xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision d0fed6c87ddc40a8bffa6f99e7433ddfc864dd83)
1 /*	$NetBSD: vfs_syscalls.c,v 1.85 1997/04/11 22:08:28 kleink 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 
64 void checkdirs __P((struct vnode *));
65 int dounmount __P((struct mount *, int, struct proc *));
66 
67 /*
68  * Virtual File System System Calls
69  */
70 
71 /*
72  * Mount a file system.
73  */
74 /* ARGSUSED */
75 int
76 sys_mount(p, v, retval)
77 	struct proc *p;
78 	void *v;
79 	register_t *retval;
80 {
81 	register struct sys_mount_args /* {
82 		syscallarg(const char *) type;
83 		syscallarg(const char *) path;
84 		syscallarg(int) flags;
85 		syscallarg(void *) data;
86 	} */ *uap = v;
87 	register struct vnode *vp;
88 	register struct mount *mp;
89 	int error, flag = 0;
90 	u_long fsindex = 0;
91 	char fstypename[MFSNAMELEN];
92 	struct vattr va;
93 	struct nameidata nd;
94 
95 	/*
96 	 * Get vnode to be covered
97 	 */
98 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
99 	    SCARG(uap, path), p);
100 	if ((error = namei(&nd)) != 0)
101 		return (error);
102 	vp = nd.ni_vp;
103 	if (SCARG(uap, flags) & MNT_UPDATE) {
104 		if ((vp->v_flag & VROOT) == 0) {
105 			vput(vp);
106 			return (EINVAL);
107 		}
108 		mp = vp->v_mount;
109 		flag = mp->mnt_flag;
110 		/*
111 		 * We only allow the filesystem to be reloaded if it
112 		 * is currently mounted read-only.
113 		 */
114 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
115 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
116 			vput(vp);
117 			return (EOPNOTSUPP);	/* Needs translation */
118 		}
119 		mp->mnt_flag |=
120 		    SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
121 		/*
122 		 * Only root, or the user that did the original mount is
123 		 * permitted to update it.
124 		 */
125 		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
126 		    (error = suser(p->p_ucred, &p->p_acflag))) {
127 			vput(vp);
128 			return (error);
129 		}
130 		/*
131 		 * Do not allow NFS export by non-root users. Silently
132 		 * enforce MNT_NOSUID and MNT_NODEV for non-root users.
133 		 */
134 		if (p->p_ucred->cr_uid != 0) {
135 			if (SCARG(uap, flags) & MNT_EXPORTED) {
136 				vput(vp);
137 				return (EPERM);
138 			}
139 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
140 		}
141 		VOP_UNLOCK(vp);
142 		goto update;
143 	}
144 	/*
145 	 * If the user is not root, ensure that they own the directory
146 	 * onto which we are attempting to mount.
147 	 */
148 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) ||
149 	    (va.va_uid != p->p_ucred->cr_uid &&
150 	     (error = suser(p->p_ucred, &p->p_acflag)))) {
151 		vput(vp);
152 		return (error);
153 	}
154 	/*
155 	 * Do not allow NFS export by non-root users. Silently
156 	 * enforce MNT_NOSUID and MNT_NODEV for non-root users.
157 	 */
158 	if (p->p_ucred->cr_uid != 0) {
159 		if (SCARG(uap, flags) & MNT_EXPORTED) {
160 			vput(vp);
161 			return (EPERM);
162 		}
163 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
164 	}
165 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
166 		return (error);
167 	if (vp->v_type != VDIR) {
168 		vput(vp);
169 		return (ENOTDIR);
170 	}
171 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
172 	if (error) {
173 #if defined(COMPAT_09) || defined(COMPAT_43)
174 		/*
175 		 * Historically filesystem types were identified by number.
176 		 * If we get an integer for the filesystem type instead of a
177 		 * string, we check to see if it matches one of the historic
178 		 * filesystem types.
179 		 */
180 		fsindex = (u_long)SCARG(uap, type);
181 		if (fsindex >= nvfssw || vfssw[fsindex] == NULL) {
182 			vput(vp);
183 			return (ENODEV);
184 		}
185 		strncpy(fstypename, vfssw[fsindex]->vfs_name, MFSNAMELEN);
186 #else
187 		vput(vp);
188 		return (error);
189 #endif
190 	}
191 #ifdef	COMPAT_10
192 	/* Accept `ufs' as an alias for `ffs'. */
193 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
194 		strncpy(fstypename, "ffs", MFSNAMELEN);
195 #endif
196 	for (fsindex = 0; fsindex < nvfssw; fsindex++)
197 		if (vfssw[fsindex] != NULL &&
198 		    !strncmp(vfssw[fsindex]->vfs_name, fstypename, MFSNAMELEN))
199 			break;
200 	if (fsindex >= nvfssw) {
201 		vput(vp);
202 		return (ENODEV);
203 	}
204 	if (vp->v_mountedhere != NULL) {
205 		vput(vp);
206 		return (EBUSY);
207 	}
208 
209 	/*
210 	 * Allocate and initialize the file system.
211 	 */
212 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
213 		M_MOUNT, M_WAITOK);
214 	bzero((char *)mp, (u_long)sizeof(struct mount));
215 	mp->mnt_op = vfssw[fsindex];
216 	if ((error = vfs_lock(mp)) != 0) {
217 		free((caddr_t)mp, M_MOUNT);
218 		vput(vp);
219 		return (error);
220 	}
221 	/* Do this early in case we block later. */
222 	vfssw[fsindex]->vfs_refcount++;
223 	vp->v_mountedhere = mp;
224 	mp->mnt_vnodecovered = vp;
225 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
226 update:
227 	/*
228 	 * Set the mount level flags.
229 	 */
230 	if (SCARG(uap, flags) & MNT_RDONLY)
231 		mp->mnt_flag |= MNT_RDONLY;
232 	else if (mp->mnt_flag & MNT_RDONLY)
233 		mp->mnt_flag |= MNT_WANTRDWR;
234 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
235 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
236 	    MNT_NOATIME);
237 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
238 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
239 	    MNT_NOCOREDUMP | MNT_NOATIME);
240 	/*
241 	 * Mount the filesystem.
242 	 */
243 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
244 	if (mp->mnt_flag & MNT_UPDATE) {
245 		vrele(vp);
246 		if (mp->mnt_flag & MNT_WANTRDWR)
247 			mp->mnt_flag &= ~MNT_RDONLY;
248 		mp->mnt_flag &=~
249 		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
250 		if (error)
251 			mp->mnt_flag = flag;
252 		return (error);
253 	}
254 	/*
255 	 * Put the new filesystem on the mount list after root.
256 	 */
257 	cache_purge(vp);
258 	if (!error) {
259 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
260 		checkdirs(vp);
261 		VOP_UNLOCK(vp);
262 		vfs_unlock(mp);
263 		(void) VFS_STATFS(mp, &mp->mnt_stat, p);
264 		error = VFS_START(mp, 0, p);
265 	} else {
266 		mp->mnt_vnodecovered->v_mountedhere = (struct mount *)0;
267 		vfssw[fsindex]->vfs_refcount--;
268 		vfs_unlock(mp);
269 		free((caddr_t)mp, M_MOUNT);
270 		vput(vp);
271 	}
272 	return (error);
273 }
274 
275 /*
276  * Scan all active processes to see if any of them have a current
277  * or root directory onto which the new filesystem has just been
278  * mounted. If so, replace them with the new mount point.
279  */
280 void
281 checkdirs(olddp)
282 	struct vnode *olddp;
283 {
284 	struct filedesc *fdp;
285 	struct vnode *newdp;
286 	struct proc *p;
287 
288 	if (olddp->v_usecount == 1)
289 		return;
290 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
291 		panic("mount: lost mount");
292 	for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
293 		fdp = p->p_fd;
294 		if (fdp->fd_cdir == olddp) {
295 			vrele(fdp->fd_cdir);
296 			VREF(newdp);
297 			fdp->fd_cdir = newdp;
298 		}
299 		if (fdp->fd_rdir == olddp) {
300 			vrele(fdp->fd_rdir);
301 			VREF(newdp);
302 			fdp->fd_rdir = newdp;
303 		}
304 	}
305 	if (rootvnode == olddp) {
306 		vrele(rootvnode);
307 		VREF(newdp);
308 		rootvnode = newdp;
309 	}
310 	vput(newdp);
311 }
312 
313 /*
314  * Unmount a file system.
315  *
316  * Note: unmount takes a path to the vnode mounted on as argument,
317  * not special file (as before).
318  */
319 /* ARGSUSED */
320 int
321 sys_unmount(p, v, retval)
322 	struct proc *p;
323 	void *v;
324 	register_t *retval;
325 {
326 	register struct sys_unmount_args /* {
327 		syscallarg(const char *) path;
328 		syscallarg(int) flags;
329 	} */ *uap = v;
330 	register struct vnode *vp;
331 	struct mount *mp;
332 	int error;
333 	struct nameidata nd;
334 
335 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
336 	    SCARG(uap, path), p);
337 	if ((error = namei(&nd)) != 0)
338 		return (error);
339 	vp = nd.ni_vp;
340 	mp = vp->v_mount;
341 
342 	/*
343 	 * Only root, or the user that did the original mount is
344 	 * permitted to unmount this filesystem.
345 	 */
346 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
347 	    (error = suser(p->p_ucred, &p->p_acflag))) {
348 		vput(vp);
349 		return (error);
350 	}
351 
352 	/*
353 	 * Don't allow unmounting the root file system.
354 	 */
355 	if (mp->mnt_flag & MNT_ROOTFS) {
356 		vput(vp);
357 		return (EINVAL);
358 	}
359 
360 	/*
361 	 * Must be the root of the filesystem
362 	 */
363 	if ((vp->v_flag & VROOT) == 0) {
364 		vput(vp);
365 		return (EINVAL);
366 	}
367 	vput(vp);
368 
369 	if (vfs_busy(mp))
370 		return (EBUSY);
371 
372 	return (dounmount(mp, SCARG(uap, flags), p));
373 }
374 
375 /*
376  * Do the actual file system unmount. File system is assumed to have been
377  * marked busy by the caller.
378  */
379 int
380 dounmount(mp, flags, p)
381 	register struct mount *mp;
382 	int flags;
383 	struct proc *p;
384 {
385 	struct vnode *coveredvp;
386 	int error;
387 
388 	coveredvp = mp->mnt_vnodecovered;
389 	mp->mnt_flag |= MNT_UNMOUNT;
390 	if ((error = vfs_lock(mp)) != 0) {
391 		vfs_unbusy(mp);
392 		return (error);
393 	}
394 
395 	mp->mnt_flag &=~ MNT_ASYNC;
396 	vnode_pager_umount(mp);	/* release cached vnodes */
397 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
398 	if ((error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0 ||
399 	    (flags & MNT_FORCE))
400 		error = VFS_UNMOUNT(mp, flags, p);
401 	mp->mnt_flag &= ~MNT_UNMOUNT;
402 	vfs_unbusy(mp);
403 	if (error) {
404 		vfs_unlock(mp);
405 	} else {
406 		CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
407 		if (coveredvp != NULLVP) {
408 			vrele(coveredvp);
409 			coveredvp->v_mountedhere = (struct mount *)0;
410 		}
411 		mp->mnt_op->vfs_refcount--;
412 		vfs_unlock(mp);
413 		if (mp->mnt_vnodelist.lh_first != NULL)
414 			panic("unmount: dangling vnode");
415 		free((caddr_t)mp, M_MOUNT);
416 	}
417 	return (error);
418 }
419 
420 /*
421  * Sync each mounted filesystem.
422  */
423 #ifdef DEBUG
424 int syncprt = 0;
425 struct ctldebug debug0 = { "syncprt", &syncprt };
426 #endif
427 
428 /* ARGSUSED */
429 int
430 sys_sync(p, v, retval)
431 	struct proc *p;
432 	void *v;
433 	register_t *retval;
434 {
435 	register struct mount *mp, *nmp;
436 	int asyncflag;
437 
438 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
439 		/*
440 		 * Get the prev pointer in case we hang on vfs_busy
441 		 * while we are being unmounted.
442 		 */
443 		nmp = mp->mnt_list.cqe_prev;
444 		/*
445 		 * The lock check below is to avoid races with mount
446 		 * and unmount.
447 		 */
448 		if ((mp->mnt_flag & (MNT_MLOCK|MNT_RDONLY|MNT_MPBUSY)) == 0 &&
449 		    !vfs_busy(mp)) {
450 			asyncflag = mp->mnt_flag & MNT_ASYNC;
451 			mp->mnt_flag &= ~MNT_ASYNC;
452 			vnode_pager_sync(mp);
453 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
454 			if (asyncflag)
455 				mp->mnt_flag |= MNT_ASYNC;
456 			/*
457 			 * Get the prev pointer again, as the prior filesystem
458 			 * might have been unmounted while we were sync'ing.
459 			 */
460 			nmp = mp->mnt_list.cqe_prev;
461 			vfs_unbusy(mp);
462 		}
463 	}
464 #ifdef DEBUG
465 	if (syncprt)
466 		vfs_bufstats();
467 #endif /* DEBUG */
468 	return (0);
469 }
470 
471 /*
472  * Change filesystem quotas.
473  */
474 /* ARGSUSED */
475 int
476 sys_quotactl(p, v, retval)
477 	struct proc *p;
478 	void *v;
479 	register_t *retval;
480 {
481 	register struct sys_quotactl_args /* {
482 		syscallarg(const char *) path;
483 		syscallarg(int) cmd;
484 		syscallarg(int) uid;
485 		syscallarg(caddr_t) arg;
486 	} */ *uap = v;
487 	register struct mount *mp;
488 	int error;
489 	struct nameidata nd;
490 
491 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
492 	if ((error = namei(&nd)) != 0)
493 		return (error);
494 	mp = nd.ni_vp->v_mount;
495 	vrele(nd.ni_vp);
496 	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
497 	    SCARG(uap, arg), p));
498 }
499 
500 /*
501  * Get filesystem statistics.
502  */
503 /* ARGSUSED */
504 int
505 sys_statfs(p, v, retval)
506 	struct proc *p;
507 	void *v;
508 	register_t *retval;
509 {
510 	register struct sys_statfs_args /* {
511 		syscallarg(const char *) path;
512 		syscallarg(struct statfs *) buf;
513 	} */ *uap = v;
514 	register struct mount *mp;
515 	register struct statfs *sp;
516 	int error;
517 	struct nameidata nd;
518 
519 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
520 	if ((error = namei(&nd)) != 0)
521 		return (error);
522 	mp = nd.ni_vp->v_mount;
523 	sp = &mp->mnt_stat;
524 	vrele(nd.ni_vp);
525 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
526 		return (error);
527 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
528 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
529 }
530 
531 /*
532  * Get filesystem statistics.
533  */
534 /* ARGSUSED */
535 int
536 sys_fstatfs(p, v, retval)
537 	struct proc *p;
538 	void *v;
539 	register_t *retval;
540 {
541 	register struct sys_fstatfs_args /* {
542 		syscallarg(int) fd;
543 		syscallarg(struct statfs *) buf;
544 	} */ *uap = v;
545 	struct file *fp;
546 	struct mount *mp;
547 	register struct statfs *sp;
548 	int error;
549 
550 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
551 		return (error);
552 	mp = ((struct vnode *)fp->f_data)->v_mount;
553 	sp = &mp->mnt_stat;
554 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
555 		return (error);
556 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
557 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
558 }
559 
560 /*
561  * Get statistics on all filesystems.
562  */
563 int
564 sys_getfsstat(p, v, retval)
565 	struct proc *p;
566 	void *v;
567 	register_t *retval;
568 {
569 	register struct sys_getfsstat_args /* {
570 		syscallarg(struct statfs *) buf;
571 		syscallarg(long) bufsize;
572 		syscallarg(int) flags;
573 	} */ *uap = v;
574 	register struct mount *mp, *nmp;
575 	register struct statfs *sp;
576 	caddr_t sfsp;
577 	long count, maxcount, error;
578 
579 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
580 	sfsp = (caddr_t)SCARG(uap, buf);
581 	for (count = 0,
582 	     mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
583 		nmp = mp->mnt_list.cqe_next;
584 		if (sfsp && count < maxcount &&
585 		    ((mp->mnt_flag & MNT_MLOCK) == 0)) {
586 			sp = &mp->mnt_stat;
587 			/*
588 			 * If MNT_NOWAIT is specified, do not refresh the
589 			 * fsstat cache. MNT_WAIT overrides MNT_NOWAIT.
590 			 */
591 			if (((SCARG(uap, flags) & MNT_NOWAIT) == 0 ||
592 			    (SCARG(uap, flags) & MNT_WAIT)) &&
593 			    (error = VFS_STATFS(mp, sp, p)))
594 				continue;
595 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
596 			error = copyout(sp, sfsp, sizeof(*sp));
597 			if (error)
598 				return (error);
599 			sfsp += sizeof(*sp);
600 		}
601 		count++;
602 	}
603 	if (sfsp && count > maxcount)
604 		*retval = maxcount;
605 	else
606 		*retval = count;
607 	return (0);
608 }
609 
610 /*
611  * Change current working directory to a given file descriptor.
612  */
613 /* ARGSUSED */
614 int
615 sys_fchdir(p, v, retval)
616 	struct proc *p;
617 	void *v;
618 	register_t *retval;
619 {
620 	struct sys_fchdir_args /* {
621 		syscallarg(int) fd;
622 	} */ *uap = v;
623 	register struct filedesc *fdp = p->p_fd;
624 	struct vnode *vp, *tdp;
625 	struct mount *mp;
626 	struct file *fp;
627 	int error;
628 
629 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
630 		return (error);
631 	vp = (struct vnode *)fp->f_data;
632 	VREF(vp);
633 	VOP_LOCK(vp);
634 	if (vp->v_type != VDIR)
635 		error = ENOTDIR;
636 	else
637 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
638 	while (!error && (mp = vp->v_mountedhere) != NULL) {
639 		if (mp->mnt_flag & MNT_MLOCK) {
640 			mp->mnt_flag |= MNT_MWAIT;
641 			sleep((caddr_t)mp, PVFS);
642 			continue;
643 		}
644 		if ((error = VFS_ROOT(mp, &tdp)) != 0)
645 			break;
646 		vput(vp);
647 		vp = tdp;
648 	}
649 	VOP_UNLOCK(vp);
650 	if (error) {
651 		vrele(vp);
652 		return (error);
653 	}
654 	vrele(fdp->fd_cdir);
655 	fdp->fd_cdir = vp;
656 	return (0);
657 }
658 
659 /*
660  * Change current working directory (``.'').
661  */
662 /* ARGSUSED */
663 int
664 sys_chdir(p, v, retval)
665 	struct proc *p;
666 	void *v;
667 	register_t *retval;
668 {
669 	struct sys_chdir_args /* {
670 		syscallarg(const char *) path;
671 	} */ *uap = v;
672 	register struct filedesc *fdp = p->p_fd;
673 	int error;
674 	struct nameidata nd;
675 
676 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
677 	    SCARG(uap, path), p);
678 	if ((error = change_dir(&nd, p)) != 0)
679 		return (error);
680 	vrele(fdp->fd_cdir);
681 	fdp->fd_cdir = nd.ni_vp;
682 	return (0);
683 }
684 
685 /*
686  * Change notion of root (``/'') directory.
687  */
688 /* ARGSUSED */
689 int
690 sys_chroot(p, v, retval)
691 	struct proc *p;
692 	void *v;
693 	register_t *retval;
694 {
695 	struct sys_chroot_args /* {
696 		syscallarg(const char *) path;
697 	} */ *uap = v;
698 	register struct filedesc *fdp = p->p_fd;
699 	int error;
700 	struct nameidata nd;
701 
702 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
703 		return (error);
704 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
705 	    SCARG(uap, path), p);
706 	if ((error = change_dir(&nd, p)) != 0)
707 		return (error);
708 	if (fdp->fd_rdir != NULL)
709 		vrele(fdp->fd_rdir);
710 	fdp->fd_rdir = nd.ni_vp;
711 	return (0);
712 }
713 
714 /*
715  * Common routine for chroot and chdir.
716  */
717 static int
718 change_dir(ndp, p)
719 	register struct nameidata *ndp;
720 	struct proc *p;
721 {
722 	struct vnode *vp;
723 	int error;
724 
725 	if ((error = namei(ndp)) != 0)
726 		return (error);
727 	vp = ndp->ni_vp;
728 	if (vp->v_type != VDIR)
729 		error = ENOTDIR;
730 	else
731 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
732 	VOP_UNLOCK(vp);
733 	if (error)
734 		vrele(vp);
735 	return (error);
736 }
737 
738 /*
739  * Check permissions, allocate an open file structure,
740  * and call the device open routine if any.
741  */
742 int
743 sys_open(p, v, retval)
744 	struct proc *p;
745 	void *v;
746 	register_t *retval;
747 {
748 	register struct sys_open_args /* {
749 		syscallarg(const char *) path;
750 		syscallarg(int) flags;
751 		syscallarg(int) mode;
752 	} */ *uap = v;
753 	register struct filedesc *fdp = p->p_fd;
754 	register struct file *fp;
755 	register struct vnode *vp;
756 	int flags, cmode;
757 	struct file *nfp;
758 	int type, indx, error;
759 	struct flock lf;
760 	struct nameidata nd;
761 	extern struct fileops vnops;
762 
763 	if ((error = falloc(p, &nfp, &indx)) != 0)
764 		return (error);
765 	fp = nfp;
766 	flags = FFLAGS(SCARG(uap, flags));
767 	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
768 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
769 	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
770 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
771 		ffree(fp);
772 		if ((error == ENODEV || error == ENXIO) &&
773 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
774 		    (error =
775 			dupfdopen(fdp, indx, p->p_dupfd, flags, error)) == 0) {
776 			*retval = indx;
777 			return (0);
778 		}
779 		if (error == ERESTART)
780 			error = EINTR;
781 		fdp->fd_ofiles[indx] = NULL;
782 		return (error);
783 	}
784 	p->p_dupfd = 0;
785 	vp = nd.ni_vp;
786 	fp->f_flag = flags & FMASK;
787 	fp->f_type = DTYPE_VNODE;
788 	fp->f_ops = &vnops;
789 	fp->f_data = (caddr_t)vp;
790 	if (flags & (O_EXLOCK | O_SHLOCK)) {
791 		lf.l_whence = SEEK_SET;
792 		lf.l_start = 0;
793 		lf.l_len = 0;
794 		if (flags & O_EXLOCK)
795 			lf.l_type = F_WRLCK;
796 		else
797 			lf.l_type = F_RDLCK;
798 		type = F_FLOCK;
799 		if ((flags & FNONBLOCK) == 0)
800 			type |= F_WAIT;
801 		VOP_UNLOCK(vp);
802 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
803 		if (error) {
804 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
805 			ffree(fp);
806 			fdp->fd_ofiles[indx] = NULL;
807 			return (error);
808 		}
809 		VOP_LOCK(vp);
810 		fp->f_flag |= FHASLOCK;
811 	}
812 	VOP_UNLOCK(vp);
813 	*retval = indx;
814 	return (0);
815 }
816 
817 /*
818  * Create a special file.
819  */
820 /* ARGSUSED */
821 int
822 sys_mknod(p, v, retval)
823 	struct proc *p;
824 	void *v;
825 	register_t *retval;
826 {
827 	register struct sys_mknod_args /* {
828 		syscallarg(const char *) path;
829 		syscallarg(int) mode;
830 		syscallarg(int) dev;
831 	} */ *uap = v;
832 	register struct vnode *vp;
833 	struct vattr vattr;
834 	int error;
835 	int whiteout = 0;
836 	struct nameidata nd;
837 
838 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
839 		return (error);
840 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
841 	if ((error = namei(&nd)) != 0)
842 		return (error);
843 	vp = nd.ni_vp;
844 	if (vp != NULL)
845 		error = EEXIST;
846 	else {
847 		VATTR_NULL(&vattr);
848 		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
849 		vattr.va_rdev = SCARG(uap, dev);
850 		whiteout = 0;
851 
852 		switch (SCARG(uap, mode) & S_IFMT) {
853 		case S_IFMT:	/* used by badsect to flag bad sectors */
854 			vattr.va_type = VBAD;
855 			break;
856 		case S_IFCHR:
857 			vattr.va_type = VCHR;
858 			break;
859 		case S_IFBLK:
860 			vattr.va_type = VBLK;
861 			break;
862 		case S_IFWHT:
863 			whiteout = 1;
864 			break;
865 		default:
866 			error = EINVAL;
867 			break;
868 		}
869 	}
870 	if (!error) {
871 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
872 		if (whiteout) {
873 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
874 			if (error)
875 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
876 			vput(nd.ni_dvp);
877 		} else {
878 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
879 						&nd.ni_cnd, &vattr);
880 		}
881 	} else {
882 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
883 		if (nd.ni_dvp == vp)
884 			vrele(nd.ni_dvp);
885 		else
886 			vput(nd.ni_dvp);
887 		if (vp)
888 			vrele(vp);
889 	}
890 	return (error);
891 }
892 
893 /*
894  * Create a named pipe.
895  */
896 /* ARGSUSED */
897 int
898 sys_mkfifo(p, v, retval)
899 	struct proc *p;
900 	void *v;
901 	register_t *retval;
902 {
903 #ifndef FIFO
904 	return (EOPNOTSUPP);
905 #else
906 	register struct sys_mkfifo_args /* {
907 		syscallarg(const char *) path;
908 		syscallarg(int) mode;
909 	} */ *uap = v;
910 	struct vattr vattr;
911 	int error;
912 	struct nameidata nd;
913 
914 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
915 	if ((error = namei(&nd)) != 0)
916 		return (error);
917 	if (nd.ni_vp != NULL) {
918 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
919 		if (nd.ni_dvp == nd.ni_vp)
920 			vrele(nd.ni_dvp);
921 		else
922 			vput(nd.ni_dvp);
923 		vrele(nd.ni_vp);
924 		return (EEXIST);
925 	}
926 	VATTR_NULL(&vattr);
927 	vattr.va_type = VFIFO;
928 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
929 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
930 	return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
931 #endif /* FIFO */
932 }
933 
934 /*
935  * Make a hard file link.
936  */
937 /* ARGSUSED */
938 int
939 sys_link(p, v, retval)
940 	struct proc *p;
941 	void *v;
942 	register_t *retval;
943 {
944 	register struct sys_link_args /* {
945 		syscallarg(const char *) path;
946 		syscallarg(const char *) link;
947 	} */ *uap = v;
948 	register struct vnode *vp;
949 	struct nameidata nd;
950 	int error;
951 
952 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
953 	if ((error = namei(&nd)) != 0)
954 		return (error);
955 	vp = nd.ni_vp;
956 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
957 	if ((error = namei(&nd)) != 0)
958 		goto out;
959 	if (nd.ni_vp) {
960 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
961 		if (nd.ni_dvp == nd.ni_vp)
962 			vrele(nd.ni_dvp);
963 		else
964 			vput(nd.ni_dvp);
965 		vrele(nd.ni_vp);
966 		error = EEXIST;
967 		goto out;
968 	}
969 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
970 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
971 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
972 out:
973 	vrele(vp);
974 	return (error);
975 }
976 
977 /*
978  * Make a symbolic link.
979  */
980 /* ARGSUSED */
981 int
982 sys_symlink(p, v, retval)
983 	struct proc *p;
984 	void *v;
985 	register_t *retval;
986 {
987 	register struct sys_symlink_args /* {
988 		syscallarg(const char *) path;
989 		syscallarg(const char *) link;
990 	} */ *uap = v;
991 	struct vattr vattr;
992 	char *path;
993 	int error;
994 	struct nameidata nd;
995 
996 	MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
997 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
998 	if (error)
999 		goto out;
1000 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1001 	if ((error = namei(&nd)) != 0)
1002 		goto out;
1003 	if (nd.ni_vp) {
1004 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1005 		if (nd.ni_dvp == nd.ni_vp)
1006 			vrele(nd.ni_dvp);
1007 		else
1008 			vput(nd.ni_dvp);
1009 		vrele(nd.ni_vp);
1010 		error = EEXIST;
1011 		goto out;
1012 	}
1013 	VATTR_NULL(&vattr);
1014 	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1015 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1016 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1017 out:
1018 	FREE(path, M_NAMEI);
1019 	return (error);
1020 }
1021 
1022 /*
1023  * Delete a whiteout from the filesystem.
1024  */
1025 /* ARGSUSED */
1026 int
1027 sys_undelete(p, v, retval)
1028 	struct proc *p;
1029 	void *v;
1030 	register_t *retval;
1031 {
1032 	register struct sys_undelete_args /* {
1033 		syscallarg(const char *) path;
1034 	} */ *uap = v;
1035 	int error;
1036 	struct nameidata nd;
1037 
1038 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1039 	    SCARG(uap, path), p);
1040 	error = namei(&nd);
1041 	if (error)
1042 		return (error);
1043 
1044 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1045 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1046 		if (nd.ni_dvp == nd.ni_vp)
1047 			vrele(nd.ni_dvp);
1048 		else
1049 			vput(nd.ni_dvp);
1050 		if (nd.ni_vp)
1051 			vrele(nd.ni_vp);
1052 		return (EEXIST);
1053 	}
1054 
1055 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1056 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1057 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1058 	vput(nd.ni_dvp);
1059 	return (error);
1060 }
1061 
1062 /*
1063  * Delete a name from the filesystem.
1064  */
1065 /* ARGSUSED */
1066 int
1067 sys_unlink(p, v, retval)
1068 	struct proc *p;
1069 	void *v;
1070 	register_t *retval;
1071 {
1072 	struct sys_unlink_args /* {
1073 		syscallarg(const char *) path;
1074 	} */ *uap = v;
1075 	register struct vnode *vp;
1076 	int error;
1077 	struct nameidata nd;
1078 
1079 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1080 	    SCARG(uap, path), p);
1081 	if ((error = namei(&nd)) != 0)
1082 		return (error);
1083 	vp = nd.ni_vp;
1084 
1085 	/*
1086 	 * The root of a mounted filesystem cannot be deleted.
1087 	 */
1088 	if (vp->v_flag & VROOT) {
1089 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1090 		if (nd.ni_dvp == vp)
1091 			vrele(nd.ni_dvp);
1092 		else
1093 			vput(nd.ni_dvp);
1094 		vput(vp);
1095 		error = EBUSY;
1096 		goto out;
1097 	}
1098 
1099 	(void)vnode_pager_uncache(vp);
1100 
1101 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1102 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1103 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1104 out:
1105 	return (error);
1106 }
1107 
1108 /*
1109  * Reposition read/write file offset.
1110  */
1111 int
1112 sys_lseek(p, v, retval)
1113 	struct proc *p;
1114 	void *v;
1115 	register_t *retval;
1116 {
1117 	register struct sys_lseek_args /* {
1118 		syscallarg(int) fd;
1119 		syscallarg(int) pad;
1120 		syscallarg(off_t) offset;
1121 		syscallarg(int) whence;
1122 	} */ *uap = v;
1123 	struct ucred *cred = p->p_ucred;
1124 	register struct filedesc *fdp = p->p_fd;
1125 	register struct file *fp;
1126 	struct vnode *vp;
1127 	struct vattr vattr;
1128 	register off_t newoff;
1129 	int error;
1130 
1131 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1132 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
1133 		return (EBADF);
1134 
1135 	vp = (struct vnode *)fp->f_data;
1136 	if (fp->f_type != DTYPE_VNODE
1137 #ifdef FIFO
1138 	    || vp->v_type == VFIFO
1139 #endif
1140 	)
1141 		return (ESPIPE);
1142 
1143 	switch (SCARG(uap, whence)) {
1144 	case L_INCR:
1145 		newoff = fp->f_offset + SCARG(uap, offset);
1146 		break;
1147 	case L_XTND:
1148 		error = VOP_GETATTR(vp, &vattr, cred, p);
1149 		if (error)
1150 			return (error);
1151 		newoff = SCARG(uap, offset) + vattr.va_size;
1152 		break;
1153 	case L_SET:
1154 		newoff = SCARG(uap, offset);
1155 		break;
1156 	default:
1157 		return (EINVAL);
1158 	}
1159 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1160 		return (error);
1161 
1162 	*(off_t *)retval = fp->f_offset = newoff;
1163 	return (0);
1164 }
1165 
1166 /*
1167  * Check access permissions.
1168  */
1169 int
1170 sys_access(p, v, retval)
1171 	struct proc *p;
1172 	void *v;
1173 	register_t *retval;
1174 {
1175 	register struct sys_access_args /* {
1176 		syscallarg(const char *) path;
1177 		syscallarg(int) flags;
1178 	} */ *uap = v;
1179 	register struct ucred *cred = p->p_ucred;
1180 	register struct vnode *vp;
1181 	int error, flags, t_gid, t_uid;
1182 	struct nameidata nd;
1183 
1184 	t_uid = cred->cr_uid;
1185 	t_gid = cred->cr_gid;
1186 	cred->cr_uid = p->p_cred->p_ruid;
1187 	cred->cr_gid = p->p_cred->p_rgid;
1188 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1189 	    SCARG(uap, path), p);
1190 	if ((error = namei(&nd)) != 0)
1191 		goto out1;
1192 	vp = nd.ni_vp;
1193 
1194 	/* Flags == 0 means only check for existence. */
1195 	if (SCARG(uap, flags)) {
1196 		flags = 0;
1197 		if (SCARG(uap, flags) & R_OK)
1198 			flags |= VREAD;
1199 		if (SCARG(uap, flags) & W_OK)
1200 			flags |= VWRITE;
1201 		if (SCARG(uap, flags) & X_OK)
1202 			flags |= VEXEC;
1203 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1204 			error = VOP_ACCESS(vp, flags, cred, p);
1205 	}
1206 	vput(vp);
1207 out1:
1208 	cred->cr_uid = t_uid;
1209 	cred->cr_gid = t_gid;
1210 	return (error);
1211 }
1212 
1213 /*
1214  * Get file status; this version follows links.
1215  */
1216 /* ARGSUSED */
1217 int
1218 sys_stat(p, v, retval)
1219 	struct proc *p;
1220 	void *v;
1221 	register_t *retval;
1222 {
1223 	register struct sys_stat_args /* {
1224 		syscallarg(const char *) path;
1225 		syscallarg(struct stat *) ub;
1226 	} */ *uap = v;
1227 	struct stat sb;
1228 	int error;
1229 	struct nameidata nd;
1230 
1231 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1232 	    SCARG(uap, path), p);
1233 	if ((error = namei(&nd)) != 0)
1234 		return (error);
1235 	error = vn_stat(nd.ni_vp, &sb, p);
1236 	vput(nd.ni_vp);
1237 	if (error)
1238 		return (error);
1239 	error = copyout(&sb, SCARG(uap, ub), sizeof (sb));
1240 	return (error);
1241 }
1242 
1243 /*
1244  * Get file status; this version does not follow links.
1245  */
1246 /* ARGSUSED */
1247 int
1248 sys_lstat(p, v, retval)
1249 	struct proc *p;
1250 	void *v;
1251 	register_t *retval;
1252 {
1253 	register struct sys_lstat_args /* {
1254 		syscallarg(const char *) path;
1255 		syscallarg(struct stat *) ub;
1256 	} */ *uap = v;
1257 	struct stat sb;
1258 	int error;
1259 	struct nameidata nd;
1260 
1261 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1262 	    SCARG(uap, path), p);
1263 	if ((error = namei(&nd)) != 0)
1264 		return (error);
1265 	error = vn_stat(nd.ni_vp, &sb, p);
1266 	vput(nd.ni_vp);
1267 	if (error)
1268 		return (error);
1269 	error = copyout(&sb, SCARG(uap, ub), sizeof (sb));
1270 	return (error);
1271 }
1272 
1273 /*
1274  * Get configurable pathname variables.
1275  */
1276 /* ARGSUSED */
1277 int
1278 sys_pathconf(p, v, retval)
1279 	struct proc *p;
1280 	void *v;
1281 	register_t *retval;
1282 {
1283 	register struct sys_pathconf_args /* {
1284 		syscallarg(const char *) path;
1285 		syscallarg(int) name;
1286 	} */ *uap = v;
1287 	int error;
1288 	struct nameidata nd;
1289 
1290 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1291 	    SCARG(uap, path), p);
1292 	if ((error = namei(&nd)) != 0)
1293 		return (error);
1294 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1295 	vput(nd.ni_vp);
1296 	return (error);
1297 }
1298 
1299 /*
1300  * Return target name of a symbolic link.
1301  */
1302 /* ARGSUSED */
1303 int
1304 sys_readlink(p, v, retval)
1305 	struct proc *p;
1306 	void *v;
1307 	register_t *retval;
1308 {
1309 	register struct sys_readlink_args /* {
1310 		syscallarg(const char *) path;
1311 		syscallarg(char *) buf;
1312 		syscallarg(int) count;
1313 	} */ *uap = v;
1314 	register struct vnode *vp;
1315 	struct iovec aiov;
1316 	struct uio auio;
1317 	int error;
1318 	struct nameidata nd;
1319 
1320 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1321 	    SCARG(uap, path), p);
1322 	if ((error = namei(&nd)) != 0)
1323 		return (error);
1324 	vp = nd.ni_vp;
1325 	if (vp->v_type != VLNK)
1326 		error = EINVAL;
1327 	else {
1328 		aiov.iov_base = SCARG(uap, buf);
1329 		aiov.iov_len = SCARG(uap, count);
1330 		auio.uio_iov = &aiov;
1331 		auio.uio_iovcnt = 1;
1332 		auio.uio_offset = 0;
1333 		auio.uio_rw = UIO_READ;
1334 		auio.uio_segflg = UIO_USERSPACE;
1335 		auio.uio_procp = p;
1336 		auio.uio_resid = SCARG(uap, count);
1337 		error = VOP_READLINK(vp, &auio, p->p_ucred);
1338 	}
1339 	vput(vp);
1340 	*retval = SCARG(uap, count) - auio.uio_resid;
1341 	return (error);
1342 }
1343 
1344 /*
1345  * Change flags of a file given a path name.
1346  */
1347 /* ARGSUSED */
1348 int
1349 sys_chflags(p, v, retval)
1350 	struct proc *p;
1351 	void *v;
1352 	register_t *retval;
1353 {
1354 	register struct sys_chflags_args /* {
1355 		syscallarg(const char *) path;
1356 		syscallarg(u_long) flags;
1357 	} */ *uap = v;
1358 	register struct vnode *vp;
1359 	struct vattr vattr;
1360 	int error;
1361 	struct nameidata nd;
1362 
1363 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1364 	if ((error = namei(&nd)) != 0)
1365 		return (error);
1366 	vp = nd.ni_vp;
1367 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1368 	VOP_LOCK(vp);
1369 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1370 		error = EROFS;
1371 	else {
1372 		VATTR_NULL(&vattr);
1373 		vattr.va_flags = SCARG(uap, flags);
1374 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1375 	}
1376 	vput(vp);
1377 	return (error);
1378 }
1379 
1380 /*
1381  * Change flags of a file given a file descriptor.
1382  */
1383 /* ARGSUSED */
1384 int
1385 sys_fchflags(p, v, retval)
1386 	struct proc *p;
1387 	void *v;
1388 	register_t *retval;
1389 {
1390 	register struct sys_fchflags_args /* {
1391 		syscallarg(int) fd;
1392 		syscallarg(u_long) flags;
1393 	} */ *uap = v;
1394 	struct vattr vattr;
1395 	struct vnode *vp;
1396 	struct file *fp;
1397 	int error;
1398 
1399 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1400 		return (error);
1401 	vp = (struct vnode *)fp->f_data;
1402 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1403 	VOP_LOCK(vp);
1404 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1405 		error = EROFS;
1406 	else {
1407 		VATTR_NULL(&vattr);
1408 		vattr.va_flags = SCARG(uap, flags);
1409 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1410 	}
1411 	VOP_UNLOCK(vp);
1412 	return (error);
1413 }
1414 
1415 /*
1416  * Change mode of a file given path name.
1417  */
1418 /* ARGSUSED */
1419 int
1420 sys_chmod(p, v, retval)
1421 	struct proc *p;
1422 	void *v;
1423 	register_t *retval;
1424 {
1425 	register struct sys_chmod_args /* {
1426 		syscallarg(const char *) path;
1427 		syscallarg(int) mode;
1428 	} */ *uap = v;
1429 	register struct vnode *vp;
1430 	struct vattr vattr;
1431 	int error;
1432 	struct nameidata nd;
1433 
1434 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1435 	if ((error = namei(&nd)) != 0)
1436 		return (error);
1437 	vp = nd.ni_vp;
1438 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1439 	VOP_LOCK(vp);
1440 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1441 		error = EROFS;
1442 	else {
1443 		VATTR_NULL(&vattr);
1444 		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1445 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1446 	}
1447 	vput(vp);
1448 	return (error);
1449 }
1450 
1451 /*
1452  * Change mode of a file given a file descriptor.
1453  */
1454 /* ARGSUSED */
1455 int
1456 sys_fchmod(p, v, retval)
1457 	struct proc *p;
1458 	void *v;
1459 	register_t *retval;
1460 {
1461 	register struct sys_fchmod_args /* {
1462 		syscallarg(int) fd;
1463 		syscallarg(int) mode;
1464 	} */ *uap = v;
1465 	struct vattr vattr;
1466 	struct vnode *vp;
1467 	struct file *fp;
1468 	int error;
1469 
1470 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1471 		return (error);
1472 	vp = (struct vnode *)fp->f_data;
1473 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1474 	VOP_LOCK(vp);
1475 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1476 		error = EROFS;
1477 	else {
1478 		VATTR_NULL(&vattr);
1479 		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1480 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1481 	}
1482 	VOP_UNLOCK(vp);
1483 	return (error);
1484 }
1485 
1486 /*
1487  * Set ownership given a path name.
1488  */
1489 /* ARGSUSED */
1490 int
1491 sys_chown(p, v, retval)
1492 	struct proc *p;
1493 	void *v;
1494 	register_t *retval;
1495 {
1496 	register struct sys_chown_args /* {
1497 		syscallarg(const char *) path;
1498 		syscallarg(uid_t) uid;
1499 		syscallarg(gid_t) gid;
1500 	} */ *uap = v;
1501 	register struct vnode *vp;
1502 	struct vattr vattr;
1503 	int error;
1504 	struct nameidata nd;
1505 
1506 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1507 	if ((error = namei(&nd)) != 0)
1508 		return (error);
1509 	vp = nd.ni_vp;
1510 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1511 	VOP_LOCK(vp);
1512 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1513 		error = EROFS;
1514 	else {
1515 		VATTR_NULL(&vattr);
1516 		vattr.va_uid = SCARG(uap, uid);
1517 		vattr.va_gid = SCARG(uap, gid);
1518 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1519 	}
1520 	vput(vp);
1521 	return (error);
1522 }
1523 
1524 /*
1525  * Set ownership given a file descriptor.
1526  */
1527 /* ARGSUSED */
1528 int
1529 sys_fchown(p, v, retval)
1530 	struct proc *p;
1531 	void *v;
1532 	register_t *retval;
1533 {
1534 	register struct sys_fchown_args /* {
1535 		syscallarg(int) fd;
1536 		syscallarg(uid_t) uid;
1537 		syscallarg(gid_t) gid;
1538 	} */ *uap = v;
1539 	register struct vnode *vp;
1540 	struct vattr vattr;
1541 	int error;
1542 	struct file *fp;
1543 
1544 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1545 		return (error);
1546 	vp = (struct vnode *)fp->f_data;
1547 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1548 	VOP_LOCK(vp);
1549 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1550 		error = EROFS;
1551 	else {
1552 		VATTR_NULL(&vattr);
1553 		vattr.va_uid = SCARG(uap, uid);
1554 		vattr.va_gid = SCARG(uap, gid);
1555 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1556 	}
1557 	VOP_UNLOCK(vp);
1558 	return (error);
1559 }
1560 
1561 /*
1562  * Set the access and modification times given a path name.
1563  */
1564 /* ARGSUSED */
1565 int
1566 sys_utimes(p, v, retval)
1567 	struct proc *p;
1568 	void *v;
1569 	register_t *retval;
1570 {
1571 	register struct sys_utimes_args /* {
1572 		syscallarg(const char *) path;
1573 		syscallarg(const struct timeval *) tptr;
1574 	} */ *uap = v;
1575 	register struct vnode *vp;
1576 	struct timeval tv[2];
1577 	struct vattr vattr;
1578 	int error;
1579 	struct nameidata nd;
1580 
1581 	VATTR_NULL(&vattr);
1582 	if (SCARG(uap, tptr) == NULL) {
1583 		microtime(&tv[0]);
1584 		tv[1] = tv[0];
1585 		vattr.va_vaflags |= VA_UTIMES_NULL;
1586 	} else {
1587 		error = copyin(SCARG(uap, tptr), tv, sizeof (tv));
1588 		if (error)
1589 			return (error);
1590 	}
1591 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1592 	if ((error = namei(&nd)) != 0)
1593 		return (error);
1594 	vp = nd.ni_vp;
1595 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1596 	VOP_LOCK(vp);
1597 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1598 		error = EROFS;
1599 	else {
1600 		vattr.va_atime.tv_sec = tv[0].tv_sec;
1601 		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
1602 		vattr.va_mtime.tv_sec = tv[1].tv_sec;
1603 		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
1604 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1605 	}
1606 	vput(vp);
1607 	return (error);
1608 }
1609 
1610 /*
1611  * Set the access and modification times given a file descriptor.
1612  */
1613 /* ARGSUSED */
1614 int
1615 sys_futimes(p, v, retval)
1616 	struct proc *p;
1617 	void *v;
1618 	register_t *retval;
1619 {
1620 	register struct sys_futimes_args /* {
1621 		syscallarg(int) fd;
1622 		syscallarg(const struct timeval *) tptr;
1623 	} */ *uap = v;
1624 	register struct vnode *vp;
1625 	struct timeval tv[2];
1626 	struct vattr vattr;
1627 	int error;
1628 	struct file *fp;
1629 
1630 	VATTR_NULL(&vattr);
1631 	if (SCARG(uap, tptr) == NULL) {
1632 		microtime(&tv[0]);
1633 		tv[1] = tv[0];
1634 		vattr.va_vaflags |= VA_UTIMES_NULL;
1635 	} else {
1636 		error = copyin(SCARG(uap, tptr), tv, sizeof (tv));
1637 		if (error)
1638 			return (error);
1639 	}
1640 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1641 		return (error);
1642 	vp = (struct vnode *)fp->f_data;
1643 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1644 	VOP_LOCK(vp);
1645 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1646 		error = EROFS;
1647 	else {
1648 		vattr.va_atime.tv_sec = tv[0].tv_sec;
1649 		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
1650 		vattr.va_mtime.tv_sec = tv[1].tv_sec;
1651 		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
1652 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1653 	}
1654 	VOP_UNLOCK(vp);
1655 	return (error);
1656 }
1657 
1658 /*
1659  * Truncate a file given its path name.
1660  */
1661 /* ARGSUSED */
1662 int
1663 sys_truncate(p, v, retval)
1664 	struct proc *p;
1665 	void *v;
1666 	register_t *retval;
1667 {
1668 	register struct sys_truncate_args /* {
1669 		syscallarg(const char *) path;
1670 		syscallarg(int) pad;
1671 		syscallarg(off_t) length;
1672 	} */ *uap = v;
1673 	register struct vnode *vp;
1674 	struct vattr vattr;
1675 	int error;
1676 	struct nameidata nd;
1677 
1678 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1679 	if ((error = namei(&nd)) != 0)
1680 		return (error);
1681 	vp = nd.ni_vp;
1682 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1683 	VOP_LOCK(vp);
1684 	if (vp->v_type == VDIR)
1685 		error = EISDIR;
1686 	else if ((error = vn_writechk(vp)) == 0 &&
1687 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
1688 		VATTR_NULL(&vattr);
1689 		vattr.va_size = SCARG(uap, length);
1690 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1691 	}
1692 	vput(vp);
1693 	return (error);
1694 }
1695 
1696 /*
1697  * Truncate a file given a file descriptor.
1698  */
1699 /* ARGSUSED */
1700 int
1701 sys_ftruncate(p, v, retval)
1702 	struct proc *p;
1703 	void *v;
1704 	register_t *retval;
1705 {
1706 	register struct sys_ftruncate_args /* {
1707 		syscallarg(int) fd;
1708 		syscallarg(int) pad;
1709 		syscallarg(off_t) length;
1710 	} */ *uap = v;
1711 	struct vattr vattr;
1712 	struct vnode *vp;
1713 	struct file *fp;
1714 	int error;
1715 
1716 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1717 		return (error);
1718 	if ((fp->f_flag & FWRITE) == 0)
1719 		return (EINVAL);
1720 	vp = (struct vnode *)fp->f_data;
1721 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1722 	VOP_LOCK(vp);
1723 	if (vp->v_type == VDIR)
1724 		error = EISDIR;
1725 	else if ((error = vn_writechk(vp)) == 0) {
1726 		VATTR_NULL(&vattr);
1727 		vattr.va_size = SCARG(uap, length);
1728 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
1729 	}
1730 	VOP_UNLOCK(vp);
1731 	return (error);
1732 }
1733 
1734 /*
1735  * Sync an open file.
1736  */
1737 /* ARGSUSED */
1738 int
1739 sys_fsync(p, v, retval)
1740 	struct proc *p;
1741 	void *v;
1742 	register_t *retval;
1743 {
1744 	struct sys_fsync_args /* {
1745 		syscallarg(int) fd;
1746 	} */ *uap = v;
1747 	register struct vnode *vp;
1748 	struct file *fp;
1749 	int error;
1750 
1751 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1752 		return (error);
1753 	vp = (struct vnode *)fp->f_data;
1754 	VOP_LOCK(vp);
1755 	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
1756 	VOP_UNLOCK(vp);
1757 	return (error);
1758 }
1759 
1760 /*
1761  * Rename files.  Source and destination must either both be directories,
1762  * or both not be directories.  If target is a directory, it must be empty.
1763  */
1764 /* ARGSUSED */
1765 int
1766 sys_rename(p, v, retval)
1767 	struct proc *p;
1768 	void *v;
1769 	register_t *retval;
1770 {
1771 	register struct sys_rename_args /* {
1772 		syscallarg(const char *) from;
1773 		syscallarg(const char *) to;
1774 	} */ *uap = v;
1775 	register struct vnode *tvp, *fvp, *tdvp;
1776 	struct nameidata fromnd, tond;
1777 	int error;
1778 
1779 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
1780 	    SCARG(uap, from), p);
1781 	if ((error = namei(&fromnd)) != 0)
1782 		return (error);
1783 	fvp = fromnd.ni_vp;
1784 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
1785 	    UIO_USERSPACE, SCARG(uap, to), p);
1786 	if ((error = namei(&tond)) != 0) {
1787 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1788 		vrele(fromnd.ni_dvp);
1789 		vrele(fvp);
1790 		goto out1;
1791 	}
1792 	tdvp = tond.ni_dvp;
1793 	tvp = tond.ni_vp;
1794 	if (tvp != NULL) {
1795 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1796 			error = ENOTDIR;
1797 			goto out;
1798 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1799 			error = EISDIR;
1800 			goto out;
1801 		}
1802 	}
1803 	if (fvp == tdvp)
1804 		error = EINVAL;
1805 	/*
1806 	 * If source is the same as the destination (that is the
1807 	 * same inode number with the same name in the same directory),
1808 	 * then there is nothing to do.
1809 	 */
1810 	if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1811 	    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1812 	    !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1813 	      fromnd.ni_cnd.cn_namelen))
1814 		error = -1;
1815 out:
1816 	if (!error) {
1817 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
1818 		if (fromnd.ni_dvp != tdvp)
1819 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1820 		if (tvp) {
1821 			(void)vnode_pager_uncache(tvp);
1822 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
1823 		}
1824 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1825 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1826 	} else {
1827 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1828 		if (tdvp == tvp)
1829 			vrele(tdvp);
1830 		else
1831 			vput(tdvp);
1832 		if (tvp)
1833 			vput(tvp);
1834 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1835 		vrele(fromnd.ni_dvp);
1836 		vrele(fvp);
1837 	}
1838 	vrele(tond.ni_startdir);
1839 	FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1840 out1:
1841 	if (fromnd.ni_startdir)
1842 		vrele(fromnd.ni_startdir);
1843 	FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1844 	return (error == -1 ? 0 : error);
1845 }
1846 
1847 /*
1848  * Make a directory file.
1849  */
1850 /* ARGSUSED */
1851 int
1852 sys_mkdir(p, v, retval)
1853 	struct proc *p;
1854 	void *v;
1855 	register_t *retval;
1856 {
1857 	register struct sys_mkdir_args /* {
1858 		syscallarg(const char *) path;
1859 		syscallarg(int) mode;
1860 	} */ *uap = v;
1861 	register struct vnode *vp;
1862 	struct vattr vattr;
1863 	int error;
1864 	struct nameidata nd;
1865 
1866 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1867 	if ((error = namei(&nd)) != 0)
1868 		return (error);
1869 	vp = nd.ni_vp;
1870 	if (vp != NULL) {
1871 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1872 		if (nd.ni_dvp == vp)
1873 			vrele(nd.ni_dvp);
1874 		else
1875 			vput(nd.ni_dvp);
1876 		vrele(vp);
1877 		return (EEXIST);
1878 	}
1879 	VATTR_NULL(&vattr);
1880 	vattr.va_type = VDIR;
1881 	vattr.va_mode = (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_fd->fd_cmask;
1882 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1883 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1884 	if (!error)
1885 		vput(nd.ni_vp);
1886 	return (error);
1887 }
1888 
1889 /*
1890  * Remove a directory file.
1891  */
1892 /* ARGSUSED */
1893 int
1894 sys_rmdir(p, v, retval)
1895 	struct proc *p;
1896 	void *v;
1897 	register_t *retval;
1898 {
1899 	struct sys_rmdir_args /* {
1900 		syscallarg(const char *) path;
1901 	} */ *uap = v;
1902 	register struct vnode *vp;
1903 	int error;
1904 	struct nameidata nd;
1905 
1906 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1907 	    SCARG(uap, path), p);
1908 	if ((error = namei(&nd)) != 0)
1909 		return (error);
1910 	vp = nd.ni_vp;
1911 	if (vp->v_type != VDIR) {
1912 		error = ENOTDIR;
1913 		goto out;
1914 	}
1915 	/*
1916 	 * No rmdir "." please.
1917 	 */
1918 	if (nd.ni_dvp == vp) {
1919 		error = EINVAL;
1920 		goto out;
1921 	}
1922 	/*
1923 	 * The root of a mounted filesystem cannot be deleted.
1924 	 */
1925 	if (vp->v_flag & VROOT)
1926 		error = EBUSY;
1927 out:
1928 	if (!error) {
1929 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1930 		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1931 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1932 	} else {
1933 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1934 		if (nd.ni_dvp == vp)
1935 			vrele(nd.ni_dvp);
1936 		else
1937 			vput(nd.ni_dvp);
1938 		vput(vp);
1939 	}
1940 	return (error);
1941 }
1942 
1943 /*
1944  * Read a block of directory entries in a file system independent format.
1945  */
1946 int
1947 sys_getdirentries(p, v, retval)
1948 	struct proc *p;
1949 	void *v;
1950 	register_t *retval;
1951 {
1952 	register struct sys_getdirentries_args /* {
1953 		syscallarg(int) fd;
1954 		syscallarg(char *) buf;
1955 		syscallarg(u_int) count;
1956 		syscallarg(long *) basep;
1957 	} */ *uap = v;
1958 	register struct vnode *vp;
1959 	struct file *fp;
1960 	struct uio auio;
1961 	struct iovec aiov;
1962 	long loff;
1963 	int error, eofflag;
1964 
1965 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1966 		return (error);
1967 	if ((fp->f_flag & FREAD) == 0)
1968 		return (EBADF);
1969 	vp = (struct vnode *)fp->f_data;
1970 unionread:
1971 	if (vp->v_type != VDIR)
1972 		return (EINVAL);
1973 	aiov.iov_base = SCARG(uap, buf);
1974 	aiov.iov_len = SCARG(uap, count);
1975 	auio.uio_iov = &aiov;
1976 	auio.uio_iovcnt = 1;
1977 	auio.uio_rw = UIO_READ;
1978 	auio.uio_segflg = UIO_USERSPACE;
1979 	auio.uio_procp = p;
1980 	auio.uio_resid = SCARG(uap, count);
1981 	VOP_LOCK(vp);
1982 	loff = auio.uio_offset = fp->f_offset;
1983 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, (u_long *)0, 0);
1984 	fp->f_offset = auio.uio_offset;
1985 	VOP_UNLOCK(vp);
1986 	if (error)
1987 		return (error);
1988 
1989 #ifdef UNION
1990 {
1991 	extern int (**union_vnodeop_p) __P((void *));
1992 	extern struct vnode *union_dircache __P((struct vnode *));
1993 
1994 	if ((SCARG(uap, count) == auio.uio_resid) &&
1995 	    (vp->v_op == union_vnodeop_p)) {
1996 		struct vnode *lvp;
1997 
1998 		lvp = union_dircache(vp);
1999 		if (lvp != NULLVP) {
2000 			struct vattr va;
2001 
2002 			/*
2003 			 * If the directory is opaque,
2004 			 * then don't show lower entries
2005 			 */
2006 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
2007 			if (va.va_flags & OPAQUE) {
2008 				vput(lvp);
2009 				lvp = NULL;
2010 			}
2011 		}
2012 
2013 		if (lvp != NULLVP) {
2014 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
2015 			VOP_UNLOCK(lvp);
2016 
2017 			if (error) {
2018 				vrele(lvp);
2019 				return (error);
2020 			}
2021 			fp->f_data = (caddr_t) lvp;
2022 			fp->f_offset = 0;
2023 			error = vn_close(vp, FREAD, fp->f_cred, p);
2024 			if (error)
2025 				return (error);
2026 			vp = lvp;
2027 			goto unionread;
2028 		}
2029 	}
2030 }
2031 #endif /* UNION */
2032 
2033 	if ((SCARG(uap, count) == auio.uio_resid) &&
2034 	    (vp->v_flag & VROOT) &&
2035 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
2036 		struct vnode *tvp = vp;
2037 		vp = vp->v_mount->mnt_vnodecovered;
2038 		VREF(vp);
2039 		fp->f_data = (caddr_t) vp;
2040 		fp->f_offset = 0;
2041 		vrele(tvp);
2042 		goto unionread;
2043 	}
2044 	error = copyout(&loff, SCARG(uap, basep), sizeof(long));
2045 	*retval = SCARG(uap, count) - auio.uio_resid;
2046 	return (error);
2047 }
2048 
2049 /*
2050  * Set the mode mask for creation of filesystem nodes.
2051  */
2052 int
2053 sys_umask(p, v, retval)
2054 	struct proc *p;
2055 	void *v;
2056 	register_t *retval;
2057 {
2058 	struct sys_umask_args /* {
2059 		syscallarg(int) newmask;
2060 	} */ *uap = v;
2061 	register struct filedesc *fdp;
2062 
2063 	fdp = p->p_fd;
2064 	*retval = fdp->fd_cmask;
2065 	fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
2066 	return (0);
2067 }
2068 
2069 /*
2070  * Void all references to file by ripping underlying filesystem
2071  * away from vnode.
2072  */
2073 /* ARGSUSED */
2074 int
2075 sys_revoke(p, v, retval)
2076 	struct proc *p;
2077 	void *v;
2078 	register_t *retval;
2079 {
2080 	register struct sys_revoke_args /* {
2081 		syscallarg(const char *) path;
2082 	} */ *uap = v;
2083 	register struct vnode *vp;
2084 	struct vattr vattr;
2085 	int error;
2086 	struct nameidata nd;
2087 
2088 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2089 	if ((error = namei(&nd)) != 0)
2090 		return (error);
2091 	vp = nd.ni_vp;
2092 	if (vp->v_type != VCHR && vp->v_type != VBLK) {
2093 		error = EINVAL;
2094 		goto out;
2095 	}
2096 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2097 		goto out;
2098 	if (p->p_ucred->cr_uid != vattr.va_uid &&
2099 	    (error = suser(p->p_ucred, &p->p_acflag)))
2100 		goto out;
2101 	if (vp->v_usecount > 1 || (vp->v_flag & VALIASED))
2102 		vgoneall(vp);
2103 out:
2104 	vrele(vp);
2105 	return (error);
2106 }
2107 
2108 /*
2109  * Convert a user file descriptor to a kernel file entry.
2110  */
2111 int
2112 getvnode(fdp, fd, fpp)
2113 	struct filedesc *fdp;
2114 	int fd;
2115 	struct file **fpp;
2116 {
2117 	struct vnode *vp;
2118 	struct file *fp;
2119 
2120 	if ((u_int)fd >= fdp->fd_nfiles ||
2121 	    (fp = fdp->fd_ofiles[fd]) == NULL)
2122 		return (EBADF);
2123 	if (fp->f_type != DTYPE_VNODE)
2124 		return (EINVAL);
2125 	vp = (struct vnode *)fp->f_data;
2126 	if (vp->v_type == VBAD)
2127 		return (EBADF);
2128 	*fpp = fp;
2129 	return (0);
2130 }
2131