xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 4b896b232495b7a9b8b94a1cf1e21873296d53b8)
1 /*	$NetBSD: vfs_syscalls.c,v 1.209 2004/05/25 14:54:57 hannken Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.209 2004/05/25 14:54:57 hannken Exp $");
41 
42 #include "opt_compat_netbsd.h"
43 #include "opt_compat_43.h"
44 #include "opt_ktrace.h"
45 #include "fss.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/namei.h>
50 #include <sys/filedesc.h>
51 #include <sys/kernel.h>
52 #include <sys/file.h>
53 #include <sys/stat.h>
54 #include <sys/vnode.h>
55 #include <sys/mount.h>
56 #include <sys/proc.h>
57 #include <sys/uio.h>
58 #include <sys/malloc.h>
59 #include <sys/dirent.h>
60 #include <sys/sysctl.h>
61 #include <sys/sa.h>
62 #include <sys/syscallargs.h>
63 #ifdef KTRACE
64 #include <sys/ktrace.h>
65 #endif
66 
67 #include <miscfs/genfs/genfs.h>
68 #include <miscfs/syncfs/syncfs.h>
69 
70 #if NFSS > 0
71 #include <dev/fssvar.h>
72 #endif
73 
74 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
75 
76 static int change_dir(struct nameidata *, struct proc *);
77 static int change_flags(struct vnode *, u_long, struct proc *);
78 static int change_mode(struct vnode *, int, struct proc *p);
79 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int);
80 static int change_utimes(struct vnode *vp, const struct timeval *,
81 	       struct proc *p);
82 static int rename_files(const char *, const char *, struct proc *, int);
83 
84 void checkdirs(struct vnode *);
85 
86 int dovfsusermount = 0;
87 
88 /*
89  * Virtual File System System Calls
90  */
91 
92 /*
93  * Mount a file system.
94  */
95 
96 #if defined(COMPAT_09) || defined(COMPAT_43)
97 /*
98  * This table is used to maintain compatibility with 4.3BSD
99  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
100  *
101  * Do not modify this table. It should only contain filesystems
102  * supported by NetBSD 0.9 and 4.3BSD.
103  */
104 const char * const mountcompatnames[] = {
105 	NULL,		/* 0 = MOUNT_NONE */
106 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
107 	MOUNT_NFS,	/* 2 */
108 	MOUNT_MFS,	/* 3 */
109 	MOUNT_MSDOS,	/* 4 */
110 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
111 	MOUNT_FDESC,	/* 6 */
112 	MOUNT_KERNFS,	/* 7 */
113 	NULL,		/* 8 = MOUNT_DEVFS */
114 	MOUNT_AFS,	/* 9 */
115 };
116 const int nmountcompatnames = sizeof(mountcompatnames) /
117     sizeof(mountcompatnames[0]);
118 #endif /* COMPAT_09 || COMPAT_43 */
119 
120 /* ARGSUSED */
121 int
122 sys_mount(l, v, retval)
123 	struct lwp *l;
124 	void *v;
125 	register_t *retval;
126 {
127 	struct sys_mount_args /* {
128 		syscallarg(const char *) type;
129 		syscallarg(const char *) path;
130 		syscallarg(int) flags;
131 		syscallarg(void *) data;
132 	} */ *uap = v;
133 	struct proc *p = l->l_proc;
134 	struct vnode *vp;
135 	struct mount *mp;
136 	int error, flag = 0;
137 	char fstypename[MFSNAMELEN];
138 	struct vattr va;
139 	struct nameidata nd;
140 	struct vfsops *vfs;
141 
142 	if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
143 	    (error = suser(p->p_ucred, &p->p_acflag)))
144 		return (error);
145 	/*
146 	 * Get vnode to be covered
147 	 */
148 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
149 	    SCARG(uap, path), p);
150 	if ((error = namei(&nd)) != 0)
151 		return (error);
152 	vp = nd.ni_vp;
153 	/*
154 	 * A lookup in VFS_MOUNT might result in an attempt to
155 	 * lock this vnode again, so make the lock recursive.
156 	 */
157 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
158 	if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
159 		if ((vp->v_flag & VROOT) == 0) {
160 			vput(vp);
161 			return (EINVAL);
162 		}
163 		mp = vp->v_mount;
164 		flag = mp->mnt_flag;
165 		vfs = mp->mnt_op;
166 		/*
167 		 * We only allow the filesystem to be reloaded if it
168 		 * is currently mounted read-only.
169 		 */
170 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
171 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
172 			vput(vp);
173 			return (EOPNOTSUPP);	/* Needs translation */
174 		}
175 		/*
176 		 * In "highly secure" mode, don't let the caller do anything
177 		 * but downgrade a filesystem from read-write to read-only.
178 		 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
179 		 */
180 		if (securelevel >= 2 &&
181 		    SCARG(uap, flags) != MNT_GETARGS &&
182 		    SCARG(uap, flags) !=
183 		    (mp->mnt_flag | MNT_RDONLY |
184 		     MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
185 			vput(vp);
186 			return (EPERM);
187 		}
188 		mp->mnt_flag |= SCARG(uap, flags) &
189 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
190 		/*
191 		 * Only root, or the user that did the original mount is
192 		 * permitted to update it.
193 		 */
194 		if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
195 		    mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
196 		    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
197 			vput(vp);
198 			return (error);
199 		}
200 		/*
201 		 * Do not allow NFS export by non-root users. For non-root
202 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
203 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
204 		 */
205 		if (p->p_ucred->cr_uid != 0) {
206 			if (SCARG(uap, flags) & MNT_EXPORTED) {
207 				vput(vp);
208 				return (EPERM);
209 			}
210 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
211 			if (flag & MNT_NOEXEC)
212 				SCARG(uap, flags) |= MNT_NOEXEC;
213 		}
214 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
215 			vput(vp);
216 			return (EPERM);
217 		}
218 		goto update;
219 	} else {
220 		if (securelevel >= 2) {
221 			vput(vp);
222 			return (EPERM);
223 		}
224 	}
225 	/*
226 	 * If the user is not root, ensure that they own the directory
227 	 * onto which we are attempting to mount.
228 	 */
229 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
230 	    (va.va_uid != p->p_ucred->cr_uid &&
231 		(error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
232 		vput(vp);
233 		return (error);
234 	}
235 	/*
236 	 * Do not allow NFS export by non-root users. For non-root users,
237 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
238 	 * mount point is already MNT_NOEXEC.
239 	 */
240 	if (p->p_ucred->cr_uid != 0) {
241 		if (SCARG(uap, flags) & MNT_EXPORTED) {
242 			vput(vp);
243 			return (EPERM);
244 		}
245 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
246 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
247 			SCARG(uap, flags) |= MNT_NOEXEC;
248 	}
249 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) {
250 		vput(vp);
251 		return (error);
252 	}
253 	if (vp->v_type != VDIR) {
254 		vput(vp);
255 		return (ENOTDIR);
256 	}
257 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
258 	if (error) {
259 #if defined(COMPAT_09) || defined(COMPAT_43)
260 		/*
261 		 * Historically filesystem types were identified by number.
262 		 * If we get an integer for the filesystem type instead of a
263 		 * string, we check to see if it matches one of the historic
264 		 * filesystem types.
265 		 */
266 		u_long fsindex = (u_long)SCARG(uap, type);
267 		if (fsindex >= nmountcompatnames ||
268 		    mountcompatnames[fsindex] == NULL) {
269 			vput(vp);
270 			return (ENODEV);
271 		}
272 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
273 #else
274 		vput(vp);
275 		return (error);
276 #endif
277 	}
278 #ifdef	COMPAT_10
279 	/* Accept `ufs' as an alias for `ffs'. */
280 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
281 		strncpy(fstypename, "ffs", MFSNAMELEN);
282 #endif
283 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
284 		vput(vp);
285 		return (ENODEV);
286 	}
287 	if (vp->v_mountedhere != NULL) {
288 		vput(vp);
289 		return (EBUSY);
290 	}
291 
292 	/*
293 	 * Allocate and initialize the file system.
294 	 */
295 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
296 		M_MOUNT, M_WAITOK);
297 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
298 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
299 	simple_lock_init(&mp->mnt_slock);
300 	(void)vfs_busy(mp, LK_NOWAIT, 0);
301 	mp->mnt_op = vfs;
302 	vfs->vfs_refcount++;
303 	mp->mnt_vnodecovered = vp;
304 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
305 	mp->mnt_unmounter = NULL;
306 
307 	/*
308 	 * The underlying file system may refuse the mount for
309 	 * various reasons.  Allow the user to force it to happen.
310 	 */
311 	mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE;
312  update:
313 	/*
314 	 * Set the mount level flags.
315 	 */
316 	if (SCARG(uap, flags) & MNT_RDONLY)
317 		mp->mnt_flag |= MNT_RDONLY;
318 	else if (mp->mnt_flag & MNT_RDONLY)
319 		mp->mnt_iflag |= IMNT_WANTRDWR;
320 	mp->mnt_flag &=
321 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
322 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
323 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
324 	mp->mnt_flag |= SCARG(uap, flags) &
325 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
326 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
327 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
328 	    MNT_IGNORE);
329 	/*
330 	 * Mount the filesystem.
331 	 */
332 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
333 	if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
334 		if (mp->mnt_iflag & IMNT_WANTRDWR)
335 			mp->mnt_flag &= ~MNT_RDONLY;
336 		if (error || (mp->mnt_flag & MNT_GETARGS))
337 			mp->mnt_flag = flag;
338 		mp->mnt_flag &=~
339 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
340 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
341 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
342 			if (mp->mnt_syncer == NULL)
343 				error = vfs_allocate_syncvnode(mp);
344 		} else {
345 			if (mp->mnt_syncer != NULL)
346 				vfs_deallocate_syncvnode(mp);
347 		}
348 		vfs_unbusy(mp);
349 		VOP_UNLOCK(vp, 0);
350 		vrele(vp);
351 		return (error);
352 	}
353 	/*
354 	 * Put the new filesystem on the mount list after root.
355 	 */
356 	cache_purge(vp);
357 	if (!error) {
358 		mp->mnt_flag &=~
359 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
360 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
361 		vp->v_mountedhere = mp;
362 		simple_lock(&mountlist_slock);
363 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
364 		simple_unlock(&mountlist_slock);
365 		checkdirs(vp);
366 		VOP_UNLOCK(vp, 0);
367 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
368 			error = vfs_allocate_syncvnode(mp);
369 		vfs_unbusy(mp);
370 		(void) VFS_STATVFS(mp, &mp->mnt_stat, p);
371 		if ((error = VFS_START(mp, 0, p)))
372 			vrele(vp);
373 	} else {
374 		vp->v_mountedhere = (struct mount *)0;
375 		vfs->vfs_refcount--;
376 		vfs_unbusy(mp);
377 		free(mp, M_MOUNT);
378 		vput(vp);
379 	}
380 	return (error);
381 }
382 
383 /*
384  * Scan all active processes to see if any of them have a current
385  * or root directory onto which the new filesystem has just been
386  * mounted. If so, replace them with the new mount point.
387  */
388 void
389 checkdirs(olddp)
390 	struct vnode *olddp;
391 {
392 	struct cwdinfo *cwdi;
393 	struct vnode *newdp;
394 	struct proc *p;
395 
396 	if (olddp->v_usecount == 1)
397 		return;
398 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
399 		panic("mount: lost mount");
400 	proclist_lock_read();
401 	LIST_FOREACH(p, &allproc, p_list) {
402 		cwdi = p->p_cwdi;
403 		if (cwdi->cwdi_cdir == olddp) {
404 			vrele(cwdi->cwdi_cdir);
405 			VREF(newdp);
406 			cwdi->cwdi_cdir = newdp;
407 		}
408 		if (cwdi->cwdi_rdir == olddp) {
409 			vrele(cwdi->cwdi_rdir);
410 			VREF(newdp);
411 			cwdi->cwdi_rdir = newdp;
412 		}
413 	}
414 	proclist_unlock_read();
415 	if (rootvnode == olddp) {
416 		vrele(rootvnode);
417 		VREF(newdp);
418 		rootvnode = newdp;
419 	}
420 	vput(newdp);
421 }
422 
423 /*
424  * Unmount a file system.
425  *
426  * Note: unmount takes a path to the vnode mounted on as argument,
427  * not special file (as before).
428  */
429 /* ARGSUSED */
430 int
431 sys_unmount(l, v, retval)
432 	struct lwp *l;
433 	void *v;
434 	register_t *retval;
435 {
436 	struct sys_unmount_args /* {
437 		syscallarg(const char *) path;
438 		syscallarg(int) flags;
439 	} */ *uap = v;
440 	struct proc *p = l->l_proc;
441 	struct vnode *vp;
442 	struct mount *mp;
443 	int error;
444 	struct nameidata nd;
445 
446 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
447 	    SCARG(uap, path), p);
448 	if ((error = namei(&nd)) != 0)
449 		return (error);
450 	vp = nd.ni_vp;
451 	mp = vp->v_mount;
452 
453 	/*
454 	 * Only root, or the user that did the original mount is
455 	 * permitted to unmount this filesystem.
456 	 */
457 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
458 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
459 		vput(vp);
460 		return (error);
461 	}
462 
463 	/*
464 	 * Don't allow unmounting the root file system.
465 	 */
466 	if (mp->mnt_flag & MNT_ROOTFS) {
467 		vput(vp);
468 		return (EINVAL);
469 	}
470 
471 	/*
472 	 * Must be the root of the filesystem
473 	 */
474 	if ((vp->v_flag & VROOT) == 0) {
475 		vput(vp);
476 		return (EINVAL);
477 	}
478 	vput(vp);
479 
480 	/*
481 	 * XXX Freeze syncer.  Must do this before locking the
482 	 * mount point.  See dounmount() for details.
483 	 */
484 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
485 
486 	if (vfs_busy(mp, 0, 0)) {
487 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
488 		return (EBUSY);
489 	}
490 
491 	return (dounmount(mp, SCARG(uap, flags), p));
492 }
493 
494 /*
495  * Do the actual file system unmount. File system is assumed to have been
496  * marked busy by the caller.
497  */
498 int
499 dounmount(mp, flags, p)
500 	struct mount *mp;
501 	int flags;
502 	struct proc *p;
503 {
504 	struct vnode *coveredvp;
505 	int error;
506 	int async;
507 	int used_syncer;
508 
509 	simple_lock(&mountlist_slock);
510 	vfs_unbusy(mp);
511 	used_syncer = (mp->mnt_syncer != NULL);
512 
513 	/*
514 	 * XXX Syncer must be frozen when we get here.  This should really
515 	 * be done on a per-mountpoint basis, but especially the softdep
516 	 * code possibly called from the syncer doens't exactly work on a
517 	 * per-mountpoint basis, so the softdep code would become a maze
518 	 * of vfs_busy() calls.
519 	 *
520 	 * The caller of dounmount() must acquire syncer_lock because
521 	 * the syncer itself acquires locks in syncer_lock -> vfs_busy
522 	 * order, and we must preserve that order to avoid deadlock.
523 	 *
524 	 * So, if the file system did not use the syncer, now is
525 	 * the time to release the syncer_lock.
526 	 */
527 	if (used_syncer == 0)
528 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
529 
530 	mp->mnt_iflag |= IMNT_UNMOUNT;
531 	mp->mnt_unmounter = p;
532 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
533 	vn_start_write(NULL, &mp, V_WAIT);
534 
535 	if (mp->mnt_flag & MNT_EXPUBLIC)
536 		vfs_setpublicfs(NULL, NULL, NULL);
537 	async = mp->mnt_flag & MNT_ASYNC;
538 	mp->mnt_flag &= ~MNT_ASYNC;
539 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
540 	if (mp->mnt_syncer != NULL)
541 		vfs_deallocate_syncvnode(mp);
542 	error = 0;
543 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
544 #if NFSS > 0
545 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
546 #endif
547 		if (error == 0)
548 			error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
549 	}
550 	if (error == 0 || (flags & MNT_FORCE))
551 		error = VFS_UNMOUNT(mp, flags, p);
552 	vn_finished_write(mp, 0);
553 	simple_lock(&mountlist_slock);
554 	if (error) {
555 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
556 			(void) vfs_allocate_syncvnode(mp);
557 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
558 		mp->mnt_unmounter = NULL;
559 		mp->mnt_flag |= async;
560 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
561 		    &mountlist_slock);
562 		if (used_syncer)
563 			lockmgr(&syncer_lock, LK_RELEASE, NULL);
564 		simple_lock(&mp->mnt_slock);
565 		while (mp->mnt_wcnt > 0) {
566 			wakeup(mp);
567 			ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
568 				0, &mp->mnt_slock);
569 		}
570 		simple_unlock(&mp->mnt_slock);
571 		return (error);
572 	}
573 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
574 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
575 		coveredvp->v_mountedhere = NULL;
576 		vrele(coveredvp);
577 	}
578 	mp->mnt_op->vfs_refcount--;
579 	if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
580 		panic("unmount: dangling vnode");
581 	mp->mnt_iflag |= IMNT_GONE;
582 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
583 	if (used_syncer)
584 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
585 	simple_lock(&mp->mnt_slock);
586 	while (mp->mnt_wcnt > 0) {
587 		wakeup(mp);
588 		ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
589 	}
590 	simple_unlock(&mp->mnt_slock);
591 	free(mp, M_MOUNT);
592 	return (0);
593 }
594 
595 /*
596  * Sync each mounted filesystem.
597  */
598 #ifdef DEBUG
599 int syncprt = 0;
600 struct ctldebug debug0 = { "syncprt", &syncprt };
601 #endif
602 
603 /* ARGSUSED */
604 int
605 sys_sync(l, v, retval)
606 	struct lwp *l;
607 	void *v;
608 	register_t *retval;
609 {
610 	struct mount *mp, *nmp;
611 	int asyncflag;
612 	struct proc *p = l == NULL ? &proc0 : l->l_proc;
613 
614 	simple_lock(&mountlist_slock);
615 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
616 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
617 			nmp = mp->mnt_list.cqe_prev;
618 			continue;
619 		}
620 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
621 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
622 			asyncflag = mp->mnt_flag & MNT_ASYNC;
623 			mp->mnt_flag &= ~MNT_ASYNC;
624 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
625 			if (asyncflag)
626 				 mp->mnt_flag |= MNT_ASYNC;
627 			vn_finished_write(mp, 0);
628 		}
629 		simple_lock(&mountlist_slock);
630 		nmp = mp->mnt_list.cqe_prev;
631 		vfs_unbusy(mp);
632 
633 	}
634 	simple_unlock(&mountlist_slock);
635 #ifdef DEBUG
636 	if (syncprt)
637 		vfs_bufstats();
638 #endif /* DEBUG */
639 	return (0);
640 }
641 
642 /*
643  * Change filesystem quotas.
644  */
645 /* ARGSUSED */
646 int
647 sys_quotactl(l, v, retval)
648 	struct lwp *l;
649 	void *v;
650 	register_t *retval;
651 {
652 	struct sys_quotactl_args /* {
653 		syscallarg(const char *) path;
654 		syscallarg(int) cmd;
655 		syscallarg(int) uid;
656 		syscallarg(caddr_t) arg;
657 	} */ *uap = v;
658 	struct proc *p = l->l_proc;
659 	struct mount *mp;
660 	int error;
661 	struct nameidata nd;
662 
663 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
664 	if ((error = namei(&nd)) != 0)
665 		return (error);
666 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
667 	vrele(nd.ni_vp);
668 	if (error)
669 		return (error);
670 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
671 	    SCARG(uap, arg), p);
672 	vn_finished_write(mp, 0);
673 	return (error);
674 }
675 
676 int
677 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags,
678     int root)
679 {
680 	struct cwdinfo *cwdi = p->p_cwdi;
681 	int error = 0;
682 
683 	/*
684 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
685 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
686 	 * overrides MNT_NOWAIT.
687 	 */
688 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
689 	    (flags != MNT_WAIT && flags != 0)) {
690 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
691 		goto done;
692 	}
693 
694 	if ((error = VFS_STATVFS(mp, sp, p)) != 0) {
695 		return error;
696 	}
697 
698 	if (cwdi->cwdi_rdir == NULL)
699 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
700 done:
701 	if (cwdi->cwdi_rdir != NULL) {
702 		size_t len;
703 		char *bp;
704 		char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
705 		if (!path)
706 			return ENOMEM;
707 
708 		bp = path + MAXPATHLEN;
709 		*--bp = '\0';
710 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
711 		    MAXPATHLEN / 2, 0, p);
712 		if (error) {
713 			free(path, M_TEMP);
714 			return error;
715 		}
716 		len = strlen(bp);
717 		/*
718 		 * for mount points that are below our root, we can see
719 		 * them, so we fix up the pathname and return them. The
720 		 * rest we cannot see, so we don't allow viewing the
721 		 * data.
722 		 */
723 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
724 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
725 			    sizeof(sp->f_mntonname));
726 			if (sp->f_mntonname[0] == '\0')
727 				(void)strlcpy(sp->f_mntonname, "/",
728 				    sizeof(sp->f_mntonname));
729 		} else {
730 			if (root)
731 				(void)strlcpy(sp->f_mntonname, "/",
732 				    sizeof(sp->f_mntonname));
733 			else
734 				error = EPERM;
735 		}
736 		free(path, M_TEMP);
737 	}
738 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
739 	return error;
740 }
741 
742 /*
743  * Get filesystem statistics.
744  */
745 /* ARGSUSED */
746 int
747 sys_statvfs1(l, v, retval)
748 	struct lwp *l;
749 	void *v;
750 	register_t *retval;
751 {
752 	struct sys_statvfs1_args /* {
753 		syscallarg(const char *) path;
754 		syscallarg(struct statvfs *) buf;
755 		syscallarg(int) flags;
756 	} */ *uap = v;
757 	struct proc *p = l->l_proc;
758 	struct mount *mp;
759 	struct statvfs sbuf;
760 	int error;
761 	struct nameidata nd;
762 
763 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
764 	if ((error = namei(&nd)) != 0)
765 		return error;
766 	mp = nd.ni_vp->v_mount;
767 	vrele(nd.ni_vp);
768 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
769 		return error;
770 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
771 }
772 
773 /*
774  * Get filesystem statistics.
775  */
776 /* ARGSUSED */
777 int
778 sys_fstatvfs1(l, v, retval)
779 	struct lwp *l;
780 	void *v;
781 	register_t *retval;
782 {
783 	struct sys_fstatvfs1_args /* {
784 		syscallarg(int) fd;
785 		syscallarg(struct statvfs *) buf;
786 		syscallarg(int) flags;
787 	} */ *uap = v;
788 	struct proc *p = l->l_proc;
789 	struct file *fp;
790 	struct mount *mp;
791 	struct statvfs sbuf;
792 	int error;
793 
794 	/* getvnode() will use the descriptor for us */
795 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
796 		return (error);
797 	mp = ((struct vnode *)fp->f_data)->v_mount;
798 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
799 		goto out;
800 	error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
801  out:
802 	FILE_UNUSE(fp, p);
803 	return error;
804 }
805 
806 
807 /*
808  * Get statistics on all filesystems.
809  */
810 int
811 sys_getvfsstat(l, v, retval)
812 	struct lwp *l;
813 	void *v;
814 	register_t *retval;
815 {
816 	struct sys_getvfsstat_args /* {
817 		syscallarg(struct statvfs *) buf;
818 		syscallarg(size_t) bufsize;
819 		syscallarg(int) flags;
820 	} */ *uap = v;
821 	int root = 0;
822 	struct proc *p = l->l_proc;
823 	struct mount *mp, *nmp;
824 	struct statvfs sbuf;
825 	struct statvfs *sfsp;
826 	size_t count, maxcount;
827 	int error = 0;
828 
829 	maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs);
830 	sfsp = SCARG(uap, buf);
831 	simple_lock(&mountlist_slock);
832 	count = 0;
833 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
834 	     mp = nmp) {
835 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
836 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
837 			continue;
838 		}
839 		if (sfsp && count < maxcount) {
840 			error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0);
841 			if (error) {
842 				simple_lock(&mountlist_slock);
843 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
844 				vfs_unbusy(mp);
845 				continue;
846 			}
847 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
848 			if (error) {
849 				vfs_unbusy(mp);
850 				return (error);
851 			}
852 			sfsp++;
853 			root |= strcmp(sbuf.f_mntonname, "/") == 0;
854 		}
855 		count++;
856 		simple_lock(&mountlist_slock);
857 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
858 		vfs_unbusy(mp);
859 	}
860 	simple_unlock(&mountlist_slock);
861 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
862 		/*
863 		 * fake a root entry
864 		 */
865 		if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p,
866 		    SCARG(uap, flags), 1)) != 0)
867 			return error;
868 		if (sfsp)
869 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
870 		count++;
871 	}
872 	if (sfsp && count > maxcount)
873 		*retval = maxcount;
874 	else
875 		*retval = count;
876 	return error;
877 }
878 
879 /*
880  * Change current working directory to a given file descriptor.
881  */
882 /* ARGSUSED */
883 int
884 sys_fchdir(l, v, retval)
885 	struct lwp *l;
886 	void *v;
887 	register_t *retval;
888 {
889 	struct sys_fchdir_args /* {
890 		syscallarg(int) fd;
891 	} */ *uap = v;
892 	struct proc *p = l->l_proc;
893 	struct filedesc *fdp = p->p_fd;
894 	struct cwdinfo *cwdi = p->p_cwdi;
895 	struct vnode *vp, *tdp;
896 	struct mount *mp;
897 	struct file *fp;
898 	int error;
899 
900 	/* getvnode() will use the descriptor for us */
901 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
902 		return (error);
903 	vp = (struct vnode *)fp->f_data;
904 
905 	VREF(vp);
906 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
907 	if (vp->v_type != VDIR)
908 		error = ENOTDIR;
909 	else
910 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
911 	while (!error && (mp = vp->v_mountedhere) != NULL) {
912 		if (vfs_busy(mp, 0, 0))
913 			continue;
914 		error = VFS_ROOT(mp, &tdp);
915 		vfs_unbusy(mp);
916 		if (error)
917 			break;
918 		vput(vp);
919 		vp = tdp;
920 	}
921 	if (error) {
922 		vput(vp);
923 		goto out;
924 	}
925 	VOP_UNLOCK(vp, 0);
926 
927 	/*
928 	 * Disallow changing to a directory not under the process's
929 	 * current root directory (if there is one).
930 	 */
931 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
932 		vrele(vp);
933 		error = EPERM;	/* operation not permitted */
934 		goto out;
935 	}
936 
937 	vrele(cwdi->cwdi_cdir);
938 	cwdi->cwdi_cdir = vp;
939  out:
940 	FILE_UNUSE(fp, p);
941 	return (error);
942 }
943 
944 /*
945  * Change this process's notion of the root directory to a given file descriptor.
946  */
947 
948 int
949 sys_fchroot(l, v, retval)
950 	struct lwp *l;
951 	void *v;
952 	register_t *retval;
953 {
954 	struct sys_fchroot_args *uap = v;
955 	struct proc *p = l->l_proc;
956 	struct filedesc *fdp = p->p_fd;
957 	struct cwdinfo *cwdi = p->p_cwdi;
958 	struct vnode	*vp;
959 	struct file	*fp;
960 	int		 error;
961 
962 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
963 		return error;
964 	/* getvnode() will use the descriptor for us */
965 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
966 		return error;
967 	vp = (struct vnode *) fp->f_data;
968 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
969 	if (vp->v_type != VDIR)
970 		error = ENOTDIR;
971 	else
972 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
973 	VOP_UNLOCK(vp, 0);
974 	if (error)
975 		goto out;
976 	VREF(vp);
977 
978 	/*
979 	 * Prevent escaping from chroot by putting the root under
980 	 * the working directory.  Silently chdir to / if we aren't
981 	 * already there.
982 	 */
983 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
984 		/*
985 		 * XXX would be more failsafe to change directory to a
986 		 * deadfs node here instead
987 		 */
988 		vrele(cwdi->cwdi_cdir);
989 		VREF(vp);
990 		cwdi->cwdi_cdir = vp;
991 	}
992 
993 	if (cwdi->cwdi_rdir != NULL)
994 		vrele(cwdi->cwdi_rdir);
995 	cwdi->cwdi_rdir = vp;
996  out:
997 	FILE_UNUSE(fp, p);
998 	return (error);
999 }
1000 
1001 
1002 
1003 /*
1004  * Change current working directory (``.'').
1005  */
1006 /* ARGSUSED */
1007 int
1008 sys_chdir(l, v, retval)
1009 	struct lwp *l;
1010 	void *v;
1011 	register_t *retval;
1012 {
1013 	struct sys_chdir_args /* {
1014 		syscallarg(const char *) path;
1015 	} */ *uap = v;
1016 	struct proc *p = l->l_proc;
1017 	struct cwdinfo *cwdi = p->p_cwdi;
1018 	int error;
1019 	struct nameidata nd;
1020 
1021 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1022 	    SCARG(uap, path), p);
1023 	if ((error = change_dir(&nd, p)) != 0)
1024 		return (error);
1025 	vrele(cwdi->cwdi_cdir);
1026 	cwdi->cwdi_cdir = nd.ni_vp;
1027 	return (0);
1028 }
1029 
1030 /*
1031  * Change notion of root (``/'') directory.
1032  */
1033 /* ARGSUSED */
1034 int
1035 sys_chroot(l, v, retval)
1036 	struct lwp *l;
1037 	void *v;
1038 	register_t *retval;
1039 {
1040 	struct sys_chroot_args /* {
1041 		syscallarg(const char *) path;
1042 	} */ *uap = v;
1043 	struct proc *p = l->l_proc;
1044 	struct cwdinfo *cwdi = p->p_cwdi;
1045 	struct vnode *vp;
1046 	int error;
1047 	struct nameidata nd;
1048 
1049 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1050 		return (error);
1051 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1052 	    SCARG(uap, path), p);
1053 	if ((error = change_dir(&nd, p)) != 0)
1054 		return (error);
1055 	if (cwdi->cwdi_rdir != NULL)
1056 		vrele(cwdi->cwdi_rdir);
1057 	vp = nd.ni_vp;
1058 	cwdi->cwdi_rdir = vp;
1059 
1060 	/*
1061 	 * Prevent escaping from chroot by putting the root under
1062 	 * the working directory.  Silently chdir to / if we aren't
1063 	 * already there.
1064 	 */
1065 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1066 		/*
1067 		 * XXX would be more failsafe to change directory to a
1068 		 * deadfs node here instead
1069 		 */
1070 		vrele(cwdi->cwdi_cdir);
1071 		VREF(vp);
1072 		cwdi->cwdi_cdir = vp;
1073 	}
1074 
1075 	return (0);
1076 }
1077 
1078 /*
1079  * Common routine for chroot and chdir.
1080  */
1081 static int
1082 change_dir(ndp, p)
1083 	struct nameidata *ndp;
1084 	struct proc *p;
1085 {
1086 	struct vnode *vp;
1087 	int error;
1088 
1089 	if ((error = namei(ndp)) != 0)
1090 		return (error);
1091 	vp = ndp->ni_vp;
1092 	if (vp->v_type != VDIR)
1093 		error = ENOTDIR;
1094 	else
1095 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
1096 
1097 	if (error)
1098 		vput(vp);
1099 	else
1100 		VOP_UNLOCK(vp, 0);
1101 	return (error);
1102 }
1103 
1104 /*
1105  * Check permissions, allocate an open file structure,
1106  * and call the device open routine if any.
1107  */
1108 int
1109 sys_open(l, v, retval)
1110 	struct lwp *l;
1111 	void *v;
1112 	register_t *retval;
1113 {
1114 	struct sys_open_args /* {
1115 		syscallarg(const char *) path;
1116 		syscallarg(int) flags;
1117 		syscallarg(int) mode;
1118 	} */ *uap = v;
1119 	struct proc *p = l->l_proc;
1120 	struct cwdinfo *cwdi = p->p_cwdi;
1121 	struct filedesc *fdp = p->p_fd;
1122 	struct file *fp;
1123 	struct vnode *vp;
1124 	int flags, cmode;
1125 	int type, indx, error;
1126 	struct flock lf;
1127 	struct nameidata nd;
1128 
1129 	flags = FFLAGS(SCARG(uap, flags));
1130 	if ((flags & (FREAD | FWRITE)) == 0)
1131 		return (EINVAL);
1132 	/* falloc() will use the file descriptor for us */
1133 	if ((error = falloc(p, &fp, &indx)) != 0)
1134 		return (error);
1135 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1136 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1137 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1138 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1139 		FILE_UNUSE(fp, p);
1140 		fdp->fd_ofiles[indx] = NULL;
1141 		ffree(fp);
1142 		if ((error == ENODEV || error == ENXIO) &&
1143 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1144 		    (error =
1145 			dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) {
1146 			*retval = indx;
1147 			return (0);
1148 		}
1149 		if (error == ERESTART)
1150 			error = EINTR;
1151 		fdremove(fdp, indx);
1152 		return (error);
1153 	}
1154 	l->l_dupfd = 0;
1155 	vp = nd.ni_vp;
1156 	fp->f_flag = flags & FMASK;
1157 	fp->f_type = DTYPE_VNODE;
1158 	fp->f_ops = &vnops;
1159 	fp->f_data = vp;
1160 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1161 		lf.l_whence = SEEK_SET;
1162 		lf.l_start = 0;
1163 		lf.l_len = 0;
1164 		if (flags & O_EXLOCK)
1165 			lf.l_type = F_WRLCK;
1166 		else
1167 			lf.l_type = F_RDLCK;
1168 		type = F_FLOCK;
1169 		if ((flags & FNONBLOCK) == 0)
1170 			type |= F_WAIT;
1171 		VOP_UNLOCK(vp, 0);
1172 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1173 		if (error) {
1174 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1175 			FILE_UNUSE(fp, p);
1176 			ffree(fp);
1177 			fdremove(fdp, indx);
1178 			return (error);
1179 		}
1180 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1181 		fp->f_flag |= FHASLOCK;
1182 	}
1183 	VOP_UNLOCK(vp, 0);
1184 	*retval = indx;
1185 	FILE_SET_MATURE(fp);
1186 	FILE_UNUSE(fp, p);
1187 	return (0);
1188 }
1189 
1190 /*
1191  * Get file handle system call
1192  */
1193 int
1194 sys_getfh(l, v, retval)
1195 	struct lwp *l;
1196 	void *v;
1197 	register_t *retval;
1198 {
1199 	struct sys_getfh_args /* {
1200 		syscallarg(char *) fname;
1201 		syscallarg(fhandle_t *) fhp;
1202 	} */ *uap = v;
1203 	struct proc *p = l->l_proc;
1204 	struct vnode *vp;
1205 	fhandle_t fh;
1206 	int error;
1207 	struct nameidata nd;
1208 
1209 	/*
1210 	 * Must be super user
1211 	 */
1212 	error = suser(p->p_ucred, &p->p_acflag);
1213 	if (error)
1214 		return (error);
1215 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1216 	    SCARG(uap, fname), p);
1217 	error = namei(&nd);
1218 	if (error)
1219 		return (error);
1220 	vp = nd.ni_vp;
1221 	memset(&fh, 0, sizeof(fh));
1222 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1223 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1224 	vput(vp);
1225 	if (error)
1226 		return (error);
1227 	error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1228 	return (error);
1229 }
1230 
1231 /*
1232  * Open a file given a file handle.
1233  *
1234  * Check permissions, allocate an open file structure,
1235  * and call the device open routine if any.
1236  */
1237 int
1238 sys_fhopen(l, v, retval)
1239 	struct lwp *l;
1240 	void *v;
1241 	register_t *retval;
1242 {
1243 	struct sys_fhopen_args /* {
1244 		syscallarg(const fhandle_t *) fhp;
1245 		syscallarg(int) flags;
1246 	} */ *uap = v;
1247 	struct proc *p = l->l_proc;
1248 	struct filedesc *fdp = p->p_fd;
1249 	struct file *fp;
1250 	struct vnode *vp = NULL;
1251 	struct mount *mp;
1252 	struct ucred *cred = p->p_ucred;
1253 	int flags;
1254 	struct file *nfp;
1255 	int type, indx, error=0;
1256 	struct flock lf;
1257 	struct vattr va;
1258 	fhandle_t fh;
1259 
1260 	/*
1261 	 * Must be super user
1262 	 */
1263 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1264 		return (error);
1265 
1266 	flags = FFLAGS(SCARG(uap, flags));
1267 	if ((flags & (FREAD | FWRITE)) == 0)
1268 		return (EINVAL);
1269 	if ((flags & O_CREAT))
1270 		return (EINVAL);
1271 	/* falloc() will use the file descriptor for us */
1272 	if ((error = falloc(p, &nfp, &indx)) != 0)
1273 		return (error);
1274 	fp = nfp;
1275 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1276 		goto bad;
1277 
1278 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1279 		error = ESTALE;
1280 		goto bad;
1281 	}
1282 
1283 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1284 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1285 		goto bad;
1286 	}
1287 
1288 	/* Now do an effective vn_open */
1289 
1290 	if (vp->v_type == VSOCK) {
1291 		error = EOPNOTSUPP;
1292 		goto bad;
1293 	}
1294 	if (flags & FREAD) {
1295 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1296 			goto bad;
1297 	}
1298 	if (flags & (FWRITE | O_TRUNC)) {
1299 		if (vp->v_type == VDIR) {
1300 			error = EISDIR;
1301 			goto bad;
1302 		}
1303 		if ((error = vn_writechk(vp)) != 0 ||
1304 		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1305 			goto bad;
1306 	}
1307 	if (flags & O_TRUNC) {
1308 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
1309 			goto bad;
1310 		VOP_UNLOCK(vp, 0);			/* XXX */
1311 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1312 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1313 		VATTR_NULL(&va);
1314 		va.va_size = 0;
1315 		error = VOP_SETATTR(vp, &va, cred, p);
1316 		vn_finished_write(mp, 0);
1317 		if (error)
1318 			goto bad;
1319 	}
1320 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1321 		goto bad;
1322 	if (vp->v_type == VREG &&
1323 	    uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1324 		error = EIO;
1325 		goto bad;
1326 	}
1327 	if (flags & FWRITE)
1328 		vp->v_writecount++;
1329 
1330 	/* done with modified vn_open, now finish what sys_open does. */
1331 
1332 	fp->f_flag = flags & FMASK;
1333 	fp->f_type = DTYPE_VNODE;
1334 	fp->f_ops = &vnops;
1335 	fp->f_data = vp;
1336 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1337 		lf.l_whence = SEEK_SET;
1338 		lf.l_start = 0;
1339 		lf.l_len = 0;
1340 		if (flags & O_EXLOCK)
1341 			lf.l_type = F_WRLCK;
1342 		else
1343 			lf.l_type = F_RDLCK;
1344 		type = F_FLOCK;
1345 		if ((flags & FNONBLOCK) == 0)
1346 			type |= F_WAIT;
1347 		VOP_UNLOCK(vp, 0);
1348 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1349 		if (error) {
1350 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1351 			FILE_UNUSE(fp, p);
1352 			ffree(fp);
1353 			fdremove(fdp, indx);
1354 			return (error);
1355 		}
1356 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1357 		fp->f_flag |= FHASLOCK;
1358 	}
1359 	VOP_UNLOCK(vp, 0);
1360 	*retval = indx;
1361 	FILE_SET_MATURE(fp);
1362 	FILE_UNUSE(fp, p);
1363 	return (0);
1364 
1365 bad:
1366 	FILE_UNUSE(fp, p);
1367 	ffree(fp);
1368 	fdremove(fdp, indx);
1369 	if (vp != NULL)
1370 		vput(vp);
1371 	return (error);
1372 }
1373 
1374 /* ARGSUSED */
1375 int
1376 sys_fhstat(l, v, retval)
1377 	struct lwp *l;
1378 	void *v;
1379 	register_t *retval;
1380 {
1381 	struct sys_fhstat_args /* {
1382 		syscallarg(const fhandle_t *) fhp;
1383 		syscallarg(struct stat *) sb;
1384 	} */ *uap = v;
1385 	struct proc *p = l->l_proc;
1386 	struct stat sb;
1387 	int error;
1388 	fhandle_t fh;
1389 	struct mount *mp;
1390 	struct vnode *vp;
1391 
1392 	/*
1393 	 * Must be super user
1394 	 */
1395 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1396 		return (error);
1397 
1398 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1399 		return (error);
1400 
1401 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1402 		return (ESTALE);
1403 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1404 		return (error);
1405 	error = vn_stat(vp, &sb, p);
1406 	vput(vp);
1407 	if (error)
1408 		return (error);
1409 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1410 	return (error);
1411 }
1412 
1413 /* ARGSUSED */
1414 int
1415 sys_fhstatvfs1(l, v, retval)
1416 	struct lwp *l;
1417 	void *v;
1418 	register_t *retval;
1419 {
1420 	struct sys_fhstatvfs1_args /*
1421 		syscallarg(const fhandle_t *) fhp;
1422 		syscallarg(struct statvfs *) buf;
1423 		syscallarg(int)	flags;
1424 	} */ *uap = v;
1425 	struct proc *p = l->l_proc;
1426 	struct statvfs sbuf;
1427 	fhandle_t fh;
1428 	struct mount *mp;
1429 	struct vnode *vp;
1430 	int error;
1431 
1432 	/*
1433 	 * Must be super user
1434 	 */
1435 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1436 		return error;
1437 
1438 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1439 		return error;
1440 
1441 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1442 		return ESTALE;
1443 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1444 		return error;
1445 
1446 	mp = vp->v_mount;
1447 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) {
1448 		vput(vp);
1449 		return error;
1450 	}
1451 	vput(vp);
1452 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
1453 }
1454 
1455 /*
1456  * Create a special file.
1457  */
1458 /* ARGSUSED */
1459 int
1460 sys_mknod(l, v, retval)
1461 	struct lwp *l;
1462 	void *v;
1463 	register_t *retval;
1464 {
1465 	struct sys_mknod_args /* {
1466 		syscallarg(const char *) path;
1467 		syscallarg(int) mode;
1468 		syscallarg(int) dev;
1469 	} */ *uap = v;
1470 	struct proc *p = l->l_proc;
1471 	struct vnode *vp;
1472 	struct mount *mp;
1473 	struct vattr vattr;
1474 	int error;
1475 	int whiteout = 0;
1476 	struct nameidata nd;
1477 
1478 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1479 		return (error);
1480 restart:
1481 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1482 	if ((error = namei(&nd)) != 0)
1483 		return (error);
1484 	vp = nd.ni_vp;
1485 	if (vp != NULL)
1486 		error = EEXIST;
1487 	else {
1488 		VATTR_NULL(&vattr);
1489 		vattr.va_mode =
1490 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1491 		vattr.va_rdev = SCARG(uap, dev);
1492 		whiteout = 0;
1493 
1494 		switch (SCARG(uap, mode) & S_IFMT) {
1495 		case S_IFMT:	/* used by badsect to flag bad sectors */
1496 			vattr.va_type = VBAD;
1497 			break;
1498 		case S_IFCHR:
1499 			vattr.va_type = VCHR;
1500 			break;
1501 		case S_IFBLK:
1502 			vattr.va_type = VBLK;
1503 			break;
1504 		case S_IFWHT:
1505 			whiteout = 1;
1506 			break;
1507 		default:
1508 			error = EINVAL;
1509 			break;
1510 		}
1511 	}
1512 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1513 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1514 		if (nd.ni_dvp == vp)
1515 			vrele(nd.ni_dvp);
1516 		else
1517 			vput(nd.ni_dvp);
1518 		if (vp)
1519 			vrele(vp);
1520 		if ((error = vn_start_write(NULL, &mp,
1521 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1522 			return (error);
1523 		goto restart;
1524 	}
1525 	if (!error) {
1526 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1527 		if (whiteout) {
1528 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1529 			if (error)
1530 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1531 			vput(nd.ni_dvp);
1532 		} else {
1533 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1534 						&nd.ni_cnd, &vattr);
1535 			if (error == 0)
1536 				vput(nd.ni_vp);
1537 		}
1538 	} else {
1539 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1540 		if (nd.ni_dvp == vp)
1541 			vrele(nd.ni_dvp);
1542 		else
1543 			vput(nd.ni_dvp);
1544 		if (vp)
1545 			vrele(vp);
1546 	}
1547 	vn_finished_write(mp, 0);
1548 	return (error);
1549 }
1550 
1551 /*
1552  * Create a named pipe.
1553  */
1554 /* ARGSUSED */
1555 int
1556 sys_mkfifo(l, v, retval)
1557 	struct lwp *l;
1558 	void *v;
1559 	register_t *retval;
1560 {
1561 	struct sys_mkfifo_args /* {
1562 		syscallarg(const char *) path;
1563 		syscallarg(int) mode;
1564 	} */ *uap = v;
1565 	struct proc *p = l->l_proc;
1566 	struct mount *mp;
1567 	struct vattr vattr;
1568 	int error;
1569 	struct nameidata nd;
1570 
1571 restart:
1572 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1573 	if ((error = namei(&nd)) != 0)
1574 		return (error);
1575 	if (nd.ni_vp != NULL) {
1576 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1577 		if (nd.ni_dvp == nd.ni_vp)
1578 			vrele(nd.ni_dvp);
1579 		else
1580 			vput(nd.ni_dvp);
1581 		vrele(nd.ni_vp);
1582 		return (EEXIST);
1583 	}
1584 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1585 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1586 		if (nd.ni_dvp == nd.ni_vp)
1587 			vrele(nd.ni_dvp);
1588 		else
1589 			vput(nd.ni_dvp);
1590 		if (nd.ni_vp)
1591 			vrele(nd.ni_vp);
1592 		if ((error = vn_start_write(NULL, &mp,
1593 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1594 			return (error);
1595 		goto restart;
1596 	}
1597 	VATTR_NULL(&vattr);
1598 	vattr.va_type = VFIFO;
1599 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1600 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1601 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1602 	if (error == 0)
1603 		vput(nd.ni_vp);
1604 	vn_finished_write(mp, 0);
1605 	return (error);
1606 }
1607 
1608 /*
1609  * Make a hard file link.
1610  */
1611 /* ARGSUSED */
1612 int
1613 sys_link(l, v, retval)
1614 	struct lwp *l;
1615 	void *v;
1616 	register_t *retval;
1617 {
1618 	struct sys_link_args /* {
1619 		syscallarg(const char *) path;
1620 		syscallarg(const char *) link;
1621 	} */ *uap = v;
1622 	struct proc *p = l->l_proc;
1623 	struct vnode *vp;
1624 	struct mount *mp;
1625 	struct nameidata nd;
1626 	int error;
1627 
1628 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1629 	if ((error = namei(&nd)) != 0)
1630 		return (error);
1631 	vp = nd.ni_vp;
1632 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
1633 		vrele(vp);
1634 		return (error);
1635 	}
1636 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1637 	if ((error = namei(&nd)) != 0)
1638 		goto out;
1639 	if (nd.ni_vp) {
1640 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1641 		if (nd.ni_dvp == nd.ni_vp)
1642 			vrele(nd.ni_dvp);
1643 		else
1644 			vput(nd.ni_dvp);
1645 		vrele(nd.ni_vp);
1646 		error = EEXIST;
1647 		goto out;
1648 	}
1649 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1650 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1651 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1652 out:
1653 	vrele(vp);
1654 	vn_finished_write(mp, 0);
1655 	return (error);
1656 }
1657 
1658 /*
1659  * Make a symbolic link.
1660  */
1661 /* ARGSUSED */
1662 int
1663 sys_symlink(l, v, retval)
1664 	struct lwp *l;
1665 	void *v;
1666 	register_t *retval;
1667 {
1668 	struct sys_symlink_args /* {
1669 		syscallarg(const char *) path;
1670 		syscallarg(const char *) link;
1671 	} */ *uap = v;
1672 	struct proc *p = l->l_proc;
1673 	struct mount *mp;
1674 	struct vattr vattr;
1675 	char *path;
1676 	int error;
1677 	struct nameidata nd;
1678 
1679 	path = PNBUF_GET();
1680 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1681 	if (error)
1682 		goto out;
1683 restart:
1684 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1685 	if ((error = namei(&nd)) != 0)
1686 		goto out;
1687 	if (nd.ni_vp) {
1688 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1689 		if (nd.ni_dvp == nd.ni_vp)
1690 			vrele(nd.ni_dvp);
1691 		else
1692 			vput(nd.ni_dvp);
1693 		vrele(nd.ni_vp);
1694 		error = EEXIST;
1695 		goto out;
1696 	}
1697 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1698 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1699 		if (nd.ni_dvp == nd.ni_vp)
1700 			vrele(nd.ni_dvp);
1701 		else
1702 			vput(nd.ni_dvp);
1703 		if ((error = vn_start_write(NULL, &mp,
1704 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1705 			return (error);
1706 		goto restart;
1707 	}
1708 	VATTR_NULL(&vattr);
1709 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1710 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1711 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1712 	if (error == 0)
1713 		vput(nd.ni_vp);
1714 	vn_finished_write(mp, 0);
1715 out:
1716 	PNBUF_PUT(path);
1717 	return (error);
1718 }
1719 
1720 /*
1721  * Delete a whiteout from the filesystem.
1722  */
1723 /* ARGSUSED */
1724 int
1725 sys_undelete(l, v, retval)
1726 	struct lwp *l;
1727 	void *v;
1728 	register_t *retval;
1729 {
1730 	struct sys_undelete_args /* {
1731 		syscallarg(const char *) path;
1732 	} */ *uap = v;
1733 	struct proc *p = l->l_proc;
1734 	int error;
1735 	struct mount *mp;
1736 	struct nameidata nd;
1737 
1738 restart:
1739 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1740 	    SCARG(uap, path), p);
1741 	error = namei(&nd);
1742 	if (error)
1743 		return (error);
1744 
1745 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1746 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1747 		if (nd.ni_dvp == nd.ni_vp)
1748 			vrele(nd.ni_dvp);
1749 		else
1750 			vput(nd.ni_dvp);
1751 		if (nd.ni_vp)
1752 			vrele(nd.ni_vp);
1753 		return (EEXIST);
1754 	}
1755 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1756 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1757 		if (nd.ni_dvp == nd.ni_vp)
1758 			vrele(nd.ni_dvp);
1759 		else
1760 			vput(nd.ni_dvp);
1761 		if ((error = vn_start_write(NULL, &mp,
1762 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1763 			return (error);
1764 		goto restart;
1765 	}
1766 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1767 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1768 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1769 	vput(nd.ni_dvp);
1770 	vn_finished_write(mp, 0);
1771 	return (error);
1772 }
1773 
1774 /*
1775  * Delete a name from the filesystem.
1776  */
1777 /* ARGSUSED */
1778 int
1779 sys_unlink(l, v, retval)
1780 	struct lwp *l;
1781 	void *v;
1782 	register_t *retval;
1783 {
1784 	struct sys_unlink_args /* {
1785 		syscallarg(const char *) path;
1786 	} */ *uap = v;
1787 	struct proc *p = l->l_proc;
1788 	struct mount *mp;
1789 	struct vnode *vp;
1790 	int error;
1791 	struct nameidata nd;
1792 
1793 restart:
1794 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1795 	    SCARG(uap, path), p);
1796 	if ((error = namei(&nd)) != 0)
1797 		return (error);
1798 	vp = nd.ni_vp;
1799 
1800 	/*
1801 	 * The root of a mounted filesystem cannot be deleted.
1802 	 */
1803 	if (vp->v_flag & VROOT) {
1804 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1805 		if (nd.ni_dvp == vp)
1806 			vrele(nd.ni_dvp);
1807 		else
1808 			vput(nd.ni_dvp);
1809 		vput(vp);
1810 		error = EBUSY;
1811 		goto out;
1812 	}
1813 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1814 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1815 		if (nd.ni_dvp == vp)
1816 			vrele(nd.ni_dvp);
1817 		else
1818 			vput(nd.ni_dvp);
1819 		vput(vp);
1820 		if ((error = vn_start_write(NULL, &mp,
1821 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1822 			return (error);
1823 		goto restart;
1824 	}
1825 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1826 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1827 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1828 	vn_finished_write(mp, 0);
1829 out:
1830 	return (error);
1831 }
1832 
1833 /*
1834  * Reposition read/write file offset.
1835  */
1836 int
1837 sys_lseek(l, v, retval)
1838 	struct lwp *l;
1839 	void *v;
1840 	register_t *retval;
1841 {
1842 	struct sys_lseek_args /* {
1843 		syscallarg(int) fd;
1844 		syscallarg(int) pad;
1845 		syscallarg(off_t) offset;
1846 		syscallarg(int) whence;
1847 	} */ *uap = v;
1848 	struct proc *p = l->l_proc;
1849 	struct ucred *cred = p->p_ucred;
1850 	struct filedesc *fdp = p->p_fd;
1851 	struct file *fp;
1852 	struct vnode *vp;
1853 	struct vattr vattr;
1854 	off_t newoff;
1855 	int error;
1856 
1857 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1858 		return (EBADF);
1859 
1860 	FILE_USE(fp);
1861 
1862 	vp = (struct vnode *)fp->f_data;
1863 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1864 		error = ESPIPE;
1865 		goto out;
1866 	}
1867 
1868 	switch (SCARG(uap, whence)) {
1869 	case SEEK_CUR:
1870 		newoff = fp->f_offset + SCARG(uap, offset);
1871 		break;
1872 	case SEEK_END:
1873 		error = VOP_GETATTR(vp, &vattr, cred, p);
1874 		if (error)
1875 			goto out;
1876 		newoff = SCARG(uap, offset) + vattr.va_size;
1877 		break;
1878 	case SEEK_SET:
1879 		newoff = SCARG(uap, offset);
1880 		break;
1881 	default:
1882 		error = EINVAL;
1883 		goto out;
1884 	}
1885 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1886 		goto out;
1887 
1888 	*(off_t *)retval = fp->f_offset = newoff;
1889  out:
1890 	FILE_UNUSE(fp, p);
1891 	return (error);
1892 }
1893 
1894 /*
1895  * Positional read system call.
1896  */
1897 int
1898 sys_pread(l, v, retval)
1899 	struct lwp *l;
1900 	void *v;
1901 	register_t *retval;
1902 {
1903 	struct sys_pread_args /* {
1904 		syscallarg(int) fd;
1905 		syscallarg(void *) buf;
1906 		syscallarg(size_t) nbyte;
1907 		syscallarg(off_t) offset;
1908 	} */ *uap = v;
1909 	struct proc *p = l->l_proc;
1910 	struct filedesc *fdp = p->p_fd;
1911 	struct file *fp;
1912 	struct vnode *vp;
1913 	off_t offset;
1914 	int error, fd = SCARG(uap, fd);
1915 
1916 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1917 		return (EBADF);
1918 
1919 	if ((fp->f_flag & FREAD) == 0) {
1920 		simple_unlock(&fp->f_slock);
1921 		return (EBADF);
1922 	}
1923 
1924 	FILE_USE(fp);
1925 
1926 	vp = (struct vnode *)fp->f_data;
1927 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1928 		error = ESPIPE;
1929 		goto out;
1930 	}
1931 
1932 	offset = SCARG(uap, offset);
1933 
1934 	/*
1935 	 * XXX This works because no file systems actually
1936 	 * XXX take any action on the seek operation.
1937 	 */
1938 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1939 		goto out;
1940 
1941 	/* dofileread() will unuse the descriptor for us */
1942 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1943 	    &offset, 0, retval));
1944 
1945  out:
1946 	FILE_UNUSE(fp, p);
1947 	return (error);
1948 }
1949 
1950 /*
1951  * Positional scatter read system call.
1952  */
1953 int
1954 sys_preadv(l, v, retval)
1955 	struct lwp *l;
1956 	void *v;
1957 	register_t *retval;
1958 {
1959 	struct sys_preadv_args /* {
1960 		syscallarg(int) fd;
1961 		syscallarg(const struct iovec *) iovp;
1962 		syscallarg(int) iovcnt;
1963 		syscallarg(off_t) offset;
1964 	} */ *uap = v;
1965 	struct proc *p = l->l_proc;
1966 	struct filedesc *fdp = p->p_fd;
1967 	struct file *fp;
1968 	struct vnode *vp;
1969 	off_t offset;
1970 	int error, fd = SCARG(uap, fd);
1971 
1972 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1973 		return (EBADF);
1974 
1975 	if ((fp->f_flag & FREAD) == 0) {
1976 		simple_unlock(&fp->f_slock);
1977 		return (EBADF);
1978 	}
1979 
1980 	FILE_USE(fp);
1981 
1982 	vp = (struct vnode *)fp->f_data;
1983 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1984 		error = ESPIPE;
1985 		goto out;
1986 	}
1987 
1988 	offset = SCARG(uap, offset);
1989 
1990 	/*
1991 	 * XXX This works because no file systems actually
1992 	 * XXX take any action on the seek operation.
1993 	 */
1994 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1995 		goto out;
1996 
1997 	/* dofilereadv() will unuse the descriptor for us */
1998 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1999 	    &offset, 0, retval));
2000 
2001  out:
2002 	FILE_UNUSE(fp, p);
2003 	return (error);
2004 }
2005 
2006 /*
2007  * Positional write system call.
2008  */
2009 int
2010 sys_pwrite(l, v, retval)
2011 	struct lwp *l;
2012 	void *v;
2013 	register_t *retval;
2014 {
2015 	struct sys_pwrite_args /* {
2016 		syscallarg(int) fd;
2017 		syscallarg(const void *) buf;
2018 		syscallarg(size_t) nbyte;
2019 		syscallarg(off_t) offset;
2020 	} */ *uap = v;
2021 	struct proc *p = l->l_proc;
2022 	struct filedesc *fdp = p->p_fd;
2023 	struct file *fp;
2024 	struct vnode *vp;
2025 	off_t offset;
2026 	int error, fd = SCARG(uap, fd);
2027 
2028 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2029 		return (EBADF);
2030 
2031 	if ((fp->f_flag & FWRITE) == 0) {
2032 		simple_unlock(&fp->f_slock);
2033 		return (EBADF);
2034 	}
2035 
2036 	FILE_USE(fp);
2037 
2038 	vp = (struct vnode *)fp->f_data;
2039 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2040 		error = ESPIPE;
2041 		goto out;
2042 	}
2043 
2044 	offset = SCARG(uap, offset);
2045 
2046 	/*
2047 	 * XXX This works because no file systems actually
2048 	 * XXX take any action on the seek operation.
2049 	 */
2050 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2051 		goto out;
2052 
2053 	/* dofilewrite() will unuse the descriptor for us */
2054 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2055 	    &offset, 0, retval));
2056 
2057  out:
2058 	FILE_UNUSE(fp, p);
2059 	return (error);
2060 }
2061 
2062 /*
2063  * Positional gather write system call.
2064  */
2065 int
2066 sys_pwritev(l, v, retval)
2067 	struct lwp *l;
2068 	void *v;
2069 	register_t *retval;
2070 {
2071 	struct sys_pwritev_args /* {
2072 		syscallarg(int) fd;
2073 		syscallarg(const struct iovec *) iovp;
2074 		syscallarg(int) iovcnt;
2075 		syscallarg(off_t) offset;
2076 	} */ *uap = v;
2077 	struct proc *p = l->l_proc;
2078 	struct filedesc *fdp = p->p_fd;
2079 	struct file *fp;
2080 	struct vnode *vp;
2081 	off_t offset;
2082 	int error, fd = SCARG(uap, fd);
2083 
2084 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2085 		return (EBADF);
2086 
2087 	if ((fp->f_flag & FWRITE) == 0) {
2088 		simple_unlock(&fp->f_slock);
2089 		return (EBADF);
2090 	}
2091 
2092 	FILE_USE(fp);
2093 
2094 	vp = (struct vnode *)fp->f_data;
2095 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2096 		error = ESPIPE;
2097 		goto out;
2098 	}
2099 
2100 	offset = SCARG(uap, offset);
2101 
2102 	/*
2103 	 * XXX This works because no file systems actually
2104 	 * XXX take any action on the seek operation.
2105 	 */
2106 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2107 		goto out;
2108 
2109 	/* dofilewritev() will unuse the descriptor for us */
2110 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2111 	    &offset, 0, retval));
2112 
2113  out:
2114 	FILE_UNUSE(fp, p);
2115 	return (error);
2116 }
2117 
2118 /*
2119  * Check access permissions.
2120  */
2121 int
2122 sys_access(l, v, retval)
2123 	struct lwp *l;
2124 	void *v;
2125 	register_t *retval;
2126 {
2127 	struct sys_access_args /* {
2128 		syscallarg(const char *) path;
2129 		syscallarg(int) flags;
2130 	} */ *uap = v;
2131 	struct proc *p = l->l_proc;
2132 	struct ucred *cred;
2133 	struct vnode *vp;
2134 	int error, flags;
2135 	struct nameidata nd;
2136 
2137 	cred = crdup(p->p_ucred);
2138 	cred->cr_uid = p->p_cred->p_ruid;
2139 	cred->cr_gid = p->p_cred->p_rgid;
2140 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2141 	    SCARG(uap, path), p);
2142 	/* Override default credentials */
2143 	nd.ni_cnd.cn_cred = cred;
2144 	if ((error = namei(&nd)) != 0)
2145 		goto out;
2146 	vp = nd.ni_vp;
2147 
2148 	/* Flags == 0 means only check for existence. */
2149 	if (SCARG(uap, flags)) {
2150 		flags = 0;
2151 		if (SCARG(uap, flags) & R_OK)
2152 			flags |= VREAD;
2153 		if (SCARG(uap, flags) & W_OK)
2154 			flags |= VWRITE;
2155 		if (SCARG(uap, flags) & X_OK)
2156 			flags |= VEXEC;
2157 
2158 		error = VOP_ACCESS(vp, flags, cred, p);
2159 		if (!error && (flags & VWRITE))
2160 			error = vn_writechk(vp);
2161 	}
2162 	vput(vp);
2163 out:
2164 	crfree(cred);
2165 	return (error);
2166 }
2167 
2168 /*
2169  * Get file status; this version follows links.
2170  */
2171 /* ARGSUSED */
2172 int
2173 sys___stat13(l, v, retval)
2174 	struct lwp *l;
2175 	void *v;
2176 	register_t *retval;
2177 {
2178 	struct sys___stat13_args /* {
2179 		syscallarg(const char *) path;
2180 		syscallarg(struct stat *) ub;
2181 	} */ *uap = v;
2182 	struct proc *p = l->l_proc;
2183 	struct stat sb;
2184 	int error;
2185 	struct nameidata nd;
2186 
2187 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2188 	    SCARG(uap, path), p);
2189 	if ((error = namei(&nd)) != 0)
2190 		return (error);
2191 	error = vn_stat(nd.ni_vp, &sb, p);
2192 	vput(nd.ni_vp);
2193 	if (error)
2194 		return (error);
2195 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2196 	return (error);
2197 }
2198 
2199 /*
2200  * Get file status; this version does not follow links.
2201  */
2202 /* ARGSUSED */
2203 int
2204 sys___lstat13(l, v, retval)
2205 	struct lwp *l;
2206 	void *v;
2207 	register_t *retval;
2208 {
2209 	struct sys___lstat13_args /* {
2210 		syscallarg(const char *) path;
2211 		syscallarg(struct stat *) ub;
2212 	} */ *uap = v;
2213 	struct proc *p = l->l_proc;
2214 	struct stat sb;
2215 	int error;
2216 	struct nameidata nd;
2217 
2218 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2219 	    SCARG(uap, path), p);
2220 	if ((error = namei(&nd)) != 0)
2221 		return (error);
2222 	error = vn_stat(nd.ni_vp, &sb, p);
2223 	vput(nd.ni_vp);
2224 	if (error)
2225 		return (error);
2226 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2227 	return (error);
2228 }
2229 
2230 /*
2231  * Get configurable pathname variables.
2232  */
2233 /* ARGSUSED */
2234 int
2235 sys_pathconf(l, v, retval)
2236 	struct lwp *l;
2237 	void *v;
2238 	register_t *retval;
2239 {
2240 	struct sys_pathconf_args /* {
2241 		syscallarg(const char *) path;
2242 		syscallarg(int) name;
2243 	} */ *uap = v;
2244 	struct proc *p = l->l_proc;
2245 	int error;
2246 	struct nameidata nd;
2247 
2248 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2249 	    SCARG(uap, path), p);
2250 	if ((error = namei(&nd)) != 0)
2251 		return (error);
2252 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2253 	vput(nd.ni_vp);
2254 	return (error);
2255 }
2256 
2257 /*
2258  * Return target name of a symbolic link.
2259  */
2260 /* ARGSUSED */
2261 int
2262 sys_readlink(l, v, retval)
2263 	struct lwp *l;
2264 	void *v;
2265 	register_t *retval;
2266 {
2267 	struct sys_readlink_args /* {
2268 		syscallarg(const char *) path;
2269 		syscallarg(char *) buf;
2270 		syscallarg(size_t) count;
2271 	} */ *uap = v;
2272 	struct proc *p = l->l_proc;
2273 	struct vnode *vp;
2274 	struct iovec aiov;
2275 	struct uio auio;
2276 	int error;
2277 	struct nameidata nd;
2278 
2279 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2280 	    SCARG(uap, path), p);
2281 	if ((error = namei(&nd)) != 0)
2282 		return (error);
2283 	vp = nd.ni_vp;
2284 	if (vp->v_type != VLNK)
2285 		error = EINVAL;
2286 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2287 	    (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2288 		aiov.iov_base = SCARG(uap, buf);
2289 		aiov.iov_len = SCARG(uap, count);
2290 		auio.uio_iov = &aiov;
2291 		auio.uio_iovcnt = 1;
2292 		auio.uio_offset = 0;
2293 		auio.uio_rw = UIO_READ;
2294 		auio.uio_segflg = UIO_USERSPACE;
2295 		auio.uio_procp = p;
2296 		auio.uio_resid = SCARG(uap, count);
2297 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2298 	}
2299 	vput(vp);
2300 	*retval = SCARG(uap, count) - auio.uio_resid;
2301 	return (error);
2302 }
2303 
2304 /*
2305  * Change flags of a file given a path name.
2306  */
2307 /* ARGSUSED */
2308 int
2309 sys_chflags(l, v, retval)
2310 	struct lwp *l;
2311 	void *v;
2312 	register_t *retval;
2313 {
2314 	struct sys_chflags_args /* {
2315 		syscallarg(const char *) path;
2316 		syscallarg(u_long) flags;
2317 	} */ *uap = v;
2318 	struct proc *p = l->l_proc;
2319 	struct vnode *vp;
2320 	int error;
2321 	struct nameidata nd;
2322 
2323 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2324 	if ((error = namei(&nd)) != 0)
2325 		return (error);
2326 	vp = nd.ni_vp;
2327 	error = change_flags(vp, SCARG(uap, flags), p);
2328 	vput(vp);
2329 	return (error);
2330 }
2331 
2332 /*
2333  * Change flags of a file given a file descriptor.
2334  */
2335 /* ARGSUSED */
2336 int
2337 sys_fchflags(l, v, retval)
2338 	struct lwp *l;
2339 	void *v;
2340 	register_t *retval;
2341 {
2342 	struct sys_fchflags_args /* {
2343 		syscallarg(int) fd;
2344 		syscallarg(u_long) flags;
2345 	} */ *uap = v;
2346 	struct proc *p = l->l_proc;
2347 	struct vnode *vp;
2348 	struct file *fp;
2349 	int error;
2350 
2351 	/* getvnode() will use the descriptor for us */
2352 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2353 		return (error);
2354 	vp = (struct vnode *)fp->f_data;
2355 	error = change_flags(vp, SCARG(uap, flags), p);
2356 	VOP_UNLOCK(vp, 0);
2357 	FILE_UNUSE(fp, p);
2358 	return (error);
2359 }
2360 
2361 /*
2362  * Change flags of a file given a path name; this version does
2363  * not follow links.
2364  */
2365 int
2366 sys_lchflags(l, v, retval)
2367 	struct lwp *l;
2368 	void *v;
2369 	register_t *retval;
2370 {
2371 	struct sys_lchflags_args /* {
2372 		syscallarg(const char *) path;
2373 		syscallarg(u_long) flags;
2374 	} */ *uap = v;
2375 	struct proc *p = l->l_proc;
2376 	struct vnode *vp;
2377 	int error;
2378 	struct nameidata nd;
2379 
2380 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2381 	if ((error = namei(&nd)) != 0)
2382 		return (error);
2383 	vp = nd.ni_vp;
2384 	error = change_flags(vp, SCARG(uap, flags), p);
2385 	vput(vp);
2386 	return (error);
2387 }
2388 
2389 /*
2390  * Common routine to change flags of a file.
2391  */
2392 int
2393 change_flags(vp, flags, p)
2394 	struct vnode *vp;
2395 	u_long flags;
2396 	struct proc *p;
2397 {
2398 	struct mount *mp;
2399 	struct vattr vattr;
2400 	int error;
2401 
2402 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2403 		return (error);
2404 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2405 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2406 	/*
2407 	 * Non-superusers cannot change the flags on devices, even if they
2408 	 * own them.
2409 	 */
2410 	if (suser(p->p_ucred, &p->p_acflag) != 0) {
2411 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2412 			goto out;
2413 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2414 			error = EINVAL;
2415 			goto out;
2416 		}
2417 	}
2418 	VATTR_NULL(&vattr);
2419 	vattr.va_flags = flags;
2420 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2421 out:
2422 	vn_finished_write(mp, 0);
2423 	return (error);
2424 }
2425 
2426 /*
2427  * Change mode of a file given path name; this version follows links.
2428  */
2429 /* ARGSUSED */
2430 int
2431 sys_chmod(l, v, retval)
2432 	struct lwp *l;
2433 	void *v;
2434 	register_t *retval;
2435 {
2436 	struct sys_chmod_args /* {
2437 		syscallarg(const char *) path;
2438 		syscallarg(int) mode;
2439 	} */ *uap = v;
2440 	struct proc *p = l->l_proc;
2441 	int error;
2442 	struct nameidata nd;
2443 
2444 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2445 	if ((error = namei(&nd)) != 0)
2446 		return (error);
2447 
2448 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2449 
2450 	vrele(nd.ni_vp);
2451 	return (error);
2452 }
2453 
2454 /*
2455  * Change mode of a file given a file descriptor.
2456  */
2457 /* ARGSUSED */
2458 int
2459 sys_fchmod(l, v, retval)
2460 	struct lwp *l;
2461 	void *v;
2462 	register_t *retval;
2463 {
2464 	struct sys_fchmod_args /* {
2465 		syscallarg(int) fd;
2466 		syscallarg(int) mode;
2467 	} */ *uap = v;
2468 	struct proc *p = l->l_proc;
2469 	struct file *fp;
2470 	int error;
2471 
2472 	/* getvnode() will use the descriptor for us */
2473 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2474 		return (error);
2475 
2476 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2477 	FILE_UNUSE(fp, p);
2478 	return (error);
2479 }
2480 
2481 /*
2482  * Change mode of a file given path name; this version does not follow links.
2483  */
2484 /* ARGSUSED */
2485 int
2486 sys_lchmod(l, v, retval)
2487 	struct lwp *l;
2488 	void *v;
2489 	register_t *retval;
2490 {
2491 	struct sys_lchmod_args /* {
2492 		syscallarg(const char *) path;
2493 		syscallarg(int) mode;
2494 	} */ *uap = v;
2495 	struct proc *p = l->l_proc;
2496 	int error;
2497 	struct nameidata nd;
2498 
2499 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2500 	if ((error = namei(&nd)) != 0)
2501 		return (error);
2502 
2503 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2504 
2505 	vrele(nd.ni_vp);
2506 	return (error);
2507 }
2508 
2509 /*
2510  * Common routine to set mode given a vnode.
2511  */
2512 static int
2513 change_mode(vp, mode, p)
2514 	struct vnode *vp;
2515 	int mode;
2516 	struct proc *p;
2517 {
2518 	struct mount *mp;
2519 	struct vattr vattr;
2520 	int error;
2521 
2522 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2523 		return (error);
2524 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2525 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2526 	VATTR_NULL(&vattr);
2527 	vattr.va_mode = mode & ALLPERMS;
2528 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2529 	VOP_UNLOCK(vp, 0);
2530 	vn_finished_write(mp, 0);
2531 	return (error);
2532 }
2533 
2534 /*
2535  * Set ownership given a path name; this version follows links.
2536  */
2537 /* ARGSUSED */
2538 int
2539 sys_chown(l, v, retval)
2540 	struct lwp *l;
2541 	void *v;
2542 	register_t *retval;
2543 {
2544 	struct sys_chown_args /* {
2545 		syscallarg(const char *) path;
2546 		syscallarg(uid_t) uid;
2547 		syscallarg(gid_t) gid;
2548 	} */ *uap = v;
2549 	struct proc *p = l->l_proc;
2550 	int error;
2551 	struct nameidata nd;
2552 
2553 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2554 	if ((error = namei(&nd)) != 0)
2555 		return (error);
2556 
2557 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2558 
2559 	vrele(nd.ni_vp);
2560 	return (error);
2561 }
2562 
2563 /*
2564  * Set ownership given a path name; this version follows links.
2565  * Provides POSIX semantics.
2566  */
2567 /* ARGSUSED */
2568 int
2569 sys___posix_chown(l, v, retval)
2570 	struct lwp *l;
2571 	void *v;
2572 	register_t *retval;
2573 {
2574 	struct sys_chown_args /* {
2575 		syscallarg(const char *) path;
2576 		syscallarg(uid_t) uid;
2577 		syscallarg(gid_t) gid;
2578 	} */ *uap = v;
2579 	struct proc *p = l->l_proc;
2580 	int error;
2581 	struct nameidata nd;
2582 
2583 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2584 	if ((error = namei(&nd)) != 0)
2585 		return (error);
2586 
2587 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2588 
2589 	vrele(nd.ni_vp);
2590 	return (error);
2591 }
2592 
2593 /*
2594  * Set ownership given a file descriptor.
2595  */
2596 /* ARGSUSED */
2597 int
2598 sys_fchown(l, v, retval)
2599 	struct lwp *l;
2600 	void *v;
2601 	register_t *retval;
2602 {
2603 	struct sys_fchown_args /* {
2604 		syscallarg(int) fd;
2605 		syscallarg(uid_t) uid;
2606 		syscallarg(gid_t) gid;
2607 	} */ *uap = v;
2608 	struct proc *p = l->l_proc;
2609 	int error;
2610 	struct file *fp;
2611 
2612 	/* getvnode() will use the descriptor for us */
2613 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2614 		return (error);
2615 
2616 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2617 	    SCARG(uap, gid), p, 0);
2618 	FILE_UNUSE(fp, p);
2619 	return (error);
2620 }
2621 
2622 /*
2623  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2624  */
2625 /* ARGSUSED */
2626 int
2627 sys___posix_fchown(l, v, retval)
2628 	struct lwp *l;
2629 	void *v;
2630 	register_t *retval;
2631 {
2632 	struct sys_fchown_args /* {
2633 		syscallarg(int) fd;
2634 		syscallarg(uid_t) uid;
2635 		syscallarg(gid_t) gid;
2636 	} */ *uap = v;
2637 	struct proc *p = l->l_proc;
2638 	int error;
2639 	struct file *fp;
2640 
2641 	/* getvnode() will use the descriptor for us */
2642 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2643 		return (error);
2644 
2645 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2646 	    SCARG(uap, gid), p, 1);
2647 	FILE_UNUSE(fp, p);
2648 	return (error);
2649 }
2650 
2651 /*
2652  * Set ownership given a path name; this version does not follow links.
2653  */
2654 /* ARGSUSED */
2655 int
2656 sys_lchown(l, v, retval)
2657 	struct lwp *l;
2658 	void *v;
2659 	register_t *retval;
2660 {
2661 	struct sys_lchown_args /* {
2662 		syscallarg(const char *) path;
2663 		syscallarg(uid_t) uid;
2664 		syscallarg(gid_t) gid;
2665 	} */ *uap = v;
2666 	struct proc *p = l->l_proc;
2667 	int error;
2668 	struct nameidata nd;
2669 
2670 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2671 	if ((error = namei(&nd)) != 0)
2672 		return (error);
2673 
2674 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2675 
2676 	vrele(nd.ni_vp);
2677 	return (error);
2678 }
2679 
2680 /*
2681  * Set ownership given a path name; this version does not follow links.
2682  * Provides POSIX/XPG semantics.
2683  */
2684 /* ARGSUSED */
2685 int
2686 sys___posix_lchown(l, v, retval)
2687 	struct lwp *l;
2688 	void *v;
2689 	register_t *retval;
2690 {
2691 	struct sys_lchown_args /* {
2692 		syscallarg(const char *) path;
2693 		syscallarg(uid_t) uid;
2694 		syscallarg(gid_t) gid;
2695 	} */ *uap = v;
2696 	struct proc *p = l->l_proc;
2697 	int error;
2698 	struct nameidata nd;
2699 
2700 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2701 	if ((error = namei(&nd)) != 0)
2702 		return (error);
2703 
2704 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2705 
2706 	vrele(nd.ni_vp);
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Common routine to set ownership given a vnode.
2712  */
2713 static int
2714 change_owner(vp, uid, gid, p, posix_semantics)
2715 	struct vnode *vp;
2716 	uid_t uid;
2717 	gid_t gid;
2718 	struct proc *p;
2719 	int posix_semantics;
2720 {
2721 	struct mount *mp;
2722 	struct vattr vattr;
2723 	mode_t newmode;
2724 	int error;
2725 
2726 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2727 		return (error);
2728 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2729 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2730 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2731 		goto out;
2732 
2733 #define CHANGED(x) ((int)(x) != -1)
2734 	newmode = vattr.va_mode;
2735 	if (posix_semantics) {
2736 		/*
2737 		 * POSIX/XPG semantics: if the caller is not the super-user,
2738 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2739 		 * the XPG consider the behaviour for calls by the super-user
2740 		 * implementation-defined; we leave the set-user-id and set-
2741 		 * group-id settings intact in that case.
2742 		 */
2743 		if (suser(p->p_ucred, NULL) != 0)
2744 			newmode &= ~(S_ISUID | S_ISGID);
2745 	} else {
2746 		/*
2747 		 * NetBSD semantics: when changing owner and/or group,
2748 		 * clear the respective bit(s).
2749 		 */
2750 		if (CHANGED(uid))
2751 			newmode &= ~S_ISUID;
2752 		if (CHANGED(gid))
2753 			newmode &= ~S_ISGID;
2754 	}
2755 	/* Update va_mode iff altered. */
2756 	if (vattr.va_mode == newmode)
2757 		newmode = VNOVAL;
2758 
2759 	VATTR_NULL(&vattr);
2760 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2761 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2762 	vattr.va_mode = newmode;
2763 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2764 #undef CHANGED
2765 
2766 out:
2767 	VOP_UNLOCK(vp, 0);
2768 	vn_finished_write(mp, 0);
2769 	return (error);
2770 }
2771 
2772 /*
2773  * Set the access and modification times given a path name; this
2774  * version follows links.
2775  */
2776 /* ARGSUSED */
2777 int
2778 sys_utimes(l, v, retval)
2779 	struct lwp *l;
2780 	void *v;
2781 	register_t *retval;
2782 {
2783 	struct sys_utimes_args /* {
2784 		syscallarg(const char *) path;
2785 		syscallarg(const struct timeval *) tptr;
2786 	} */ *uap = v;
2787 	struct proc *p = l->l_proc;
2788 	int error;
2789 	struct nameidata nd;
2790 
2791 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2792 	if ((error = namei(&nd)) != 0)
2793 		return (error);
2794 
2795 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2796 
2797 	vrele(nd.ni_vp);
2798 	return (error);
2799 }
2800 
2801 /*
2802  * Set the access and modification times given a file descriptor.
2803  */
2804 /* ARGSUSED */
2805 int
2806 sys_futimes(l, v, retval)
2807 	struct lwp *l;
2808 	void *v;
2809 	register_t *retval;
2810 {
2811 	struct sys_futimes_args /* {
2812 		syscallarg(int) fd;
2813 		syscallarg(const struct timeval *) tptr;
2814 	} */ *uap = v;
2815 	struct proc *p = l->l_proc;
2816 	int error;
2817 	struct file *fp;
2818 
2819 	/* getvnode() will use the descriptor for us */
2820 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2821 		return (error);
2822 
2823 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2824 	FILE_UNUSE(fp, p);
2825 	return (error);
2826 }
2827 
2828 /*
2829  * Set the access and modification times given a path name; this
2830  * version does not follow links.
2831  */
2832 /* ARGSUSED */
2833 int
2834 sys_lutimes(l, v, retval)
2835 	struct lwp *l;
2836 	void *v;
2837 	register_t *retval;
2838 {
2839 	struct sys_lutimes_args /* {
2840 		syscallarg(const char *) path;
2841 		syscallarg(const struct timeval *) tptr;
2842 	} */ *uap = v;
2843 	struct proc *p = l->l_proc;
2844 	int error;
2845 	struct nameidata nd;
2846 
2847 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2848 	if ((error = namei(&nd)) != 0)
2849 		return (error);
2850 
2851 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2852 
2853 	vrele(nd.ni_vp);
2854 	return (error);
2855 }
2856 
2857 /*
2858  * Common routine to set access and modification times given a vnode.
2859  */
2860 static int
2861 change_utimes(vp, tptr, p)
2862 	struct vnode *vp;
2863 	const struct timeval *tptr;
2864 	struct proc *p;
2865 {
2866 	struct timeval tv[2];
2867 	struct mount *mp;
2868 	struct vattr vattr;
2869 	int error;
2870 
2871 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2872 		return (error);
2873 	VATTR_NULL(&vattr);
2874 	if (tptr == NULL) {
2875 		microtime(&tv[0]);
2876 		tv[1] = tv[0];
2877 		vattr.va_vaflags |= VA_UTIMES_NULL;
2878 	} else {
2879 		error = copyin(tptr, tv, sizeof(tv));
2880 		if (error)
2881 			goto out;
2882 	}
2883 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2884 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2885 	vattr.va_atime.tv_sec = tv[0].tv_sec;
2886 	vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2887 	vattr.va_mtime.tv_sec = tv[1].tv_sec;
2888 	vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2889 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2890 	VOP_UNLOCK(vp, 0);
2891 out:
2892 	vn_finished_write(mp, 0);
2893 	return (error);
2894 }
2895 
2896 /*
2897  * Truncate a file given its path name.
2898  */
2899 /* ARGSUSED */
2900 int
2901 sys_truncate(l, v, retval)
2902 	struct lwp *l;
2903 	void *v;
2904 	register_t *retval;
2905 {
2906 	struct sys_truncate_args /* {
2907 		syscallarg(const char *) path;
2908 		syscallarg(int) pad;
2909 		syscallarg(off_t) length;
2910 	} */ *uap = v;
2911 	struct proc *p = l->l_proc;
2912 	struct vnode *vp;
2913 	struct mount *mp;
2914 	struct vattr vattr;
2915 	int error;
2916 	struct nameidata nd;
2917 
2918 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2919 	if ((error = namei(&nd)) != 0)
2920 		return (error);
2921 	vp = nd.ni_vp;
2922 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2923 		vrele(vp);
2924 		return (error);
2925 	}
2926 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2927 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2928 	if (vp->v_type == VDIR)
2929 		error = EISDIR;
2930 	else if ((error = vn_writechk(vp)) == 0 &&
2931 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2932 		VATTR_NULL(&vattr);
2933 		vattr.va_size = SCARG(uap, length);
2934 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2935 	}
2936 	vput(vp);
2937 	vn_finished_write(mp, 0);
2938 	return (error);
2939 }
2940 
2941 /*
2942  * Truncate a file given a file descriptor.
2943  */
2944 /* ARGSUSED */
2945 int
2946 sys_ftruncate(l, v, retval)
2947 	struct lwp *l;
2948 	void *v;
2949 	register_t *retval;
2950 {
2951 	struct sys_ftruncate_args /* {
2952 		syscallarg(int) fd;
2953 		syscallarg(int) pad;
2954 		syscallarg(off_t) length;
2955 	} */ *uap = v;
2956 	struct proc *p = l->l_proc;
2957 	struct mount *mp;
2958 	struct vattr vattr;
2959 	struct vnode *vp;
2960 	struct file *fp;
2961 	int error;
2962 
2963 	/* getvnode() will use the descriptor for us */
2964 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2965 		return (error);
2966 	if ((fp->f_flag & FWRITE) == 0) {
2967 		error = EINVAL;
2968 		goto out;
2969 	}
2970 	vp = (struct vnode *)fp->f_data;
2971 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2972 		FILE_UNUSE(fp, p);
2973 		return (error);
2974 	}
2975 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2976 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2977 	if (vp->v_type == VDIR)
2978 		error = EISDIR;
2979 	else if ((error = vn_writechk(vp)) == 0) {
2980 		VATTR_NULL(&vattr);
2981 		vattr.va_size = SCARG(uap, length);
2982 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2983 	}
2984 	VOP_UNLOCK(vp, 0);
2985 	vn_finished_write(mp, 0);
2986  out:
2987 	FILE_UNUSE(fp, p);
2988 	return (error);
2989 }
2990 
2991 /*
2992  * Sync an open file.
2993  */
2994 /* ARGSUSED */
2995 int
2996 sys_fsync(l, v, retval)
2997 	struct lwp *l;
2998 	void *v;
2999 	register_t *retval;
3000 {
3001 	struct sys_fsync_args /* {
3002 		syscallarg(int) fd;
3003 	} */ *uap = v;
3004 	struct proc *p = l->l_proc;
3005 	struct vnode *vp;
3006 	struct mount *mp;
3007 	struct file *fp;
3008 	int error;
3009 
3010 	/* getvnode() will use the descriptor for us */
3011 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3012 		return (error);
3013 	vp = (struct vnode *)fp->f_data;
3014 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3015 		FILE_UNUSE(fp, p);
3016 		return (error);
3017 	}
3018 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3019 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
3020 	if (error == 0 && bioops.io_fsync != NULL &&
3021 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3022 		(*bioops.io_fsync)(vp);
3023 	VOP_UNLOCK(vp, 0);
3024 	vn_finished_write(mp, 0);
3025 	FILE_UNUSE(fp, p);
3026 	return (error);
3027 }
3028 
3029 /*
3030  * Sync a range of file data.  API modeled after that found in AIX.
3031  *
3032  * FDATASYNC indicates that we need only save enough metadata to be able
3033  * to re-read the written data.  Note we duplicate AIX's requirement that
3034  * the file be open for writing.
3035  */
3036 /* ARGSUSED */
3037 int
3038 sys_fsync_range(l, v, retval)
3039 	struct lwp *l;
3040 	void *v;
3041 	register_t *retval;
3042 {
3043 	struct sys_fsync_range_args /* {
3044 		syscallarg(int) fd;
3045 		syscallarg(int) flags;
3046 		syscallarg(off_t) start;
3047 		syscallarg(int) length;
3048 	} */ *uap = v;
3049 	struct proc *p = l->l_proc;
3050 	struct vnode *vp;
3051 	struct file *fp;
3052 	int flags, nflags;
3053 	off_t s, e, len;
3054 	int error;
3055 
3056 	/* getvnode() will use the descriptor for us */
3057 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3058 		return (error);
3059 
3060 	if ((fp->f_flag & FWRITE) == 0) {
3061 		FILE_UNUSE(fp, p);
3062 		return (EBADF);
3063 	}
3064 
3065 	flags = SCARG(uap, flags);
3066 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3067 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3068 		return (EINVAL);
3069 	}
3070 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3071 	if (flags & FDATASYNC)
3072 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3073 	else
3074 		nflags = FSYNC_WAIT;
3075 
3076 	len = SCARG(uap, length);
3077 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3078 	if (len) {
3079 		s = SCARG(uap, start);
3080 		e = s + len;
3081 		if (e < s) {
3082 			FILE_UNUSE(fp, p);
3083 			return (EINVAL);
3084 		}
3085 	} else {
3086 		e = 0;
3087 		s = 0;
3088 	}
3089 
3090 	vp = (struct vnode *)fp->f_data;
3091 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3092 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p);
3093 
3094 	if (error == 0 && bioops.io_fsync != NULL &&
3095 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3096 		(*bioops.io_fsync)(vp);
3097 
3098 	VOP_UNLOCK(vp, 0);
3099 	FILE_UNUSE(fp, p);
3100 	return (error);
3101 }
3102 
3103 /*
3104  * Sync the data of an open file.
3105  */
3106 /* ARGSUSED */
3107 int
3108 sys_fdatasync(l, v, retval)
3109 	struct lwp *l;
3110 	void *v;
3111 	register_t *retval;
3112 {
3113 	struct sys_fdatasync_args /* {
3114 		syscallarg(int) fd;
3115 	} */ *uap = v;
3116 	struct proc *p = l->l_proc;
3117 	struct vnode *vp;
3118 	struct file *fp;
3119 	int error;
3120 
3121 	/* getvnode() will use the descriptor for us */
3122 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3123 		return (error);
3124 	if ((fp->f_flag & FWRITE) == 0) {
3125 		FILE_UNUSE(fp, p);
3126 		return (EBADF);
3127 	}
3128 	vp = (struct vnode *)fp->f_data;
3129 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3130 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
3131 	VOP_UNLOCK(vp, 0);
3132 	FILE_UNUSE(fp, p);
3133 	return (error);
3134 }
3135 
3136 /*
3137  * Rename files, (standard) BSD semantics frontend.
3138  */
3139 /* ARGSUSED */
3140 int
3141 sys_rename(l, v, retval)
3142 	struct lwp *l;
3143 	void *v;
3144 	register_t *retval;
3145 {
3146 	struct sys_rename_args /* {
3147 		syscallarg(const char *) from;
3148 		syscallarg(const char *) to;
3149 	} */ *uap = v;
3150 	struct proc *p = l->l_proc;
3151 
3152 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
3153 }
3154 
3155 /*
3156  * Rename files, POSIX semantics frontend.
3157  */
3158 /* ARGSUSED */
3159 int
3160 sys___posix_rename(l, v, retval)
3161 	struct lwp *l;
3162 	void *v;
3163 	register_t *retval;
3164 {
3165 	struct sys___posix_rename_args /* {
3166 		syscallarg(const char *) from;
3167 		syscallarg(const char *) to;
3168 	} */ *uap = v;
3169 	struct proc *p = l->l_proc;
3170 
3171 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
3172 }
3173 
3174 /*
3175  * Rename files.  Source and destination must either both be directories,
3176  * or both not be directories.  If target is a directory, it must be empty.
3177  * If `from' and `to' refer to the same object, the value of the `retain'
3178  * argument is used to determine whether `from' will be
3179  *
3180  * (retain == 0)	deleted unless `from' and `to' refer to the same
3181  *			object in the file system's name space (BSD).
3182  * (retain == 1)	always retained (POSIX).
3183  */
3184 static int
3185 rename_files(from, to, p, retain)
3186 	const char *from, *to;
3187 	struct proc *p;
3188 	int retain;
3189 {
3190 	struct mount *mp = NULL;
3191 	struct vnode *tvp, *fvp, *tdvp;
3192 	struct nameidata fromnd, tond;
3193 	int error;
3194 
3195 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
3196 	    from, p);
3197 	if ((error = namei(&fromnd)) != 0)
3198 		return (error);
3199 	fvp = fromnd.ni_vp;
3200 	error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH);
3201 	if (error != 0) {
3202 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3203 		vrele(fromnd.ni_dvp);
3204 		vrele(fvp);
3205 		if (fromnd.ni_startdir)
3206 			vrele(fromnd.ni_startdir);
3207 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3208 		return (error);
3209 	}
3210 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3211 	    (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p);
3212 	if ((error = namei(&tond)) != 0) {
3213 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3214 		vrele(fromnd.ni_dvp);
3215 		vrele(fvp);
3216 		goto out1;
3217 	}
3218 	tdvp = tond.ni_dvp;
3219 	tvp = tond.ni_vp;
3220 
3221 	if (tvp != NULL) {
3222 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3223 			error = ENOTDIR;
3224 			goto out;
3225 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3226 			error = EISDIR;
3227 			goto out;
3228 		}
3229 	}
3230 
3231 	if (fvp == tdvp)
3232 		error = EINVAL;
3233 
3234 	/*
3235 	 * Source and destination refer to the same object.
3236 	 */
3237 	if (fvp == tvp) {
3238 		if (retain)
3239 			error = -1;
3240 		else if (fromnd.ni_dvp == tdvp &&
3241 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3242 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3243 		          tond.ni_cnd.cn_nameptr,
3244 		          fromnd.ni_cnd.cn_namelen))
3245 		error = -1;
3246 	}
3247 
3248 out:
3249 	if (!error) {
3250 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
3251 		if (fromnd.ni_dvp != tdvp)
3252 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3253 		if (tvp) {
3254 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
3255 		}
3256 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3257 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3258 	} else {
3259 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3260 		if (tdvp == tvp)
3261 			vrele(tdvp);
3262 		else
3263 			vput(tdvp);
3264 		if (tvp)
3265 			vput(tvp);
3266 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3267 		vrele(fromnd.ni_dvp);
3268 		vrele(fvp);
3269 	}
3270 	vrele(tond.ni_startdir);
3271 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3272 out1:
3273 	vn_finished_write(mp, 0);
3274 	if (fromnd.ni_startdir)
3275 		vrele(fromnd.ni_startdir);
3276 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3277 	return (error == -1 ? 0 : error);
3278 }
3279 
3280 /*
3281  * Make a directory file.
3282  */
3283 /* ARGSUSED */
3284 int
3285 sys_mkdir(l, v, retval)
3286 	struct lwp *l;
3287 	void *v;
3288 	register_t *retval;
3289 {
3290 	struct sys_mkdir_args /* {
3291 		syscallarg(const char *) path;
3292 		syscallarg(int) mode;
3293 	} */ *uap = v;
3294 	struct proc *p = l->l_proc;
3295 	struct mount *mp;
3296 	struct vnode *vp;
3297 	struct vattr vattr;
3298 	int error;
3299 	struct nameidata nd;
3300 
3301 restart:
3302 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE,
3303 	    SCARG(uap, path), p);
3304 	if ((error = namei(&nd)) != 0)
3305 		return (error);
3306 	vp = nd.ni_vp;
3307 	if (vp != NULL) {
3308 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3309 		if (nd.ni_dvp == vp)
3310 			vrele(nd.ni_dvp);
3311 		else
3312 			vput(nd.ni_dvp);
3313 		vrele(vp);
3314 		return (EEXIST);
3315 	}
3316 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3317 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3318 		if (nd.ni_dvp == vp)
3319 			vrele(nd.ni_dvp);
3320 		else
3321 			vput(nd.ni_dvp);
3322 		if ((error = vn_start_write(NULL, &mp,
3323 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3324 			return (error);
3325 		goto restart;
3326 	}
3327 	VATTR_NULL(&vattr);
3328 	vattr.va_type = VDIR;
3329 	vattr.va_mode =
3330 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3331 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3332 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3333 	if (!error)
3334 		vput(nd.ni_vp);
3335 	vn_finished_write(mp, 0);
3336 	return (error);
3337 }
3338 
3339 /*
3340  * Remove a directory file.
3341  */
3342 /* ARGSUSED */
3343 int
3344 sys_rmdir(l, v, retval)
3345 	struct lwp *l;
3346 	void *v;
3347 	register_t *retval;
3348 {
3349 	struct sys_rmdir_args /* {
3350 		syscallarg(const char *) path;
3351 	} */ *uap = v;
3352 	struct proc *p = l->l_proc;
3353 	struct mount *mp;
3354 	struct vnode *vp;
3355 	int error;
3356 	struct nameidata nd;
3357 
3358 restart:
3359 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3360 	    SCARG(uap, path), p);
3361 	if ((error = namei(&nd)) != 0)
3362 		return (error);
3363 	vp = nd.ni_vp;
3364 	if (vp->v_type != VDIR) {
3365 		error = ENOTDIR;
3366 		goto out;
3367 	}
3368 	/*
3369 	 * No rmdir "." please.
3370 	 */
3371 	if (nd.ni_dvp == vp) {
3372 		error = EINVAL;
3373 		goto out;
3374 	}
3375 	/*
3376 	 * The root of a mounted filesystem cannot be deleted.
3377 	 */
3378 	if (vp->v_flag & VROOT) {
3379 		error = EBUSY;
3380 		goto out;
3381 	}
3382 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3383 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3384 		if (nd.ni_dvp == vp)
3385 			vrele(nd.ni_dvp);
3386 		else
3387 			vput(nd.ni_dvp);
3388 		vput(vp);
3389 		if ((error = vn_start_write(NULL, &mp,
3390 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3391 			return (error);
3392 		goto restart;
3393 	}
3394 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3395 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3396 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3397 	vn_finished_write(mp, 0);
3398 	return (error);
3399 
3400 out:
3401 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3402 	if (nd.ni_dvp == vp)
3403 		vrele(nd.ni_dvp);
3404 	else
3405 		vput(nd.ni_dvp);
3406 	vput(vp);
3407 	return (error);
3408 }
3409 
3410 /*
3411  * Read a block of directory entries in a file system independent format.
3412  */
3413 int
3414 sys_getdents(l, v, retval)
3415 	struct lwp *l;
3416 	void *v;
3417 	register_t *retval;
3418 {
3419 	struct sys_getdents_args /* {
3420 		syscallarg(int) fd;
3421 		syscallarg(char *) buf;
3422 		syscallarg(size_t) count;
3423 	} */ *uap = v;
3424 	struct proc *p = l->l_proc;
3425 	struct file *fp;
3426 	int error, done;
3427 
3428 	/* getvnode() will use the descriptor for us */
3429 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3430 		return (error);
3431 	if ((fp->f_flag & FREAD) == 0) {
3432 		error = EBADF;
3433 		goto out;
3434 	}
3435 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3436 			SCARG(uap, count), &done, p, 0, 0);
3437 #ifdef KTRACE
3438 	if (!error && KTRPOINT(p, KTR_GENIO)) {
3439 		struct iovec iov;
3440 		iov.iov_base = SCARG(uap, buf);
3441 		iov.iov_len = done;
3442 		ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0);
3443 	}
3444 #endif
3445 	*retval = done;
3446  out:
3447 	FILE_UNUSE(fp, p);
3448 	return (error);
3449 }
3450 
3451 /*
3452  * Set the mode mask for creation of filesystem nodes.
3453  */
3454 int
3455 sys_umask(l, v, retval)
3456 	struct lwp *l;
3457 	void *v;
3458 	register_t *retval;
3459 {
3460 	struct sys_umask_args /* {
3461 		syscallarg(mode_t) newmask;
3462 	} */ *uap = v;
3463 	struct proc *p = l->l_proc;
3464 	struct cwdinfo *cwdi;
3465 
3466 	cwdi = p->p_cwdi;
3467 	*retval = cwdi->cwdi_cmask;
3468 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3469 	return (0);
3470 }
3471 
3472 /*
3473  * Void all references to file by ripping underlying filesystem
3474  * away from vnode.
3475  */
3476 /* ARGSUSED */
3477 int
3478 sys_revoke(l, v, retval)
3479 	struct lwp *l;
3480 	void *v;
3481 	register_t *retval;
3482 {
3483 	struct sys_revoke_args /* {
3484 		syscallarg(const char *) path;
3485 	} */ *uap = v;
3486 	struct proc *p = l->l_proc;
3487 	struct mount *mp;
3488 	struct vnode *vp;
3489 	struct vattr vattr;
3490 	int error;
3491 	struct nameidata nd;
3492 
3493 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3494 	if ((error = namei(&nd)) != 0)
3495 		return (error);
3496 	vp = nd.ni_vp;
3497 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3498 		goto out;
3499 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3500 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3501 		goto out;
3502 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3503 		goto out;
3504 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3505 		VOP_REVOKE(vp, REVOKEALL);
3506 	vn_finished_write(mp, 0);
3507 out:
3508 	vrele(vp);
3509 	return (error);
3510 }
3511 
3512 /*
3513  * Convert a user file descriptor to a kernel file entry.
3514  */
3515 int
3516 getvnode(fdp, fd, fpp)
3517 	struct filedesc *fdp;
3518 	int fd;
3519 	struct file **fpp;
3520 {
3521 	struct vnode *vp;
3522 	struct file *fp;
3523 
3524 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3525 		return (EBADF);
3526 
3527 	FILE_USE(fp);
3528 
3529 	if (fp->f_type != DTYPE_VNODE) {
3530 		FILE_UNUSE(fp, NULL);
3531 		return (EINVAL);
3532 	}
3533 
3534 	vp = (struct vnode *)fp->f_data;
3535 	if (vp->v_type == VBAD) {
3536 		FILE_UNUSE(fp, NULL);
3537 		return (EBADF);
3538 	}
3539 
3540 	*fpp = fp;
3541 	return (0);
3542 }
3543