xref: /netbsd-src/sys/ufs/ufs/ufs_vnops.c (revision 6fc217346bb51c463d3a5a2a7883cb56515cd6d7)
1 /*	$NetBSD: ufs_vnops.c,v 1.180 2009/10/14 09:40:27 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 1.180 2009/10/14 09:40:27 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 	VOP_UNLOCK(*vpp, 0);
215 	(*vpp)->v_type = VNON;
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 	if (vp->v_usecount > 1)
271 		UFS_ITIMES(vp, NULL, NULL, NULL);
272 	return (0);
273 }
274 
275 static int
276 ufs_check_possible(struct vnode *vp, struct inode *ip, mode_t mode)
277 {
278 #ifdef QUOTA
279 	int error;
280 #endif /* QUOTA */
281 
282 	/*
283 	 * Disallow write attempts on read-only file systems;
284 	 * unless the file is a socket, fifo, or a block or
285 	 * character device resident on the file system.
286 	 */
287 	if (mode & VWRITE) {
288 		switch (vp->v_type) {
289 		case VDIR:
290 		case VLNK:
291 		case VREG:
292 			if (vp->v_mount->mnt_flag & MNT_RDONLY)
293 				return (EROFS);
294 #ifdef QUOTA
295 			fstrans_start(vp->v_mount, FSTRANS_SHARED);
296 			error = getinoquota(ip);
297 			fstrans_done(vp->v_mount);
298 			if (error != 0)
299 				return error;
300 #endif
301 			break;
302 		case VBAD:
303 		case VBLK:
304 		case VCHR:
305 		case VSOCK:
306 		case VFIFO:
307 		case VNON:
308 		default:
309 			break;
310 		}
311 	}
312 
313 	/* If it is a snapshot, nobody gets access to it. */
314 	if ((ip->i_flags & SF_SNAPSHOT))
315 		return (EPERM);
316 	/* If immutable bit set, nobody gets to write it. */
317 	if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
318 		return (EPERM);
319 
320 	return 0;
321 }
322 
323 static int
324 ufs_check_permitted(struct vnode *vp, struct inode *ip, mode_t mode,
325     kauth_cred_t cred)
326 {
327 
328 	return genfs_can_access(vp->v_type, ip->i_mode & ALLPERMS, ip->i_uid,
329 	    ip->i_gid, mode, cred);
330 }
331 
332 int
333 ufs_access(void *v)
334 {
335 	struct vop_access_args /* {
336 		struct vnode	*a_vp;
337 		int		a_mode;
338 		kauth_cred_t	a_cred;
339 	} */ *ap = v;
340 	struct vnode	*vp;
341 	struct inode	*ip;
342 	mode_t		mode;
343 	int		error;
344 
345 	vp = ap->a_vp;
346 	ip = VTOI(vp);
347 	mode = ap->a_mode;
348 
349 	error = ufs_check_possible(vp, ip, mode);
350 	if (error)
351 		return error;
352 
353 	error = ufs_check_permitted(vp, ip, mode, ap->a_cred);
354 
355 	return error;
356 }
357 
358 /* ARGSUSED */
359 int
360 ufs_getattr(void *v)
361 {
362 	struct vop_getattr_args /* {
363 		struct vnode	*a_vp;
364 		struct vattr	*a_vap;
365 		kauth_cred_t	a_cred;
366 	} */ *ap = v;
367 	struct vnode	*vp;
368 	struct inode	*ip;
369 	struct vattr	*vap;
370 
371 	vp = ap->a_vp;
372 	ip = VTOI(vp);
373 	vap = ap->a_vap;
374 	UFS_ITIMES(vp, NULL, NULL, NULL);
375 
376 	/*
377 	 * Copy from inode table
378 	 */
379 	vap->va_fsid = ip->i_dev;
380 	vap->va_fileid = ip->i_number;
381 	vap->va_mode = ip->i_mode & ALLPERMS;
382 	vap->va_nlink = ip->i_nlink;
383 	vap->va_uid = ip->i_uid;
384 	vap->va_gid = ip->i_gid;
385 	vap->va_size = vp->v_size;
386 	if (ip->i_ump->um_fstype == UFS1) {
387 		vap->va_rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev,
388 		    UFS_MPNEEDSWAP(ip->i_ump));
389 		vap->va_atime.tv_sec = ip->i_ffs1_atime;
390 		vap->va_atime.tv_nsec = ip->i_ffs1_atimensec;
391 		vap->va_mtime.tv_sec = ip->i_ffs1_mtime;
392 		vap->va_mtime.tv_nsec = ip->i_ffs1_mtimensec;
393 		vap->va_ctime.tv_sec = ip->i_ffs1_ctime;
394 		vap->va_ctime.tv_nsec = ip->i_ffs1_ctimensec;
395 		vap->va_birthtime.tv_sec = 0;
396 		vap->va_birthtime.tv_nsec = 0;
397 		vap->va_bytes = dbtob((u_quad_t)ip->i_ffs1_blocks);
398 	} else {
399 		vap->va_rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev,
400 		    UFS_MPNEEDSWAP(ip->i_ump));
401 		vap->va_atime.tv_sec = ip->i_ffs2_atime;
402 		vap->va_atime.tv_nsec = ip->i_ffs2_atimensec;
403 		vap->va_mtime.tv_sec = ip->i_ffs2_mtime;
404 		vap->va_mtime.tv_nsec = ip->i_ffs2_mtimensec;
405 		vap->va_ctime.tv_sec = ip->i_ffs2_ctime;
406 		vap->va_ctime.tv_nsec = ip->i_ffs2_ctimensec;
407 		vap->va_birthtime.tv_sec = ip->i_ffs2_birthtime;
408 		vap->va_birthtime.tv_nsec = ip->i_ffs2_birthnsec;
409 		vap->va_bytes = dbtob(ip->i_ffs2_blocks);
410 	}
411 	vap->va_gen = ip->i_gen;
412 	vap->va_flags = ip->i_flags;
413 
414 	/* this doesn't belong here */
415 	if (vp->v_type == VBLK)
416 		vap->va_blocksize = BLKDEV_IOSIZE;
417 	else if (vp->v_type == VCHR)
418 		vap->va_blocksize = MAXBSIZE;
419 	else
420 		vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
421 	vap->va_type = vp->v_type;
422 	vap->va_filerev = ip->i_modrev;
423 	return (0);
424 }
425 
426 /*
427  * Set attribute vnode op. called from several syscalls
428  */
429 int
430 ufs_setattr(void *v)
431 {
432 	struct vop_setattr_args /* {
433 		struct vnode	*a_vp;
434 		struct vattr	*a_vap;
435 		kauth_cred_t	a_cred;
436 	} */ *ap = v;
437 	struct vattr	*vap;
438 	struct vnode	*vp;
439 	struct inode	*ip;
440 	kauth_cred_t	cred;
441 	struct lwp	*l;
442 	int		error;
443 
444 	vap = ap->a_vap;
445 	vp = ap->a_vp;
446 	ip = VTOI(vp);
447 	cred = ap->a_cred;
448 	l = curlwp;
449 
450 	/*
451 	 * Check for unsettable attributes.
452 	 */
453 	if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
454 	    (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
455 	    (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
456 	    ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
457 		return (EINVAL);
458 	}
459 
460 	fstrans_start(vp->v_mount, FSTRANS_SHARED);
461 
462 	if (vap->va_flags != VNOVAL) {
463 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
464 			error = EROFS;
465 			goto out;
466 		}
467 		if (kauth_cred_geteuid(cred) != ip->i_uid &&
468 		    (error = kauth_authorize_generic(cred,
469 		    KAUTH_GENERIC_ISSUSER, NULL)))
470 			goto out;
471 		if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
472 		    NULL) == 0) {
473 			if ((ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) &&
474 			    kauth_authorize_system(l->l_cred,
475 			     KAUTH_SYSTEM_CHSYSFLAGS, 0, NULL, NULL, NULL)) {
476 				error = EPERM;
477 				goto out;
478 			}
479 			/* Snapshot flag cannot be set or cleared */
480 			if ((vap->va_flags & SF_SNAPSHOT) !=
481 			    (ip->i_flags & SF_SNAPSHOT)) {
482 				error = EPERM;
483 				goto out;
484 			}
485 			error = UFS_WAPBL_BEGIN(vp->v_mount);
486 			if (error)
487 				goto out;
488 			ip->i_flags = vap->va_flags;
489 			DIP_ASSIGN(ip, flags, ip->i_flags);
490 		} else {
491 			if ((ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) ||
492 			    (vap->va_flags & UF_SETTABLE) != vap->va_flags) {
493 				error = EPERM;
494 				goto out;
495 			}
496 			if ((ip->i_flags & SF_SETTABLE) !=
497 			    (vap->va_flags & SF_SETTABLE)) {
498 				error = EPERM;
499 				goto out;
500 			}
501 			error = UFS_WAPBL_BEGIN(vp->v_mount);
502 			if (error)
503 				goto out;
504 			ip->i_flags &= SF_SETTABLE;
505 			ip->i_flags |= (vap->va_flags & UF_SETTABLE);
506 			DIP_ASSIGN(ip, flags, ip->i_flags);
507 		}
508 		ip->i_flag |= IN_CHANGE;
509 		UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
510 		UFS_WAPBL_END(vp->v_mount);
511 		if (vap->va_flags & (IMMUTABLE | APPEND)) {
512 			error = 0;
513 			goto out;
514 		}
515 	}
516 	if (ip->i_flags & (IMMUTABLE | APPEND)) {
517 		error = EPERM;
518 		goto out;
519 	}
520 	/*
521 	 * Go through the fields and update iff not VNOVAL.
522 	 */
523 	if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
524 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
525 			error = EROFS;
526 			goto out;
527 		}
528 		error = UFS_WAPBL_BEGIN(vp->v_mount);
529 		if (error)
530 			goto out;
531 		error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, l);
532 		UFS_WAPBL_END(vp->v_mount);
533 		if (error)
534 			goto out;
535 	}
536 	if (vap->va_size != VNOVAL) {
537 		/*
538 		 * Disallow write attempts on read-only file systems;
539 		 * unless the file is a socket, fifo, or a block or
540 		 * character device resident on the file system.
541 		 */
542 		switch (vp->v_type) {
543 		case VDIR:
544 			error = EISDIR;
545 			goto out;
546 		case VCHR:
547 		case VBLK:
548 		case VFIFO:
549 			break;
550 		case VREG:
551 			if (vp->v_mount->mnt_flag & MNT_RDONLY) {
552 				error = EROFS;
553 				goto out;
554 			}
555 			if ((ip->i_flags & SF_SNAPSHOT) != 0) {
556 				error = EPERM;
557 				goto out;
558 			}
559 			error = UFS_WAPBL_BEGIN(vp->v_mount);
560 			if (error)
561 				goto out;
562 			/*
563 			 * When journaling, only truncate one indirect block
564 			 * at a time.
565 			 */
566 			if (vp->v_mount->mnt_wapbl) {
567 				uint64_t incr = MNINDIR(ip->i_ump) <<
568 				    vp->v_mount->mnt_fs_bshift; /* Power of 2 */
569 				uint64_t base = NDADDR <<
570 				    vp->v_mount->mnt_fs_bshift;
571 				while (!error && ip->i_size > base + incr &&
572 				    ip->i_size > vap->va_size + incr) {
573 					/*
574 					 * round down to next full indirect
575 					 * block boundary.
576 					 */
577 					uint64_t nsize = base +
578 					    ((ip->i_size - base - 1) &
579 					    ~(incr - 1));
580 					error = UFS_TRUNCATE(vp, nsize, 0,
581 					    cred);
582 					if (error == 0) {
583 						UFS_WAPBL_END(vp->v_mount);
584 						error =
585 						   UFS_WAPBL_BEGIN(vp->v_mount);
586 					}
587 				}
588 			}
589 			if (!error)
590 				error = UFS_TRUNCATE(vp, vap->va_size, 0, cred);
591 			UFS_WAPBL_END(vp->v_mount);
592 			if (error)
593 				goto out;
594 			break;
595 		default:
596 			error = EOPNOTSUPP;
597 			goto out;
598 		}
599 	}
600 	ip = VTOI(vp);
601 	if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL ||
602 	    vap->va_birthtime.tv_sec != VNOVAL) {
603 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
604 			error = EROFS;
605 			goto out;
606 		}
607 		if ((ip->i_flags & SF_SNAPSHOT) != 0) {
608 			error = EPERM;
609 			goto out;
610 		}
611 		error = genfs_can_chtimes(vp, vap->va_vaflags, ip->i_uid, cred);
612 		if (error)
613 			goto out;
614 		error = UFS_WAPBL_BEGIN(vp->v_mount);
615 		if (error)
616 			goto out;
617 		if (vap->va_atime.tv_sec != VNOVAL)
618 			if (!(vp->v_mount->mnt_flag & MNT_NOATIME))
619 				ip->i_flag |= IN_ACCESS;
620 		if (vap->va_mtime.tv_sec != VNOVAL)
621 			ip->i_flag |= IN_CHANGE | IN_UPDATE;
622 		if (vap->va_birthtime.tv_sec != VNOVAL &&
623 		    ip->i_ump->um_fstype == UFS2) {
624 			ip->i_ffs2_birthtime = vap->va_birthtime.tv_sec;
625 			ip->i_ffs2_birthnsec = vap->va_birthtime.tv_nsec;
626 		}
627 		error = UFS_UPDATE(vp, &vap->va_atime, &vap->va_mtime, 0);
628 		UFS_WAPBL_END(vp->v_mount);
629 		if (error)
630 			goto out;
631 	}
632 	error = 0;
633 	if (vap->va_mode != (mode_t)VNOVAL) {
634 		if (vp->v_mount->mnt_flag & MNT_RDONLY) {
635 			error = EROFS;
636 			goto out;
637 		}
638 		if ((ip->i_flags & SF_SNAPSHOT) != 0 &&
639 		    (vap->va_mode & (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP |
640 		     S_IXOTH | S_IWOTH))) {
641 			error = EPERM;
642 			goto out;
643 		}
644 		error = UFS_WAPBL_BEGIN(vp->v_mount);
645 		if (error)
646 			goto out;
647 		error = ufs_chmod(vp, (int)vap->va_mode, cred, l);
648 		UFS_WAPBL_END(vp->v_mount);
649 	}
650 	VN_KNOTE(vp, NOTE_ATTRIB);
651 out:
652 	fstrans_done(vp->v_mount);
653 	return (error);
654 }
655 
656 /*
657  * Change the mode on a file.
658  * Inode must be locked before calling.
659  */
660 static int
661 ufs_chmod(struct vnode *vp, int mode, kauth_cred_t cred, struct lwp *l)
662 {
663 	struct inode	*ip;
664 	int		error;
665 
666 	UFS_WAPBL_JLOCK_ASSERT(vp->v_mount);
667 
668 	ip = VTOI(vp);
669 
670 	error = genfs_can_chmod(vp, cred, ip->i_uid, ip->i_gid, mode);
671 	if (error)
672 		return (error);
673 
674 	ip->i_mode &= ~ALLPERMS;
675 	ip->i_mode |= (mode & ALLPERMS);
676 	ip->i_flag |= IN_CHANGE;
677 	DIP_ASSIGN(ip, mode, ip->i_mode);
678 	UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
679 	return (0);
680 }
681 
682 /*
683  * Perform chown operation on inode ip;
684  * inode must be locked prior to call.
685  */
686 static int
687 ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, kauth_cred_t cred,
688     	struct lwp *l)
689 {
690 	struct inode	*ip;
691 	int		error = 0;
692 #ifdef QUOTA
693 	uid_t		ouid;
694 	gid_t		ogid;
695 	int64_t		change;
696 #endif
697 	ip = VTOI(vp);
698 	error = 0;
699 
700 	if (uid == (uid_t)VNOVAL)
701 		uid = ip->i_uid;
702 	if (gid == (gid_t)VNOVAL)
703 		gid = ip->i_gid;
704 
705 	error = genfs_can_chown(vp, cred, ip->i_uid, ip->i_gid, uid, gid);
706 	if (error)
707 		return (error);
708 
709 #ifdef QUOTA
710 	ogid = ip->i_gid;
711 	ouid = ip->i_uid;
712 	change = DIP(ip, blocks);
713 	(void) chkdq(ip, -change, cred, 0);
714 	(void) chkiq(ip, -1, cred, 0);
715 #endif
716 	ip->i_gid = gid;
717 	DIP_ASSIGN(ip, gid, gid);
718 	ip->i_uid = uid;
719 	DIP_ASSIGN(ip, uid, uid);
720 #ifdef QUOTA
721 	if ((error = chkdq(ip, change, cred, 0)) == 0) {
722 		if ((error = chkiq(ip, 1, cred, 0)) == 0)
723 			goto good;
724 		else
725 			(void) chkdq(ip, -change, cred, FORCE);
726 	}
727 	ip->i_gid = ogid;
728 	DIP_ASSIGN(ip, gid, ogid);
729 	ip->i_uid = ouid;
730 	DIP_ASSIGN(ip, uid, ouid);
731 	(void) chkdq(ip, change, cred, FORCE);
732 	(void) chkiq(ip, 1, cred, FORCE);
733 	return (error);
734  good:
735 #endif /* QUOTA */
736 	ip->i_flag |= IN_CHANGE;
737 	UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
738 	return (0);
739 }
740 
741 int
742 ufs_remove(void *v)
743 {
744 	struct vop_remove_args /* {
745 		struct vnode		*a_dvp;
746 		struct vnode		*a_vp;
747 		struct componentname	*a_cnp;
748 	} */ *ap = v;
749 	struct vnode	*vp, *dvp;
750 	struct inode	*ip;
751 	int		error;
752 
753 	vp = ap->a_vp;
754 	dvp = ap->a_dvp;
755 	ip = VTOI(vp);
756 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
757 	if (vp->v_type == VDIR || (ip->i_flags & (IMMUTABLE | APPEND)) ||
758 	    (VTOI(dvp)->i_flags & APPEND))
759 		error = EPERM;
760 	else {
761 		error = UFS_WAPBL_BEGIN(dvp->v_mount);
762 		if (error == 0) {
763 			error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
764 			UFS_WAPBL_END(dvp->v_mount);
765 		}
766 	}
767 	VN_KNOTE(vp, NOTE_DELETE);
768 	VN_KNOTE(dvp, NOTE_WRITE);
769 	if (dvp == vp)
770 		vrele(vp);
771 	else
772 		vput(vp);
773 	vput(dvp);
774 	fstrans_done(dvp->v_mount);
775 	return (error);
776 }
777 
778 /*
779  * link vnode call
780  */
781 int
782 ufs_link(void *v)
783 {
784 	struct vop_link_args /* {
785 		struct vnode *a_dvp;
786 		struct vnode *a_vp;
787 		struct componentname *a_cnp;
788 	} */ *ap = v;
789 	struct vnode		*vp, *dvp;
790 	struct componentname	*cnp;
791 	struct inode		*ip;
792 	struct direct		*newdir;
793 	int			error;
794 
795 	dvp = ap->a_dvp;
796 	vp = ap->a_vp;
797 	cnp = ap->a_cnp;
798 #ifdef DIAGNOSTIC
799 	if ((cnp->cn_flags & HASBUF) == 0)
800 		panic("ufs_link: no name");
801 #endif
802 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
803 	if (vp->v_type == VDIR) {
804 		VOP_ABORTOP(dvp, cnp);
805 		error = EPERM;
806 		goto out2;
807 	}
808 	if (dvp->v_mount != vp->v_mount) {
809 		VOP_ABORTOP(dvp, cnp);
810 		error = EXDEV;
811 		goto out2;
812 	}
813 	if (dvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE))) {
814 		VOP_ABORTOP(dvp, cnp);
815 		goto out2;
816 	}
817 	ip = VTOI(vp);
818 	if ((nlink_t)ip->i_nlink >= LINK_MAX) {
819 		VOP_ABORTOP(dvp, cnp);
820 		error = EMLINK;
821 		goto out1;
822 	}
823 	if (ip->i_flags & (IMMUTABLE | APPEND)) {
824 		VOP_ABORTOP(dvp, cnp);
825 		error = EPERM;
826 		goto out1;
827 	}
828 	error = UFS_WAPBL_BEGIN(vp->v_mount);
829 	if (error) {
830 		VOP_ABORTOP(dvp, cnp);
831 		goto out1;
832 	}
833 	ip->i_nlink++;
834 	DIP_ASSIGN(ip, nlink, ip->i_nlink);
835 	ip->i_flag |= IN_CHANGE;
836 	error = UFS_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
837 	if (!error) {
838 		newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
839 		ufs_makedirentry(ip, cnp, newdir);
840 		error = ufs_direnter(dvp, vp, newdir, cnp, NULL);
841 		pool_cache_put(ufs_direct_cache, newdir);
842 	}
843 	if (error) {
844 		ip->i_nlink--;
845 		DIP_ASSIGN(ip, nlink, ip->i_nlink);
846 		ip->i_flag |= IN_CHANGE;
847 		UFS_WAPBL_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
848 	}
849 	PNBUF_PUT(cnp->cn_pnbuf);
850 	UFS_WAPBL_END(vp->v_mount);
851  out1:
852 	if (dvp != vp)
853 		VOP_UNLOCK(vp, 0);
854  out2:
855 	VN_KNOTE(vp, NOTE_LINK);
856 	VN_KNOTE(dvp, NOTE_WRITE);
857 	vput(dvp);
858 	fstrans_done(dvp->v_mount);
859 	return (error);
860 }
861 
862 /*
863  * whiteout vnode call
864  */
865 int
866 ufs_whiteout(void *v)
867 {
868 	struct vop_whiteout_args /* {
869 		struct vnode		*a_dvp;
870 		struct componentname	*a_cnp;
871 		int			a_flags;
872 	} */ *ap = v;
873 	struct vnode		*dvp = ap->a_dvp;
874 	struct componentname	*cnp = ap->a_cnp;
875 	struct direct		*newdir;
876 	int			error;
877 	struct ufsmount		*ump = VFSTOUFS(dvp->v_mount);
878 
879 	error = 0;
880 	switch (ap->a_flags) {
881 	case LOOKUP:
882 		/* 4.4 format directories support whiteout operations */
883 		if (ump->um_maxsymlinklen > 0)
884 			return (0);
885 		return (EOPNOTSUPP);
886 
887 	case CREATE:
888 		/* create a new directory whiteout */
889 		fstrans_start(dvp->v_mount, FSTRANS_SHARED);
890 #ifdef DIAGNOSTIC
891 		if ((cnp->cn_flags & SAVENAME) == 0)
892 			panic("ufs_whiteout: missing name");
893 		if (ump->um_maxsymlinklen <= 0)
894 			panic("ufs_whiteout: old format filesystem");
895 #endif
896 
897 		newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
898 		newdir->d_ino = WINO;
899 		newdir->d_namlen = cnp->cn_namelen;
900 		memcpy(newdir->d_name, cnp->cn_nameptr,
901 		    (size_t)cnp->cn_namelen);
902 		newdir->d_name[cnp->cn_namelen] = '\0';
903 		newdir->d_type = DT_WHT;
904 		error = ufs_direnter(dvp, NULL, newdir, cnp, NULL);
905 		pool_cache_put(ufs_direct_cache, newdir);
906 		break;
907 
908 	case DELETE:
909 		/* remove an existing directory whiteout */
910 		fstrans_start(dvp->v_mount, FSTRANS_SHARED);
911 #ifdef DIAGNOSTIC
912 		if (ump->um_maxsymlinklen <= 0)
913 			panic("ufs_whiteout: old format filesystem");
914 #endif
915 
916 		cnp->cn_flags &= ~DOWHITEOUT;
917 		error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
918 		break;
919 	default:
920 		panic("ufs_whiteout: unknown op");
921 		/* NOTREACHED */
922 	}
923 	if (cnp->cn_flags & HASBUF) {
924 		PNBUF_PUT(cnp->cn_pnbuf);
925 		cnp->cn_flags &= ~HASBUF;
926 	}
927 	fstrans_done(dvp->v_mount);
928 	return (error);
929 }
930 
931 
932 /*
933  * Rename vnode operation
934  * 	rename("foo", "bar");
935  * is essentially
936  *	unlink("bar");
937  *	link("foo", "bar");
938  *	unlink("foo");
939  * but ``atomically''.  Can't do full commit without saving state in the
940  * inode on disk which isn't feasible at this time.  Best we can do is
941  * always guarantee the target exists.
942  *
943  * Basic algorithm is:
944  *
945  * 1) Bump link count on source while we're linking it to the
946  *    target.  This also ensure the inode won't be deleted out
947  *    from underneath us while we work (it may be truncated by
948  *    a concurrent `trunc' or `open' for creation).
949  * 2) Link source to destination.  If destination already exists,
950  *    delete it first.
951  * 3) Unlink source reference to inode if still around. If a
952  *    directory was moved and the parent of the destination
953  *    is different from the source, patch the ".." entry in the
954  *    directory.
955  */
956 int
957 ufs_rename(void *v)
958 {
959 	struct vop_rename_args  /* {
960 		struct vnode		*a_fdvp;
961 		struct vnode		*a_fvp;
962 		struct componentname	*a_fcnp;
963 		struct vnode		*a_tdvp;
964 		struct vnode		*a_tvp;
965 		struct componentname	*a_tcnp;
966 	} */ *ap = v;
967 	struct vnode		*tvp, *tdvp, *fvp, *fdvp;
968 	struct componentname	*tcnp, *fcnp;
969 	struct inode		*ip, *xp, *dp;
970 	struct mount		*mp;
971 	struct direct		*newdir;
972 	int			doingdirectory, oldparent, newparent, error;
973 
974 #ifdef WAPBL
975 	if (ap->a_tdvp->v_mount->mnt_wapbl)
976 		return wapbl_ufs_rename(v);
977 #endif
978 
979 	tvp = ap->a_tvp;
980 	tdvp = ap->a_tdvp;
981 	fvp = ap->a_fvp;
982 	fdvp = ap->a_fdvp;
983 	tcnp = ap->a_tcnp;
984 	fcnp = ap->a_fcnp;
985 	doingdirectory = oldparent = newparent = error = 0;
986 
987 #ifdef DIAGNOSTIC
988 	if ((tcnp->cn_flags & HASBUF) == 0 ||
989 	    (fcnp->cn_flags & HASBUF) == 0)
990 		panic("ufs_rename: no name");
991 #endif
992 	/*
993 	 * Check for cross-device rename.
994 	 */
995 	if ((fvp->v_mount != tdvp->v_mount) ||
996 	    (tvp && (fvp->v_mount != tvp->v_mount))) {
997 		error = EXDEV;
998  abortit:
999 		VOP_ABORTOP(tdvp, tcnp); /* XXX, why not in NFS? */
1000 		if (tdvp == tvp)
1001 			vrele(tdvp);
1002 		else
1003 			vput(tdvp);
1004 		if (tvp)
1005 			vput(tvp);
1006 		VOP_ABORTOP(fdvp, fcnp); /* XXX, why not in NFS? */
1007 		vrele(fdvp);
1008 		vrele(fvp);
1009 		return (error);
1010 	}
1011 
1012 	/*
1013 	 * Check if just deleting a link name.
1014 	 */
1015 	if (tvp && ((VTOI(tvp)->i_flags & (IMMUTABLE | APPEND)) ||
1016 	    (VTOI(tdvp)->i_flags & APPEND))) {
1017 		error = EPERM;
1018 		goto abortit;
1019 	}
1020 	if (fvp == tvp) {
1021 		if (fvp->v_type == VDIR) {
1022 			error = EINVAL;
1023 			goto abortit;
1024 		}
1025 
1026 		/* Release destination completely. */
1027 		VOP_ABORTOP(tdvp, tcnp);
1028 		vput(tdvp);
1029 		vput(tvp);
1030 
1031 		/* Delete source. */
1032 		vrele(fvp);
1033 		fcnp->cn_flags &= ~(MODMASK | SAVESTART);
1034 		fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1035 		fcnp->cn_nameiop = DELETE;
1036 		vn_lock(fdvp, LK_EXCLUSIVE | LK_RETRY);
1037 		if ((error = relookup(fdvp, &fvp, fcnp))) {
1038 			vput(fdvp);
1039 			return (error);
1040 		}
1041 		return (VOP_REMOVE(fdvp, fvp, fcnp));
1042 	}
1043 	if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
1044 		goto abortit;
1045 	dp = VTOI(fdvp);
1046 	ip = VTOI(fvp);
1047 	if ((nlink_t) ip->i_nlink >= LINK_MAX) {
1048 		VOP_UNLOCK(fvp, 0);
1049 		error = EMLINK;
1050 		goto abortit;
1051 	}
1052 	if ((ip->i_flags & (IMMUTABLE | APPEND)) ||
1053 		(dp->i_flags & APPEND)) {
1054 		VOP_UNLOCK(fvp, 0);
1055 		error = EPERM;
1056 		goto abortit;
1057 	}
1058 	if ((ip->i_mode & IFMT) == IFDIR) {
1059 		/*
1060 		 * Avoid ".", "..", and aliases of "." for obvious reasons.
1061 		 */
1062 		if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
1063 		    dp == ip ||
1064 		    (fcnp->cn_flags & ISDOTDOT) ||
1065 		    (tcnp->cn_flags & ISDOTDOT) ||
1066 		    (ip->i_flag & IN_RENAME)) {
1067 			VOP_UNLOCK(fvp, 0);
1068 			error = EINVAL;
1069 			goto abortit;
1070 		}
1071 		ip->i_flag |= IN_RENAME;
1072 		oldparent = dp->i_number;
1073 		doingdirectory = 1;
1074 	}
1075 	VN_KNOTE(fdvp, NOTE_WRITE);		/* XXXLUKEM/XXX: right place? */
1076 
1077 	/*
1078 	 * When the target exists, both the directory
1079 	 * and target vnodes are returned locked.
1080 	 */
1081 	dp = VTOI(tdvp);
1082 	xp = NULL;
1083 	if (tvp)
1084 		xp = VTOI(tvp);
1085 
1086 	mp = fdvp->v_mount;
1087 	fstrans_start(mp, FSTRANS_SHARED);
1088 
1089 	/*
1090 	 * 1) Bump link count while we're moving stuff
1091 	 *    around.  If we crash somewhere before
1092 	 *    completing our work, the link count
1093 	 *    may be wrong, but correctable.
1094 	 */
1095 	ip->i_nlink++;
1096 	DIP_ASSIGN(ip, nlink, ip->i_nlink);
1097 	ip->i_flag |= IN_CHANGE;
1098 	if ((error = UFS_UPDATE(fvp, NULL, NULL, UPDATE_DIROP)) != 0) {
1099 		VOP_UNLOCK(fvp, 0);
1100 		goto bad;
1101 	}
1102 
1103 	/*
1104 	 * If ".." must be changed (ie the directory gets a new
1105 	 * parent) then the source directory must not be in the
1106 	 * directory hierarchy above the target, as this would
1107 	 * orphan everything below the source directory. Also
1108 	 * the user must have write permission in the source so
1109 	 * as to be able to change "..". We must repeat the call
1110 	 * to namei, as the parent directory is unlocked by the
1111 	 * call to checkpath().
1112 	 */
1113 	error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
1114 	VOP_UNLOCK(fvp, 0);
1115 	if (oldparent != dp->i_number)
1116 		newparent = dp->i_number;
1117 	if (doingdirectory && newparent) {
1118 		if (error)	/* write access check above */
1119 			goto bad;
1120 		if (xp != NULL)
1121 			vput(tvp);
1122 		vref(tdvp);	/* compensate for the ref checkpath loses */
1123 		if ((error = ufs_checkpath(ip, dp, tcnp->cn_cred)) != 0) {
1124 			vrele(tdvp);
1125 			goto out;
1126 		}
1127 		tcnp->cn_flags &= ~SAVESTART;
1128 		vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY);
1129 		error = relookup(tdvp, &tvp, tcnp);
1130 		if (error != 0) {
1131 			vput(tdvp);
1132 			goto out;
1133 		}
1134 		dp = VTOI(tdvp);
1135 		xp = NULL;
1136 		if (tvp)
1137 			xp = VTOI(tvp);
1138 	}
1139 	/*
1140 	 * 2) If target doesn't exist, link the target
1141 	 *    to the source and unlink the source.
1142 	 *    Otherwise, rewrite the target directory
1143 	 *    entry to reference the source inode and
1144 	 *    expunge the original entry's existence.
1145 	 */
1146 	if (xp == NULL) {
1147 		if (dp->i_dev != ip->i_dev)
1148 			panic("rename: EXDEV");
1149 		/*
1150 		 * Account for ".." in new directory.
1151 		 * When source and destination have the same
1152 		 * parent we don't fool with the link count.
1153 		 */
1154 		if (doingdirectory && newparent) {
1155 			if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1156 				error = EMLINK;
1157 				goto bad;
1158 			}
1159 			dp->i_nlink++;
1160 			DIP_ASSIGN(dp, nlink, dp->i_nlink);
1161 			dp->i_flag |= IN_CHANGE;
1162 			if ((error = UFS_UPDATE(tdvp, NULL, NULL,
1163 			    UPDATE_DIROP)) != 0) {
1164 				dp->i_nlink--;
1165 				DIP_ASSIGN(dp, nlink, dp->i_nlink);
1166 				dp->i_flag |= IN_CHANGE;
1167 				goto bad;
1168 			}
1169 		}
1170 		newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
1171 		ufs_makedirentry(ip, tcnp, newdir);
1172 		error = ufs_direnter(tdvp, NULL, newdir, tcnp, NULL);
1173 		pool_cache_put(ufs_direct_cache, newdir);
1174 		if (error != 0) {
1175 			if (doingdirectory && newparent) {
1176 				dp->i_nlink--;
1177 				DIP_ASSIGN(dp, nlink, dp->i_nlink);
1178 				dp->i_flag |= IN_CHANGE;
1179 				(void)UFS_UPDATE(tdvp, NULL, NULL,
1180 						 UPDATE_WAIT|UPDATE_DIROP);
1181 			}
1182 			goto bad;
1183 		}
1184 		VN_KNOTE(tdvp, NOTE_WRITE);
1185 		vput(tdvp);
1186 	} else {
1187 		if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1188 			panic("rename: EXDEV");
1189 		/*
1190 		 * Short circuit rename(foo, foo).
1191 		 */
1192 		if (xp->i_number == ip->i_number)
1193 			panic("rename: same file");
1194 		/*
1195 		 * If the parent directory is "sticky", then the user must
1196 		 * own the parent directory, or the destination of the rename,
1197 		 * otherwise the destination may not be changed (except by
1198 		 * root). This implements append-only directories.
1199 		 */
1200 		if ((dp->i_mode & S_ISTXT) &&
1201 		    kauth_authorize_generic(tcnp->cn_cred,
1202 		     KAUTH_GENERIC_ISSUSER, NULL) != 0 &&
1203 		    kauth_cred_geteuid(tcnp->cn_cred) != dp->i_uid &&
1204 		    xp->i_uid != kauth_cred_geteuid(tcnp->cn_cred)) {
1205 			error = EPERM;
1206 			goto bad;
1207 		}
1208 		/*
1209 		 * Target must be empty if a directory and have no links
1210 		 * to it. Also, ensure source and target are compatible
1211 		 * (both directories, or both not directories).
1212 		 */
1213 		if ((xp->i_mode & IFMT) == IFDIR) {
1214 			if (xp->i_nlink > 2 ||
1215 			    !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1216 				error = ENOTEMPTY;
1217 				goto bad;
1218 			}
1219 			if (!doingdirectory) {
1220 				error = ENOTDIR;
1221 				goto bad;
1222 			}
1223 			cache_purge(tdvp);
1224 		} else if (doingdirectory) {
1225 			error = EISDIR;
1226 			goto bad;
1227 		}
1228 		if ((error = ufs_dirrewrite(dp, xp, ip->i_number,
1229 		    IFTODT(ip->i_mode), doingdirectory && newparent ?
1230 		    newparent : doingdirectory, IN_CHANGE | IN_UPDATE)) != 0)
1231 			goto bad;
1232 		if (doingdirectory) {
1233 			/*
1234 			 * Truncate inode. The only stuff left in the directory
1235 			 * is "." and "..". The "." reference is inconsequential
1236 			 * since we are quashing it. We have removed the "."
1237 			 * reference and the reference in the parent directory,
1238 			 * but there may be other hard links.
1239 			 */
1240 			if (!newparent) {
1241 				dp->i_nlink--;
1242 				DIP_ASSIGN(dp, nlink, dp->i_nlink);
1243 				dp->i_flag |= IN_CHANGE;
1244 			}
1245 			xp->i_nlink--;
1246 			DIP_ASSIGN(xp, nlink, xp->i_nlink);
1247 			xp->i_flag |= IN_CHANGE;
1248 			if ((error = UFS_TRUNCATE(tvp, (off_t)0, IO_SYNC,
1249 			    tcnp->cn_cred)))
1250 				goto bad;
1251 		}
1252 		VN_KNOTE(tdvp, NOTE_WRITE);
1253 		vput(tdvp);
1254 		VN_KNOTE(tvp, NOTE_DELETE);
1255 		vput(tvp);
1256 		xp = NULL;
1257 	}
1258 
1259 	/*
1260 	 * 3) Unlink the source.
1261 	 */
1262 	fcnp->cn_flags &= ~(MODMASK | SAVESTART);
1263 	fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1264 	vn_lock(fdvp, LK_EXCLUSIVE | LK_RETRY);
1265 	if ((error = relookup(fdvp, &fvp, fcnp))) {
1266 		vput(fdvp);
1267 		vrele(ap->a_fvp);
1268 		goto out2;
1269 	}
1270 	if (fvp != NULL) {
1271 		xp = VTOI(fvp);
1272 		dp = VTOI(fdvp);
1273 	} else {
1274 		/*
1275 		 * From name has disappeared.
1276 		 */
1277 		if (doingdirectory)
1278 			panic("rename: lost dir entry");
1279 		vrele(ap->a_fvp);
1280 		error = 0;
1281 		goto out2;
1282 	}
1283 	/*
1284 	 * Ensure that the directory entry still exists and has not
1285 	 * changed while the new name has been entered. If the source is
1286 	 * a file then the entry may have been unlinked or renamed. In
1287 	 * either case there is no further work to be done. If the source
1288 	 * is a directory then it cannot have been rmdir'ed; The IRENAME
1289 	 * flag ensures that it cannot be moved by another rename or removed
1290 	 * by a rmdir.
1291 	 */
1292 	if (xp != ip) {
1293 		if (doingdirectory)
1294 			panic("rename: lost dir entry");
1295 	} else {
1296 		/*
1297 		 * If the source is a directory with a
1298 		 * new parent, the link count of the old
1299 		 * parent directory must be decremented
1300 		 * and ".." set to point to the new parent.
1301 		 */
1302 		if (doingdirectory && newparent) {
1303 			KASSERT(dp != NULL);
1304 			xp->i_offset = mastertemplate.dot_reclen;
1305 			ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0, IN_CHANGE);
1306 			cache_purge(fdvp);
1307 		}
1308 		error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1309 		xp->i_flag &= ~IN_RENAME;
1310 	}
1311 	VN_KNOTE(fvp, NOTE_RENAME);
1312 	if (dp)
1313 		vput(fdvp);
1314 	if (xp)
1315 		vput(fvp);
1316 	vrele(ap->a_fvp);
1317 	goto out2;
1318 
1319 	/* exit routines from steps 1 & 2 */
1320  bad:
1321 	if (xp)
1322 		vput(ITOV(xp));
1323 	vput(ITOV(dp));
1324  out:
1325 	if (doingdirectory)
1326 		ip->i_flag &= ~IN_RENAME;
1327 	if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
1328 		ip->i_nlink--;
1329 		DIP_ASSIGN(ip, nlink, ip->i_nlink);
1330 		ip->i_flag |= IN_CHANGE;
1331 		ip->i_flag &= ~IN_RENAME;
1332 		vput(fvp);
1333 	} else
1334 		vrele(fvp);
1335 	vrele(fdvp);
1336 
1337 	/* exit routines from step 3 */
1338  out2:
1339 	fstrans_done(mp);
1340 	return (error);
1341 }
1342 
1343 int
1344 ufs_mkdir(void *v)
1345 {
1346 	struct vop_mkdir_args /* {
1347 		struct vnode		*a_dvp;
1348 		struct vnode		**a_vpp;
1349 		struct componentname	*a_cnp;
1350 		struct vattr		*a_vap;
1351 	} */ *ap = v;
1352 	struct vnode		*dvp = ap->a_dvp, *tvp;
1353 	struct vattr		*vap = ap->a_vap;
1354 	struct componentname	*cnp = ap->a_cnp;
1355 	struct inode		*ip, *dp = VTOI(dvp);
1356 	struct buf		*bp;
1357 	struct dirtemplate	dirtemplate;
1358 	struct direct		*newdir;
1359 	int			error, dmode;
1360 	struct ufsmount		*ump = dp->i_ump;
1361 	int			dirblksiz = ump->um_dirblksiz;
1362 
1363 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
1364 
1365 #ifdef DIAGNOSTIC
1366 	if ((cnp->cn_flags & HASBUF) == 0)
1367 		panic("ufs_mkdir: no name");
1368 #endif
1369 	if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1370 		error = EMLINK;
1371 		goto out;
1372 	}
1373 	dmode = vap->va_mode & ACCESSPERMS;
1374 	dmode |= IFDIR;
1375 	/*
1376 	 * Must simulate part of ufs_makeinode here to acquire the inode,
1377 	 * but not have it entered in the parent directory. The entry is
1378 	 * made later after writing "." and ".." entries.
1379 	 */
1380 	if ((error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, ap->a_vpp)) != 0)
1381 		goto out;
1382 	error = UFS_WAPBL_BEGIN(ap->a_dvp->v_mount);
1383 	if (error)
1384 		goto out;
1385 	tvp = *ap->a_vpp;
1386 	ip = VTOI(tvp);
1387 	ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
1388 	DIP_ASSIGN(ip, uid, ip->i_uid);
1389 	ip->i_gid = dp->i_gid;
1390 	DIP_ASSIGN(ip, gid, ip->i_gid);
1391 #ifdef QUOTA
1392 	if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1393 		PNBUF_PUT(cnp->cn_pnbuf);
1394 		UFS_VFREE(tvp, ip->i_number, dmode);
1395 		UFS_WAPBL_END(dvp->v_mount);
1396 		fstrans_done(dvp->v_mount);
1397 		vput(tvp);
1398 		vput(dvp);
1399 		return (error);
1400 	}
1401 #endif
1402 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1403 	ip->i_mode = dmode;
1404 	DIP_ASSIGN(ip, mode, dmode);
1405 	tvp->v_type = VDIR;	/* Rest init'd in getnewvnode(). */
1406 	ip->i_nlink = 2;
1407 	DIP_ASSIGN(ip, nlink, 2);
1408 	if (cnp->cn_flags & ISWHITEOUT) {
1409 		ip->i_flags |= UF_OPAQUE;
1410 		DIP_ASSIGN(ip, flags, ip->i_flags);
1411 	}
1412 
1413 	/*
1414 	 * Bump link count in parent directory to reflect work done below.
1415 	 * Should be done before reference is created so cleanup is
1416 	 * possible if we crash.
1417 	 */
1418 	dp->i_nlink++;
1419 	DIP_ASSIGN(dp, nlink, dp->i_nlink);
1420 	dp->i_flag |= IN_CHANGE;
1421 	if ((error = UFS_UPDATE(dvp, NULL, NULL, UPDATE_DIROP)) != 0)
1422 		goto bad;
1423 
1424 	/*
1425 	 * Initialize directory with "." and ".." from static template.
1426 	 */
1427 	dirtemplate = mastertemplate;
1428 	dirtemplate.dotdot_reclen = dirblksiz - dirtemplate.dot_reclen;
1429 	dirtemplate.dot_ino = ufs_rw32(ip->i_number, UFS_MPNEEDSWAP(ump));
1430 	dirtemplate.dotdot_ino = ufs_rw32(dp->i_number, UFS_MPNEEDSWAP(ump));
1431 	dirtemplate.dot_reclen = ufs_rw16(dirtemplate.dot_reclen,
1432 	    UFS_MPNEEDSWAP(ump));
1433 	dirtemplate.dotdot_reclen = ufs_rw16(dirtemplate.dotdot_reclen,
1434 	    UFS_MPNEEDSWAP(ump));
1435 	if (ump->um_maxsymlinklen <= 0) {
1436 #if BYTE_ORDER == LITTLE_ENDIAN
1437 		if (UFS_MPNEEDSWAP(ump) == 0)
1438 #else
1439 		if (UFS_MPNEEDSWAP(ump) != 0)
1440 #endif
1441 		{
1442 			dirtemplate.dot_type = dirtemplate.dot_namlen;
1443 			dirtemplate.dotdot_type = dirtemplate.dotdot_namlen;
1444 			dirtemplate.dot_namlen = dirtemplate.dotdot_namlen = 0;
1445 		} else
1446 			dirtemplate.dot_type = dirtemplate.dotdot_type = 0;
1447 	}
1448 	if ((error = UFS_BALLOC(tvp, (off_t)0, dirblksiz, cnp->cn_cred,
1449 	    B_CLRBUF, &bp)) != 0)
1450 		goto bad;
1451 	ip->i_size = dirblksiz;
1452 	DIP_ASSIGN(ip, size, dirblksiz);
1453 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
1454 	uvm_vnp_setsize(tvp, ip->i_size);
1455 	memcpy((void *)bp->b_data, (void *)&dirtemplate, sizeof dirtemplate);
1456 
1457 	/*
1458 	 * Directory set up, now install it's entry in the parent directory.
1459 	 * We must write out the buffer containing the new directory body
1460 	 * before entering the new name in the parent.
1461 	 */
1462 	if ((error = VOP_BWRITE(bp)) != 0)
1463 		goto bad;
1464 	if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0) {
1465 		goto bad;
1466 	}
1467 	newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
1468 	ufs_makedirentry(ip, cnp, newdir);
1469 	error = ufs_direnter(dvp, tvp, newdir, cnp, bp);
1470 	pool_cache_put(ufs_direct_cache, newdir);
1471  bad:
1472 	if (error == 0) {
1473 		VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1474 		UFS_WAPBL_END(dvp->v_mount);
1475 	} else {
1476 		dp->i_nlink--;
1477 		DIP_ASSIGN(dp, nlink, dp->i_nlink);
1478 		dp->i_flag |= IN_CHANGE;
1479 		UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
1480 		/*
1481 		 * No need to do an explicit UFS_TRUNCATE here, vrele will
1482 		 * do this for us because we set the link count to 0.
1483 		 */
1484 		ip->i_nlink = 0;
1485 		DIP_ASSIGN(ip, nlink, 0);
1486 		ip->i_flag |= IN_CHANGE;
1487 		/* If IN_ADIROP, account for it */
1488 		UFS_UNMARK_VNODE(tvp);
1489 		UFS_WAPBL_UPDATE(tvp, NULL, NULL, UPDATE_DIROP);
1490 		UFS_WAPBL_END(dvp->v_mount);
1491 		vput(tvp);
1492 	}
1493  out:
1494 	PNBUF_PUT(cnp->cn_pnbuf);
1495 	fstrans_done(dvp->v_mount);
1496 	vput(dvp);
1497 	return (error);
1498 }
1499 
1500 int
1501 ufs_rmdir(void *v)
1502 {
1503 	struct vop_rmdir_args /* {
1504 		struct vnode		*a_dvp;
1505 		struct vnode		*a_vp;
1506 		struct componentname	*a_cnp;
1507 	} */ *ap = v;
1508 	struct vnode		*vp, *dvp;
1509 	struct componentname	*cnp;
1510 	struct inode		*ip, *dp;
1511 	int			error;
1512 
1513 	vp = ap->a_vp;
1514 	dvp = ap->a_dvp;
1515 	cnp = ap->a_cnp;
1516 	ip = VTOI(vp);
1517 	dp = VTOI(dvp);
1518 	/*
1519 	 * No rmdir "." or of mounted directories please.
1520 	 */
1521 	if (dp == ip || vp->v_mountedhere != NULL) {
1522 		if (dp == ip)
1523 			vrele(vp);
1524 		else
1525 			vput(vp);
1526 		vput(vp);
1527 		return (EINVAL);
1528 	}
1529 
1530 	fstrans_start(dvp->v_mount, FSTRANS_SHARED);
1531 
1532 	/*
1533 	 * Do not remove a directory that is in the process of being renamed.
1534 	 * Verify that the directory is empty (and valid). (Rmdir ".." won't
1535 	 * be valid since ".." will contain a reference to the current
1536 	 * directory and thus be non-empty.)
1537 	 */
1538 	error = 0;
1539 	if (ip->i_flag & IN_RENAME) {
1540 		error = EINVAL;
1541 		goto out;
1542 	}
1543 	if (ip->i_nlink != 2 ||
1544 	    !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1545 		error = ENOTEMPTY;
1546 		goto out;
1547 	}
1548 	if ((dp->i_flags & APPEND) ||
1549 		(ip->i_flags & (IMMUTABLE | APPEND))) {
1550 		error = EPERM;
1551 		goto out;
1552 	}
1553 	error = UFS_WAPBL_BEGIN(dvp->v_mount);
1554 	if (error)
1555 		goto out;
1556 	/*
1557 	 * Delete reference to directory before purging
1558 	 * inode.  If we crash in between, the directory
1559 	 * will be reattached to lost+found,
1560 	 */
1561 	error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1562 	if (error) {
1563 		UFS_WAPBL_END(dvp->v_mount);
1564 		goto out;
1565 	}
1566 	VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
1567 	cache_purge(dvp);
1568 	/*
1569 	 * Truncate inode.  The only stuff left in the directory is "." and
1570 	 * "..".  The "." reference is inconsequential since we're quashing
1571 	 * it.
1572 	 */
1573 	dp->i_nlink--;
1574 	DIP_ASSIGN(dp, nlink, dp->i_nlink);
1575 	dp->i_flag |= IN_CHANGE;
1576 	UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
1577 	ip->i_nlink--;
1578 	DIP_ASSIGN(ip, nlink, ip->i_nlink);
1579 	ip->i_flag |= IN_CHANGE;
1580 	error = UFS_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
1581 	cache_purge(vp);
1582 	/*
1583 	 * Unlock the log while we still have reference to unlinked
1584 	 * directory vp so that it will not get locked for recycling
1585 	 */
1586 	UFS_WAPBL_END(dvp->v_mount);
1587 #ifdef UFS_DIRHASH
1588 	if (ip->i_dirhash != NULL)
1589 		ufsdirhash_free(ip);
1590 #endif
1591  out:
1592 	VN_KNOTE(vp, NOTE_DELETE);
1593 	vput(dvp);
1594 	vput(vp);
1595 	fstrans_done(dvp->v_mount);
1596 	return (error);
1597 }
1598 
1599 /*
1600  * symlink -- make a symbolic link
1601  */
1602 int
1603 ufs_symlink(void *v)
1604 {
1605 	struct vop_symlink_args /* {
1606 		struct vnode		*a_dvp;
1607 		struct vnode		**a_vpp;
1608 		struct componentname	*a_cnp;
1609 		struct vattr		*a_vap;
1610 		char			*a_target;
1611 	} */ *ap = v;
1612 	struct vnode	*vp, **vpp;
1613 	struct inode	*ip;
1614 	int		len, error;
1615 
1616 	vpp = ap->a_vpp;
1617 	/*
1618 	 * UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
1619 	 * ufs_makeinode
1620 	 */
1621 	fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
1622 	error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1623 			      vpp, ap->a_cnp);
1624 	if (error)
1625 		goto out;
1626 	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
1627 	vp = *vpp;
1628 	len = strlen(ap->a_target);
1629 	ip = VTOI(vp);
1630 	if (len < ip->i_ump->um_maxsymlinklen) {
1631 		memcpy((char *)SHORTLINK(ip), ap->a_target, len);
1632 		ip->i_size = len;
1633 		DIP_ASSIGN(ip, size, len);
1634 		uvm_vnp_setsize(vp, ip->i_size);
1635 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
1636 		UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
1637 	} else
1638 		error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1639 		    UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
1640 		    ap->a_cnp->cn_cred, NULL, NULL);
1641 	UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
1642 	if (error)
1643 		vput(vp);
1644 out:
1645 	fstrans_done(ap->a_dvp->v_mount);
1646 	return (error);
1647 }
1648 
1649 /*
1650  * Vnode op for reading directories.
1651  *
1652  * This routine handles converting from the on-disk directory format
1653  * "struct direct" to the in-memory format "struct dirent" as well as
1654  * byte swapping the entries if necessary.
1655  */
1656 int
1657 ufs_readdir(void *v)
1658 {
1659 	struct vop_readdir_args /* {
1660 		struct vnode	*a_vp;
1661 		struct uio	*a_uio;
1662 		kauth_cred_t	a_cred;
1663 		int		*a_eofflag;
1664 		off_t		**a_cookies;
1665 		int		*ncookies;
1666 	} */ *ap = v;
1667 	struct vnode	*vp = ap->a_vp;
1668 	struct direct	*cdp, *ecdp;
1669 	struct dirent	*ndp;
1670 	char		*cdbuf, *ndbuf, *endp;
1671 	struct uio	auio, *uio;
1672 	struct iovec	aiov;
1673 	int		error;
1674 	size_t		count, ccount, rcount;
1675 	off_t		off, *ccp;
1676 	off_t		startoff;
1677 	size_t		skipbytes;
1678 	struct ufsmount	*ump = VFSTOUFS(vp->v_mount);
1679 	int nswap = UFS_MPNEEDSWAP(ump);
1680 #if BYTE_ORDER == LITTLE_ENDIAN
1681 	int needswap = ump->um_maxsymlinklen <= 0 && nswap == 0;
1682 #else
1683 	int needswap = ump->um_maxsymlinklen <= 0 && nswap != 0;
1684 #endif
1685 	uio = ap->a_uio;
1686 	count = uio->uio_resid;
1687 	rcount = count - ((uio->uio_offset + count) & (ump->um_dirblksiz - 1));
1688 
1689 	if (rcount < _DIRENT_MINSIZE(cdp) || count < _DIRENT_MINSIZE(ndp))
1690 		return EINVAL;
1691 
1692 	startoff = uio->uio_offset & ~(ump->um_dirblksiz - 1);
1693 	skipbytes = uio->uio_offset - startoff;
1694 	rcount += skipbytes;
1695 
1696 	auio.uio_iov = &aiov;
1697 	auio.uio_iovcnt = 1;
1698 	auio.uio_offset = startoff;
1699 	auio.uio_resid = rcount;
1700 	UIO_SETUP_SYSSPACE(&auio);
1701 	auio.uio_rw = UIO_READ;
1702 	cdbuf = malloc(rcount, M_TEMP, M_WAITOK);
1703 	aiov.iov_base = cdbuf;
1704 	aiov.iov_len = rcount;
1705 	error = VOP_READ(vp, &auio, 0, ap->a_cred);
1706 	if (error != 0) {
1707 		free(cdbuf, M_TEMP);
1708 		return error;
1709 	}
1710 
1711 	rcount -= auio.uio_resid;
1712 
1713 	cdp = (struct direct *)(void *)cdbuf;
1714 	ecdp = (struct direct *)(void *)&cdbuf[rcount];
1715 
1716 	ndbuf = malloc(count, M_TEMP, M_WAITOK);
1717 	ndp = (struct dirent *)(void *)ndbuf;
1718 	endp = &ndbuf[count];
1719 
1720 	off = uio->uio_offset;
1721 	if (ap->a_cookies) {
1722 		ccount = rcount / _DIRENT_RECLEN(cdp, 1);
1723 		ccp = *(ap->a_cookies) = malloc(ccount * sizeof(*ccp),
1724 		    M_TEMP, M_WAITOK);
1725 	} else {
1726 		/* XXX: GCC */
1727 		ccount = 0;
1728 		ccp = NULL;
1729 	}
1730 
1731 	while (cdp < ecdp) {
1732 		cdp->d_reclen = ufs_rw16(cdp->d_reclen, nswap);
1733 		if (skipbytes > 0) {
1734 			if (cdp->d_reclen <= skipbytes) {
1735 				skipbytes -= cdp->d_reclen;
1736 				cdp = _DIRENT_NEXT(cdp);
1737 				continue;
1738 			}
1739 			/*
1740 			 * invalid cookie.
1741 			 */
1742 			error = EINVAL;
1743 			goto out;
1744 		}
1745 		if (cdp->d_reclen == 0) {
1746 			struct dirent *ondp = ndp;
1747 			ndp->d_reclen = _DIRENT_MINSIZE(ndp);
1748 			ndp = _DIRENT_NEXT(ndp);
1749 			ondp->d_reclen = 0;
1750 			cdp = ecdp;
1751 			break;
1752 		}
1753 		if (needswap) {
1754 			ndp->d_type = cdp->d_namlen;
1755 			ndp->d_namlen = cdp->d_type;
1756 		} else {
1757 			ndp->d_type = cdp->d_type;
1758 			ndp->d_namlen = cdp->d_namlen;
1759 		}
1760 		ndp->d_reclen = _DIRENT_RECLEN(ndp, ndp->d_namlen);
1761 		if ((char *)(void *)ndp + ndp->d_reclen +
1762 		    _DIRENT_MINSIZE(ndp) > endp)
1763 			break;
1764 		ndp->d_fileno = ufs_rw32(cdp->d_ino, nswap);
1765 		(void)memcpy(ndp->d_name, cdp->d_name, ndp->d_namlen);
1766 		memset(&ndp->d_name[ndp->d_namlen], 0,
1767 		    ndp->d_reclen - _DIRENT_NAMEOFF(ndp) - ndp->d_namlen);
1768 		off += cdp->d_reclen;
1769 		if (ap->a_cookies) {
1770 			KASSERT(ccp - *(ap->a_cookies) < ccount);
1771 			*(ccp++) = off;
1772 		}
1773 		ndp = _DIRENT_NEXT(ndp);
1774 		cdp = _DIRENT_NEXT(cdp);
1775 	}
1776 
1777 	count = ((char *)(void *)ndp - ndbuf);
1778 	error = uiomove(ndbuf, count, uio);
1779 out:
1780 	if (ap->a_cookies) {
1781 		if (error) {
1782 			free(*(ap->a_cookies), M_TEMP);
1783 			*(ap->a_cookies) = NULL;
1784 			*(ap->a_ncookies) = 0;
1785 		} else {
1786 			*ap->a_ncookies = ccp - *(ap->a_cookies);
1787 		}
1788 	}
1789 	uio->uio_offset = off;
1790 	free(ndbuf, M_TEMP);
1791 	free(cdbuf, M_TEMP);
1792 	*ap->a_eofflag = VTOI(vp)->i_size <= uio->uio_offset;
1793 	return error;
1794 }
1795 
1796 /*
1797  * Return target name of a symbolic link
1798  */
1799 int
1800 ufs_readlink(void *v)
1801 {
1802 	struct vop_readlink_args /* {
1803 		struct vnode	*a_vp;
1804 		struct uio	*a_uio;
1805 		kauth_cred_t	a_cred;
1806 	} */ *ap = v;
1807 	struct vnode	*vp = ap->a_vp;
1808 	struct inode	*ip = VTOI(vp);
1809 	struct ufsmount	*ump = VFSTOUFS(vp->v_mount);
1810 	int		isize;
1811 
1812 	isize = ip->i_size;
1813 	if (isize < ump->um_maxsymlinklen ||
1814 	    (ump->um_maxsymlinklen == 0 && DIP(ip, blocks) == 0)) {
1815 		uiomove((char *)SHORTLINK(ip), isize, ap->a_uio);
1816 		return (0);
1817 	}
1818 	return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1819 }
1820 
1821 /*
1822  * Calculate the logical to physical mapping if not done already,
1823  * then call the device strategy routine.
1824  */
1825 int
1826 ufs_strategy(void *v)
1827 {
1828 	struct vop_strategy_args /* {
1829 		struct vnode *a_vp;
1830 		struct buf *a_bp;
1831 	} */ *ap = v;
1832 	struct buf	*bp;
1833 	struct vnode	*vp;
1834 	struct inode	*ip;
1835 	struct mount	*mp;
1836 	int		error;
1837 
1838 	bp = ap->a_bp;
1839 	vp = ap->a_vp;
1840 	ip = VTOI(vp);
1841 	if (vp->v_type == VBLK || vp->v_type == VCHR)
1842 		panic("ufs_strategy: spec");
1843 	KASSERT(bp->b_bcount != 0);
1844 	if (bp->b_blkno == bp->b_lblkno) {
1845 		error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno,
1846 				 NULL);
1847 		if (error) {
1848 			bp->b_error = error;
1849 			biodone(bp);
1850 			return (error);
1851 		}
1852 		if (bp->b_blkno == -1) /* no valid data */
1853 			clrbuf(bp);
1854 	}
1855 	if (bp->b_blkno < 0) { /* block is not on disk */
1856 		biodone(bp);
1857 		return (0);
1858 	}
1859 	vp = ip->i_devvp;
1860 
1861 	error = VOP_STRATEGY(vp, bp);
1862 	if (error)
1863 		return error;
1864 
1865 	if (!BUF_ISREAD(bp))
1866 		return 0;
1867 
1868 	mp = wapbl_vptomp(vp);
1869 	if (mp == NULL || mp->mnt_wapbl_replay == NULL ||
1870 	    !WAPBL_REPLAY_ISOPEN(mp) ||
1871 	    !WAPBL_REPLAY_CAN_READ(mp, bp->b_blkno, bp->b_bcount))
1872 		return 0;
1873 
1874 	error = biowait(bp);
1875 	if (error)
1876 		return error;
1877 
1878 	error = WAPBL_REPLAY_READ(mp, bp->b_data, bp->b_blkno, bp->b_bcount);
1879 	if (error) {
1880 		mutex_enter(&bufcache_lock);
1881 		SET(bp->b_cflags, BC_INVAL);
1882 		mutex_exit(&bufcache_lock);
1883 	}
1884 	return error;
1885 }
1886 
1887 /*
1888  * Print out the contents of an inode.
1889  */
1890 int
1891 ufs_print(void *v)
1892 {
1893 	struct vop_print_args /* {
1894 		struct vnode	*a_vp;
1895 	} */ *ap = v;
1896 	struct vnode	*vp;
1897 	struct inode	*ip;
1898 
1899 	vp = ap->a_vp;
1900 	ip = VTOI(vp);
1901 	printf("tag VT_UFS, ino %llu, on dev %llu, %llu",
1902 	    (unsigned long long)ip->i_number,
1903 	    (unsigned long long)major(ip->i_dev),
1904 	    (unsigned long long)minor(ip->i_dev));
1905 	printf(" flags 0x%x, nlink %d\n",
1906 	    ip->i_flag, ip->i_nlink);
1907 	printf("\tmode 0%o, owner %d, group %d, size %qd",
1908 	    ip->i_mode, ip->i_uid, ip->i_gid,
1909 	    (long long)ip->i_size);
1910 	if (vp->v_type == VFIFO)
1911 		fifo_printinfo(vp);
1912 	printf("\n");
1913 	return (0);
1914 }
1915 
1916 /*
1917  * Read wrapper for special devices.
1918  */
1919 int
1920 ufsspec_read(void *v)
1921 {
1922 	struct vop_read_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 access flag.
1931 	 */
1932 	if ((ap->a_vp->v_mount->mnt_flag & MNT_NODEVMTIME) == 0)
1933 		VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1934 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_read), ap));
1935 }
1936 
1937 /*
1938  * Write wrapper for special devices.
1939  */
1940 int
1941 ufsspec_write(void *v)
1942 {
1943 	struct vop_write_args /* {
1944 		struct vnode	*a_vp;
1945 		struct uio	*a_uio;
1946 		int		a_ioflag;
1947 		kauth_cred_t	a_cred;
1948 	} */ *ap = v;
1949 
1950 	/*
1951 	 * Set update and change flags.
1952 	 */
1953 	if ((ap->a_vp->v_mount->mnt_flag & MNT_NODEVMTIME) == 0)
1954 		VTOI(ap->a_vp)->i_flag |= IN_MODIFY;
1955 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_write), ap));
1956 }
1957 
1958 /*
1959  * Close wrapper for special devices.
1960  *
1961  * Update the times on the inode then do device close.
1962  */
1963 int
1964 ufsspec_close(void *v)
1965 {
1966 	struct vop_close_args /* {
1967 		struct vnode	*a_vp;
1968 		int		a_fflag;
1969 		kauth_cred_t	a_cred;
1970 	} */ *ap = v;
1971 	struct vnode	*vp;
1972 	struct inode	*ip;
1973 
1974 	vp = ap->a_vp;
1975 	ip = VTOI(vp);
1976 	if (vp->v_usecount > 1)
1977 		UFS_ITIMES(vp, NULL, NULL, NULL);
1978 	return (VOCALL (spec_vnodeop_p, VOFFSET(vop_close), ap));
1979 }
1980 
1981 /*
1982  * Read wrapper for fifo's
1983  */
1984 int
1985 ufsfifo_read(void *v)
1986 {
1987 	struct vop_read_args /* {
1988 		struct vnode	*a_vp;
1989 		struct uio	*a_uio;
1990 		int		a_ioflag;
1991 		kauth_cred_t	a_cred;
1992 	} */ *ap = v;
1993 
1994 	/*
1995 	 * Set access flag.
1996 	 */
1997 	VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1998 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_read), ap));
1999 }
2000 
2001 /*
2002  * Write wrapper for fifo's.
2003  */
2004 int
2005 ufsfifo_write(void *v)
2006 {
2007 	struct vop_write_args /* {
2008 		struct vnode	*a_vp;
2009 		struct uio	*a_uio;
2010 		int		a_ioflag;
2011 		kauth_cred_t	a_cred;
2012 	} */ *ap = v;
2013 
2014 	/*
2015 	 * Set update and change flags.
2016 	 */
2017 	VTOI(ap->a_vp)->i_flag |= IN_MODIFY;
2018 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_write), ap));
2019 }
2020 
2021 /*
2022  * Close wrapper for fifo's.
2023  *
2024  * Update the times on the inode then do device close.
2025  */
2026 int
2027 ufsfifo_close(void *v)
2028 {
2029 	struct vop_close_args /* {
2030 		struct vnode	*a_vp;
2031 		int		a_fflag;
2032 		kauth_cred_t	a_cred;
2033 	} */ *ap = v;
2034 	struct vnode	*vp;
2035 	struct inode	*ip;
2036 
2037 	vp = ap->a_vp;
2038 	ip = VTOI(vp);
2039 	if (ap->a_vp->v_usecount > 1)
2040 		UFS_ITIMES(vp, NULL, NULL, NULL);
2041 	return (VOCALL (fifo_vnodeop_p, VOFFSET(vop_close), ap));
2042 }
2043 
2044 /*
2045  * Return POSIX pathconf information applicable to ufs filesystems.
2046  */
2047 int
2048 ufs_pathconf(void *v)
2049 {
2050 	struct vop_pathconf_args /* {
2051 		struct vnode	*a_vp;
2052 		int		a_name;
2053 		register_t	*a_retval;
2054 	} */ *ap = v;
2055 
2056 	switch (ap->a_name) {
2057 	case _PC_LINK_MAX:
2058 		*ap->a_retval = LINK_MAX;
2059 		return (0);
2060 	case _PC_NAME_MAX:
2061 		*ap->a_retval = NAME_MAX;
2062 		return (0);
2063 	case _PC_PATH_MAX:
2064 		*ap->a_retval = PATH_MAX;
2065 		return (0);
2066 	case _PC_PIPE_BUF:
2067 		*ap->a_retval = PIPE_BUF;
2068 		return (0);
2069 	case _PC_CHOWN_RESTRICTED:
2070 		*ap->a_retval = 1;
2071 		return (0);
2072 	case _PC_NO_TRUNC:
2073 		*ap->a_retval = 1;
2074 		return (0);
2075 	case _PC_SYNC_IO:
2076 		*ap->a_retval = 1;
2077 		return (0);
2078 	case _PC_FILESIZEBITS:
2079 		*ap->a_retval = 42;
2080 		return (0);
2081 	case _PC_SYMLINK_MAX:
2082 		*ap->a_retval = MAXPATHLEN;
2083 		return (0);
2084 	case _PC_2_SYMLINKS:
2085 		*ap->a_retval = 1;
2086 		return (0);
2087 	default:
2088 		return (EINVAL);
2089 	}
2090 	/* NOTREACHED */
2091 }
2092 
2093 /*
2094  * Advisory record locking support
2095  */
2096 int
2097 ufs_advlock(void *v)
2098 {
2099 	struct vop_advlock_args /* {
2100 		struct vnode	*a_vp;
2101 		void *		a_id;
2102 		int		a_op;
2103 		struct flock	*a_fl;
2104 		int		a_flags;
2105 	} */ *ap = v;
2106 	struct inode *ip;
2107 
2108 	ip = VTOI(ap->a_vp);
2109 	return lf_advlock(ap, &ip->i_lockf, ip->i_size);
2110 }
2111 
2112 /*
2113  * Initialize the vnode associated with a new inode, handle aliased
2114  * vnodes.
2115  */
2116 void
2117 ufs_vinit(struct mount *mntp, int (**specops)(void *), int (**fifoops)(void *),
2118 	struct vnode **vpp)
2119 {
2120 	struct timeval	tv;
2121 	struct inode	*ip;
2122 	struct vnode	*vp;
2123 	dev_t		rdev;
2124 	struct ufsmount	*ump;
2125 
2126 	vp = *vpp;
2127 	ip = VTOI(vp);
2128 	switch(vp->v_type = IFTOVT(ip->i_mode)) {
2129 	case VCHR:
2130 	case VBLK:
2131 		vp->v_op = specops;
2132 		ump = ip->i_ump;
2133 		if (ump->um_fstype == UFS1)
2134 			rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev,
2135 			    UFS_MPNEEDSWAP(ump));
2136 		else
2137 			rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev,
2138 			    UFS_MPNEEDSWAP(ump));
2139 		spec_node_init(vp, rdev);
2140 		break;
2141 	case VFIFO:
2142 		vp->v_op = fifoops;
2143 		break;
2144 	case VNON:
2145 	case VBAD:
2146 	case VSOCK:
2147 	case VLNK:
2148 	case VDIR:
2149 	case VREG:
2150 		break;
2151 	}
2152 	if (ip->i_number == ROOTINO)
2153                 vp->v_vflag |= VV_ROOT;
2154 	/*
2155 	 * Initialize modrev times
2156 	 */
2157 	getmicrouptime(&tv);
2158 	ip->i_modrev = (uint64_t)(uint)tv.tv_sec << 32
2159 			| tv.tv_usec * 4294u;
2160 	*vpp = vp;
2161 }
2162 
2163 /*
2164  * Allocate a new inode.
2165  */
2166 int
2167 ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
2168 	struct componentname *cnp)
2169 {
2170 	struct inode	*ip, *pdir;
2171 	struct direct	*newdir;
2172 	struct vnode	*tvp;
2173 	int		error, ismember = 0;
2174 
2175 	UFS_WAPBL_JUNLOCK_ASSERT(dvp->v_mount);
2176 
2177 	pdir = VTOI(dvp);
2178 #ifdef DIAGNOSTIC
2179 	if ((cnp->cn_flags & HASBUF) == 0)
2180 		panic("ufs_makeinode: no name");
2181 #endif
2182 	if ((mode & IFMT) == 0)
2183 		mode |= IFREG;
2184 
2185 	if ((error = UFS_VALLOC(dvp, mode, cnp->cn_cred, vpp)) != 0) {
2186 		PNBUF_PUT(cnp->cn_pnbuf);
2187 		vput(dvp);
2188 		return (error);
2189 	}
2190 	tvp = *vpp;
2191 	ip = VTOI(tvp);
2192 	ip->i_gid = pdir->i_gid;
2193 	DIP_ASSIGN(ip, gid, ip->i_gid);
2194 	ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
2195 	DIP_ASSIGN(ip, uid, ip->i_uid);
2196 	error = UFS_WAPBL_BEGIN1(dvp->v_mount, dvp);
2197 	if (error) {
2198 		/*
2199 		 * Note, we can't VOP_VFREE(tvp) here like we should
2200 		 * because we can't write to the disk.  Instead, we leave
2201 		 * the vnode dangling from the journal.
2202 		 */
2203 		vput(tvp);
2204 		PNBUF_PUT(cnp->cn_pnbuf);
2205 		vput(dvp);
2206 		return (error);
2207 	}
2208 #ifdef QUOTA
2209 	if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2210 		UFS_VFREE(tvp, ip->i_number, mode);
2211 		UFS_WAPBL_END1(dvp->v_mount, dvp);
2212 		vput(tvp);
2213 		PNBUF_PUT(cnp->cn_pnbuf);
2214 		vput(dvp);
2215 		return (error);
2216 	}
2217 #endif
2218 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2219 	ip->i_mode = mode;
2220 	DIP_ASSIGN(ip, mode, mode);
2221 	tvp->v_type = IFTOVT(mode);	/* Rest init'd in getnewvnode(). */
2222 	ip->i_nlink = 1;
2223 	DIP_ASSIGN(ip, nlink, 1);
2224 	if ((ip->i_mode & ISGID) && (kauth_cred_ismember_gid(cnp->cn_cred,
2225 	    ip->i_gid, &ismember) != 0 || !ismember) &&
2226 	    kauth_authorize_generic(cnp->cn_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2227 		ip->i_mode &= ~ISGID;
2228 		DIP_ASSIGN(ip, mode, ip->i_mode);
2229 	}
2230 
2231 	if (cnp->cn_flags & ISWHITEOUT) {
2232 		ip->i_flags |= UF_OPAQUE;
2233 		DIP_ASSIGN(ip, flags, ip->i_flags);
2234 	}
2235 
2236 	/*
2237 	 * Make sure inode goes to disk before directory entry.
2238 	 */
2239 	if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0)
2240 		goto bad;
2241 	newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
2242 	ufs_makedirentry(ip, cnp, newdir);
2243 	error = ufs_direnter(dvp, tvp, newdir, cnp, NULL);
2244 	pool_cache_put(ufs_direct_cache, newdir);
2245 	if (error)
2246 		goto bad;
2247 	if ((cnp->cn_flags & SAVESTART) == 0)
2248 		PNBUF_PUT(cnp->cn_pnbuf);
2249 	vput(dvp);
2250 	*vpp = tvp;
2251 	return (0);
2252 
2253  bad:
2254 	/*
2255 	 * Write error occurred trying to update the inode
2256 	 * or the directory so must deallocate the inode.
2257 	 */
2258 	ip->i_nlink = 0;
2259 	DIP_ASSIGN(ip, nlink, 0);
2260 	ip->i_flag |= IN_CHANGE;
2261 	/* If IN_ADIROP, account for it */
2262 	UFS_UNMARK_VNODE(tvp);
2263 	UFS_WAPBL_UPDATE(tvp, NULL, NULL, 0);
2264 	tvp->v_type = VNON;		/* explodes later if VBLK */
2265 	UFS_WAPBL_END1(dvp->v_mount, dvp);
2266 	vput(tvp);
2267 	PNBUF_PUT(cnp->cn_pnbuf);
2268 	vput(dvp);
2269 	return (error);
2270 }
2271 
2272 /*
2273  * Allocate len bytes at offset off.
2274  */
2275 int
2276 ufs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags,
2277     kauth_cred_t cred)
2278 {
2279         struct inode *ip = VTOI(vp);
2280         int error, delta, bshift, bsize;
2281         UVMHIST_FUNC("ufs_gop_alloc"); UVMHIST_CALLED(ubchist);
2282 
2283         error = 0;
2284         bshift = vp->v_mount->mnt_fs_bshift;
2285         bsize = 1 << bshift;
2286 
2287         delta = off & (bsize - 1);
2288         off -= delta;
2289         len += delta;
2290 
2291         while (len > 0) {
2292                 bsize = MIN(bsize, len);
2293 
2294                 error = UFS_BALLOC(vp, off, bsize, cred, flags, NULL);
2295                 if (error) {
2296                         goto out;
2297                 }
2298 
2299                 /*
2300                  * increase file size now, UFS_BALLOC() requires that
2301                  * EOF be up-to-date before each call.
2302                  */
2303 
2304                 if (ip->i_size < off + bsize) {
2305                         UVMHIST_LOG(ubchist, "vp %p old 0x%x new 0x%x",
2306                             vp, ip->i_size, off + bsize, 0);
2307                         ip->i_size = off + bsize;
2308 			DIP_ASSIGN(ip, size, ip->i_size);
2309                 }
2310 
2311                 off += bsize;
2312                 len -= bsize;
2313         }
2314 
2315 out:
2316 	UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
2317 	return error;
2318 }
2319 
2320 void
2321 ufs_gop_markupdate(struct vnode *vp, int flags)
2322 {
2323 	u_int32_t mask = 0;
2324 
2325 	if ((flags & GOP_UPDATE_ACCESSED) != 0) {
2326 		mask = IN_ACCESS;
2327 	}
2328 	if ((flags & GOP_UPDATE_MODIFIED) != 0) {
2329 		if (vp->v_type == VREG) {
2330 			mask |= IN_CHANGE | IN_UPDATE;
2331 		} else {
2332 			mask |= IN_MODIFY;
2333 		}
2334 	}
2335 	if (mask) {
2336 		struct inode *ip = VTOI(vp);
2337 
2338 		ip->i_flag |= mask;
2339 	}
2340 }
2341