xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision fad4c9f71477ae11cea2ee75ec82151ac770a534)
1 /*	$NetBSD: vfs_syscalls.c,v 1.243 2006/06/17 07:06:51 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 1.243 2006/06/17 07:06:51 yamt Exp $");
41 
42 #include "opt_compat_netbsd.h"
43 #include "opt_compat_43.h"
44 #include "opt_ktrace.h"
45 #include "opt_verified_exec.h"
46 #include "fss.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/namei.h>
51 #include <sys/filedesc.h>
52 #include <sys/kernel.h>
53 #include <sys/file.h>
54 #include <sys/stat.h>
55 #include <sys/vnode.h>
56 #include <sys/mount.h>
57 #include <sys/proc.h>
58 #include <sys/uio.h>
59 #include <sys/malloc.h>
60 #include <sys/dirent.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 #ifdef VERIFIED_EXEC
68 #include <sys/verified_exec.h>
69 #endif /* VERIFIED_EXEC */
70 #include <sys/kauth.h>
71 
72 #include <miscfs/genfs/genfs.h>
73 #include <miscfs/syncfs/syncfs.h>
74 
75 #ifdef COMPAT_30
76 #include "opt_nfsserver.h"
77 #include <nfs/rpcv2.h>
78 #include <nfs/nfsproto.h>
79 #include <nfs/nfs.h>
80 #include <nfs/nfs_var.h>
81 #endif
82 
83 #if NFSS > 0
84 #include <dev/fssvar.h>
85 #endif
86 
87 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
88 
89 static int change_dir(struct nameidata *, struct lwp *);
90 static int change_flags(struct vnode *, u_long, struct lwp *);
91 static int change_mode(struct vnode *, int, struct lwp *l);
92 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
93 static int change_utimes(struct vnode *vp, const struct timeval *,
94 	       struct lwp *l);
95 static int rename_files(const char *, const char *, struct lwp *, int);
96 
97 void checkdirs(struct vnode *);
98 
99 int dovfsusermount = 0;
100 
101 /*
102  * Virtual File System System Calls
103  */
104 
105 /*
106  * Mount a file system.
107  */
108 
109 #if defined(COMPAT_09) || defined(COMPAT_43)
110 /*
111  * This table is used to maintain compatibility with 4.3BSD
112  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
113  *
114  * Do not modify this table. It should only contain filesystems
115  * supported by NetBSD 0.9 and 4.3BSD.
116  */
117 const char * const mountcompatnames[] = {
118 	NULL,		/* 0 = MOUNT_NONE */
119 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
120 	MOUNT_NFS,	/* 2 */
121 	MOUNT_MFS,	/* 3 */
122 	MOUNT_MSDOS,	/* 4 */
123 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
124 	MOUNT_FDESC,	/* 6 */
125 	MOUNT_KERNFS,	/* 7 */
126 	NULL,		/* 8 = MOUNT_DEVFS */
127 	MOUNT_AFS,	/* 9 */
128 };
129 const int nmountcompatnames = sizeof(mountcompatnames) /
130     sizeof(mountcompatnames[0]);
131 #endif /* COMPAT_09 || COMPAT_43 */
132 
133 /* ARGSUSED */
134 int
135 sys_mount(struct lwp *l, void *v, register_t *retval)
136 {
137 	struct sys_mount_args /* {
138 		syscallarg(const char *) type;
139 		syscallarg(const char *) path;
140 		syscallarg(int) flags;
141 		syscallarg(void *) data;
142 	} */ *uap = v;
143 	struct proc *p = l->l_proc;
144 	struct vnode *vp;
145 	struct mount *mp;
146 	int error, flag = 0;
147 	char fstypename[MFSNAMELEN];
148 	struct vattr va;
149 	struct nameidata nd;
150 	struct vfsops *vfs;
151 
152 	/*
153 	 * if MNT_GETARGS is specified, it should be only flag.
154 	 */
155 
156 	if ((SCARG(uap, flags) & MNT_GETARGS) != 0 &&
157 	    (SCARG(uap, flags) & ~MNT_GETARGS) != 0) {
158 		return EINVAL;
159 	}
160 
161 	if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
162 	    (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
163 				       &p->p_acflag)))
164 		return (error);
165 	/*
166 	 * Get vnode to be covered
167 	 */
168 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
169 	    SCARG(uap, path), l);
170 	if ((error = namei(&nd)) != 0)
171 		return (error);
172 	vp = nd.ni_vp;
173 	/*
174 	 * A lookup in VFS_MOUNT might result in an attempt to
175 	 * lock this vnode again, so make the lock recursive.
176 	 */
177 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
178 	if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
179 		if ((vp->v_flag & VROOT) == 0) {
180 			vput(vp);
181 			return (EINVAL);
182 		}
183 		mp = vp->v_mount;
184 		flag = mp->mnt_flag;
185 		vfs = mp->mnt_op;
186 		/*
187 		 * We only allow the filesystem to be reloaded if it
188 		 * is currently mounted read-only.
189 		 */
190 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
191 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
192 			vput(vp);
193 			return (EOPNOTSUPP);	/* Needs translation */
194 		}
195 		/*
196 		 * In "highly secure" mode, don't let the caller do anything
197 		 * but downgrade a filesystem from read-write to read-only.
198 		 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
199 		 */
200 		if (securelevel >= 2 &&
201 		    SCARG(uap, flags) != MNT_GETARGS &&
202 		    SCARG(uap, flags) !=
203 		    (mp->mnt_flag | MNT_RDONLY |
204 		     MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
205 			vput(vp);
206 			return (EPERM);
207 		}
208 		mp->mnt_flag |= SCARG(uap, flags) &
209 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
210 		/*
211 		 * Only root, or the user that did the original mount is
212 		 * permitted to update it.
213 		 */
214 		if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
215 		    mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred) &&
216 		    (error = kauth_authorize_generic(p->p_cred,
217 					       KAUTH_GENERIC_ISSUSER,
218 					       &p->p_acflag)) != 0) {
219 			vput(vp);
220 			return (error);
221 		}
222 		/*
223 		 * Do not allow NFS export by non-root users. For non-root
224 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
225 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
226 		 */
227 		if (kauth_cred_geteuid(p->p_cred) != 0) {
228 			if (SCARG(uap, flags) & MNT_EXPORTED) {
229 				vput(vp);
230 				return (EPERM);
231 			}
232 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
233 			if (flag & MNT_NOEXEC)
234 				SCARG(uap, flags) |= MNT_NOEXEC;
235 		}
236 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
237 			vput(vp);
238 			return (EPERM);
239 		}
240 		goto update;
241 	} else {
242 		if (securelevel >= 2) {
243 			vput(vp);
244 			return (EPERM);
245 		}
246 	}
247 	/*
248 	 * If the user is not root, ensure that they own the directory
249 	 * onto which we are attempting to mount.
250 	 */
251 	if ((error = VOP_GETATTR(vp, &va, p->p_cred, l)) != 0 ||
252 	    (va.va_uid != kauth_cred_geteuid(p->p_cred) &&
253 		(error = kauth_authorize_generic(p->p_cred,
254 					   KAUTH_GENERIC_ISSUSER,
255 					   &p->p_acflag)) != 0)) {
256 		vput(vp);
257 		return (error);
258 	}
259 	/*
260 	 * Do not allow NFS export by non-root users. For non-root users,
261 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
262 	 * mount point is already MNT_NOEXEC.
263 	 */
264 	if (kauth_cred_geteuid(p->p_cred) != 0) {
265 		if (SCARG(uap, flags) & MNT_EXPORTED) {
266 			vput(vp);
267 			return (EPERM);
268 		}
269 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
270 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
271 			SCARG(uap, flags) |= MNT_NOEXEC;
272 	}
273 	if ((error = vinvalbuf(vp, V_SAVE, p->p_cred, l, 0, 0)) != 0) {
274 		vput(vp);
275 		return (error);
276 	}
277 	if (vp->v_type != VDIR) {
278 		vput(vp);
279 		return (ENOTDIR);
280 	}
281 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
282 	if (error) {
283 #if defined(COMPAT_09) || defined(COMPAT_43)
284 		/*
285 		 * Historically, filesystem types were identified by numbers.
286 		 * If we get an integer for the filesystem type instead of a
287 		 * string, we check to see if it matches one of the historic
288 		 * filesystem types.
289 		 */
290 		u_long fsindex = (u_long)SCARG(uap, type);
291 		if (fsindex >= nmountcompatnames ||
292 		    mountcompatnames[fsindex] == NULL) {
293 			vput(vp);
294 			return (ENODEV);
295 		}
296 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
297 #else
298 		vput(vp);
299 		return (error);
300 #endif
301 	}
302 #ifdef	COMPAT_10
303 	/* Accept `ufs' as an alias for `ffs'. */
304 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
305 		strncpy(fstypename, "ffs", MFSNAMELEN);
306 #endif
307 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
308 		vput(vp);
309 		return (ENODEV);
310 	}
311 	if (vp->v_mountedhere != NULL) {
312 		vput(vp);
313 		return (EBUSY);
314 	}
315 
316 	/*
317 	 * Allocate and initialize the file system.
318 	 */
319 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
320 		M_MOUNT, M_WAITOK);
321 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
322 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
323 	simple_lock_init(&mp->mnt_slock);
324 	(void)vfs_busy(mp, LK_NOWAIT, 0);
325 	mp->mnt_op = vfs;
326 	vfs->vfs_refcount++;
327 	mp->mnt_vnodecovered = vp;
328 	mp->mnt_stat.f_owner = kauth_cred_geteuid(p->p_cred);
329 	mp->mnt_unmounter = NULL;
330 	mp->mnt_leaf = mp;
331 
332 	/*
333 	 * The underlying file system may refuse the mount for
334 	 * various reasons.  Allow the user to force it to happen.
335 	 */
336 	mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE;
337  update:
338 	if ((SCARG(uap, flags) & MNT_GETARGS) == 0) {
339 		/*
340 		 * Set the mount level flags.
341 		 */
342 		if (SCARG(uap, flags) & MNT_RDONLY)
343 			mp->mnt_flag |= MNT_RDONLY;
344 		else if (mp->mnt_flag & MNT_RDONLY)
345 			mp->mnt_iflag |= IMNT_WANTRDWR;
346 		mp->mnt_flag &=
347 		  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
348 		    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
349 		    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
350 		mp->mnt_flag |= SCARG(uap, flags) &
351 		   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
352 		    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
353 		    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
354 		    MNT_IGNORE);
355 	}
356 	/*
357 	 * Mount the filesystem.
358 	 */
359 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l);
360 	if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
361 #if defined(COMPAT_30) && defined(NFSSERVER)
362 		if (mp->mnt_flag & MNT_UPDATE && error != 0) {
363 			int error2;
364 
365 			/* Update failed; let's try and see if it was an
366 			 * export request. */
367 			error2 = nfs_update_exports_30(mp, SCARG(uap, path),
368 			    SCARG(uap, data), l);
369 
370 			/* Only update error code if the export request was
371 			 * understood but some problem occurred while
372 			 * processing it. */
373 			if (error2 != EJUSTRETURN)
374 				error = error2;
375 		}
376 #endif
377 		if (mp->mnt_iflag & IMNT_WANTRDWR)
378 			mp->mnt_flag &= ~MNT_RDONLY;
379 		if (error)
380 			mp->mnt_flag = flag;
381 		mp->mnt_flag &=~
382 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
383 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
384 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
385 			if (mp->mnt_syncer == NULL)
386 				error = vfs_allocate_syncvnode(mp);
387 		} else {
388 			if (mp->mnt_syncer != NULL)
389 				vfs_deallocate_syncvnode(mp);
390 		}
391 		vfs_unbusy(mp);
392 		VOP_UNLOCK(vp, 0);
393 		vrele(vp);
394 		return (error);
395 	}
396 	/*
397 	 * Put the new filesystem on the mount list after root.
398 	 */
399 	cache_purge(vp);
400 	if (!error) {
401 		mp->mnt_flag &=~
402 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
403 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
404 		vp->v_mountedhere = mp;
405 		simple_lock(&mountlist_slock);
406 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
407 		simple_unlock(&mountlist_slock);
408 		checkdirs(vp);
409 		VOP_UNLOCK(vp, 0);
410 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
411 			error = vfs_allocate_syncvnode(mp);
412 		vfs_unbusy(mp);
413 		(void) VFS_STATVFS(mp, &mp->mnt_stat, l);
414 		if ((error = VFS_START(mp, 0, l)))
415 			vrele(vp);
416 	} else {
417 		vp->v_mountedhere = (struct mount *)0;
418 		vfs->vfs_refcount--;
419 		vfs_unbusy(mp);
420 		free(mp, M_MOUNT);
421 		vput(vp);
422 	}
423 	return (error);
424 }
425 
426 /*
427  * Scan all active processes to see if any of them have a current
428  * or root directory onto which the new filesystem has just been
429  * mounted. If so, replace them with the new mount point.
430  */
431 void
432 checkdirs(struct vnode *olddp)
433 {
434 	struct cwdinfo *cwdi;
435 	struct vnode *newdp;
436 	struct proc *p;
437 
438 	if (olddp->v_usecount == 1)
439 		return;
440 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
441 		panic("mount: lost mount");
442 	proclist_lock_read();
443 	PROCLIST_FOREACH(p, &allproc) {
444 		cwdi = p->p_cwdi;
445 		if (!cwdi)
446 			continue;
447 		if (cwdi->cwdi_cdir == olddp) {
448 			vrele(cwdi->cwdi_cdir);
449 			VREF(newdp);
450 			cwdi->cwdi_cdir = newdp;
451 		}
452 		if (cwdi->cwdi_rdir == olddp) {
453 			vrele(cwdi->cwdi_rdir);
454 			VREF(newdp);
455 			cwdi->cwdi_rdir = newdp;
456 		}
457 	}
458 	proclist_unlock_read();
459 	if (rootvnode == olddp) {
460 		vrele(rootvnode);
461 		VREF(newdp);
462 		rootvnode = newdp;
463 	}
464 	vput(newdp);
465 }
466 
467 /*
468  * Unmount a file system.
469  *
470  * Note: unmount takes a path to the vnode mounted on as argument,
471  * not special file (as before).
472  */
473 /* ARGSUSED */
474 int
475 sys_unmount(struct lwp *l, void *v, register_t *retval)
476 {
477 	struct sys_unmount_args /* {
478 		syscallarg(const char *) path;
479 		syscallarg(int) flags;
480 	} */ *uap = v;
481 	struct proc *p = l->l_proc;
482 	struct vnode *vp;
483 	struct mount *mp;
484 	int error;
485 	struct nameidata nd;
486 
487 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
488 	    SCARG(uap, path), l);
489 	if ((error = namei(&nd)) != 0)
490 		return (error);
491 	vp = nd.ni_vp;
492 	mp = vp->v_mount;
493 
494 	/*
495 	 * Only root, or the user that did the original mount is
496 	 * permitted to unmount this filesystem.
497 	 */
498 	if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred)) &&
499 	    (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
500 				       &p->p_acflag)) != 0) {
501 		vput(vp);
502 		return (error);
503 	}
504 
505 	/*
506 	 * Don't allow unmounting the root file system.
507 	 */
508 	if (mp->mnt_flag & MNT_ROOTFS) {
509 		vput(vp);
510 		return (EINVAL);
511 	}
512 
513 	/*
514 	 * Must be the root of the filesystem
515 	 */
516 	if ((vp->v_flag & VROOT) == 0) {
517 		vput(vp);
518 		return (EINVAL);
519 	}
520 	vput(vp);
521 
522 	/*
523 	 * XXX Freeze syncer.  Must do this before locking the
524 	 * mount point.  See dounmount() for details.
525 	 */
526 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
527 
528 	if (vfs_busy(mp, 0, 0)) {
529 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
530 		return (EBUSY);
531 	}
532 
533 	return (dounmount(mp, SCARG(uap, flags), l));
534 }
535 
536 /*
537  * Do the actual file system unmount. File system is assumed to have been
538  * marked busy by the caller.
539  */
540 int
541 dounmount(struct mount *mp, int flags, struct lwp *l)
542 {
543 	struct vnode *coveredvp;
544 	int error;
545 	int async;
546 	int used_syncer;
547 
548 	simple_lock(&mountlist_slock);
549 	vfs_unbusy(mp);
550 	used_syncer = (mp->mnt_syncer != NULL);
551 
552 	/*
553 	 * XXX Syncer must be frozen when we get here.  This should really
554 	 * be done on a per-mountpoint basis, but especially the softdep
555 	 * code possibly called from the syncer doesn't exactly work on a
556 	 * per-mountpoint basis, so the softdep code would become a maze
557 	 * of vfs_busy() calls.
558 	 *
559 	 * The caller of dounmount() must acquire syncer_lock because
560 	 * the syncer itself acquires locks in syncer_lock -> vfs_busy
561 	 * order, and we must preserve that order to avoid deadlock.
562 	 *
563 	 * So, if the file system did not use the syncer, now is
564 	 * the time to release the syncer_lock.
565 	 */
566 	if (used_syncer == 0)
567 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
568 
569 	mp->mnt_iflag |= IMNT_UNMOUNT;
570 	mp->mnt_unmounter = l;
571 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
572 	vn_start_write(NULL, &mp, V_WAIT);
573 
574 	async = mp->mnt_flag & MNT_ASYNC;
575 	mp->mnt_flag &= ~MNT_ASYNC;
576 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
577 	if (mp->mnt_syncer != NULL)
578 		vfs_deallocate_syncvnode(mp);
579 	error = 0;
580 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
581 #if NFSS > 0
582 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
583 #endif
584 		if (error == 0)
585 			error = VFS_SYNC(mp, MNT_WAIT, l->l_proc->p_cred, l);
586 	}
587 	if (error == 0 || (flags & MNT_FORCE))
588 		error = VFS_UNMOUNT(mp, flags, l);
589 	vn_finished_write(mp, 0);
590 	simple_lock(&mountlist_slock);
591 	if (error) {
592 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
593 			(void) vfs_allocate_syncvnode(mp);
594 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
595 		mp->mnt_unmounter = NULL;
596 		mp->mnt_flag |= async;
597 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
598 		    &mountlist_slock);
599 		if (used_syncer)
600 			lockmgr(&syncer_lock, LK_RELEASE, NULL);
601 		simple_lock(&mp->mnt_slock);
602 		while (mp->mnt_wcnt > 0) {
603 			wakeup(mp);
604 			ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
605 				0, &mp->mnt_slock);
606 		}
607 		simple_unlock(&mp->mnt_slock);
608 		return (error);
609 	}
610 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
611 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
612 		coveredvp->v_mountedhere = NULL;
613 		vrele(coveredvp);
614 	}
615 	mp->mnt_op->vfs_refcount--;
616 	if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
617 		panic("unmount: dangling vnode");
618 	mp->mnt_iflag |= IMNT_GONE;
619 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
620 	if (used_syncer)
621 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
622 	simple_lock(&mp->mnt_slock);
623 	while (mp->mnt_wcnt > 0) {
624 		wakeup(mp);
625 		ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
626 	}
627 	simple_unlock(&mp->mnt_slock);
628 	vfs_hooks_unmount(mp);
629 	free(mp, M_MOUNT);
630 	return (0);
631 }
632 
633 /*
634  * Sync each mounted filesystem.
635  */
636 #ifdef DEBUG
637 int syncprt = 0;
638 struct ctldebug debug0 = { "syncprt", &syncprt };
639 #endif
640 
641 /* ARGSUSED */
642 int
643 sys_sync(struct lwp *l, void *v, register_t *retval)
644 {
645 	struct mount *mp, *nmp;
646 	int asyncflag;
647 	struct proc *p = l == NULL ? &proc0 : l->l_proc;
648 
649 	simple_lock(&mountlist_slock);
650 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
651 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
652 			nmp = mp->mnt_list.cqe_prev;
653 			continue;
654 		}
655 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
656 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
657 			asyncflag = mp->mnt_flag & MNT_ASYNC;
658 			mp->mnt_flag &= ~MNT_ASYNC;
659 			VFS_SYNC(mp, MNT_NOWAIT, p->p_cred, l);
660 			if (asyncflag)
661 				 mp->mnt_flag |= MNT_ASYNC;
662 			vn_finished_write(mp, 0);
663 		}
664 		simple_lock(&mountlist_slock);
665 		nmp = mp->mnt_list.cqe_prev;
666 		vfs_unbusy(mp);
667 
668 	}
669 	simple_unlock(&mountlist_slock);
670 #ifdef DEBUG
671 	if (syncprt)
672 		vfs_bufstats();
673 #endif /* DEBUG */
674 	return (0);
675 }
676 
677 /*
678  * Change filesystem quotas.
679  */
680 /* ARGSUSED */
681 int
682 sys_quotactl(struct lwp *l, void *v, register_t *retval)
683 {
684 	struct sys_quotactl_args /* {
685 		syscallarg(const char *) path;
686 		syscallarg(int) cmd;
687 		syscallarg(int) uid;
688 		syscallarg(caddr_t) arg;
689 	} */ *uap = v;
690 	struct mount *mp;
691 	int error;
692 	struct nameidata nd;
693 
694 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
695 	if ((error = namei(&nd)) != 0)
696 		return (error);
697 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
698 	vrele(nd.ni_vp);
699 	if (error)
700 		return (error);
701 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
702 	    SCARG(uap, arg), l);
703 	vn_finished_write(mp, 0);
704 	return (error);
705 }
706 
707 int
708 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
709     int root)
710 {
711 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
712 	int error = 0;
713 
714 	/*
715 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
716 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
717 	 * overrides MNT_NOWAIT.
718 	 */
719 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
720 	    (flags != MNT_WAIT && flags != 0)) {
721 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
722 		goto done;
723 	}
724 
725 	/* Get the filesystem stats now */
726 	memset(sp, 0, sizeof(*sp));
727 	if ((error = VFS_STATVFS(mp, sp, l)) != 0) {
728 		return error;
729 	}
730 
731 	if (cwdi->cwdi_rdir == NULL)
732 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
733 done:
734 	if (cwdi->cwdi_rdir != NULL) {
735 		size_t len;
736 		char *bp;
737 		char *path = PNBUF_GET();
738 		if (!path)
739 			return ENOMEM;
740 
741 		bp = path + MAXPATHLEN;
742 		*--bp = '\0';
743 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
744 		    MAXPATHLEN / 2, 0, l);
745 		if (error) {
746 			PNBUF_PUT(path);
747 			return error;
748 		}
749 		len = strlen(bp);
750 		/*
751 		 * for mount points that are below our root, we can see
752 		 * them, so we fix up the pathname and return them. The
753 		 * rest we cannot see, so we don't allow viewing the
754 		 * data.
755 		 */
756 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
757 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
758 			    sizeof(sp->f_mntonname));
759 			if (sp->f_mntonname[0] == '\0')
760 				(void)strlcpy(sp->f_mntonname, "/",
761 				    sizeof(sp->f_mntonname));
762 		} else {
763 			if (root)
764 				(void)strlcpy(sp->f_mntonname, "/",
765 				    sizeof(sp->f_mntonname));
766 			else
767 				error = EPERM;
768 		}
769 		PNBUF_PUT(path);
770 	}
771 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
772 	return error;
773 }
774 
775 /*
776  * Get filesystem statistics.
777  */
778 /* ARGSUSED */
779 int
780 sys_statvfs1(struct lwp *l, void *v, register_t *retval)
781 {
782 	struct sys_statvfs1_args /* {
783 		syscallarg(const char *) path;
784 		syscallarg(struct statvfs *) buf;
785 		syscallarg(int) flags;
786 	} */ *uap = v;
787 	struct mount *mp;
788 	struct statvfs *sb;
789 	int error;
790 	struct nameidata nd;
791 
792 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
793 	if ((error = namei(&nd)) != 0)
794 		return error;
795 	mp = nd.ni_vp->v_mount;
796 	vrele(nd.ni_vp);
797 	sb = STATVFSBUF_GET();
798 	error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1);
799 	if (error == 0) {
800 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
801 	}
802 	STATVFSBUF_PUT(sb);
803 	return error;
804 }
805 
806 /*
807  * Get filesystem statistics.
808  */
809 /* ARGSUSED */
810 int
811 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval)
812 {
813 	struct sys_fstatvfs1_args /* {
814 		syscallarg(int) fd;
815 		syscallarg(struct statvfs *) buf;
816 		syscallarg(int) flags;
817 	} */ *uap = v;
818 	struct proc *p = l->l_proc;
819 	struct file *fp;
820 	struct mount *mp;
821 	struct statvfs *sb;
822 	int error;
823 
824 	/* getvnode() will use the descriptor for us */
825 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
826 		return (error);
827 	mp = ((struct vnode *)fp->f_data)->v_mount;
828 	sb = STATVFSBUF_GET();
829 	if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0)
830 		goto out;
831 	error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
832  out:
833 	FILE_UNUSE(fp, l);
834 	STATVFSBUF_PUT(sb);
835 	return error;
836 }
837 
838 
839 /*
840  * Get statistics on all filesystems.
841  */
842 int
843 sys_getvfsstat(struct lwp *l, void *v, register_t *retval)
844 {
845 	struct sys_getvfsstat_args /* {
846 		syscallarg(struct statvfs *) buf;
847 		syscallarg(size_t) bufsize;
848 		syscallarg(int) flags;
849 	} */ *uap = v;
850 	int root = 0;
851 	struct proc *p = l->l_proc;
852 	struct mount *mp, *nmp;
853 	struct statvfs *sb;
854 	struct statvfs *sfsp;
855 	size_t count, maxcount;
856 	int error = 0;
857 
858 	sb = STATVFSBUF_GET();
859 	maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs);
860 	sfsp = SCARG(uap, buf);
861 	simple_lock(&mountlist_slock);
862 	count = 0;
863 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
864 	     mp = nmp) {
865 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
866 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
867 			continue;
868 		}
869 		if (sfsp && count < maxcount) {
870 			error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0);
871 			if (error) {
872 				simple_lock(&mountlist_slock);
873 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
874 				vfs_unbusy(mp);
875 				continue;
876 			}
877 			error = copyout(sb, sfsp, sizeof(*sfsp));
878 			if (error) {
879 				vfs_unbusy(mp);
880 				goto out;
881 			}
882 			sfsp++;
883 			root |= strcmp(sb->f_mntonname, "/") == 0;
884 		}
885 		count++;
886 		simple_lock(&mountlist_slock);
887 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
888 		vfs_unbusy(mp);
889 	}
890 	simple_unlock(&mountlist_slock);
891 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
892 		/*
893 		 * fake a root entry
894 		 */
895 		if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l,
896 		    SCARG(uap, flags), 1)) != 0)
897 			goto out;
898 		if (sfsp)
899 			error = copyout(sb, sfsp, sizeof(*sfsp));
900 		count++;
901 	}
902 	if (sfsp && count > maxcount)
903 		*retval = maxcount;
904 	else
905 		*retval = count;
906 out:
907 	STATVFSBUF_PUT(sb);
908 	return error;
909 }
910 
911 /*
912  * Change current working directory to a given file descriptor.
913  */
914 /* ARGSUSED */
915 int
916 sys_fchdir(struct lwp *l, void *v, register_t *retval)
917 {
918 	struct sys_fchdir_args /* {
919 		syscallarg(int) fd;
920 	} */ *uap = v;
921 	struct proc *p = l->l_proc;
922 	struct filedesc *fdp = p->p_fd;
923 	struct cwdinfo *cwdi = p->p_cwdi;
924 	struct vnode *vp, *tdp;
925 	struct mount *mp;
926 	struct file *fp;
927 	int error;
928 
929 	/* getvnode() will use the descriptor for us */
930 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
931 		return (error);
932 	vp = (struct vnode *)fp->f_data;
933 
934 	VREF(vp);
935 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
936 	if (vp->v_type != VDIR)
937 		error = ENOTDIR;
938 	else
939 		error = VOP_ACCESS(vp, VEXEC, p->p_cred, l);
940 	while (!error && (mp = vp->v_mountedhere) != NULL) {
941 		if (vfs_busy(mp, 0, 0))
942 			continue;
943 		error = VFS_ROOT(mp, &tdp);
944 		vfs_unbusy(mp);
945 		if (error)
946 			break;
947 		vput(vp);
948 		vp = tdp;
949 	}
950 	if (error) {
951 		vput(vp);
952 		goto out;
953 	}
954 	VOP_UNLOCK(vp, 0);
955 
956 	/*
957 	 * Disallow changing to a directory not under the process's
958 	 * current root directory (if there is one).
959 	 */
960 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
961 		vrele(vp);
962 		error = EPERM;	/* operation not permitted */
963 		goto out;
964 	}
965 
966 	vrele(cwdi->cwdi_cdir);
967 	cwdi->cwdi_cdir = vp;
968  out:
969 	FILE_UNUSE(fp, l);
970 	return (error);
971 }
972 
973 /*
974  * Change this process's notion of the root directory to a given file
975  * descriptor.
976  */
977 int
978 sys_fchroot(struct lwp *l, void *v, register_t *retval)
979 {
980 	struct sys_fchroot_args *uap = v;
981 	struct proc *p = l->l_proc;
982 	struct filedesc *fdp = p->p_fd;
983 	struct cwdinfo *cwdi = p->p_cwdi;
984 	struct vnode	*vp;
985 	struct file	*fp;
986 	int		 error;
987 
988 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
989 				       &p->p_acflag)) != 0)
990 		return error;
991 	/* getvnode() will use the descriptor for us */
992 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
993 		return error;
994 	vp = (struct vnode *) fp->f_data;
995 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
996 	if (vp->v_type != VDIR)
997 		error = ENOTDIR;
998 	else
999 		error = VOP_ACCESS(vp, VEXEC, p->p_cred, l);
1000 	VOP_UNLOCK(vp, 0);
1001 	if (error)
1002 		goto out;
1003 	VREF(vp);
1004 
1005 	/*
1006 	 * Prevent escaping from chroot by putting the root under
1007 	 * the working directory.  Silently chdir to / if we aren't
1008 	 * already there.
1009 	 */
1010 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1011 		/*
1012 		 * XXX would be more failsafe to change directory to a
1013 		 * deadfs node here instead
1014 		 */
1015 		vrele(cwdi->cwdi_cdir);
1016 		VREF(vp);
1017 		cwdi->cwdi_cdir = vp;
1018 	}
1019 
1020 	if (cwdi->cwdi_rdir != NULL)
1021 		vrele(cwdi->cwdi_rdir);
1022 	cwdi->cwdi_rdir = vp;
1023  out:
1024 	FILE_UNUSE(fp, l);
1025 	return (error);
1026 }
1027 
1028 /*
1029  * Change current working directory (``.'').
1030  */
1031 /* ARGSUSED */
1032 int
1033 sys_chdir(struct lwp *l, void *v, register_t *retval)
1034 {
1035 	struct sys_chdir_args /* {
1036 		syscallarg(const char *) path;
1037 	} */ *uap = v;
1038 	struct proc *p = l->l_proc;
1039 	struct cwdinfo *cwdi = p->p_cwdi;
1040 	int error;
1041 	struct nameidata nd;
1042 
1043 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1044 	    SCARG(uap, path), l);
1045 	if ((error = change_dir(&nd, l)) != 0)
1046 		return (error);
1047 	vrele(cwdi->cwdi_cdir);
1048 	cwdi->cwdi_cdir = nd.ni_vp;
1049 	return (0);
1050 }
1051 
1052 /*
1053  * Change notion of root (``/'') directory.
1054  */
1055 /* ARGSUSED */
1056 int
1057 sys_chroot(struct lwp *l, void *v, register_t *retval)
1058 {
1059 	struct sys_chroot_args /* {
1060 		syscallarg(const char *) path;
1061 	} */ *uap = v;
1062 	struct proc *p = l->l_proc;
1063 	struct cwdinfo *cwdi = p->p_cwdi;
1064 	struct vnode *vp;
1065 	int error;
1066 	struct nameidata nd;
1067 
1068 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1069 				       &p->p_acflag)) != 0)
1070 		return (error);
1071 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1072 	    SCARG(uap, path), l);
1073 	if ((error = change_dir(&nd, l)) != 0)
1074 		return (error);
1075 	if (cwdi->cwdi_rdir != NULL)
1076 		vrele(cwdi->cwdi_rdir);
1077 	vp = nd.ni_vp;
1078 	cwdi->cwdi_rdir = vp;
1079 
1080 	/*
1081 	 * Prevent escaping from chroot by putting the root under
1082 	 * the working directory.  Silently chdir to / if we aren't
1083 	 * already there.
1084 	 */
1085 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1086 		/*
1087 		 * XXX would be more failsafe to change directory to a
1088 		 * deadfs node here instead
1089 		 */
1090 		vrele(cwdi->cwdi_cdir);
1091 		VREF(vp);
1092 		cwdi->cwdi_cdir = vp;
1093 	}
1094 
1095 	return (0);
1096 }
1097 
1098 /*
1099  * Common routine for chroot and chdir.
1100  */
1101 static int
1102 change_dir(struct nameidata *ndp, struct lwp *l)
1103 {
1104 	struct vnode *vp;
1105 	int error;
1106 
1107 	if ((error = namei(ndp)) != 0)
1108 		return (error);
1109 	vp = ndp->ni_vp;
1110 	if (vp->v_type != VDIR)
1111 		error = ENOTDIR;
1112 	else
1113 		error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_cred, l);
1114 
1115 	if (error)
1116 		vput(vp);
1117 	else
1118 		VOP_UNLOCK(vp, 0);
1119 	return (error);
1120 }
1121 
1122 /*
1123  * Check permissions, allocate an open file structure,
1124  * and call the device open routine if any.
1125  */
1126 int
1127 sys_open(struct lwp *l, void *v, register_t *retval)
1128 {
1129 	struct sys_open_args /* {
1130 		syscallarg(const char *) path;
1131 		syscallarg(int) flags;
1132 		syscallarg(int) mode;
1133 	} */ *uap = v;
1134 	struct proc *p = l->l_proc;
1135 	struct cwdinfo *cwdi = p->p_cwdi;
1136 	struct filedesc *fdp = p->p_fd;
1137 	struct file *fp;
1138 	struct vnode *vp;
1139 	int flags, cmode;
1140 	int type, indx, error;
1141 	struct flock lf;
1142 	struct nameidata nd;
1143 
1144 	flags = FFLAGS(SCARG(uap, flags));
1145 	if ((flags & (FREAD | FWRITE)) == 0)
1146 		return (EINVAL);
1147 	/* falloc() will use the file descriptor for us */
1148 	if ((error = falloc(p, &fp, &indx)) != 0)
1149 		return (error);
1150 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1151 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
1152 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1153 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1154 		FILE_UNUSE(fp, l);
1155 		fdp->fd_ofiles[indx] = NULL;
1156 		ffree(fp);
1157 		if ((error == EDUPFD || error == EMOVEFD) &&
1158 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1159 		    (error =
1160 			dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) {
1161 			*retval = indx;
1162 			return (0);
1163 		}
1164 		if (error == ERESTART)
1165 			error = EINTR;
1166 		fdremove(fdp, indx);
1167 		return (error);
1168 	}
1169 	l->l_dupfd = 0;
1170 	vp = nd.ni_vp;
1171 	fp->f_flag = flags & FMASK;
1172 	fp->f_type = DTYPE_VNODE;
1173 	fp->f_ops = &vnops;
1174 	fp->f_data = vp;
1175 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1176 		lf.l_whence = SEEK_SET;
1177 		lf.l_start = 0;
1178 		lf.l_len = 0;
1179 		if (flags & O_EXLOCK)
1180 			lf.l_type = F_WRLCK;
1181 		else
1182 			lf.l_type = F_RDLCK;
1183 		type = F_FLOCK;
1184 		if ((flags & FNONBLOCK) == 0)
1185 			type |= F_WAIT;
1186 		VOP_UNLOCK(vp, 0);
1187 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1188 		if (error) {
1189 			(void) vn_close(vp, fp->f_flag, fp->f_cred, l);
1190 			FILE_UNUSE(fp, l);
1191 			ffree(fp);
1192 			fdremove(fdp, indx);
1193 			return (error);
1194 		}
1195 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1196 		fp->f_flag |= FHASLOCK;
1197 	}
1198 	VOP_UNLOCK(vp, 0);
1199 	*retval = indx;
1200 	FILE_SET_MATURE(fp);
1201 	FILE_UNUSE(fp, l);
1202 	return (0);
1203 }
1204 
1205 /*
1206  * vfs_composefh: compose a filehandle.
1207  */
1208 
1209 int
1210 vfs_composefh(struct vnode *vp, fhandle_t *fhp)
1211 {
1212 	struct mount *mp;
1213 	int error;
1214 
1215 	mp = vp->v_mount;
1216 	if (mp->mnt_op->vfs_vptofh == NULL) {
1217 		return EOPNOTSUPP;
1218 	}
1219 	memset(fhp, 0, sizeof(*fhp));
1220 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1221 	error = VFS_VPTOFH(vp, &fhp->fh_fid);
1222 	KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
1223 	return error;
1224 }
1225 
1226 /*
1227  * Get file handle system call
1228  */
1229 int
1230 sys_getfh(struct lwp *l, void *v, register_t *retval)
1231 {
1232 	struct sys_getfh_args /* {
1233 		syscallarg(char *) fname;
1234 		syscallarg(fhandle_t *) fhp;
1235 	} */ *uap = v;
1236 	struct proc *p = l->l_proc;
1237 	struct vnode *vp;
1238 	fhandle_t fh;
1239 	int error;
1240 	struct nameidata nd;
1241 
1242 	/*
1243 	 * Must be super user
1244 	 */
1245 	error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1246 				  &p->p_acflag);
1247 	if (error)
1248 		return (error);
1249 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1250 	    SCARG(uap, fname), l);
1251 	error = namei(&nd);
1252 	if (error)
1253 		return (error);
1254 	vp = nd.ni_vp;
1255 	error = vfs_composefh(vp, &fh);
1256 	vput(vp);
1257 	if (error)
1258 		return (error);
1259 	error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1260 	return (error);
1261 }
1262 
1263 /*
1264  * Open a file given a file handle.
1265  *
1266  * Check permissions, allocate an open file structure,
1267  * and call the device open routine if any.
1268  */
1269 int
1270 sys_fhopen(struct lwp *l, void *v, register_t *retval)
1271 {
1272 	struct sys_fhopen_args /* {
1273 		syscallarg(const fhandle_t *) fhp;
1274 		syscallarg(int) flags;
1275 	} */ *uap = v;
1276 	struct proc *p = l->l_proc;
1277 	struct filedesc *fdp = p->p_fd;
1278 	struct file *fp;
1279 	struct vnode *vp = NULL;
1280 	struct mount *mp;
1281 	kauth_cred_t cred = p->p_cred;
1282 	int flags;
1283 	struct file *nfp;
1284 	int type, indx, error=0;
1285 	struct flock lf;
1286 	struct vattr va;
1287 	fhandle_t fh;
1288 
1289 	/*
1290 	 * Must be super user
1291 	 */
1292 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1293 				       &p->p_acflag)))
1294 		return (error);
1295 
1296 	flags = FFLAGS(SCARG(uap, flags));
1297 	if ((flags & (FREAD | FWRITE)) == 0)
1298 		return (EINVAL);
1299 	if ((flags & O_CREAT))
1300 		return (EINVAL);
1301 	/* falloc() will use the file descriptor for us */
1302 	if ((error = falloc(p, &nfp, &indx)) != 0)
1303 		return (error);
1304 	fp = nfp;
1305 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1306 		goto bad;
1307 
1308 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1309 		error = ESTALE;
1310 		goto bad;
1311 	}
1312 
1313 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1314 		error = EOPNOTSUPP;
1315 		goto bad;
1316 	}
1317 
1318 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1319 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1320 		goto bad;
1321 	}
1322 
1323 	/* Now do an effective vn_open */
1324 
1325 	if (vp->v_type == VSOCK) {
1326 		error = EOPNOTSUPP;
1327 		goto bad;
1328 	}
1329 	if (flags & FREAD) {
1330 		if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0)
1331 			goto bad;
1332 	}
1333 	if (flags & (FWRITE | O_TRUNC)) {
1334 		if (vp->v_type == VDIR) {
1335 			error = EISDIR;
1336 			goto bad;
1337 		}
1338 		if ((error = vn_writechk(vp)) != 0 ||
1339 		    (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0)
1340 			goto bad;
1341 	}
1342 	if (flags & O_TRUNC) {
1343 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
1344 			goto bad;
1345 		VOP_UNLOCK(vp, 0);			/* XXX */
1346 		VOP_LEASE(vp, l, cred, LEASE_WRITE);
1347 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1348 		VATTR_NULL(&va);
1349 		va.va_size = 0;
1350 		error = VOP_SETATTR(vp, &va, cred, l);
1351 		vn_finished_write(mp, 0);
1352 		if (error)
1353 			goto bad;
1354 	}
1355 	if ((error = VOP_OPEN(vp, flags, cred, l)) != 0)
1356 		goto bad;
1357 	if (vp->v_type == VREG &&
1358 	    uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1359 		error = EIO;
1360 		goto bad;
1361 	}
1362 	if (flags & FWRITE)
1363 		vp->v_writecount++;
1364 
1365 	/* done with modified vn_open, now finish what sys_open does. */
1366 
1367 	fp->f_flag = flags & FMASK;
1368 	fp->f_type = DTYPE_VNODE;
1369 	fp->f_ops = &vnops;
1370 	fp->f_data = vp;
1371 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1372 		lf.l_whence = SEEK_SET;
1373 		lf.l_start = 0;
1374 		lf.l_len = 0;
1375 		if (flags & O_EXLOCK)
1376 			lf.l_type = F_WRLCK;
1377 		else
1378 			lf.l_type = F_RDLCK;
1379 		type = F_FLOCK;
1380 		if ((flags & FNONBLOCK) == 0)
1381 			type |= F_WAIT;
1382 		VOP_UNLOCK(vp, 0);
1383 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1384 		if (error) {
1385 			(void) vn_close(vp, fp->f_flag, fp->f_cred, l);
1386 			FILE_UNUSE(fp, l);
1387 			ffree(fp);
1388 			fdremove(fdp, indx);
1389 			return (error);
1390 		}
1391 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1392 		fp->f_flag |= FHASLOCK;
1393 	}
1394 	VOP_UNLOCK(vp, 0);
1395 	*retval = indx;
1396 	FILE_SET_MATURE(fp);
1397 	FILE_UNUSE(fp, l);
1398 	return (0);
1399 
1400 bad:
1401 	FILE_UNUSE(fp, l);
1402 	ffree(fp);
1403 	fdremove(fdp, indx);
1404 	if (vp != NULL)
1405 		vput(vp);
1406 	return (error);
1407 }
1408 
1409 /* ARGSUSED */
1410 int
1411 sys___fhstat30(struct lwp *l, void *v, register_t *retval)
1412 {
1413 	struct sys___fhstat30_args /* {
1414 		syscallarg(const fhandle_t *) fhp;
1415 		syscallarg(struct stat *) sb;
1416 	} */ *uap = v;
1417 	struct proc *p = l->l_proc;
1418 	struct stat sb;
1419 	int error;
1420 	fhandle_t fh;
1421 	struct mount *mp;
1422 	struct vnode *vp;
1423 
1424 	/*
1425 	 * Must be super user
1426 	 */
1427 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1428 				       &p->p_acflag)))
1429 		return (error);
1430 
1431 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1432 		return (error);
1433 
1434 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1435 		return (ESTALE);
1436 	if (mp->mnt_op->vfs_fhtovp == NULL)
1437 		return EOPNOTSUPP;
1438 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1439 		return (error);
1440 	error = vn_stat(vp, &sb, l);
1441 	vput(vp);
1442 	if (error)
1443 		return (error);
1444 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1445 	return (error);
1446 }
1447 
1448 /* ARGSUSED */
1449 int
1450 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval)
1451 {
1452 	struct sys_fhstatvfs1_args /* {
1453 		syscallarg(const fhandle_t *) fhp;
1454 		syscallarg(struct statvfs *) buf;
1455 		syscallarg(int)	flags;
1456 	} */ *uap = v;
1457 	struct proc *p = l->l_proc;
1458 	struct statvfs *sb;
1459 	fhandle_t fh;
1460 	struct mount *mp;
1461 	struct vnode *vp;
1462 	int error;
1463 
1464 	/*
1465 	 * Must be super user
1466 	 */
1467 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1468 				       &p->p_acflag)) != 0)
1469 		return error;
1470 
1471 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1472 		return error;
1473 
1474 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1475 		return ESTALE;
1476 	if (mp->mnt_op->vfs_fhtovp == NULL)
1477 		return EOPNOTSUPP;
1478 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1479 		return error;
1480 
1481 	mp = vp->v_mount;
1482 	sb = STATVFSBUF_GET();
1483 	if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) {
1484 		vput(vp);
1485 		goto out;
1486 	}
1487 	vput(vp);
1488 	error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1489 out:
1490 	STATVFSBUF_PUT(sb);
1491 	return error;
1492 }
1493 
1494 /*
1495  * Create a special file.
1496  */
1497 /* ARGSUSED */
1498 int
1499 sys_mknod(struct lwp *l, void *v, register_t *retval)
1500 {
1501 	struct sys_mknod_args /* {
1502 		syscallarg(const char *) path;
1503 		syscallarg(int) mode;
1504 		syscallarg(int) dev;
1505 	} */ *uap = v;
1506 	struct proc *p = l->l_proc;
1507 	struct vnode *vp;
1508 	struct mount *mp;
1509 	struct vattr vattr;
1510 	int error;
1511 	int whiteout = 0;
1512 	struct nameidata nd;
1513 
1514 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1515 				       &p->p_acflag)) != 0)
1516 		return (error);
1517 restart:
1518 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l);
1519 	if ((error = namei(&nd)) != 0)
1520 		return (error);
1521 	vp = nd.ni_vp;
1522 	if (vp != NULL)
1523 		error = EEXIST;
1524 	else {
1525 		VATTR_NULL(&vattr);
1526 		vattr.va_mode =
1527 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1528 		vattr.va_rdev = SCARG(uap, dev);
1529 		whiteout = 0;
1530 
1531 		switch (SCARG(uap, mode) & S_IFMT) {
1532 		case S_IFMT:	/* used by badsect to flag bad sectors */
1533 			vattr.va_type = VBAD;
1534 			break;
1535 		case S_IFCHR:
1536 			vattr.va_type = VCHR;
1537 			break;
1538 		case S_IFBLK:
1539 			vattr.va_type = VBLK;
1540 			break;
1541 		case S_IFWHT:
1542 			whiteout = 1;
1543 			break;
1544 		default:
1545 			error = EINVAL;
1546 			break;
1547 		}
1548 	}
1549 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1550 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1551 		if (nd.ni_dvp == vp)
1552 			vrele(nd.ni_dvp);
1553 		else
1554 			vput(nd.ni_dvp);
1555 		if (vp)
1556 			vrele(vp);
1557 		if ((error = vn_start_write(NULL, &mp,
1558 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1559 			return (error);
1560 		goto restart;
1561 	}
1562 	if (!error) {
1563 		VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
1564 		if (whiteout) {
1565 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1566 			if (error)
1567 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1568 			vput(nd.ni_dvp);
1569 		} else {
1570 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1571 						&nd.ni_cnd, &vattr);
1572 			if (error == 0)
1573 				vput(nd.ni_vp);
1574 		}
1575 	} else {
1576 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1577 		if (nd.ni_dvp == vp)
1578 			vrele(nd.ni_dvp);
1579 		else
1580 			vput(nd.ni_dvp);
1581 		if (vp)
1582 			vrele(vp);
1583 	}
1584 	vn_finished_write(mp, 0);
1585 	return (error);
1586 }
1587 
1588 /*
1589  * Create a named pipe.
1590  */
1591 /* ARGSUSED */
1592 int
1593 sys_mkfifo(struct lwp *l, void *v, register_t *retval)
1594 {
1595 	struct sys_mkfifo_args /* {
1596 		syscallarg(const char *) path;
1597 		syscallarg(int) mode;
1598 	} */ *uap = v;
1599 	struct proc *p = l->l_proc;
1600 	struct mount *mp;
1601 	struct vattr vattr;
1602 	int error;
1603 	struct nameidata nd;
1604 
1605 restart:
1606 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l);
1607 	if ((error = namei(&nd)) != 0)
1608 		return (error);
1609 	if (nd.ni_vp != NULL) {
1610 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1611 		if (nd.ni_dvp == nd.ni_vp)
1612 			vrele(nd.ni_dvp);
1613 		else
1614 			vput(nd.ni_dvp);
1615 		vrele(nd.ni_vp);
1616 		return (EEXIST);
1617 	}
1618 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1619 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1620 		if (nd.ni_dvp == nd.ni_vp)
1621 			vrele(nd.ni_dvp);
1622 		else
1623 			vput(nd.ni_dvp);
1624 		if (nd.ni_vp)
1625 			vrele(nd.ni_vp);
1626 		if ((error = vn_start_write(NULL, &mp,
1627 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1628 			return (error);
1629 		goto restart;
1630 	}
1631 	VATTR_NULL(&vattr);
1632 	vattr.va_type = VFIFO;
1633 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1634 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
1635 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1636 	if (error == 0)
1637 		vput(nd.ni_vp);
1638 	vn_finished_write(mp, 0);
1639 	return (error);
1640 }
1641 
1642 /*
1643  * Make a hard file link.
1644  */
1645 /* ARGSUSED */
1646 int
1647 sys_link(struct lwp *l, void *v, register_t *retval)
1648 {
1649 	struct sys_link_args /* {
1650 		syscallarg(const char *) path;
1651 		syscallarg(const char *) link;
1652 	} */ *uap = v;
1653 	struct proc *p = l->l_proc;
1654 	struct vnode *vp;
1655 	struct mount *mp;
1656 	struct nameidata nd;
1657 	int error;
1658 
1659 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
1660 	if ((error = namei(&nd)) != 0)
1661 		return (error);
1662 	vp = nd.ni_vp;
1663 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
1664 		vrele(vp);
1665 		return (error);
1666 	}
1667 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l);
1668 	if ((error = namei(&nd)) != 0)
1669 		goto out;
1670 	if (nd.ni_vp) {
1671 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1672 		if (nd.ni_dvp == nd.ni_vp)
1673 			vrele(nd.ni_dvp);
1674 		else
1675 			vput(nd.ni_dvp);
1676 		vrele(nd.ni_vp);
1677 		error = EEXIST;
1678 		goto out;
1679 	}
1680 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
1681 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
1682 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1683 out:
1684 	vrele(vp);
1685 	vn_finished_write(mp, 0);
1686 	return (error);
1687 }
1688 
1689 /*
1690  * Make a symbolic link.
1691  */
1692 /* ARGSUSED */
1693 int
1694 sys_symlink(struct lwp *l, void *v, register_t *retval)
1695 {
1696 	struct sys_symlink_args /* {
1697 		syscallarg(const char *) path;
1698 		syscallarg(const char *) link;
1699 	} */ *uap = v;
1700 	struct proc *p = l->l_proc;
1701 	struct mount *mp;
1702 	struct vattr vattr;
1703 	char *path;
1704 	int error;
1705 	struct nameidata nd;
1706 
1707 	path = PNBUF_GET();
1708 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1709 	if (error)
1710 		goto out;
1711 restart:
1712 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l);
1713 	if ((error = namei(&nd)) != 0)
1714 		goto out;
1715 	if (nd.ni_vp) {
1716 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1717 		if (nd.ni_dvp == nd.ni_vp)
1718 			vrele(nd.ni_dvp);
1719 		else
1720 			vput(nd.ni_dvp);
1721 		vrele(nd.ni_vp);
1722 		error = EEXIST;
1723 		goto out;
1724 	}
1725 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1726 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1727 		if (nd.ni_dvp == nd.ni_vp)
1728 			vrele(nd.ni_dvp);
1729 		else
1730 			vput(nd.ni_dvp);
1731 		if ((error = vn_start_write(NULL, &mp,
1732 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1733 			return (error);
1734 		goto restart;
1735 	}
1736 	VATTR_NULL(&vattr);
1737 	vattr.va_type = VLNK;
1738 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1739 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
1740 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1741 	if (error == 0)
1742 		vput(nd.ni_vp);
1743 	vn_finished_write(mp, 0);
1744 out:
1745 	PNBUF_PUT(path);
1746 	return (error);
1747 }
1748 
1749 /*
1750  * Delete a whiteout from the filesystem.
1751  */
1752 /* ARGSUSED */
1753 int
1754 sys_undelete(struct lwp *l, void *v, register_t *retval)
1755 {
1756 	struct sys_undelete_args /* {
1757 		syscallarg(const char *) path;
1758 	} */ *uap = v;
1759 	struct proc *p = l->l_proc;
1760 	int error;
1761 	struct mount *mp;
1762 	struct nameidata nd;
1763 
1764 restart:
1765 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1766 	    SCARG(uap, path), l);
1767 	error = namei(&nd);
1768 	if (error)
1769 		return (error);
1770 
1771 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1772 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1773 		if (nd.ni_dvp == nd.ni_vp)
1774 			vrele(nd.ni_dvp);
1775 		else
1776 			vput(nd.ni_dvp);
1777 		if (nd.ni_vp)
1778 			vrele(nd.ni_vp);
1779 		return (EEXIST);
1780 	}
1781 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1782 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1783 		if (nd.ni_dvp == nd.ni_vp)
1784 			vrele(nd.ni_dvp);
1785 		else
1786 			vput(nd.ni_dvp);
1787 		if ((error = vn_start_write(NULL, &mp,
1788 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1789 			return (error);
1790 		goto restart;
1791 	}
1792 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
1793 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1794 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1795 	vput(nd.ni_dvp);
1796 	vn_finished_write(mp, 0);
1797 	return (error);
1798 }
1799 
1800 /*
1801  * Delete a name from the filesystem.
1802  */
1803 /* ARGSUSED */
1804 int
1805 sys_unlink(struct lwp *l, void *v, register_t *retval)
1806 {
1807 	struct sys_unlink_args /* {
1808 		syscallarg(const char *) path;
1809 	} */ *uap = v;
1810 	struct proc *p = l->l_proc;
1811 	struct mount *mp;
1812 	struct vnode *vp;
1813 	int error;
1814 	struct nameidata nd;
1815 
1816 restart:
1817 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1818 	    SCARG(uap, path), l);
1819 	if ((error = namei(&nd)) != 0)
1820 		return (error);
1821 	vp = nd.ni_vp;
1822 
1823 	/*
1824 	 * The root of a mounted filesystem cannot be deleted.
1825 	 */
1826 	if (vp->v_flag & VROOT) {
1827 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1828 		if (nd.ni_dvp == vp)
1829 			vrele(nd.ni_dvp);
1830 		else
1831 			vput(nd.ni_dvp);
1832 		vput(vp);
1833 		error = EBUSY;
1834 		goto out;
1835 	}
1836 
1837 #ifdef VERIFIED_EXEC
1838 	/* Handle remove requests for veriexec entries. */
1839 	if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) {
1840 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1841 		if (nd.ni_dvp == vp)
1842 			vrele(nd.ni_dvp);
1843 		else
1844 			vput(nd.ni_dvp);
1845 		vput(vp);
1846 		goto out;
1847 	}
1848 #endif /* VERIFIED_EXEC */
1849 
1850 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1851 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1852 		if (nd.ni_dvp == vp)
1853 			vrele(nd.ni_dvp);
1854 		else
1855 			vput(nd.ni_dvp);
1856 		vput(vp);
1857 		if ((error = vn_start_write(NULL, &mp,
1858 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1859 			return (error);
1860 		goto restart;
1861 	}
1862 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
1863 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
1864 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1865 	vn_finished_write(mp, 0);
1866 out:
1867 	return (error);
1868 }
1869 
1870 /*
1871  * Reposition read/write file offset.
1872  */
1873 int
1874 sys_lseek(struct lwp *l, void *v, register_t *retval)
1875 {
1876 	struct sys_lseek_args /* {
1877 		syscallarg(int) fd;
1878 		syscallarg(int) pad;
1879 		syscallarg(off_t) offset;
1880 		syscallarg(int) whence;
1881 	} */ *uap = v;
1882 	struct proc *p = l->l_proc;
1883 	kauth_cred_t cred = p->p_cred;
1884 	struct filedesc *fdp = p->p_fd;
1885 	struct file *fp;
1886 	struct vnode *vp;
1887 	struct vattr vattr;
1888 	off_t newoff;
1889 	int error;
1890 
1891 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1892 		return (EBADF);
1893 
1894 	FILE_USE(fp);
1895 
1896 	vp = (struct vnode *)fp->f_data;
1897 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1898 		error = ESPIPE;
1899 		goto out;
1900 	}
1901 
1902 	switch (SCARG(uap, whence)) {
1903 	case SEEK_CUR:
1904 		newoff = fp->f_offset + SCARG(uap, offset);
1905 		break;
1906 	case SEEK_END:
1907 		error = VOP_GETATTR(vp, &vattr, cred, l);
1908 		if (error)
1909 			goto out;
1910 		newoff = SCARG(uap, offset) + vattr.va_size;
1911 		break;
1912 	case SEEK_SET:
1913 		newoff = SCARG(uap, offset);
1914 		break;
1915 	default:
1916 		error = EINVAL;
1917 		goto out;
1918 	}
1919 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1920 		goto out;
1921 
1922 	*(off_t *)retval = fp->f_offset = newoff;
1923  out:
1924 	FILE_UNUSE(fp, l);
1925 	return (error);
1926 }
1927 
1928 /*
1929  * Positional read system call.
1930  */
1931 int
1932 sys_pread(struct lwp *l, void *v, register_t *retval)
1933 {
1934 	struct sys_pread_args /* {
1935 		syscallarg(int) fd;
1936 		syscallarg(void *) buf;
1937 		syscallarg(size_t) nbyte;
1938 		syscallarg(off_t) offset;
1939 	} */ *uap = v;
1940 	struct proc *p = l->l_proc;
1941 	struct filedesc *fdp = p->p_fd;
1942 	struct file *fp;
1943 	struct vnode *vp;
1944 	off_t offset;
1945 	int error, fd = SCARG(uap, fd);
1946 
1947 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1948 		return (EBADF);
1949 
1950 	if ((fp->f_flag & FREAD) == 0) {
1951 		simple_unlock(&fp->f_slock);
1952 		return (EBADF);
1953 	}
1954 
1955 	FILE_USE(fp);
1956 
1957 	vp = (struct vnode *)fp->f_data;
1958 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1959 		error = ESPIPE;
1960 		goto out;
1961 	}
1962 
1963 	offset = SCARG(uap, offset);
1964 
1965 	/*
1966 	 * XXX This works because no file systems actually
1967 	 * XXX take any action on the seek operation.
1968 	 */
1969 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1970 		goto out;
1971 
1972 	/* dofileread() will unuse the descriptor for us */
1973 	return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1974 	    &offset, 0, retval));
1975 
1976  out:
1977 	FILE_UNUSE(fp, l);
1978 	return (error);
1979 }
1980 
1981 /*
1982  * Positional scatter read system call.
1983  */
1984 int
1985 sys_preadv(struct lwp *l, void *v, register_t *retval)
1986 {
1987 	struct sys_preadv_args /* {
1988 		syscallarg(int) fd;
1989 		syscallarg(const struct iovec *) iovp;
1990 		syscallarg(int) iovcnt;
1991 		syscallarg(off_t) offset;
1992 	} */ *uap = v;
1993 	struct proc *p = l->l_proc;
1994 	struct filedesc *fdp = p->p_fd;
1995 	struct file *fp;
1996 	struct vnode *vp;
1997 	off_t offset;
1998 	int error, fd = SCARG(uap, fd);
1999 
2000 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2001 		return (EBADF);
2002 
2003 	if ((fp->f_flag & FREAD) == 0) {
2004 		simple_unlock(&fp->f_slock);
2005 		return (EBADF);
2006 	}
2007 
2008 	FILE_USE(fp);
2009 
2010 	vp = (struct vnode *)fp->f_data;
2011 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2012 		error = ESPIPE;
2013 		goto out;
2014 	}
2015 
2016 	offset = SCARG(uap, offset);
2017 
2018 	/*
2019 	 * XXX This works because no file systems actually
2020 	 * XXX take any action on the seek operation.
2021 	 */
2022 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2023 		goto out;
2024 
2025 	/* dofilereadv() will unuse the descriptor for us */
2026 	return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2027 	    &offset, 0, retval));
2028 
2029  out:
2030 	FILE_UNUSE(fp, l);
2031 	return (error);
2032 }
2033 
2034 /*
2035  * Positional write system call.
2036  */
2037 int
2038 sys_pwrite(struct lwp *l, void *v, register_t *retval)
2039 {
2040 	struct sys_pwrite_args /* {
2041 		syscallarg(int) fd;
2042 		syscallarg(const void *) buf;
2043 		syscallarg(size_t) nbyte;
2044 		syscallarg(off_t) offset;
2045 	} */ *uap = v;
2046 	struct proc *p = l->l_proc;
2047 	struct filedesc *fdp = p->p_fd;
2048 	struct file *fp;
2049 	struct vnode *vp;
2050 	off_t offset;
2051 	int error, fd = SCARG(uap, fd);
2052 
2053 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2054 		return (EBADF);
2055 
2056 	if ((fp->f_flag & FWRITE) == 0) {
2057 		simple_unlock(&fp->f_slock);
2058 		return (EBADF);
2059 	}
2060 
2061 	FILE_USE(fp);
2062 
2063 	vp = (struct vnode *)fp->f_data;
2064 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2065 		error = ESPIPE;
2066 		goto out;
2067 	}
2068 
2069 	offset = SCARG(uap, offset);
2070 
2071 	/*
2072 	 * XXX This works because no file systems actually
2073 	 * XXX take any action on the seek operation.
2074 	 */
2075 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2076 		goto out;
2077 
2078 	/* dofilewrite() will unuse the descriptor for us */
2079 	return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2080 	    &offset, 0, retval));
2081 
2082  out:
2083 	FILE_UNUSE(fp, l);
2084 	return (error);
2085 }
2086 
2087 /*
2088  * Positional gather write system call.
2089  */
2090 int
2091 sys_pwritev(struct lwp *l, void *v, register_t *retval)
2092 {
2093 	struct sys_pwritev_args /* {
2094 		syscallarg(int) fd;
2095 		syscallarg(const struct iovec *) iovp;
2096 		syscallarg(int) iovcnt;
2097 		syscallarg(off_t) offset;
2098 	} */ *uap = v;
2099 	struct proc *p = l->l_proc;
2100 	struct filedesc *fdp = p->p_fd;
2101 	struct file *fp;
2102 	struct vnode *vp;
2103 	off_t offset;
2104 	int error, fd = SCARG(uap, fd);
2105 
2106 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2107 		return (EBADF);
2108 
2109 	if ((fp->f_flag & FWRITE) == 0) {
2110 		simple_unlock(&fp->f_slock);
2111 		return (EBADF);
2112 	}
2113 
2114 	FILE_USE(fp);
2115 
2116 	vp = (struct vnode *)fp->f_data;
2117 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2118 		error = ESPIPE;
2119 		goto out;
2120 	}
2121 
2122 	offset = SCARG(uap, offset);
2123 
2124 	/*
2125 	 * XXX This works because no file systems actually
2126 	 * XXX take any action on the seek operation.
2127 	 */
2128 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2129 		goto out;
2130 
2131 	/* dofilewritev() will unuse the descriptor for us */
2132 	return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2133 	    &offset, 0, retval));
2134 
2135  out:
2136 	FILE_UNUSE(fp, l);
2137 	return (error);
2138 }
2139 
2140 /*
2141  * Check access permissions.
2142  */
2143 int
2144 sys_access(struct lwp *l, void *v, register_t *retval)
2145 {
2146 	struct sys_access_args /* {
2147 		syscallarg(const char *) path;
2148 		syscallarg(int) flags;
2149 	} */ *uap = v;
2150 	struct proc *p = l->l_proc;
2151 	kauth_cred_t cred;
2152 	struct vnode *vp;
2153 	int error, flags;
2154 	struct nameidata nd;
2155 
2156 	cred = kauth_cred_dup(p->p_cred);
2157 	kauth_cred_seteuid(cred, kauth_cred_getuid(p->p_cred));
2158 	kauth_cred_setegid(cred, kauth_cred_getgid(p->p_cred));
2159 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2160 	    SCARG(uap, path), l);
2161 	/* Override default credentials */
2162 	nd.ni_cnd.cn_cred = cred;
2163 	if ((error = namei(&nd)) != 0)
2164 		goto out;
2165 	vp = nd.ni_vp;
2166 
2167 	/* Flags == 0 means only check for existence. */
2168 	if (SCARG(uap, flags)) {
2169 		flags = 0;
2170 		if (SCARG(uap, flags) & R_OK)
2171 			flags |= VREAD;
2172 		if (SCARG(uap, flags) & W_OK)
2173 			flags |= VWRITE;
2174 		if (SCARG(uap, flags) & X_OK)
2175 			flags |= VEXEC;
2176 
2177 		error = VOP_ACCESS(vp, flags, cred, l);
2178 		if (!error && (flags & VWRITE))
2179 			error = vn_writechk(vp);
2180 	}
2181 	vput(vp);
2182 out:
2183 	kauth_cred_free(cred);
2184 	return (error);
2185 }
2186 
2187 /*
2188  * Get file status; this version follows links.
2189  */
2190 /* ARGSUSED */
2191 int
2192 sys___stat30(struct lwp *l, void *v, register_t *retval)
2193 {
2194 	struct sys___stat30_args /* {
2195 		syscallarg(const char *) path;
2196 		syscallarg(struct stat *) ub;
2197 	} */ *uap = v;
2198 	struct stat sb;
2199 	int error;
2200 	struct nameidata nd;
2201 
2202 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2203 	    SCARG(uap, path), l);
2204 	if ((error = namei(&nd)) != 0)
2205 		return (error);
2206 	error = vn_stat(nd.ni_vp, &sb, l);
2207 	vput(nd.ni_vp);
2208 	if (error)
2209 		return (error);
2210 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2211 	return (error);
2212 }
2213 
2214 /*
2215  * Get file status; this version does not follow links.
2216  */
2217 /* ARGSUSED */
2218 int
2219 sys___lstat30(struct lwp *l, void *v, register_t *retval)
2220 {
2221 	struct sys___lstat30_args /* {
2222 		syscallarg(const char *) path;
2223 		syscallarg(struct stat *) ub;
2224 	} */ *uap = v;
2225 	struct stat sb;
2226 	int error;
2227 	struct nameidata nd;
2228 
2229 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2230 	    SCARG(uap, path), l);
2231 	if ((error = namei(&nd)) != 0)
2232 		return (error);
2233 	error = vn_stat(nd.ni_vp, &sb, l);
2234 	vput(nd.ni_vp);
2235 	if (error)
2236 		return (error);
2237 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2238 	return (error);
2239 }
2240 
2241 /*
2242  * Get configurable pathname variables.
2243  */
2244 /* ARGSUSED */
2245 int
2246 sys_pathconf(struct lwp *l, void *v, register_t *retval)
2247 {
2248 	struct sys_pathconf_args /* {
2249 		syscallarg(const char *) path;
2250 		syscallarg(int) name;
2251 	} */ *uap = v;
2252 	int error;
2253 	struct nameidata nd;
2254 
2255 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2256 	    SCARG(uap, path), l);
2257 	if ((error = namei(&nd)) != 0)
2258 		return (error);
2259 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2260 	vput(nd.ni_vp);
2261 	return (error);
2262 }
2263 
2264 /*
2265  * Return target name of a symbolic link.
2266  */
2267 /* ARGSUSED */
2268 int
2269 sys_readlink(struct lwp *l, void *v, register_t *retval)
2270 {
2271 	struct sys_readlink_args /* {
2272 		syscallarg(const char *) path;
2273 		syscallarg(char *) buf;
2274 		syscallarg(size_t) count;
2275 	} */ *uap = v;
2276 	struct proc *p = l->l_proc;
2277 	struct vnode *vp;
2278 	struct iovec aiov;
2279 	struct uio auio;
2280 	int error;
2281 	struct nameidata nd;
2282 
2283 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2284 	    SCARG(uap, path), l);
2285 	if ((error = namei(&nd)) != 0)
2286 		return (error);
2287 	vp = nd.ni_vp;
2288 	if (vp->v_type != VLNK)
2289 		error = EINVAL;
2290 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2291 	    (error = VOP_ACCESS(vp, VREAD, p->p_cred, l)) == 0) {
2292 		aiov.iov_base = SCARG(uap, buf);
2293 		aiov.iov_len = SCARG(uap, count);
2294 		auio.uio_iov = &aiov;
2295 		auio.uio_iovcnt = 1;
2296 		auio.uio_offset = 0;
2297 		auio.uio_rw = UIO_READ;
2298 		KASSERT(l == curlwp);
2299 		auio.uio_vmspace = l->l_proc->p_vmspace;
2300 		auio.uio_resid = SCARG(uap, count);
2301 		error = VOP_READLINK(vp, &auio, p->p_cred);
2302 	}
2303 	vput(vp);
2304 	*retval = SCARG(uap, count) - auio.uio_resid;
2305 	return (error);
2306 }
2307 
2308 /*
2309  * Change flags of a file given a path name.
2310  */
2311 /* ARGSUSED */
2312 int
2313 sys_chflags(struct lwp *l, void *v, register_t *retval)
2314 {
2315 	struct sys_chflags_args /* {
2316 		syscallarg(const char *) path;
2317 		syscallarg(u_long) flags;
2318 	} */ *uap = v;
2319 	struct vnode *vp;
2320 	int error;
2321 	struct nameidata nd;
2322 
2323 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2324 	if ((error = namei(&nd)) != 0)
2325 		return (error);
2326 	vp = nd.ni_vp;
2327 	error = change_flags(vp, SCARG(uap, flags), l);
2328 	vput(vp);
2329 	return (error);
2330 }
2331 
2332 /*
2333  * Change flags of a file given a file descriptor.
2334  */
2335 /* ARGSUSED */
2336 int
2337 sys_fchflags(struct lwp *l, void *v, register_t *retval)
2338 {
2339 	struct sys_fchflags_args /* {
2340 		syscallarg(int) fd;
2341 		syscallarg(u_long) flags;
2342 	} */ *uap = v;
2343 	struct proc *p = l->l_proc;
2344 	struct vnode *vp;
2345 	struct file *fp;
2346 	int error;
2347 
2348 	/* getvnode() will use the descriptor for us */
2349 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2350 		return (error);
2351 	vp = (struct vnode *)fp->f_data;
2352 	error = change_flags(vp, SCARG(uap, flags), l);
2353 	VOP_UNLOCK(vp, 0);
2354 	FILE_UNUSE(fp, l);
2355 	return (error);
2356 }
2357 
2358 /*
2359  * Change flags of a file given a path name; this version does
2360  * not follow links.
2361  */
2362 int
2363 sys_lchflags(struct lwp *l, void *v, register_t *retval)
2364 {
2365 	struct sys_lchflags_args /* {
2366 		syscallarg(const char *) path;
2367 		syscallarg(u_long) flags;
2368 	} */ *uap = v;
2369 	struct vnode *vp;
2370 	int error;
2371 	struct nameidata nd;
2372 
2373 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2374 	if ((error = namei(&nd)) != 0)
2375 		return (error);
2376 	vp = nd.ni_vp;
2377 	error = change_flags(vp, SCARG(uap, flags), l);
2378 	vput(vp);
2379 	return (error);
2380 }
2381 
2382 /*
2383  * Common routine to change flags of a file.
2384  */
2385 int
2386 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2387 {
2388 	struct proc *p = l->l_proc;
2389 	struct mount *mp;
2390 	struct vattr vattr;
2391 	int error;
2392 
2393 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2394 		return (error);
2395 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
2396 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2397 	/*
2398 	 * Non-superusers cannot change the flags on devices, even if they
2399 	 * own them.
2400 	 */
2401 	if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
2402 			      &p->p_acflag) != 0) {
2403 		if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0)
2404 			goto out;
2405 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2406 			error = EINVAL;
2407 			goto out;
2408 		}
2409 	}
2410 	VATTR_NULL(&vattr);
2411 	vattr.va_flags = flags;
2412 	error = VOP_SETATTR(vp, &vattr, p->p_cred, l);
2413 out:
2414 	vn_finished_write(mp, 0);
2415 	return (error);
2416 }
2417 
2418 /*
2419  * Change mode of a file given path name; this version follows links.
2420  */
2421 /* ARGSUSED */
2422 int
2423 sys_chmod(struct lwp *l, void *v, register_t *retval)
2424 {
2425 	struct sys_chmod_args /* {
2426 		syscallarg(const char *) path;
2427 		syscallarg(int) mode;
2428 	} */ *uap = v;
2429 	int error;
2430 	struct nameidata nd;
2431 
2432 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2433 	if ((error = namei(&nd)) != 0)
2434 		return (error);
2435 
2436 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2437 
2438 	vrele(nd.ni_vp);
2439 	return (error);
2440 }
2441 
2442 /*
2443  * Change mode of a file given a file descriptor.
2444  */
2445 /* ARGSUSED */
2446 int
2447 sys_fchmod(struct lwp *l, void *v, register_t *retval)
2448 {
2449 	struct sys_fchmod_args /* {
2450 		syscallarg(int) fd;
2451 		syscallarg(int) mode;
2452 	} */ *uap = v;
2453 	struct proc *p = l->l_proc;
2454 	struct file *fp;
2455 	int error;
2456 
2457 	/* getvnode() will use the descriptor for us */
2458 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2459 		return (error);
2460 
2461 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l);
2462 	FILE_UNUSE(fp, l);
2463 	return (error);
2464 }
2465 
2466 /*
2467  * Change mode of a file given path name; this version does not follow links.
2468  */
2469 /* ARGSUSED */
2470 int
2471 sys_lchmod(struct lwp *l, void *v, register_t *retval)
2472 {
2473 	struct sys_lchmod_args /* {
2474 		syscallarg(const char *) path;
2475 		syscallarg(int) mode;
2476 	} */ *uap = v;
2477 	int error;
2478 	struct nameidata nd;
2479 
2480 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2481 	if ((error = namei(&nd)) != 0)
2482 		return (error);
2483 
2484 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2485 
2486 	vrele(nd.ni_vp);
2487 	return (error);
2488 }
2489 
2490 /*
2491  * Common routine to set mode given a vnode.
2492  */
2493 static int
2494 change_mode(struct vnode *vp, int mode, struct lwp *l)
2495 {
2496 	struct proc *p = l->l_proc;
2497 	struct mount *mp;
2498 	struct vattr vattr;
2499 	int error;
2500 
2501 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2502 		return (error);
2503 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
2504 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2505 	VATTR_NULL(&vattr);
2506 	vattr.va_mode = mode & ALLPERMS;
2507 	error = VOP_SETATTR(vp, &vattr, p->p_cred, l);
2508 	VOP_UNLOCK(vp, 0);
2509 	vn_finished_write(mp, 0);
2510 	return (error);
2511 }
2512 
2513 /*
2514  * Set ownership given a path name; this version follows links.
2515  */
2516 /* ARGSUSED */
2517 int
2518 sys_chown(struct lwp *l, void *v, register_t *retval)
2519 {
2520 	struct sys_chown_args /* {
2521 		syscallarg(const char *) path;
2522 		syscallarg(uid_t) uid;
2523 		syscallarg(gid_t) gid;
2524 	} */ *uap = v;
2525 	int error;
2526 	struct nameidata nd;
2527 
2528 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2529 	if ((error = namei(&nd)) != 0)
2530 		return (error);
2531 
2532 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2533 
2534 	vrele(nd.ni_vp);
2535 	return (error);
2536 }
2537 
2538 /*
2539  * Set ownership given a path name; this version follows links.
2540  * Provides POSIX semantics.
2541  */
2542 /* ARGSUSED */
2543 int
2544 sys___posix_chown(struct lwp *l, void *v, register_t *retval)
2545 {
2546 	struct sys_chown_args /* {
2547 		syscallarg(const char *) path;
2548 		syscallarg(uid_t) uid;
2549 		syscallarg(gid_t) gid;
2550 	} */ *uap = v;
2551 	int error;
2552 	struct nameidata nd;
2553 
2554 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2555 	if ((error = namei(&nd)) != 0)
2556 		return (error);
2557 
2558 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2559 
2560 	vrele(nd.ni_vp);
2561 	return (error);
2562 }
2563 
2564 /*
2565  * Set ownership given a file descriptor.
2566  */
2567 /* ARGSUSED */
2568 int
2569 sys_fchown(struct lwp *l, void *v, register_t *retval)
2570 {
2571 	struct sys_fchown_args /* {
2572 		syscallarg(int) fd;
2573 		syscallarg(uid_t) uid;
2574 		syscallarg(gid_t) gid;
2575 	} */ *uap = v;
2576 	struct proc *p = l->l_proc;
2577 	int error;
2578 	struct file *fp;
2579 
2580 	/* getvnode() will use the descriptor for us */
2581 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2582 		return (error);
2583 
2584 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2585 	    SCARG(uap, gid), l, 0);
2586 	FILE_UNUSE(fp, l);
2587 	return (error);
2588 }
2589 
2590 /*
2591  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2592  */
2593 /* ARGSUSED */
2594 int
2595 sys___posix_fchown(struct lwp *l, void *v, register_t *retval)
2596 {
2597 	struct sys_fchown_args /* {
2598 		syscallarg(int) fd;
2599 		syscallarg(uid_t) uid;
2600 		syscallarg(gid_t) gid;
2601 	} */ *uap = v;
2602 	struct proc *p = l->l_proc;
2603 	int error;
2604 	struct file *fp;
2605 
2606 	/* getvnode() will use the descriptor for us */
2607 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2608 		return (error);
2609 
2610 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2611 	    SCARG(uap, gid), l, 1);
2612 	FILE_UNUSE(fp, l);
2613 	return (error);
2614 }
2615 
2616 /*
2617  * Set ownership given a path name; this version does not follow links.
2618  */
2619 /* ARGSUSED */
2620 int
2621 sys_lchown(struct lwp *l, void *v, register_t *retval)
2622 {
2623 	struct sys_lchown_args /* {
2624 		syscallarg(const char *) path;
2625 		syscallarg(uid_t) uid;
2626 		syscallarg(gid_t) gid;
2627 	} */ *uap = v;
2628 	int error;
2629 	struct nameidata nd;
2630 
2631 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2632 	if ((error = namei(&nd)) != 0)
2633 		return (error);
2634 
2635 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2636 
2637 	vrele(nd.ni_vp);
2638 	return (error);
2639 }
2640 
2641 /*
2642  * Set ownership given a path name; this version does not follow links.
2643  * Provides POSIX/XPG semantics.
2644  */
2645 /* ARGSUSED */
2646 int
2647 sys___posix_lchown(struct lwp *l, void *v, register_t *retval)
2648 {
2649 	struct sys_lchown_args /* {
2650 		syscallarg(const char *) path;
2651 		syscallarg(uid_t) uid;
2652 		syscallarg(gid_t) gid;
2653 	} */ *uap = v;
2654 	int error;
2655 	struct nameidata nd;
2656 
2657 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2658 	if ((error = namei(&nd)) != 0)
2659 		return (error);
2660 
2661 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2662 
2663 	vrele(nd.ni_vp);
2664 	return (error);
2665 }
2666 
2667 /*
2668  * Common routine to set ownership given a vnode.
2669  */
2670 static int
2671 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2672     int posix_semantics)
2673 {
2674 	struct proc *p = l->l_proc;
2675 	struct mount *mp;
2676 	struct vattr vattr;
2677 	mode_t newmode;
2678 	int error;
2679 
2680 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2681 		return (error);
2682 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
2683 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2684 	if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0)
2685 		goto out;
2686 
2687 #define CHANGED(x) ((int)(x) != -1)
2688 	newmode = vattr.va_mode;
2689 	if (posix_semantics) {
2690 		/*
2691 		 * POSIX/XPG semantics: if the caller is not the super-user,
2692 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2693 		 * the XPG consider the behaviour for calls by the super-user
2694 		 * implementation-defined; we leave the set-user-id and set-
2695 		 * group-id settings intact in that case.
2696 		 */
2697 		if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
2698 				      NULL) != 0)
2699 			newmode &= ~(S_ISUID | S_ISGID);
2700 	} else {
2701 		/*
2702 		 * NetBSD semantics: when changing owner and/or group,
2703 		 * clear the respective bit(s).
2704 		 */
2705 		if (CHANGED(uid))
2706 			newmode &= ~S_ISUID;
2707 		if (CHANGED(gid))
2708 			newmode &= ~S_ISGID;
2709 	}
2710 	/* Update va_mode iff altered. */
2711 	if (vattr.va_mode == newmode)
2712 		newmode = VNOVAL;
2713 
2714 	VATTR_NULL(&vattr);
2715 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2716 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2717 	vattr.va_mode = newmode;
2718 	error = VOP_SETATTR(vp, &vattr, p->p_cred, l);
2719 #undef CHANGED
2720 
2721 out:
2722 	VOP_UNLOCK(vp, 0);
2723 	vn_finished_write(mp, 0);
2724 	return (error);
2725 }
2726 
2727 /*
2728  * Set the access and modification times given a path name; this
2729  * version follows links.
2730  */
2731 /* ARGSUSED */
2732 int
2733 sys_utimes(struct lwp *l, void *v, register_t *retval)
2734 {
2735 	struct sys_utimes_args /* {
2736 		syscallarg(const char *) path;
2737 		syscallarg(const struct timeval *) tptr;
2738 	} */ *uap = v;
2739 	int error;
2740 	struct nameidata nd;
2741 
2742 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2743 	if ((error = namei(&nd)) != 0)
2744 		return (error);
2745 
2746 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l);
2747 
2748 	vrele(nd.ni_vp);
2749 	return (error);
2750 }
2751 
2752 /*
2753  * Set the access and modification times given a file descriptor.
2754  */
2755 /* ARGSUSED */
2756 int
2757 sys_futimes(struct lwp *l, void *v, register_t *retval)
2758 {
2759 	struct sys_futimes_args /* {
2760 		syscallarg(int) fd;
2761 		syscallarg(const struct timeval *) tptr;
2762 	} */ *uap = v;
2763 	struct proc *p = l->l_proc;
2764 	int error;
2765 	struct file *fp;
2766 
2767 	/* getvnode() will use the descriptor for us */
2768 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2769 		return (error);
2770 
2771 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l);
2772 	FILE_UNUSE(fp, l);
2773 	return (error);
2774 }
2775 
2776 /*
2777  * Set the access and modification times given a path name; this
2778  * version does not follow links.
2779  */
2780 /* ARGSUSED */
2781 int
2782 sys_lutimes(struct lwp *l, void *v, register_t *retval)
2783 {
2784 	struct sys_lutimes_args /* {
2785 		syscallarg(const char *) path;
2786 		syscallarg(const struct timeval *) tptr;
2787 	} */ *uap = v;
2788 	int error;
2789 	struct nameidata nd;
2790 
2791 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2792 	if ((error = namei(&nd)) != 0)
2793 		return (error);
2794 
2795 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l);
2796 
2797 	vrele(nd.ni_vp);
2798 	return (error);
2799 }
2800 
2801 /*
2802  * Common routine to set access and modification times given a vnode.
2803  */
2804 static int
2805 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l)
2806 {
2807 	struct proc *p = l->l_proc;
2808 	struct mount *mp;
2809 	struct vattr vattr;
2810 	int error;
2811 
2812 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2813 		return (error);
2814 	VATTR_NULL(&vattr);
2815 	if (tptr == NULL) {
2816 		nanotime(&vattr.va_atime);
2817 		vattr.va_mtime = vattr.va_atime;
2818 		vattr.va_vaflags |= VA_UTIMES_NULL;
2819 	} else {
2820 		struct timeval tv[2];
2821 
2822 		error = copyin(tptr, tv, sizeof(tv));
2823 		if (error)
2824 			goto out;
2825 		TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime);
2826 		TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime);
2827 	}
2828 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
2829 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2830 	error = VOP_SETATTR(vp, &vattr, p->p_cred, l);
2831 	VOP_UNLOCK(vp, 0);
2832 out:
2833 	vn_finished_write(mp, 0);
2834 	return (error);
2835 }
2836 
2837 /*
2838  * Truncate a file given its path name.
2839  */
2840 /* ARGSUSED */
2841 int
2842 sys_truncate(struct lwp *l, void *v, register_t *retval)
2843 {
2844 	struct sys_truncate_args /* {
2845 		syscallarg(const char *) path;
2846 		syscallarg(int) pad;
2847 		syscallarg(off_t) length;
2848 	} */ *uap = v;
2849 	struct proc *p = l->l_proc;
2850 	struct vnode *vp;
2851 	struct mount *mp;
2852 	struct vattr vattr;
2853 	int error;
2854 	struct nameidata nd;
2855 
2856 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
2857 	if ((error = namei(&nd)) != 0)
2858 		return (error);
2859 	vp = nd.ni_vp;
2860 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2861 		vrele(vp);
2862 		return (error);
2863 	}
2864 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
2865 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2866 	if (vp->v_type == VDIR)
2867 		error = EISDIR;
2868 	else if ((error = vn_writechk(vp)) == 0 &&
2869 	    (error = VOP_ACCESS(vp, VWRITE, p->p_cred, l)) == 0) {
2870 		VATTR_NULL(&vattr);
2871 		vattr.va_size = SCARG(uap, length);
2872 		error = VOP_SETATTR(vp, &vattr, p->p_cred, l);
2873 	}
2874 	vput(vp);
2875 	vn_finished_write(mp, 0);
2876 	return (error);
2877 }
2878 
2879 /*
2880  * Truncate a file given a file descriptor.
2881  */
2882 /* ARGSUSED */
2883 int
2884 sys_ftruncate(struct lwp *l, void *v, register_t *retval)
2885 {
2886 	struct sys_ftruncate_args /* {
2887 		syscallarg(int) fd;
2888 		syscallarg(int) pad;
2889 		syscallarg(off_t) length;
2890 	} */ *uap = v;
2891 	struct proc *p = l->l_proc;
2892 	struct mount *mp;
2893 	struct vattr vattr;
2894 	struct vnode *vp;
2895 	struct file *fp;
2896 	int error;
2897 
2898 	/* getvnode() will use the descriptor for us */
2899 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2900 		return (error);
2901 	if ((fp->f_flag & FWRITE) == 0) {
2902 		error = EINVAL;
2903 		goto out;
2904 	}
2905 	vp = (struct vnode *)fp->f_data;
2906 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2907 		FILE_UNUSE(fp, l);
2908 		return (error);
2909 	}
2910 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
2911 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2912 	if (vp->v_type == VDIR)
2913 		error = EISDIR;
2914 	else if ((error = vn_writechk(vp)) == 0) {
2915 		VATTR_NULL(&vattr);
2916 		vattr.va_size = SCARG(uap, length);
2917 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, l);
2918 	}
2919 	VOP_UNLOCK(vp, 0);
2920 	vn_finished_write(mp, 0);
2921  out:
2922 	FILE_UNUSE(fp, l);
2923 	return (error);
2924 }
2925 
2926 /*
2927  * Sync an open file.
2928  */
2929 /* ARGSUSED */
2930 int
2931 sys_fsync(struct lwp *l, void *v, register_t *retval)
2932 {
2933 	struct sys_fsync_args /* {
2934 		syscallarg(int) fd;
2935 	} */ *uap = v;
2936 	struct proc *p = l->l_proc;
2937 	struct vnode *vp;
2938 	struct mount *mp;
2939 	struct file *fp;
2940 	int error;
2941 
2942 	/* getvnode() will use the descriptor for us */
2943 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2944 		return (error);
2945 	vp = (struct vnode *)fp->f_data;
2946 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2947 		FILE_UNUSE(fp, l);
2948 		return (error);
2949 	}
2950 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2951 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l);
2952 	if (error == 0 && bioops.io_fsync != NULL &&
2953 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2954 		(*bioops.io_fsync)(vp, 0);
2955 	VOP_UNLOCK(vp, 0);
2956 	vn_finished_write(mp, 0);
2957 	FILE_UNUSE(fp, l);
2958 	return (error);
2959 }
2960 
2961 /*
2962  * Sync a range of file data.  API modeled after that found in AIX.
2963  *
2964  * FDATASYNC indicates that we need only save enough metadata to be able
2965  * to re-read the written data.  Note we duplicate AIX's requirement that
2966  * the file be open for writing.
2967  */
2968 /* ARGSUSED */
2969 int
2970 sys_fsync_range(struct lwp *l, void *v, register_t *retval)
2971 {
2972 	struct sys_fsync_range_args /* {
2973 		syscallarg(int) fd;
2974 		syscallarg(int) flags;
2975 		syscallarg(off_t) start;
2976 		syscallarg(off_t) length;
2977 	} */ *uap = v;
2978 	struct proc *p = l->l_proc;
2979 	struct vnode *vp;
2980 	struct file *fp;
2981 	int flags, nflags;
2982 	off_t s, e, len;
2983 	int error;
2984 
2985 	/* getvnode() will use the descriptor for us */
2986 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2987 		return (error);
2988 
2989 	if ((fp->f_flag & FWRITE) == 0) {
2990 		FILE_UNUSE(fp, l);
2991 		return (EBADF);
2992 	}
2993 
2994 	flags = SCARG(uap, flags);
2995 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
2996 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
2997 		return (EINVAL);
2998 	}
2999 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3000 	if (flags & FDATASYNC)
3001 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3002 	else
3003 		nflags = FSYNC_WAIT;
3004 	if (flags & FDISKSYNC)
3005 		nflags |= FSYNC_CACHE;
3006 
3007 	len = SCARG(uap, length);
3008 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3009 	if (len) {
3010 		s = SCARG(uap, start);
3011 		e = s + len;
3012 		if (e < s) {
3013 			FILE_UNUSE(fp, l);
3014 			return (EINVAL);
3015 		}
3016 	} else {
3017 		e = 0;
3018 		s = 0;
3019 	}
3020 
3021 	vp = (struct vnode *)fp->f_data;
3022 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3023 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l);
3024 
3025 	if (error == 0 && bioops.io_fsync != NULL &&
3026 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3027 		(*bioops.io_fsync)(vp, nflags);
3028 
3029 	VOP_UNLOCK(vp, 0);
3030 	FILE_UNUSE(fp, l);
3031 	return (error);
3032 }
3033 
3034 /*
3035  * Sync the data of an open file.
3036  */
3037 /* ARGSUSED */
3038 int
3039 sys_fdatasync(struct lwp *l, void *v, register_t *retval)
3040 {
3041 	struct sys_fdatasync_args /* {
3042 		syscallarg(int) fd;
3043 	} */ *uap = v;
3044 	struct proc *p = l->l_proc;
3045 	struct vnode *vp;
3046 	struct file *fp;
3047 	int error;
3048 
3049 	/* getvnode() will use the descriptor for us */
3050 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3051 		return (error);
3052 	if ((fp->f_flag & FWRITE) == 0) {
3053 		FILE_UNUSE(fp, l);
3054 		return (EBADF);
3055 	}
3056 	vp = (struct vnode *)fp->f_data;
3057 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3058 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l);
3059 	VOP_UNLOCK(vp, 0);
3060 	FILE_UNUSE(fp, l);
3061 	return (error);
3062 }
3063 
3064 /*
3065  * Rename files, (standard) BSD semantics frontend.
3066  */
3067 /* ARGSUSED */
3068 int
3069 sys_rename(struct lwp *l, void *v, register_t *retval)
3070 {
3071 	struct sys_rename_args /* {
3072 		syscallarg(const char *) from;
3073 		syscallarg(const char *) to;
3074 	} */ *uap = v;
3075 
3076 	return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0));
3077 }
3078 
3079 /*
3080  * Rename files, POSIX semantics frontend.
3081  */
3082 /* ARGSUSED */
3083 int
3084 sys___posix_rename(struct lwp *l, void *v, register_t *retval)
3085 {
3086 	struct sys___posix_rename_args /* {
3087 		syscallarg(const char *) from;
3088 		syscallarg(const char *) to;
3089 	} */ *uap = v;
3090 
3091 	return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1));
3092 }
3093 
3094 /*
3095  * Rename files.  Source and destination must either both be directories,
3096  * or both not be directories.  If target is a directory, it must be empty.
3097  * If `from' and `to' refer to the same object, the value of the `retain'
3098  * argument is used to determine whether `from' will be
3099  *
3100  * (retain == 0)	deleted unless `from' and `to' refer to the same
3101  *			object in the file system's name space (BSD).
3102  * (retain == 1)	always retained (POSIX).
3103  */
3104 static int
3105 rename_files(const char *from, const char *to, struct lwp *l, int retain)
3106 {
3107 	struct mount *mp = NULL;
3108 	struct vnode *tvp, *fvp, *tdvp;
3109 	struct nameidata fromnd, tond;
3110 	struct proc *p;
3111 	int error;
3112 
3113 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
3114 	    from, l);
3115 	if ((error = namei(&fromnd)) != 0)
3116 		return (error);
3117 	fvp = fromnd.ni_vp;
3118 	error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH);
3119 	if (error != 0) {
3120 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3121 		vrele(fromnd.ni_dvp);
3122 		vrele(fvp);
3123 		if (fromnd.ni_startdir)
3124 			vrele(fromnd.ni_startdir);
3125 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3126 		return (error);
3127 	}
3128 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3129 	    (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l);
3130 	if ((error = namei(&tond)) != 0) {
3131 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3132 		vrele(fromnd.ni_dvp);
3133 		vrele(fvp);
3134 		goto out1;
3135 	}
3136 	tdvp = tond.ni_dvp;
3137 	tvp = tond.ni_vp;
3138 
3139 	if (tvp != NULL) {
3140 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3141 			error = ENOTDIR;
3142 			goto out;
3143 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3144 			error = EISDIR;
3145 			goto out;
3146 		}
3147 	}
3148 
3149 	if (fvp == tdvp)
3150 		error = EINVAL;
3151 
3152 	/*
3153 	 * Source and destination refer to the same object.
3154 	 */
3155 	if (fvp == tvp) {
3156 		if (retain)
3157 			error = -1;
3158 		else if (fromnd.ni_dvp == tdvp &&
3159 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3160 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3161 		          tond.ni_cnd.cn_nameptr,
3162 		          fromnd.ni_cnd.cn_namelen))
3163 		error = -1;
3164 	}
3165 
3166 #ifdef VERIFIED_EXEC
3167 	if (!error)
3168 		error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l);
3169 #endif /* VERIFIED_EXEC */
3170 
3171 out:
3172 	p = l->l_proc;
3173 	if (!error) {
3174 		VOP_LEASE(tdvp, l, p->p_cred, LEASE_WRITE);
3175 		if (fromnd.ni_dvp != tdvp)
3176 			VOP_LEASE(fromnd.ni_dvp, l, p->p_cred, LEASE_WRITE);
3177 		if (tvp) {
3178 			VOP_LEASE(tvp, l, p->p_cred, LEASE_WRITE);
3179 		}
3180 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3181 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3182 	} else {
3183 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3184 		if (tdvp == tvp)
3185 			vrele(tdvp);
3186 		else
3187 			vput(tdvp);
3188 		if (tvp)
3189 			vput(tvp);
3190 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3191 		vrele(fromnd.ni_dvp);
3192 		vrele(fvp);
3193 	}
3194 	vrele(tond.ni_startdir);
3195 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3196 out1:
3197 	vn_finished_write(mp, 0);
3198 	if (fromnd.ni_startdir)
3199 		vrele(fromnd.ni_startdir);
3200 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3201 	return (error == -1 ? 0 : error);
3202 }
3203 
3204 /*
3205  * Make a directory file.
3206  */
3207 /* ARGSUSED */
3208 int
3209 sys_mkdir(struct lwp *l, void *v, register_t *retval)
3210 {
3211 	struct sys_mkdir_args /* {
3212 		syscallarg(const char *) path;
3213 		syscallarg(int) mode;
3214 	} */ *uap = v;
3215 	struct proc *p = l->l_proc;
3216 	struct mount *mp;
3217 	struct vnode *vp;
3218 	struct vattr vattr;
3219 	int error;
3220 	struct nameidata nd;
3221 
3222 restart:
3223 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE,
3224 	    SCARG(uap, path), l);
3225 	if ((error = namei(&nd)) != 0)
3226 		return (error);
3227 	vp = nd.ni_vp;
3228 	if (vp != NULL) {
3229 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3230 		if (nd.ni_dvp == vp)
3231 			vrele(nd.ni_dvp);
3232 		else
3233 			vput(nd.ni_dvp);
3234 		vrele(vp);
3235 		return (EEXIST);
3236 	}
3237 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3238 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3239 		if (nd.ni_dvp == vp)
3240 			vrele(nd.ni_dvp);
3241 		else
3242 			vput(nd.ni_dvp);
3243 		if ((error = vn_start_write(NULL, &mp,
3244 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3245 			return (error);
3246 		goto restart;
3247 	}
3248 	VATTR_NULL(&vattr);
3249 	vattr.va_type = VDIR;
3250 	vattr.va_mode =
3251 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3252 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
3253 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3254 	if (!error)
3255 		vput(nd.ni_vp);
3256 	vn_finished_write(mp, 0);
3257 	return (error);
3258 }
3259 
3260 /*
3261  * Remove a directory file.
3262  */
3263 /* ARGSUSED */
3264 int
3265 sys_rmdir(struct lwp *l, void *v, register_t *retval)
3266 {
3267 	struct sys_rmdir_args /* {
3268 		syscallarg(const char *) path;
3269 	} */ *uap = v;
3270 	struct proc *p = l->l_proc;
3271 	struct mount *mp;
3272 	struct vnode *vp;
3273 	int error;
3274 	struct nameidata nd;
3275 
3276 restart:
3277 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3278 	    SCARG(uap, path), l);
3279 	if ((error = namei(&nd)) != 0)
3280 		return (error);
3281 	vp = nd.ni_vp;
3282 	if (vp->v_type != VDIR) {
3283 		error = ENOTDIR;
3284 		goto out;
3285 	}
3286 	/*
3287 	 * No rmdir "." please.
3288 	 */
3289 	if (nd.ni_dvp == vp) {
3290 		error = EINVAL;
3291 		goto out;
3292 	}
3293 	/*
3294 	 * The root of a mounted filesystem cannot be deleted.
3295 	 */
3296 	if (vp->v_flag & VROOT) {
3297 		error = EBUSY;
3298 		goto out;
3299 	}
3300 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3301 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3302 		if (nd.ni_dvp == vp)
3303 			vrele(nd.ni_dvp);
3304 		else
3305 			vput(nd.ni_dvp);
3306 		vput(vp);
3307 		if ((error = vn_start_write(NULL, &mp,
3308 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3309 			return (error);
3310 		goto restart;
3311 	}
3312 	VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE);
3313 	VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE);
3314 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3315 	vn_finished_write(mp, 0);
3316 	return (error);
3317 
3318 out:
3319 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3320 	if (nd.ni_dvp == vp)
3321 		vrele(nd.ni_dvp);
3322 	else
3323 		vput(nd.ni_dvp);
3324 	vput(vp);
3325 	return (error);
3326 }
3327 
3328 /*
3329  * Read a block of directory entries in a file system independent format.
3330  */
3331 int
3332 sys___getdents30(struct lwp *l, void *v, register_t *retval)
3333 {
3334 	struct sys___getdents30_args /* {
3335 		syscallarg(int) fd;
3336 		syscallarg(char *) buf;
3337 		syscallarg(size_t) count;
3338 	} */ *uap = v;
3339 	struct proc *p = l->l_proc;
3340 	struct file *fp;
3341 	int error, done;
3342 
3343 	/* getvnode() will use the descriptor for us */
3344 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3345 		return (error);
3346 	if ((fp->f_flag & FREAD) == 0) {
3347 		error = EBADF;
3348 		goto out;
3349 	}
3350 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3351 			SCARG(uap, count), &done, l, 0, 0);
3352 #ifdef KTRACE
3353 	if (!error && KTRPOINT(p, KTR_GENIO)) {
3354 		struct iovec iov;
3355 		iov.iov_base = SCARG(uap, buf);
3356 		iov.iov_len = done;
3357 		ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0);
3358 	}
3359 #endif
3360 	*retval = done;
3361  out:
3362 	FILE_UNUSE(fp, l);
3363 	return (error);
3364 }
3365 
3366 /*
3367  * Set the mode mask for creation of filesystem nodes.
3368  */
3369 int
3370 sys_umask(struct lwp *l, void *v, register_t *retval)
3371 {
3372 	struct sys_umask_args /* {
3373 		syscallarg(mode_t) newmask;
3374 	} */ *uap = v;
3375 	struct proc *p = l->l_proc;
3376 	struct cwdinfo *cwdi;
3377 
3378 	cwdi = p->p_cwdi;
3379 	*retval = cwdi->cwdi_cmask;
3380 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3381 	return (0);
3382 }
3383 
3384 /*
3385  * Void all references to file by ripping underlying filesystem
3386  * away from vnode.
3387  */
3388 /* ARGSUSED */
3389 int
3390 sys_revoke(struct lwp *l, void *v, register_t *retval)
3391 {
3392 	struct sys_revoke_args /* {
3393 		syscallarg(const char *) path;
3394 	} */ *uap = v;
3395 	struct proc *p = l->l_proc;
3396 	struct mount *mp;
3397 	struct vnode *vp;
3398 	struct vattr vattr;
3399 	int error;
3400 	struct nameidata nd;
3401 
3402 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
3403 	if ((error = namei(&nd)) != 0)
3404 		return (error);
3405 	vp = nd.ni_vp;
3406 	if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0)
3407 		goto out;
3408 	if (kauth_cred_geteuid(p->p_cred) != vattr.va_uid &&
3409 	    (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
3410 				       &p->p_acflag)) != 0)
3411 		goto out;
3412 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3413 		goto out;
3414 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3415 		VOP_REVOKE(vp, REVOKEALL);
3416 	vn_finished_write(mp, 0);
3417 out:
3418 	vrele(vp);
3419 	return (error);
3420 }
3421 
3422 /*
3423  * Convert a user file descriptor to a kernel file entry.
3424  */
3425 int
3426 getvnode(struct filedesc *fdp, int fd, struct file **fpp)
3427 {
3428 	struct vnode *vp;
3429 	struct file *fp;
3430 
3431 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3432 		return (EBADF);
3433 
3434 	FILE_USE(fp);
3435 
3436 	if (fp->f_type != DTYPE_VNODE) {
3437 		FILE_UNUSE(fp, NULL);
3438 		return (EINVAL);
3439 	}
3440 
3441 	vp = (struct vnode *)fp->f_data;
3442 	if (vp->v_type == VBAD) {
3443 		FILE_UNUSE(fp, NULL);
3444 		return (EBADF);
3445 	}
3446 
3447 	*fpp = fp;
3448 	return (0);
3449 }
3450