xref: /netbsd/sys/kern/vfs_syscalls.c (revision c4a72b64)
1 /*	$NetBSD: vfs_syscalls.c,v 1.178 2002/10/30 22:36:46 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	8.42 (Berkeley) 7/31/95
41  */
42 
43 #include <sys/cdefs.h>
44 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.178 2002/10/30 22:36:46 kleink Exp $");
45 
46 #include "opt_compat_netbsd.h"
47 #include "opt_compat_43.h"
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/namei.h>
52 #include <sys/filedesc.h>
53 #include <sys/kernel.h>
54 #include <sys/file.h>
55 #include <sys/stat.h>
56 #include <sys/vnode.h>
57 #include <sys/mount.h>
58 #include <sys/proc.h>
59 #include <sys/uio.h>
60 #include <sys/malloc.h>
61 #include <sys/dirent.h>
62 #include <sys/sysctl.h>
63 #include <sys/syscallargs.h>
64 
65 #include <miscfs/genfs/genfs.h>
66 #include <miscfs/syncfs/syncfs.h>
67 
68 static int change_dir __P((struct nameidata *, struct proc *));
69 static int change_flags __P((struct vnode *, u_long, struct proc *));
70 static int change_mode __P((struct vnode *, int, struct proc *p));
71 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *,
72     int));
73 static int change_utimes __P((struct vnode *vp, const struct timeval *,
74 	       struct proc *p));
75 static int rename_files __P((const char *, const char *, struct proc *, int));
76 
77 void checkdirs __P((struct vnode *));
78 
79 int dovfsusermount = 0;
80 
81 /*
82  * Virtual File System System Calls
83  */
84 
85 /*
86  * Mount a file system.
87  */
88 
89 #if defined(COMPAT_09) || defined(COMPAT_43)
90 /*
91  * This table is used to maintain compatibility with 4.3BSD
92  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
93  *
94  * Do not modify this table. It should only contain filesystems
95  * supported by NetBSD 0.9 and 4.3BSD.
96  */
97 const char * const mountcompatnames[] = {
98 	NULL,		/* 0 = MOUNT_NONE */
99 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
100 	MOUNT_NFS,	/* 2 */
101 	MOUNT_MFS,	/* 3 */
102 	MOUNT_MSDOS,	/* 4 */
103 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
104 	MOUNT_FDESC,	/* 6 */
105 	MOUNT_KERNFS,	/* 7 */
106 	NULL,		/* 8 = MOUNT_DEVFS */
107 	MOUNT_AFS,	/* 9 */
108 };
109 const int nmountcompatnames = sizeof(mountcompatnames) /
110     sizeof(mountcompatnames[0]);
111 #endif /* COMPAT_09 || COMPAT_43 */
112 
113 /* ARGSUSED */
114 int
115 sys_mount(p, v, retval)
116 	struct proc *p;
117 	void *v;
118 	register_t *retval;
119 {
120 	struct sys_mount_args /* {
121 		syscallarg(const char *) type;
122 		syscallarg(const char *) path;
123 		syscallarg(int) flags;
124 		syscallarg(void *) data;
125 	} */ *uap = v;
126 	struct vnode *vp;
127 	struct mount *mp;
128 	int error, flag = 0;
129 	char fstypename[MFSNAMELEN];
130 	struct vattr va;
131 	struct nameidata nd;
132 	struct vfsops *vfs;
133 
134 	if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
135 	    (error = suser(p->p_ucred, &p->p_acflag)))
136 		return (error);
137 	/*
138 	 * Get vnode to be covered
139 	 */
140 	NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
141 	    SCARG(uap, path), p);
142 	if ((error = namei(&nd)) != 0)
143 		return (error);
144 	vp = nd.ni_vp;
145 	/*
146 	 * A lookup in VFS_MOUNT might result in an attempt to
147 	 * lock this vnode again, so make the lock resursive.
148 	 */
149 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
150 	if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
151 		if ((vp->v_flag & VROOT) == 0) {
152 			vput(vp);
153 			return (EINVAL);
154 		}
155 		mp = vp->v_mount;
156 		flag = mp->mnt_flag;
157 		vfs = mp->mnt_op;
158 		/*
159 		 * We only allow the filesystem to be reloaded if it
160 		 * is currently mounted read-only.
161 		 */
162 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
163 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
164 			vput(vp);
165 			return (EOPNOTSUPP);	/* Needs translation */
166 		}
167 		/*
168 		 * In "highly secure" mode, don't let the caller do anything
169 		 * but downgrade a filesystem from read-write to read-only.
170 		 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
171 		 */
172 		if (securelevel >= 2 &&
173 		    SCARG(uap, flags) != MNT_GETARGS &&
174 		    SCARG(uap, flags) !=
175 		    (mp->mnt_flag | MNT_RDONLY |
176 		    MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
177 			vput(vp);
178 			return (EPERM);
179 		}
180 		mp->mnt_flag |= SCARG(uap, flags) &
181 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
182 		/*
183 		 * Only root, or the user that did the original mount is
184 		 * permitted to update it.
185 		 */
186 		if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
187 		    mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
188 		    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
189 			vput(vp);
190 			return (error);
191 		}
192 		/*
193 		 * Do not allow NFS export by non-root users. For non-root
194 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
195 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
196 		 */
197 		if (p->p_ucred->cr_uid != 0) {
198 			if (SCARG(uap, flags) & MNT_EXPORTED) {
199 				vput(vp);
200 				return (EPERM);
201 			}
202 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
203 			if (flag & MNT_NOEXEC)
204 				SCARG(uap, flags) |= MNT_NOEXEC;
205 		}
206 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
207 			vput(vp);
208 			return (EPERM);
209 		}
210 		goto update;
211 	} else {
212 		if (securelevel >= 2) {
213 			vput(vp);
214 			return (EPERM);
215 		}
216 	}
217 	/*
218 	 * If the user is not root, ensure that they own the directory
219 	 * onto which we are attempting to mount.
220 	 */
221 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
222 	    (va.va_uid != p->p_ucred->cr_uid &&
223 		(error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
224 		vput(vp);
225 		return (error);
226 	}
227 	/*
228 	 * Do not allow NFS export by non-root users. For non-root users,
229 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
230 	 * mount point is already MNT_NOEXEC.
231 	 */
232 	if (p->p_ucred->cr_uid != 0) {
233 		if (SCARG(uap, flags) & MNT_EXPORTED) {
234 			vput(vp);
235 			return (EPERM);
236 		}
237 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
238 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
239 			SCARG(uap, flags) |= MNT_NOEXEC;
240 	}
241 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
242 		return (error);
243 	if (vp->v_type != VDIR) {
244 		vput(vp);
245 		return (ENOTDIR);
246 	}
247 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
248 	if (error) {
249 #if defined(COMPAT_09) || defined(COMPAT_43)
250 		/*
251 		 * Historically filesystem types were identified by number.
252 		 * If we get an integer for the filesystem type instead of a
253 		 * string, we check to see if it matches one of the historic
254 		 * filesystem types.
255 		 */
256 		u_long fsindex = (u_long)SCARG(uap, type);
257 		if (fsindex >= nmountcompatnames ||
258 		    mountcompatnames[fsindex] == NULL) {
259 			vput(vp);
260 			return (ENODEV);
261 		}
262 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
263 #else
264 		vput(vp);
265 		return (error);
266 #endif
267 	}
268 #ifdef	COMPAT_10
269 	/* Accept `ufs' as an alias for `ffs'. */
270 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
271 		strncpy(fstypename, "ffs", MFSNAMELEN);
272 #endif
273 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
274 		vput(vp);
275 		return (ENODEV);
276 	}
277 	if (vp->v_mountedhere != NULL) {
278 		vput(vp);
279 		return (EBUSY);
280 	}
281 
282 	/*
283 	 * Allocate and initialize the file system.
284 	 */
285 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
286 		M_MOUNT, M_WAITOK);
287 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
288 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
289 	(void)vfs_busy(mp, LK_NOWAIT, 0);
290 	mp->mnt_op = vfs;
291 	vfs->vfs_refcount++;
292 	mp->mnt_vnodecovered = vp;
293 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
294 	mp->mnt_unmounter = NULL;
295 update:
296 	/*
297 	 * Set the mount level flags.
298 	 */
299 	if (SCARG(uap, flags) & MNT_RDONLY)
300 		mp->mnt_flag |= MNT_RDONLY;
301 	else if (mp->mnt_flag & MNT_RDONLY)
302 		mp->mnt_flag |= MNT_WANTRDWR;
303 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
304 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
305 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
306 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
307 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
308 	    MNT_NOCOREDUMP | MNT_IGNORE | MNT_NOATIME | MNT_NODEVMTIME |
309 	    MNT_SYMPERM | MNT_SOFTDEP);
310 	/*
311 	 * Mount the filesystem.
312 	 */
313 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
314 	if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
315 		if (mp->mnt_flag & MNT_WANTRDWR)
316 			mp->mnt_flag &= ~MNT_RDONLY;
317 		if (error || (mp->mnt_flag & MNT_GETARGS))
318 			mp->mnt_flag = flag;
319 		mp->mnt_flag &=~
320 		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR |
321 		     MNT_GETARGS);
322 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
323 			if (mp->mnt_syncer == NULL)
324 				error = vfs_allocate_syncvnode(mp);
325 		} else {
326 			if (mp->mnt_syncer != NULL)
327 				vfs_deallocate_syncvnode(mp);
328 		}
329 		vfs_unbusy(mp);
330 		VOP_UNLOCK(vp, 0);
331 		vrele(vp);
332 		return (error);
333 	}
334 	/*
335 	 * Put the new filesystem on the mount list after root.
336 	 */
337 	cache_purge(vp);
338 	if (!error) {
339 		vp->v_mountedhere = mp;
340 		simple_lock(&mountlist_slock);
341 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
342 		simple_unlock(&mountlist_slock);
343 		checkdirs(vp);
344 		VOP_UNLOCK(vp, 0);
345 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
346 			error = vfs_allocate_syncvnode(mp);
347 		vfs_unbusy(mp);
348 		(void) VFS_STATFS(mp, &mp->mnt_stat, p);
349 		if ((error = VFS_START(mp, 0, p)))
350 			vrele(vp);
351 	} else {
352 		vp->v_mountedhere = (struct mount *)0;
353 		vfs->vfs_refcount--;
354 		vfs_unbusy(mp);
355 		free((caddr_t)mp, M_MOUNT);
356 		vput(vp);
357 	}
358 	return (error);
359 }
360 
361 /*
362  * Scan all active processes to see if any of them have a current
363  * or root directory onto which the new filesystem has just been
364  * mounted. If so, replace them with the new mount point.
365  */
366 void
367 checkdirs(olddp)
368 	struct vnode *olddp;
369 {
370 	struct cwdinfo *cwdi;
371 	struct vnode *newdp;
372 	struct proc *p;
373 
374 	if (olddp->v_usecount == 1)
375 		return;
376 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
377 		panic("mount: lost mount");
378 	proclist_lock_read();
379 	LIST_FOREACH(p, &allproc, p_list) {
380 		cwdi = p->p_cwdi;
381 		if (cwdi->cwdi_cdir == olddp) {
382 			vrele(cwdi->cwdi_cdir);
383 			VREF(newdp);
384 			cwdi->cwdi_cdir = newdp;
385 		}
386 		if (cwdi->cwdi_rdir == olddp) {
387 			vrele(cwdi->cwdi_rdir);
388 			VREF(newdp);
389 			cwdi->cwdi_rdir = newdp;
390 		}
391 	}
392 	proclist_unlock_read();
393 	if (rootvnode == olddp) {
394 		vrele(rootvnode);
395 		VREF(newdp);
396 		rootvnode = newdp;
397 	}
398 	vput(newdp);
399 }
400 
401 /*
402  * Unmount a file system.
403  *
404  * Note: unmount takes a path to the vnode mounted on as argument,
405  * not special file (as before).
406  */
407 /* ARGSUSED */
408 int
409 sys_unmount(p, v, retval)
410 	struct proc *p;
411 	void *v;
412 	register_t *retval;
413 {
414 	struct sys_unmount_args /* {
415 		syscallarg(const char *) path;
416 		syscallarg(int) flags;
417 	} */ *uap = v;
418 	struct vnode *vp;
419 	struct mount *mp;
420 	int error;
421 	struct nameidata nd;
422 
423 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
424 	    SCARG(uap, path), p);
425 	if ((error = namei(&nd)) != 0)
426 		return (error);
427 	vp = nd.ni_vp;
428 	mp = vp->v_mount;
429 
430 	/*
431 	 * Only root, or the user that did the original mount is
432 	 * permitted to unmount this filesystem.
433 	 */
434 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
435 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
436 		vput(vp);
437 		return (error);
438 	}
439 
440 	/*
441 	 * Don't allow unmounting the root file system.
442 	 */
443 	if (mp->mnt_flag & MNT_ROOTFS) {
444 		vput(vp);
445 		return (EINVAL);
446 	}
447 
448 	/*
449 	 * Must be the root of the filesystem
450 	 */
451 	if ((vp->v_flag & VROOT) == 0) {
452 		vput(vp);
453 		return (EINVAL);
454 	}
455 	vput(vp);
456 
457 	/*
458 	 * XXX Freeze syncer.  Must do this before locking the
459 	 * mount point.  See dounmount() for details.
460 	 */
461 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
462 
463 	if (vfs_busy(mp, 0, 0)) {
464 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
465 		return (EBUSY);
466 	}
467 
468 	return (dounmount(mp, SCARG(uap, flags), p));
469 }
470 
471 /*
472  * Do the actual file system unmount. File system is assumed to have been
473  * marked busy by the caller.
474  */
475 int
476 dounmount(mp, flags, p)
477 	struct mount *mp;
478 	int flags;
479 	struct proc *p;
480 {
481 	struct vnode *coveredvp;
482 	int error;
483 	int async;
484 	int used_syncer;
485 
486 	simple_lock(&mountlist_slock);
487 	vfs_unbusy(mp);
488 	used_syncer = (mp->mnt_syncer != NULL);
489 
490 	/*
491 	 * XXX Syncer must be frozen when we get here.  This should really
492 	 * be done on a per-mountpoint basis, but especially the softdep
493 	 * code possibly called from the syncer doens't exactly work on a
494 	 * per-mountpoint basis, so the softdep code would become a maze
495 	 * of vfs_busy() calls.
496 	 *
497 	 * The caller of dounmount() must acquire syncer_lock because
498 	 * the syncer itself acquires locks in syncer_lock -> vfs_busy
499 	 * order, and we must preserve that order to avoid deadlock.
500 	 *
501 	 * So, if the file system did not use the syncer, now is
502 	 * the time to release the syncer_lock.
503 	 */
504 	if (used_syncer == 0)
505 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
506 
507 	mp->mnt_flag |= MNT_UNMOUNT;
508 	mp->mnt_unmounter = p;
509 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
510 	if (mp->mnt_flag & MNT_EXPUBLIC)
511 		vfs_setpublicfs(NULL, NULL, NULL);
512 	async = mp->mnt_flag & MNT_ASYNC;
513 	mp->mnt_flag &= ~MNT_ASYNC;
514 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
515 	if (mp->mnt_syncer != NULL)
516 		vfs_deallocate_syncvnode(mp);
517 	if (((mp->mnt_flag & MNT_RDONLY) ||
518 	    (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
519 	    (flags & MNT_FORCE))
520 		error = VFS_UNMOUNT(mp, flags, p);
521 	simple_lock(&mountlist_slock);
522 	if (error) {
523 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
524 			(void) vfs_allocate_syncvnode(mp);
525 		mp->mnt_flag &= ~MNT_UNMOUNT;
526 		mp->mnt_unmounter = NULL;
527 		mp->mnt_flag |= async;
528 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
529 		    &mountlist_slock);
530 		if (used_syncer)
531 			lockmgr(&syncer_lock, LK_RELEASE, NULL);
532 		while (mp->mnt_wcnt > 0) {
533 			wakeup((caddr_t)mp);
534 			tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
535 		}
536 		return (error);
537 	}
538 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
539 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
540 		coveredvp->v_mountedhere = NULL;
541 		vrele(coveredvp);
542 	}
543 	mp->mnt_op->vfs_refcount--;
544 	if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
545 		panic("unmount: dangling vnode");
546 	mp->mnt_flag |= MNT_GONE;
547 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
548 	if (used_syncer)
549 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
550 	while(mp->mnt_wcnt > 0) {
551 		wakeup((caddr_t)mp);
552 		tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
553 	}
554 	free((caddr_t)mp, M_MOUNT);
555 	return (0);
556 }
557 
558 /*
559  * Sync each mounted filesystem.
560  */
561 #ifdef DEBUG
562 int syncprt = 0;
563 struct ctldebug debug0 = { "syncprt", &syncprt };
564 #endif
565 
566 /* ARGSUSED */
567 int
568 sys_sync(p, v, retval)
569 	struct proc *p;
570 	void *v;
571 	register_t *retval;
572 {
573 	struct mount *mp, *nmp;
574 	int asyncflag;
575 
576 	simple_lock(&mountlist_slock);
577 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
578 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
579 			nmp = mp->mnt_list.cqe_prev;
580 			continue;
581 		}
582 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
583 			asyncflag = mp->mnt_flag & MNT_ASYNC;
584 			mp->mnt_flag &= ~MNT_ASYNC;
585 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
586 			if (asyncflag)
587 				 mp->mnt_flag |= MNT_ASYNC;
588 		}
589 		simple_lock(&mountlist_slock);
590 		nmp = mp->mnt_list.cqe_prev;
591 		vfs_unbusy(mp);
592 
593 	}
594 	simple_unlock(&mountlist_slock);
595 #ifdef DEBUG
596 	if (syncprt)
597 		vfs_bufstats();
598 #endif /* DEBUG */
599 	return (0);
600 }
601 
602 /*
603  * Change filesystem quotas.
604  */
605 /* ARGSUSED */
606 int
607 sys_quotactl(p, v, retval)
608 	struct proc *p;
609 	void *v;
610 	register_t *retval;
611 {
612 	struct sys_quotactl_args /* {
613 		syscallarg(const char *) path;
614 		syscallarg(int) cmd;
615 		syscallarg(int) uid;
616 		syscallarg(caddr_t) arg;
617 	} */ *uap = v;
618 	struct mount *mp;
619 	int error;
620 	struct nameidata nd;
621 
622 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
623 	if ((error = namei(&nd)) != 0)
624 		return (error);
625 	mp = nd.ni_vp->v_mount;
626 	vrele(nd.ni_vp);
627 	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
628 	    SCARG(uap, arg), p));
629 }
630 
631 /*
632  * Get filesystem statistics.
633  */
634 /* ARGSUSED */
635 int
636 sys_statfs(p, v, retval)
637 	struct proc *p;
638 	void *v;
639 	register_t *retval;
640 {
641 	struct sys_statfs_args /* {
642 		syscallarg(const char *) path;
643 		syscallarg(struct statfs *) buf;
644 	} */ *uap = v;
645 	struct mount *mp;
646 	struct statfs *sp;
647 	int error;
648 	struct nameidata nd;
649 
650 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
651 	if ((error = namei(&nd)) != 0)
652 		return (error);
653 	mp = nd.ni_vp->v_mount;
654 	sp = &mp->mnt_stat;
655 	vrele(nd.ni_vp);
656 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
657 		return (error);
658 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
659 	sp->f_oflags = sp->f_flags & 0xffff;
660 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
661 }
662 
663 /*
664  * Get filesystem statistics.
665  */
666 /* ARGSUSED */
667 int
668 sys_fstatfs(p, v, retval)
669 	struct proc *p;
670 	void *v;
671 	register_t *retval;
672 {
673 	struct sys_fstatfs_args /* {
674 		syscallarg(int) fd;
675 		syscallarg(struct statfs *) buf;
676 	} */ *uap = v;
677 	struct file *fp;
678 	struct mount *mp;
679 	struct statfs *sp;
680 	int error;
681 
682 	/* getvnode() will use the descriptor for us */
683 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
684 		return (error);
685 	mp = ((struct vnode *)fp->f_data)->v_mount;
686 	sp = &mp->mnt_stat;
687 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
688 		goto out;
689 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
690 	sp->f_oflags = sp->f_flags & 0xffff;
691 	error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
692  out:
693 	FILE_UNUSE(fp, p);
694 	return (error);
695 }
696 
697 /*
698  * Get statistics on all filesystems.
699  */
700 int
701 sys_getfsstat(p, v, retval)
702 	struct proc *p;
703 	void *v;
704 	register_t *retval;
705 {
706 	struct sys_getfsstat_args /* {
707 		syscallarg(struct statfs *) buf;
708 		syscallarg(long) bufsize;
709 		syscallarg(int) flags;
710 	} */ *uap = v;
711 	struct mount *mp, *nmp;
712 	struct statfs *sp;
713 	caddr_t sfsp;
714 	long count, maxcount, error;
715 
716 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
717 	sfsp = (caddr_t)SCARG(uap, buf);
718 	simple_lock(&mountlist_slock);
719 	count = 0;
720 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
721 	     mp = nmp) {
722 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
723 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
724 			continue;
725 		}
726 		if (sfsp && count < maxcount) {
727 			sp = &mp->mnt_stat;
728 			/*
729 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
730 			 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
731 			 * overrides MNT_NOWAIT.
732 			 */
733 			if (SCARG(uap, flags) != MNT_NOWAIT &&
734 			    SCARG(uap, flags) != MNT_LAZY &&
735 			    (SCARG(uap, flags) == MNT_WAIT ||
736 			     SCARG(uap, flags) == 0) &&
737 			    (error = VFS_STATFS(mp, sp, p)) != 0) {
738 				simple_lock(&mountlist_slock);
739 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
740 				vfs_unbusy(mp);
741 				continue;
742 			}
743 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
744 			sp->f_oflags = sp->f_flags & 0xffff;
745 			error = copyout(sp, sfsp, sizeof(*sp));
746 			if (error) {
747 				vfs_unbusy(mp);
748 				return (error);
749 			}
750 			sfsp += sizeof(*sp);
751 		}
752 		count++;
753 		simple_lock(&mountlist_slock);
754 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
755 		vfs_unbusy(mp);
756 	}
757 	simple_unlock(&mountlist_slock);
758 	if (sfsp && count > maxcount)
759 		*retval = maxcount;
760 	else
761 		*retval = count;
762 	return (0);
763 }
764 
765 /*
766  * Change current working directory to a given file descriptor.
767  */
768 /* ARGSUSED */
769 int
770 sys_fchdir(p, v, retval)
771 	struct proc *p;
772 	void *v;
773 	register_t *retval;
774 {
775 	struct sys_fchdir_args /* {
776 		syscallarg(int) fd;
777 	} */ *uap = v;
778 	struct filedesc *fdp = p->p_fd;
779 	struct cwdinfo *cwdi = p->p_cwdi;
780 	struct vnode *vp, *tdp;
781 	struct mount *mp;
782 	struct file *fp;
783 	int error;
784 
785 	/* getvnode() will use the descriptor for us */
786 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
787 		return (error);
788 	vp = (struct vnode *)fp->f_data;
789 
790 	VREF(vp);
791 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
792 	if (vp->v_type != VDIR)
793 		error = ENOTDIR;
794 	else
795 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
796 	while (!error && (mp = vp->v_mountedhere) != NULL) {
797 		if (vfs_busy(mp, 0, 0))
798 			continue;
799 		error = VFS_ROOT(mp, &tdp);
800 		vfs_unbusy(mp);
801 		if (error)
802 			break;
803 		vput(vp);
804 		vp = tdp;
805 	}
806 	if (error) {
807 		vput(vp);
808 		goto out;
809 	}
810 	VOP_UNLOCK(vp, 0);
811 
812 	/*
813 	 * Disallow changing to a directory not under the process's
814 	 * current root directory (if there is one).
815 	 */
816 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
817 		vrele(vp);
818 		error = EPERM;	/* operation not permitted */
819 		goto out;
820 	}
821 
822 	vrele(cwdi->cwdi_cdir);
823 	cwdi->cwdi_cdir = vp;
824  out:
825 	FILE_UNUSE(fp, p);
826 	return (error);
827 }
828 
829 /*
830  * Change this process's notion of the root directory to a given file descriptor.
831  */
832 
833 int
834 sys_fchroot(p, v, retval)
835 	struct proc *p;
836 	void *v;
837 	register_t *retval;
838 {
839 	struct sys_fchroot_args *uap = v;
840 	struct filedesc *fdp = p->p_fd;
841 	struct cwdinfo *cwdi = p->p_cwdi;
842 	struct vnode	*vp;
843 	struct file	*fp;
844 	int		 error;
845 
846 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
847 		return error;
848 	/* getvnode() will use the descriptor for us */
849 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
850 		return error;
851 	vp = (struct vnode *) fp->f_data;
852 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
853 	if (vp->v_type != VDIR)
854 		error = ENOTDIR;
855 	else
856 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
857 	VOP_UNLOCK(vp, 0);
858 	if (error)
859 		goto out;
860 	VREF(vp);
861 
862 	/*
863 	 * Prevent escaping from chroot by putting the root under
864 	 * the working directory.  Silently chdir to / if we aren't
865 	 * already there.
866 	 */
867 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
868 		/*
869 		 * XXX would be more failsafe to change directory to a
870 		 * deadfs node here instead
871 		 */
872 		vrele(cwdi->cwdi_cdir);
873 		VREF(vp);
874 		cwdi->cwdi_cdir = vp;
875 	}
876 
877 	if (cwdi->cwdi_rdir != NULL)
878 		vrele(cwdi->cwdi_rdir);
879 	cwdi->cwdi_rdir = vp;
880  out:
881 	FILE_UNUSE(fp, p);
882 	return (error);
883 }
884 
885 
886 
887 /*
888  * Change current working directory (``.'').
889  */
890 /* ARGSUSED */
891 int
892 sys_chdir(p, v, retval)
893 	struct proc *p;
894 	void *v;
895 	register_t *retval;
896 {
897 	struct sys_chdir_args /* {
898 		syscallarg(const char *) path;
899 	} */ *uap = v;
900 	struct cwdinfo *cwdi = p->p_cwdi;
901 	int error;
902 	struct nameidata nd;
903 
904 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
905 	    SCARG(uap, path), p);
906 	if ((error = change_dir(&nd, p)) != 0)
907 		return (error);
908 	vrele(cwdi->cwdi_cdir);
909 	cwdi->cwdi_cdir = nd.ni_vp;
910 	return (0);
911 }
912 
913 /*
914  * Change notion of root (``/'') directory.
915  */
916 /* ARGSUSED */
917 int
918 sys_chroot(p, v, retval)
919 	struct proc *p;
920 	void *v;
921 	register_t *retval;
922 {
923 	struct sys_chroot_args /* {
924 		syscallarg(const char *) path;
925 	} */ *uap = v;
926 	struct cwdinfo *cwdi = p->p_cwdi;
927 	struct vnode *vp;
928 	int error;
929 	struct nameidata nd;
930 
931 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
932 		return (error);
933 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
934 	    SCARG(uap, path), p);
935 	if ((error = change_dir(&nd, p)) != 0)
936 		return (error);
937 	if (cwdi->cwdi_rdir != NULL)
938 		vrele(cwdi->cwdi_rdir);
939 	vp = nd.ni_vp;
940 	cwdi->cwdi_rdir = vp;
941 
942 	/*
943 	 * Prevent escaping from chroot by putting the root under
944 	 * the working directory.  Silently chdir to / if we aren't
945 	 * already there.
946 	 */
947 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
948 		/*
949 		 * XXX would be more failsafe to change directory to a
950 		 * deadfs node here instead
951 		 */
952 		vrele(cwdi->cwdi_cdir);
953 		VREF(vp);
954 		cwdi->cwdi_cdir = vp;
955 	}
956 
957 	return (0);
958 }
959 
960 /*
961  * Common routine for chroot and chdir.
962  */
963 static int
964 change_dir(ndp, p)
965 	struct nameidata *ndp;
966 	struct proc *p;
967 {
968 	struct vnode *vp;
969 	int error;
970 
971 	if ((error = namei(ndp)) != 0)
972 		return (error);
973 	vp = ndp->ni_vp;
974 	if (vp->v_type != VDIR)
975 		error = ENOTDIR;
976 	else
977 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
978 
979 	if (error)
980 		vput(vp);
981 	else
982 		VOP_UNLOCK(vp, 0);
983 	return (error);
984 }
985 
986 /*
987  * Check permissions, allocate an open file structure,
988  * and call the device open routine if any.
989  */
990 int
991 sys_open(p, v, retval)
992 	struct proc *p;
993 	void *v;
994 	register_t *retval;
995 {
996 	struct sys_open_args /* {
997 		syscallarg(const char *) path;
998 		syscallarg(int) flags;
999 		syscallarg(int) mode;
1000 	} */ *uap = v;
1001 	struct cwdinfo *cwdi = p->p_cwdi;
1002 	struct filedesc *fdp = p->p_fd;
1003 	struct file *fp;
1004 	struct vnode *vp;
1005 	int flags, cmode;
1006 	int type, indx, error;
1007 	struct flock lf;
1008 	struct nameidata nd;
1009 
1010 	flags = FFLAGS(SCARG(uap, flags));
1011 	if ((flags & (FREAD | FWRITE)) == 0)
1012 		return (EINVAL);
1013 	/* falloc() will use the file descriptor for us */
1014 	if ((error = falloc(p, &fp, &indx)) != 0)
1015 		return (error);
1016 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1017 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1018 	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
1019 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1020 		FILE_UNUSE(fp, p);
1021 		ffree(fp);
1022 		if ((error == ENODEV || error == ENXIO) &&
1023 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
1024 		    (error =
1025 			dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1026 			*retval = indx;
1027 			return (0);
1028 		}
1029 		if (error == ERESTART)
1030 			error = EINTR;
1031 		fdremove(fdp, indx);
1032 		return (error);
1033 	}
1034 	p->p_dupfd = 0;
1035 	vp = nd.ni_vp;
1036 	fp->f_flag = flags & FMASK;
1037 	fp->f_type = DTYPE_VNODE;
1038 	fp->f_ops = &vnops;
1039 	fp->f_data = (caddr_t)vp;
1040 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1041 		lf.l_whence = SEEK_SET;
1042 		lf.l_start = 0;
1043 		lf.l_len = 0;
1044 		if (flags & O_EXLOCK)
1045 			lf.l_type = F_WRLCK;
1046 		else
1047 			lf.l_type = F_RDLCK;
1048 		type = F_FLOCK;
1049 		if ((flags & FNONBLOCK) == 0)
1050 			type |= F_WAIT;
1051 		VOP_UNLOCK(vp, 0);
1052 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1053 		if (error) {
1054 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1055 			FILE_UNUSE(fp, p);
1056 			ffree(fp);
1057 			fdremove(fdp, indx);
1058 			return (error);
1059 		}
1060 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1061 		fp->f_flag |= FHASLOCK;
1062 	}
1063 	VOP_UNLOCK(vp, 0);
1064 	*retval = indx;
1065 	FILE_SET_MATURE(fp);
1066 	FILE_UNUSE(fp, p);
1067 	return (0);
1068 }
1069 
1070 /*
1071  * Get file handle system call
1072  */
1073 int
1074 sys_getfh(p, v, retval)
1075 	struct proc *p;
1076 	void *v;
1077 	register_t *retval;
1078 {
1079 	struct sys_getfh_args /* {
1080 		syscallarg(char *) fname;
1081 		syscallarg(fhandle_t *) fhp;
1082 	} */ *uap = v;
1083 	struct vnode *vp;
1084 	fhandle_t fh;
1085 	int error;
1086 	struct nameidata nd;
1087 
1088 	/*
1089 	 * Must be super user
1090 	 */
1091 	error = suser(p->p_ucred, &p->p_acflag);
1092 	if (error)
1093 		return (error);
1094 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1095 	    SCARG(uap, fname), p);
1096 	error = namei(&nd);
1097 	if (error)
1098 		return (error);
1099 	vp = nd.ni_vp;
1100 	memset((caddr_t)&fh, 0, sizeof(fh));
1101 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1102 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1103 	vput(vp);
1104 	if (error)
1105 		return (error);
1106 	error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1107 	return (error);
1108 }
1109 
1110 /*
1111  * Open a file given a file handle.
1112  *
1113  * Check permissions, allocate an open file structure,
1114  * and call the device open routine if any.
1115  */
1116 int
1117 sys_fhopen(p, v, retval)
1118 	struct proc *p;
1119 	void *v;
1120 	register_t *retval;
1121 {
1122 	struct sys_fhopen_args /* {
1123 		syscallarg(const fhandle_t *) fhp;
1124 		syscallarg(int) flags;
1125 	} */ *uap = v;
1126 	struct filedesc *fdp = p->p_fd;
1127 	struct file *fp;
1128 	struct vnode *vp = NULL;
1129 	struct mount *mp;
1130 	struct ucred *cred = p->p_ucred;
1131 	int flags;
1132 	struct file *nfp;
1133 	int type, indx, error=0;
1134 	struct flock lf;
1135 	struct vattr va;
1136 	fhandle_t fh;
1137 
1138 	/*
1139 	 * Must be super user
1140 	 */
1141 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1142 		return (error);
1143 
1144 	flags = FFLAGS(SCARG(uap, flags));
1145 	if ((flags & (FREAD | FWRITE)) == 0)
1146 		return (EINVAL);
1147 	if ((flags & O_CREAT))
1148 		return (EINVAL);
1149 	/* falloc() will use the file descriptor for us */
1150 	if ((error = falloc(p, &nfp, &indx)) != 0)
1151 		return (error);
1152 	fp = nfp;
1153 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1154 		goto bad;
1155 
1156 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1157 		error = ESTALE;
1158 		goto bad;
1159 	}
1160 
1161 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1162 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1163 		goto bad;
1164 	}
1165 
1166 	/* Now do an effective vn_open */
1167 
1168 	if (vp->v_type == VSOCK) {
1169 		error = EOPNOTSUPP;
1170 		goto bad;
1171 	}
1172 	if (flags & FREAD) {
1173 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1174 			goto bad;
1175 	}
1176 	if (flags & (FWRITE | O_TRUNC)) {
1177 		if (vp->v_type == VDIR) {
1178 			error = EISDIR;
1179 			goto bad;
1180 		}
1181 		if ((error = vn_writechk(vp)) != 0 ||
1182 		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1183 			goto bad;
1184 	}
1185 	if (flags & O_TRUNC) {
1186 		VOP_UNLOCK(vp, 0);			/* XXX */
1187 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1188 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1189 		VATTR_NULL(&va);
1190 		va.va_size = 0;
1191 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1192 			goto bad;
1193 	}
1194 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1195 		goto bad;
1196 	if (vp->v_type == VREG &&
1197 	    uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1198 		error = EIO;
1199 		goto bad;
1200 	}
1201 	if (flags & FWRITE)
1202 		vp->v_writecount++;
1203 
1204 	/* done with modified vn_open, now finish what sys_open does. */
1205 
1206 	fp->f_flag = flags & FMASK;
1207 	fp->f_type = DTYPE_VNODE;
1208 	fp->f_ops = &vnops;
1209 	fp->f_data = (caddr_t)vp;
1210 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1211 		lf.l_whence = SEEK_SET;
1212 		lf.l_start = 0;
1213 		lf.l_len = 0;
1214 		if (flags & O_EXLOCK)
1215 			lf.l_type = F_WRLCK;
1216 		else
1217 			lf.l_type = F_RDLCK;
1218 		type = F_FLOCK;
1219 		if ((flags & FNONBLOCK) == 0)
1220 			type |= F_WAIT;
1221 		VOP_UNLOCK(vp, 0);
1222 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1223 		if (error) {
1224 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1225 			FILE_UNUSE(fp, p);
1226 			ffree(fp);
1227 			fdremove(fdp, indx);
1228 			return (error);
1229 		}
1230 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1231 		fp->f_flag |= FHASLOCK;
1232 	}
1233 	VOP_UNLOCK(vp, 0);
1234 	*retval = indx;
1235 	FILE_SET_MATURE(fp);
1236 	FILE_UNUSE(fp, p);
1237 	return (0);
1238 
1239 bad:
1240 	FILE_UNUSE(fp, p);
1241 	ffree(fp);
1242 	fdremove(fdp, indx);
1243 	if (vp != NULL)
1244 		vput(vp);
1245 	return (error);
1246 }
1247 
1248 /* ARGSUSED */
1249 int
1250 sys_fhstat(p, v, retval)
1251 	struct proc *p;
1252 	void *v;
1253 	register_t *retval;
1254 {
1255 	struct sys_fhstat_args /* {
1256 		syscallarg(const fhandle_t *) fhp;
1257 		syscallarg(struct stat *) sb;
1258 	} */ *uap = v;
1259 	struct stat sb;
1260 	int error;
1261 	fhandle_t fh;
1262 	struct mount *mp;
1263 	struct vnode *vp;
1264 
1265 	/*
1266 	 * Must be super user
1267 	 */
1268 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1269 		return (error);
1270 
1271 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1272 		return (error);
1273 
1274 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1275 		return (ESTALE);
1276 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1277 		return (error);
1278 	error = vn_stat(vp, &sb, p);
1279 	vput(vp);
1280 	if (error)
1281 		return (error);
1282 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1283 	return (error);
1284 }
1285 
1286 /* ARGSUSED */
1287 int
1288 sys_fhstatfs(p, v, retval)
1289 	struct proc *p;
1290 	void *v;
1291 	register_t *retval;
1292 {
1293 	struct sys_fhstatfs_args /*
1294 		syscallarg(const fhandle_t *) fhp;
1295 		syscallarg(struct statfs *) buf;
1296 	} */ *uap = v;
1297 	struct statfs sp;
1298 	fhandle_t fh;
1299 	struct mount *mp;
1300 	struct vnode *vp;
1301 	int error;
1302 
1303 	/*
1304 	 * Must be super user
1305 	 */
1306 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1307 		return (error);
1308 
1309 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1310 		return (error);
1311 
1312 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1313 		return (ESTALE);
1314 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1315 		return (error);
1316 	mp = vp->v_mount;
1317 	vput(vp);
1318 	if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1319 		return (error);
1320 	sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1321 	sp.f_oflags = sp.f_flags & 0xffff;
1322 	return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1323 }
1324 
1325 /*
1326  * Create a special file.
1327  */
1328 /* ARGSUSED */
1329 int
1330 sys_mknod(p, v, retval)
1331 	struct proc *p;
1332 	void *v;
1333 	register_t *retval;
1334 {
1335 	struct sys_mknod_args /* {
1336 		syscallarg(const char *) path;
1337 		syscallarg(int) mode;
1338 		syscallarg(int) dev;
1339 	} */ *uap = v;
1340 	struct vnode *vp;
1341 	struct vattr vattr;
1342 	int error;
1343 	int whiteout = 0;
1344 	struct nameidata nd;
1345 
1346 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1347 		return (error);
1348 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1349 	if ((error = namei(&nd)) != 0)
1350 		return (error);
1351 	vp = nd.ni_vp;
1352 	if (vp != NULL)
1353 		error = EEXIST;
1354 	else {
1355 		VATTR_NULL(&vattr);
1356 		vattr.va_mode =
1357 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1358 		vattr.va_rdev = SCARG(uap, dev);
1359 		whiteout = 0;
1360 
1361 		switch (SCARG(uap, mode) & S_IFMT) {
1362 		case S_IFMT:	/* used by badsect to flag bad sectors */
1363 			vattr.va_type = VBAD;
1364 			break;
1365 		case S_IFCHR:
1366 			vattr.va_type = VCHR;
1367 			break;
1368 		case S_IFBLK:
1369 			vattr.va_type = VBLK;
1370 			break;
1371 		case S_IFWHT:
1372 			whiteout = 1;
1373 			break;
1374 		default:
1375 			error = EINVAL;
1376 			break;
1377 		}
1378 	}
1379 	if (!error) {
1380 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1381 		if (whiteout) {
1382 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1383 			if (error)
1384 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1385 			vput(nd.ni_dvp);
1386 		} else {
1387 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1388 						&nd.ni_cnd, &vattr);
1389 			if (error == 0)
1390 				vput(nd.ni_vp);
1391 		}
1392 	} else {
1393 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1394 		if (nd.ni_dvp == vp)
1395 			vrele(nd.ni_dvp);
1396 		else
1397 			vput(nd.ni_dvp);
1398 		if (vp)
1399 			vrele(vp);
1400 	}
1401 	return (error);
1402 }
1403 
1404 /*
1405  * Create a named pipe.
1406  */
1407 /* ARGSUSED */
1408 int
1409 sys_mkfifo(p, v, retval)
1410 	struct proc *p;
1411 	void *v;
1412 	register_t *retval;
1413 {
1414 	struct sys_mkfifo_args /* {
1415 		syscallarg(const char *) path;
1416 		syscallarg(int) mode;
1417 	} */ *uap = v;
1418 	struct vattr vattr;
1419 	int error;
1420 	struct nameidata nd;
1421 
1422 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1423 	if ((error = namei(&nd)) != 0)
1424 		return (error);
1425 	if (nd.ni_vp != NULL) {
1426 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1427 		if (nd.ni_dvp == nd.ni_vp)
1428 			vrele(nd.ni_dvp);
1429 		else
1430 			vput(nd.ni_dvp);
1431 		vrele(nd.ni_vp);
1432 		return (EEXIST);
1433 	}
1434 	VATTR_NULL(&vattr);
1435 	vattr.va_type = VFIFO;
1436 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1437 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1438 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1439 	if (error == 0)
1440 		vput(nd.ni_vp);
1441 	return (error);
1442 }
1443 
1444 /*
1445  * Make a hard file link.
1446  */
1447 /* ARGSUSED */
1448 int
1449 sys_link(p, v, retval)
1450 	struct proc *p;
1451 	void *v;
1452 	register_t *retval;
1453 {
1454 	struct sys_link_args /* {
1455 		syscallarg(const char *) path;
1456 		syscallarg(const char *) link;
1457 	} */ *uap = v;
1458 	struct vnode *vp;
1459 	struct nameidata nd;
1460 	int error;
1461 
1462 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1463 	if ((error = namei(&nd)) != 0)
1464 		return (error);
1465 	vp = nd.ni_vp;
1466 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1467 	if ((error = namei(&nd)) != 0)
1468 		goto out;
1469 	if (nd.ni_vp) {
1470 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1471 		if (nd.ni_dvp == nd.ni_vp)
1472 			vrele(nd.ni_dvp);
1473 		else
1474 			vput(nd.ni_dvp);
1475 		vrele(nd.ni_vp);
1476 		error = EEXIST;
1477 		goto out;
1478 	}
1479 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1480 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1481 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1482 out:
1483 	vrele(vp);
1484 	return (error);
1485 }
1486 
1487 /*
1488  * Make a symbolic link.
1489  */
1490 /* ARGSUSED */
1491 int
1492 sys_symlink(p, v, retval)
1493 	struct proc *p;
1494 	void *v;
1495 	register_t *retval;
1496 {
1497 	struct sys_symlink_args /* {
1498 		syscallarg(const char *) path;
1499 		syscallarg(const char *) link;
1500 	} */ *uap = v;
1501 	struct vattr vattr;
1502 	char *path;
1503 	int error;
1504 	struct nameidata nd;
1505 
1506 	path = PNBUF_GET();
1507 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1508 	if (error)
1509 		goto out;
1510 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1511 	if ((error = namei(&nd)) != 0)
1512 		goto out;
1513 	if (nd.ni_vp) {
1514 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1515 		if (nd.ni_dvp == nd.ni_vp)
1516 			vrele(nd.ni_dvp);
1517 		else
1518 			vput(nd.ni_dvp);
1519 		vrele(nd.ni_vp);
1520 		error = EEXIST;
1521 		goto out;
1522 	}
1523 	VATTR_NULL(&vattr);
1524 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1525 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1526 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1527 	if (error == 0)
1528 		vput(nd.ni_vp);
1529 out:
1530 	PNBUF_PUT(path);
1531 	return (error);
1532 }
1533 
1534 /*
1535  * Delete a whiteout from the filesystem.
1536  */
1537 /* ARGSUSED */
1538 int
1539 sys_undelete(p, v, retval)
1540 	struct proc *p;
1541 	void *v;
1542 	register_t *retval;
1543 {
1544 	struct sys_undelete_args /* {
1545 		syscallarg(const char *) path;
1546 	} */ *uap = v;
1547 	int error;
1548 	struct nameidata nd;
1549 
1550 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1551 	    SCARG(uap, path), p);
1552 	error = namei(&nd);
1553 	if (error)
1554 		return (error);
1555 
1556 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1557 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1558 		if (nd.ni_dvp == nd.ni_vp)
1559 			vrele(nd.ni_dvp);
1560 		else
1561 			vput(nd.ni_dvp);
1562 		if (nd.ni_vp)
1563 			vrele(nd.ni_vp);
1564 		return (EEXIST);
1565 	}
1566 
1567 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1568 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1569 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1570 	vput(nd.ni_dvp);
1571 	return (error);
1572 }
1573 
1574 /*
1575  * Delete a name from the filesystem.
1576  */
1577 /* ARGSUSED */
1578 int
1579 sys_unlink(p, v, retval)
1580 	struct proc *p;
1581 	void *v;
1582 	register_t *retval;
1583 {
1584 	struct sys_unlink_args /* {
1585 		syscallarg(const char *) path;
1586 	} */ *uap = v;
1587 	struct vnode *vp;
1588 	int error;
1589 	struct nameidata nd;
1590 
1591 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1592 	    SCARG(uap, path), p);
1593 	if ((error = namei(&nd)) != 0)
1594 		return (error);
1595 	vp = nd.ni_vp;
1596 
1597 	/*
1598 	 * The root of a mounted filesystem cannot be deleted.
1599 	 */
1600 	if (vp->v_flag & VROOT) {
1601 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1602 		if (nd.ni_dvp == vp)
1603 			vrele(nd.ni_dvp);
1604 		else
1605 			vput(nd.ni_dvp);
1606 		vput(vp);
1607 		error = EBUSY;
1608 		goto out;
1609 	}
1610 
1611 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1612 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1613 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1614 out:
1615 	return (error);
1616 }
1617 
1618 /*
1619  * Reposition read/write file offset.
1620  */
1621 int
1622 sys_lseek(p, v, retval)
1623 	struct proc *p;
1624 	void *v;
1625 	register_t *retval;
1626 {
1627 	struct sys_lseek_args /* {
1628 		syscallarg(int) fd;
1629 		syscallarg(int) pad;
1630 		syscallarg(off_t) offset;
1631 		syscallarg(int) whence;
1632 	} */ *uap = v;
1633 	struct ucred *cred = p->p_ucred;
1634 	struct filedesc *fdp = p->p_fd;
1635 	struct file *fp;
1636 	struct vnode *vp;
1637 	struct vattr vattr;
1638 	off_t newoff;
1639 	int error;
1640 
1641 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1642 		return (EBADF);
1643 
1644 	FILE_USE(fp);
1645 
1646 	vp = (struct vnode *)fp->f_data;
1647 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1648 		error = ESPIPE;
1649 		goto out;
1650 	}
1651 
1652 	switch (SCARG(uap, whence)) {
1653 	case SEEK_CUR:
1654 		newoff = fp->f_offset + SCARG(uap, offset);
1655 		break;
1656 	case SEEK_END:
1657 		error = VOP_GETATTR(vp, &vattr, cred, p);
1658 		if (error)
1659 			goto out;
1660 		newoff = SCARG(uap, offset) + vattr.va_size;
1661 		break;
1662 	case SEEK_SET:
1663 		newoff = SCARG(uap, offset);
1664 		break;
1665 	default:
1666 		error = EINVAL;
1667 		goto out;
1668 	}
1669 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1670 		goto out;
1671 
1672 	*(off_t *)retval = fp->f_offset = newoff;
1673  out:
1674 	FILE_UNUSE(fp, p);
1675 	return (error);
1676 }
1677 
1678 /*
1679  * Positional read system call.
1680  */
1681 int
1682 sys_pread(p, v, retval)
1683 	struct proc *p;
1684 	void *v;
1685 	register_t *retval;
1686 {
1687 	struct sys_pread_args /* {
1688 		syscallarg(int) fd;
1689 		syscallarg(void *) buf;
1690 		syscallarg(size_t) nbyte;
1691 		syscallarg(off_t) offset;
1692 	} */ *uap = v;
1693 	struct filedesc *fdp = p->p_fd;
1694 	struct file *fp;
1695 	struct vnode *vp;
1696 	off_t offset;
1697 	int error, fd = SCARG(uap, fd);
1698 
1699 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1700 		return (EBADF);
1701 
1702 	if ((fp->f_flag & FREAD) == 0)
1703 		return (EBADF);
1704 
1705 	FILE_USE(fp);
1706 
1707 	vp = (struct vnode *)fp->f_data;
1708 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1709 		error = ESPIPE;
1710 		goto out;
1711 	}
1712 
1713 	offset = SCARG(uap, offset);
1714 
1715 	/*
1716 	 * XXX This works because no file systems actually
1717 	 * XXX take any action on the seek operation.
1718 	 */
1719 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1720 		goto out;
1721 
1722 	/* dofileread() will unuse the descriptor for us */
1723 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1724 	    &offset, 0, retval));
1725 
1726  out:
1727 	FILE_UNUSE(fp, p);
1728 	return (error);
1729 }
1730 
1731 /*
1732  * Positional scatter read system call.
1733  */
1734 int
1735 sys_preadv(p, v, retval)
1736 	struct proc *p;
1737 	void *v;
1738 	register_t *retval;
1739 {
1740 	struct sys_preadv_args /* {
1741 		syscallarg(int) fd;
1742 		syscallarg(const struct iovec *) iovp;
1743 		syscallarg(int) iovcnt;
1744 		syscallarg(off_t) offset;
1745 	} */ *uap = v;
1746 	struct filedesc *fdp = p->p_fd;
1747 	struct file *fp;
1748 	struct vnode *vp;
1749 	off_t offset;
1750 	int error, fd = SCARG(uap, fd);
1751 
1752 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1753 		return (EBADF);
1754 
1755 	if ((fp->f_flag & FREAD) == 0)
1756 		return (EBADF);
1757 
1758 	FILE_USE(fp);
1759 
1760 	vp = (struct vnode *)fp->f_data;
1761 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1762 		error = ESPIPE;
1763 		goto out;
1764 	}
1765 
1766 	offset = SCARG(uap, offset);
1767 
1768 	/*
1769 	 * XXX This works because no file systems actually
1770 	 * XXX take any action on the seek operation.
1771 	 */
1772 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1773 		goto out;
1774 
1775 	/* dofilereadv() will unuse the descriptor for us */
1776 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1777 	    &offset, 0, retval));
1778 
1779  out:
1780 	FILE_UNUSE(fp, p);
1781 	return (error);
1782 }
1783 
1784 /*
1785  * Positional write system call.
1786  */
1787 int
1788 sys_pwrite(p, v, retval)
1789 	struct proc *p;
1790 	void *v;
1791 	register_t *retval;
1792 {
1793 	struct sys_pwrite_args /* {
1794 		syscallarg(int) fd;
1795 		syscallarg(const void *) buf;
1796 		syscallarg(size_t) nbyte;
1797 		syscallarg(off_t) offset;
1798 	} */ *uap = v;
1799 	struct filedesc *fdp = p->p_fd;
1800 	struct file *fp;
1801 	struct vnode *vp;
1802 	off_t offset;
1803 	int error, fd = SCARG(uap, fd);
1804 
1805 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1806 		return (EBADF);
1807 
1808 	if ((fp->f_flag & FWRITE) == 0)
1809 		return (EBADF);
1810 
1811 	FILE_USE(fp);
1812 
1813 	vp = (struct vnode *)fp->f_data;
1814 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1815 		error = ESPIPE;
1816 		goto out;
1817 	}
1818 
1819 	offset = SCARG(uap, offset);
1820 
1821 	/*
1822 	 * XXX This works because no file systems actually
1823 	 * XXX take any action on the seek operation.
1824 	 */
1825 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1826 		goto out;
1827 
1828 	/* dofilewrite() will unuse the descriptor for us */
1829 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1830 	    &offset, 0, retval));
1831 
1832  out:
1833 	FILE_UNUSE(fp, p);
1834 	return (error);
1835 }
1836 
1837 /*
1838  * Positional gather write system call.
1839  */
1840 int
1841 sys_pwritev(p, v, retval)
1842 	struct proc *p;
1843 	void *v;
1844 	register_t *retval;
1845 {
1846 	struct sys_pwritev_args /* {
1847 		syscallarg(int) fd;
1848 		syscallarg(const struct iovec *) iovp;
1849 		syscallarg(int) iovcnt;
1850 		syscallarg(off_t) offset;
1851 	} */ *uap = v;
1852 	struct filedesc *fdp = p->p_fd;
1853 	struct file *fp;
1854 	struct vnode *vp;
1855 	off_t offset;
1856 	int error, fd = SCARG(uap, fd);
1857 
1858 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1859 		return (EBADF);
1860 
1861 	if ((fp->f_flag & FWRITE) == 0)
1862 		return (EBADF);
1863 
1864 	FILE_USE(fp);
1865 
1866 	vp = (struct vnode *)fp->f_data;
1867 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1868 		error = ESPIPE;
1869 		goto out;
1870 	}
1871 
1872 	offset = SCARG(uap, offset);
1873 
1874 	/*
1875 	 * XXX This works because no file systems actually
1876 	 * XXX take any action on the seek operation.
1877 	 */
1878 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1879 		goto out;
1880 
1881 	/* dofilewritev() will unuse the descriptor for us */
1882 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1883 	    &offset, 0, retval));
1884 
1885  out:
1886 	FILE_UNUSE(fp, p);
1887 	return (error);
1888 }
1889 
1890 /*
1891  * Check access permissions.
1892  */
1893 int
1894 sys_access(p, v, retval)
1895 	struct proc *p;
1896 	void *v;
1897 	register_t *retval;
1898 {
1899 	struct sys_access_args /* {
1900 		syscallarg(const char *) path;
1901 		syscallarg(int) flags;
1902 	} */ *uap = v;
1903 	struct ucred *cred = crget();
1904 	struct vnode *vp;
1905 	int error, flags;
1906 	struct nameidata nd;
1907 
1908 	(void)memcpy(cred, p->p_ucred, sizeof(*cred));
1909 	cred->cr_ref = 1;
1910 	cred->cr_uid = p->p_cred->p_ruid;
1911 	cred->cr_gid = p->p_cred->p_rgid;
1912 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1913 	    SCARG(uap, path), p);
1914 	/* Override default credentials */
1915 	nd.ni_cnd.cn_cred = cred;
1916 	if ((error = namei(&nd)) != 0)
1917 		goto out;
1918 	vp = nd.ni_vp;
1919 
1920 	/* Flags == 0 means only check for existence. */
1921 	if (SCARG(uap, flags)) {
1922 		flags = 0;
1923 		if (SCARG(uap, flags) & R_OK)
1924 			flags |= VREAD;
1925 		if (SCARG(uap, flags) & W_OK)
1926 			flags |= VWRITE;
1927 		if (SCARG(uap, flags) & X_OK)
1928 			flags |= VEXEC;
1929 
1930 		error = VOP_ACCESS(vp, flags, cred, p);
1931 		if (!error && (flags & VWRITE))
1932 			error = vn_writechk(vp);
1933 	}
1934 	vput(vp);
1935 out:
1936 	crfree(cred);
1937 	return (error);
1938 }
1939 
1940 /*
1941  * Get file status; this version follows links.
1942  */
1943 /* ARGSUSED */
1944 int
1945 sys___stat13(p, v, retval)
1946 	struct proc *p;
1947 	void *v;
1948 	register_t *retval;
1949 {
1950 	struct sys___stat13_args /* {
1951 		syscallarg(const char *) path;
1952 		syscallarg(struct stat *) ub;
1953 	} */ *uap = v;
1954 	struct stat sb;
1955 	int error;
1956 	struct nameidata nd;
1957 
1958 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1959 	    SCARG(uap, path), p);
1960 	if ((error = namei(&nd)) != 0)
1961 		return (error);
1962 	error = vn_stat(nd.ni_vp, &sb, p);
1963 	vput(nd.ni_vp);
1964 	if (error)
1965 		return (error);
1966 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1967 	return (error);
1968 }
1969 
1970 /*
1971  * Get file status; this version does not follow links.
1972  */
1973 /* ARGSUSED */
1974 int
1975 sys___lstat13(p, v, retval)
1976 	struct proc *p;
1977 	void *v;
1978 	register_t *retval;
1979 {
1980 	struct sys___lstat13_args /* {
1981 		syscallarg(const char *) path;
1982 		syscallarg(struct stat *) ub;
1983 	} */ *uap = v;
1984 	struct stat sb;
1985 	int error;
1986 	struct nameidata nd;
1987 
1988 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1989 	    SCARG(uap, path), p);
1990 	if ((error = namei(&nd)) != 0)
1991 		return (error);
1992 	error = vn_stat(nd.ni_vp, &sb, p);
1993 	vput(nd.ni_vp);
1994 	if (error)
1995 		return (error);
1996 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1997 	return (error);
1998 }
1999 
2000 /*
2001  * Get configurable pathname variables.
2002  */
2003 /* ARGSUSED */
2004 int
2005 sys_pathconf(p, v, retval)
2006 	struct proc *p;
2007 	void *v;
2008 	register_t *retval;
2009 {
2010 	struct sys_pathconf_args /* {
2011 		syscallarg(const char *) path;
2012 		syscallarg(int) name;
2013 	} */ *uap = v;
2014 	int error;
2015 	struct nameidata nd;
2016 
2017 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2018 	    SCARG(uap, path), p);
2019 	if ((error = namei(&nd)) != 0)
2020 		return (error);
2021 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2022 	vput(nd.ni_vp);
2023 	return (error);
2024 }
2025 
2026 /*
2027  * Return target name of a symbolic link.
2028  */
2029 /* ARGSUSED */
2030 int
2031 sys_readlink(p, v, retval)
2032 	struct proc *p;
2033 	void *v;
2034 	register_t *retval;
2035 {
2036 	struct sys_readlink_args /* {
2037 		syscallarg(const char *) path;
2038 		syscallarg(char *) buf;
2039 		syscallarg(size_t) count;
2040 	} */ *uap = v;
2041 	struct vnode *vp;
2042 	struct iovec aiov;
2043 	struct uio auio;
2044 	int error;
2045 	struct nameidata nd;
2046 
2047 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2048 	    SCARG(uap, path), p);
2049 	if ((error = namei(&nd)) != 0)
2050 		return (error);
2051 	vp = nd.ni_vp;
2052 	if (vp->v_type != VLNK)
2053 		error = EINVAL;
2054 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2055 	    (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2056 		aiov.iov_base = SCARG(uap, buf);
2057 		aiov.iov_len = SCARG(uap, count);
2058 		auio.uio_iov = &aiov;
2059 		auio.uio_iovcnt = 1;
2060 		auio.uio_offset = 0;
2061 		auio.uio_rw = UIO_READ;
2062 		auio.uio_segflg = UIO_USERSPACE;
2063 		auio.uio_procp = p;
2064 		auio.uio_resid = SCARG(uap, count);
2065 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2066 	}
2067 	vput(vp);
2068 	*retval = SCARG(uap, count) - auio.uio_resid;
2069 	return (error);
2070 }
2071 
2072 /*
2073  * Change flags of a file given a path name.
2074  */
2075 /* ARGSUSED */
2076 int
2077 sys_chflags(p, v, retval)
2078 	struct proc *p;
2079 	void *v;
2080 	register_t *retval;
2081 {
2082 	struct sys_chflags_args /* {
2083 		syscallarg(const char *) path;
2084 		syscallarg(u_long) flags;
2085 	} */ *uap = v;
2086 	struct vnode *vp;
2087 	int error;
2088 	struct nameidata nd;
2089 
2090 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2091 	if ((error = namei(&nd)) != 0)
2092 		return (error);
2093 	vp = nd.ni_vp;
2094 	error = change_flags(vp, SCARG(uap, flags), p);
2095 	vput(vp);
2096 	return (error);
2097 }
2098 
2099 /*
2100  * Change flags of a file given a file descriptor.
2101  */
2102 /* ARGSUSED */
2103 int
2104 sys_fchflags(p, v, retval)
2105 	struct proc *p;
2106 	void *v;
2107 	register_t *retval;
2108 {
2109 	struct sys_fchflags_args /* {
2110 		syscallarg(int) fd;
2111 		syscallarg(u_long) flags;
2112 	} */ *uap = v;
2113 	struct vnode *vp;
2114 	struct file *fp;
2115 	int error;
2116 
2117 	/* getvnode() will use the descriptor for us */
2118 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2119 		return (error);
2120 	vp = (struct vnode *)fp->f_data;
2121 	error = change_flags(vp, SCARG(uap, flags), p);
2122 	VOP_UNLOCK(vp, 0);
2123 	FILE_UNUSE(fp, p);
2124 	return (error);
2125 }
2126 
2127 /*
2128  * Change flags of a file given a path name; this version does
2129  * not follow links.
2130  */
2131 int
2132 sys_lchflags(p, v, retval)
2133 	struct proc *p;
2134 	void *v;
2135 	register_t *retval;
2136 {
2137 	struct sys_lchflags_args /* {
2138 		syscallarg(const char *) path;
2139 		syscallarg(u_long) flags;
2140 	} */ *uap = v;
2141 	struct vnode *vp;
2142 	int error;
2143 	struct nameidata nd;
2144 
2145 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2146 	if ((error = namei(&nd)) != 0)
2147 		return (error);
2148 	vp = nd.ni_vp;
2149 	error = change_flags(vp, SCARG(uap, flags), p);
2150 	vput(vp);
2151 	return (error);
2152 }
2153 
2154 /*
2155  * Common routine to change flags of a file.
2156  */
2157 int
2158 change_flags(vp, flags, p)
2159 	struct vnode *vp;
2160 	u_long flags;
2161 	struct proc *p;
2162 {
2163 	struct vattr vattr;
2164 	int error;
2165 
2166 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2167 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2168 	/*
2169 	 * Non-superusers cannot change the flags on devices, even if they
2170 	 * own them.
2171 	 */
2172 	if (suser(p->p_ucred, &p->p_acflag) != 0) {
2173 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2174 			goto out;
2175 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2176 			error = EINVAL;
2177 			goto out;
2178 		}
2179 	}
2180 	VATTR_NULL(&vattr);
2181 	vattr.va_flags = flags;
2182 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2183 out:
2184 	return (error);
2185 }
2186 
2187 /*
2188  * Change mode of a file given path name; this version follows links.
2189  */
2190 /* ARGSUSED */
2191 int
2192 sys_chmod(p, v, retval)
2193 	struct proc *p;
2194 	void *v;
2195 	register_t *retval;
2196 {
2197 	struct sys_chmod_args /* {
2198 		syscallarg(const char *) path;
2199 		syscallarg(int) mode;
2200 	} */ *uap = v;
2201 	int error;
2202 	struct nameidata nd;
2203 
2204 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2205 	if ((error = namei(&nd)) != 0)
2206 		return (error);
2207 
2208 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2209 
2210 	vrele(nd.ni_vp);
2211 	return (error);
2212 }
2213 
2214 /*
2215  * Change mode of a file given a file descriptor.
2216  */
2217 /* ARGSUSED */
2218 int
2219 sys_fchmod(p, v, retval)
2220 	struct proc *p;
2221 	void *v;
2222 	register_t *retval;
2223 {
2224 	struct sys_fchmod_args /* {
2225 		syscallarg(int) fd;
2226 		syscallarg(int) mode;
2227 	} */ *uap = v;
2228 	struct file *fp;
2229 	int error;
2230 
2231 	/* getvnode() will use the descriptor for us */
2232 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2233 		return (error);
2234 
2235 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2236 	FILE_UNUSE(fp, p);
2237 	return (error);
2238 }
2239 
2240 /*
2241  * Change mode of a file given path name; this version does not follow links.
2242  */
2243 /* ARGSUSED */
2244 int
2245 sys_lchmod(p, v, retval)
2246 	struct proc *p;
2247 	void *v;
2248 	register_t *retval;
2249 {
2250 	struct sys_lchmod_args /* {
2251 		syscallarg(const char *) path;
2252 		syscallarg(int) mode;
2253 	} */ *uap = v;
2254 	int error;
2255 	struct nameidata nd;
2256 
2257 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2258 	if ((error = namei(&nd)) != 0)
2259 		return (error);
2260 
2261 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2262 
2263 	vrele(nd.ni_vp);
2264 	return (error);
2265 }
2266 
2267 /*
2268  * Common routine to set mode given a vnode.
2269  */
2270 static int
2271 change_mode(vp, mode, p)
2272 	struct vnode *vp;
2273 	int mode;
2274 	struct proc *p;
2275 {
2276 	struct vattr vattr;
2277 	int error;
2278 
2279 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2280 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2281 	VATTR_NULL(&vattr);
2282 	vattr.va_mode = mode & ALLPERMS;
2283 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2284 	VOP_UNLOCK(vp, 0);
2285 	return (error);
2286 }
2287 
2288 /*
2289  * Set ownership given a path name; this version follows links.
2290  */
2291 /* ARGSUSED */
2292 int
2293 sys_chown(p, v, retval)
2294 	struct proc *p;
2295 	void *v;
2296 	register_t *retval;
2297 {
2298 	struct sys_chown_args /* {
2299 		syscallarg(const char *) path;
2300 		syscallarg(uid_t) uid;
2301 		syscallarg(gid_t) gid;
2302 	} */ *uap = v;
2303 	int error;
2304 	struct nameidata nd;
2305 
2306 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2307 	if ((error = namei(&nd)) != 0)
2308 		return (error);
2309 
2310 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2311 
2312 	vrele(nd.ni_vp);
2313 	return (error);
2314 }
2315 
2316 /*
2317  * Set ownership given a path name; this version follows links.
2318  * Provides POSIX semantics.
2319  */
2320 /* ARGSUSED */
2321 int
2322 sys___posix_chown(p, v, retval)
2323 	struct proc *p;
2324 	void *v;
2325 	register_t *retval;
2326 {
2327 	struct sys_chown_args /* {
2328 		syscallarg(const char *) path;
2329 		syscallarg(uid_t) uid;
2330 		syscallarg(gid_t) gid;
2331 	} */ *uap = v;
2332 	int error;
2333 	struct nameidata nd;
2334 
2335 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2336 	if ((error = namei(&nd)) != 0)
2337 		return (error);
2338 
2339 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2340 
2341 	vrele(nd.ni_vp);
2342 	return (error);
2343 }
2344 
2345 /*
2346  * Set ownership given a file descriptor.
2347  */
2348 /* ARGSUSED */
2349 int
2350 sys_fchown(p, v, retval)
2351 	struct proc *p;
2352 	void *v;
2353 	register_t *retval;
2354 {
2355 	struct sys_fchown_args /* {
2356 		syscallarg(int) fd;
2357 		syscallarg(uid_t) uid;
2358 		syscallarg(gid_t) gid;
2359 	} */ *uap = v;
2360 	int error;
2361 	struct file *fp;
2362 
2363 	/* getvnode() will use the descriptor for us */
2364 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2365 		return (error);
2366 
2367 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2368 	    SCARG(uap, gid), p, 0);
2369 	FILE_UNUSE(fp, p);
2370 	return (error);
2371 }
2372 
2373 /*
2374  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2375  */
2376 /* ARGSUSED */
2377 int
2378 sys___posix_fchown(p, v, retval)
2379 	struct proc *p;
2380 	void *v;
2381 	register_t *retval;
2382 {
2383 	struct sys_fchown_args /* {
2384 		syscallarg(int) fd;
2385 		syscallarg(uid_t) uid;
2386 		syscallarg(gid_t) gid;
2387 	} */ *uap = v;
2388 	int error;
2389 	struct file *fp;
2390 
2391 	/* getvnode() will use the descriptor for us */
2392 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2393 		return (error);
2394 
2395 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2396 	    SCARG(uap, gid), p, 1);
2397 	FILE_UNUSE(fp, p);
2398 	return (error);
2399 }
2400 
2401 /*
2402  * Set ownership given a path name; this version does not follow links.
2403  */
2404 /* ARGSUSED */
2405 int
2406 sys_lchown(p, v, retval)
2407 	struct proc *p;
2408 	void *v;
2409 	register_t *retval;
2410 {
2411 	struct sys_lchown_args /* {
2412 		syscallarg(const char *) path;
2413 		syscallarg(uid_t) uid;
2414 		syscallarg(gid_t) gid;
2415 	} */ *uap = v;
2416 	int error;
2417 	struct nameidata nd;
2418 
2419 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2420 	if ((error = namei(&nd)) != 0)
2421 		return (error);
2422 
2423 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2424 
2425 	vrele(nd.ni_vp);
2426 	return (error);
2427 }
2428 
2429 /*
2430  * Set ownership given a path name; this version does not follow links.
2431  * Provides POSIX/XPG semantics.
2432  */
2433 /* ARGSUSED */
2434 int
2435 sys___posix_lchown(p, v, retval)
2436 	struct proc *p;
2437 	void *v;
2438 	register_t *retval;
2439 {
2440 	struct sys_lchown_args /* {
2441 		syscallarg(const char *) path;
2442 		syscallarg(uid_t) uid;
2443 		syscallarg(gid_t) gid;
2444 	} */ *uap = v;
2445 	int error;
2446 	struct nameidata nd;
2447 
2448 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2449 	if ((error = namei(&nd)) != 0)
2450 		return (error);
2451 
2452 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2453 
2454 	vrele(nd.ni_vp);
2455 	return (error);
2456 }
2457 
2458 /*
2459  * Common routine to set ownership given a vnode.
2460  */
2461 static int
2462 change_owner(vp, uid, gid, p, posix_semantics)
2463 	struct vnode *vp;
2464 	uid_t uid;
2465 	gid_t gid;
2466 	struct proc *p;
2467 	int posix_semantics;
2468 {
2469 	struct vattr vattr;
2470 	mode_t newmode;
2471 	int error;
2472 
2473 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2474 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2475 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2476 		goto out;
2477 
2478 #define CHANGED(x) ((int)(x) != -1)
2479 	newmode = vattr.va_mode;
2480 	if (posix_semantics) {
2481 		/*
2482 		 * POSIX/XPG semantics: if the caller is not the super-user,
2483 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2484 		 * the XPG consider the behaviour for calls by the super-user
2485 		 * implementation-defined; we leave the set-user-id and set-
2486 		 * group-id settings intact in that case.
2487 		 */
2488 		if (suser(p->p_ucred, NULL) != 0)
2489 			newmode &= ~(S_ISUID | S_ISGID);
2490 	} else {
2491 		/*
2492 		 * NetBSD semantics: when changing owner and/or group,
2493 		 * clear the respective bit(s).
2494 		 */
2495 		if (CHANGED(uid))
2496 			newmode &= ~S_ISUID;
2497 		if (CHANGED(gid))
2498 			newmode &= ~S_ISGID;
2499 	}
2500 	/* Update va_mode iff altered. */
2501 	if (vattr.va_mode == newmode)
2502 		newmode = VNOVAL;
2503 
2504 	VATTR_NULL(&vattr);
2505 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2506 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2507 	vattr.va_mode = newmode;
2508 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2509 #undef CHANGED
2510 
2511 out:
2512 	VOP_UNLOCK(vp, 0);
2513 	return (error);
2514 }
2515 
2516 /*
2517  * Set the access and modification times given a path name; this
2518  * version follows links.
2519  */
2520 /* ARGSUSED */
2521 int
2522 sys_utimes(p, v, retval)
2523 	struct proc *p;
2524 	void *v;
2525 	register_t *retval;
2526 {
2527 	struct sys_utimes_args /* {
2528 		syscallarg(const char *) path;
2529 		syscallarg(const struct timeval *) tptr;
2530 	} */ *uap = v;
2531 	int error;
2532 	struct nameidata nd;
2533 
2534 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2535 	if ((error = namei(&nd)) != 0)
2536 		return (error);
2537 
2538 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2539 
2540 	vrele(nd.ni_vp);
2541 	return (error);
2542 }
2543 
2544 /*
2545  * Set the access and modification times given a file descriptor.
2546  */
2547 /* ARGSUSED */
2548 int
2549 sys_futimes(p, v, retval)
2550 	struct proc *p;
2551 	void *v;
2552 	register_t *retval;
2553 {
2554 	struct sys_futimes_args /* {
2555 		syscallarg(int) fd;
2556 		syscallarg(const struct timeval *) tptr;
2557 	} */ *uap = v;
2558 	int error;
2559 	struct file *fp;
2560 
2561 	/* getvnode() will use the descriptor for us */
2562 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2563 		return (error);
2564 
2565 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2566 	FILE_UNUSE(fp, p);
2567 	return (error);
2568 }
2569 
2570 /*
2571  * Set the access and modification times given a path name; this
2572  * version does not follow links.
2573  */
2574 /* ARGSUSED */
2575 int
2576 sys_lutimes(p, v, retval)
2577 	struct proc *p;
2578 	void *v;
2579 	register_t *retval;
2580 {
2581 	struct sys_lutimes_args /* {
2582 		syscallarg(const char *) path;
2583 		syscallarg(const struct timeval *) tptr;
2584 	} */ *uap = v;
2585 	int error;
2586 	struct nameidata nd;
2587 
2588 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2589 	if ((error = namei(&nd)) != 0)
2590 		return (error);
2591 
2592 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2593 
2594 	vrele(nd.ni_vp);
2595 	return (error);
2596 }
2597 
2598 /*
2599  * Common routine to set access and modification times given a vnode.
2600  */
2601 static int
2602 change_utimes(vp, tptr, p)
2603 	struct vnode *vp;
2604 	const struct timeval *tptr;
2605 	struct proc *p;
2606 {
2607 	struct timeval tv[2];
2608 	struct vattr vattr;
2609 	int error;
2610 
2611 	VATTR_NULL(&vattr);
2612 	if (tptr == NULL) {
2613 		microtime(&tv[0]);
2614 		tv[1] = tv[0];
2615 		vattr.va_vaflags |= VA_UTIMES_NULL;
2616 	} else {
2617 		error = copyin(tptr, tv, sizeof(tv));
2618 		if (error)
2619 			return (error);
2620 	}
2621 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2622 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2623 	vattr.va_atime.tv_sec = tv[0].tv_sec;
2624 	vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2625 	vattr.va_mtime.tv_sec = tv[1].tv_sec;
2626 	vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2627 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2628 	VOP_UNLOCK(vp, 0);
2629 	return (error);
2630 }
2631 
2632 /*
2633  * Truncate a file given its path name.
2634  */
2635 /* ARGSUSED */
2636 int
2637 sys_truncate(p, v, retval)
2638 	struct proc *p;
2639 	void *v;
2640 	register_t *retval;
2641 {
2642 	struct sys_truncate_args /* {
2643 		syscallarg(const char *) path;
2644 		syscallarg(int) pad;
2645 		syscallarg(off_t) length;
2646 	} */ *uap = v;
2647 	struct vnode *vp;
2648 	struct vattr vattr;
2649 	int error;
2650 	struct nameidata nd;
2651 
2652 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2653 	if ((error = namei(&nd)) != 0)
2654 		return (error);
2655 	vp = nd.ni_vp;
2656 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2657 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2658 	if (vp->v_type == VDIR)
2659 		error = EISDIR;
2660 	else if ((error = vn_writechk(vp)) == 0 &&
2661 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2662 		VATTR_NULL(&vattr);
2663 		vattr.va_size = SCARG(uap, length);
2664 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2665 	}
2666 	vput(vp);
2667 	return (error);
2668 }
2669 
2670 /*
2671  * Truncate a file given a file descriptor.
2672  */
2673 /* ARGSUSED */
2674 int
2675 sys_ftruncate(p, v, retval)
2676 	struct proc *p;
2677 	void *v;
2678 	register_t *retval;
2679 {
2680 	struct sys_ftruncate_args /* {
2681 		syscallarg(int) fd;
2682 		syscallarg(int) pad;
2683 		syscallarg(off_t) length;
2684 	} */ *uap = v;
2685 	struct vattr vattr;
2686 	struct vnode *vp;
2687 	struct file *fp;
2688 	int error;
2689 
2690 	/* getvnode() will use the descriptor for us */
2691 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2692 		return (error);
2693 	if ((fp->f_flag & FWRITE) == 0) {
2694 		error = EINVAL;
2695 		goto out;
2696 	}
2697 	vp = (struct vnode *)fp->f_data;
2698 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2699 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2700 	if (vp->v_type == VDIR)
2701 		error = EISDIR;
2702 	else if ((error = vn_writechk(vp)) == 0) {
2703 		VATTR_NULL(&vattr);
2704 		vattr.va_size = SCARG(uap, length);
2705 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2706 	}
2707 	VOP_UNLOCK(vp, 0);
2708  out:
2709 	FILE_UNUSE(fp, p);
2710 	return (error);
2711 }
2712 
2713 /*
2714  * Sync an open file.
2715  */
2716 /* ARGSUSED */
2717 int
2718 sys_fsync(p, v, retval)
2719 	struct proc *p;
2720 	void *v;
2721 	register_t *retval;
2722 {
2723 	struct sys_fsync_args /* {
2724 		syscallarg(int) fd;
2725 	} */ *uap = v;
2726 	struct vnode *vp;
2727 	struct file *fp;
2728 	int error;
2729 
2730 	/* getvnode() will use the descriptor for us */
2731 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2732 		return (error);
2733 	vp = (struct vnode *)fp->f_data;
2734 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2735 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2736 	if (error == 0 && bioops.io_fsync != NULL &&
2737 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2738 		(*bioops.io_fsync)(vp);
2739 	VOP_UNLOCK(vp, 0);
2740 	FILE_UNUSE(fp, p);
2741 	return (error);
2742 }
2743 
2744 /*
2745  * Sync the data of an open file.
2746  */
2747 /* ARGSUSED */
2748 int
2749 sys_fdatasync(p, v, retval)
2750 	struct proc *p;
2751 	void *v;
2752 	register_t *retval;
2753 {
2754 	struct sys_fdatasync_args /* {
2755 		syscallarg(int) fd;
2756 	} */ *uap = v;
2757 	struct vnode *vp;
2758 	struct file *fp;
2759 	int error;
2760 
2761 	/* getvnode() will use the descriptor for us */
2762 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2763 		return (error);
2764 	vp = (struct vnode *)fp->f_data;
2765 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2766 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2767 	VOP_UNLOCK(vp, 0);
2768 	FILE_UNUSE(fp, p);
2769 	return (error);
2770 }
2771 
2772 /*
2773  * Rename files, (standard) BSD semantics frontend.
2774  */
2775 /* ARGSUSED */
2776 int
2777 sys_rename(p, v, retval)
2778 	struct proc *p;
2779 	void *v;
2780 	register_t *retval;
2781 {
2782 	struct sys_rename_args /* {
2783 		syscallarg(const char *) from;
2784 		syscallarg(const char *) to;
2785 	} */ *uap = v;
2786 
2787 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2788 }
2789 
2790 /*
2791  * Rename files, POSIX semantics frontend.
2792  */
2793 /* ARGSUSED */
2794 int
2795 sys___posix_rename(p, v, retval)
2796 	struct proc *p;
2797 	void *v;
2798 	register_t *retval;
2799 {
2800 	struct sys___posix_rename_args /* {
2801 		syscallarg(const char *) from;
2802 		syscallarg(const char *) to;
2803 	} */ *uap = v;
2804 
2805 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2806 }
2807 
2808 /*
2809  * Rename files.  Source and destination must either both be directories,
2810  * or both not be directories.  If target is a directory, it must be empty.
2811  * If `from' and `to' refer to the same object, the value of the `retain'
2812  * argument is used to determine whether `from' will be
2813  *
2814  * (retain == 0)	deleted unless `from' and `to' refer to the same
2815  *			object in the file system's name space (BSD).
2816  * (retain == 1)	always retained (POSIX).
2817  */
2818 static int
2819 rename_files(from, to, p, retain)
2820 	const char *from, *to;
2821 	struct proc *p;
2822 	int retain;
2823 {
2824 	struct vnode *tvp, *fvp, *tdvp;
2825 	struct nameidata fromnd, tond;
2826 	int error;
2827 
2828 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2829 	    from, p);
2830 	if ((error = namei(&fromnd)) != 0)
2831 		return (error);
2832 	fvp = fromnd.ni_vp;
2833 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2834 	    UIO_USERSPACE, to, p);
2835 	if ((error = namei(&tond)) != 0) {
2836 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2837 		vrele(fromnd.ni_dvp);
2838 		vrele(fvp);
2839 		goto out1;
2840 	}
2841 	tdvp = tond.ni_dvp;
2842 	tvp = tond.ni_vp;
2843 
2844 	if (tvp != NULL) {
2845 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2846 			error = ENOTDIR;
2847 			goto out;
2848 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2849 			error = EISDIR;
2850 			goto out;
2851 		}
2852 	}
2853 
2854 	if (fvp == tdvp)
2855 		error = EINVAL;
2856 
2857 	/*
2858 	 * Source and destination refer to the same object.
2859 	 */
2860 	if (fvp == tvp) {
2861 		if (retain)
2862 			error = -1;
2863 		else if (fromnd.ni_dvp == tdvp &&
2864 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2865 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
2866 		          tond.ni_cnd.cn_nameptr,
2867 		          fromnd.ni_cnd.cn_namelen))
2868 		error = -1;
2869 	}
2870 
2871 out:
2872 	if (!error) {
2873 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2874 		if (fromnd.ni_dvp != tdvp)
2875 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2876 		if (tvp) {
2877 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2878 		}
2879 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2880 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2881 	} else {
2882 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2883 		if (tdvp == tvp)
2884 			vrele(tdvp);
2885 		else
2886 			vput(tdvp);
2887 		if (tvp)
2888 			vput(tvp);
2889 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2890 		vrele(fromnd.ni_dvp);
2891 		vrele(fvp);
2892 	}
2893 	vrele(tond.ni_startdir);
2894 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2895 out1:
2896 	if (fromnd.ni_startdir)
2897 		vrele(fromnd.ni_startdir);
2898 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2899 	return (error == -1 ? 0 : error);
2900 }
2901 
2902 /*
2903  * Make a directory file.
2904  */
2905 /* ARGSUSED */
2906 int
2907 sys_mkdir(p, v, retval)
2908 	struct proc *p;
2909 	void *v;
2910 	register_t *retval;
2911 {
2912 	struct sys_mkdir_args /* {
2913 		syscallarg(const char *) path;
2914 		syscallarg(int) mode;
2915 	} */ *uap = v;
2916 	struct vnode *vp;
2917 	struct vattr vattr;
2918 	int error;
2919 	struct nameidata nd;
2920 
2921 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2922 	if ((error = namei(&nd)) != 0)
2923 		return (error);
2924 	vp = nd.ni_vp;
2925 	if (vp != NULL) {
2926 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2927 		if (nd.ni_dvp == vp)
2928 			vrele(nd.ni_dvp);
2929 		else
2930 			vput(nd.ni_dvp);
2931 		vrele(vp);
2932 		return (EEXIST);
2933 	}
2934 	VATTR_NULL(&vattr);
2935 	vattr.va_type = VDIR;
2936 	vattr.va_mode =
2937 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2938 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2939 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2940 	if (!error)
2941 		vput(nd.ni_vp);
2942 	return (error);
2943 }
2944 
2945 /*
2946  * Remove a directory file.
2947  */
2948 /* ARGSUSED */
2949 int
2950 sys_rmdir(p, v, retval)
2951 	struct proc *p;
2952 	void *v;
2953 	register_t *retval;
2954 {
2955 	struct sys_rmdir_args /* {
2956 		syscallarg(const char *) path;
2957 	} */ *uap = v;
2958 	struct vnode *vp;
2959 	int error;
2960 	struct nameidata nd;
2961 
2962 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2963 	    SCARG(uap, path), p);
2964 	if ((error = namei(&nd)) != 0)
2965 		return (error);
2966 	vp = nd.ni_vp;
2967 	if (vp->v_type != VDIR) {
2968 		error = ENOTDIR;
2969 		goto out;
2970 	}
2971 	/*
2972 	 * No rmdir "." please.
2973 	 */
2974 	if (nd.ni_dvp == vp) {
2975 		error = EINVAL;
2976 		goto out;
2977 	}
2978 	/*
2979 	 * The root of a mounted filesystem cannot be deleted.
2980 	 */
2981 	if (vp->v_flag & VROOT)
2982 		error = EBUSY;
2983 out:
2984 	if (!error) {
2985 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2986 		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2987 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2988 	} else {
2989 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2990 		if (nd.ni_dvp == vp)
2991 			vrele(nd.ni_dvp);
2992 		else
2993 			vput(nd.ni_dvp);
2994 		vput(vp);
2995 	}
2996 	return (error);
2997 }
2998 
2999 /*
3000  * Read a block of directory entries in a file system independent format.
3001  */
3002 int
3003 sys_getdents(p, v, retval)
3004 	struct proc *p;
3005 	void *v;
3006 	register_t *retval;
3007 {
3008 	struct sys_getdents_args /* {
3009 		syscallarg(int) fd;
3010 		syscallarg(char *) buf;
3011 		syscallarg(size_t) count;
3012 	} */ *uap = v;
3013 	struct file *fp;
3014 	int error, done;
3015 
3016 	/* getvnode() will use the descriptor for us */
3017 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3018 		return (error);
3019 	if ((fp->f_flag & FREAD) == 0) {
3020 		error = EBADF;
3021 		goto out;
3022 	}
3023 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3024 			SCARG(uap, count), &done, p, 0, 0);
3025 	*retval = done;
3026  out:
3027 	FILE_UNUSE(fp, p);
3028 	return (error);
3029 }
3030 
3031 /*
3032  * Set the mode mask for creation of filesystem nodes.
3033  */
3034 int
3035 sys_umask(p, v, retval)
3036 	struct proc *p;
3037 	void *v;
3038 	register_t *retval;
3039 {
3040 	struct sys_umask_args /* {
3041 		syscallarg(mode_t) newmask;
3042 	} */ *uap = v;
3043 	struct cwdinfo *cwdi;
3044 
3045 	cwdi = p->p_cwdi;
3046 	*retval = cwdi->cwdi_cmask;
3047 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3048 	return (0);
3049 }
3050 
3051 /*
3052  * Void all references to file by ripping underlying filesystem
3053  * away from vnode.
3054  */
3055 /* ARGSUSED */
3056 int
3057 sys_revoke(p, v, retval)
3058 	struct proc *p;
3059 	void *v;
3060 	register_t *retval;
3061 {
3062 	struct sys_revoke_args /* {
3063 		syscallarg(const char *) path;
3064 	} */ *uap = v;
3065 	struct vnode *vp;
3066 	struct vattr vattr;
3067 	int error;
3068 	struct nameidata nd;
3069 
3070 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3071 	if ((error = namei(&nd)) != 0)
3072 		return (error);
3073 	vp = nd.ni_vp;
3074 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3075 		goto out;
3076 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3077 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3078 		goto out;
3079 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3080 		VOP_REVOKE(vp, REVOKEALL);
3081 out:
3082 	vrele(vp);
3083 	return (error);
3084 }
3085 
3086 /*
3087  * Convert a user file descriptor to a kernel file entry.
3088  */
3089 int
3090 getvnode(fdp, fd, fpp)
3091 	struct filedesc *fdp;
3092 	int fd;
3093 	struct file **fpp;
3094 {
3095 	struct vnode *vp;
3096 	struct file *fp;
3097 
3098 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3099 		return (EBADF);
3100 
3101 	FILE_USE(fp);
3102 
3103 	if (fp->f_type != DTYPE_VNODE) {
3104 		FILE_UNUSE(fp, NULL);
3105 		return (EINVAL);
3106 	}
3107 
3108 	vp = (struct vnode *)fp->f_data;
3109 	if (vp->v_type == VBAD) {
3110 		FILE_UNUSE(fp, NULL);
3111 		return (EBADF);
3112 	}
3113 
3114 	*fpp = fp;
3115 	return (0);
3116 }
3117