xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision fdecd6a253f999ae92b139670d9e15cc9df4497c)
1 /*	$NetBSD: vfs_syscalls.c,v 1.91 1997/06/24 23:44:57 fvdl 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. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	@(#)vfs_syscalls.c	8.28 (Berkeley) 12/10/94
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/namei.h>
46 #include <sys/filedesc.h>
47 #include <sys/kernel.h>
48 #include <sys/file.h>
49 #include <sys/stat.h>
50 #include <sys/vnode.h>
51 #include <sys/mount.h>
52 #include <sys/proc.h>
53 #include <sys/uio.h>
54 #include <sys/malloc.h>
55 #include <sys/dirent.h>
56 
57 #include <sys/syscallargs.h>
58 
59 #include <vm/vm.h>
60 #include <sys/sysctl.h>
61 
62 static int change_dir __P((struct nameidata *, struct proc *));
63 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *));
64 static int rename_files __P((const char *, const char *, struct proc *, int));
65 
66 void checkdirs __P((struct vnode *));
67 int dounmount __P((struct mount *, int, struct proc *));
68 
69 /*
70  * Virtual File System System Calls
71  */
72 
73 /*
74  * Mount a file system.
75  */
76 /* ARGSUSED */
77 int
78 sys_mount(p, v, retval)
79 	struct proc *p;
80 	void *v;
81 	register_t *retval;
82 {
83 	register struct sys_mount_args /* {
84 		syscallarg(const char *) type;
85 		syscallarg(const char *) path;
86 		syscallarg(int) flags;
87 		syscallarg(void *) data;
88 	} */ *uap = v;
89 	register struct vnode *vp;
90 	register struct mount *mp;
91 	int error, flag = 0;
92 	u_long fsindex = 0;
93 	char fstypename[MFSNAMELEN];
94 	struct vattr va;
95 	struct nameidata nd;
96 
97 	/*
98 	 * Get vnode to be covered
99 	 */
100 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
101 	    SCARG(uap, path), p);
102 	if ((error = namei(&nd)) != 0)
103 		return (error);
104 	vp = nd.ni_vp;
105 	if (SCARG(uap, flags) & MNT_UPDATE) {
106 		if ((vp->v_flag & VROOT) == 0) {
107 			vput(vp);
108 			return (EINVAL);
109 		}
110 		mp = vp->v_mount;
111 		flag = mp->mnt_flag;
112 		/*
113 		 * We only allow the filesystem to be reloaded if it
114 		 * is currently mounted read-only.
115 		 */
116 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
117 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
118 			vput(vp);
119 			return (EOPNOTSUPP);	/* Needs translation */
120 		}
121 		mp->mnt_flag |=
122 		    SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
123 		/*
124 		 * Only root, or the user that did the original mount is
125 		 * permitted to update it.
126 		 */
127 		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
128 		    (error = suser(p->p_ucred, &p->p_acflag))) {
129 			vput(vp);
130 			return (error);
131 		}
132 		/*
133 		 * Do not allow NFS export by non-root users. Silently
134 		 * enforce MNT_NOSUID and MNT_NODEV for non-root users.
135 		 */
136 		if (p->p_ucred->cr_uid != 0) {
137 			if (SCARG(uap, flags) & MNT_EXPORTED) {
138 				vput(vp);
139 				return (EPERM);
140 			}
141 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
142 		}
143 		VOP_UNLOCK(vp);
144 		goto update;
145 	}
146 	/*
147 	 * If the user is not root, ensure that they own the directory
148 	 * onto which we are attempting to mount.
149 	 */
150 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) ||
151 	    (va.va_uid != p->p_ucred->cr_uid &&
152 	     (error = suser(p->p_ucred, &p->p_acflag)))) {
153 		vput(vp);
154 		return (error);
155 	}
156 	/*
157 	 * Do not allow NFS export by non-root users. Silently
158 	 * enforce MNT_NOSUID and MNT_NODEV for non-root users.
159 	 */
160 	if (p->p_ucred->cr_uid != 0) {
161 		if (SCARG(uap, flags) & MNT_EXPORTED) {
162 			vput(vp);
163 			return (EPERM);
164 		}
165 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
166 	}
167 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
168 		return (error);
169 	if (vp->v_type != VDIR) {
170 		vput(vp);
171 		return (ENOTDIR);
172 	}
173 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
174 	if (error) {
175 #if defined(COMPAT_09) || defined(COMPAT_43)
176 		/*
177 		 * Historically filesystem types were identified by number.
178 		 * If we get an integer for the filesystem type instead of a
179 		 * string, we check to see if it matches one of the historic
180 		 * filesystem types.
181 		 */
182 		fsindex = (u_long)SCARG(uap, type);
183 		if (fsindex >= nvfssw || vfssw[fsindex] == NULL) {
184 			vput(vp);
185 			return (ENODEV);
186 		}
187 		strncpy(fstypename, vfssw[fsindex]->vfs_name, MFSNAMELEN);
188 #else
189 		vput(vp);
190 		return (error);
191 #endif
192 	}
193 #ifdef	COMPAT_10
194 	/* Accept `ufs' as an alias for `ffs'. */
195 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
196 		strncpy(fstypename, "ffs", MFSNAMELEN);
197 #endif
198 	for (fsindex = 0; fsindex < nvfssw; fsindex++)
199 		if (vfssw[fsindex] != NULL &&
200 		    !strncmp(vfssw[fsindex]->vfs_name, fstypename, MFSNAMELEN))
201 			break;
202 	if (fsindex >= nvfssw) {
203 		vput(vp);
204 		return (ENODEV);
205 	}
206 	if (vp->v_mountedhere != NULL) {
207 		vput(vp);
208 		return (EBUSY);
209 	}
210 
211 	/*
212 	 * Allocate and initialize the file system.
213 	 */
214 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
215 		M_MOUNT, M_WAITOK);
216 	bzero((char *)mp, (u_long)sizeof(struct mount));
217 	mp->mnt_op = vfssw[fsindex];
218 	if ((error = vfs_lock(mp)) != 0) {
219 		free((caddr_t)mp, M_MOUNT);
220 		vput(vp);
221 		return (error);
222 	}
223 	/* Do this early in case we block later. */
224 	vfssw[fsindex]->vfs_refcount++;
225 	vp->v_mountedhere = mp;
226 	mp->mnt_vnodecovered = vp;
227 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
228 update:
229 	/*
230 	 * Set the mount level flags.
231 	 */
232 	if (SCARG(uap, flags) & MNT_RDONLY)
233 		mp->mnt_flag |= MNT_RDONLY;
234 	else if (mp->mnt_flag & MNT_RDONLY)
235 		mp->mnt_flag |= MNT_WANTRDWR;
236 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
237 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
238 	    MNT_NOATIME);
239 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
240 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
241 	    MNT_NOCOREDUMP | MNT_NOATIME);
242 	/*
243 	 * Mount the filesystem.
244 	 */
245 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
246 	if (mp->mnt_flag & MNT_UPDATE) {
247 		vrele(vp);
248 		if (mp->mnt_flag & MNT_WANTRDWR)
249 			mp->mnt_flag &= ~MNT_RDONLY;
250 		mp->mnt_flag &=~
251 		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
252 		if (error)
253 			mp->mnt_flag = flag;
254 		return (error);
255 	}
256 	/*
257 	 * Put the new filesystem on the mount list after root.
258 	 */
259 	cache_purge(vp);
260 	if (!error) {
261 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
262 		checkdirs(vp);
263 		VOP_UNLOCK(vp);
264 		vfs_unlock(mp);
265 		(void) VFS_STATFS(mp, &mp->mnt_stat, p);
266 		error = VFS_START(mp, 0, p);
267 	} else {
268 		mp->mnt_vnodecovered->v_mountedhere = (struct mount *)0;
269 		vfssw[fsindex]->vfs_refcount--;
270 		vfs_unlock(mp);
271 		free((caddr_t)mp, M_MOUNT);
272 		vput(vp);
273 	}
274 	return (error);
275 }
276 
277 /*
278  * Scan all active processes to see if any of them have a current
279  * or root directory onto which the new filesystem has just been
280  * mounted. If so, replace them with the new mount point.
281  */
282 void
283 checkdirs(olddp)
284 	struct vnode *olddp;
285 {
286 	struct filedesc *fdp;
287 	struct vnode *newdp;
288 	struct proc *p;
289 
290 	if (olddp->v_usecount == 1)
291 		return;
292 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
293 		panic("mount: lost mount");
294 	for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
295 		fdp = p->p_fd;
296 		if (fdp->fd_cdir == olddp) {
297 			vrele(fdp->fd_cdir);
298 			VREF(newdp);
299 			fdp->fd_cdir = newdp;
300 		}
301 		if (fdp->fd_rdir == olddp) {
302 			vrele(fdp->fd_rdir);
303 			VREF(newdp);
304 			fdp->fd_rdir = newdp;
305 		}
306 	}
307 	if (rootvnode == olddp) {
308 		vrele(rootvnode);
309 		VREF(newdp);
310 		rootvnode = newdp;
311 	}
312 	vput(newdp);
313 }
314 
315 /*
316  * Unmount a file system.
317  *
318  * Note: unmount takes a path to the vnode mounted on as argument,
319  * not special file (as before).
320  */
321 /* ARGSUSED */
322 int
323 sys_unmount(p, v, retval)
324 	struct proc *p;
325 	void *v;
326 	register_t *retval;
327 {
328 	register struct sys_unmount_args /* {
329 		syscallarg(const char *) path;
330 		syscallarg(int) flags;
331 	} */ *uap = v;
332 	register struct vnode *vp;
333 	struct mount *mp;
334 	int error;
335 	struct nameidata nd;
336 
337 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
338 	    SCARG(uap, path), p);
339 	if ((error = namei(&nd)) != 0)
340 		return (error);
341 	vp = nd.ni_vp;
342 	mp = vp->v_mount;
343 
344 	/*
345 	 * Only root, or the user that did the original mount is
346 	 * permitted to unmount this filesystem.
347 	 */
348 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
349 	    (error = suser(p->p_ucred, &p->p_acflag))) {
350 		vput(vp);
351 		return (error);
352 	}
353 
354 	/*
355 	 * Don't allow unmounting the root file system.
356 	 */
357 	if (mp->mnt_flag & MNT_ROOTFS) {
358 		vput(vp);
359 		return (EINVAL);
360 	}
361 
362 	/*
363 	 * Must be the root of the filesystem
364 	 */
365 	if ((vp->v_flag & VROOT) == 0) {
366 		vput(vp);
367 		return (EINVAL);
368 	}
369 	vput(vp);
370 
371 	if (vfs_busy(mp))
372 		return (EBUSY);
373 
374 	return (dounmount(mp, SCARG(uap, flags), p));
375 }
376 
377 /*
378  * Do the actual file system unmount. File system is assumed to have been
379  * marked busy by the caller.
380  */
381 int
382 dounmount(mp, flags, p)
383 	register struct mount *mp;
384 	int flags;
385 	struct proc *p;
386 {
387 	struct vnode *coveredvp;
388 	int error;
389 
390 	coveredvp = mp->mnt_vnodecovered;
391 	mp->mnt_flag |= MNT_UNMOUNT;
392 	if ((error = vfs_lock(mp)) != 0) {
393 		vfs_unbusy(mp);
394 		return (error);
395 	}
396 
397 	if (mp->mnt_flag & MNT_EXPUBLIC)
398 		vfs_setpublicfs(NULL, NULL, NULL);
399 
400 	mp->mnt_flag &=~ MNT_ASYNC;
401 	vnode_pager_umount(mp);	/* release cached vnodes */
402 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
403 	if ((error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0 ||
404 	    (flags & MNT_FORCE))
405 		error = VFS_UNMOUNT(mp, flags, p);
406 	mp->mnt_flag &= ~MNT_UNMOUNT;
407 	vfs_unbusy(mp);
408 	if (error) {
409 		vfs_unlock(mp);
410 	} else {
411 		CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
412 		if (coveredvp != NULLVP) {
413 			vrele(coveredvp);
414 			coveredvp->v_mountedhere = (struct mount *)0;
415 		}
416 		mp->mnt_op->vfs_refcount--;
417 		vfs_unlock(mp);
418 		if (mp->mnt_vnodelist.lh_first != NULL)
419 			panic("unmount: dangling vnode");
420 		free((caddr_t)mp, M_MOUNT);
421 	}
422 	return (error);
423 }
424 
425 /*
426  * Sync each mounted filesystem.
427  */
428 #ifdef DEBUG
429 int syncprt = 0;
430 struct ctldebug debug0 = { "syncprt", &syncprt };
431 #endif
432 
433 /* ARGSUSED */
434 int
435 sys_sync(p, v, retval)
436 	struct proc *p;
437 	void *v;
438 	register_t *retval;
439 {
440 	register struct mount *mp, *nmp;
441 	int asyncflag;
442 
443 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
444 		/*
445 		 * Get the prev pointer in case we hang on vfs_busy
446 		 * while we are being unmounted.
447 		 */
448 		nmp = mp->mnt_list.cqe_prev;
449 		/*
450 		 * The lock check below is to avoid races with mount
451 		 * and unmount.
452 		 */
453 		if ((mp->mnt_flag & (MNT_MLOCK|MNT_RDONLY|MNT_MPBUSY)) == 0 &&
454 		    !vfs_busy(mp)) {
455 			asyncflag = mp->mnt_flag & MNT_ASYNC;
456 			mp->mnt_flag &= ~MNT_ASYNC;
457 			vnode_pager_sync(mp);
458 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
459 			if (asyncflag)
460 				mp->mnt_flag |= MNT_ASYNC;
461 			/*
462 			 * Get the prev pointer again, as the prior filesystem
463 			 * might have been unmounted while we were sync'ing.
464 			 */
465 			nmp = mp->mnt_list.cqe_prev;
466 			vfs_unbusy(mp);
467 		}
468 	}
469 #ifdef DEBUG
470 	if (syncprt)
471 		vfs_bufstats();
472 #endif /* DEBUG */
473 	return (0);
474 }
475 
476 /*
477  * Change filesystem quotas.
478  */
479 /* ARGSUSED */
480 int
481 sys_quotactl(p, v, retval)
482 	struct proc *p;
483 	void *v;
484 	register_t *retval;
485 {
486 	register struct sys_quotactl_args /* {
487 		syscallarg(const char *) path;
488 		syscallarg(int) cmd;
489 		syscallarg(int) uid;
490 		syscallarg(caddr_t) arg;
491 	} */ *uap = v;
492 	register struct mount *mp;
493 	int error;
494 	struct nameidata nd;
495 
496 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
497 	if ((error = namei(&nd)) != 0)
498 		return (error);
499 	mp = nd.ni_vp->v_mount;
500 	vrele(nd.ni_vp);
501 	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
502 	    SCARG(uap, arg), p));
503 }
504 
505 /*
506  * Get filesystem statistics.
507  */
508 /* ARGSUSED */
509 int
510 sys_statfs(p, v, retval)
511 	struct proc *p;
512 	void *v;
513 	register_t *retval;
514 {
515 	register struct sys_statfs_args /* {
516 		syscallarg(const char *) path;
517 		syscallarg(struct statfs *) buf;
518 	} */ *uap = v;
519 	register struct mount *mp;
520 	register struct statfs *sp;
521 	int error;
522 	struct nameidata nd;
523 
524 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
525 	if ((error = namei(&nd)) != 0)
526 		return (error);
527 	mp = nd.ni_vp->v_mount;
528 	sp = &mp->mnt_stat;
529 	vrele(nd.ni_vp);
530 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
531 		return (error);
532 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
533 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
534 }
535 
536 /*
537  * Get filesystem statistics.
538  */
539 /* ARGSUSED */
540 int
541 sys_fstatfs(p, v, retval)
542 	struct proc *p;
543 	void *v;
544 	register_t *retval;
545 {
546 	register struct sys_fstatfs_args /* {
547 		syscallarg(int) fd;
548 		syscallarg(struct statfs *) buf;
549 	} */ *uap = v;
550 	struct file *fp;
551 	struct mount *mp;
552 	register struct statfs *sp;
553 	int error;
554 
555 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
556 		return (error);
557 	mp = ((struct vnode *)fp->f_data)->v_mount;
558 	sp = &mp->mnt_stat;
559 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
560 		return (error);
561 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
562 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
563 }
564 
565 /*
566  * Get statistics on all filesystems.
567  */
568 int
569 sys_getfsstat(p, v, retval)
570 	struct proc *p;
571 	void *v;
572 	register_t *retval;
573 {
574 	register struct sys_getfsstat_args /* {
575 		syscallarg(struct statfs *) buf;
576 		syscallarg(long) bufsize;
577 		syscallarg(int) flags;
578 	} */ *uap = v;
579 	register struct mount *mp, *nmp;
580 	register struct statfs *sp;
581 	caddr_t sfsp;
582 	long count, maxcount, error;
583 
584 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
585 	sfsp = (caddr_t)SCARG(uap, buf);
586 	for (count = 0,
587 	     mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
588 		nmp = mp->mnt_list.cqe_next;
589 		if (sfsp && count < maxcount &&
590 		    ((mp->mnt_flag & MNT_MLOCK) == 0)) {
591 			sp = &mp->mnt_stat;
592 			/*
593 			 * If MNT_NOWAIT is specified, do not refresh the
594 			 * fsstat cache. MNT_WAIT overrides MNT_NOWAIT.
595 			 */
596 			if (((SCARG(uap, flags) & MNT_NOWAIT) == 0 ||
597 			    (SCARG(uap, flags) & MNT_WAIT)) &&
598 			    (error = VFS_STATFS(mp, sp, p)))
599 				continue;
600 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
601 			error = copyout(sp, sfsp, sizeof(*sp));
602 			if (error)
603 				return (error);
604 			sfsp += sizeof(*sp);
605 		}
606 		count++;
607 	}
608 	if (sfsp && count > maxcount)
609 		*retval = maxcount;
610 	else
611 		*retval = count;
612 	return (0);
613 }
614 
615 /*
616  * Change current working directory to a given file descriptor.
617  */
618 /* ARGSUSED */
619 int
620 sys_fchdir(p, v, retval)
621 	struct proc *p;
622 	void *v;
623 	register_t *retval;
624 {
625 	struct sys_fchdir_args /* {
626 		syscallarg(int) fd;
627 	} */ *uap = v;
628 	register struct filedesc *fdp = p->p_fd;
629 	struct vnode *vp, *tdp;
630 	struct mount *mp;
631 	struct file *fp;
632 	int error;
633 
634 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
635 		return (error);
636 	vp = (struct vnode *)fp->f_data;
637 	VREF(vp);
638 	VOP_LOCK(vp);
639 	if (vp->v_type != VDIR)
640 		error = ENOTDIR;
641 	else
642 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
643 	while (!error && (mp = vp->v_mountedhere) != NULL) {
644 		if (mp->mnt_flag & MNT_MLOCK) {
645 			mp->mnt_flag |= MNT_MWAIT;
646 			sleep((caddr_t)mp, PVFS);
647 			continue;
648 		}
649 		if ((error = VFS_ROOT(mp, &tdp)) != 0)
650 			break;
651 		vput(vp);
652 		vp = tdp;
653 	}
654 	VOP_UNLOCK(vp);
655 	if (error) {
656 		vrele(vp);
657 		return (error);
658 	}
659 	vrele(fdp->fd_cdir);
660 	fdp->fd_cdir = vp;
661 	return (0);
662 }
663 
664 /*
665  * Change current working directory (``.'').
666  */
667 /* ARGSUSED */
668 int
669 sys_chdir(p, v, retval)
670 	struct proc *p;
671 	void *v;
672 	register_t *retval;
673 {
674 	struct sys_chdir_args /* {
675 		syscallarg(const char *) path;
676 	} */ *uap = v;
677 	register struct filedesc *fdp = p->p_fd;
678 	int error;
679 	struct nameidata nd;
680 
681 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
682 	    SCARG(uap, path), p);
683 	if ((error = change_dir(&nd, p)) != 0)
684 		return (error);
685 	vrele(fdp->fd_cdir);
686 	fdp->fd_cdir = nd.ni_vp;
687 	return (0);
688 }
689 
690 /*
691  * Change notion of root (``/'') directory.
692  */
693 /* ARGSUSED */
694 int
695 sys_chroot(p, v, retval)
696 	struct proc *p;
697 	void *v;
698 	register_t *retval;
699 {
700 	struct sys_chroot_args /* {
701 		syscallarg(const char *) path;
702 	} */ *uap = v;
703 	register struct filedesc *fdp = p->p_fd;
704 	int error;
705 	struct nameidata nd;
706 
707 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
708 		return (error);
709 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
710 	    SCARG(uap, path), p);
711 	if ((error = change_dir(&nd, p)) != 0)
712 		return (error);
713 	if (fdp->fd_rdir != NULL)
714 		vrele(fdp->fd_rdir);
715 	fdp->fd_rdir = nd.ni_vp;
716 	return (0);
717 }
718 
719 /*
720  * Common routine for chroot and chdir.
721  */
722 static int
723 change_dir(ndp, p)
724 	register struct nameidata *ndp;
725 	struct proc *p;
726 {
727 	struct vnode *vp;
728 	int error;
729 
730 	if ((error = namei(ndp)) != 0)
731 		return (error);
732 	vp = ndp->ni_vp;
733 	if (vp->v_type != VDIR)
734 		error = ENOTDIR;
735 	else
736 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
737 	VOP_UNLOCK(vp);
738 	if (error)
739 		vrele(vp);
740 	return (error);
741 }
742 
743 /*
744  * Check permissions, allocate an open file structure,
745  * and call the device open routine if any.
746  */
747 int
748 sys_open(p, v, retval)
749 	struct proc *p;
750 	void *v;
751 	register_t *retval;
752 {
753 	register struct sys_open_args /* {
754 		syscallarg(const char *) path;
755 		syscallarg(int) flags;
756 		syscallarg(int) mode;
757 	} */ *uap = v;
758 	register struct filedesc *fdp = p->p_fd;
759 	register struct file *fp;
760 	register struct vnode *vp;
761 	int flags, cmode;
762 	struct file *nfp;
763 	int type, indx, error;
764 	struct flock lf;
765 	struct nameidata nd;
766 	extern struct fileops vnops;
767 
768 	if ((error = falloc(p, &nfp, &indx)) != 0)
769 		return (error);
770 	fp = nfp;
771 	flags = FFLAGS(SCARG(uap, flags));
772 	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
773 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
774 	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
775 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
776 		ffree(fp);
777 		if ((error == ENODEV || error == ENXIO) &&
778 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
779 		    (error =
780 			dupfdopen(fdp, indx, p->p_dupfd, flags, error)) == 0) {
781 			*retval = indx;
782 			return (0);
783 		}
784 		if (error == ERESTART)
785 			error = EINTR;
786 		fdp->fd_ofiles[indx] = NULL;
787 		return (error);
788 	}
789 	p->p_dupfd = 0;
790 	vp = nd.ni_vp;
791 	fp->f_flag = flags & FMASK;
792 	fp->f_type = DTYPE_VNODE;
793 	fp->f_ops = &vnops;
794 	fp->f_data = (caddr_t)vp;
795 	if (flags & (O_EXLOCK | O_SHLOCK)) {
796 		lf.l_whence = SEEK_SET;
797 		lf.l_start = 0;
798 		lf.l_len = 0;
799 		if (flags & O_EXLOCK)
800 			lf.l_type = F_WRLCK;
801 		else
802 			lf.l_type = F_RDLCK;
803 		type = F_FLOCK;
804 		if ((flags & FNONBLOCK) == 0)
805 			type |= F_WAIT;
806 		VOP_UNLOCK(vp);
807 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
808 		if (error) {
809 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
810 			ffree(fp);
811 			fdp->fd_ofiles[indx] = NULL;
812 			return (error);
813 		}
814 		VOP_LOCK(vp);
815 		fp->f_flag |= FHASLOCK;
816 	}
817 	VOP_UNLOCK(vp);
818 	*retval = indx;
819 	return (0);
820 }
821 
822 /*
823  * Create a special file.
824  */
825 /* ARGSUSED */
826 int
827 sys_mknod(p, v, retval)
828 	struct proc *p;
829 	void *v;
830 	register_t *retval;
831 {
832 	register struct sys_mknod_args /* {
833 		syscallarg(const char *) path;
834 		syscallarg(int) mode;
835 		syscallarg(int) dev;
836 	} */ *uap = v;
837 	register struct vnode *vp;
838 	struct vattr vattr;
839 	int error;
840 	int whiteout = 0;
841 	struct nameidata nd;
842 
843 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
844 		return (error);
845 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
846 	if ((error = namei(&nd)) != 0)
847 		return (error);
848 	vp = nd.ni_vp;
849 	if (vp != NULL)
850 		error = EEXIST;
851 	else {
852 		VATTR_NULL(&vattr);
853 		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
854 		vattr.va_rdev = SCARG(uap, dev);
855 		whiteout = 0;
856 
857 		switch (SCARG(uap, mode) & S_IFMT) {
858 		case S_IFMT:	/* used by badsect to flag bad sectors */
859 			vattr.va_type = VBAD;
860 			break;
861 		case S_IFCHR:
862 			vattr.va_type = VCHR;
863 			break;
864 		case S_IFBLK:
865 			vattr.va_type = VBLK;
866 			break;
867 		case S_IFWHT:
868 			whiteout = 1;
869 			break;
870 		default:
871 			error = EINVAL;
872 			break;
873 		}
874 	}
875 	if (!error) {
876 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
877 		if (whiteout) {
878 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
879 			if (error)
880 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
881 			vput(nd.ni_dvp);
882 		} else {
883 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
884 						&nd.ni_cnd, &vattr);
885 		}
886 	} else {
887 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
888 		if (nd.ni_dvp == vp)
889 			vrele(nd.ni_dvp);
890 		else
891 			vput(nd.ni_dvp);
892 		if (vp)
893 			vrele(vp);
894 	}
895 	return (error);
896 }
897 
898 /*
899  * Create a named pipe.
900  */
901 /* ARGSUSED */
902 int
903 sys_mkfifo(p, v, retval)
904 	struct proc *p;
905 	void *v;
906 	register_t *retval;
907 {
908 #ifndef FIFO
909 	return (EOPNOTSUPP);
910 #else
911 	register struct sys_mkfifo_args /* {
912 		syscallarg(const char *) path;
913 		syscallarg(int) mode;
914 	} */ *uap = v;
915 	struct vattr vattr;
916 	int error;
917 	struct nameidata nd;
918 
919 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
920 	if ((error = namei(&nd)) != 0)
921 		return (error);
922 	if (nd.ni_vp != NULL) {
923 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
924 		if (nd.ni_dvp == nd.ni_vp)
925 			vrele(nd.ni_dvp);
926 		else
927 			vput(nd.ni_dvp);
928 		vrele(nd.ni_vp);
929 		return (EEXIST);
930 	}
931 	VATTR_NULL(&vattr);
932 	vattr.va_type = VFIFO;
933 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
934 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
935 	return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
936 #endif /* FIFO */
937 }
938 
939 /*
940  * Make a hard file link.
941  */
942 /* ARGSUSED */
943 int
944 sys_link(p, v, retval)
945 	struct proc *p;
946 	void *v;
947 	register_t *retval;
948 {
949 	register struct sys_link_args /* {
950 		syscallarg(const char *) path;
951 		syscallarg(const char *) link;
952 	} */ *uap = v;
953 	register struct vnode *vp;
954 	struct nameidata nd;
955 	int error;
956 
957 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
958 	if ((error = namei(&nd)) != 0)
959 		return (error);
960 	vp = nd.ni_vp;
961 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
962 	if ((error = namei(&nd)) != 0)
963 		goto out;
964 	if (nd.ni_vp) {
965 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
966 		if (nd.ni_dvp == nd.ni_vp)
967 			vrele(nd.ni_dvp);
968 		else
969 			vput(nd.ni_dvp);
970 		vrele(nd.ni_vp);
971 		error = EEXIST;
972 		goto out;
973 	}
974 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
975 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
976 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
977 out:
978 	vrele(vp);
979 	return (error);
980 }
981 
982 /*
983  * Make a symbolic link.
984  */
985 /* ARGSUSED */
986 int
987 sys_symlink(p, v, retval)
988 	struct proc *p;
989 	void *v;
990 	register_t *retval;
991 {
992 	register struct sys_symlink_args /* {
993 		syscallarg(const char *) path;
994 		syscallarg(const char *) link;
995 	} */ *uap = v;
996 	struct vattr vattr;
997 	char *path;
998 	int error;
999 	struct nameidata nd;
1000 
1001 	MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1002 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1003 	if (error)
1004 		goto out;
1005 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1006 	if ((error = namei(&nd)) != 0)
1007 		goto out;
1008 	if (nd.ni_vp) {
1009 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1010 		if (nd.ni_dvp == nd.ni_vp)
1011 			vrele(nd.ni_dvp);
1012 		else
1013 			vput(nd.ni_dvp);
1014 		vrele(nd.ni_vp);
1015 		error = EEXIST;
1016 		goto out;
1017 	}
1018 	VATTR_NULL(&vattr);
1019 	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1020 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1021 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1022 out:
1023 	FREE(path, M_NAMEI);
1024 	return (error);
1025 }
1026 
1027 /*
1028  * Delete a whiteout from the filesystem.
1029  */
1030 /* ARGSUSED */
1031 int
1032 sys_undelete(p, v, retval)
1033 	struct proc *p;
1034 	void *v;
1035 	register_t *retval;
1036 {
1037 	register struct sys_undelete_args /* {
1038 		syscallarg(const char *) path;
1039 	} */ *uap = v;
1040 	int error;
1041 	struct nameidata nd;
1042 
1043 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1044 	    SCARG(uap, path), p);
1045 	error = namei(&nd);
1046 	if (error)
1047 		return (error);
1048 
1049 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1050 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1051 		if (nd.ni_dvp == nd.ni_vp)
1052 			vrele(nd.ni_dvp);
1053 		else
1054 			vput(nd.ni_dvp);
1055 		if (nd.ni_vp)
1056 			vrele(nd.ni_vp);
1057 		return (EEXIST);
1058 	}
1059 
1060 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1061 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1062 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1063 	vput(nd.ni_dvp);
1064 	return (error);
1065 }
1066 
1067 /*
1068  * Delete a name from the filesystem.
1069  */
1070 /* ARGSUSED */
1071 int
1072 sys_unlink(p, v, retval)
1073 	struct proc *p;
1074 	void *v;
1075 	register_t *retval;
1076 {
1077 	struct sys_unlink_args /* {
1078 		syscallarg(const char *) path;
1079 	} */ *uap = v;
1080 	register struct vnode *vp;
1081 	int error;
1082 	struct nameidata nd;
1083 
1084 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1085 	    SCARG(uap, path), p);
1086 	if ((error = namei(&nd)) != 0)
1087 		return (error);
1088 	vp = nd.ni_vp;
1089 
1090 	/*
1091 	 * The root of a mounted filesystem cannot be deleted.
1092 	 */
1093 	if (vp->v_flag & VROOT) {
1094 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1095 		if (nd.ni_dvp == vp)
1096 			vrele(nd.ni_dvp);
1097 		else
1098 			vput(nd.ni_dvp);
1099 		vput(vp);
1100 		error = EBUSY;
1101 		goto out;
1102 	}
1103 
1104 	(void)vnode_pager_uncache(vp);
1105 
1106 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1107 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1108 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1109 out:
1110 	return (error);
1111 }
1112 
1113 /*
1114  * Reposition read/write file offset.
1115  */
1116 int
1117 sys_lseek(p, v, retval)
1118 	struct proc *p;
1119 	void *v;
1120 	register_t *retval;
1121 {
1122 	register struct sys_lseek_args /* {
1123 		syscallarg(int) fd;
1124 		syscallarg(int) pad;
1125 		syscallarg(off_t) offset;
1126 		syscallarg(int) whence;
1127 	} */ *uap = v;
1128 	struct ucred *cred = p->p_ucred;
1129 	register struct filedesc *fdp = p->p_fd;
1130 	register struct file *fp;
1131 	struct vnode *vp;
1132 	struct vattr vattr;
1133 	register off_t newoff;
1134 	int error;
1135 
1136 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1137 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
1138 		return (EBADF);
1139 
1140 	vp = (struct vnode *)fp->f_data;
1141 	if (fp->f_type != DTYPE_VNODE
1142 #ifdef FIFO
1143 	    || vp->v_type == VFIFO
1144 #endif
1145 	)
1146 		return (ESPIPE);
1147 
1148 	switch (SCARG(uap, whence)) {
1149 	case L_INCR:
1150 		newoff = fp->f_offset + SCARG(uap, offset);
1151 		break;
1152 	case L_XTND:
1153 		error = VOP_GETATTR(vp, &vattr, cred, p);
1154 		if (error)
1155 			return (error);
1156 		newoff = SCARG(uap, offset) + vattr.va_size;
1157 		break;
1158 	case L_SET:
1159 		newoff = SCARG(uap, offset);
1160 		break;
1161 	default:
1162 		return (EINVAL);
1163 	}
1164 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1165 		return (error);
1166 
1167 	*(off_t *)retval = fp->f_offset = newoff;
1168 	return (0);
1169 }
1170 
1171 /*
1172  * Check access permissions.
1173  */
1174 int
1175 sys_access(p, v, retval)
1176 	struct proc *p;
1177 	void *v;
1178 	register_t *retval;
1179 {
1180 	register struct sys_access_args /* {
1181 		syscallarg(const char *) path;
1182 		syscallarg(int) flags;
1183 	} */ *uap = v;
1184 	register struct ucred *cred = p->p_ucred;
1185 	register struct vnode *vp;
1186 	int error, flags, t_gid, t_uid;
1187 	struct nameidata nd;
1188 
1189 	t_uid = cred->cr_uid;
1190 	t_gid = cred->cr_gid;
1191 	cred->cr_uid = p->p_cred->p_ruid;
1192 	cred->cr_gid = p->p_cred->p_rgid;
1193 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1194 	    SCARG(uap, path), p);
1195 	if ((error = namei(&nd)) != 0)
1196 		goto out1;
1197 	vp = nd.ni_vp;
1198 
1199 	/* Flags == 0 means only check for existence. */
1200 	if (SCARG(uap, flags)) {
1201 		flags = 0;
1202 		if (SCARG(uap, flags) & R_OK)
1203 			flags |= VREAD;
1204 		if (SCARG(uap, flags) & W_OK)
1205 			flags |= VWRITE;
1206 		if (SCARG(uap, flags) & X_OK)
1207 			flags |= VEXEC;
1208 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1209 			error = VOP_ACCESS(vp, flags, cred, p);
1210 	}
1211 	vput(vp);
1212 out1:
1213 	cred->cr_uid = t_uid;
1214 	cred->cr_gid = t_gid;
1215 	return (error);
1216 }
1217 
1218 /*
1219  * Get file status; this version follows links.
1220  */
1221 /* ARGSUSED */
1222 int
1223 sys_stat(p, v, retval)
1224 	struct proc *p;
1225 	void *v;
1226 	register_t *retval;
1227 {
1228 	register struct sys_stat_args /* {
1229 		syscallarg(const char *) path;
1230 		syscallarg(struct stat *) ub;
1231 	} */ *uap = v;
1232 	struct stat sb;
1233 	int error;
1234 	struct nameidata nd;
1235 
1236 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1237 	    SCARG(uap, path), p);
1238 	if ((error = namei(&nd)) != 0)
1239 		return (error);
1240 	error = vn_stat(nd.ni_vp, &sb, p);
1241 	vput(nd.ni_vp);
1242 	if (error)
1243 		return (error);
1244 	error = copyout(&sb, SCARG(uap, ub), sizeof (sb));
1245 	return (error);
1246 }
1247 
1248 /*
1249  * Get file status; this version does not follow links.
1250  */
1251 /* ARGSUSED */
1252 int
1253 sys_lstat(p, v, retval)
1254 	struct proc *p;
1255 	void *v;
1256 	register_t *retval;
1257 {
1258 	register struct sys_lstat_args /* {
1259 		syscallarg(const char *) path;
1260 		syscallarg(struct stat *) ub;
1261 	} */ *uap = v;
1262 	struct stat sb;
1263 	int error;
1264 	struct nameidata nd;
1265 
1266 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1267 	    SCARG(uap, path), p);
1268 	if ((error = namei(&nd)) != 0)
1269 		return (error);
1270 	error = vn_stat(nd.ni_vp, &sb, p);
1271 	vput(nd.ni_vp);
1272 	if (error)
1273 		return (error);
1274 	error = copyout(&sb, SCARG(uap, ub), sizeof (sb));
1275 	return (error);
1276 }
1277 
1278 /*
1279  * Get configurable pathname variables.
1280  */
1281 /* ARGSUSED */
1282 int
1283 sys_pathconf(p, v, retval)
1284 	struct proc *p;
1285 	void *v;
1286 	register_t *retval;
1287 {
1288 	register struct sys_pathconf_args /* {
1289 		syscallarg(const char *) path;
1290 		syscallarg(int) name;
1291 	} */ *uap = v;
1292 	int error;
1293 	struct nameidata nd;
1294 
1295 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1296 	    SCARG(uap, path), p);
1297 	if ((error = namei(&nd)) != 0)
1298 		return (error);
1299 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1300 	vput(nd.ni_vp);
1301 	return (error);
1302 }
1303 
1304 /*
1305  * Return target name of a symbolic link.
1306  */
1307 /* ARGSUSED */
1308 int
1309 sys_readlink(p, v, retval)
1310 	struct proc *p;
1311 	void *v;
1312 	register_t *retval;
1313 {
1314 	register struct sys_readlink_args /* {
1315 		syscallarg(const char *) path;
1316 		syscallarg(char *) buf;
1317 		syscallarg(int) count;
1318 	} */ *uap = v;
1319 	register struct vnode *vp;
1320 	struct iovec aiov;
1321 	struct uio auio;
1322 	int error;
1323 	struct nameidata nd;
1324 
1325 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1326 	    SCARG(uap, path), p);
1327 	if ((error = namei(&nd)) != 0)
1328 		return (error);
1329 	vp = nd.ni_vp;
1330 	if (vp->v_type != VLNK)
1331 		error = EINVAL;
1332 	else {
1333 		aiov.iov_base = SCARG(uap, buf);
1334 		aiov.iov_len = SCARG(uap, count);
1335 		auio.uio_iov = &aiov;
1336 		auio.uio_iovcnt = 1;
1337 		auio.uio_offset = 0;
1338 		auio.uio_rw = UIO_READ;
1339 		auio.uio_segflg = UIO_USERSPACE;
1340 		auio.uio_procp = p;
1341 		auio.uio_resid = SCARG(uap, count);
1342 		error = VOP_READLINK(vp, &auio, p->p_ucred);
1343 	}
1344 	vput(vp);
1345 	*retval = SCARG(uap, count) - auio.uio_resid;
1346 	return (error);
1347 }
1348 
1349 /*
1350  * Change flags of a file given a path name.
1351  */
1352 /* ARGSUSED */
1353 int
1354 sys_chflags(p, v, retval)
1355 	struct proc *p;
1356 	void *v;
1357 	register_t *retval;
1358 {
1359 	register struct sys_chflags_args /* {
1360 		syscallarg(const char *) path;
1361 		syscallarg(u_long) flags;
1362 	} */ *uap = v;
1363 	register struct vnode *vp;
1364 	struct vattr vattr;
1365 	int error;
1366 	struct nameidata nd;
1367 
1368 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1369 	if ((error = namei(&nd)) != 0)
1370 		return (error);
1371 	vp = nd.ni_vp;
1372 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1373 	VOP_LOCK(vp);
1374 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1375 		error = EROFS;
1376 	else {
1377 		VATTR_NULL(&vattr);
1378 		vattr.va_flags = SCARG(uap, flags);
1379 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1380 	}
1381 	vput(vp);
1382 	return (error);
1383 }
1384 
1385 /*
1386  * Change flags of a file given a file descriptor.
1387  */
1388 /* ARGSUSED */
1389 int
1390 sys_fchflags(p, v, retval)
1391 	struct proc *p;
1392 	void *v;
1393 	register_t *retval;
1394 {
1395 	register struct sys_fchflags_args /* {
1396 		syscallarg(int) fd;
1397 		syscallarg(u_long) flags;
1398 	} */ *uap = v;
1399 	struct vattr vattr;
1400 	struct vnode *vp;
1401 	struct file *fp;
1402 	int error;
1403 
1404 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1405 		return (error);
1406 	vp = (struct vnode *)fp->f_data;
1407 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1408 	VOP_LOCK(vp);
1409 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1410 		error = EROFS;
1411 	else {
1412 		VATTR_NULL(&vattr);
1413 		vattr.va_flags = SCARG(uap, flags);
1414 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1415 	}
1416 	VOP_UNLOCK(vp);
1417 	return (error);
1418 }
1419 
1420 /*
1421  * Change mode of a file given path name.
1422  */
1423 /* ARGSUSED */
1424 int
1425 sys_chmod(p, v, retval)
1426 	struct proc *p;
1427 	void *v;
1428 	register_t *retval;
1429 {
1430 	register struct sys_chmod_args /* {
1431 		syscallarg(const char *) path;
1432 		syscallarg(int) mode;
1433 	} */ *uap = v;
1434 	register struct vnode *vp;
1435 	struct vattr vattr;
1436 	int error;
1437 	struct nameidata nd;
1438 
1439 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1440 	if ((error = namei(&nd)) != 0)
1441 		return (error);
1442 	vp = nd.ni_vp;
1443 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1444 	VOP_LOCK(vp);
1445 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1446 		error = EROFS;
1447 	else {
1448 		VATTR_NULL(&vattr);
1449 		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1450 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1451 	}
1452 	vput(vp);
1453 	return (error);
1454 }
1455 
1456 /*
1457  * Change mode of a file given a file descriptor.
1458  */
1459 /* ARGSUSED */
1460 int
1461 sys_fchmod(p, v, retval)
1462 	struct proc *p;
1463 	void *v;
1464 	register_t *retval;
1465 {
1466 	register struct sys_fchmod_args /* {
1467 		syscallarg(int) fd;
1468 		syscallarg(int) mode;
1469 	} */ *uap = v;
1470 	struct vattr vattr;
1471 	struct vnode *vp;
1472 	struct file *fp;
1473 	int error;
1474 
1475 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1476 		return (error);
1477 	vp = (struct vnode *)fp->f_data;
1478 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1479 	VOP_LOCK(vp);
1480 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1481 		error = EROFS;
1482 	else {
1483 		VATTR_NULL(&vattr);
1484 		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1485 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1486 	}
1487 	VOP_UNLOCK(vp);
1488 	return (error);
1489 }
1490 
1491 /*
1492  * Set ownership given a path name.
1493  */
1494 /* ARGSUSED */
1495 int
1496 sys_chown(p, v, retval)
1497 	struct proc *p;
1498 	void *v;
1499 	register_t *retval;
1500 {
1501 	register struct sys_chown_args /* {
1502 		syscallarg(const char *) path;
1503 		syscallarg(uid_t) uid;
1504 		syscallarg(gid_t) gid;
1505 	} */ *uap = v;
1506 	int error;
1507 	struct nameidata nd;
1508 
1509 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1510 	if ((error = namei(&nd)) != 0)
1511 		return (error);
1512 
1513 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p);
1514 
1515 	vrele(nd.ni_vp);
1516 	return (error);
1517 }
1518 
1519 /*
1520  * Set ownership given a file descriptor.
1521  */
1522 /* ARGSUSED */
1523 int
1524 sys_fchown(p, v, retval)
1525 	struct proc *p;
1526 	void *v;
1527 	register_t *retval;
1528 {
1529 	register struct sys_fchown_args /* {
1530 		syscallarg(int) fd;
1531 		syscallarg(uid_t) uid;
1532 		syscallarg(gid_t) gid;
1533 	} */ *uap = v;
1534 	int error;
1535 	struct file *fp;
1536 
1537 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1538 		return (error);
1539 
1540 	return (change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
1541 	    SCARG(uap, gid), p));
1542 }
1543 
1544 /*
1545  * Common routine to set ownership given a vnode.
1546  */
1547 static int
1548 change_owner(vp, uid, gid, p)
1549 	register struct vnode *vp;
1550 	uid_t uid;
1551 	gid_t gid;
1552 	struct proc *p;
1553 {
1554 	struct vattr vattr;
1555 	mode_t newmode = VNOVAL;
1556 	int error;
1557 
1558 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1559 	VOP_LOCK(vp);
1560 	if (vp->v_mount->mnt_flag & MNT_RDONLY) {
1561 		error = EROFS;
1562 		goto out;
1563 	}
1564 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)))
1565 		goto out;
1566 
1567 	/* Clear (S_ISUID | S_ISGID) bits: alter va_mode only if necessary. */
1568 	if (vattr.va_mode & (S_ISUID | S_ISGID))
1569 		newmode = vattr.va_mode & ~(S_ISUID | S_ISGID);
1570 
1571 	VATTR_NULL(&vattr);
1572 	vattr.va_uid = uid;
1573 	vattr.va_gid = gid;
1574 	vattr.va_mode = newmode;
1575 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1576 
1577 out:
1578 	VOP_UNLOCK(vp);
1579 	return (error);
1580 }
1581 
1582 /*
1583  * Set the access and modification times given a path name.
1584  */
1585 /* ARGSUSED */
1586 int
1587 sys_utimes(p, v, retval)
1588 	struct proc *p;
1589 	void *v;
1590 	register_t *retval;
1591 {
1592 	register struct sys_utimes_args /* {
1593 		syscallarg(const char *) path;
1594 		syscallarg(const struct timeval *) tptr;
1595 	} */ *uap = v;
1596 	register struct vnode *vp;
1597 	struct timeval tv[2];
1598 	struct vattr vattr;
1599 	int error;
1600 	struct nameidata nd;
1601 
1602 	VATTR_NULL(&vattr);
1603 	if (SCARG(uap, tptr) == NULL) {
1604 		microtime(&tv[0]);
1605 		tv[1] = tv[0];
1606 		vattr.va_vaflags |= VA_UTIMES_NULL;
1607 	} else {
1608 		error = copyin(SCARG(uap, tptr), tv, sizeof (tv));
1609 		if (error)
1610 			return (error);
1611 	}
1612 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1613 	if ((error = namei(&nd)) != 0)
1614 		return (error);
1615 	vp = nd.ni_vp;
1616 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1617 	VOP_LOCK(vp);
1618 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1619 		error = EROFS;
1620 	else {
1621 		vattr.va_atime.tv_sec = tv[0].tv_sec;
1622 		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
1623 		vattr.va_mtime.tv_sec = tv[1].tv_sec;
1624 		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
1625 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1626 	}
1627 	vput(vp);
1628 	return (error);
1629 }
1630 
1631 /*
1632  * Set the access and modification times given a file descriptor.
1633  */
1634 /* ARGSUSED */
1635 int
1636 sys_futimes(p, v, retval)
1637 	struct proc *p;
1638 	void *v;
1639 	register_t *retval;
1640 {
1641 	register struct sys_futimes_args /* {
1642 		syscallarg(int) fd;
1643 		syscallarg(const struct timeval *) tptr;
1644 	} */ *uap = v;
1645 	register struct vnode *vp;
1646 	struct timeval tv[2];
1647 	struct vattr vattr;
1648 	int error;
1649 	struct file *fp;
1650 
1651 	VATTR_NULL(&vattr);
1652 	if (SCARG(uap, tptr) == NULL) {
1653 		microtime(&tv[0]);
1654 		tv[1] = tv[0];
1655 		vattr.va_vaflags |= VA_UTIMES_NULL;
1656 	} else {
1657 		error = copyin(SCARG(uap, tptr), tv, sizeof (tv));
1658 		if (error)
1659 			return (error);
1660 	}
1661 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1662 		return (error);
1663 	vp = (struct vnode *)fp->f_data;
1664 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1665 	VOP_LOCK(vp);
1666 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1667 		error = EROFS;
1668 	else {
1669 		vattr.va_atime.tv_sec = tv[0].tv_sec;
1670 		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
1671 		vattr.va_mtime.tv_sec = tv[1].tv_sec;
1672 		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
1673 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1674 	}
1675 	VOP_UNLOCK(vp);
1676 	return (error);
1677 }
1678 
1679 /*
1680  * Truncate a file given its path name.
1681  */
1682 /* ARGSUSED */
1683 int
1684 sys_truncate(p, v, retval)
1685 	struct proc *p;
1686 	void *v;
1687 	register_t *retval;
1688 {
1689 	register struct sys_truncate_args /* {
1690 		syscallarg(const char *) path;
1691 		syscallarg(int) pad;
1692 		syscallarg(off_t) length;
1693 	} */ *uap = v;
1694 	register struct vnode *vp;
1695 	struct vattr vattr;
1696 	int error;
1697 	struct nameidata nd;
1698 
1699 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1700 	if ((error = namei(&nd)) != 0)
1701 		return (error);
1702 	vp = nd.ni_vp;
1703 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1704 	VOP_LOCK(vp);
1705 	if (vp->v_type == VDIR)
1706 		error = EISDIR;
1707 	else if ((error = vn_writechk(vp)) == 0 &&
1708 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
1709 		VATTR_NULL(&vattr);
1710 		vattr.va_size = SCARG(uap, length);
1711 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1712 	}
1713 	vput(vp);
1714 	return (error);
1715 }
1716 
1717 /*
1718  * Truncate a file given a file descriptor.
1719  */
1720 /* ARGSUSED */
1721 int
1722 sys_ftruncate(p, v, retval)
1723 	struct proc *p;
1724 	void *v;
1725 	register_t *retval;
1726 {
1727 	register struct sys_ftruncate_args /* {
1728 		syscallarg(int) fd;
1729 		syscallarg(int) pad;
1730 		syscallarg(off_t) length;
1731 	} */ *uap = v;
1732 	struct vattr vattr;
1733 	struct vnode *vp;
1734 	struct file *fp;
1735 	int error;
1736 
1737 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1738 		return (error);
1739 	if ((fp->f_flag & FWRITE) == 0)
1740 		return (EINVAL);
1741 	vp = (struct vnode *)fp->f_data;
1742 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1743 	VOP_LOCK(vp);
1744 	if (vp->v_type == VDIR)
1745 		error = EISDIR;
1746 	else if ((error = vn_writechk(vp)) == 0) {
1747 		VATTR_NULL(&vattr);
1748 		vattr.va_size = SCARG(uap, length);
1749 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
1750 	}
1751 	VOP_UNLOCK(vp);
1752 	return (error);
1753 }
1754 
1755 /*
1756  * Sync an open file.
1757  */
1758 /* ARGSUSED */
1759 int
1760 sys_fsync(p, v, retval)
1761 	struct proc *p;
1762 	void *v;
1763 	register_t *retval;
1764 {
1765 	struct sys_fsync_args /* {
1766 		syscallarg(int) fd;
1767 	} */ *uap = v;
1768 	register struct vnode *vp;
1769 	struct file *fp;
1770 	int error;
1771 
1772 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1773 		return (error);
1774 	vp = (struct vnode *)fp->f_data;
1775 	VOP_LOCK(vp);
1776 	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
1777 	VOP_UNLOCK(vp);
1778 	return (error);
1779 }
1780 
1781 /*
1782  * Rename files, (standard) BSD semantics frontend.
1783  */
1784 /* ARGSUSED */
1785 int
1786 sys_rename(p, v, retval)
1787 	struct proc *p;
1788 	void *v;
1789 	register_t *retval;
1790 {
1791 	register struct sys_rename_args /* {
1792 		syscallarg(const char *) from;
1793 		syscallarg(const char *) to;
1794 	} */ *uap = v;
1795 
1796 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
1797 }
1798 
1799 /*
1800  * Rename files, POSIX semantics frontend.
1801  */
1802 /* ARGSUSED */
1803 int
1804 sys_posix_rename(p, v, retval)
1805 	struct proc *p;
1806 	void *v;
1807 	register_t *retval;
1808 {
1809 	register struct sys_posix_rename_args /* {
1810 		syscallarg(const char *) from;
1811 		syscallarg(const char *) to;
1812 	} */ *uap = v;
1813 
1814 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
1815 }
1816 
1817 /*
1818  * Rename files.  Source and destination must either both be directories,
1819  * or both not be directories.  If target is a directory, it must be empty.
1820  * If `from' and `to' refer to the same object, the value of the `retain'
1821  * argument is used to determine whether `from' will be
1822  *
1823  * (retain == 0)	deleted unless `from' and `to' refer to the same
1824  *			object in the file system's name space (BSD).
1825  * (retain == 1)	always retained (POSIX).
1826  */
1827 static int
1828 rename_files(from, to, p, retain)
1829 	const char *from, *to;
1830 	struct proc *p;
1831 	int retain;
1832 {
1833 	register struct vnode *tvp, *fvp, *tdvp;
1834 	struct nameidata fromnd, tond;
1835 	int error;
1836 
1837 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
1838 	    from, p);
1839 	if ((error = namei(&fromnd)) != 0)
1840 		return (error);
1841 	fvp = fromnd.ni_vp;
1842 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
1843 	    UIO_USERSPACE, to, p);
1844 	if ((error = namei(&tond)) != 0) {
1845 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1846 		vrele(fromnd.ni_dvp);
1847 		vrele(fvp);
1848 		goto out1;
1849 	}
1850 	tdvp = tond.ni_dvp;
1851 	tvp = tond.ni_vp;
1852 
1853 	if (tvp != NULL) {
1854 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1855 			error = ENOTDIR;
1856 			goto out;
1857 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1858 			error = EISDIR;
1859 			goto out;
1860 		}
1861 	}
1862 
1863 	if (fvp == tdvp)
1864 		error = EINVAL;
1865 
1866 	/*
1867 	 * Source and destination refer to the same object.
1868 	 */
1869 	if (fvp == tvp) {
1870 		if (retain)
1871 			error = -1;
1872 		else if (fromnd.ni_dvp == tdvp &&
1873 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1874 		    !bcmp(fromnd.ni_cnd.cn_nameptr,
1875 		          tond.ni_cnd.cn_nameptr,
1876 		          fromnd.ni_cnd.cn_namelen))
1877 		error = -1;
1878 	}
1879 
1880 out:
1881 	if (!error) {
1882 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
1883 		if (fromnd.ni_dvp != tdvp)
1884 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1885 		if (tvp) {
1886 			(void)vnode_pager_uncache(tvp);
1887 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
1888 		}
1889 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1890 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1891 	} else {
1892 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1893 		if (tdvp == tvp)
1894 			vrele(tdvp);
1895 		else
1896 			vput(tdvp);
1897 		if (tvp)
1898 			vput(tvp);
1899 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1900 		vrele(fromnd.ni_dvp);
1901 		vrele(fvp);
1902 	}
1903 	vrele(tond.ni_startdir);
1904 	FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1905 out1:
1906 	if (fromnd.ni_startdir)
1907 		vrele(fromnd.ni_startdir);
1908 	FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1909 	return (error == -1 ? 0 : error);
1910 }
1911 
1912 /*
1913  * Make a directory file.
1914  */
1915 /* ARGSUSED */
1916 int
1917 sys_mkdir(p, v, retval)
1918 	struct proc *p;
1919 	void *v;
1920 	register_t *retval;
1921 {
1922 	register struct sys_mkdir_args /* {
1923 		syscallarg(const char *) path;
1924 		syscallarg(int) mode;
1925 	} */ *uap = v;
1926 	register struct vnode *vp;
1927 	struct vattr vattr;
1928 	int error;
1929 	struct nameidata nd;
1930 
1931 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1932 	if ((error = namei(&nd)) != 0)
1933 		return (error);
1934 	vp = nd.ni_vp;
1935 	if (vp != NULL) {
1936 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1937 		if (nd.ni_dvp == vp)
1938 			vrele(nd.ni_dvp);
1939 		else
1940 			vput(nd.ni_dvp);
1941 		vrele(vp);
1942 		return (EEXIST);
1943 	}
1944 	VATTR_NULL(&vattr);
1945 	vattr.va_type = VDIR;
1946 	vattr.va_mode = (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_fd->fd_cmask;
1947 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1948 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1949 	if (!error)
1950 		vput(nd.ni_vp);
1951 	return (error);
1952 }
1953 
1954 /*
1955  * Remove a directory file.
1956  */
1957 /* ARGSUSED */
1958 int
1959 sys_rmdir(p, v, retval)
1960 	struct proc *p;
1961 	void *v;
1962 	register_t *retval;
1963 {
1964 	struct sys_rmdir_args /* {
1965 		syscallarg(const char *) path;
1966 	} */ *uap = v;
1967 	register struct vnode *vp;
1968 	int error;
1969 	struct nameidata nd;
1970 
1971 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1972 	    SCARG(uap, path), p);
1973 	if ((error = namei(&nd)) != 0)
1974 		return (error);
1975 	vp = nd.ni_vp;
1976 	if (vp->v_type != VDIR) {
1977 		error = ENOTDIR;
1978 		goto out;
1979 	}
1980 	/*
1981 	 * No rmdir "." please.
1982 	 */
1983 	if (nd.ni_dvp == vp) {
1984 		error = EINVAL;
1985 		goto out;
1986 	}
1987 	/*
1988 	 * The root of a mounted filesystem cannot be deleted.
1989 	 */
1990 	if (vp->v_flag & VROOT)
1991 		error = EBUSY;
1992 out:
1993 	if (!error) {
1994 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1995 		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1996 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1997 	} else {
1998 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1999 		if (nd.ni_dvp == vp)
2000 			vrele(nd.ni_dvp);
2001 		else
2002 			vput(nd.ni_dvp);
2003 		vput(vp);
2004 	}
2005 	return (error);
2006 }
2007 
2008 /*
2009  * Read a block of directory entries in a file system independent format.
2010  */
2011 int
2012 sys_getdirentries(p, v, retval)
2013 	struct proc *p;
2014 	void *v;
2015 	register_t *retval;
2016 {
2017 	register struct sys_getdirentries_args /* {
2018 		syscallarg(int) fd;
2019 		syscallarg(char *) buf;
2020 		syscallarg(u_int) count;
2021 		syscallarg(long *) basep;
2022 	} */ *uap = v;
2023 	register struct vnode *vp;
2024 	struct file *fp;
2025 	struct uio auio;
2026 	struct iovec aiov;
2027 	long loff;
2028 	int error, eofflag;
2029 
2030 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2031 		return (error);
2032 	if ((fp->f_flag & FREAD) == 0)
2033 		return (EBADF);
2034 	vp = (struct vnode *)fp->f_data;
2035 unionread:
2036 	if (vp->v_type != VDIR)
2037 		return (EINVAL);
2038 	aiov.iov_base = SCARG(uap, buf);
2039 	aiov.iov_len = SCARG(uap, count);
2040 	auio.uio_iov = &aiov;
2041 	auio.uio_iovcnt = 1;
2042 	auio.uio_rw = UIO_READ;
2043 	auio.uio_segflg = UIO_USERSPACE;
2044 	auio.uio_procp = p;
2045 	auio.uio_resid = SCARG(uap, count);
2046 	VOP_LOCK(vp);
2047 	loff = auio.uio_offset = fp->f_offset;
2048 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, (u_long *)0, 0);
2049 	fp->f_offset = auio.uio_offset;
2050 	VOP_UNLOCK(vp);
2051 	if (error)
2052 		return (error);
2053 
2054 #ifdef UNION
2055 {
2056 	extern int (**union_vnodeop_p) __P((void *));
2057 	extern struct vnode *union_dircache __P((struct vnode *));
2058 
2059 	if ((SCARG(uap, count) == auio.uio_resid) &&
2060 	    (vp->v_op == union_vnodeop_p)) {
2061 		struct vnode *lvp;
2062 
2063 		lvp = union_dircache(vp);
2064 		if (lvp != NULLVP) {
2065 			struct vattr va;
2066 
2067 			/*
2068 			 * If the directory is opaque,
2069 			 * then don't show lower entries
2070 			 */
2071 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
2072 			if (va.va_flags & OPAQUE) {
2073 				vput(lvp);
2074 				lvp = NULL;
2075 			}
2076 		}
2077 
2078 		if (lvp != NULLVP) {
2079 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
2080 			VOP_UNLOCK(lvp);
2081 
2082 			if (error) {
2083 				vrele(lvp);
2084 				return (error);
2085 			}
2086 			fp->f_data = (caddr_t) lvp;
2087 			fp->f_offset = 0;
2088 			error = vn_close(vp, FREAD, fp->f_cred, p);
2089 			if (error)
2090 				return (error);
2091 			vp = lvp;
2092 			goto unionread;
2093 		}
2094 	}
2095 }
2096 #endif /* UNION */
2097 
2098 	if ((SCARG(uap, count) == auio.uio_resid) &&
2099 	    (vp->v_flag & VROOT) &&
2100 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
2101 		struct vnode *tvp = vp;
2102 		vp = vp->v_mount->mnt_vnodecovered;
2103 		VREF(vp);
2104 		fp->f_data = (caddr_t) vp;
2105 		fp->f_offset = 0;
2106 		vrele(tvp);
2107 		goto unionread;
2108 	}
2109 	error = copyout(&loff, SCARG(uap, basep), sizeof(long));
2110 	*retval = SCARG(uap, count) - auio.uio_resid;
2111 	return (error);
2112 }
2113 
2114 /*
2115  * Set the mode mask for creation of filesystem nodes.
2116  */
2117 int
2118 sys_umask(p, v, retval)
2119 	struct proc *p;
2120 	void *v;
2121 	register_t *retval;
2122 {
2123 	struct sys_umask_args /* {
2124 		syscallarg(int) newmask;
2125 	} */ *uap = v;
2126 	register struct filedesc *fdp;
2127 
2128 	fdp = p->p_fd;
2129 	*retval = fdp->fd_cmask;
2130 	fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
2131 	return (0);
2132 }
2133 
2134 /*
2135  * Void all references to file by ripping underlying filesystem
2136  * away from vnode.
2137  */
2138 /* ARGSUSED */
2139 int
2140 sys_revoke(p, v, retval)
2141 	struct proc *p;
2142 	void *v;
2143 	register_t *retval;
2144 {
2145 	register struct sys_revoke_args /* {
2146 		syscallarg(const char *) path;
2147 	} */ *uap = v;
2148 	register struct vnode *vp;
2149 	struct vattr vattr;
2150 	int error;
2151 	struct nameidata nd;
2152 
2153 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2154 	if ((error = namei(&nd)) != 0)
2155 		return (error);
2156 	vp = nd.ni_vp;
2157 	if (vp->v_type != VCHR && vp->v_type != VBLK) {
2158 		error = EINVAL;
2159 		goto out;
2160 	}
2161 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2162 		goto out;
2163 	if (p->p_ucred->cr_uid != vattr.va_uid &&
2164 	    (error = suser(p->p_ucred, &p->p_acflag)))
2165 		goto out;
2166 	if (vp->v_usecount > 1 || (vp->v_flag & VALIASED))
2167 		vgoneall(vp);
2168 out:
2169 	vrele(vp);
2170 	return (error);
2171 }
2172 
2173 /*
2174  * Convert a user file descriptor to a kernel file entry.
2175  */
2176 int
2177 getvnode(fdp, fd, fpp)
2178 	struct filedesc *fdp;
2179 	int fd;
2180 	struct file **fpp;
2181 {
2182 	struct vnode *vp;
2183 	struct file *fp;
2184 
2185 	if ((u_int)fd >= fdp->fd_nfiles ||
2186 	    (fp = fdp->fd_ofiles[fd]) == NULL)
2187 		return (EBADF);
2188 	if (fp->f_type != DTYPE_VNODE)
2189 		return (EINVAL);
2190 	vp = (struct vnode *)fp->f_data;
2191 	if (vp->v_type == VBAD)
2192 		return (EBADF);
2193 	*fpp = fp;
2194 	return (0);
2195 }
2196