ufs_vnops.c revision 38292
1/*
2 * Copyright (c) 1982, 1986, 1989, 1993, 1995
3 *	The Regents of the University of California.  All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *	This product includes software developed by the University of
21 *	California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 *    may be used to endorse or promote products derived from this software
24 *    without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 *	@(#)ufs_vnops.c	8.27 (Berkeley) 5/27/95
39 * $Id: ufs_vnops.c,v 1.98 1998/08/12 20:46:47 julian Exp $
40 */
41
42#include "opt_quota.h"
43#include "opt_suiddir.h"
44
45#include <sys/param.h>
46#include <sys/systm.h>
47#include <sys/namei.h>
48#include <sys/kernel.h>
49#include <sys/fcntl.h>
50#include <sys/stat.h>
51#include <sys/buf.h>
52#include <sys/proc.h>
53#include <sys/mount.h>
54#include <sys/unistd.h>
55#include <sys/vnode.h>
56#include <sys/malloc.h>
57#include <sys/dirent.h>
58#include <sys/lockf.h>
59#include <sys/poll.h>
60
61#include <vm/vm_zone.h>
62#include <vm/vm.h>
63#include <vm/vm_extern.h>
64
65#include <miscfs/specfs/specdev.h>
66#include <miscfs/fifofs/fifo.h>
67
68#include <ufs/ufs/quota.h>
69#include <ufs/ufs/inode.h>
70#include <ufs/ufs/dir.h>
71#include <ufs/ufs/ufsmount.h>
72#include <ufs/ufs/ufs_extern.h>
73
74static int ufs_abortop __P((struct vop_abortop_args *));
75static int ufs_access __P((struct vop_access_args *));
76static int ufs_advlock __P((struct vop_advlock_args *));
77static int ufs_chmod __P((struct vnode *, int, struct ucred *, struct proc *));
78static int ufs_chown __P((struct vnode *, uid_t, gid_t, struct ucred *, struct proc *));
79static int ufs_close __P((struct vop_close_args *));
80static int ufs_create __P((struct vop_create_args *));
81static int ufs_getattr __P((struct vop_getattr_args *));
82static int ufs_link __P((struct vop_link_args *));
83static int ufs_makeinode __P((int mode, struct vnode *, struct vnode **, struct componentname *));
84static int ufs_missingop __P((struct vop_generic_args *ap));
85static int ufs_mkdir __P((struct vop_mkdir_args *));
86static int ufs_mknod __P((struct vop_mknod_args *));
87static int ufs_mmap __P((struct vop_mmap_args *));
88static int ufs_open __P((struct vop_open_args *));
89static int ufs_pathconf __P((struct vop_pathconf_args *));
90static int ufs_print __P((struct vop_print_args *));
91static int ufs_readdir __P((struct vop_readdir_args *));
92static int ufs_readlink __P((struct vop_readlink_args *));
93static int ufs_remove __P((struct vop_remove_args *));
94static int ufs_rename __P((struct vop_rename_args *));
95static int ufs_rmdir __P((struct vop_rmdir_args *));
96static int ufs_setattr __P((struct vop_setattr_args *));
97static int ufs_strategy __P((struct vop_strategy_args *));
98static int ufs_symlink __P((struct vop_symlink_args *));
99static int ufs_whiteout __P((struct vop_whiteout_args *));
100static int ufsfifo_close __P((struct vop_close_args *));
101static int ufsfifo_read __P((struct vop_read_args *));
102static int ufsfifo_write __P((struct vop_write_args *));
103static int ufsspec_close __P((struct vop_close_args *));
104static int ufsspec_read __P((struct vop_read_args *));
105static int ufsspec_write __P((struct vop_write_args *));
106
107union _qcvt {
108	int64_t qcvt;
109	int32_t val[2];
110};
111#define SETHIGH(q, h) { \
112	union _qcvt tmp; \
113	tmp.qcvt = (q); \
114	tmp.val[_QUAD_HIGHWORD] = (h); \
115	(q) = tmp.qcvt; \
116}
117#define SETLOW(q, l) { \
118	union _qcvt tmp; \
119	tmp.qcvt = (q); \
120	tmp.val[_QUAD_LOWWORD] = (l); \
121	(q) = tmp.qcvt; \
122}
123
124/*
125 * A virgin directory (no blushing please).
126 */
127static struct dirtemplate mastertemplate = {
128	0, 12, DT_DIR, 1, ".",
129	0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
130};
131static struct odirtemplate omastertemplate = {
132	0, 12, 1, ".",
133	0, DIRBLKSIZ - 12, 2, ".."
134};
135
136void
137ufs_itimes(vp)
138	struct vnode *vp;
139{
140	struct inode *ip;
141	time_t tv_sec;
142
143	ip = VTOI(vp);
144	if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
145		return;
146	if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
147		tv_sec = time_second;
148		if ((vp->v_type == VBLK || vp->v_type == VCHR) &&
149		    !DOINGSOFTDEP(vp))
150			ip->i_flag |= IN_LAZYMOD;
151		else
152			ip->i_flag |= IN_MODIFIED;
153		if (ip->i_flag & IN_ACCESS)
154			ip->i_atime = tv_sec;
155		if (ip->i_flag & IN_UPDATE) {
156			ip->i_mtime = tv_sec;
157			ip->i_modrev++;
158		}
159		if (ip->i_flag & IN_CHANGE)
160			ip->i_ctime = tv_sec;
161	}
162	ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
163}
164
165/*
166 * Create a regular file
167 */
168int
169ufs_create(ap)
170	struct vop_create_args /* {
171		struct vnode *a_dvp;
172		struct vnode **a_vpp;
173		struct componentname *a_cnp;
174		struct vattr *a_vap;
175	} */ *ap;
176{
177	int error;
178
179	error =
180	    ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
181	    ap->a_dvp, ap->a_vpp, ap->a_cnp);
182	if (error)
183		return (error);
184	VN_POLLEVENT(ap->a_dvp, POLLWRITE);
185	return (0);
186}
187
188/*
189 * Mknod vnode call
190 */
191/* ARGSUSED */
192int
193ufs_mknod(ap)
194	struct vop_mknod_args /* {
195		struct vnode *a_dvp;
196		struct vnode **a_vpp;
197		struct componentname *a_cnp;
198		struct vattr *a_vap;
199	} */ *ap;
200{
201	struct vattr *vap = ap->a_vap;
202	struct vnode **vpp = ap->a_vpp;
203	struct inode *ip;
204	int error;
205
206	error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
207	    ap->a_dvp, vpp, ap->a_cnp);
208	if (error)
209		return (error);
210	VN_POLLEVENT(ap->a_dvp, POLLWRITE);
211	ip = VTOI(*vpp);
212	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
213	if (vap->va_rdev != VNOVAL) {
214		/*
215		 * Want to be able to use this to make badblock
216		 * inodes, so don't truncate the dev number.
217		 */
218		ip->i_rdev = vap->va_rdev;
219	}
220	/*
221	 * Remove inode so that it will be reloaded by VFS_VGET and
222	 * checked to see if it is an alias of an existing entry in
223	 * the inode cache.
224	 */
225	vput(*vpp);
226	(*vpp)->v_type = VNON;
227	vgone(*vpp);
228	*vpp = 0;
229	return (0);
230}
231
232/*
233 * Open called.
234 *
235 * Nothing to do.
236 */
237/* ARGSUSED */
238int
239ufs_open(ap)
240	struct vop_open_args /* {
241		struct vnode *a_vp;
242		int  a_mode;
243		struct ucred *a_cred;
244		struct proc *a_p;
245	} */ *ap;
246{
247
248	/*
249	 * Files marked append-only must be opened for appending.
250	 */
251	if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
252	    (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
253		return (EPERM);
254	return (0);
255}
256
257/*
258 * Close called.
259 *
260 * Update the times on the inode.
261 */
262/* ARGSUSED */
263int
264ufs_close(ap)
265	struct vop_close_args /* {
266		struct vnode *a_vp;
267		int  a_fflag;
268		struct ucred *a_cred;
269		struct proc *a_p;
270	} */ *ap;
271{
272	register struct vnode *vp = ap->a_vp;
273
274	simple_lock(&vp->v_interlock);
275	if (vp->v_usecount > 1)
276		ufs_itimes(vp);
277	simple_unlock(&vp->v_interlock);
278	return (0);
279}
280
281int
282ufs_access(ap)
283	struct vop_access_args /* {
284		struct vnode *a_vp;
285		int  a_mode;
286		struct ucred *a_cred;
287		struct proc *a_p;
288	} */ *ap;
289{
290	struct vnode *vp = ap->a_vp;
291	struct inode *ip = VTOI(vp);
292	struct ucred *cred = ap->a_cred;
293	mode_t mask, mode = ap->a_mode;
294	register gid_t *gp;
295	int i;
296#ifdef QUOTA
297	int error;
298#endif
299
300	/*
301	 * Disallow write attempts on read-only file systems;
302	 * unless the file is a socket, fifo, or a block or
303	 * character device resident on the file system.
304	 */
305	if (mode & VWRITE) {
306		switch (vp->v_type) {
307		case VDIR:
308		case VLNK:
309		case VREG:
310			if (vp->v_mount->mnt_flag & MNT_RDONLY)
311				return (EROFS);
312#ifdef QUOTA
313			if (error = getinoquota(ip))
314				return (error);
315#endif
316			break;
317		default:
318			break;
319		}
320	}
321
322	/* If immutable bit set, nobody gets to write it. */
323	if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
324		return (EPERM);
325
326	/* Otherwise, user id 0 always gets access. */
327	if (cred->cr_uid == 0)
328		return (0);
329
330	mask = 0;
331
332	/* Otherwise, check the owner. */
333	if (cred->cr_uid == ip->i_uid) {
334		if (mode & VEXEC)
335			mask |= S_IXUSR;
336		if (mode & VREAD)
337			mask |= S_IRUSR;
338		if (mode & VWRITE)
339			mask |= S_IWUSR;
340		return ((ip->i_mode & mask) == mask ? 0 : EACCES);
341	}
342
343	/* Otherwise, check the groups. */
344	for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
345		if (ip->i_gid == *gp) {
346			if (mode & VEXEC)
347				mask |= S_IXGRP;
348			if (mode & VREAD)
349				mask |= S_IRGRP;
350			if (mode & VWRITE)
351				mask |= S_IWGRP;
352			return ((ip->i_mode & mask) == mask ? 0 : EACCES);
353		}
354
355	/* Otherwise, check everyone else. */
356	if (mode & VEXEC)
357		mask |= S_IXOTH;
358	if (mode & VREAD)
359		mask |= S_IROTH;
360	if (mode & VWRITE)
361		mask |= S_IWOTH;
362	return ((ip->i_mode & mask) == mask ? 0 : EACCES);
363}
364
365/* ARGSUSED */
366int
367ufs_getattr(ap)
368	struct vop_getattr_args /* {
369		struct vnode *a_vp;
370		struct vattr *a_vap;
371		struct ucred *a_cred;
372		struct proc *a_p;
373	} */ *ap;
374{
375	register struct vnode *vp = ap->a_vp;
376	register struct inode *ip = VTOI(vp);
377	register struct vattr *vap = ap->a_vap;
378
379	ufs_itimes(vp);
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 & ~IFMT;
386	vap->va_nlink = ip->i_effnlink;
387	vap->va_uid = ip->i_uid;
388	vap->va_gid = ip->i_gid;
389	vap->va_rdev = (dev_t)ip->i_rdev;
390	vap->va_size = ip->i_din.di_size;
391	vap->va_atime.tv_sec = ip->i_atime;
392	vap->va_atime.tv_nsec = ip->i_atimensec;
393	vap->va_mtime.tv_sec = ip->i_mtime;
394	vap->va_mtime.tv_nsec = ip->i_mtimensec;
395	vap->va_ctime.tv_sec = ip->i_ctime;
396	vap->va_ctime.tv_nsec = ip->i_ctimensec;
397	vap->va_flags = ip->i_flags;
398	vap->va_gen = ip->i_gen;
399	/* this doesn't belong here */
400	if (vp->v_type == VBLK)
401		vap->va_blocksize = BLKDEV_IOSIZE;
402	else if (vp->v_type == VCHR)
403		vap->va_blocksize = MAXBSIZE;
404	else
405		vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
406	vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
407	vap->va_type = IFTOVT(ip->i_mode);
408	vap->va_filerev = ip->i_modrev;
409	return (0);
410}
411
412/*
413 * Set attribute vnode op. called from several syscalls
414 */
415int
416ufs_setattr(ap)
417	struct vop_setattr_args /* {
418		struct vnode *a_vp;
419		struct vattr *a_vap;
420		struct ucred *a_cred;
421		struct proc *a_p;
422	} */ *ap;
423{
424	struct vattr *vap = ap->a_vap;
425	struct vnode *vp = ap->a_vp;
426	struct inode *ip = VTOI(vp);
427	struct ucred *cred = ap->a_cred;
428	struct proc *p = ap->a_p;
429	int error;
430
431	/*
432	 * Check for unsettable attributes.
433	 */
434	if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
435	    (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
436	    (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
437	    ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
438		return (EINVAL);
439	}
440	if (vap->va_flags != VNOVAL) {
441		if (vp->v_mount->mnt_flag & MNT_RDONLY)
442			return (EROFS);
443		if (cred->cr_uid != ip->i_uid &&
444		    (error = suser(cred, &p->p_acflag)))
445			return (error);
446		if (cred->cr_uid == 0) {
447			if ((ip->i_flags
448			    & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
449			    securelevel > 0)
450				return (EPERM);
451			ip->i_flags = vap->va_flags;
452		} else {
453			if (ip->i_flags
454			    & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
455			    (vap->va_flags & UF_SETTABLE) != vap->va_flags)
456				return (EPERM);
457			ip->i_flags &= SF_SETTABLE;
458			ip->i_flags |= (vap->va_flags & UF_SETTABLE);
459		}
460		ip->i_flag |= IN_CHANGE;
461		if (vap->va_flags & (IMMUTABLE | APPEND))
462			return (0);
463	}
464	if (ip->i_flags & (IMMUTABLE | APPEND))
465		return (EPERM);
466	/*
467	 * Go through the fields and update iff not VNOVAL.
468	 */
469	if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
470		if (vp->v_mount->mnt_flag & MNT_RDONLY)
471			return (EROFS);
472		if (error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, p))
473			return (error);
474	}
475	if (vap->va_size != VNOVAL) {
476		/*
477		 * Disallow write attempts on read-only file systems;
478		 * unless the file is a socket, fifo, or a block or
479		 * character device resident on the file system.
480		 */
481		switch (vp->v_type) {
482		case VDIR:
483			return (EISDIR);
484		case VLNK:
485		case VREG:
486			if (vp->v_mount->mnt_flag & MNT_RDONLY)
487				return (EROFS);
488			break;
489		default:
490			break;
491		}
492		if (error = UFS_TRUNCATE(vp, vap->va_size, 0, cred, p))
493			return (error);
494	}
495	ip = VTOI(vp);
496	if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
497		if (vp->v_mount->mnt_flag & MNT_RDONLY)
498			return (EROFS);
499		if (cred->cr_uid != ip->i_uid &&
500		    (error = suser(cred, &p->p_acflag)) &&
501		    ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
502		    (error = VOP_ACCESS(vp, VWRITE, cred, p))))
503			return (error);
504		if (vap->va_atime.tv_sec != VNOVAL)
505			ip->i_flag |= IN_ACCESS;
506		if (vap->va_mtime.tv_sec != VNOVAL)
507			ip->i_flag |= IN_CHANGE | IN_UPDATE;
508		ufs_itimes(vp);
509		if (vap->va_mtime.tv_sec != VNOVAL)
510			ip->i_atime = vap->va_atime.tv_sec;
511		if (vap->va_atime.tv_sec != VNOVAL)
512			ip->i_mtime = vap->va_mtime.tv_sec;
513		error = UFS_UPDATE(vp, (struct timeval *)0,
514		    (struct timeval *)0, 0);
515		if (error)
516			return (error);
517	}
518	error = 0;
519	if (vap->va_mode != (mode_t)VNOVAL) {
520		if (vp->v_mount->mnt_flag & MNT_RDONLY)
521			return (EROFS);
522		error = ufs_chmod(vp, (int)vap->va_mode, cred, p);
523	}
524	VN_POLLEVENT(vp, POLLATTRIB);
525	return (error);
526}
527
528/*
529 * Change the mode on a file.
530 * Inode must be locked before calling.
531 */
532static int
533ufs_chmod(vp, mode, cred, p)
534	register struct vnode *vp;
535	register int mode;
536	register struct ucred *cred;
537	struct proc *p;
538{
539	register struct inode *ip = VTOI(vp);
540	int error;
541
542	if (cred->cr_uid != ip->i_uid) {
543	    error = suser(cred, &p->p_acflag);
544	    if (error)
545		return (error);
546	}
547	if (cred->cr_uid) {
548		if (vp->v_type != VDIR && (mode & S_ISTXT))
549			return (EFTYPE);
550		if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
551			return (EPERM);
552	}
553	ip->i_mode &= ~ALLPERMS;
554	ip->i_mode |= (mode & ALLPERMS);
555	ip->i_flag |= IN_CHANGE;
556	return (0);
557}
558
559/*
560 * Perform chown operation on inode ip;
561 * inode must be locked prior to call.
562 */
563static int
564ufs_chown(vp, uid, gid, cred, p)
565	register struct vnode *vp;
566	uid_t uid;
567	gid_t gid;
568	struct ucred *cred;
569	struct proc *p;
570{
571	register struct inode *ip = VTOI(vp);
572	uid_t ouid;
573	gid_t ogid;
574	int error = 0;
575#ifdef QUOTA
576	register int i;
577	long change;
578#endif
579
580	if (uid == (uid_t)VNOVAL)
581		uid = ip->i_uid;
582	if (gid == (gid_t)VNOVAL)
583		gid = ip->i_gid;
584	/*
585	 * If we don't own the file, are trying to change the owner
586	 * of the file, or are not a member of the target group,
587	 * the caller must be superuser or the call fails.
588	 */
589	if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
590	    (gid != ip->i_gid && !groupmember((gid_t)gid, cred))) &&
591	    (error = suser(cred, &p->p_acflag)))
592		return (error);
593	ogid = ip->i_gid;
594	ouid = ip->i_uid;
595#ifdef QUOTA
596	if (error = getinoquota(ip))
597		return (error);
598	if (ouid == uid) {
599		dqrele(vp, ip->i_dquot[USRQUOTA]);
600		ip->i_dquot[USRQUOTA] = NODQUOT;
601	}
602	if (ogid == gid) {
603		dqrele(vp, ip->i_dquot[GRPQUOTA]);
604		ip->i_dquot[GRPQUOTA] = NODQUOT;
605	}
606	change = ip->i_blocks;
607	(void) chkdq(ip, -change, cred, CHOWN);
608	(void) chkiq(ip, -1, cred, CHOWN);
609	for (i = 0; i < MAXQUOTAS; i++) {
610		dqrele(vp, ip->i_dquot[i]);
611		ip->i_dquot[i] = NODQUOT;
612	}
613#endif
614	ip->i_gid = gid;
615	ip->i_uid = uid;
616#ifdef QUOTA
617	if ((error = getinoquota(ip)) == 0) {
618		if (ouid == uid) {
619			dqrele(vp, ip->i_dquot[USRQUOTA]);
620			ip->i_dquot[USRQUOTA] = NODQUOT;
621		}
622		if (ogid == gid) {
623			dqrele(vp, ip->i_dquot[GRPQUOTA]);
624			ip->i_dquot[GRPQUOTA] = NODQUOT;
625		}
626		if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
627			if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
628				goto good;
629			else
630				(void) chkdq(ip, -change, cred, CHOWN|FORCE);
631		}
632		for (i = 0; i < MAXQUOTAS; i++) {
633			dqrele(vp, ip->i_dquot[i]);
634			ip->i_dquot[i] = NODQUOT;
635		}
636	}
637	ip->i_gid = ogid;
638	ip->i_uid = ouid;
639	if (getinoquota(ip) == 0) {
640		if (ouid == uid) {
641			dqrele(vp, ip->i_dquot[USRQUOTA]);
642			ip->i_dquot[USRQUOTA] = NODQUOT;
643		}
644		if (ogid == gid) {
645			dqrele(vp, ip->i_dquot[GRPQUOTA]);
646			ip->i_dquot[GRPQUOTA] = NODQUOT;
647		}
648		(void) chkdq(ip, change, cred, FORCE|CHOWN);
649		(void) chkiq(ip, 1, cred, FORCE|CHOWN);
650		(void) getinoquota(ip);
651	}
652	return (error);
653good:
654	if (getinoquota(ip))
655		panic("ufs_chown: lost quota");
656#endif /* QUOTA */
657	ip->i_flag |= IN_CHANGE;
658	if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
659		ip->i_mode &= ~(ISUID | ISGID);
660	return (0);
661}
662
663/*
664 * Mmap a file
665 *
666 * NB Currently unsupported.
667 */
668/* ARGSUSED */
669int
670ufs_mmap(ap)
671	struct vop_mmap_args /* {
672		struct vnode *a_vp;
673		int  a_fflags;
674		struct ucred *a_cred;
675		struct proc *a_p;
676	} */ *ap;
677{
678
679	return (EINVAL);
680}
681
682int
683ufs_remove(ap)
684	struct vop_remove_args /* {
685		struct vnode *a_dvp;
686		struct vnode *a_vp;
687		struct componentname *a_cnp;
688	} */ *ap;
689{
690	struct inode *ip;
691	struct vnode *vp = ap->a_vp;
692	struct vnode *dvp = ap->a_dvp;
693	int error;
694
695	ip = VTOI(vp);
696	if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
697	    (VTOI(dvp)->i_flags & APPEND)) {
698		error = EPERM;
699		goto out;
700	}
701	error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
702	VN_POLLEVENT(vp, POLLNLINK);
703	VN_POLLEVENT(dvp, POLLWRITE);
704out:
705	return (error);
706}
707
708/*
709 * link vnode call
710 */
711int
712ufs_link(ap)
713	struct vop_link_args /* {
714		struct vnode *a_tdvp;
715		struct vnode *a_vp;
716		struct componentname *a_cnp;
717	} */ *ap;
718{
719	struct vnode *vp = ap->a_vp;
720	struct vnode *tdvp = ap->a_tdvp;
721	struct componentname *cnp = ap->a_cnp;
722	struct proc *p = cnp->cn_proc;
723	struct inode *ip;
724	struct timeval tv;
725	struct direct newdir;
726	int error;
727
728#ifdef DIAGNOSTIC
729	if ((cnp->cn_flags & HASBUF) == 0)
730		panic("ufs_link: no name");
731#endif
732	if (tdvp->v_mount != vp->v_mount) {
733		VOP_ABORTOP(tdvp, cnp);
734		error = EXDEV;
735		goto out2;
736	}
737	if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, p))) {
738		VOP_ABORTOP(tdvp, cnp);
739		goto out2;
740	}
741	ip = VTOI(vp);
742	if ((nlink_t)ip->i_nlink >= LINK_MAX) {
743		VOP_ABORTOP(tdvp, cnp);
744		error = EMLINK;
745		goto out1;
746	}
747	if (ip->i_flags & (IMMUTABLE | APPEND)) {
748		VOP_ABORTOP(tdvp, cnp);
749		error = EPERM;
750		goto out1;
751	}
752	ip->i_effnlink++;
753	ip->i_nlink++;
754	ip->i_flag |= IN_CHANGE;
755	if (DOINGSOFTDEP(vp))
756		softdep_increase_linkcnt(ip);
757	getmicrotime(&tv);
758	error = UFS_UPDATE(vp, &tv, &tv, !DOINGSOFTDEP(vp));
759	if (!error) {
760		ufs_makedirentry(ip, cnp, &newdir);
761		error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
762	}
763
764	if (error) {
765		ip->i_effnlink--;
766		ip->i_nlink--;
767		ip->i_flag |= IN_CHANGE;
768	}
769	zfree(namei_zone, cnp->cn_pnbuf);
770out1:
771	if (tdvp != vp)
772		VOP_UNLOCK(vp, 0, p);
773out2:
774	VN_POLLEVENT(vp, POLLNLINK);
775	VN_POLLEVENT(tdvp, POLLWRITE);
776	return (error);
777}
778
779/*
780 * whiteout vnode call
781 */
782int
783ufs_whiteout(ap)
784	struct vop_whiteout_args /* {
785		struct vnode *a_dvp;
786		struct componentname *a_cnp;
787		int a_flags;
788	} */ *ap;
789{
790	struct vnode *dvp = ap->a_dvp;
791	struct componentname *cnp = ap->a_cnp;
792	struct direct newdir;
793	int error = 0;
794
795	switch (ap->a_flags) {
796	case LOOKUP:
797		/* 4.4 format directories support whiteout operations */
798		if (dvp->v_mount->mnt_maxsymlinklen > 0)
799			return (0);
800		return (EOPNOTSUPP);
801
802	case CREATE:
803		/* create a new directory whiteout */
804#ifdef DIAGNOSTIC
805		if ((cnp->cn_flags & SAVENAME) == 0)
806			panic("ufs_whiteout: missing name");
807		if (dvp->v_mount->mnt_maxsymlinklen <= 0)
808			panic("ufs_whiteout: old format filesystem");
809#endif
810
811		newdir.d_ino = WINO;
812		newdir.d_namlen = cnp->cn_namelen;
813		bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
814		newdir.d_type = DT_WHT;
815		error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
816		break;
817
818	case DELETE:
819		/* remove an existing directory whiteout */
820#ifdef DIAGNOSTIC
821		if (dvp->v_mount->mnt_maxsymlinklen <= 0)
822			panic("ufs_whiteout: old format filesystem");
823#endif
824
825		cnp->cn_flags &= ~DOWHITEOUT;
826		error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
827		break;
828	default:
829		panic("ufs_whiteout: unknown op");
830	}
831	if (cnp->cn_flags & HASBUF) {
832		zfree(namei_zone, cnp->cn_pnbuf);
833		cnp->cn_flags &= ~HASBUF;
834	}
835	return (error);
836}
837
838/*
839 * Rename system call.
840 * 	rename("foo", "bar");
841 * is essentially
842 *	unlink("bar");
843 *	link("foo", "bar");
844 *	unlink("foo");
845 * but ``atomically''.  Can't do full commit without saving state in the
846 * inode on disk which isn't feasible at this time.  Best we can do is
847 * always guarantee the target exists.
848 *
849 * Basic algorithm is:
850 *
851 * 1) Bump link count on source while we're linking it to the
852 *    target.  This also ensure the inode won't be deleted out
853 *    from underneath us while we work (it may be truncated by
854 *    a concurrent `trunc' or `open' for creation).
855 * 2) Link source to destination.  If destination already exists,
856 *    delete it first.
857 * 3) Unlink source reference to inode if still around. If a
858 *    directory was moved and the parent of the destination
859 *    is different from the source, patch the ".." entry in the
860 *    directory.
861 */
862int
863ufs_rename(ap)
864	struct vop_rename_args  /* {
865		struct vnode *a_fdvp;
866		struct vnode *a_fvp;
867		struct componentname *a_fcnp;
868		struct vnode *a_tdvp;
869		struct vnode *a_tvp;
870		struct componentname *a_tcnp;
871	} */ *ap;
872{
873	struct vnode *tvp = ap->a_tvp;
874	register struct vnode *tdvp = ap->a_tdvp;
875	struct vnode *fvp = ap->a_fvp;
876	struct vnode *fdvp = ap->a_fdvp;
877	struct componentname *tcnp = ap->a_tcnp;
878	struct componentname *fcnp = ap->a_fcnp;
879	struct proc *p = fcnp->cn_proc;
880	struct inode *ip, *xp, *dp;
881	struct direct newdir;
882	struct timeval tv;
883	int doingdirectory = 0, oldparent = 0, newparent = 0;
884	int error = 0;
885
886#ifdef DIAGNOSTIC
887	if ((tcnp->cn_flags & HASBUF) == 0 ||
888	    (fcnp->cn_flags & HASBUF) == 0)
889		panic("ufs_rename: no name");
890#endif
891	/*
892	 * Check for cross-device rename.
893	 */
894	if ((fvp->v_mount != tdvp->v_mount) ||
895	    (tvp && (fvp->v_mount != tvp->v_mount))) {
896		error = EXDEV;
897abortit:
898		VOP_ABORTOP(tdvp, tcnp); /* XXX, why not in NFS? */
899		if (tdvp == tvp)
900			vrele(tdvp);
901		else
902			vput(tdvp);
903		if (tvp)
904			vput(tvp);
905		VOP_ABORTOP(fdvp, fcnp); /* XXX, why not in NFS? */
906		vrele(fdvp);
907		vrele(fvp);
908		return (error);
909	}
910
911	if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
912	    (VTOI(tdvp)->i_flags & APPEND))) {
913		error = EPERM;
914		goto abortit;
915	}
916
917	/*
918	 * Check if just deleting a link name or if we've lost a race.
919	 * If another process completes the same rename after we've looked
920	 * up the source and have blocked looking up the target, then the
921	 * source and target inodes may be identical now although the
922	 * names were never linked.
923	 */
924	if (fvp == tvp) {
925		if (fvp->v_type == VDIR) {
926			/*
927			 * Linked directories are impossible, so we must
928			 * have lost the race.  Pretend that the rename
929			 * completed before the lookup.
930			 */
931#ifdef UFS_RENAME_DEBUG
932			printf("ufs_rename: fvp == tvp for directories\n");
933#endif
934			error = ENOENT;
935			goto abortit;
936		}
937
938		/* Release destination completely. */
939		VOP_ABORTOP(tdvp, tcnp);
940		vput(tdvp);
941		vput(tvp);
942
943		/*
944		 * Delete source.  There is another race now that everything
945		 * is unlocked, but this doesn't cause any new complications.
946		 * Relookup() may find a file that is unrelated to the
947		 * original one, or it may fail.  Too bad.
948		 */
949		vrele(fdvp);
950		vrele(fvp);
951		fcnp->cn_flags &= ~MODMASK;
952		fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
953		if ((fcnp->cn_flags & SAVESTART) == 0)
954			panic("ufs_rename: lost from startdir");
955		fcnp->cn_nameiop = DELETE;
956		VREF(fdvp);
957		error = relookup(fdvp, &fvp, fcnp);
958		if (error == 0)
959			vrele(fdvp);
960		if (fvp == NULL) {
961#ifdef UFS_RENAME_DEBUG
962			printf("ufs_rename: from name disappeared\n");
963#endif
964			return (ENOENT);
965		}
966		error = VOP_REMOVE(fdvp, fvp, fcnp);
967		if (fdvp == fvp)
968			vrele(fdvp);
969		else
970			vput(fdvp);
971		vput(fvp);
972		return (error);
973	}
974	if (error = vn_lock(fvp, LK_EXCLUSIVE, p))
975		goto abortit;
976	dp = VTOI(fdvp);
977	ip = VTOI(fvp);
978	if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
979	    || (dp->i_flags & APPEND)) {
980		VOP_UNLOCK(fvp, 0, p);
981		error = EPERM;
982		goto abortit;
983	}
984	if ((ip->i_mode & IFMT) == IFDIR) {
985		/*
986		 * Avoid ".", "..", and aliases of "." for obvious reasons.
987		 */
988		if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
989		    dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
990		    (ip->i_flag & IN_RENAME)) {
991			VOP_UNLOCK(fvp, 0, p);
992			error = EINVAL;
993			goto abortit;
994		}
995		ip->i_flag |= IN_RENAME;
996		oldparent = dp->i_number;
997		doingdirectory = 1;
998	}
999	VN_POLLEVENT(fdvp, POLLWRITE);
1000	vrele(fdvp);
1001
1002	/*
1003	 * When the target exists, both the directory
1004	 * and target vnodes are returned locked.
1005	 */
1006	dp = VTOI(tdvp);
1007	xp = NULL;
1008	if (tvp)
1009		xp = VTOI(tvp);
1010
1011	/*
1012	 * 1) Bump link count while we're moving stuff
1013	 *    around.  If we crash somewhere before
1014	 *    completing our work, the link count
1015	 *    may be wrong, but correctable.
1016	 */
1017	ip->i_effnlink++;
1018	ip->i_nlink++;
1019	ip->i_flag |= IN_CHANGE;
1020	if (DOINGSOFTDEP(fvp))
1021		softdep_increase_linkcnt(ip);
1022	getmicrotime(&tv);
1023	if (error = UFS_UPDATE(fvp, &tv, &tv, !DOINGSOFTDEP(fvp))) {
1024		VOP_UNLOCK(fvp, 0, p);
1025		goto bad;
1026	}
1027
1028	/*
1029	 * If ".." must be changed (ie the directory gets a new
1030	 * parent) then the source directory must not be in the
1031	 * directory heirarchy above the target, as this would
1032	 * orphan everything below the source directory. Also
1033	 * the user must have write permission in the source so
1034	 * as to be able to change "..". We must repeat the call
1035	 * to namei, as the parent directory is unlocked by the
1036	 * call to checkpath().
1037	 */
1038	error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_proc);
1039	VOP_UNLOCK(fvp, 0, p);
1040	if (oldparent != dp->i_number)
1041		newparent = dp->i_number;
1042	if (doingdirectory && newparent) {
1043		if (error)	/* write access check above */
1044			goto bad;
1045		if (xp != NULL)
1046			vput(tvp);
1047		error = ufs_checkpath(ip, dp, tcnp->cn_cred);
1048		if (error)
1049			goto out;
1050		if ((tcnp->cn_flags & SAVESTART) == 0)
1051			panic("ufs_rename: lost to startdir");
1052		VREF(tdvp);
1053		error = relookup(tdvp, &tvp, tcnp);
1054		if (error)
1055			goto out;
1056		vrele(tdvp);
1057		dp = VTOI(tdvp);
1058		xp = NULL;
1059		if (tvp)
1060			xp = VTOI(tvp);
1061	}
1062	/*
1063	 * 2) If target doesn't exist, link the target
1064	 *    to the source and unlink the source.
1065	 *    Otherwise, rewrite the target directory
1066	 *    entry to reference the source inode and
1067	 *    expunge the original entry's existence.
1068	 */
1069	if (xp == NULL) {
1070		if (dp->i_dev != ip->i_dev)
1071			panic("ufs_rename: EXDEV");
1072		/*
1073		 * Account for ".." in new directory.
1074		 * When source and destination have the same
1075		 * parent we don't fool with the link count.
1076		 */
1077		if (doingdirectory && newparent) {
1078			if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1079				error = EMLINK;
1080				goto bad;
1081			}
1082			dp->i_effnlink++;
1083			dp->i_nlink++;
1084			dp->i_flag |= IN_CHANGE;
1085			if (DOINGSOFTDEP(tdvp))
1086				softdep_increase_linkcnt(dp);
1087			error = UFS_UPDATE(tdvp, &tv, &tv, !DOINGSOFTDEP(tdvp));
1088			if (error)
1089				goto bad;
1090		}
1091		ufs_makedirentry(ip, tcnp, &newdir);
1092		error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
1093		if (error) {
1094			if (doingdirectory && newparent) {
1095				dp->i_effnlink--;
1096				dp->i_nlink--;
1097				dp->i_flag |= IN_CHANGE;
1098				(void)UFS_UPDATE(tdvp, &tv, &tv, 1);
1099			}
1100			goto bad;
1101		}
1102		VN_POLLEVENT(tdvp, POLLWRITE);
1103		vput(tdvp);
1104	} else {
1105		if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
1106			panic("ufs_rename: EXDEV");
1107		/*
1108		 * Short circuit rename(foo, foo).
1109		 */
1110		if (xp->i_number == ip->i_number)
1111			panic("ufs_rename: same file");
1112		/*
1113		 * If the parent directory is "sticky", then the user must
1114		 * own the parent directory, or the destination of the rename,
1115		 * otherwise the destination may not be changed (except by
1116		 * root). This implements append-only directories.
1117		 */
1118		if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
1119		    tcnp->cn_cred->cr_uid != dp->i_uid &&
1120		    xp->i_uid != tcnp->cn_cred->cr_uid) {
1121			error = EPERM;
1122			goto bad;
1123		}
1124		/*
1125		 * Target must be empty if a directory and have no links
1126		 * to it. Also, ensure source and target are compatible
1127		 * (both directories, or both not directories).
1128		 */
1129		if ((xp->i_mode&IFMT) == IFDIR) {
1130			if ((xp->i_effnlink > 2) ||
1131			    !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
1132				error = ENOTEMPTY;
1133				goto bad;
1134			}
1135			if (!doingdirectory) {
1136				error = ENOTDIR;
1137				goto bad;
1138			}
1139			cache_purge(tdvp);
1140		} else if (doingdirectory) {
1141			error = EISDIR;
1142			goto bad;
1143		}
1144		error = ufs_dirrewrite(dp, xp, ip->i_number,
1145		    IFTODT(ip->i_mode),
1146		    (doingdirectory && newparent) ? newparent : doingdirectory);
1147		if (error)
1148			goto bad;
1149		if (doingdirectory) {
1150			if (!newparent) {
1151				dp->i_effnlink--;
1152				dp->i_flag |= IN_CHANGE;
1153			}
1154			xp->i_effnlink--;
1155			xp->i_flag |= IN_CHANGE;
1156		}
1157		VN_POLLEVENT(tdvp, POLLWRITE);
1158		if (doingdirectory && !DOINGSOFTDEP(tvp)) {
1159			/*
1160			 * Truncate inode. The only stuff left in the directory
1161			 * is "." and "..". The "." reference is inconsequential
1162			 * since we are quashing it. We have removed the "."
1163			 * reference and the reference in the parent directory,
1164			 * but there may be other hard links. The soft
1165			 * dependency code will arrange to do these operations
1166			 * after the parent directory entry has been deleted on
1167			 * disk, so when running with that code we avoid doing
1168			 * them now.
1169			 */
1170			if (!newparent)
1171				dp->i_nlink--;
1172			xp->i_nlink--;
1173			if ((error = UFS_TRUNCATE(tvp, (off_t)0, IO_SYNC,
1174			    tcnp->cn_cred, tcnp->cn_proc)) != 0)
1175				goto bad;
1176		}
1177		vput(tdvp);
1178		VN_POLLEVENT(tvp, POLLNLINK); /* XXX this right? */
1179		vput(tvp);
1180		xp = NULL;
1181	}
1182
1183	/*
1184	 * 3) Unlink the source.
1185	 */
1186	fcnp->cn_flags &= ~MODMASK;
1187	fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1188	if ((fcnp->cn_flags & SAVESTART) == 0)
1189		panic("ufs_rename: lost from startdir");
1190	VREF(fdvp);
1191	error = relookup(fdvp, &fvp, fcnp);
1192	if (error == 0)
1193		vrele(fdvp);
1194	if (fvp != NULL) {
1195		xp = VTOI(fvp);
1196		dp = VTOI(fdvp);
1197	} else {
1198		/*
1199		 * From name has disappeared.
1200		 */
1201		if (doingdirectory)
1202			panic("ufs_rename: lost dir entry");
1203		vrele(ap->a_fvp);
1204		return (0);
1205	}
1206	/*
1207	 * Ensure that the directory entry still exists and has not
1208	 * changed while the new name has been entered. If the source is
1209	 * a file then the entry may have been unlinked or renamed. In
1210	 * either case there is no further work to be done. If the source
1211	 * is a directory then it cannot have been rmdir'ed; the IN_RENAME
1212	 * flag ensures that it cannot be moved by another rename or removed
1213	 * by a rmdir.
1214	 */
1215	if (xp != ip) {
1216		if (doingdirectory)
1217			panic("ufs_rename: lost dir entry");
1218	} else {
1219		/*
1220		 * If the source is a directory with a
1221		 * new parent, the link count of the old
1222		 * parent directory must be decremented
1223		 * and ".." set to point to the new parent.
1224		 */
1225		if (doingdirectory && newparent) {
1226			xp->i_offset = mastertemplate.dot_reclen;
1227			ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
1228			cache_purge(fdvp);
1229		}
1230		error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
1231		xp->i_flag &= ~IN_RENAME;
1232	}
1233	if (dp)
1234		vput(fdvp);
1235	if (xp)
1236		vput(fvp);
1237	vrele(ap->a_fvp);
1238	return (error);
1239
1240bad:
1241	if (xp)
1242		vput(ITOV(xp));
1243	vput(ITOV(dp));
1244out:
1245	if (doingdirectory)
1246		ip->i_flag &= ~IN_RENAME;
1247	if (vn_lock(fvp, LK_EXCLUSIVE, p) == 0) {
1248		ip->i_effnlink--;
1249		ip->i_nlink--;
1250		ip->i_flag |= IN_CHANGE;
1251		ip->i_flag &= ~IN_RENAME;
1252		vput(fvp);
1253	} else
1254		vrele(fvp);
1255	return (error);
1256}
1257
1258/*
1259 * Mkdir system call
1260 */
1261int
1262ufs_mkdir(ap)
1263	struct vop_mkdir_args /* {
1264		struct vnode *a_dvp;
1265		struct vnode **a_vpp;
1266		struct componentname *a_cnp;
1267		struct vattr *a_vap;
1268	} */ *ap;
1269{
1270	register struct vnode *dvp = ap->a_dvp;
1271	register struct vattr *vap = ap->a_vap;
1272	register struct componentname *cnp = ap->a_cnp;
1273	register struct inode *ip, *dp;
1274	struct vnode *tvp;
1275	struct buf *bp;
1276	struct dirtemplate dirtemplate, *dtp;
1277	struct direct newdir;
1278	struct timeval tv;
1279	int error, dmode;
1280	long blkoff;
1281
1282#ifdef DIAGNOSTIC
1283	if ((cnp->cn_flags & HASBUF) == 0)
1284		panic("ufs_mkdir: no name");
1285#endif
1286	dp = VTOI(dvp);
1287	if ((nlink_t)dp->i_nlink >= LINK_MAX) {
1288		error = EMLINK;
1289		goto out;
1290	}
1291	dmode = vap->va_mode & 0777;
1292	dmode |= IFDIR;
1293	/*
1294	 * Must simulate part of ufs_makeinode here to acquire the inode,
1295	 * but not have it entered in the parent directory. The entry is
1296	 * made later after writing "." and ".." entries.
1297	 */
1298	error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
1299	if (error)
1300		goto out;
1301	ip = VTOI(tvp);
1302	ip->i_gid = dp->i_gid;
1303#ifdef SUIDDIR
1304	{
1305#ifdef QUOTA
1306		struct ucred ucred, *ucp;
1307		ucp = cnp->cn_cred;
1308#endif			I
1309		/*
1310		 * If we are hacking owners here, (only do this where told to)
1311		 * and we are not giving it TOO root, (would subvert quotas)
1312		 * then go ahead and give it to the other user.
1313		 * The new directory also inherits the SUID bit.
1314		 * If user's UID and dir UID are the same,
1315		 * 'give it away' so that the SUID is still forced on.
1316		 */
1317		if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1318		    (dp->i_mode & ISUID) && dp->i_uid) {
1319			dmode |= ISUID;
1320			ip->i_uid = dp->i_uid;
1321#ifdef QUOTA
1322			if (dp->i_uid != cnp->cn_cred->cr_uid) {
1323				/*
1324				 * Make sure the correct user gets charged
1325				 * for the space.
1326				 * Make a dummy credential for the victim.
1327				 * XXX This seems to never be accessed out of
1328				 * our context so a stack variable is ok.
1329				 */
1330				ucred.cr_ref = 1;
1331				ucred.cr_uid = ip->i_uid;
1332				ucred.cr_ngroups = 1;
1333				ucred.cr_groups[0] = dp->i_gid;
1334				ucp = &ucred;
1335			}
1336#endif
1337		} else
1338			ip->i_uid = cnp->cn_cred->cr_uid;
1339#ifdef QUOTA
1340		if ((error = getinoquota(ip)) ||
1341	    	    (error = chkiq(ip, 1, ucp, 0))) {
1342			zfree(namei_zone, cnp->cn_pnbuf);
1343			UFS_VFREE(tvp, ip->i_number, dmode);
1344			vput(tvp);
1345			return (error);
1346		}
1347#endif
1348	}
1349#else	/* !SUIDDIR */
1350	ip->i_uid = cnp->cn_cred->cr_uid;
1351#ifdef QUOTA
1352	if ((error = getinoquota(ip)) ||
1353	    (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
1354		zfree(namei_zone, cnp->cn_pnbuf);
1355		UFS_VFREE(tvp, ip->i_number, dmode);
1356		vput(tvp);
1357		return (error);
1358	}
1359#endif
1360#endif	/* !SUIDDIR */
1361	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1362	ip->i_mode = dmode;
1363	tvp->v_type = VDIR;	/* Rest init'd in getnewvnode(). */
1364	ip->i_effnlink = 2;
1365	ip->i_nlink = 2;
1366	if (DOINGSOFTDEP(tvp))
1367		softdep_increase_linkcnt(ip);
1368	if (cnp->cn_flags & ISWHITEOUT)
1369		ip->i_flags |= UF_OPAQUE;
1370
1371	/*
1372	 * Bump link count in parent directory to reflect work done below.
1373	 * Should be done before reference is created so cleanup is
1374	 * possible if we crash.
1375	 */
1376	dp->i_effnlink++;
1377	dp->i_nlink++;
1378	dp->i_flag |= IN_CHANGE;
1379	if (DOINGSOFTDEP(dvp))
1380		softdep_increase_linkcnt(dp);
1381	getmicrotime(&tv);
1382        error = UFS_UPDATE(tvp, &tv, &tv, !DOINGSOFTDEP(dvp));
1383	if (error)
1384		goto bad;
1385
1386	/*
1387	 * Initialize directory with "." and ".." from static template.
1388	 */
1389	if (dvp->v_mount->mnt_maxsymlinklen > 0
1390	)
1391		dtp = &mastertemplate;
1392	else
1393		dtp = (struct dirtemplate *)&omastertemplate;
1394	dirtemplate = *dtp;
1395	dirtemplate.dot_ino = ip->i_number;
1396	dirtemplate.dotdot_ino = dp->i_number;
1397	if ((error = VOP_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
1398	    B_CLRBUF, &bp)) != 0)
1399		goto bad;
1400	ip->i_size = DIRBLKSIZ;
1401	ip->i_flag |= IN_CHANGE | IN_UPDATE;
1402	vnode_pager_setsize(tvp, (u_long)ip->i_size);
1403	bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
1404	if (DOINGSOFTDEP(tvp)) {
1405		/*
1406		 * Ensure that the entire newly allocated block is a
1407		 * valid directory so that future growth within the
1408		 * block does not have to ensure that the block is
1409		 * written before the inode.
1410		 */
1411		blkoff = DIRBLKSIZ;
1412		while (blkoff < bp->b_bcount) {
1413			((struct direct *)
1414			   (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
1415			blkoff += DIRBLKSIZ;
1416		}
1417	}
1418	if ((error = UFS_UPDATE(tvp, &tv, &tv, !DOINGSOFTDEP(tvp))) != 0) {
1419		(void)VOP_BWRITE(bp);
1420		goto bad;
1421	}
1422	VN_POLLEVENT(dvp, POLLWRITE); /* XXX right place? */
1423	/*
1424	 * Directory set up, now install its entry in the parent directory.
1425	 *
1426	 * If we are not doing soft dependencies, then we must write out the
1427	 * buffer containing the new directory body before entering the new
1428	 * name in the parent. If we are doing soft dependencies, then the
1429	 * buffer containing the new directory body will be passed to and
1430	 * released in the soft dependency code after the code has attached
1431	 * an appropriate ordering dependency to the buffer which ensures that
1432	 * the buffer is written before the new name is written in the parent.
1433	 */
1434	if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp)) != 0))
1435		goto bad;
1436	ufs_makedirentry(ip, cnp, &newdir);
1437	error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
1438
1439bad:
1440	if (error == 0) {
1441		*ap->a_vpp = tvp;
1442	} else {
1443		dp->i_effnlink--;
1444		dp->i_nlink--;
1445		dp->i_flag |= IN_CHANGE;
1446		/*
1447		 * No need to do an explicit VOP_TRUNCATE here, vrele will
1448		 * do this for us because we set the link count to 0.
1449		 */
1450		ip->i_effnlink = 0;
1451		ip->i_nlink = 0;
1452		ip->i_flag |= IN_CHANGE;
1453		vput(tvp);
1454	}
1455out:
1456	zfree(namei_zone, cnp->cn_pnbuf);
1457	return (error);
1458}
1459
1460/*
1461 * Rmdir system call.
1462 */
1463int
1464ufs_rmdir(ap)
1465	struct vop_rmdir_args /* {
1466		struct vnode *a_dvp;
1467		struct vnode *a_vp;
1468		struct componentname *a_cnp;
1469	} */ *ap;
1470{
1471	struct vnode *vp = ap->a_vp;
1472	struct vnode *dvp = ap->a_dvp;
1473	struct componentname *cnp = ap->a_cnp;
1474	struct inode *ip, *dp;
1475	int error;
1476
1477	ip = VTOI(vp);
1478	dp = VTOI(dvp);
1479
1480	/*
1481	 * Do not remove a directory that is in the process of being renamed.
1482	 * Verify the directory is empty (and valid). Rmdir ".." will not be
1483	 * valid since ".." will contain a reference to the current directory
1484	 * and thus be non-empty.
1485	 */
1486	error = 0;
1487	if (ip->i_flag & IN_RENAME) {
1488		error = EINVAL;
1489		goto out;
1490	}
1491	if (ip->i_effnlink != 2 ||
1492	    !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1493		error = ENOTEMPTY;
1494		goto out;
1495	}
1496	if ((dp->i_flags & APPEND)
1497	    || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1498		error = EPERM;
1499		goto out;
1500	}
1501	/*
1502	 * Delete reference to directory before purging
1503	 * inode.  If we crash in between, the directory
1504	 * will be reattached to lost+found,
1505	 */
1506	error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
1507	if (error)
1508		goto out;
1509	VN_POLLEVENT(dvp, POLLWRITE|POLLNLINK);
1510	cache_purge(dvp);
1511	/*
1512	 * Truncate inode. The only stuff left in the directory is "." and
1513	 * "..". The "." reference is inconsequential since we are quashing
1514	 * it. We have removed the "." reference and the reference in the
1515	 * parent directory, but there may be other hard links. So,
1516	 * ufs_dirremove will set the UF_IMMUTABLE flag to ensure that no
1517	 * new entries are made. The soft dependency code will arrange to
1518	 * do these operations after the parent directory entry has been
1519	 * deleted on disk, so when running with that code we avoid doing
1520	 * them now.
1521	 */
1522	dp->i_effnlink--;
1523	dp->i_flag |= IN_CHANGE;
1524	ip->i_effnlink--;
1525	ip->i_flag |= IN_CHANGE;
1526	if (!DOINGSOFTDEP(vp)) {
1527		dp->i_nlink--;
1528		ip->i_nlink--;
1529		error = UFS_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred,
1530		    cnp->cn_proc);
1531	}
1532	cache_purge(vp);
1533out:
1534	VN_POLLEVENT(vp, POLLNLINK);
1535	return (error);
1536}
1537
1538/*
1539 * symlink -- make a symbolic link
1540 */
1541int
1542ufs_symlink(ap)
1543	struct vop_symlink_args /* {
1544		struct vnode *a_dvp;
1545		struct vnode **a_vpp;
1546		struct componentname *a_cnp;
1547		struct vattr *a_vap;
1548		char *a_target;
1549	} */ *ap;
1550{
1551	register struct vnode *vp, **vpp = ap->a_vpp;
1552	register struct inode *ip;
1553	int len, error;
1554
1555	error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1556	    vpp, ap->a_cnp);
1557	if (error)
1558		return (error);
1559	VN_POLLEVENT(ap->a_dvp, POLLWRITE);
1560	vp = *vpp;
1561	len = strlen(ap->a_target);
1562	if (len < vp->v_mount->mnt_maxsymlinklen) {
1563		ip = VTOI(vp);
1564		bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1565		ip->i_size = len;
1566		ip->i_flag |= IN_CHANGE | IN_UPDATE;
1567	} else
1568		error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1569		    UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred, (int *)0,
1570		    (struct proc *)0);
1571	vput(vp);
1572	return (error);
1573}
1574
1575/*
1576 * Vnode op for reading directories.
1577 *
1578 * The routine below assumes that the on-disk format of a directory
1579 * is the same as that defined by <sys/dirent.h>. If the on-disk
1580 * format changes, then it will be necessary to do a conversion
1581 * from the on-disk format that read returns to the format defined
1582 * by <sys/dirent.h>.
1583 */
1584int
1585ufs_readdir(ap)
1586	struct vop_readdir_args /* {
1587		struct vnode *a_vp;
1588		struct uio *a_uio;
1589		struct ucred *a_cred;
1590		int *a_eofflag;
1591		int *ncookies;
1592		u_long **a_cookies;
1593	} */ *ap;
1594{
1595	register struct uio *uio = ap->a_uio;
1596	int error;
1597	size_t count, lost;
1598	off_t off;
1599
1600	if (ap->a_ncookies != NULL)
1601		/*
1602		 * Ensure that the block is aligned.  The caller can use
1603		 * the cookies to determine where in the block to start.
1604		 */
1605		uio->uio_offset &= ~(DIRBLKSIZ - 1);
1606	off = uio->uio_offset;
1607	count = uio->uio_resid;
1608	/* Make sure we don't return partial entries. */
1609	count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
1610	if (count <= 0)
1611		return (EINVAL);
1612	lost = uio->uio_resid - count;
1613	uio->uio_resid = count;
1614	uio->uio_iov->iov_len = count;
1615#	if (BYTE_ORDER == LITTLE_ENDIAN)
1616		if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
1617			error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1618		} else {
1619			struct dirent *dp, *edp;
1620			struct uio auio;
1621			struct iovec aiov;
1622			caddr_t dirbuf;
1623			int readcnt;
1624			u_char tmp;
1625
1626			auio = *uio;
1627			auio.uio_iov = &aiov;
1628			auio.uio_iovcnt = 1;
1629			auio.uio_segflg = UIO_SYSSPACE;
1630			aiov.iov_len = count;
1631			MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
1632			aiov.iov_base = dirbuf;
1633			error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
1634			if (error == 0) {
1635				readcnt = count - auio.uio_resid;
1636				edp = (struct dirent *)&dirbuf[readcnt];
1637				for (dp = (struct dirent *)dirbuf; dp < edp; ) {
1638					tmp = dp->d_namlen;
1639					dp->d_namlen = dp->d_type;
1640					dp->d_type = tmp;
1641					if (dp->d_reclen > 0) {
1642						dp = (struct dirent *)
1643						    ((char *)dp + dp->d_reclen);
1644					} else {
1645						error = EIO;
1646						break;
1647					}
1648				}
1649				if (dp >= edp)
1650					error = uiomove(dirbuf, readcnt, uio);
1651			}
1652			FREE(dirbuf, M_TEMP);
1653		}
1654#	else
1655		error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
1656#	endif
1657	if (!error && ap->a_ncookies != NULL) {
1658		struct dirent* dpStart;
1659		struct dirent* dpEnd;
1660		struct dirent* dp;
1661		int ncookies;
1662		u_long *cookies;
1663		u_long *cookiep;
1664
1665		if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1666			panic("ufs_readdir: unexpected uio from NFS server");
1667		dpStart = (struct dirent *)
1668		     (uio->uio_iov->iov_base - (uio->uio_offset - off));
1669		dpEnd = (struct dirent *) uio->uio_iov->iov_base;
1670		for (dp = dpStart, ncookies = 0;
1671		     dp < dpEnd;
1672		     dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
1673			ncookies++;
1674		MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
1675		    M_WAITOK);
1676		for (dp = dpStart, cookiep = cookies;
1677		     dp < dpEnd;
1678		     dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
1679			off += dp->d_reclen;
1680			*cookiep++ = (u_long) off;
1681		}
1682		*ap->a_ncookies = ncookies;
1683		*ap->a_cookies = cookies;
1684	}
1685	uio->uio_resid += lost;
1686	if (ap->a_eofflag)
1687	    *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
1688	return (error);
1689}
1690
1691/*
1692 * Return target name of a symbolic link
1693 */
1694int
1695ufs_readlink(ap)
1696	struct vop_readlink_args /* {
1697		struct vnode *a_vp;
1698		struct uio *a_uio;
1699		struct ucred *a_cred;
1700	} */ *ap;
1701{
1702	register struct vnode *vp = ap->a_vp;
1703	register struct inode *ip = VTOI(vp);
1704	int isize;
1705
1706	isize = ip->i_size;
1707	if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1708	    (ip->i_din.di_blocks == 0)) {	/* XXX - for old fastlink support */
1709		uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1710		return (0);
1711	}
1712	return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1713}
1714
1715/*
1716 * Ufs abort op, called after namei() when a CREATE/DELETE isn't actually
1717 * done. If a buffer has been saved in anticipation of a CREATE, delete it.
1718 */
1719/* ARGSUSED */
1720int
1721ufs_abortop(ap)
1722	struct vop_abortop_args /* {
1723		struct vnode *a_dvp;
1724		struct componentname *a_cnp;
1725	} */ *ap;
1726{
1727	if ((ap->a_cnp->cn_flags & (HASBUF | SAVESTART)) == HASBUF)
1728		zfree(namei_zone, ap->a_cnp->cn_pnbuf);
1729	return (0);
1730}
1731
1732/*
1733 * Calculate the logical to physical mapping if not done already,
1734 * then call the device strategy routine.
1735 */
1736int
1737ufs_strategy(ap)
1738	struct vop_strategy_args /* {
1739		struct vnode *a_vp;
1740		struct buf *a_bp;
1741	} */ *ap;
1742{
1743	register struct buf *bp = ap->a_bp;
1744	register struct vnode *vp = ap->a_vp;
1745	register struct inode *ip;
1746	int error;
1747
1748	ip = VTOI(vp);
1749	if (vp->v_type == VBLK || vp->v_type == VCHR)
1750		panic("ufs_strategy: spec");
1751	if (bp->b_blkno == bp->b_lblkno) {
1752		error = VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
1753		if (error) {
1754			bp->b_error = error;
1755			bp->b_flags |= B_ERROR;
1756			biodone(bp);
1757			return (error);
1758		}
1759		if ((long)bp->b_blkno == -1)
1760			vfs_bio_clrbuf(bp);
1761	}
1762	if ((long)bp->b_blkno == -1) {
1763		biodone(bp);
1764		return (0);
1765	}
1766	vp = ip->i_devvp;
1767	bp->b_dev = vp->v_rdev;
1768	VOP_STRATEGY(vp, bp);
1769	return (0);
1770}
1771
1772/*
1773 * Print out the contents of an inode.
1774 */
1775int
1776ufs_print(ap)
1777	struct vop_print_args /* {
1778		struct vnode *a_vp;
1779	} */ *ap;
1780{
1781	register struct vnode *vp = ap->a_vp;
1782	register struct inode *ip = VTOI(vp);
1783
1784	printf("tag VT_UFS, ino %lu, on dev %#lx (%d, %d)",
1785	    (u_long)ip->i_number, (u_long)ip->i_dev, major(ip->i_dev),
1786	    minor(ip->i_dev));
1787	if (vp->v_type == VFIFO)
1788		fifo_printinfo(vp);
1789	lockmgr_printinfo(&ip->i_lock);
1790	printf("\n");
1791	return (0);
1792}
1793
1794/*
1795 * Read wrapper for special devices.
1796 */
1797int
1798ufsspec_read(ap)
1799	struct vop_read_args /* {
1800		struct vnode *a_vp;
1801		struct uio *a_uio;
1802		int  a_ioflag;
1803		struct ucred *a_cred;
1804	} */ *ap;
1805{
1806	int error, resid;
1807	struct uio *uio;
1808
1809	uio = ap->a_uio;
1810	resid = uio->uio_resid;
1811	error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap);
1812	if (uio->uio_resid != resid)
1813		VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1814	return (error);
1815}
1816
1817/*
1818 * Write wrapper for special devices.
1819 */
1820int
1821ufsspec_write(ap)
1822	struct vop_write_args /* {
1823		struct vnode *a_vp;
1824		struct uio *a_uio;
1825		int  a_ioflag;
1826		struct ucred *a_cred;
1827	} */ *ap;
1828{
1829	int error, resid;
1830	struct uio *uio;
1831
1832	uio = ap->a_uio;
1833	resid = uio->uio_resid;
1834	error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap);
1835	if (uio->uio_resid != resid)
1836		VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1837	return (error);
1838}
1839
1840/*
1841 * Close wrapper for special devices.
1842 *
1843 * Update the times on the inode then do device close.
1844 */
1845int
1846ufsspec_close(ap)
1847	struct vop_close_args /* {
1848		struct vnode *a_vp;
1849		int  a_fflag;
1850		struct ucred *a_cred;
1851		struct proc *a_p;
1852	} */ *ap;
1853{
1854	struct vnode *vp = ap->a_vp;
1855
1856	simple_lock(&vp->v_interlock);
1857	if (vp->v_usecount > 1)
1858		ufs_itimes(vp);
1859	simple_unlock(&vp->v_interlock);
1860	return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
1861}
1862
1863/*
1864 * Read wrapper for fifos.
1865 */
1866int
1867ufsfifo_read(ap)
1868	struct vop_read_args /* {
1869		struct vnode *a_vp;
1870		struct uio *a_uio;
1871		int  a_ioflag;
1872		struct ucred *a_cred;
1873	} */ *ap;
1874{
1875	int error, resid;
1876	struct uio *uio;
1877
1878	uio = ap->a_uio;
1879	resid = uio->uio_resid;
1880	error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap);
1881	if (uio->uio_resid != resid &&
1882	    (ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0)
1883		VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1884	return (error);
1885}
1886
1887/*
1888 * Write wrapper for fifos.
1889 */
1890int
1891ufsfifo_write(ap)
1892	struct vop_write_args /* {
1893		struct vnode *a_vp;
1894		struct uio *a_uio;
1895		int  a_ioflag;
1896		struct ucred *a_cred;
1897	} */ *ap;
1898{
1899	int error, resid;
1900	struct uio *uio;
1901
1902	uio = ap->a_uio;
1903	resid = uio->uio_resid;
1904	error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap);
1905	if (uio->uio_resid != resid)
1906		VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1907	return (error);
1908}
1909
1910/*
1911 * Close wrapper for fifos.
1912 *
1913 * Update the times on the inode then do device close.
1914 */
1915int
1916ufsfifo_close(ap)
1917	struct vop_close_args /* {
1918		struct vnode *a_vp;
1919		int  a_fflag;
1920		struct ucred *a_cred;
1921		struct proc *a_p;
1922	} */ *ap;
1923{
1924	struct vnode *vp = ap->a_vp;
1925
1926	simple_lock(&vp->v_interlock);
1927	if (vp->v_usecount > 1)
1928		ufs_itimes(vp);
1929	simple_unlock(&vp->v_interlock);
1930	return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
1931}
1932
1933/*
1934 * Return POSIX pathconf information applicable to ufs filesystems.
1935 */
1936int
1937ufs_pathconf(ap)
1938	struct vop_pathconf_args /* {
1939		struct vnode *a_vp;
1940		int a_name;
1941		int *a_retval;
1942	} */ *ap;
1943{
1944
1945	switch (ap->a_name) {
1946	case _PC_LINK_MAX:
1947		*ap->a_retval = LINK_MAX;
1948		return (0);
1949	case _PC_NAME_MAX:
1950		*ap->a_retval = NAME_MAX;
1951		return (0);
1952	case _PC_PATH_MAX:
1953		*ap->a_retval = PATH_MAX;
1954		return (0);
1955	case _PC_PIPE_BUF:
1956		*ap->a_retval = PIPE_BUF;
1957		return (0);
1958	case _PC_CHOWN_RESTRICTED:
1959		*ap->a_retval = 1;
1960		return (0);
1961	case _PC_NO_TRUNC:
1962		*ap->a_retval = 1;
1963		return (0);
1964	default:
1965		return (EINVAL);
1966	}
1967	/* NOTREACHED */
1968}
1969
1970/*
1971 * Advisory record locking support
1972 */
1973int
1974ufs_advlock(ap)
1975	struct vop_advlock_args /* {
1976		struct vnode *a_vp;
1977		caddr_t  a_id;
1978		int  a_op;
1979		struct flock *a_fl;
1980		int  a_flags;
1981	} */ *ap;
1982{
1983	register struct inode *ip = VTOI(ap->a_vp);
1984
1985	return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
1986}
1987
1988/*
1989 * Initialize the vnode associated with a new inode, handle aliased
1990 * vnodes.
1991 */
1992int
1993ufs_vinit(mntp, specops, fifoops, vpp)
1994	struct mount *mntp;
1995	vop_t **specops;
1996	vop_t **fifoops;
1997	struct vnode **vpp;
1998{
1999	struct inode *ip;
2000	struct vnode *vp, *nvp;
2001	struct timeval tv;
2002
2003	vp = *vpp;
2004	ip = VTOI(vp);
2005	switch(vp->v_type = IFTOVT(ip->i_mode)) {
2006	case VCHR:
2007	case VBLK:
2008		vp->v_op = specops;
2009		nvp = checkalias(vp, ip->i_rdev, mntp);
2010		if (nvp) {
2011			/*
2012			 * Discard unneeded vnode, but save its inode.
2013			 * Note that the lock is carried over in the inode
2014			 * to the replacement vnode.
2015			 */
2016			nvp->v_data = vp->v_data;
2017			vp->v_data = NULL;
2018			vp->v_op = spec_vnodeop_p;
2019			vrele(vp);
2020			vgone(vp);
2021			/*
2022			 * Reinitialize aliased inode.
2023			 */
2024			vp = nvp;
2025			ip->i_vnode = vp;
2026		}
2027		break;
2028	case VFIFO:
2029		vp->v_op = fifoops;
2030		break;
2031	default:
2032		break;
2033
2034	}
2035	if (ip->i_number == ROOTINO)
2036		vp->v_flag |= VROOT;
2037	/*
2038	 * Initialize modrev times
2039	 */
2040	getmicrouptime(&tv);
2041	SETHIGH(ip->i_modrev, tv.tv_sec);
2042	SETLOW(ip->i_modrev, tv.tv_usec * 4294);
2043	*vpp = vp;
2044	return (0);
2045}
2046
2047/*
2048 * Allocate a new inode.
2049 */
2050int
2051ufs_makeinode(mode, dvp, vpp, cnp)
2052	int mode;
2053	struct vnode *dvp;
2054	struct vnode **vpp;
2055	struct componentname *cnp;
2056{
2057	register struct inode *ip, *pdir;
2058	struct direct newdir;
2059	struct timeval tv;
2060	struct vnode *tvp;
2061	int error;
2062
2063	pdir = VTOI(dvp);
2064#ifdef DIAGNOSTIC
2065	if ((cnp->cn_flags & HASBUF) == 0)
2066		panic("ufs_makeinode: no name");
2067#endif
2068	*vpp = NULL;
2069	if ((mode & IFMT) == 0)
2070		mode |= IFREG;
2071
2072	error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
2073	if (error) {
2074		zfree(namei_zone, cnp->cn_pnbuf);
2075		return (error);
2076	}
2077	ip = VTOI(tvp);
2078	ip->i_gid = pdir->i_gid;
2079#ifdef SUIDDIR
2080	{
2081#ifdef QUOTA
2082		struct ucred ucred, *ucp;
2083		ucp = cnp->cn_cred;
2084#endif			I
2085		/*
2086		 * If we are not the owner of the directory,
2087		 * and we are hacking owners here, (only do this where told to)
2088		 * and we are not giving it TOO root, (would subvert quotas)
2089		 * then go ahead and give it to the other user.
2090		 * Note that this drops off the execute bits for security.
2091		 */
2092		if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
2093		    (pdir->i_mode & ISUID) &&
2094		    (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
2095			ip->i_uid = pdir->i_uid;
2096			mode &= ~07111;
2097#ifdef QUOTA
2098			/*
2099			 * Make sure the correct user gets charged
2100			 * for the space.
2101			 * Quickly knock up a dummy credential for the victim.
2102			 * XXX This seems to never be accessed out of our
2103			 * context so a stack variable is ok.
2104			 */
2105			ucred.cr_ref = 1;
2106			ucred.cr_uid = ip->i_uid;
2107			ucred.cr_ngroups = 1;
2108			ucred.cr_groups[0] = pdir->i_gid;
2109			ucp = &ucred;
2110#endif
2111		} else
2112			ip->i_uid = cnp->cn_cred->cr_uid;
2113
2114#ifdef QUOTA
2115		if ((error = getinoquota(ip)) ||
2116	    	    (error = chkiq(ip, 1, ucp, 0))) {
2117			zfree(namei_zone, cnp->cn_pnbuf);
2118			UFS_VFREE(tvp, ip->i_number, mode);
2119			vput(tvp);
2120			return (error);
2121		}
2122#endif
2123	}
2124#else	/* !SUIDDIR */
2125	ip->i_uid = cnp->cn_cred->cr_uid;
2126#ifdef QUOTA
2127	if ((error = getinoquota(ip)) ||
2128	    (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
2129		zfree(namei_zone, cnp->cn_pnbuf);
2130		UFS_VFREE(tvp, ip->i_number, mode);
2131		vput(tvp);
2132		return (error);
2133	}
2134#endif
2135#endif	/* !SUIDDIR */
2136	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
2137	ip->i_mode = mode;
2138	tvp->v_type = IFTOVT(mode);	/* Rest init'd in getnewvnode(). */
2139	ip->i_effnlink = 1;
2140	ip->i_nlink = 1;
2141	if (DOINGSOFTDEP(tvp))
2142		softdep_increase_linkcnt(ip);
2143	if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
2144	    suser(cnp->cn_cred, NULL))
2145		ip->i_mode &= ~ISGID;
2146
2147	if (cnp->cn_flags & ISWHITEOUT)
2148		ip->i_flags |= UF_OPAQUE;
2149
2150	/*
2151	 * Make sure inode goes to disk before directory entry.
2152	 */
2153	getmicrotime(&tv);
2154	error = UFS_UPDATE(tvp, &tv, &tv, !DOINGSOFTDEP(tvp));
2155	if (error)
2156		goto bad;
2157	ufs_makedirentry(ip, cnp, &newdir);
2158	error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
2159	if (error)
2160		goto bad;
2161
2162	if ((cnp->cn_flags & SAVESTART) == 0)
2163		zfree(namei_zone, cnp->cn_pnbuf);
2164	*vpp = tvp;
2165	return (0);
2166
2167bad:
2168	/*
2169	 * Write error occurred trying to update the inode
2170	 * or the directory so must deallocate the inode.
2171	 */
2172	zfree(namei_zone, cnp->cn_pnbuf);
2173	ip->i_effnlink = 0;
2174	ip->i_nlink = 0;
2175	ip->i_flag |= IN_CHANGE;
2176	vput(tvp);
2177	return (error);
2178}
2179
2180static int
2181ufs_missingop(ap)
2182	struct vop_generic_args *ap;
2183{
2184
2185	panic("no vop function for %s in ufs child", ap->a_desc->vdesc_name);
2186	return (EOPNOTSUPP);
2187}
2188
2189/* Global vfs data structures for ufs. */
2190static vop_t **ufs_vnodeop_p;
2191static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
2192	{ &vop_default_desc,		(vop_t *) vop_defaultop },
2193	{ &vop_fsync_desc,		(vop_t *) ufs_missingop },
2194	{ &vop_read_desc,		(vop_t *) ufs_missingop },
2195	{ &vop_reallocblks_desc,	(vop_t *) ufs_missingop },
2196	{ &vop_write_desc,		(vop_t *) ufs_missingop },
2197	{ &vop_abortop_desc,		(vop_t *) ufs_abortop },
2198	{ &vop_access_desc,		(vop_t *) ufs_access },
2199	{ &vop_advlock_desc,		(vop_t *) ufs_advlock },
2200	{ &vop_bmap_desc,		(vop_t *) ufs_bmap },
2201	{ &vop_cachedlookup_desc,	(vop_t *) ufs_lookup },
2202	{ &vop_close_desc,		(vop_t *) ufs_close },
2203	{ &vop_create_desc,		(vop_t *) ufs_create },
2204	{ &vop_getattr_desc,		(vop_t *) ufs_getattr },
2205	{ &vop_inactive_desc,		(vop_t *) ufs_inactive },
2206	{ &vop_islocked_desc,		(vop_t *) vop_stdislocked },
2207	{ &vop_link_desc,		(vop_t *) ufs_link },
2208	{ &vop_lock_desc,		(vop_t *) vop_stdlock },
2209	{ &vop_lookup_desc,		(vop_t *) vfs_cache_lookup },
2210	{ &vop_mkdir_desc,		(vop_t *) ufs_mkdir },
2211	{ &vop_mknod_desc,		(vop_t *) ufs_mknod },
2212	{ &vop_mmap_desc,		(vop_t *) ufs_mmap },
2213	{ &vop_open_desc,		(vop_t *) ufs_open },
2214	{ &vop_pathconf_desc,		(vop_t *) ufs_pathconf },
2215	{ &vop_poll_desc,		(vop_t *) vop_stdpoll },
2216	{ &vop_print_desc,		(vop_t *) ufs_print },
2217	{ &vop_readdir_desc,		(vop_t *) ufs_readdir },
2218	{ &vop_readlink_desc,		(vop_t *) ufs_readlink },
2219	{ &vop_reclaim_desc,		(vop_t *) ufs_reclaim },
2220	{ &vop_remove_desc,		(vop_t *) ufs_remove },
2221	{ &vop_rename_desc,		(vop_t *) ufs_rename },
2222	{ &vop_rmdir_desc,		(vop_t *) ufs_rmdir },
2223	{ &vop_setattr_desc,		(vop_t *) ufs_setattr },
2224	{ &vop_strategy_desc,		(vop_t *) ufs_strategy },
2225	{ &vop_symlink_desc,		(vop_t *) ufs_symlink },
2226	{ &vop_unlock_desc,		(vop_t *) vop_stdunlock },
2227	{ &vop_whiteout_desc,		(vop_t *) ufs_whiteout },
2228	{ NULL, NULL }
2229};
2230static struct vnodeopv_desc ufs_vnodeop_opv_desc =
2231	{ &ufs_vnodeop_p, ufs_vnodeop_entries };
2232
2233static vop_t **ufs_specop_p;
2234static struct vnodeopv_entry_desc ufs_specop_entries[] = {
2235	{ &vop_default_desc,		(vop_t *) spec_vnoperate },
2236	{ &vop_fsync_desc,		(vop_t *) ufs_missingop },
2237	{ &vop_access_desc,		(vop_t *) ufs_access },
2238	{ &vop_close_desc,		(vop_t *) ufsspec_close },
2239	{ &vop_getattr_desc,		(vop_t *) ufs_getattr },
2240	{ &vop_inactive_desc,		(vop_t *) ufs_inactive },
2241	{ &vop_islocked_desc,		(vop_t *) vop_stdislocked },
2242	{ &vop_lock_desc,		(vop_t *) vop_stdlock },
2243	{ &vop_print_desc,		(vop_t *) ufs_print },
2244	{ &vop_read_desc,		(vop_t *) ufsspec_read },
2245	{ &vop_reclaim_desc,		(vop_t *) ufs_reclaim },
2246	{ &vop_setattr_desc,		(vop_t *) ufs_setattr },
2247	{ &vop_unlock_desc,		(vop_t *) vop_stdunlock },
2248	{ &vop_write_desc,		(vop_t *) ufsspec_write },
2249	{ NULL, NULL }
2250};
2251static struct vnodeopv_desc ufs_specop_opv_desc =
2252	{ &ufs_specop_p, ufs_specop_entries };
2253
2254static vop_t **ufs_fifoop_p;
2255static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
2256	{ &vop_default_desc,		(vop_t *) fifo_vnoperate },
2257	{ &vop_fsync_desc,		(vop_t *) ufs_missingop },
2258	{ &vop_access_desc,		(vop_t *) ufs_access },
2259	{ &vop_close_desc,		(vop_t *) ufsfifo_close },
2260	{ &vop_getattr_desc,		(vop_t *) ufs_getattr },
2261	{ &vop_inactive_desc,		(vop_t *) ufs_inactive },
2262	{ &vop_islocked_desc,		(vop_t *) vop_stdislocked },
2263	{ &vop_lock_desc,		(vop_t *) vop_stdlock },
2264	{ &vop_print_desc,		(vop_t *) ufs_print },
2265	{ &vop_read_desc,		(vop_t *) ufsfifo_read },
2266	{ &vop_reclaim_desc,		(vop_t *) ufs_reclaim },
2267	{ &vop_setattr_desc,		(vop_t *) ufs_setattr },
2268	{ &vop_unlock_desc,		(vop_t *) vop_stdunlock },
2269	{ &vop_write_desc,		(vop_t *) ufsfifo_write },
2270	{ NULL, NULL }
2271};
2272static struct vnodeopv_desc ufs_fifoop_opv_desc =
2273	{ &ufs_fifoop_p, ufs_fifoop_entries };
2274
2275VNODEOP_SET(ufs_vnodeop_opv_desc);
2276VNODEOP_SET(ufs_specop_opv_desc);
2277VNODEOP_SET(ufs_fifoop_opv_desc);
2278
2279int
2280ufs_vnoperate(ap)
2281	struct vop_generic_args /* {
2282		struct vnodeop_desc *a_desc;
2283	} */ *ap;
2284{
2285	return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap));
2286}
2287
2288int
2289ufs_vnoperatefifo(ap)
2290	struct vop_generic_args /* {
2291		struct vnodeop_desc *a_desc;
2292	} */ *ap;
2293{
2294	return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap));
2295}
2296
2297int
2298ufs_vnoperatespec(ap)
2299	struct vop_generic_args /* {
2300		struct vnodeop_desc *a_desc;
2301	} */ *ap;
2302{
2303	return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap));
2304}
2305
2306
2307