (xp)
1300 		vput(fvp);
1301 	vrele(ap->a_fvp);
1302 	goto out2;
1303 
1304 	/* exit routines from steps 1 & 2 */
1305  bad:
1306 	if (xp)
1307 		vput(ITOV(xp));
1308 	vput(ITOV(dp));
1309  out:
1310 	if (doingdirectory)
1311 		ip->i_flag &= ~IN_RENAME;
1312 	if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
1313 		ip->i_nlink--;
1314 		DIP_ASSIGN(ip, nlink, ip->i_nlink);
1315 		ip->i_flag |= IN_CHANGE;
1316 		ip->i_flag &= ~IN_RENAME;
1317 		vput(fvp);
1318 	} else
1319 		vrele(fvp);
1320 	vrele(fdvp);
1321 
1322 	/* exit routines from step 3 */
1323  out2:
1324 	fstrans_done(mp);
1325 	return (error);
1326 }
1327 
1328 int
1329 ufs_mkdir(void *v)
1330 {
1331 	struct vop_mkdir_args /* {
1332 		struct vnode		*a_dvp;
1333 		struct vnode		**a_vpp;
1334 		struct componentname	*a_cnp;
1335 		struct vattr		*a_vap;
1336 	} */ *ap = v;
1337 	struct vnode		*dvp = ap->a_dvp, *tvp;
1338 	struct vattr		*vap = ap->a_vap;
1339 	struct componentname	*cnp = ap->a_cnp;
1340 	struct inode		*ip, *dp = VTOI(dvp);
1341 	struct buf		*bp;
1342 	struct dirtemplate	dirtemplate;
1343 	struct direct		*newdir;
1344 	int			error, dmode;
1345 	struct ufsmount		*ump = dp->i_ump;
1346 	int			dirblksiz = ump->um_dirblksiz;
1347 
1348 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
1349 
1350 	if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1351 		error = EMLINK;
1352 		goto out;
1353 	}
1354 	dmode = vap->va_mode & ACCESSPERMS;
1355 	dmode |= IFDIR;
1356 	/*
1357 	 * Must simulate part of ufs_makeinode here to acquire the inode,
1358 	 * but not have it entered in the parent directory. The entry is
1359 	 * made later after writing "." and ".." entries.
1360 	 */
1361 	if ((error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, ap->a_vpp)) != 0)
1362 		goto out;
1363 	error = UFS_WAPBL_BEGIN(ap->a_dvp->v_mount);
1364 	if (error)
1365 		goto out;
1366 	tvp = *ap->a_vpp;
1367 	ip = VTOI(tvp);
1368 	ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
1369 	DIP_ASSIGN(ip, uid, ip->i_uid);
1370 	ip->i_gid = dp->i_gid;
1371 	DIP_ASSIGN(ip, gid, ip->i_gid);
1372 #if defined(QUOTA) || defined(QUOTA2)
1373 	if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1374 		UFS_VFREE(tvp, ip->i_number, dmode);
1375 		UFS_WAPBL_END(dvp->v_mount);
1376 		fstrans_done(dvp->v_mount);
1377 		vput(tvp);
1378 		vput(dvp);
1379 		return (error);
1380 	}
1381 #endif
1382 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1383 	ip->i_mode = dmode;
1384 	DIP_ASSIGN(ip, mode, dmode);
1385 	tvp->v_type = VDIR;	/* Rest init'd in getnewvnode(). */
1386 	ip->i_nlink = 2;
1387 	DIP_ASSIGN(ip, nlink, 2);
1388 	if (cnp->cn_flags & ISWHITEOUT) {
1389 		ip->i_flags |= UF_OPAQUE;
1390 		DIP_ASSIGN(ip, flags, ip->i_flags);
1391 	}
1392 
1393 	/*
1394 	 * Bump link count in parent directory to reflect work done below.
1395 	 * Should be done before reference is created so cleanup is
1396 	 * possible if we crash.
1397 	 */
1398 	dp->i_nlink++;
1399 	DIP_ASSIGN(dp, nlink, dp->i_nlink);
1400 	dp->i_flag |= IN_CHANGE;
1401 	if ((error = UFS_UPDATE(dvp, NULL, NULL, UPDATE_DIROP)) != 0)
1402 		goto bad;
1403 
1404 	/*
1405 	 * Initialize directory with "." and ".." from static template.
1406 	 */
1407 	dirtemplate = mastertemplate;
1408 	dirtemplate.dotdot_reclen = dirblksiz - dirtemplate.dot_reclen;
1409 	dirtemplate.dot_ino = ufs_rw32(ip->i_number, UFS_MPNEEDSWAP(ump));
1410 	dirtemplate.dotdot_ino = ufs_rw32(dp->i_number, UFS_MPNEEDSWAP(ump));
1411 	dirtemplate.dot_reclen = ufs_rw16(dirtemplate.dot_reclen,
1412 	    UFS_MPNEEDSWAP(ump));
1413 	dirtemplate.dotdot_reclen = ufs_rw16(dirtemplate.dotdot_reclen,
1414 	    UFS_MPNEEDSWAP(ump));
1415 	if (ump->um_maxsymlinklen <= 0) {
1416 #if BYTE_ORDER == LITTLE_ENDIAN
1417 		if (UFS_MPNEEDSWAP(ump) == 0)
1418 #else
1419 		if (UFS_MPNEEDSWAP(ump) != 0)
1420 #endif
1421 		{
1422 			dirtemplate.dot_type = dirtemplate.dot_namlen;
1423 			dirtemplate.dotdot_type = dirtemplate.dotdot_namlen;
1424 			dirtemplate.dot_namlen = dirtemplate.dotdot_namlen = 0;
1425 		} else
1426 			dirtemplate.dot_type = dirtemplate.dotdot_type = 0;
1427 	}
1428 	if ((error = UFS_BALLOC(tvp, (off_t)0, dirblksiz, cnp->cn_cred,
1429 	    B_CLRBUF, &bp)) != 0)
1430 		goto bad;
1431 	ip->i_size = dirblksiz;
1432 	DIP_ASSIGN(ip, size, dirblksiz);
1433 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
1434 	uvm_vnp_setsize(tvp, ip->i_size);
1435 	memcpy((void *)bp->b_data, (void *)&dirtemplate, sizeof dirtemplate);
1436 
1437 	/*
1438 	 * Directory set up, now install it's entry in the parent directory.
1439 	 * We must write out the buffer containing the new directory body
1440 	 * before entering the new name in the parent.
1441 	 */
1442 	if ((error = VOP_BWRITE(bp)) != 0)
1443 		goto bad;
1444 	if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0) {
1445 		goto bad;
1446 	}
1447 	newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
1448 	ufs_makedirentry(ip, cnp, newdir);
1449 	error = ufs_direnter(dvp, tvp, newdir, cnp, bp);
1450 	pool_cache_put(ufs_direct_cache, newdir);
1451  bad:
1452 	if (error == 0) {
1453 		VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1454 		UFS_WAPBL_END(dvp->v_mount);
1455 	} else {
1456 		dp->i_nlink--;
1457 		DIP_ASSIGN(dp, nlink, dp->i_nlink);
1458 		dp->i_flag |= IN_CHANGE;
1459 		UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
1460 		/*
1461 		 * No need to do an explicit UFS_TRUNCATE here, vrele will
1462 		 * do this for us because we set the link count to 0.
1463 		 */
1464 		ip->i_nlink = 0;
1465 		DIP_ASSIGN(ip, nlink, 0);
1466 		ip->i_flag |= IN_CHANGE;
1467 		/* If IN_ADIROP, account for it */
1468 		UFS_UNMARK_VNODE(tvp);
1469 		UFS_WAPBL_UPDATE(tvp, NULL, NULL, UPDATE_DIROP);
1470 		UFS_WAPBL_END(dvp->v_mount);
1471 		vput(tvp);
1472 	}
1473  out:
1474 	fstrans_done(dvp->v_mount);
1475 	vput(dvp);
1476 	return (error);
1477 }
1478 
1479 int
1480 ufs_rmdir(void *v)
1481 {
1482 	struct vop_rmdir_args /* {
1483 		struct vnode		*a_dvp;
1484 		struct vnode		*a_vp;
1485 		struct componentname	*a_cnp;
1486 	} */ *ap = v;
1487 	struct vnode		*vp, *dvp;
1488 	struct componentname	*cnp;
1489 	struct inode		*ip, *dp;
1490 	int			error;
1491 
1492 	vp = ap->a_vp;
1493 	dvp = ap->a_dvp;
1494 	cnp = ap->a_cnp;
1495 	ip = VTOI(vp);
1496 	dp = VTOI(dvp);
1497 	/*
1498 	 * No rmdir "." or of mounted directories please.
1499 	 */
1500 	if (dp == ip || vp->v_mountedhere != NULL) {
1501 		if (dp == ip)
1502 			vrele(vp);
1503 		else
1504 			vput(vp);
1505 		vput(vp);
1506 		return (EINVAL);
1507 	}
1508 
1509 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
1510 
1511 	/*
1512 	 * Do not remove a directory that is in the process of being renamed.
1513 	 * Verify that the directory is empty (and valid). (Rmdir ".." won't
1514 	 * be valid since ".." will contain a reference to the current
1515 	 * directory and thus be non-empty.)
1516 	 */
1517 	error = 0;
1518 	if (ip->i_flag & IN_RENAME) {
1519 		error = EINVAL;
1520 		goto out;
1521 	}
1522 	if (ip->i_nlink != 2 ||
1523 	    !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1524 		error = ENOTEMPTY;
1525 		goto out;
1526 	}
1527 	if ((dp->i_flags & APPEND) ||
1528 		(ip->i_flags & (IMMUTABLE | APPEND))) {
1529 		error = EPERM;
1530 		goto out;
1531 	}
1532 	error = UFS_WAPBL_BEGIN(dvp->v_mount);
1533 	if (error)
1534 		goto out;
1535 	/*
1536 	 * Delete reference to directory before purging
1537 	 * inode.  If we crash in between, the directory
1538 	 * will be reattached to lost+found,
1539 	 */
1540 	error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1541 	if (error) {
1542 		UFS_WAPBL_END(dvp->v_mount);
1543 		goto out;
1544 	}
1545 	VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1546 	cache_purge(dvp);
1547 	/*
1548 	 * Truncate inode.  The only stuff left in the directory is "." and
1549 	 * "..".  The "." reference is inconsequential since we're quashing
1550 	 * it.
1551 	 */
1552 	dp->i_nlink--;
1553 	DIP_ASSIGN(dp, nlink, dp->i_nlink);
1554 	dp->i_flag |= IN_CHANGE;
1555 	UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
1556 	ip->i_nlink--;
1557 	DIP_ASSIGN(ip, nlink, ip->i_nlink);
1558 	ip->i_flag |= IN_CHANGE;
1559 	error = UFS_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
1560 	cache_purge(vp);
1561 	/*
1562 	 * Unlock the log while we still have reference to unlinked
1563 	 * directory vp so that it will not get locked for recycling
1564 	 */
1565 	UFS_WAPBL_END(dvp->v_mount);
1566 #ifdef UFS_DIRHASH
1567 	if (ip->i_dirhash != NULL)
1568 		ufsdirhash_free(ip);
1569 #endif
1570  out:
1571 	VN_KNOTE(vp, NOTE_DELETE);
1572 	vput(dvp);
1573 	vput(vp);
1574 	fstrans_done(dvp->v_mount);
1575 	return (error);
1576 }
1577 
1578 /*
1579  * symlink -- make a symbolic link
1580  */
1581 int
1582 ufs_symlink(void *v)
1583 {
1584 	struct vop_symlink_args /* {
1585 		struct vnode		*a_dvp;
1586 		struct vnode		**a_vpp;
1587 		struct componentname	*a_cnp;
1588 		struct vattr		*a_vap;
1589 		char			*a_target;
1590 	} */ *ap = v;
1591 	struct vnode	*vp, **vpp;
1592 	struct inode	*ip;
1593 	int		len, error;
1594 
1595 	vpp = ap->a_vpp;
1596 	/*
1597 	 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
1598 	 * ufs_makeinode
1599 	 */
1600 	fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
1601 	error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1602 			      vpp, ap->a_cnp);
1603 	if (error)
1604 		goto out;
1605 	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1606 	vp = *vpp;
1607 	len = strlen(ap->a_target);
1608 	ip = VTOI(vp);
1609 	if (len < ip->i_ump->um_maxsymlinklen) {
1610 		memcpy((char *)SHORTLINK(ip), ap->a_target, len);
1611 		ip->i_size = len;
1612 		DIP_ASSIGN(ip, size, len);
1613 		uvm_vnp_setsize(vp, ip->i_size);
1614 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
1615 		UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
1616 	} else
1617 		error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1618 		    UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
1619 		    ap->a_cnp->cn_cred, NULL, NULL);
1620 	UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
1621 	if (error)
1622 		vput(vp);
1623 out:
1624 	fstrans_done(ap->a_dvp->v_mount);
1625 	return (error);
1626 }
1627 
1628 /*
1629  * Vnode op for reading directories.
1630  *
1631  * This routine handles converting from the on-disk directory format
1632  * "struct direct" to the in-memory format "struct dirent" as well as
1633  * byte swapping the entries if necessary.
1634  */
1635 int
1636 ufs_readdir(void *v)
1637 {
1638 	struct vop_readdir_args /* {
1639 		struct vnode	*a_vp;
1640 		struct uio	*a_uio;
1641 		kauth_cred_t	a_cred;
1642 		int		*a_eofflag;
1643 		off_t		**a_cookies;
1644 		int		*ncookies;
1645 	} */ *ap = v;
1646 	struct vnode	*vp = ap->a_vp;
1647 	struct direct	*cdp, *ecdp;
1648 	struct dirent	*ndp;
1649 	char		*cdbuf, *ndbuf, *endp;
1650 	struct uio	auio, *uio;
1651 	struct iovec	aiov;
1652 	int		error;
1653 	size_t		count, ccount, rcount;
1654 	off_t		off, *ccp;
1655 	off_t		startoff;
1656 	size_t		skipbytes;
1657 	struct ufsmount	*ump = VFSTOUFS(vp->v_mount);
1658 	int nswap = UFS_MPNEEDSWAP(ump);
1659 #if BYTE_ORDER == LITTLE_ENDIAN
1660 	int needswap = ump->um_maxsymlinklen <= 0 && nswap == 0;
1661 #else
1662 	int needswap = ump->um_maxsymlinklen <= 0 && nswap != 0;
1663 #endif
1664 	uio = ap->a_uio;
1665 	count = uio->uio_resid;
1666 	rcount = count - ((uio->uio_offset + count) & (ump->um_dirblksiz - 1));
1667 
1668 	if (rcount < _DIRENT_MINSIZE(cdp) || count < _DIRENT_MINSIZE(ndp))
1669 		return EINVAL;
1670 
1671 	startoff = uio->uio_offset & ~(ump->um_dirblksiz - 1);
1672 	skipbytes = uio->uio_offset - startoff;
1673 	rcount += skipbytes;
1674 
1675 	auio.uio_iov = &aiov;
1676 	auio.uio_iovcnt = 1;
1677 	auio.uio_offset = startoff;
1678 	auio.uio_resid = rcount;
1679 	UIO_SETUP_SYSSPACE(&auio);
1680 	auio.uio_rw = UIO_READ;
1681 	cdbuf = malloc(rcount, M_TEMP, M_WAITOK);
1682 	aiov.iov_base = cdbuf;
1683 	aiov.iov_len = rcount;
1684 	error = VOP_READ(vp, &auio, 0, ap->a_cred);
1685 	if (error != 0) {
1686 		free(cdbuf, M_TEMP);
1687 		return error;
1688 	}
1689 
1690 	rcount -= auio.uio_resid;
1691 
1692 	cdp = (struct direct *)(void *)cdbuf;
1693 	ecdp = (struct direct *)(void *)&cdbuf[rcount];
1694 
1695 	ndbuf = malloc(count, M_TEMP, M_WAITOK);
1696 	ndp = (struct dirent *)(void *)ndbuf;
1697 	endp = &ndbuf[count];
1698 
1699 	off = uio->uio_offset;
1700 	if (ap->a_cookies) {
1701 		ccount = rcount / _DIRENT_RECLEN(cdp, 1);
1702 		ccp = *(ap->a_cookies) = malloc(ccount * sizeof(*ccp),
1703 		    M_TEMP, M_WAITOK);
1704 	} else {
1705 		/* XXX: GCC */
1706 		ccount = 0;
1707 		ccp = NULL;
1708 	}
1709 
1710 	while (cdp < ecdp) {
1711 		cdp->d_reclen = ufs_rw16(cdp->d_reclen, nswap);
1712 		if (skipbytes > 0) {
1713 			if (cdp->d_reclen <= skipbytes) {
1714 				skipbytes -= cdp->d_reclen;
1715 				cdp = _DIRENT_NEXT(cdp);
1716 				continue;
1717 			}
1718 			/*
1719 			 * invalid cookie.
1720 			 */
1721 			error = EINVAL;
1722 			goto out;
1723 		}
1724 		if (cdp->d_reclen == 0) {
1725 			struct dirent *ondp = ndp;
1726 			ndp->d_reclen = _DIRENT_MINSIZE(ndp);
1727 			ndp = _DIRENT_NEXT(ndp);
1728 			ondp->d_reclen = 0;
1729 			cdp = ecdp;
1730 			break;
1731 		}
1732 		if (needswap) {
1733 			ndp->d_type = cdp->d_namlen;
1734 			ndp->d_namlen = cdp->d_type;
1735 		} else {
1736 			ndp->d_type = cdp->d_type;
1737 			ndp->d_namlen = cdp->d_namlen;
1738 		}
1739 		ndp->d_reclen = _DIRENT_RECLEN(ndp, ndp->d_namlen);
1740 		if ((char *)(void *)ndp + ndp->d_reclen +
1741 		    _DIRENT_MINSIZE(ndp) > endp)
1742 			break;
1743 		ndp->d_fileno = ufs_rw32(cdp->d_ino, nswap);
1744 		(void)memcpy(ndp->d_name, cdp->d_name, ndp->d_namlen);
1745 		memset(&ndp->d_name[ndp->d_namlen], 0,
1746 		    ndp->d_reclen - _DIRENT_NAMEOFF(ndp) - ndp->d_namlen);
1747 		off += cdp->d_reclen;
1748 		if (ap->a_cookies) {
1749 			KASSERT(ccp - *(ap->a_cookies) < ccount);
1750 			*(ccp++) = off;
1751 		}
1752 		ndp = _DIRENT_NEXT(ndp);
1753 		cdp = _DIRENT_NEXT(cdp);
1754 	}
1755 
1756 	count = ((char *)(void *)ndp - ndbuf);
1757 	error = uiomove(ndbuf, count, uio);
1758 out:
1759 	if (ap->a_cookies) {
1760 		if (error) {
1761 			free(*(ap->a_cookies), M_TEMP);
1762 			*(ap->a_cookies) = NULL;
1763 			*(ap->a_ncookies) = 0;
1764 		} else {
1765 			*ap->a_ncookies = ccp - *(ap->a_cookies);
1766 		}
1767 	}
1768 	uio->uio_offset = off;
1769 	free(ndbuf, M_TEMP);
1770 	free(cdbuf, M_TEMP);
1771 	*ap->a_eofflag = VTOI(vp)->i_size <= uio->uio_offset;
1772 	return error;
1773 }
1774 
1775 /*
1776  * Return target name of a symbolic link
1777  */
1778 int
1779 ufs_readlink(void *v)
1780 {
1781 	struct vop_readlink_args /* {
1782 		struct vnode	*a_vp;
1783 		struct uio	*a_uio;
1784 		kauth_cred_t	a_cred;
1785 	} */ *ap = v;
1786 	struct vnode	*vp = ap->a_vp;
1787 	struct inode	*ip = VTOI(vp);
1788 	struct ufsmount	*ump = VFSTOUFS(vp->v_mount);
1789 	int		isize;
1790 
1791 	isize = ip->i_size;
1792 	if (isize < ump->um_maxsymlinklen ||
1793 	    (ump->um_maxsymlinklen == 0 && DIP(ip, blocks) == 0)) {
1794 		uiomove((char *)SHORTLINK(ip), isize, ap->a_uio);
1795 		return (0);
1796 	}
1797 	return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1798 }
1799 
1800 /*
1801  * Calculate the logical to physical mapping if not done already,
1802  * then call the device strategy routine.
1803  */
1804 int
1805 ufs_strategy(void *v)
1806 {
1807 	struct vop_strategy_args /* {
1808 		struct vnode *a_vp;
1809 		struct buf *a_bp;
1810 	} */ *ap = v;
1811 	struct buf	*bp;
1812 	struct vnode	*vp;
1813 	struct inode	*ip;
1814 	struct mount	*mp;
1815 	int		error;
1816 
1817 	bp = ap->a_bp;
1818 	vp = ap->a_vp;
1819 	ip = VTOI(vp);
1820 	if (vp->v_type == VBLK || vp->v_type == VCHR)
1821 		panic("ufs_strategy: spec");
1822 	KASSERT(bp->b_bcount != 0);
1823 	if (bp->b_blkno == bp->b_lblkno) {
1824 		error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno,
1825 				 NULL);
1826 		if (error) {
1827 			bp->b_error = error;
1828 			biodone(bp);
1829 			return (error);
1830 		}
1831 		if (bp->b_blkno == -1) /* no valid data */
1832 			clrbuf(bp);
1833 	}
1834 	if (bp->b_blkno < 0) { /* block is not on disk */
1835 		biodone(bp);
1836 		return (0);
1837 	}
1838 	vp = ip->i_devvp;
1839 
1840 	error = VOP_STRATEGY(vp, bp);
1841 	if (error)
1842 		return error;
1843 
1844 	if (!BUF_ISREAD(bp))
1845 		return 0;
1846 
1847 	mp = wapbl_vptomp(vp);
1848 	if (mp == NULL || mp->mnt_wapbl_replay == NULL ||
1849 	    !WAPBL_REPLAY_ISOPEN(mp) ||
1850 	    !WAPBL_REPLAY_CAN_READ(mp, bp->b_blkno, bp->b_bcount))
1851 		return 0;
1852 
1853 	error = biowait(bp);
1854 	if (error)
1855 		return error;
1856 
1857 	error = WAPBL_REPLAY_READ(mp, bp->b_data, bp->b_blkno, bp->b_bcount);
1858 	if (error) {
1859 		mutex_enter(&bufcache_lock);
1860 		SET(bp->b_cflags, BC_INVAL);
1861 		mutex_exit(&bufcache_lock);
1862 	}
1863 	return error;
1864 }
1865 
1866 /*
1867  * Print out the contents of an inode.
1868  */
1869 int
1870 ufs_print(void *v)
1871 {
1872 	struct vop_print_args /* {
1873 		struct vnode	*a_vp;
1874 	} */ *ap = v;
1875 	struct vnode	*vp;
1876 	struct inode	*ip;
1877 
1878 	vp = ap->a_vp;
1879 	ip = VTOI(vp);
1880 	printf("tag VT_UFS, ino %llu, on dev %llu, %llu",
1881 	    (unsigned long long)ip->i_number,
1882 	    (unsigned long long)major(ip->i_dev),
1883 	    (unsigned long long)minor(ip->i_dev));
1884 	printf(" flags 0x%x, nlink %d\n",
1885 	    ip->i_flag, ip->i_nlink);
1886 	printf("\tmode 0%o, owner %d, group %d, size %qd",
1887 	    ip->i_mode, ip->i_uid, ip->i_gid,
1888 	    (long long)ip->i_size);
1889 	if (vp->v_type == VFIFO)
1890 		VOCALL(fifo_vnodeop_p, VOFFSET(vop_print), v);
1891 	printf("\n");
1892 	return (0);
1893 }
1894 
1895 /*
1896  * Read wrapper for special devices.
1897  */
1898 int
1899 ufsspec_read(void *v)
1900 {
1901 	struct vop_read_args /* {
1902 		struct vnode	*a_vp;
1903 		struct uio	*a_uio;
1904 		int		a_ioflag;
1905 		kauth_cred_t	a_cred;
1906 	} */ *ap = v;
1907 
1908 	/*
1909 	 * Set access flag.
1910 	 */
1911 	if ((ap->a_vp->v_mount->mnt_flag & MNT_NODEVMTIME) == 0)
1912 		VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1913 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_read), ap));
1914 }
1915 
1916 /*
1917  * Write wrapper for special devices.
1918  */
1919 int
1920 ufsspec_write(void *v)
1921 {
1922 	struct vop_write_args /* {
1923 		struct vnode	*a_vp;
1924 		struct uio	*a_uio;
1925 		int		a_ioflag;
1926 		kauth_cred_t	a_cred;
1927 	} */ *ap = v;
1928 
1929 	/*
1930 	 * Set update and change flags.
1931 	 */
1932 	if ((ap->a_vp->v_mount->mnt_flag & MNT_NODEVMTIME) == 0)
1933 		VTOI(ap->a_vp)->i_flag |= IN_MODIFY;
1934 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_write), ap));
1935 }
1936 
1937 /*
1938  * Close wrapper for special devices.
1939  *
1940  * Update the times on the inode then do device close.
1941  */
1942 int
1943 ufsspec_close(void *v)
1944 {
1945 	struct vop_close_args /* {
1946 		struct vnode	*a_vp;
1947 		int		a_fflag;
1948 		kauth_cred_t	a_cred;
1949 	} */ *ap = v;
1950 	struct vnode	*vp;
1951 	struct inode	*ip;
1952 
1953 	vp = ap->a_vp;
1954 	ip = VTOI(vp);
1955 	if (vp->v_usecount > 1)
1956 		UFS_ITIMES(vp, NULL, NULL, NULL);
1957 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_close), ap));
1958 }
1959 
1960 /*
1961  * Read wrapper for fifo's
1962  */
1963 int
1964 ufsfifo_read(void *v)
1965 {
1966 	struct vop_read_args /* {
1967 		struct vnode	*a_vp;
1968 		struct uio	*a_uio;
1969 		int		a_ioflag;
1970 		kauth_cred_t	a_cred;
1971 	} */ *ap = v;
1972 
1973 	/*
1974 	 * Set access flag.
1975 	 */
1976 	VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1977 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_read), ap));
1978 }
1979 
1980 /*
1981  * Write wrapper for fifo's.
1982  */
1983 int
1984 ufsfifo_write(void *v)
1985 {
1986 	struct vop_write_args /* {
1987 		struct vnode	*a_vp;
1988 		struct uio	*a_uio;
1989 		int		a_ioflag;
1990 		kauth_cred_t	a_cred;
1991 	} */ *ap = v;
1992 
1993 	/*
1994 	 * Set update and change flags.
1995 	 */
1996 	VTOI(ap->a_vp)->i_flag |= IN_MODIFY;
1997 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_write), ap));
1998 }
1999 
2000 /*
2001  * Close wrapper for fifo's.
2002  *
2003  * Update the times on the inode then do device close.
2004  */
2005 int
2006 ufsfifo_close(void *v)
2007 {
2008 	struct vop_close_args /* {
2009 		struct vnode	*a_vp;
2010 		int		a_fflag;
2011 		kauth_cred_t	a_cred;
2012 	} */ *ap = v;
2013 	struct vnode	*vp;
2014 	struct inode	*ip;
2015 
2016 	vp = ap->a_vp;
2017 	ip = VTOI(vp);
2018 	if (ap->a_vp->v_usecount > 1)
2019 		UFS_ITIMES(vp, NULL, NULL, NULL);
2020 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_close), ap));
2021 }
2022 
2023 /*
2024  * Return POSIX pathconf information applicable to ufs filesystems.
2025  */
2026 int
2027 ufs_pathconf(void *v)
2028 {
2029 	struct vop_pathconf_args /* {
2030 		struct vnode	*a_vp;
2031 		int		a_name;
2032 		register_t	*a_retval;
2033 	} */ *ap = v;
2034 
2035 	switch (ap->a_name) {
2036 	case _PC_LINK_MAX:
2037 		*ap->a_retval = LINK_MAX;
2038 		return (0);
2039 	case _PC_NAME_MAX:
2040 		*ap->a_retval = NAME_MAX;
2041 		return (0);
2042 	case _PC_PATH_MAX:
2043 		*ap->a_retval = PATH_MAX;
2044 		return (0);
2045 	case _PC_PIPE_BUF:
2046 		*ap->a_retval = PIPE_BUF;
2047 		return (0);
2048 	case _PC_CHOWN_RESTRICTED:
2049 		*ap->a_retval = 1;
2050 		return (0);
2051 	case _PC_NO_TRUNC:
2052 		*ap->a_retval = 1;
2053 		return (0);
2054 	case _PC_SYNC_IO:
2055 		*ap->a_retval = 1;
2056 		return (0);
2057 	case _PC_FILESIZEBITS:
2058 		*ap->a_retval = 42;
2059 		return (0);
2060 	case _PC_SYMLINK_MAX:
2061 		*ap->a_retval = MAXPATHLEN;
2062 		return (0);
2063 	case _PC_2_SYMLINKS:
2064 		*ap->a_retval = 1;
2065 		return (0);
2066 	default:
2067 		return (EINVAL);
2068 	}
2069 	/* NOTREACHED */
2070 }
2071 
2072 /*
2073  * Advisory record locking support
2074  */
2075 int
2076 ufs_advlock(void *v)
2077 {
2078 	struct vop_advlock_args /* {
2079 		struct vnode	*a_vp;
2080 		void *		a_id;
2081 		int		a_op;
2082 		struct flock	*a_fl;
2083 		int		a_flags;
2084 	} */ *ap = v;
2085 	struct inode *ip;
2086 
2087 	ip = VTOI(ap->a_vp);
2088 	return lf_advlock(ap, &ip->i_lockf, ip->i_size);
2089 }
2090 
2091 /*
2092  * Initialize the vnode associated with a new inode, handle aliased
2093  * vnodes.
2094  */
2095 void
2096 ufs_vinit(struct mount *mntp, int (**specops)(void *), int (**fifoops)(void *),
2097 	struct vnode **vpp)
2098 {
2099 	struct timeval	tv;
2100 	struct inode	*ip;
2101 	struct vnode	*vp;
2102 	dev_t		rdev;
2103 	struct ufsmount	*ump;
2104 
2105 	vp = *vpp;
2106 	ip = VTOI(vp);
2107 	switch(vp->v_type = IFTOVT(ip->i_mode)) {
2108 	case VCHR:
2109 	case VBLK:
2110 		vp->v_op = specops;
2111 		ump = ip->i_ump;
2112 		if (ump->um_fstype == UFS1)
2113 			rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev,
2114 			    UFS_MPNEEDSWAP(ump));
2115 		else
2116 			rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev,
2117 			    UFS_MPNEEDSWAP(ump));
2118 		spec_node_init(vp, rdev);
2119 		break;
2120 	case VFIFO:
2121 		vp->v_op = fifoops;
2122 		break;
2123 	case VNON:
2124 	case VBAD:
2125 	case VSOCK:
2126 	case VLNK:
2127 	case VDIR:
2128 	case VREG:
2129 		break;
2130 	}
2131 	if (ip->i_number == ROOTINO)
2132                 vp->v_vflag |= VV_ROOT;
2133 	/*
2134 	 * Initialize modrev times
2135 	 */
2136 	getmicrouptime(&tv);
2137 	ip->i_modrev = (uint64_t)(uint)tv.tv_sec << 32
2138 			| tv.tv_usec * 4294u;
2139 	*vpp = vp;
2140 }
2141 
2142 /*
2143  * Allocate a new inode.
2144  */
2145 int
2146 ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
2147 	struct componentname *cnp)
2148 {
2149 	struct inode	*ip, *pdir;
2150 	struct direct	*newdir;
2151 	struct vnode	*tvp;
2152 	int		error, ismember = 0;
2153 
2154 	UFS_WAPBL_JUNLOCK_ASSERT(dvp->v_mount);
2155 
2156 	pdir = VTOI(dvp);
2157 
2158 	if ((mode & IFMT) == 0)
2159 		mode |= IFREG;
2160 
2161 	if ((error = UFS_VALLOC(dvp, mode, cnp->cn_cred, vpp)) != 0) {
2162 		vput(dvp);
2163 		return (error);
2164 	}
2165 	tvp = *vpp;
2166 	ip = VTOI(tvp);
2167 	ip->i_gid = pdir->i_gid;
2168 	DIP_ASSIGN(ip, gid, ip->i_gid);
2169 	ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
2170 	DIP_ASSIGN(ip, uid, ip->i_uid);
2171 	error = UFS_WAPBL_BEGIN1(dvp->v_mount, dvp);
2172 	if (error) {
2173 		/*
2174 		 * Note, we can't VOP_VFREE(tvp) here like we should
2175 		 * because we can't write to the disk.  Instead, we leave
2176 		 * the vnode dangling from the journal.
2177 		 */
2178 		vput(tvp);
2179 		vput(dvp);
2180 		return (error);
2181 	}
2182 #if defined(QUOTA) || defined(QUOTA2)
2183 	if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2184 		UFS_VFREE(tvp, ip->i_number, mode);
2185 		UFS_WAPBL_END1(dvp->v_mount, dvp);
2186 		vput(tvp);
2187 		vput(dvp);
2188 		return (error);
2189 	}
2190 #endif
2191 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2192 	ip->i_mode = mode;
2193 	DIP_ASSIGN(ip, mode, mode);
2194 	tvp->v_type = IFTOVT(mode);	/* Rest init'd in getnewvnode(). */
2195 	ip->i_nlink = 1;
2196 	DIP_ASSIGN(ip, nlink, 1);
2197 	if ((ip->i_mode & ISGID) && (kauth_cred_ismember_gid(cnp->cn_cred,
2198 	    ip->i_gid, &ismember) != 0 || !ismember) &&
2199 	    kauth_authorize_generic(cnp->cn_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2200 		ip->i_mode &= ~ISGID;
2201 		DIP_ASSIGN(ip, mode, ip->i_mode);
2202 	}
2203 
2204 	if (cnp->cn_flags & ISWHITEOUT) {
2205 		ip->i_flags |= UF_OPAQUE;
2206 		DIP_ASSIGN(ip, flags, ip->i_flags);
2207 	}
2208 
2209 	/*
2210 	 * Make sure inode goes to disk before directory entry.
2211 	 */
2212 	if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0)
2213 		goto bad;
2214 	newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
2215 	ufs_makedirentry(ip, cnp, newdir);
2216 	error = ufs_direnter(dvp, tvp, newdir, cnp, NULL);
2217 	pool_cache_put(ufs_direct_cache, newdir);
2218 	if (error)
2219 		goto bad;
2220 	vput(dvp);
2221 	*vpp = tvp;
2222 	return (0);
2223 
2224  bad:
2225 	/*
2226 	 * Write error occurred trying to update the inode
2227 	 * or the directory so must deallocate the inode.
2228 	 */
2229 	ip->i_nlink = 0;
2230 	DIP_ASSIGN(ip, nlink, 0);
2231 	ip->i_flag |= IN_CHANGE;
2232 	/* If IN_ADIROP, account for it */
2233 	UFS_UNMARK_VNODE(tvp);
2234 	UFS_WAPBL_UPDATE(tvp, NULL, NULL, 0);
2235 	tvp->v_type = VNON;		/* explodes later if VBLK */
2236 	UFS_WAPBL_END1(dvp->v_mount, dvp);
2237 	vput(tvp);
2238 	vput(dvp);
2239 	return (error);
2240 }
2241 
2242 /*
2243  * Allocate len bytes at offset off.
2244  */
2245 int
2246 ufs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags,
2247     kauth_cred_t cred)
2248 {
2249         struct inode *ip = VTOI(vp);
2250         int error, delta, bshift, bsize;
2251         UVMHIST_FUNC("ufs_gop_alloc"); UVMHIST_CALLED(ubchist);
2252 
2253         error = 0;
2254         bshift = vp->v_mount->mnt_fs_bshift;
2255         bsize = 1 << bshift;
2256 
2257         delta = off & (bsize - 1);
2258         off -= delta;
2259         len += delta;
2260 
2261         while (len > 0) {
2262                 bsize = MIN(bsize, len);
2263 
2264                 error = UFS_BALLOC(vp, off, bsize, cred, flags, NULL);
2265                 if (error) {
2266                         goto out;
2267                 }
2268 
2269                 /*
2270                  * increase file size now, UFS_BALLOC() requires that
2271                  * EOF be up-to-date before each call.
2272                  */
2273 
2274                 if (ip->i_size < off + bsize) {
2275                         UVMHIST_LOG(ubchist, "vp %p old 0x%x new 0x%x",
2276                             vp, ip->i_size, off + bsize, 0);
2277                         ip->i_size = off + bsize;
2278 			DIP_ASSIGN(ip, size, ip->i_size);
2279                 }
2280 
2281                 off += bsize;
2282                 len -= bsize;
2283         }
2284 
2285 out:
2286 	UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
2287 	return error;
2288 }
2289 
2290 void
2291 ufs_gop_markupdate(struct vnode *vp, int flags)
2292 {
2293 	u_int32_t mask = 0;
2294 
2295 	if ((flags & GOP_UPDATE_ACCESSED) != 0) {
2296 		mask = IN_ACCESS;
2297 	}
2298 	if ((flags & GOP_UPDATE_MODIFIED) != 0) {
2299 		if (vp->v_type == VREG) {
2300 			mask |= IN_CHANGE | IN_UPDATE;
2301 		} else {
2302 			mask |= IN_MODIFY;
2303 		}
2304 	}
2305 	if (mask) {
2306 		struct inode *ip = VTOI(vp);
2307 
2308 		ip->i_flag |= mask;
2309 	}
2310 }
2311