xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 20e85ad185ab16980f1219a557c42e057edb42ea)
1 /*	$NetBSD: vfs_syscalls.c,v 1.218 2005/04/06 13:49:31 yamt 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 1.218 2005/04/06 13:49:31 yamt 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/extattr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sa.h>
63 #include <sys/syscallargs.h>
64 #ifdef KTRACE
65 #include <sys/ktrace.h>
66 #endif
67 
68 #include <miscfs/genfs/genfs.h>
69 #include <miscfs/syncfs/syncfs.h>
70 
71 #if NFSS > 0
72 #include <dev/fssvar.h>
73 #endif
74 
75 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
76 
77 static int change_dir(struct nameidata *, struct proc *);
78 static int change_flags(struct vnode *, u_long, struct proc *);
79 static int change_mode(struct vnode *, int, struct proc *p);
80 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int);
81 static int change_utimes(struct vnode *vp, const struct timeval *,
82 	       struct proc *p);
83 static int rename_files(const char *, const char *, struct proc *, int);
84 
85 void checkdirs(struct vnode *);
86 
87 int dovfsusermount = 0;
88 
89 /*
90  * Virtual File System System Calls
91  */
92 
93 /*
94  * Mount a file system.
95  */
96 
97 #if defined(COMPAT_09) || defined(COMPAT_43)
98 /*
99  * This table is used to maintain compatibility with 4.3BSD
100  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
101  *
102  * Do not modify this table. It should only contain filesystems
103  * supported by NetBSD 0.9 and 4.3BSD.
104  */
105 const char * const mountcompatnames[] = {
106 	NULL,		/* 0 = MOUNT_NONE */
107 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
108 	MOUNT_NFS,	/* 2 */
109 	MOUNT_MFS,	/* 3 */
110 	MOUNT_MSDOS,	/* 4 */
111 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
112 	MOUNT_FDESC,	/* 6 */
113 	MOUNT_KERNFS,	/* 7 */
114 	NULL,		/* 8 = MOUNT_DEVFS */
115 	MOUNT_AFS,	/* 9 */
116 };
117 const int nmountcompatnames = sizeof(mountcompatnames) /
118     sizeof(mountcompatnames[0]);
119 #endif /* COMPAT_09 || COMPAT_43 */
120 
121 /* ARGSUSED */
122 int
123 sys_mount(l, v, retval)
124 	struct lwp *l;
125 	void *v;
126 	register_t *retval;
127 {
128 	struct sys_mount_args /* {
129 		syscallarg(const char *) type;
130 		syscallarg(const char *) path;
131 		syscallarg(int) flags;
132 		syscallarg(void *) data;
133 	} */ *uap = v;
134 	struct proc *p = l->l_proc;
135 	struct vnode *vp;
136 	struct mount *mp;
137 	int error, flag = 0;
138 	char fstypename[MFSNAMELEN];
139 	struct vattr va;
140 	struct nameidata nd;
141 	struct vfsops *vfs;
142 
143 	/*
144 	 * if MNT_GETARGS is specified, it should be only flag.
145 	 */
146 
147 	if ((SCARG(uap, flags) & MNT_GETARGS) != 0 &&
148 	    (SCARG(uap, flags) & ~MNT_GETARGS) != 0) {
149 		return EINVAL;
150 	}
151 
152 	if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
153 	    (error = suser(p->p_ucred, &p->p_acflag)))
154 		return (error);
155 	/*
156 	 * Get vnode to be covered
157 	 */
158 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
159 	    SCARG(uap, path), p);
160 	if ((error = namei(&nd)) != 0)
161 		return (error);
162 	vp = nd.ni_vp;
163 	/*
164 	 * A lookup in VFS_MOUNT might result in an attempt to
165 	 * lock this vnode again, so make the lock recursive.
166 	 */
167 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
168 	if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
169 		if ((vp->v_flag & VROOT) == 0) {
170 			vput(vp);
171 			return (EINVAL);
172 		}
173 		mp = vp->v_mount;
174 		flag = mp->mnt_flag;
175 		vfs = mp->mnt_op;
176 		/*
177 		 * We only allow the filesystem to be reloaded if it
178 		 * is currently mounted read-only.
179 		 */
180 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
181 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
182 			vput(vp);
183 			return (EOPNOTSUPP);	/* Needs translation */
184 		}
185 		/*
186 		 * In "highly secure" mode, don't let the caller do anything
187 		 * but downgrade a filesystem from read-write to read-only.
188 		 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
189 		 */
190 		if (securelevel >= 2 &&
191 		    SCARG(uap, flags) != MNT_GETARGS &&
192 		    SCARG(uap, flags) !=
193 		    (mp->mnt_flag | MNT_RDONLY |
194 		     MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
195 			vput(vp);
196 			return (EPERM);
197 		}
198 		mp->mnt_flag |= SCARG(uap, flags) &
199 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
200 		/*
201 		 * Only root, or the user that did the original mount is
202 		 * permitted to update it.
203 		 */
204 		if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
205 		    mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
206 		    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
207 			vput(vp);
208 			return (error);
209 		}
210 		/*
211 		 * Do not allow NFS export by non-root users. For non-root
212 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
213 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
214 		 */
215 		if (p->p_ucred->cr_uid != 0) {
216 			if (SCARG(uap, flags) & MNT_EXPORTED) {
217 				vput(vp);
218 				return (EPERM);
219 			}
220 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
221 			if (flag & MNT_NOEXEC)
222 				SCARG(uap, flags) |= MNT_NOEXEC;
223 		}
224 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
225 			vput(vp);
226 			return (EPERM);
227 		}
228 		goto update;
229 	} else {
230 		if (securelevel >= 2) {
231 			vput(vp);
232 			return (EPERM);
233 		}
234 	}
235 	/*
236 	 * If the user is not root, ensure that they own the directory
237 	 * onto which we are attempting to mount.
238 	 */
239 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
240 	    (va.va_uid != p->p_ucred->cr_uid &&
241 		(error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
242 		vput(vp);
243 		return (error);
244 	}
245 	/*
246 	 * Do not allow NFS export by non-root users. For non-root users,
247 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
248 	 * mount point is already MNT_NOEXEC.
249 	 */
250 	if (p->p_ucred->cr_uid != 0) {
251 		if (SCARG(uap, flags) & MNT_EXPORTED) {
252 			vput(vp);
253 			return (EPERM);
254 		}
255 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
256 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
257 			SCARG(uap, flags) |= MNT_NOEXEC;
258 	}
259 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) {
260 		vput(vp);
261 		return (error);
262 	}
263 	if (vp->v_type != VDIR) {
264 		vput(vp);
265 		return (ENOTDIR);
266 	}
267 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
268 	if (error) {
269 #if defined(COMPAT_09) || defined(COMPAT_43)
270 		/*
271 		 * Historically filesystem types were identified by number.
272 		 * If we get an integer for the filesystem type instead of a
273 		 * string, we check to see if it matches one of the historic
274 		 * filesystem types.
275 		 */
276 		u_long fsindex = (u_long)SCARG(uap, type);
277 		if (fsindex >= nmountcompatnames ||
278 		    mountcompatnames[fsindex] == NULL) {
279 			vput(vp);
280 			return (ENODEV);
281 		}
282 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
283 #else
284 		vput(vp);
285 		return (error);
286 #endif
287 	}
288 #ifdef	COMPAT_10
289 	/* Accept `ufs' as an alias for `ffs'. */
290 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
291 		strncpy(fstypename, "ffs", MFSNAMELEN);
292 #endif
293 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
294 		vput(vp);
295 		return (ENODEV);
296 	}
297 	if (vp->v_mountedhere != NULL) {
298 		vput(vp);
299 		return (EBUSY);
300 	}
301 
302 	/*
303 	 * Allocate and initialize the file system.
304 	 */
305 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
306 		M_MOUNT, M_WAITOK);
307 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
308 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
309 	simple_lock_init(&mp->mnt_slock);
310 	(void)vfs_busy(mp, LK_NOWAIT, 0);
311 	mp->mnt_op = vfs;
312 	vfs->vfs_refcount++;
313 	mp->mnt_vnodecovered = vp;
314 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
315 	mp->mnt_unmounter = NULL;
316 	mp->mnt_leaf = mp;
317 
318 	/*
319 	 * The underlying file system may refuse the mount for
320 	 * various reasons.  Allow the user to force it to happen.
321 	 */
322 	mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE;
323  update:
324 	if ((SCARG(uap, flags) & MNT_GETARGS) == 0) {
325 		/*
326 		 * Set the mount level flags.
327 		 */
328 		if (SCARG(uap, flags) & MNT_RDONLY)
329 			mp->mnt_flag |= MNT_RDONLY;
330 		else if (mp->mnt_flag & MNT_RDONLY)
331 			mp->mnt_iflag |= IMNT_WANTRDWR;
332 		mp->mnt_flag &=
333 		  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
334 		    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
335 		    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
336 		mp->mnt_flag |= SCARG(uap, flags) &
337 		   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
338 		    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
339 		    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
340 		    MNT_IGNORE);
341 	}
342 	/*
343 	 * Mount the filesystem.
344 	 */
345 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
346 	if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
347 		if (mp->mnt_iflag & IMNT_WANTRDWR)
348 			mp->mnt_flag &= ~MNT_RDONLY;
349 		if (error)
350 			mp->mnt_flag = flag;
351 		mp->mnt_flag &=~
352 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
353 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
354 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
355 			if (mp->mnt_syncer == NULL)
356 				error = vfs_allocate_syncvnode(mp);
357 		} else {
358 			if (mp->mnt_syncer != NULL)
359 				vfs_deallocate_syncvnode(mp);
360 		}
361 		vfs_unbusy(mp);
362 		VOP_UNLOCK(vp, 0);
363 		vrele(vp);
364 		return (error);
365 	}
366 	/*
367 	 * Put the new filesystem on the mount list after root.
368 	 */
369 	cache_purge(vp);
370 	if (!error) {
371 		mp->mnt_flag &=~
372 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
373 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
374 		vp->v_mountedhere = mp;
375 		simple_lock(&mountlist_slock);
376 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
377 		simple_unlock(&mountlist_slock);
378 		checkdirs(vp);
379 		VOP_UNLOCK(vp, 0);
380 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
381 			error = vfs_allocate_syncvnode(mp);
382 		vfs_unbusy(mp);
383 		(void) VFS_STATVFS(mp, &mp->mnt_stat, p);
384 		if ((error = VFS_START(mp, 0, p)))
385 			vrele(vp);
386 	} else {
387 		vp->v_mountedhere = (struct mount *)0;
388 		vfs->vfs_refcount--;
389 		vfs_unbusy(mp);
390 		free(mp, M_MOUNT);
391 		vput(vp);
392 	}
393 	return (error);
394 }
395 
396 /*
397  * Scan all active processes to see if any of them have a current
398  * or root directory onto which the new filesystem has just been
399  * mounted. If so, replace them with the new mount point.
400  */
401 void
402 checkdirs(olddp)
403 	struct vnode *olddp;
404 {
405 	struct cwdinfo *cwdi;
406 	struct vnode *newdp;
407 	struct proc *p;
408 
409 	if (olddp->v_usecount == 1)
410 		return;
411 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
412 		panic("mount: lost mount");
413 	proclist_lock_read();
414 	PROCLIST_FOREACH(p, &allproc) {
415 		cwdi = p->p_cwdi;
416 		if (!cwdi)
417 			continue;
418 		if (cwdi->cwdi_cdir == olddp) {
419 			vrele(cwdi->cwdi_cdir);
420 			VREF(newdp);
421 			cwdi->cwdi_cdir = newdp;
422 		}
423 		if (cwdi->cwdi_rdir == olddp) {
424 			vrele(cwdi->cwdi_rdir);
425 			VREF(newdp);
426 			cwdi->cwdi_rdir = newdp;
427 		}
428 	}
429 	proclist_unlock_read();
430 	if (rootvnode == olddp) {
431 		vrele(rootvnode);
432 		VREF(newdp);
433 		rootvnode = newdp;
434 	}
435 	vput(newdp);
436 }
437 
438 /*
439  * Unmount a file system.
440  *
441  * Note: unmount takes a path to the vnode mounted on as argument,
442  * not special file (as before).
443  */
444 /* ARGSUSED */
445 int
446 sys_unmount(l, v, retval)
447 	struct lwp *l;
448 	void *v;
449 	register_t *retval;
450 {
451 	struct sys_unmount_args /* {
452 		syscallarg(const char *) path;
453 		syscallarg(int) flags;
454 	} */ *uap = v;
455 	struct proc *p = l->l_proc;
456 	struct vnode *vp;
457 	struct mount *mp;
458 	int error;
459 	struct nameidata nd;
460 
461 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
462 	    SCARG(uap, path), p);
463 	if ((error = namei(&nd)) != 0)
464 		return (error);
465 	vp = nd.ni_vp;
466 	mp = vp->v_mount;
467 
468 	/*
469 	 * Only root, or the user that did the original mount is
470 	 * permitted to unmount this filesystem.
471 	 */
472 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
473 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
474 		vput(vp);
475 		return (error);
476 	}
477 
478 	/*
479 	 * Don't allow unmounting the root file system.
480 	 */
481 	if (mp->mnt_flag & MNT_ROOTFS) {
482 		vput(vp);
483 		return (EINVAL);
484 	}
485 
486 	/*
487 	 * Must be the root of the filesystem
488 	 */
489 	if ((vp->v_flag & VROOT) == 0) {
490 		vput(vp);
491 		return (EINVAL);
492 	}
493 	vput(vp);
494 
495 	/*
496 	 * XXX Freeze syncer.  Must do this before locking the
497 	 * mount point.  See dounmount() for details.
498 	 */
499 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
500 
501 	if (vfs_busy(mp, 0, 0)) {
502 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
503 		return (EBUSY);
504 	}
505 
506 	return (dounmount(mp, SCARG(uap, flags), p));
507 }
508 
509 /*
510  * Do the actual file system unmount. File system is assumed to have been
511  * marked busy by the caller.
512  */
513 int
514 dounmount(mp, flags, p)
515 	struct mount *mp;
516 	int flags;
517 	struct proc *p;
518 {
519 	struct vnode *coveredvp;
520 	int error;
521 	int async;
522 	int used_syncer;
523 
524 	simple_lock(&mountlist_slock);
525 	vfs_unbusy(mp);
526 	used_syncer = (mp->mnt_syncer != NULL);
527 
528 	/*
529 	 * XXX Syncer must be frozen when we get here.  This should really
530 	 * be done on a per-mountpoint basis, but especially the softdep
531 	 * code possibly called from the syncer doens't exactly work on a
532 	 * per-mountpoint basis, so the softdep code would become a maze
533 	 * of vfs_busy() calls.
534 	 *
535 	 * The caller of dounmount() must acquire syncer_lock because
536 	 * the syncer itself acquires locks in syncer_lock -> vfs_busy
537 	 * order, and we must preserve that order to avoid deadlock.
538 	 *
539 	 * So, if the file system did not use the syncer, now is
540 	 * the time to release the syncer_lock.
541 	 */
542 	if (used_syncer == 0)
543 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
544 
545 	mp->mnt_iflag |= IMNT_UNMOUNT;
546 	mp->mnt_unmounter = p;
547 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
548 	vn_start_write(NULL, &mp, V_WAIT);
549 
550 	if (mp->mnt_flag & MNT_EXPUBLIC)
551 		vfs_setpublicfs(NULL, NULL, NULL);
552 	async = mp->mnt_flag & MNT_ASYNC;
553 	mp->mnt_flag &= ~MNT_ASYNC;
554 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
555 	if (mp->mnt_syncer != NULL)
556 		vfs_deallocate_syncvnode(mp);
557 	error = 0;
558 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
559 #if NFSS > 0
560 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
561 #endif
562 		if (error == 0)
563 			error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
564 	}
565 	if (error == 0 || (flags & MNT_FORCE))
566 		error = VFS_UNMOUNT(mp, flags, p);
567 	vn_finished_write(mp, 0);
568 	simple_lock(&mountlist_slock);
569 	if (error) {
570 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
571 			(void) vfs_allocate_syncvnode(mp);
572 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
573 		mp->mnt_unmounter = NULL;
574 		mp->mnt_flag |= async;
575 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
576 		    &mountlist_slock);
577 		if (used_syncer)
578 			lockmgr(&syncer_lock, LK_RELEASE, NULL);
579 		simple_lock(&mp->mnt_slock);
580 		while (mp->mnt_wcnt > 0) {
581 			wakeup(mp);
582 			ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
583 				0, &mp->mnt_slock);
584 		}
585 		simple_unlock(&mp->mnt_slock);
586 		return (error);
587 	}
588 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
589 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
590 		coveredvp->v_mountedhere = NULL;
591 		vrele(coveredvp);
592 	}
593 	mp->mnt_op->vfs_refcount--;
594 	if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
595 		panic("unmount: dangling vnode");
596 	mp->mnt_iflag |= IMNT_GONE;
597 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
598 	if (used_syncer)
599 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
600 	simple_lock(&mp->mnt_slock);
601 	while (mp->mnt_wcnt > 0) {
602 		wakeup(mp);
603 		ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
604 	}
605 	simple_unlock(&mp->mnt_slock);
606 	free(mp, M_MOUNT);
607 	return (0);
608 }
609 
610 /*
611  * Sync each mounted filesystem.
612  */
613 #ifdef DEBUG
614 int syncprt = 0;
615 struct ctldebug debug0 = { "syncprt", &syncprt };
616 #endif
617 
618 /* ARGSUSED */
619 int
620 sys_sync(l, v, retval)
621 	struct lwp *l;
622 	void *v;
623 	register_t *retval;
624 {
625 	struct mount *mp, *nmp;
626 	int asyncflag;
627 	struct proc *p = l == NULL ? &proc0 : l->l_proc;
628 
629 	simple_lock(&mountlist_slock);
630 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
631 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
632 			nmp = mp->mnt_list.cqe_prev;
633 			continue;
634 		}
635 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
636 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
637 			asyncflag = mp->mnt_flag & MNT_ASYNC;
638 			mp->mnt_flag &= ~MNT_ASYNC;
639 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
640 			if (asyncflag)
641 				 mp->mnt_flag |= MNT_ASYNC;
642 			vn_finished_write(mp, 0);
643 		}
644 		simple_lock(&mountlist_slock);
645 		nmp = mp->mnt_list.cqe_prev;
646 		vfs_unbusy(mp);
647 
648 	}
649 	simple_unlock(&mountlist_slock);
650 #ifdef DEBUG
651 	if (syncprt)
652 		vfs_bufstats();
653 #endif /* DEBUG */
654 	return (0);
655 }
656 
657 /*
658  * Change filesystem quotas.
659  */
660 /* ARGSUSED */
661 int
662 sys_quotactl(l, v, retval)
663 	struct lwp *l;
664 	void *v;
665 	register_t *retval;
666 {
667 	struct sys_quotactl_args /* {
668 		syscallarg(const char *) path;
669 		syscallarg(int) cmd;
670 		syscallarg(int) uid;
671 		syscallarg(caddr_t) arg;
672 	} */ *uap = v;
673 	struct proc *p = l->l_proc;
674 	struct mount *mp;
675 	int error;
676 	struct nameidata nd;
677 
678 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
679 	if ((error = namei(&nd)) != 0)
680 		return (error);
681 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
682 	vrele(nd.ni_vp);
683 	if (error)
684 		return (error);
685 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
686 	    SCARG(uap, arg), p);
687 	vn_finished_write(mp, 0);
688 	return (error);
689 }
690 
691 int
692 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags,
693     int root)
694 {
695 	struct cwdinfo *cwdi = p->p_cwdi;
696 	int error = 0;
697 
698 	/*
699 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
700 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
701 	 * overrides MNT_NOWAIT.
702 	 */
703 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
704 	    (flags != MNT_WAIT && flags != 0)) {
705 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
706 		goto done;
707 	}
708 
709 	/* Get the filesystem stats now */
710 	memset(sp, 0, sizeof(*sp));
711 	if ((error = VFS_STATVFS(mp, sp, p)) != 0) {
712 		return error;
713 	}
714 
715 	if (cwdi->cwdi_rdir == NULL)
716 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
717 done:
718 	if (cwdi->cwdi_rdir != NULL) {
719 		size_t len;
720 		char *bp;
721 		char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
722 		if (!path)
723 			return ENOMEM;
724 
725 		bp = path + MAXPATHLEN;
726 		*--bp = '\0';
727 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
728 		    MAXPATHLEN / 2, 0, p);
729 		if (error) {
730 			free(path, M_TEMP);
731 			return error;
732 		}
733 		len = strlen(bp);
734 		/*
735 		 * for mount points that are below our root, we can see
736 		 * them, so we fix up the pathname and return them. The
737 		 * rest we cannot see, so we don't allow viewing the
738 		 * data.
739 		 */
740 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
741 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
742 			    sizeof(sp->f_mntonname));
743 			if (sp->f_mntonname[0] == '\0')
744 				(void)strlcpy(sp->f_mntonname, "/",
745 				    sizeof(sp->f_mntonname));
746 		} else {
747 			if (root)
748 				(void)strlcpy(sp->f_mntonname, "/",
749 				    sizeof(sp->f_mntonname));
750 			else
751 				error = EPERM;
752 		}
753 		free(path, M_TEMP);
754 	}
755 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
756 	return error;
757 }
758 
759 /*
760  * Get filesystem statistics.
761  */
762 /* ARGSUSED */
763 int
764 sys_statvfs1(l, v, retval)
765 	struct lwp *l;
766 	void *v;
767 	register_t *retval;
768 {
769 	struct sys_statvfs1_args /* {
770 		syscallarg(const char *) path;
771 		syscallarg(struct statvfs *) buf;
772 		syscallarg(int) flags;
773 	} */ *uap = v;
774 	struct proc *p = l->l_proc;
775 	struct mount *mp;
776 	struct statvfs sbuf;
777 	int error;
778 	struct nameidata nd;
779 
780 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
781 	if ((error = namei(&nd)) != 0)
782 		return error;
783 	mp = nd.ni_vp->v_mount;
784 	vrele(nd.ni_vp);
785 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
786 		return error;
787 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
788 }
789 
790 /*
791  * Get filesystem statistics.
792  */
793 /* ARGSUSED */
794 int
795 sys_fstatvfs1(l, v, retval)
796 	struct lwp *l;
797 	void *v;
798 	register_t *retval;
799 {
800 	struct sys_fstatvfs1_args /* {
801 		syscallarg(int) fd;
802 		syscallarg(struct statvfs *) buf;
803 		syscallarg(int) flags;
804 	} */ *uap = v;
805 	struct proc *p = l->l_proc;
806 	struct file *fp;
807 	struct mount *mp;
808 	struct statvfs sbuf;
809 	int error;
810 
811 	/* getvnode() will use the descriptor for us */
812 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
813 		return (error);
814 	mp = ((struct vnode *)fp->f_data)->v_mount;
815 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
816 		goto out;
817 	error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
818  out:
819 	FILE_UNUSE(fp, p);
820 	return error;
821 }
822 
823 
824 /*
825  * Get statistics on all filesystems.
826  */
827 int
828 sys_getvfsstat(l, v, retval)
829 	struct lwp *l;
830 	void *v;
831 	register_t *retval;
832 {
833 	struct sys_getvfsstat_args /* {
834 		syscallarg(struct statvfs *) buf;
835 		syscallarg(size_t) bufsize;
836 		syscallarg(int) flags;
837 	} */ *uap = v;
838 	int root = 0;
839 	struct proc *p = l->l_proc;
840 	struct mount *mp, *nmp;
841 	struct statvfs sbuf;
842 	struct statvfs *sfsp;
843 	size_t count, maxcount;
844 	int error = 0;
845 
846 	maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs);
847 	sfsp = SCARG(uap, buf);
848 	simple_lock(&mountlist_slock);
849 	count = 0;
850 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
851 	     mp = nmp) {
852 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
853 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
854 			continue;
855 		}
856 		if (sfsp && count < maxcount) {
857 			error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0);
858 			if (error) {
859 				simple_lock(&mountlist_slock);
860 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
861 				vfs_unbusy(mp);
862 				continue;
863 			}
864 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
865 			if (error) {
866 				vfs_unbusy(mp);
867 				return (error);
868 			}
869 			sfsp++;
870 			root |= strcmp(sbuf.f_mntonname, "/") == 0;
871 		}
872 		count++;
873 		simple_lock(&mountlist_slock);
874 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
875 		vfs_unbusy(mp);
876 	}
877 	simple_unlock(&mountlist_slock);
878 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
879 		/*
880 		 * fake a root entry
881 		 */
882 		if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p,
883 		    SCARG(uap, flags), 1)) != 0)
884 			return error;
885 		if (sfsp)
886 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
887 		count++;
888 	}
889 	if (sfsp && count > maxcount)
890 		*retval = maxcount;
891 	else
892 		*retval = count;
893 	return error;
894 }
895 
896 /*
897  * Change current working directory to a given file descriptor.
898  */
899 /* ARGSUSED */
900 int
901 sys_fchdir(l, v, retval)
902 	struct lwp *l;
903 	void *v;
904 	register_t *retval;
905 {
906 	struct sys_fchdir_args /* {
907 		syscallarg(int) fd;
908 	} */ *uap = v;
909 	struct proc *p = l->l_proc;
910 	struct filedesc *fdp = p->p_fd;
911 	struct cwdinfo *cwdi = p->p_cwdi;
912 	struct vnode *vp, *tdp;
913 	struct mount *mp;
914 	struct file *fp;
915 	int error;
916 
917 	/* getvnode() will use the descriptor for us */
918 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
919 		return (error);
920 	vp = (struct vnode *)fp->f_data;
921 
922 	VREF(vp);
923 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
924 	if (vp->v_type != VDIR)
925 		error = ENOTDIR;
926 	else
927 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
928 	while (!error && (mp = vp->v_mountedhere) != NULL) {
929 		if (vfs_busy(mp, 0, 0))
930 			continue;
931 		error = VFS_ROOT(mp, &tdp);
932 		vfs_unbusy(mp);
933 		if (error)
934 			break;
935 		vput(vp);
936 		vp = tdp;
937 	}
938 	if (error) {
939 		vput(vp);
940 		goto out;
941 	}
942 	VOP_UNLOCK(vp, 0);
943 
944 	/*
945 	 * Disallow changing to a directory not under the process's
946 	 * current root directory (if there is one).
947 	 */
948 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
949 		vrele(vp);
950 		error = EPERM;	/* operation not permitted */
951 		goto out;
952 	}
953 
954 	vrele(cwdi->cwdi_cdir);
955 	cwdi->cwdi_cdir = vp;
956  out:
957 	FILE_UNUSE(fp, p);
958 	return (error);
959 }
960 
961 /*
962  * Change this process's notion of the root directory to a given file descriptor.
963  */
964 
965 int
966 sys_fchroot(l, v, retval)
967 	struct lwp *l;
968 	void *v;
969 	register_t *retval;
970 {
971 	struct sys_fchroot_args *uap = v;
972 	struct proc *p = l->l_proc;
973 	struct filedesc *fdp = p->p_fd;
974 	struct cwdinfo *cwdi = p->p_cwdi;
975 	struct vnode	*vp;
976 	struct file	*fp;
977 	int		 error;
978 
979 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
980 		return error;
981 	/* getvnode() will use the descriptor for us */
982 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
983 		return error;
984 	vp = (struct vnode *) fp->f_data;
985 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
986 	if (vp->v_type != VDIR)
987 		error = ENOTDIR;
988 	else
989 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
990 	VOP_UNLOCK(vp, 0);
991 	if (error)
992 		goto out;
993 	VREF(vp);
994 
995 	/*
996 	 * Prevent escaping from chroot by putting the root under
997 	 * the working directory.  Silently chdir to / if we aren't
998 	 * already there.
999 	 */
1000 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1001 		/*
1002 		 * XXX would be more failsafe to change directory to a
1003 		 * deadfs node here instead
1004 		 */
1005 		vrele(cwdi->cwdi_cdir);
1006 		VREF(vp);
1007 		cwdi->cwdi_cdir = vp;
1008 	}
1009 
1010 	if (cwdi->cwdi_rdir != NULL)
1011 		vrele(cwdi->cwdi_rdir);
1012 	cwdi->cwdi_rdir = vp;
1013  out:
1014 	FILE_UNUSE(fp, p);
1015 	return (error);
1016 }
1017 
1018 
1019 
1020 /*
1021  * Change current working directory (``.'').
1022  */
1023 /* ARGSUSED */
1024 int
1025 sys_chdir(l, v, retval)
1026 	struct lwp *l;
1027 	void *v;
1028 	register_t *retval;
1029 {
1030 	struct sys_chdir_args /* {
1031 		syscallarg(const char *) path;
1032 	} */ *uap = v;
1033 	struct proc *p = l->l_proc;
1034 	struct cwdinfo *cwdi = p->p_cwdi;
1035 	int error;
1036 	struct nameidata nd;
1037 
1038 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1039 	    SCARG(uap, path), p);
1040 	if ((error = change_dir(&nd, p)) != 0)
1041 		return (error);
1042 	vrele(cwdi->cwdi_cdir);
1043 	cwdi->cwdi_cdir = nd.ni_vp;
1044 	return (0);
1045 }
1046 
1047 /*
1048  * Change notion of root (``/'') directory.
1049  */
1050 /* ARGSUSED */
1051 int
1052 sys_chroot(l, v, retval)
1053 	struct lwp *l;
1054 	void *v;
1055 	register_t *retval;
1056 {
1057 	struct sys_chroot_args /* {
1058 		syscallarg(const char *) path;
1059 	} */ *uap = v;
1060 	struct proc *p = l->l_proc;
1061 	struct cwdinfo *cwdi = p->p_cwdi;
1062 	struct vnode *vp;
1063 	int error;
1064 	struct nameidata nd;
1065 
1066 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1067 		return (error);
1068 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1069 	    SCARG(uap, path), p);
1070 	if ((error = change_dir(&nd, p)) != 0)
1071 		return (error);
1072 	if (cwdi->cwdi_rdir != NULL)
1073 		vrele(cwdi->cwdi_rdir);
1074 	vp = nd.ni_vp;
1075 	cwdi->cwdi_rdir = vp;
1076 
1077 	/*
1078 	 * Prevent escaping from chroot by putting the root under
1079 	 * the working directory.  Silently chdir to / if we aren't
1080 	 * already there.
1081 	 */
1082 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1083 		/*
1084 		 * XXX would be more failsafe to change directory to a
1085 		 * deadfs node here instead
1086 		 */
1087 		vrele(cwdi->cwdi_cdir);
1088 		VREF(vp);
1089 		cwdi->cwdi_cdir = vp;
1090 	}
1091 
1092 	return (0);
1093 }
1094 
1095 /*
1096  * Common routine for chroot and chdir.
1097  */
1098 static int
1099 change_dir(ndp, p)
1100 	struct nameidata *ndp;
1101 	struct proc *p;
1102 {
1103 	struct vnode *vp;
1104 	int error;
1105 
1106 	if ((error = namei(ndp)) != 0)
1107 		return (error);
1108 	vp = ndp->ni_vp;
1109 	if (vp->v_type != VDIR)
1110 		error = ENOTDIR;
1111 	else
1112 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
1113 
1114 	if (error)
1115 		vput(vp);
1116 	else
1117 		VOP_UNLOCK(vp, 0);
1118 	return (error);
1119 }
1120 
1121 /*
1122  * Check permissions, allocate an open file structure,
1123  * and call the device open routine if any.
1124  */
1125 int
1126 sys_open(l, v, retval)
1127 	struct lwp *l;
1128 	void *v;
1129 	register_t *retval;
1130 {
1131 	struct sys_open_args /* {
1132 		syscallarg(const char *) path;
1133 		syscallarg(int) flags;
1134 		syscallarg(int) mode;
1135 	} */ *uap = v;
1136 	struct proc *p = l->l_proc;
1137 	struct cwdinfo *cwdi = p->p_cwdi;
1138 	struct filedesc *fdp = p->p_fd;
1139 	struct file *fp;
1140 	struct vnode *vp;
1141 	int flags, cmode;
1142 	int type, indx, error;
1143 	struct flock lf;
1144 	struct nameidata nd;
1145 
1146 	flags = FFLAGS(SCARG(uap, flags));
1147 	if ((flags & (FREAD | FWRITE)) == 0)
1148 		return (EINVAL);
1149 	/* falloc() will use the file descriptor for us */
1150 	if ((error = falloc(p, &fp, &indx)) != 0)
1151 		return (error);
1152 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1153 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1154 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1155 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1156 		FILE_UNUSE(fp, p);
1157 		fdp->fd_ofiles[indx] = NULL;
1158 		ffree(fp);
1159 		if ((error == EDUPFD || error == EMOVEFD) &&
1160 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1161 		    (error =
1162 			dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) {
1163 			*retval = indx;
1164 			return (0);
1165 		}
1166 		if (error == ERESTART)
1167 			error = EINTR;
1168 		fdremove(fdp, indx);
1169 		return (error);
1170 	}
1171 	l->l_dupfd = 0;
1172 	vp = nd.ni_vp;
1173 	fp->f_flag = flags & FMASK;
1174 	fp->f_type = DTYPE_VNODE;
1175 	fp->f_ops = &vnops;
1176 	fp->f_data = vp;
1177 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1178 		lf.l_whence = SEEK_SET;
1179 		lf.l_start = 0;
1180 		lf.l_len = 0;
1181 		if (flags & O_EXLOCK)
1182 			lf.l_type = F_WRLCK;
1183 		else
1184 			lf.l_type = F_RDLCK;
1185 		type = F_FLOCK;
1186 		if ((flags & FNONBLOCK) == 0)
1187 			type |= F_WAIT;
1188 		VOP_UNLOCK(vp, 0);
1189 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1190 		if (error) {
1191 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1192 			FILE_UNUSE(fp, p);
1193 			ffree(fp);
1194 			fdremove(fdp, indx);
1195 			return (error);
1196 		}
1197 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1198 		fp->f_flag |= FHASLOCK;
1199 	}
1200 	VOP_UNLOCK(vp, 0);
1201 	*retval = indx;
1202 	FILE_SET_MATURE(fp);
1203 	FILE_UNUSE(fp, p);
1204 	return (0);
1205 }
1206 
1207 /*
1208  * Get file handle system call
1209  */
1210 int
1211 sys_getfh(l, v, retval)
1212 	struct lwp *l;
1213 	void *v;
1214 	register_t *retval;
1215 {
1216 	struct sys_getfh_args /* {
1217 		syscallarg(char *) fname;
1218 		syscallarg(fhandle_t *) fhp;
1219 	} */ *uap = v;
1220 	struct proc *p = l->l_proc;
1221 	struct vnode *vp;
1222 	fhandle_t fh;
1223 	int error;
1224 	struct nameidata nd;
1225 
1226 	/*
1227 	 * Must be super user
1228 	 */
1229 	error = suser(p->p_ucred, &p->p_acflag);
1230 	if (error)
1231 		return (error);
1232 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1233 	    SCARG(uap, fname), p);
1234 	error = namei(&nd);
1235 	if (error)
1236 		return (error);
1237 	vp = nd.ni_vp;
1238 	memset(&fh, 0, sizeof(fh));
1239 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1240 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1241 	vput(vp);
1242 	if (error)
1243 		return (error);
1244 	error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1245 	return (error);
1246 }
1247 
1248 /*
1249  * Open a file given a file handle.
1250  *
1251  * Check permissions, allocate an open file structure,
1252  * and call the device open routine if any.
1253  */
1254 int
1255 sys_fhopen(l, v, retval)
1256 	struct lwp *l;
1257 	void *v;
1258 	register_t *retval;
1259 {
1260 	struct sys_fhopen_args /* {
1261 		syscallarg(const fhandle_t *) fhp;
1262 		syscallarg(int) flags;
1263 	} */ *uap = v;
1264 	struct proc *p = l->l_proc;
1265 	struct filedesc *fdp = p->p_fd;
1266 	struct file *fp;
1267 	struct vnode *vp = NULL;
1268 	struct mount *mp;
1269 	struct ucred *cred = p->p_ucred;
1270 	int flags;
1271 	struct file *nfp;
1272 	int type, indx, error=0;
1273 	struct flock lf;
1274 	struct vattr va;
1275 	fhandle_t fh;
1276 
1277 	/*
1278 	 * Must be super user
1279 	 */
1280 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1281 		return (error);
1282 
1283 	flags = FFLAGS(SCARG(uap, flags));
1284 	if ((flags & (FREAD | FWRITE)) == 0)
1285 		return (EINVAL);
1286 	if ((flags & O_CREAT))
1287 		return (EINVAL);
1288 	/* falloc() will use the file descriptor for us */
1289 	if ((error = falloc(p, &nfp, &indx)) != 0)
1290 		return (error);
1291 	fp = nfp;
1292 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1293 		goto bad;
1294 
1295 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1296 		error = ESTALE;
1297 		goto bad;
1298 	}
1299 
1300 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1301 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1302 		goto bad;
1303 	}
1304 
1305 	/* Now do an effective vn_open */
1306 
1307 	if (vp->v_type == VSOCK) {
1308 		error = EOPNOTSUPP;
1309 		goto bad;
1310 	}
1311 	if (flags & FREAD) {
1312 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1313 			goto bad;
1314 	}
1315 	if (flags & (FWRITE | O_TRUNC)) {
1316 		if (vp->v_type == VDIR) {
1317 			error = EISDIR;
1318 			goto bad;
1319 		}
1320 		if ((error = vn_writechk(vp)) != 0 ||
1321 		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1322 			goto bad;
1323 	}
1324 	if (flags & O_TRUNC) {
1325 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
1326 			goto bad;
1327 		VOP_UNLOCK(vp, 0);			/* XXX */
1328 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1329 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1330 		VATTR_NULL(&va);
1331 		va.va_size = 0;
1332 		error = VOP_SETATTR(vp, &va, cred, p);
1333 		vn_finished_write(mp, 0);
1334 		if (error)
1335 			goto bad;
1336 	}
1337 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1338 		goto bad;
1339 	if (vp->v_type == VREG &&
1340 	    uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1341 		error = EIO;
1342 		goto bad;
1343 	}
1344 	if (flags & FWRITE)
1345 		vp->v_writecount++;
1346 
1347 	/* done with modified vn_open, now finish what sys_open does. */
1348 
1349 	fp->f_flag = flags & FMASK;
1350 	fp->f_type = DTYPE_VNODE;
1351 	fp->f_ops = &vnops;
1352 	fp->f_data = vp;
1353 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1354 		lf.l_whence = SEEK_SET;
1355 		lf.l_start = 0;
1356 		lf.l_len = 0;
1357 		if (flags & O_EXLOCK)
1358 			lf.l_type = F_WRLCK;
1359 		else
1360 			lf.l_type = F_RDLCK;
1361 		type = F_FLOCK;
1362 		if ((flags & FNONBLOCK) == 0)
1363 			type |= F_WAIT;
1364 		VOP_UNLOCK(vp, 0);
1365 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1366 		if (error) {
1367 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1368 			FILE_UNUSE(fp, p);
1369 			ffree(fp);
1370 			fdremove(fdp, indx);
1371 			return (error);
1372 		}
1373 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1374 		fp->f_flag |= FHASLOCK;
1375 	}
1376 	VOP_UNLOCK(vp, 0);
1377 	*retval = indx;
1378 	FILE_SET_MATURE(fp);
1379 	FILE_UNUSE(fp, p);
1380 	return (0);
1381 
1382 bad:
1383 	FILE_UNUSE(fp, p);
1384 	ffree(fp);
1385 	fdremove(fdp, indx);
1386 	if (vp != NULL)
1387 		vput(vp);
1388 	return (error);
1389 }
1390 
1391 /* ARGSUSED */
1392 int
1393 sys_fhstat(l, v, retval)
1394 	struct lwp *l;
1395 	void *v;
1396 	register_t *retval;
1397 {
1398 	struct sys_fhstat_args /* {
1399 		syscallarg(const fhandle_t *) fhp;
1400 		syscallarg(struct stat *) sb;
1401 	} */ *uap = v;
1402 	struct proc *p = l->l_proc;
1403 	struct stat sb;
1404 	int error;
1405 	fhandle_t fh;
1406 	struct mount *mp;
1407 	struct vnode *vp;
1408 
1409 	/*
1410 	 * Must be super user
1411 	 */
1412 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1413 		return (error);
1414 
1415 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1416 		return (error);
1417 
1418 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1419 		return (ESTALE);
1420 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1421 		return (error);
1422 	error = vn_stat(vp, &sb, p);
1423 	vput(vp);
1424 	if (error)
1425 		return (error);
1426 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1427 	return (error);
1428 }
1429 
1430 /* ARGSUSED */
1431 int
1432 sys_fhstatvfs1(l, v, retval)
1433 	struct lwp *l;
1434 	void *v;
1435 	register_t *retval;
1436 {
1437 	struct sys_fhstatvfs1_args /*
1438 		syscallarg(const fhandle_t *) fhp;
1439 		syscallarg(struct statvfs *) buf;
1440 		syscallarg(int)	flags;
1441 	} */ *uap = v;
1442 	struct proc *p = l->l_proc;
1443 	struct statvfs sbuf;
1444 	fhandle_t fh;
1445 	struct mount *mp;
1446 	struct vnode *vp;
1447 	int error;
1448 
1449 	/*
1450 	 * Must be super user
1451 	 */
1452 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1453 		return error;
1454 
1455 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1456 		return error;
1457 
1458 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1459 		return ESTALE;
1460 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1461 		return error;
1462 
1463 	mp = vp->v_mount;
1464 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) {
1465 		vput(vp);
1466 		return error;
1467 	}
1468 	vput(vp);
1469 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
1470 }
1471 
1472 /*
1473  * Create a special file.
1474  */
1475 /* ARGSUSED */
1476 int
1477 sys_mknod(l, v, retval)
1478 	struct lwp *l;
1479 	void *v;
1480 	register_t *retval;
1481 {
1482 	struct sys_mknod_args /* {
1483 		syscallarg(const char *) path;
1484 		syscallarg(int) mode;
1485 		syscallarg(int) dev;
1486 	} */ *uap = v;
1487 	struct proc *p = l->l_proc;
1488 	struct vnode *vp;
1489 	struct mount *mp;
1490 	struct vattr vattr;
1491 	int error;
1492 	int whiteout = 0;
1493 	struct nameidata nd;
1494 
1495 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1496 		return (error);
1497 restart:
1498 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1499 	if ((error = namei(&nd)) != 0)
1500 		return (error);
1501 	vp = nd.ni_vp;
1502 	if (vp != NULL)
1503 		error = EEXIST;
1504 	else {
1505 		VATTR_NULL(&vattr);
1506 		vattr.va_mode =
1507 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1508 		vattr.va_rdev = SCARG(uap, dev);
1509 		whiteout = 0;
1510 
1511 		switch (SCARG(uap, mode) & S_IFMT) {
1512 		case S_IFMT:	/* used by badsect to flag bad sectors */
1513 			vattr.va_type = VBAD;
1514 			break;
1515 		case S_IFCHR:
1516 			vattr.va_type = VCHR;
1517 			break;
1518 		case S_IFBLK:
1519 			vattr.va_type = VBLK;
1520 			break;
1521 		case S_IFWHT:
1522 			whiteout = 1;
1523 			break;
1524 		default:
1525 			error = EINVAL;
1526 			break;
1527 		}
1528 	}
1529 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1530 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1531 		if (nd.ni_dvp == vp)
1532 			vrele(nd.ni_dvp);
1533 		else
1534 			vput(nd.ni_dvp);
1535 		if (vp)
1536 			vrele(vp);
1537 		if ((error = vn_start_write(NULL, &mp,
1538 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1539 			return (error);
1540 		goto restart;
1541 	}
1542 	if (!error) {
1543 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1544 		if (whiteout) {
1545 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1546 			if (error)
1547 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1548 			vput(nd.ni_dvp);
1549 		} else {
1550 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1551 						&nd.ni_cnd, &vattr);
1552 			if (error == 0)
1553 				vput(nd.ni_vp);
1554 		}
1555 	} else {
1556 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1557 		if (nd.ni_dvp == vp)
1558 			vrele(nd.ni_dvp);
1559 		else
1560 			vput(nd.ni_dvp);
1561 		if (vp)
1562 			vrele(vp);
1563 	}
1564 	vn_finished_write(mp, 0);
1565 	return (error);
1566 }
1567 
1568 /*
1569  * Create a named pipe.
1570  */
1571 /* ARGSUSED */
1572 int
1573 sys_mkfifo(l, v, retval)
1574 	struct lwp *l;
1575 	void *v;
1576 	register_t *retval;
1577 {
1578 	struct sys_mkfifo_args /* {
1579 		syscallarg(const char *) path;
1580 		syscallarg(int) mode;
1581 	} */ *uap = v;
1582 	struct proc *p = l->l_proc;
1583 	struct mount *mp;
1584 	struct vattr vattr;
1585 	int error;
1586 	struct nameidata nd;
1587 
1588 restart:
1589 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1590 	if ((error = namei(&nd)) != 0)
1591 		return (error);
1592 	if (nd.ni_vp != NULL) {
1593 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1594 		if (nd.ni_dvp == nd.ni_vp)
1595 			vrele(nd.ni_dvp);
1596 		else
1597 			vput(nd.ni_dvp);
1598 		vrele(nd.ni_vp);
1599 		return (EEXIST);
1600 	}
1601 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1602 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1603 		if (nd.ni_dvp == nd.ni_vp)
1604 			vrele(nd.ni_dvp);
1605 		else
1606 			vput(nd.ni_dvp);
1607 		if (nd.ni_vp)
1608 			vrele(nd.ni_vp);
1609 		if ((error = vn_start_write(NULL, &mp,
1610 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1611 			return (error);
1612 		goto restart;
1613 	}
1614 	VATTR_NULL(&vattr);
1615 	vattr.va_type = VFIFO;
1616 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1617 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1618 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1619 	if (error == 0)
1620 		vput(nd.ni_vp);
1621 	vn_finished_write(mp, 0);
1622 	return (error);
1623 }
1624 
1625 /*
1626  * Make a hard file link.
1627  */
1628 /* ARGSUSED */
1629 int
1630 sys_link(l, v, retval)
1631 	struct lwp *l;
1632 	void *v;
1633 	register_t *retval;
1634 {
1635 	struct sys_link_args /* {
1636 		syscallarg(const char *) path;
1637 		syscallarg(const char *) link;
1638 	} */ *uap = v;
1639 	struct proc *p = l->l_proc;
1640 	struct vnode *vp;
1641 	struct mount *mp;
1642 	struct nameidata nd;
1643 	int error;
1644 
1645 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1646 	if ((error = namei(&nd)) != 0)
1647 		return (error);
1648 	vp = nd.ni_vp;
1649 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
1650 		vrele(vp);
1651 		return (error);
1652 	}
1653 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1654 	if ((error = namei(&nd)) != 0)
1655 		goto out;
1656 	if (nd.ni_vp) {
1657 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1658 		if (nd.ni_dvp == nd.ni_vp)
1659 			vrele(nd.ni_dvp);
1660 		else
1661 			vput(nd.ni_dvp);
1662 		vrele(nd.ni_vp);
1663 		error = EEXIST;
1664 		goto out;
1665 	}
1666 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1667 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1668 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1669 out:
1670 	vrele(vp);
1671 	vn_finished_write(mp, 0);
1672 	return (error);
1673 }
1674 
1675 /*
1676  * Make a symbolic link.
1677  */
1678 /* ARGSUSED */
1679 int
1680 sys_symlink(l, v, retval)
1681 	struct lwp *l;
1682 	void *v;
1683 	register_t *retval;
1684 {
1685 	struct sys_symlink_args /* {
1686 		syscallarg(const char *) path;
1687 		syscallarg(const char *) link;
1688 	} */ *uap = v;
1689 	struct proc *p = l->l_proc;
1690 	struct mount *mp;
1691 	struct vattr vattr;
1692 	char *path;
1693 	int error;
1694 	struct nameidata nd;
1695 
1696 	path = PNBUF_GET();
1697 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1698 	if (error)
1699 		goto out;
1700 restart:
1701 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1702 	if ((error = namei(&nd)) != 0)
1703 		goto out;
1704 	if (nd.ni_vp) {
1705 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1706 		if (nd.ni_dvp == nd.ni_vp)
1707 			vrele(nd.ni_dvp);
1708 		else
1709 			vput(nd.ni_dvp);
1710 		vrele(nd.ni_vp);
1711 		error = EEXIST;
1712 		goto out;
1713 	}
1714 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1715 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1716 		if (nd.ni_dvp == nd.ni_vp)
1717 			vrele(nd.ni_dvp);
1718 		else
1719 			vput(nd.ni_dvp);
1720 		if ((error = vn_start_write(NULL, &mp,
1721 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1722 			return (error);
1723 		goto restart;
1724 	}
1725 	VATTR_NULL(&vattr);
1726 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1727 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1728 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1729 	if (error == 0)
1730 		vput(nd.ni_vp);
1731 	vn_finished_write(mp, 0);
1732 out:
1733 	PNBUF_PUT(path);
1734 	return (error);
1735 }
1736 
1737 /*
1738  * Delete a whiteout from the filesystem.
1739  */
1740 /* ARGSUSED */
1741 int
1742 sys_undelete(l, v, retval)
1743 	struct lwp *l;
1744 	void *v;
1745 	register_t *retval;
1746 {
1747 	struct sys_undelete_args /* {
1748 		syscallarg(const char *) path;
1749 	} */ *uap = v;
1750 	struct proc *p = l->l_proc;
1751 	int error;
1752 	struct mount *mp;
1753 	struct nameidata nd;
1754 
1755 restart:
1756 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1757 	    SCARG(uap, path), p);
1758 	error = namei(&nd);
1759 	if (error)
1760 		return (error);
1761 
1762 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1763 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1764 		if (nd.ni_dvp == nd.ni_vp)
1765 			vrele(nd.ni_dvp);
1766 		else
1767 			vput(nd.ni_dvp);
1768 		if (nd.ni_vp)
1769 			vrele(nd.ni_vp);
1770 		return (EEXIST);
1771 	}
1772 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1773 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1774 		if (nd.ni_dvp == nd.ni_vp)
1775 			vrele(nd.ni_dvp);
1776 		else
1777 			vput(nd.ni_dvp);
1778 		if ((error = vn_start_write(NULL, &mp,
1779 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1780 			return (error);
1781 		goto restart;
1782 	}
1783 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1784 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1785 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1786 	vput(nd.ni_dvp);
1787 	vn_finished_write(mp, 0);
1788 	return (error);
1789 }
1790 
1791 /*
1792  * Delete a name from the filesystem.
1793  */
1794 /* ARGSUSED */
1795 int
1796 sys_unlink(l, v, retval)
1797 	struct lwp *l;
1798 	void *v;
1799 	register_t *retval;
1800 {
1801 	struct sys_unlink_args /* {
1802 		syscallarg(const char *) path;
1803 	} */ *uap = v;
1804 	struct proc *p = l->l_proc;
1805 	struct mount *mp;
1806 	struct vnode *vp;
1807 	int error;
1808 	struct nameidata nd;
1809 
1810 restart:
1811 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1812 	    SCARG(uap, path), p);
1813 	if ((error = namei(&nd)) != 0)
1814 		return (error);
1815 	vp = nd.ni_vp;
1816 
1817 	/*
1818 	 * The root of a mounted filesystem cannot be deleted.
1819 	 */
1820 	if (vp->v_flag & VROOT) {
1821 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1822 		if (nd.ni_dvp == vp)
1823 			vrele(nd.ni_dvp);
1824 		else
1825 			vput(nd.ni_dvp);
1826 		vput(vp);
1827 		error = EBUSY;
1828 		goto out;
1829 	}
1830 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1831 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1832 		if (nd.ni_dvp == vp)
1833 			vrele(nd.ni_dvp);
1834 		else
1835 			vput(nd.ni_dvp);
1836 		vput(vp);
1837 		if ((error = vn_start_write(NULL, &mp,
1838 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1839 			return (error);
1840 		goto restart;
1841 	}
1842 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1843 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1844 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1845 	vn_finished_write(mp, 0);
1846 out:
1847 	return (error);
1848 }
1849 
1850 /*
1851  * Reposition read/write file offset.
1852  */
1853 int
1854 sys_lseek(l, v, retval)
1855 	struct lwp *l;
1856 	void *v;
1857 	register_t *retval;
1858 {
1859 	struct sys_lseek_args /* {
1860 		syscallarg(int) fd;
1861 		syscallarg(int) pad;
1862 		syscallarg(off_t) offset;
1863 		syscallarg(int) whence;
1864 	} */ *uap = v;
1865 	struct proc *p = l->l_proc;
1866 	struct ucred *cred = p->p_ucred;
1867 	struct filedesc *fdp = p->p_fd;
1868 	struct file *fp;
1869 	struct vnode *vp;
1870 	struct vattr vattr;
1871 	off_t newoff;
1872 	int error;
1873 
1874 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1875 		return (EBADF);
1876 
1877 	FILE_USE(fp);
1878 
1879 	vp = (struct vnode *)fp->f_data;
1880 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1881 		error = ESPIPE;
1882 		goto out;
1883 	}
1884 
1885 	switch (SCARG(uap, whence)) {
1886 	case SEEK_CUR:
1887 		newoff = fp->f_offset + SCARG(uap, offset);
1888 		break;
1889 	case SEEK_END:
1890 		error = VOP_GETATTR(vp, &vattr, cred, p);
1891 		if (error)
1892 			goto out;
1893 		newoff = SCARG(uap, offset) + vattr.va_size;
1894 		break;
1895 	case SEEK_SET:
1896 		newoff = SCARG(uap, offset);
1897 		break;
1898 	default:
1899 		error = EINVAL;
1900 		goto out;
1901 	}
1902 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1903 		goto out;
1904 
1905 	*(off_t *)retval = fp->f_offset = newoff;
1906  out:
1907 	FILE_UNUSE(fp, p);
1908 	return (error);
1909 }
1910 
1911 /*
1912  * Positional read system call.
1913  */
1914 int
1915 sys_pread(l, v, retval)
1916 	struct lwp *l;
1917 	void *v;
1918 	register_t *retval;
1919 {
1920 	struct sys_pread_args /* {
1921 		syscallarg(int) fd;
1922 		syscallarg(void *) buf;
1923 		syscallarg(size_t) nbyte;
1924 		syscallarg(off_t) offset;
1925 	} */ *uap = v;
1926 	struct proc *p = l->l_proc;
1927 	struct filedesc *fdp = p->p_fd;
1928 	struct file *fp;
1929 	struct vnode *vp;
1930 	off_t offset;
1931 	int error, fd = SCARG(uap, fd);
1932 
1933 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1934 		return (EBADF);
1935 
1936 	if ((fp->f_flag & FREAD) == 0) {
1937 		simple_unlock(&fp->f_slock);
1938 		return (EBADF);
1939 	}
1940 
1941 	FILE_USE(fp);
1942 
1943 	vp = (struct vnode *)fp->f_data;
1944 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1945 		error = ESPIPE;
1946 		goto out;
1947 	}
1948 
1949 	offset = SCARG(uap, offset);
1950 
1951 	/*
1952 	 * XXX This works because no file systems actually
1953 	 * XXX take any action on the seek operation.
1954 	 */
1955 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1956 		goto out;
1957 
1958 	/* dofileread() will unuse the descriptor for us */
1959 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1960 	    &offset, 0, retval));
1961 
1962  out:
1963 	FILE_UNUSE(fp, p);
1964 	return (error);
1965 }
1966 
1967 /*
1968  * Positional scatter read system call.
1969  */
1970 int
1971 sys_preadv(l, v, retval)
1972 	struct lwp *l;
1973 	void *v;
1974 	register_t *retval;
1975 {
1976 	struct sys_preadv_args /* {
1977 		syscallarg(int) fd;
1978 		syscallarg(const struct iovec *) iovp;
1979 		syscallarg(int) iovcnt;
1980 		syscallarg(off_t) offset;
1981 	} */ *uap = v;
1982 	struct proc *p = l->l_proc;
1983 	struct filedesc *fdp = p->p_fd;
1984 	struct file *fp;
1985 	struct vnode *vp;
1986 	off_t offset;
1987 	int error, fd = SCARG(uap, fd);
1988 
1989 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1990 		return (EBADF);
1991 
1992 	if ((fp->f_flag & FREAD) == 0) {
1993 		simple_unlock(&fp->f_slock);
1994 		return (EBADF);
1995 	}
1996 
1997 	FILE_USE(fp);
1998 
1999 	vp = (struct vnode *)fp->f_data;
2000 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2001 		error = ESPIPE;
2002 		goto out;
2003 	}
2004 
2005 	offset = SCARG(uap, offset);
2006 
2007 	/*
2008 	 * XXX This works because no file systems actually
2009 	 * XXX take any action on the seek operation.
2010 	 */
2011 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2012 		goto out;
2013 
2014 	/* dofilereadv() will unuse the descriptor for us */
2015 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2016 	    &offset, 0, retval));
2017 
2018  out:
2019 	FILE_UNUSE(fp, p);
2020 	return (error);
2021 }
2022 
2023 /*
2024  * Positional write system call.
2025  */
2026 int
2027 sys_pwrite(l, v, retval)
2028 	struct lwp *l;
2029 	void *v;
2030 	register_t *retval;
2031 {
2032 	struct sys_pwrite_args /* {
2033 		syscallarg(int) fd;
2034 		syscallarg(const void *) buf;
2035 		syscallarg(size_t) nbyte;
2036 		syscallarg(off_t) offset;
2037 	} */ *uap = v;
2038 	struct proc *p = l->l_proc;
2039 	struct filedesc *fdp = p->p_fd;
2040 	struct file *fp;
2041 	struct vnode *vp;
2042 	off_t offset;
2043 	int error, fd = SCARG(uap, fd);
2044 
2045 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2046 		return (EBADF);
2047 
2048 	if ((fp->f_flag & FWRITE) == 0) {
2049 		simple_unlock(&fp->f_slock);
2050 		return (EBADF);
2051 	}
2052 
2053 	FILE_USE(fp);
2054 
2055 	vp = (struct vnode *)fp->f_data;
2056 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2057 		error = ESPIPE;
2058 		goto out;
2059 	}
2060 
2061 	offset = SCARG(uap, offset);
2062 
2063 	/*
2064 	 * XXX This works because no file systems actually
2065 	 * XXX take any action on the seek operation.
2066 	 */
2067 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2068 		goto out;
2069 
2070 	/* dofilewrite() will unuse the descriptor for us */
2071 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2072 	    &offset, 0, retval));
2073 
2074  out:
2075 	FILE_UNUSE(fp, p);
2076 	return (error);
2077 }
2078 
2079 /*
2080  * Positional gather write system call.
2081  */
2082 int
2083 sys_pwritev(l, v, retval)
2084 	struct lwp *l;
2085 	void *v;
2086 	register_t *retval;
2087 {
2088 	struct sys_pwritev_args /* {
2089 		syscallarg(int) fd;
2090 		syscallarg(const struct iovec *) iovp;
2091 		syscallarg(int) iovcnt;
2092 		syscallarg(off_t) offset;
2093 	} */ *uap = v;
2094 	struct proc *p = l->l_proc;
2095 	struct filedesc *fdp = p->p_fd;
2096 	struct file *fp;
2097 	struct vnode *vp;
2098 	off_t offset;
2099 	int error, fd = SCARG(uap, fd);
2100 
2101 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2102 		return (EBADF);
2103 
2104 	if ((fp->f_flag & FWRITE) == 0) {
2105 		simple_unlock(&fp->f_slock);
2106 		return (EBADF);
2107 	}
2108 
2109 	FILE_USE(fp);
2110 
2111 	vp = (struct vnode *)fp->f_data;
2112 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2113 		error = ESPIPE;
2114 		goto out;
2115 	}
2116 
2117 	offset = SCARG(uap, offset);
2118 
2119 	/*
2120 	 * XXX This works because no file systems actually
2121 	 * XXX take any action on the seek operation.
2122 	 */
2123 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2124 		goto out;
2125 
2126 	/* dofilewritev() will unuse the descriptor for us */
2127 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2128 	    &offset, 0, retval));
2129 
2130  out:
2131 	FILE_UNUSE(fp, p);
2132 	return (error);
2133 }
2134 
2135 /*
2136  * Check access permissions.
2137  */
2138 int
2139 sys_access(l, v, retval)
2140 	struct lwp *l;
2141 	void *v;
2142 	register_t *retval;
2143 {
2144 	struct sys_access_args /* {
2145 		syscallarg(const char *) path;
2146 		syscallarg(int) flags;
2147 	} */ *uap = v;
2148 	struct proc *p = l->l_proc;
2149 	struct ucred *cred;
2150 	struct vnode *vp;
2151 	int error, flags;
2152 	struct nameidata nd;
2153 
2154 	cred = crdup(p->p_ucred);
2155 	cred->cr_uid = p->p_cred->p_ruid;
2156 	cred->cr_gid = p->p_cred->p_rgid;
2157 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2158 	    SCARG(uap, path), p);
2159 	/* Override default credentials */
2160 	nd.ni_cnd.cn_cred = cred;
2161 	if ((error = namei(&nd)) != 0)
2162 		goto out;
2163 	vp = nd.ni_vp;
2164 
2165 	/* Flags == 0 means only check for existence. */
2166 	if (SCARG(uap, flags)) {
2167 		flags = 0;
2168 		if (SCARG(uap, flags) & R_OK)
2169 			flags |= VREAD;
2170 		if (SCARG(uap, flags) & W_OK)
2171 			flags |= VWRITE;
2172 		if (SCARG(uap, flags) & X_OK)
2173 			flags |= VEXEC;
2174 
2175 		error = VOP_ACCESS(vp, flags, cred, p);
2176 		if (!error && (flags & VWRITE))
2177 			error = vn_writechk(vp);
2178 	}
2179 	vput(vp);
2180 out:
2181 	crfree(cred);
2182 	return (error);
2183 }
2184 
2185 /*
2186  * Get file status; this version follows links.
2187  */
2188 /* ARGSUSED */
2189 int
2190 sys___stat13(l, v, retval)
2191 	struct lwp *l;
2192 	void *v;
2193 	register_t *retval;
2194 {
2195 	struct sys___stat13_args /* {
2196 		syscallarg(const char *) path;
2197 		syscallarg(struct stat *) ub;
2198 	} */ *uap = v;
2199 	struct proc *p = l->l_proc;
2200 	struct stat sb;
2201 	int error;
2202 	struct nameidata nd;
2203 
2204 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2205 	    SCARG(uap, path), p);
2206 	if ((error = namei(&nd)) != 0)
2207 		return (error);
2208 	error = vn_stat(nd.ni_vp, &sb, p);
2209 	vput(nd.ni_vp);
2210 	if (error)
2211 		return (error);
2212 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2213 	return (error);
2214 }
2215 
2216 /*
2217  * Get file status; this version does not follow links.
2218  */
2219 /* ARGSUSED */
2220 int
2221 sys___lstat13(l, v, retval)
2222 	struct lwp *l;
2223 	void *v;
2224 	register_t *retval;
2225 {
2226 	struct sys___lstat13_args /* {
2227 		syscallarg(const char *) path;
2228 		syscallarg(struct stat *) ub;
2229 	} */ *uap = v;
2230 	struct proc *p = l->l_proc;
2231 	struct stat sb;
2232 	int error;
2233 	struct nameidata nd;
2234 
2235 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2236 	    SCARG(uap, path), p);
2237 	if ((error = namei(&nd)) != 0)
2238 		return (error);
2239 	error = vn_stat(nd.ni_vp, &sb, p);
2240 	vput(nd.ni_vp);
2241 	if (error)
2242 		return (error);
2243 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2244 	return (error);
2245 }
2246 
2247 /*
2248  * Get configurable pathname variables.
2249  */
2250 /* ARGSUSED */
2251 int
2252 sys_pathconf(l, v, retval)
2253 	struct lwp *l;
2254 	void *v;
2255 	register_t *retval;
2256 {
2257 	struct sys_pathconf_args /* {
2258 		syscallarg(const char *) path;
2259 		syscallarg(int) name;
2260 	} */ *uap = v;
2261 	struct proc *p = l->l_proc;
2262 	int error;
2263 	struct nameidata nd;
2264 
2265 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2266 	    SCARG(uap, path), p);
2267 	if ((error = namei(&nd)) != 0)
2268 		return (error);
2269 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2270 	vput(nd.ni_vp);
2271 	return (error);
2272 }
2273 
2274 /*
2275  * Return target name of a symbolic link.
2276  */
2277 /* ARGSUSED */
2278 int
2279 sys_readlink(l, v, retval)
2280 	struct lwp *l;
2281 	void *v;
2282 	register_t *retval;
2283 {
2284 	struct sys_readlink_args /* {
2285 		syscallarg(const char *) path;
2286 		syscallarg(char *) buf;
2287 		syscallarg(size_t) count;
2288 	} */ *uap = v;
2289 	struct proc *p = l->l_proc;
2290 	struct vnode *vp;
2291 	struct iovec aiov;
2292 	struct uio auio;
2293 	int error;
2294 	struct nameidata nd;
2295 
2296 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2297 	    SCARG(uap, path), p);
2298 	if ((error = namei(&nd)) != 0)
2299 		return (error);
2300 	vp = nd.ni_vp;
2301 	if (vp->v_type != VLNK)
2302 		error = EINVAL;
2303 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2304 	    (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2305 		aiov.iov_base = SCARG(uap, buf);
2306 		aiov.iov_len = SCARG(uap, count);
2307 		auio.uio_iov = &aiov;
2308 		auio.uio_iovcnt = 1;
2309 		auio.uio_offset = 0;
2310 		auio.uio_rw = UIO_READ;
2311 		auio.uio_segflg = UIO_USERSPACE;
2312 		auio.uio_procp = p;
2313 		auio.uio_resid = SCARG(uap, count);
2314 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2315 	}
2316 	vput(vp);
2317 	*retval = SCARG(uap, count) - auio.uio_resid;
2318 	return (error);
2319 }
2320 
2321 /*
2322  * Change flags of a file given a path name.
2323  */
2324 /* ARGSUSED */
2325 int
2326 sys_chflags(l, v, retval)
2327 	struct lwp *l;
2328 	void *v;
2329 	register_t *retval;
2330 {
2331 	struct sys_chflags_args /* {
2332 		syscallarg(const char *) path;
2333 		syscallarg(u_long) flags;
2334 	} */ *uap = v;
2335 	struct proc *p = l->l_proc;
2336 	struct vnode *vp;
2337 	int error;
2338 	struct nameidata nd;
2339 
2340 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2341 	if ((error = namei(&nd)) != 0)
2342 		return (error);
2343 	vp = nd.ni_vp;
2344 	error = change_flags(vp, SCARG(uap, flags), p);
2345 	vput(vp);
2346 	return (error);
2347 }
2348 
2349 /*
2350  * Change flags of a file given a file descriptor.
2351  */
2352 /* ARGSUSED */
2353 int
2354 sys_fchflags(l, v, retval)
2355 	struct lwp *l;
2356 	void *v;
2357 	register_t *retval;
2358 {
2359 	struct sys_fchflags_args /* {
2360 		syscallarg(int) fd;
2361 		syscallarg(u_long) flags;
2362 	} */ *uap = v;
2363 	struct proc *p = l->l_proc;
2364 	struct vnode *vp;
2365 	struct file *fp;
2366 	int error;
2367 
2368 	/* getvnode() will use the descriptor for us */
2369 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2370 		return (error);
2371 	vp = (struct vnode *)fp->f_data;
2372 	error = change_flags(vp, SCARG(uap, flags), p);
2373 	VOP_UNLOCK(vp, 0);
2374 	FILE_UNUSE(fp, p);
2375 	return (error);
2376 }
2377 
2378 /*
2379  * Change flags of a file given a path name; this version does
2380  * not follow links.
2381  */
2382 int
2383 sys_lchflags(l, v, retval)
2384 	struct lwp *l;
2385 	void *v;
2386 	register_t *retval;
2387 {
2388 	struct sys_lchflags_args /* {
2389 		syscallarg(const char *) path;
2390 		syscallarg(u_long) flags;
2391 	} */ *uap = v;
2392 	struct proc *p = l->l_proc;
2393 	struct vnode *vp;
2394 	int error;
2395 	struct nameidata nd;
2396 
2397 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2398 	if ((error = namei(&nd)) != 0)
2399 		return (error);
2400 	vp = nd.ni_vp;
2401 	error = change_flags(vp, SCARG(uap, flags), p);
2402 	vput(vp);
2403 	return (error);
2404 }
2405 
2406 /*
2407  * Common routine to change flags of a file.
2408  */
2409 int
2410 change_flags(vp, flags, p)
2411 	struct vnode *vp;
2412 	u_long flags;
2413 	struct proc *p;
2414 {
2415 	struct mount *mp;
2416 	struct vattr vattr;
2417 	int error;
2418 
2419 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2420 		return (error);
2421 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2422 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2423 	/*
2424 	 * Non-superusers cannot change the flags on devices, even if they
2425 	 * own them.
2426 	 */
2427 	if (suser(p->p_ucred, &p->p_acflag) != 0) {
2428 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2429 			goto out;
2430 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2431 			error = EINVAL;
2432 			goto out;
2433 		}
2434 	}
2435 	VATTR_NULL(&vattr);
2436 	vattr.va_flags = flags;
2437 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2438 out:
2439 	vn_finished_write(mp, 0);
2440 	return (error);
2441 }
2442 
2443 /*
2444  * Change mode of a file given path name; this version follows links.
2445  */
2446 /* ARGSUSED */
2447 int
2448 sys_chmod(l, v, retval)
2449 	struct lwp *l;
2450 	void *v;
2451 	register_t *retval;
2452 {
2453 	struct sys_chmod_args /* {
2454 		syscallarg(const char *) path;
2455 		syscallarg(int) mode;
2456 	} */ *uap = v;
2457 	struct proc *p = l->l_proc;
2458 	int error;
2459 	struct nameidata nd;
2460 
2461 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2462 	if ((error = namei(&nd)) != 0)
2463 		return (error);
2464 
2465 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2466 
2467 	vrele(nd.ni_vp);
2468 	return (error);
2469 }
2470 
2471 /*
2472  * Change mode of a file given a file descriptor.
2473  */
2474 /* ARGSUSED */
2475 int
2476 sys_fchmod(l, v, retval)
2477 	struct lwp *l;
2478 	void *v;
2479 	register_t *retval;
2480 {
2481 	struct sys_fchmod_args /* {
2482 		syscallarg(int) fd;
2483 		syscallarg(int) mode;
2484 	} */ *uap = v;
2485 	struct proc *p = l->l_proc;
2486 	struct file *fp;
2487 	int error;
2488 
2489 	/* getvnode() will use the descriptor for us */
2490 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2491 		return (error);
2492 
2493 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2494 	FILE_UNUSE(fp, p);
2495 	return (error);
2496 }
2497 
2498 /*
2499  * Change mode of a file given path name; this version does not follow links.
2500  */
2501 /* ARGSUSED */
2502 int
2503 sys_lchmod(l, v, retval)
2504 	struct lwp *l;
2505 	void *v;
2506 	register_t *retval;
2507 {
2508 	struct sys_lchmod_args /* {
2509 		syscallarg(const char *) path;
2510 		syscallarg(int) mode;
2511 	} */ *uap = v;
2512 	struct proc *p = l->l_proc;
2513 	int error;
2514 	struct nameidata nd;
2515 
2516 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2517 	if ((error = namei(&nd)) != 0)
2518 		return (error);
2519 
2520 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2521 
2522 	vrele(nd.ni_vp);
2523 	return (error);
2524 }
2525 
2526 /*
2527  * Common routine to set mode given a vnode.
2528  */
2529 static int
2530 change_mode(vp, mode, p)
2531 	struct vnode *vp;
2532 	int mode;
2533 	struct proc *p;
2534 {
2535 	struct mount *mp;
2536 	struct vattr vattr;
2537 	int error;
2538 
2539 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2540 		return (error);
2541 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2542 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2543 	VATTR_NULL(&vattr);
2544 	vattr.va_mode = mode & ALLPERMS;
2545 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2546 	VOP_UNLOCK(vp, 0);
2547 	vn_finished_write(mp, 0);
2548 	return (error);
2549 }
2550 
2551 /*
2552  * Set ownership given a path name; this version follows links.
2553  */
2554 /* ARGSUSED */
2555 int
2556 sys_chown(l, v, retval)
2557 	struct lwp *l;
2558 	void *v;
2559 	register_t *retval;
2560 {
2561 	struct sys_chown_args /* {
2562 		syscallarg(const char *) path;
2563 		syscallarg(uid_t) uid;
2564 		syscallarg(gid_t) gid;
2565 	} */ *uap = v;
2566 	struct proc *p = l->l_proc;
2567 	int error;
2568 	struct nameidata nd;
2569 
2570 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2571 	if ((error = namei(&nd)) != 0)
2572 		return (error);
2573 
2574 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2575 
2576 	vrele(nd.ni_vp);
2577 	return (error);
2578 }
2579 
2580 /*
2581  * Set ownership given a path name; this version follows links.
2582  * Provides POSIX semantics.
2583  */
2584 /* ARGSUSED */
2585 int
2586 sys___posix_chown(l, v, retval)
2587 	struct lwp *l;
2588 	void *v;
2589 	register_t *retval;
2590 {
2591 	struct sys_chown_args /* {
2592 		syscallarg(const char *) path;
2593 		syscallarg(uid_t) uid;
2594 		syscallarg(gid_t) gid;
2595 	} */ *uap = v;
2596 	struct proc *p = l->l_proc;
2597 	int error;
2598 	struct nameidata nd;
2599 
2600 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2601 	if ((error = namei(&nd)) != 0)
2602 		return (error);
2603 
2604 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2605 
2606 	vrele(nd.ni_vp);
2607 	return (error);
2608 }
2609 
2610 /*
2611  * Set ownership given a file descriptor.
2612  */
2613 /* ARGSUSED */
2614 int
2615 sys_fchown(l, v, retval)
2616 	struct lwp *l;
2617 	void *v;
2618 	register_t *retval;
2619 {
2620 	struct sys_fchown_args /* {
2621 		syscallarg(int) fd;
2622 		syscallarg(uid_t) uid;
2623 		syscallarg(gid_t) gid;
2624 	} */ *uap = v;
2625 	struct proc *p = l->l_proc;
2626 	int error;
2627 	struct file *fp;
2628 
2629 	/* getvnode() will use the descriptor for us */
2630 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2631 		return (error);
2632 
2633 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2634 	    SCARG(uap, gid), p, 0);
2635 	FILE_UNUSE(fp, p);
2636 	return (error);
2637 }
2638 
2639 /*
2640  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2641  */
2642 /* ARGSUSED */
2643 int
2644 sys___posix_fchown(l, v, retval)
2645 	struct lwp *l;
2646 	void *v;
2647 	register_t *retval;
2648 {
2649 	struct sys_fchown_args /* {
2650 		syscallarg(int) fd;
2651 		syscallarg(uid_t) uid;
2652 		syscallarg(gid_t) gid;
2653 	} */ *uap = v;
2654 	struct proc *p = l->l_proc;
2655 	int error;
2656 	struct file *fp;
2657 
2658 	/* getvnode() will use the descriptor for us */
2659 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2660 		return (error);
2661 
2662 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2663 	    SCARG(uap, gid), p, 1);
2664 	FILE_UNUSE(fp, p);
2665 	return (error);
2666 }
2667 
2668 /*
2669  * Set ownership given a path name; this version does not follow links.
2670  */
2671 /* ARGSUSED */
2672 int
2673 sys_lchown(l, v, retval)
2674 	struct lwp *l;
2675 	void *v;
2676 	register_t *retval;
2677 {
2678 	struct sys_lchown_args /* {
2679 		syscallarg(const char *) path;
2680 		syscallarg(uid_t) uid;
2681 		syscallarg(gid_t) gid;
2682 	} */ *uap = v;
2683 	struct proc *p = l->l_proc;
2684 	int error;
2685 	struct nameidata nd;
2686 
2687 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2688 	if ((error = namei(&nd)) != 0)
2689 		return (error);
2690 
2691 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2692 
2693 	vrele(nd.ni_vp);
2694 	return (error);
2695 }
2696 
2697 /*
2698  * Set ownership given a path name; this version does not follow links.
2699  * Provides POSIX/XPG semantics.
2700  */
2701 /* ARGSUSED */
2702 int
2703 sys___posix_lchown(l, v, retval)
2704 	struct lwp *l;
2705 	void *v;
2706 	register_t *retval;
2707 {
2708 	struct sys_lchown_args /* {
2709 		syscallarg(const char *) path;
2710 		syscallarg(uid_t) uid;
2711 		syscallarg(gid_t) gid;
2712 	} */ *uap = v;
2713 	struct proc *p = l->l_proc;
2714 	int error;
2715 	struct nameidata nd;
2716 
2717 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2718 	if ((error = namei(&nd)) != 0)
2719 		return (error);
2720 
2721 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2722 
2723 	vrele(nd.ni_vp);
2724 	return (error);
2725 }
2726 
2727 /*
2728  * Common routine to set ownership given a vnode.
2729  */
2730 static int
2731 change_owner(vp, uid, gid, p, posix_semantics)
2732 	struct vnode *vp;
2733 	uid_t uid;
2734 	gid_t gid;
2735 	struct proc *p;
2736 	int posix_semantics;
2737 {
2738 	struct mount *mp;
2739 	struct vattr vattr;
2740 	mode_t newmode;
2741 	int error;
2742 
2743 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2744 		return (error);
2745 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2746 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2747 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2748 		goto out;
2749 
2750 #define CHANGED(x) ((int)(x) != -1)
2751 	newmode = vattr.va_mode;
2752 	if (posix_semantics) {
2753 		/*
2754 		 * POSIX/XPG semantics: if the caller is not the super-user,
2755 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2756 		 * the XPG consider the behaviour for calls by the super-user
2757 		 * implementation-defined; we leave the set-user-id and set-
2758 		 * group-id settings intact in that case.
2759 		 */
2760 		if (suser(p->p_ucred, NULL) != 0)
2761 			newmode &= ~(S_ISUID | S_ISGID);
2762 	} else {
2763 		/*
2764 		 * NetBSD semantics: when changing owner and/or group,
2765 		 * clear the respective bit(s).
2766 		 */
2767 		if (CHANGED(uid))
2768 			newmode &= ~S_ISUID;
2769 		if (CHANGED(gid))
2770 			newmode &= ~S_ISGID;
2771 	}
2772 	/* Update va_mode iff altered. */
2773 	if (vattr.va_mode == newmode)
2774 		newmode = VNOVAL;
2775 
2776 	VATTR_NULL(&vattr);
2777 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2778 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2779 	vattr.va_mode = newmode;
2780 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2781 #undef CHANGED
2782 
2783 out:
2784 	VOP_UNLOCK(vp, 0);
2785 	vn_finished_write(mp, 0);
2786 	return (error);
2787 }
2788 
2789 /*
2790  * Set the access and modification times given a path name; this
2791  * version follows links.
2792  */
2793 /* ARGSUSED */
2794 int
2795 sys_utimes(l, v, retval)
2796 	struct lwp *l;
2797 	void *v;
2798 	register_t *retval;
2799 {
2800 	struct sys_utimes_args /* {
2801 		syscallarg(const char *) path;
2802 		syscallarg(const struct timeval *) tptr;
2803 	} */ *uap = v;
2804 	struct proc *p = l->l_proc;
2805 	int error;
2806 	struct nameidata nd;
2807 
2808 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2809 	if ((error = namei(&nd)) != 0)
2810 		return (error);
2811 
2812 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2813 
2814 	vrele(nd.ni_vp);
2815 	return (error);
2816 }
2817 
2818 /*
2819  * Set the access and modification times given a file descriptor.
2820  */
2821 /* ARGSUSED */
2822 int
2823 sys_futimes(l, v, retval)
2824 	struct lwp *l;
2825 	void *v;
2826 	register_t *retval;
2827 {
2828 	struct sys_futimes_args /* {
2829 		syscallarg(int) fd;
2830 		syscallarg(const struct timeval *) tptr;
2831 	} */ *uap = v;
2832 	struct proc *p = l->l_proc;
2833 	int error;
2834 	struct file *fp;
2835 
2836 	/* getvnode() will use the descriptor for us */
2837 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2838 		return (error);
2839 
2840 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2841 	FILE_UNUSE(fp, p);
2842 	return (error);
2843 }
2844 
2845 /*
2846  * Set the access and modification times given a path name; this
2847  * version does not follow links.
2848  */
2849 /* ARGSUSED */
2850 int
2851 sys_lutimes(l, v, retval)
2852 	struct lwp *l;
2853 	void *v;
2854 	register_t *retval;
2855 {
2856 	struct sys_lutimes_args /* {
2857 		syscallarg(const char *) path;
2858 		syscallarg(const struct timeval *) tptr;
2859 	} */ *uap = v;
2860 	struct proc *p = l->l_proc;
2861 	int error;
2862 	struct nameidata nd;
2863 
2864 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2865 	if ((error = namei(&nd)) != 0)
2866 		return (error);
2867 
2868 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2869 
2870 	vrele(nd.ni_vp);
2871 	return (error);
2872 }
2873 
2874 /*
2875  * Common routine to set access and modification times given a vnode.
2876  */
2877 static int
2878 change_utimes(vp, tptr, p)
2879 	struct vnode *vp;
2880 	const struct timeval *tptr;
2881 	struct proc *p;
2882 {
2883 	struct timeval tv[2];
2884 	struct mount *mp;
2885 	struct vattr vattr;
2886 	int error;
2887 
2888 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2889 		return (error);
2890 	VATTR_NULL(&vattr);
2891 	if (tptr == NULL) {
2892 		microtime(&tv[0]);
2893 		tv[1] = tv[0];
2894 		vattr.va_vaflags |= VA_UTIMES_NULL;
2895 	} else {
2896 		error = copyin(tptr, tv, sizeof(tv));
2897 		if (error)
2898 			goto out;
2899 	}
2900 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2901 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2902 	vattr.va_atime.tv_sec = tv[0].tv_sec;
2903 	vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2904 	vattr.va_mtime.tv_sec = tv[1].tv_sec;
2905 	vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2906 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2907 	VOP_UNLOCK(vp, 0);
2908 out:
2909 	vn_finished_write(mp, 0);
2910 	return (error);
2911 }
2912 
2913 /*
2914  * Truncate a file given its path name.
2915  */
2916 /* ARGSUSED */
2917 int
2918 sys_truncate(l, v, retval)
2919 	struct lwp *l;
2920 	void *v;
2921 	register_t *retval;
2922 {
2923 	struct sys_truncate_args /* {
2924 		syscallarg(const char *) path;
2925 		syscallarg(int) pad;
2926 		syscallarg(off_t) length;
2927 	} */ *uap = v;
2928 	struct proc *p = l->l_proc;
2929 	struct vnode *vp;
2930 	struct mount *mp;
2931 	struct vattr vattr;
2932 	int error;
2933 	struct nameidata nd;
2934 
2935 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2936 	if ((error = namei(&nd)) != 0)
2937 		return (error);
2938 	vp = nd.ni_vp;
2939 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2940 		vrele(vp);
2941 		return (error);
2942 	}
2943 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2944 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2945 	if (vp->v_type == VDIR)
2946 		error = EISDIR;
2947 	else if ((error = vn_writechk(vp)) == 0 &&
2948 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2949 		VATTR_NULL(&vattr);
2950 		vattr.va_size = SCARG(uap, length);
2951 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2952 	}
2953 	vput(vp);
2954 	vn_finished_write(mp, 0);
2955 	return (error);
2956 }
2957 
2958 /*
2959  * Truncate a file given a file descriptor.
2960  */
2961 /* ARGSUSED */
2962 int
2963 sys_ftruncate(l, v, retval)
2964 	struct lwp *l;
2965 	void *v;
2966 	register_t *retval;
2967 {
2968 	struct sys_ftruncate_args /* {
2969 		syscallarg(int) fd;
2970 		syscallarg(int) pad;
2971 		syscallarg(off_t) length;
2972 	} */ *uap = v;
2973 	struct proc *p = l->l_proc;
2974 	struct mount *mp;
2975 	struct vattr vattr;
2976 	struct vnode *vp;
2977 	struct file *fp;
2978 	int error;
2979 
2980 	/* getvnode() will use the descriptor for us */
2981 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2982 		return (error);
2983 	if ((fp->f_flag & FWRITE) == 0) {
2984 		error = EINVAL;
2985 		goto out;
2986 	}
2987 	vp = (struct vnode *)fp->f_data;
2988 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2989 		FILE_UNUSE(fp, p);
2990 		return (error);
2991 	}
2992 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2993 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2994 	if (vp->v_type == VDIR)
2995 		error = EISDIR;
2996 	else if ((error = vn_writechk(vp)) == 0) {
2997 		VATTR_NULL(&vattr);
2998 		vattr.va_size = SCARG(uap, length);
2999 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
3000 	}
3001 	VOP_UNLOCK(vp, 0);
3002 	vn_finished_write(mp, 0);
3003  out:
3004 	FILE_UNUSE(fp, p);
3005 	return (error);
3006 }
3007 
3008 /*
3009  * Sync an open file.
3010  */
3011 /* ARGSUSED */
3012 int
3013 sys_fsync(l, v, retval)
3014 	struct lwp *l;
3015 	void *v;
3016 	register_t *retval;
3017 {
3018 	struct sys_fsync_args /* {
3019 		syscallarg(int) fd;
3020 	} */ *uap = v;
3021 	struct proc *p = l->l_proc;
3022 	struct vnode *vp;
3023 	struct mount *mp;
3024 	struct file *fp;
3025 	int error;
3026 
3027 	/* getvnode() will use the descriptor for us */
3028 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3029 		return (error);
3030 	vp = (struct vnode *)fp->f_data;
3031 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3032 		FILE_UNUSE(fp, p);
3033 		return (error);
3034 	}
3035 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3036 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
3037 	if (error == 0 && bioops.io_fsync != NULL &&
3038 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3039 		(*bioops.io_fsync)(vp, 0);
3040 	VOP_UNLOCK(vp, 0);
3041 	vn_finished_write(mp, 0);
3042 	FILE_UNUSE(fp, p);
3043 	return (error);
3044 }
3045 
3046 /*
3047  * Sync a range of file data.  API modeled after that found in AIX.
3048  *
3049  * FDATASYNC indicates that we need only save enough metadata to be able
3050  * to re-read the written data.  Note we duplicate AIX's requirement that
3051  * the file be open for writing.
3052  */
3053 /* ARGSUSED */
3054 int
3055 sys_fsync_range(l, v, retval)
3056 	struct lwp *l;
3057 	void *v;
3058 	register_t *retval;
3059 {
3060 	struct sys_fsync_range_args /* {
3061 		syscallarg(int) fd;
3062 		syscallarg(int) flags;
3063 		syscallarg(off_t) start;
3064 		syscallarg(int) length;
3065 	} */ *uap = v;
3066 	struct proc *p = l->l_proc;
3067 	struct vnode *vp;
3068 	struct file *fp;
3069 	int flags, nflags;
3070 	off_t s, e, len;
3071 	int error;
3072 
3073 	/* getvnode() will use the descriptor for us */
3074 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3075 		return (error);
3076 
3077 	if ((fp->f_flag & FWRITE) == 0) {
3078 		FILE_UNUSE(fp, p);
3079 		return (EBADF);
3080 	}
3081 
3082 	flags = SCARG(uap, flags);
3083 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3084 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3085 		return (EINVAL);
3086 	}
3087 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3088 	if (flags & FDATASYNC)
3089 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3090 	else
3091 		nflags = FSYNC_WAIT;
3092 	if (flags & FDISKSYNC)
3093 		nflags |= FSYNC_CACHE;
3094 
3095 	len = SCARG(uap, length);
3096 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3097 	if (len) {
3098 		s = SCARG(uap, start);
3099 		e = s + len;
3100 		if (e < s) {
3101 			FILE_UNUSE(fp, p);
3102 			return (EINVAL);
3103 		}
3104 	} else {
3105 		e = 0;
3106 		s = 0;
3107 	}
3108 
3109 	vp = (struct vnode *)fp->f_data;
3110 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3111 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p);
3112 
3113 	if (error == 0 && bioops.io_fsync != NULL &&
3114 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3115 		(*bioops.io_fsync)(vp, nflags);
3116 
3117 	VOP_UNLOCK(vp, 0);
3118 	FILE_UNUSE(fp, p);
3119 	return (error);
3120 }
3121 
3122 /*
3123  * Sync the data of an open file.
3124  */
3125 /* ARGSUSED */
3126 int
3127 sys_fdatasync(l, v, retval)
3128 	struct lwp *l;
3129 	void *v;
3130 	register_t *retval;
3131 {
3132 	struct sys_fdatasync_args /* {
3133 		syscallarg(int) fd;
3134 	} */ *uap = v;
3135 	struct proc *p = l->l_proc;
3136 	struct vnode *vp;
3137 	struct file *fp;
3138 	int error;
3139 
3140 	/* getvnode() will use the descriptor for us */
3141 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3142 		return (error);
3143 	if ((fp->f_flag & FWRITE) == 0) {
3144 		FILE_UNUSE(fp, p);
3145 		return (EBADF);
3146 	}
3147 	vp = (struct vnode *)fp->f_data;
3148 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3149 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
3150 	VOP_UNLOCK(vp, 0);
3151 	FILE_UNUSE(fp, p);
3152 	return (error);
3153 }
3154 
3155 /*
3156  * Rename files, (standard) BSD semantics frontend.
3157  */
3158 /* ARGSUSED */
3159 int
3160 sys_rename(l, v, retval)
3161 	struct lwp *l;
3162 	void *v;
3163 	register_t *retval;
3164 {
3165 	struct sys_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, 0));
3172 }
3173 
3174 /*
3175  * Rename files, POSIX semantics frontend.
3176  */
3177 /* ARGSUSED */
3178 int
3179 sys___posix_rename(l, v, retval)
3180 	struct lwp *l;
3181 	void *v;
3182 	register_t *retval;
3183 {
3184 	struct sys___posix_rename_args /* {
3185 		syscallarg(const char *) from;
3186 		syscallarg(const char *) to;
3187 	} */ *uap = v;
3188 	struct proc *p = l->l_proc;
3189 
3190 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
3191 }
3192 
3193 /*
3194  * Rename files.  Source and destination must either both be directories,
3195  * or both not be directories.  If target is a directory, it must be empty.
3196  * If `from' and `to' refer to the same object, the value of the `retain'
3197  * argument is used to determine whether `from' will be
3198  *
3199  * (retain == 0)	deleted unless `from' and `to' refer to the same
3200  *			object in the file system's name space (BSD).
3201  * (retain == 1)	always retained (POSIX).
3202  */
3203 static int
3204 rename_files(from, to, p, retain)
3205 	const char *from, *to;
3206 	struct proc *p;
3207 	int retain;
3208 {
3209 	struct mount *mp = NULL;
3210 	struct vnode *tvp, *fvp, *tdvp;
3211 	struct nameidata fromnd, tond;
3212 	int error;
3213 
3214 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
3215 	    from, p);
3216 	if ((error = namei(&fromnd)) != 0)
3217 		return (error);
3218 	fvp = fromnd.ni_vp;
3219 	error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH);
3220 	if (error != 0) {
3221 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3222 		vrele(fromnd.ni_dvp);
3223 		vrele(fvp);
3224 		if (fromnd.ni_startdir)
3225 			vrele(fromnd.ni_startdir);
3226 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3227 		return (error);
3228 	}
3229 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3230 	    (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p);
3231 	if ((error = namei(&tond)) != 0) {
3232 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3233 		vrele(fromnd.ni_dvp);
3234 		vrele(fvp);
3235 		goto out1;
3236 	}
3237 	tdvp = tond.ni_dvp;
3238 	tvp = tond.ni_vp;
3239 
3240 	if (tvp != NULL) {
3241 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3242 			error = ENOTDIR;
3243 			goto out;
3244 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3245 			error = EISDIR;
3246 			goto out;
3247 		}
3248 	}
3249 
3250 	if (fvp == tdvp)
3251 		error = EINVAL;
3252 
3253 	/*
3254 	 * Source and destination refer to the same object.
3255 	 */
3256 	if (fvp == tvp) {
3257 		if (retain)
3258 			error = -1;
3259 		else if (fromnd.ni_dvp == tdvp &&
3260 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3261 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3262 		          tond.ni_cnd.cn_nameptr,
3263 		          fromnd.ni_cnd.cn_namelen))
3264 		error = -1;
3265 	}
3266 
3267 out:
3268 	if (!error) {
3269 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
3270 		if (fromnd.ni_dvp != tdvp)
3271 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3272 		if (tvp) {
3273 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
3274 		}
3275 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3276 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3277 	} else {
3278 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3279 		if (tdvp == tvp)
3280 			vrele(tdvp);
3281 		else
3282 			vput(tdvp);
3283 		if (tvp)
3284 			vput(tvp);
3285 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3286 		vrele(fromnd.ni_dvp);
3287 		vrele(fvp);
3288 	}
3289 	vrele(tond.ni_startdir);
3290 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3291 out1:
3292 	vn_finished_write(mp, 0);
3293 	if (fromnd.ni_startdir)
3294 		vrele(fromnd.ni_startdir);
3295 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3296 	return (error == -1 ? 0 : error);
3297 }
3298 
3299 /*
3300  * Make a directory file.
3301  */
3302 /* ARGSUSED */
3303 int
3304 sys_mkdir(l, v, retval)
3305 	struct lwp *l;
3306 	void *v;
3307 	register_t *retval;
3308 {
3309 	struct sys_mkdir_args /* {
3310 		syscallarg(const char *) path;
3311 		syscallarg(int) mode;
3312 	} */ *uap = v;
3313 	struct proc *p = l->l_proc;
3314 	struct mount *mp;
3315 	struct vnode *vp;
3316 	struct vattr vattr;
3317 	int error;
3318 	struct nameidata nd;
3319 
3320 restart:
3321 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE,
3322 	    SCARG(uap, path), p);
3323 	if ((error = namei(&nd)) != 0)
3324 		return (error);
3325 	vp = nd.ni_vp;
3326 	if (vp != NULL) {
3327 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3328 		if (nd.ni_dvp == vp)
3329 			vrele(nd.ni_dvp);
3330 		else
3331 			vput(nd.ni_dvp);
3332 		vrele(vp);
3333 		return (EEXIST);
3334 	}
3335 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3336 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3337 		if (nd.ni_dvp == vp)
3338 			vrele(nd.ni_dvp);
3339 		else
3340 			vput(nd.ni_dvp);
3341 		if ((error = vn_start_write(NULL, &mp,
3342 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3343 			return (error);
3344 		goto restart;
3345 	}
3346 	VATTR_NULL(&vattr);
3347 	vattr.va_type = VDIR;
3348 	vattr.va_mode =
3349 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3350 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3351 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3352 	if (!error)
3353 		vput(nd.ni_vp);
3354 	vn_finished_write(mp, 0);
3355 	return (error);
3356 }
3357 
3358 /*
3359  * Remove a directory file.
3360  */
3361 /* ARGSUSED */
3362 int
3363 sys_rmdir(l, v, retval)
3364 	struct lwp *l;
3365 	void *v;
3366 	register_t *retval;
3367 {
3368 	struct sys_rmdir_args /* {
3369 		syscallarg(const char *) path;
3370 	} */ *uap = v;
3371 	struct proc *p = l->l_proc;
3372 	struct mount *mp;
3373 	struct vnode *vp;
3374 	int error;
3375 	struct nameidata nd;
3376 
3377 restart:
3378 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3379 	    SCARG(uap, path), p);
3380 	if ((error = namei(&nd)) != 0)
3381 		return (error);
3382 	vp = nd.ni_vp;
3383 	if (vp->v_type != VDIR) {
3384 		error = ENOTDIR;
3385 		goto out;
3386 	}
3387 	/*
3388 	 * No rmdir "." please.
3389 	 */
3390 	if (nd.ni_dvp == vp) {
3391 		error = EINVAL;
3392 		goto out;
3393 	}
3394 	/*
3395 	 * The root of a mounted filesystem cannot be deleted.
3396 	 */
3397 	if (vp->v_flag & VROOT) {
3398 		error = EBUSY;
3399 		goto out;
3400 	}
3401 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3402 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3403 		if (nd.ni_dvp == vp)
3404 			vrele(nd.ni_dvp);
3405 		else
3406 			vput(nd.ni_dvp);
3407 		vput(vp);
3408 		if ((error = vn_start_write(NULL, &mp,
3409 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3410 			return (error);
3411 		goto restart;
3412 	}
3413 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3414 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3415 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3416 	vn_finished_write(mp, 0);
3417 	return (error);
3418 
3419 out:
3420 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3421 	if (nd.ni_dvp == vp)
3422 		vrele(nd.ni_dvp);
3423 	else
3424 		vput(nd.ni_dvp);
3425 	vput(vp);
3426 	return (error);
3427 }
3428 
3429 /*
3430  * Read a block of directory entries in a file system independent format.
3431  */
3432 int
3433 sys_getdents(l, v, retval)
3434 	struct lwp *l;
3435 	void *v;
3436 	register_t *retval;
3437 {
3438 	struct sys_getdents_args /* {
3439 		syscallarg(int) fd;
3440 		syscallarg(char *) buf;
3441 		syscallarg(size_t) count;
3442 	} */ *uap = v;
3443 	struct proc *p = l->l_proc;
3444 	struct file *fp;
3445 	int error, done;
3446 
3447 	/* getvnode() will use the descriptor for us */
3448 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3449 		return (error);
3450 	if ((fp->f_flag & FREAD) == 0) {
3451 		error = EBADF;
3452 		goto out;
3453 	}
3454 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3455 			SCARG(uap, count), &done, p, 0, 0);
3456 #ifdef KTRACE
3457 	if (!error && KTRPOINT(p, KTR_GENIO)) {
3458 		struct iovec iov;
3459 		iov.iov_base = SCARG(uap, buf);
3460 		iov.iov_len = done;
3461 		ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0);
3462 	}
3463 #endif
3464 	*retval = done;
3465  out:
3466 	FILE_UNUSE(fp, p);
3467 	return (error);
3468 }
3469 
3470 /*
3471  * Set the mode mask for creation of filesystem nodes.
3472  */
3473 int
3474 sys_umask(l, v, retval)
3475 	struct lwp *l;
3476 	void *v;
3477 	register_t *retval;
3478 {
3479 	struct sys_umask_args /* {
3480 		syscallarg(mode_t) newmask;
3481 	} */ *uap = v;
3482 	struct proc *p = l->l_proc;
3483 	struct cwdinfo *cwdi;
3484 
3485 	cwdi = p->p_cwdi;
3486 	*retval = cwdi->cwdi_cmask;
3487 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3488 	return (0);
3489 }
3490 
3491 /*
3492  * Void all references to file by ripping underlying filesystem
3493  * away from vnode.
3494  */
3495 /* ARGSUSED */
3496 int
3497 sys_revoke(l, v, retval)
3498 	struct lwp *l;
3499 	void *v;
3500 	register_t *retval;
3501 {
3502 	struct sys_revoke_args /* {
3503 		syscallarg(const char *) path;
3504 	} */ *uap = v;
3505 	struct proc *p = l->l_proc;
3506 	struct mount *mp;
3507 	struct vnode *vp;
3508 	struct vattr vattr;
3509 	int error;
3510 	struct nameidata nd;
3511 
3512 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3513 	if ((error = namei(&nd)) != 0)
3514 		return (error);
3515 	vp = nd.ni_vp;
3516 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3517 		goto out;
3518 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3519 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3520 		goto out;
3521 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3522 		goto out;
3523 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3524 		VOP_REVOKE(vp, REVOKEALL);
3525 	vn_finished_write(mp, 0);
3526 out:
3527 	vrele(vp);
3528 	return (error);
3529 }
3530 
3531 /*
3532  * Convert a user file descriptor to a kernel file entry.
3533  */
3534 int
3535 getvnode(fdp, fd, fpp)
3536 	struct filedesc *fdp;
3537 	int fd;
3538 	struct file **fpp;
3539 {
3540 	struct vnode *vp;
3541 	struct file *fp;
3542 
3543 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3544 		return (EBADF);
3545 
3546 	FILE_USE(fp);
3547 
3548 	if (fp->f_type != DTYPE_VNODE) {
3549 		FILE_UNUSE(fp, NULL);
3550 		return (EINVAL);
3551 	}
3552 
3553 	vp = (struct vnode *)fp->f_data;
3554 	if (vp->v_type == VBAD) {
3555 		FILE_UNUSE(fp, NULL);
3556 		return (EBADF);
3557 	}
3558 
3559 	*fpp = fp;
3560 	return (0);
3561 }
3562 
3563 /*
3564  * Push extended attribute configuration information into the VFS.
3565  *
3566  * NOTE: Not all file systems that support extended attributes will
3567  * require the use of this system call.
3568  */
3569 int
3570 sys_extattrctl(struct lwp *l, void *v, register_t *retval)
3571 {
3572 	struct sys_extattrctl_args /* {
3573 		syscallarg(const char *) path;
3574 		syscallarg(int) cmd;
3575 		syscallarg(const char *) filename;
3576 		syscallarg(int) attrnamespace;
3577 		syscallarg(const char *) attrname;
3578 	} */ *uap = v;
3579 	struct proc *p = l->l_proc;
3580 	struct vnode *vp;
3581 	struct nameidata nd;
3582 	struct mount *mp;
3583 	char attrname[EXTATTR_MAXNAMELEN];
3584 	int error;
3585 
3586 	if (SCARG(uap, attrname) != NULL) {
3587 		error = copyinstr(SCARG(uap, attrname), attrname,
3588 		    sizeof(attrname), NULL);
3589 		if (error)
3590 			return (error);
3591 	}
3592 
3593 	vp = NULL;
3594 	if (SCARG(uap, filename) != NULL) {
3595 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3596 		    SCARG(uap, filename), p);
3597 		error = namei(&nd);
3598 		if (error)
3599 			return (error);
3600 		vp = nd.ni_vp;
3601 	}
3602 
3603 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3604 	error = namei(&nd);
3605 	if (error) {
3606 		if (vp != NULL)
3607 			vput(vp);
3608 		return (error);
3609 	}
3610 
3611 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
3612 	if (error) {
3613 		if (vp != NULL)
3614 			vput(vp);
3615 		return (error);
3616 	}
3617 
3618 	error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), vp,
3619 	    SCARG(uap, attrnamespace),
3620 	    SCARG(uap, attrname) != NULL ? attrname : NULL, p);
3621 
3622 	vn_finished_write(mp, 0);
3623 
3624 	if (vp != NULL)
3625 		vrele(vp);
3626 
3627 	return (error);
3628 }
3629 
3630 /*
3631  * Set a named extended attribute on a file or directory.
3632  */
3633 static int
3634 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3635     const void *data, size_t nbytes, struct proc *p, register_t *retval)
3636 {
3637 	struct mount *mp;
3638 	struct uio auio;
3639 	struct iovec aiov;
3640 	ssize_t cnt;
3641 	int error;
3642 
3643 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3644 	if (error)
3645 		return (error);
3646 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3647 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3648 
3649 	aiov.iov_base = (caddr_t) data;		/* XXX kills const */
3650 	aiov.iov_len = nbytes;
3651 	auio.uio_iov = &aiov;
3652 	auio.uio_iovcnt = 1;
3653 	auio.uio_offset = 0;
3654 	if (nbytes > INT_MAX) {
3655 		error = EINVAL;
3656 		goto done;
3657 	}
3658 	auio.uio_resid = nbytes;
3659 	auio.uio_rw = UIO_WRITE;
3660 	auio.uio_segflg = UIO_USERSPACE;
3661 	auio.uio_procp = p;
3662 	cnt = nbytes;
3663 
3664 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
3665 	    p->p_ucred, p);
3666 	cnt -= auio.uio_resid;
3667 	retval[0] = cnt;
3668 
3669  done:
3670 	VOP_UNLOCK(vp, 0);
3671 	vn_finished_write(mp, 0);
3672 	return (error);
3673 }
3674 
3675 int
3676 sys_extattr_set_fd(struct lwp *l, void *v, register_t *retval)
3677 {
3678 	struct sys_extattr_set_fd_args /* {
3679 		syscallarg(int) fd;
3680 		syscallarg(int) attrnamespace;
3681 		syscallarg(const char *) attrname;
3682 		syscallarg(const void *) data;
3683 		syscallarg(size_t) nbytes;
3684 	} */ *uap = v;
3685 	struct proc *p = l->l_proc;
3686 	struct file *fp;
3687 	struct vnode *vp;
3688 	char attrname[EXTATTR_MAXNAMELEN];
3689 	int error;
3690 
3691 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3692 	    NULL);
3693 	if (error)
3694 		return (error);
3695 
3696 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3697 	if (error)
3698 		return (error);
3699 	vp = (struct vnode *) fp->f_data;
3700 
3701 	error = extattr_set_vp(vp, SCARG(uap, attrnamespace), attrname,
3702 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3703 
3704 	FILE_UNUSE(fp, p);
3705 	return (error);
3706 }
3707 
3708 int
3709 sys_extattr_set_file(struct lwp *l, void *v, register_t *retval)
3710 {
3711 	struct sys_extattr_set_file_args /* {
3712 		syscallarg(const char *) path;
3713 		syscallarg(int) attrnamespace;
3714 		syscallarg(const char *) attrname;
3715 		syscallarg(const void *) data;
3716 		syscallarg(size_t) nbytes;
3717 	} */ *uap = v;
3718 	struct proc *p = l->l_proc;
3719 	struct nameidata nd;
3720 	char attrname[EXTATTR_MAXNAMELEN];
3721 	int error;
3722 
3723 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3724 	    NULL);
3725 	if (error)
3726 		return (error);
3727 
3728 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3729 	error = namei(&nd);
3730 	if (error)
3731 		return (error);
3732 
3733 	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3734 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3735 
3736 	vrele(nd.ni_vp);
3737 	return (error);
3738 }
3739 
3740 int
3741 sys_extattr_set_link(struct lwp *l, void *v, register_t *retval)
3742 {
3743 	struct sys_extattr_set_link_args /* {
3744 		syscallarg(const char *) path;
3745 		syscallarg(int) attrnamespace;
3746 		syscallarg(const char *) attrname;
3747 		syscallarg(const void *) data;
3748 		syscallarg(size_t) nbytes;
3749 	} */ *uap = v;
3750 	struct proc *p = l->l_proc;
3751 	struct nameidata nd;
3752 	char attrname[EXTATTR_MAXNAMELEN];
3753 	int error;
3754 
3755 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3756 	    NULL);
3757 	if (error)
3758 		return (error);
3759 
3760 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3761 	error = namei(&nd);
3762 	if (error)
3763 		return (error);
3764 
3765 	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3766 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3767 
3768 	vrele(nd.ni_vp);
3769 	return (error);
3770 }
3771 
3772 /*
3773  * Get a named extended attribute on a file or directory.
3774  */
3775 static int
3776 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3777     void *data, size_t nbytes, struct proc *p, register_t *retval)
3778 {
3779 	struct uio auio, *auiop;
3780 	struct iovec aiov;
3781 	ssize_t cnt;
3782 	size_t size, *sizep;
3783 	int error;
3784 
3785 	VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
3786 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3787 
3788 	/*
3789 	 * Slightly unusual semantics: if the user provides a NULL data
3790 	 * pointer, they don't want to receive the data, just the maximum
3791 	 * read length.
3792 	 */
3793 	auiop = NULL;
3794 	sizep = NULL;
3795 	cnt = 0;
3796 	if (data != NULL) {
3797 		aiov.iov_base = data;
3798 		aiov.iov_len = nbytes;
3799 		auio.uio_iov = &aiov;
3800 		auio.uio_offset = 0;
3801 		if (nbytes > INT_MAX) {
3802 			error = EINVAL;
3803 			goto done;
3804 		}
3805 		auio.uio_resid = nbytes;
3806 		auio.uio_rw = UIO_READ;
3807 		auio.uio_segflg = UIO_USERSPACE;
3808 		auio.uio_procp = p;
3809 		auiop = &auio;
3810 		cnt = nbytes;
3811 	} else
3812 		sizep = &size;
3813 
3814 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, auiop, sizep,
3815 	    p->p_ucred, p);
3816 
3817 	if (auiop != NULL) {
3818 		cnt -= auio.uio_resid;
3819 		retval[0] = cnt;
3820 	} else
3821 		retval[0] = size;
3822 
3823  done:
3824 	VOP_UNLOCK(vp, 0);
3825 	return (error);
3826 }
3827 
3828 int
3829 sys_extattr_get_fd(struct lwp *l, void *v, register_t *retval)
3830 {
3831 	struct sys_extattr_get_fd_args /* {
3832 		syscallarg(int) fd;
3833 		syscallarg(int) attrnamespace;
3834 		syscallarg(const char *) attrname;
3835 		syscallarg(void *) data;
3836 		syscallarg(size_t) nbytes;
3837 	} */ *uap = v;
3838 	struct proc *p = l->l_proc;
3839 	struct file *fp;
3840 	struct vnode *vp;
3841 	char attrname[EXTATTR_MAXNAMELEN];
3842 	int error;
3843 
3844 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3845 	    NULL);
3846 	if (error)
3847 		return (error);
3848 
3849 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3850 	if (error)
3851 		return (error);
3852 	vp = (struct vnode *) fp->f_data;
3853 
3854 	error = extattr_get_vp(vp, SCARG(uap, attrnamespace), attrname,
3855 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3856 
3857 	FILE_UNUSE(fp, p);
3858 	return (error);
3859 }
3860 
3861 int
3862 sys_extattr_get_file(struct lwp *l, void *v, register_t *retval)
3863 {
3864 	struct sys_extattr_get_file_args /* {
3865 		syscallarg(const char *) path;
3866 		syscallarg(int) attrnamespace;
3867 		syscallarg(const char *) attrname;
3868 		syscallarg(void *) data;
3869 		syscallarg(size_t) nbytes;
3870 	} */ *uap = v;
3871 	struct proc *p = l->l_proc;
3872 	struct nameidata nd;
3873 	char attrname[EXTATTR_MAXNAMELEN];
3874 	int error;
3875 
3876 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3877 	    NULL);
3878 	if (error)
3879 		return (error);
3880 
3881 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3882 	error = namei(&nd);
3883 	if (error)
3884 		return (error);
3885 
3886 	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3887 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3888 
3889 	vrele(nd.ni_vp);
3890 	return (error);
3891 }
3892 
3893 int
3894 sys_extattr_get_link(struct lwp *l, void *v, register_t *retval)
3895 {
3896 	struct sys_extattr_get_link_args /* {
3897 		syscallarg(const char *) path;
3898 		syscallarg(int) attrnamespace;
3899 		syscallarg(const char *) attrname;
3900 		syscallarg(void *) data;
3901 		syscallarg(size_t) nbytes;
3902 	} */ *uap = v;
3903 	struct proc *p = l->l_proc;
3904 	struct nameidata nd;
3905 	char attrname[EXTATTR_MAXNAMELEN];
3906 	int error;
3907 
3908 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3909 	    NULL);
3910 	if (error)
3911 		return (error);
3912 
3913 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3914 	error = namei(&nd);
3915 	if (error)
3916 		return (error);
3917 
3918 	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3919 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3920 
3921 	vrele(nd.ni_vp);
3922 	return (error);
3923 }
3924 
3925 /*
3926  * Delete a named extended attribute on a file or directory.
3927  */
3928 static int
3929 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3930     struct proc *p)
3931 {
3932 	struct mount *mp;
3933 	int error;
3934 
3935 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3936 	if (error)
3937 		return (error);
3938 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3939 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3940 
3941 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, p->p_ucred, p);
3942 	if (error == EOPNOTSUPP)
3943 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
3944 		    p->p_ucred, p);
3945 
3946 	VOP_UNLOCK(vp, 0);
3947 	vn_finished_write(mp, 0);
3948 	return (error);
3949 }
3950 
3951 int
3952 sys_extattr_delete_fd(struct lwp *l, void *v, register_t *retval)
3953 {
3954 	struct sys_extattr_delete_fd_args /* {
3955 		syscallarg(int) fd;
3956 		syscallarg(int) attrnamespace;
3957 		syscallarg(const char *) attrname;
3958 	} */ *uap = v;
3959 	struct proc *p = l->l_proc;
3960 	struct file *fp;
3961 	struct vnode *vp;
3962 	char attrname[EXTATTR_MAXNAMELEN];
3963 	int error;
3964 
3965 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3966 	    NULL);
3967 	if (error)
3968 		return (error);
3969 
3970 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3971 	if (error)
3972 		return (error);
3973 	vp = (struct vnode *) fp->f_data;
3974 
3975 	error = extattr_delete_vp(vp, SCARG(uap, attrnamespace), attrname, p);
3976 
3977 	FILE_UNUSE(fp, p);
3978 	return (error);
3979 }
3980 
3981 int
3982 sys_extattr_delete_file(struct lwp *l, void *v, register_t *retval)
3983 {
3984 	struct sys_extattr_delete_file_args /* {
3985 		syscallarg(const char *) path;
3986 		syscallarg(int) attrnamespace;
3987 		syscallarg(const char *) attrname;
3988 	} */ *uap = v;
3989 	struct proc *p = l->l_proc;
3990 	struct nameidata nd;
3991 	char attrname[EXTATTR_MAXNAMELEN];
3992 	int error;
3993 
3994 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3995 	    NULL);
3996 	if (error)
3997 		return (error);
3998 
3999 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4000 	error = namei(&nd);
4001 	if (error)
4002 		return (error);
4003 
4004 	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4005 	    p);
4006 
4007 	vrele(nd.ni_vp);
4008 	return (error);
4009 }
4010 
4011 int
4012 sys_extattr_delete_link(struct lwp *l, void *v, register_t *retval)
4013 {
4014 	struct sys_extattr_delete_link_args /* {
4015 		syscallarg(const char *) path;
4016 		syscallarg(int) attrnamespace;
4017 		syscallarg(const char *) attrname;
4018 	} */ *uap = v;
4019 	struct proc *p = l->l_proc;
4020 	struct nameidata nd;
4021 	char attrname[EXTATTR_MAXNAMELEN];
4022 	int error;
4023 
4024 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
4025 	    NULL);
4026 	if (error)
4027 		return (error);
4028 
4029 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4030 	error = namei(&nd);
4031 	if (error)
4032 		return (error);
4033 
4034 	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4035 	    p);
4036 
4037 	vrele(nd.ni_vp);
4038 	return (error);
4039 }
4040 
4041 /*
4042  * Retrieve a list of extended attributes on a file or directory.
4043  */
4044 static int
4045 extattr_list_vp(struct vnode *vp, int attrnamespace, void *data, size_t nbytes,
4046     struct proc *p, register_t *retval)
4047 {
4048 	struct uio auio, *auiop;
4049 	size_t size, *sizep;
4050 	struct iovec aiov;
4051 	ssize_t cnt;
4052 	int error;
4053 
4054 	VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
4055 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4056 
4057 	auiop = NULL;
4058 	sizep = NULL;
4059 	cnt = 0;
4060 	if (data != NULL) {
4061 		aiov.iov_base = data;
4062 		aiov.iov_len = nbytes;
4063 		auio.uio_iov = &aiov;
4064 		auio.uio_offset = 0;
4065 		if (nbytes > INT_MAX) {
4066 			error = EINVAL;
4067 			goto done;
4068 		}
4069 		auio.uio_resid = nbytes;
4070 		auio.uio_rw = UIO_READ;
4071 		auio.uio_segflg = UIO_USERSPACE;
4072 		auio.uio_procp = p;
4073 		auiop = &auio;
4074 		cnt = nbytes;
4075 	} else
4076 		sizep = &size;
4077 
4078 	error = VOP_LISTEXTATTR(vp, attrnamespace, auiop, sizep,
4079 	    p->p_ucred, p);
4080 
4081 	if (auiop != NULL) {
4082 		cnt -= auio.uio_resid;
4083 		retval[0] = cnt;
4084 	} else
4085 		retval[0] = size;
4086 
4087  done:
4088 	VOP_UNLOCK(vp, 0);
4089 	return (error);
4090 }
4091 
4092 int
4093 sys_extattr_list_fd(struct lwp *l, void *v, register_t *retval)
4094 {
4095 	struct sys_extattr_list_fd_args /* {
4096 		syscallarg(int) fd;
4097 		syscallarg(int) attrnamespace;
4098 		syscallarg(void *) data;
4099 		syscallarg(size_t) nbytes;
4100 	} */ *uap = v;
4101 	struct proc *p = l->l_proc;
4102 	struct file *fp;
4103 	struct vnode *vp;
4104 	int error;
4105 
4106 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
4107 	if (error)
4108 		return (error);
4109 	vp = (struct vnode *) fp->f_data;
4110 
4111 	error = extattr_list_vp(vp, SCARG(uap, attrnamespace),
4112 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4113 
4114 	FILE_UNUSE(fp, p);
4115 	return (error);
4116 }
4117 
4118 int
4119 sys_extattr_list_file(struct lwp *l, void *v, register_t *retval)
4120 {
4121 	struct sys_extattr_list_file_args /* {
4122 		syscallarg(const char *) path;
4123 		syscallarg(int) attrnamespace;
4124 		syscallarg(void *) data;
4125 		syscallarg(size_t) nbytes;
4126 	} */ *uap = v;
4127 	struct proc *p = l->l_proc;
4128 	struct nameidata nd;
4129 	int error;
4130 
4131 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4132 	error = namei(&nd);
4133 	if (error)
4134 		return (error);
4135 
4136 	error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4137 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4138 
4139 	vrele(nd.ni_vp);
4140 	return (error);
4141 }
4142 
4143 int
4144 sys_extattr_list_link(struct lwp *l, void *v, register_t *retval)
4145 {
4146 	struct sys_extattr_list_link_args /* {
4147 		syscallarg(const char *) path;
4148 		syscallarg(int) attrnamespace;
4149 		syscallarg(void *) data;
4150 		syscallarg(size_t) nbytes;
4151 	} */ *uap = v;
4152 	struct proc *p = l->l_proc;
4153 	struct nameidata nd;
4154 	int error;
4155 
4156 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4157 	error = namei(&nd);
4158 	if (error)
4159 		return (error);
4160 
4161 	error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4162 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4163 
4164 	vrele(nd.ni_vp);
4165 	return (error);
4166 }
4167