xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 481fca6e59249d8ffcf24fef7cfbe7b131bfb080)
1 /*	$NetBSD: vfs_syscalls.c,v 1.160 2000/07/09 00:59:03 mycroft 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 "opt_compat_netbsd.h"
44 #include "opt_compat_43.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/namei.h>
49 #include <sys/filedesc.h>
50 #include <sys/kernel.h>
51 #include <sys/file.h>
52 #include <sys/stat.h>
53 #include <sys/vnode.h>
54 #include <sys/mount.h>
55 #include <sys/proc.h>
56 #include <sys/uio.h>
57 #include <sys/malloc.h>
58 #include <sys/dirent.h>
59 #include <sys/syscallargs.h>
60 
61 #include <miscfs/genfs/genfs.h>
62 #include <miscfs/syncfs/syncfs.h>
63 
64 #include <uvm/uvm_extern.h>
65 
66 #include <sys/sysctl.h>
67 
68 static int change_dir __P((struct nameidata *, struct proc *));
69 static int change_mode __P((struct vnode *, int, struct proc *p));
70 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *,
71     int));
72 static int change_utimes __P((struct vnode *vp, const struct timeval *,
73 	       struct proc *p));
74 static int rename_files __P((const char *, const char *, struct proc *, int));
75 
76 void checkdirs __P((struct vnode *));
77 
78 int dovfsusermount = 0;
79 
80 /*
81  * Virtual File System System Calls
82  */
83 
84 /*
85  * Mount a file system.
86  */
87 
88 /*
89  * This table is used to maintain compatibility with 4.3BSD
90  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
91  *
92  * Also note that not all of these had actual numbers in 4.3BSD
93  * or NetBSD 0.9!
94  */
95 const char *mountcompatnames[] = {
96 	NULL,		/* 0 = MOUNT_NONE */
97 	MOUNT_FFS,	/* 1 */
98 	MOUNT_NFS,	/* 2 */
99 	MOUNT_MFS,	/* 3 */
100 	MOUNT_MSDOS,	/* 4 */
101 	MOUNT_LFS,	/* 5 */
102 	NULL,		/* 6 = MOUNT_LOFS */
103 	MOUNT_FDESC,	/* 7 */
104 	MOUNT_PORTAL,	/* 8 */
105 	MOUNT_NULL,	/* 9 */
106 	MOUNT_UMAP,	/* 10 */
107 	MOUNT_KERNFS,	/* 11 */
108 	MOUNT_PROCFS,	/* 12 */
109 	MOUNT_AFS,	/* 13 */
110 	MOUNT_CD9660,	/* 14 = MOUNT_ISOFS */
111 	MOUNT_UNION,	/* 15 */
112 	MOUNT_ADOSFS,	/* 16 */
113 	MOUNT_EXT2FS,	/* 17 */
114 	MOUNT_CODA,	/* 18 */
115 	MOUNT_FILECORE,	/* 19 */
116 	MOUNT_NTFS,	/* 20 */
117 };
118 const int nmountcompatnames = sizeof(mountcompatnames) /
119     sizeof(mountcompatnames[0]);
120 
121 /* ARGSUSED */
122 int
123 sys_mount(p, v, retval)
124 	struct proc *p;
125 	void *v;
126 	register_t *retval;
127 {
128 	struct sys_mount_args /* {
129 		syscallarg(const char *) type;
130 		syscallarg(const char *) path;
131 		syscallarg(int) flags;
132 		syscallarg(void *) data;
133 	} */ *uap = v;
134 	struct vnode *vp;
135 	struct mount *mp;
136 	int error, flag = 0;
137 	char fstypename[MFSNAMELEN];
138 	struct vattr va;
139 	struct nameidata nd;
140 	struct vfsops *vfs;
141 
142 	if (dovfsusermount == 0 && (error = suser(p->p_ucred, &p->p_acflag)))
143 		return (error);
144 	/*
145 	 * Get vnode to be covered
146 	 */
147 	NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
148 	    SCARG(uap, path), p);
149 	if ((error = namei(&nd)) != 0)
150 		return (error);
151 	vp = nd.ni_vp;
152 	/*
153 	 * A lookup in VFS_MOUNT might result in an attempt to
154 	 * lock this vnode again, so make the lock resursive.
155 	 */
156 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
157 	if (SCARG(uap, flags) & MNT_UPDATE) {
158 		if ((vp->v_flag & VROOT) == 0) {
159 			vput(vp);
160 			return (EINVAL);
161 		}
162 		mp = vp->v_mount;
163 		flag = mp->mnt_flag;
164 		vfs = mp->mnt_op;
165 		/*
166 		 * We only allow the filesystem to be reloaded if it
167 		 * is currently mounted read-only.
168 		 */
169 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
170 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
171 			vput(vp);
172 			return (EOPNOTSUPP);	/* Needs translation */
173 		}
174 		/*
175 		 * In "highly secure" mode, don't let the caller do anything
176 		 * but downgrade a filesystem from read-write to read-only.
177 		 * (see also below; MNT_UPDATE is required.)
178 		 */
179 		if (securelevel >= 2 &&
180 		    (SCARG(uap, flags) !=
181 		    (mp->mnt_flag | MNT_RDONLY |
182 		    MNT_RELOAD | MNT_FORCE | MNT_UPDATE))) {
183 			vput(vp);
184 			return (EPERM);
185 		}
186 		mp->mnt_flag |=
187 		    SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
188 		/*
189 		 * Only root, or the user that did the original mount is
190 		 * permitted to update it.
191 		 */
192 		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
193 		    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
194 			vput(vp);
195 			return (error);
196 		}
197 		/*
198 		 * Do not allow NFS export by non-root users. For non-root
199 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
200 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
201 		 */
202 		if (p->p_ucred->cr_uid != 0) {
203 			if (SCARG(uap, flags) & MNT_EXPORTED) {
204 				vput(vp);
205 				return (EPERM);
206 			}
207 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
208 			if (flag & MNT_NOEXEC)
209 				SCARG(uap, flags) |= MNT_NOEXEC;
210 		}
211 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
212 			vput(vp);
213 			return (EPERM);
214 		}
215 		VOP_UNLOCK(vp, 0);
216 		goto update;
217 	} else {
218 		if (securelevel >= 2) {
219 			vput(vp);
220 			return (EPERM);
221 		}
222 	}
223 	/*
224 	 * If the user is not root, ensure that they own the directory
225 	 * onto which we are attempting to mount.
226 	 */
227 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
228 	    (va.va_uid != p->p_ucred->cr_uid &&
229 		(error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
230 		vput(vp);
231 		return (error);
232 	}
233 	/*
234 	 * Do not allow NFS export by non-root users. For non-root users,
235 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
236 	 * mount point is already MNT_NOEXEC.
237 	 */
238 	if (p->p_ucred->cr_uid != 0) {
239 		if (SCARG(uap, flags) & MNT_EXPORTED) {
240 			vput(vp);
241 			return (EPERM);
242 		}
243 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
244 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
245 			SCARG(uap, flags) |= MNT_NOEXEC;
246 	}
247 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
248 		return (error);
249 	if (vp->v_type != VDIR) {
250 		vput(vp);
251 		return (ENOTDIR);
252 	}
253 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
254 	if (error) {
255 #if defined(COMPAT_09) || defined(COMPAT_43)
256 		/*
257 		 * Historically filesystem types were identified by number.
258 		 * If we get an integer for the filesystem type instead of a
259 		 * string, we check to see if it matches one of the historic
260 		 * filesystem types.
261 		 */
262 		u_long fsindex = (u_long)SCARG(uap, type);
263 		if (fsindex >= nmountcompatnames ||
264 		    mountcompatnames[fsindex] == NULL) {
265 			vput(vp);
266 			return (ENODEV);
267 		}
268 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
269 #else
270 		vput(vp);
271 		return (error);
272 #endif
273 	}
274 #ifdef	COMPAT_10
275 	/* Accept `ufs' as an alias for `ffs'. */
276 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
277 		strncpy(fstypename, "ffs", MFSNAMELEN);
278 #endif
279 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
280 		vput(vp);
281 		return (ENODEV);
282 	}
283 	if (vp->v_mountedhere != NULL) {
284 		vput(vp);
285 		return (EBUSY);
286 	}
287 
288 	/*
289 	 * Allocate and initialize the file system.
290 	 */
291 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
292 		M_MOUNT, M_WAITOK);
293 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
294 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
295 	(void)vfs_busy(mp, LK_NOWAIT, 0);
296 	mp->mnt_op = vfs;
297 	vfs->vfs_refcount++;
298 	mp->mnt_vnodecovered = vp;
299 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
300 	mp->mnt_unmounter = NULL;
301 update:
302 	/*
303 	 * Set the mount level flags.
304 	 */
305 	if (SCARG(uap, flags) & MNT_RDONLY)
306 		mp->mnt_flag |= MNT_RDONLY;
307 	else if (mp->mnt_flag & MNT_RDONLY)
308 		mp->mnt_flag |= MNT_WANTRDWR;
309 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
310 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
311 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
312 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
313 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
314 	    MNT_NOCOREDUMP | MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM |
315 	    MNT_SOFTDEP);
316 	/*
317 	 * Mount the filesystem.
318 	 */
319 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
320 	if (mp->mnt_flag & MNT_UPDATE) {
321 		vrele(vp);
322 		if (mp->mnt_flag & MNT_WANTRDWR)
323 			mp->mnt_flag &= ~MNT_RDONLY;
324 		mp->mnt_flag &=~
325 		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
326 		if (error)
327 			mp->mnt_flag = flag;
328 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
329 			if (mp->mnt_syncer == NULL)
330 				error = vfs_allocate_syncvnode(mp);
331 		} else {
332 			if (mp->mnt_syncer != NULL)
333 				vfs_deallocate_syncvnode(mp);
334 		}
335 		vfs_unbusy(mp);
336 		return (error);
337 	}
338 	/*
339 	 * Put the new filesystem on the mount list after root.
340 	 */
341 	cache_purge(vp);
342 	if (!error) {
343 		vp->v_mountedhere = mp;
344 		simple_lock(&mountlist_slock);
345 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
346 		simple_unlock(&mountlist_slock);
347 		checkdirs(vp);
348 		VOP_UNLOCK(vp, 0);
349 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
350 			error = vfs_allocate_syncvnode(mp);
351 		vfs_unbusy(mp);
352 		(void) VFS_STATFS(mp, &mp->mnt_stat, p);
353 		if ((error = VFS_START(mp, 0, p)))
354 			vrele(vp);
355 	} else {
356 		vp->v_mountedhere = (struct mount *)0;
357 		vfs->vfs_refcount--;
358 		vfs_unbusy(mp);
359 		free((caddr_t)mp, M_MOUNT);
360 		vput(vp);
361 	}
362 	return (error);
363 }
364 
365 /*
366  * Scan all active processes to see if any of them have a current
367  * or root directory onto which the new filesystem has just been
368  * mounted. If so, replace them with the new mount point.
369  */
370 void
371 checkdirs(olddp)
372 	struct vnode *olddp;
373 {
374 	struct cwdinfo *cwdi;
375 	struct vnode *newdp;
376 	struct proc *p;
377 
378 	if (olddp->v_usecount == 1)
379 		return;
380 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
381 		panic("mount: lost mount");
382 	proclist_lock_read();
383 	for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
384 		cwdi = p->p_cwdi;
385 		if (cwdi->cwdi_cdir == olddp) {
386 			vrele(cwdi->cwdi_cdir);
387 			VREF(newdp);
388 			cwdi->cwdi_cdir = newdp;
389 		}
390 		if (cwdi->cwdi_rdir == olddp) {
391 			vrele(cwdi->cwdi_rdir);
392 			VREF(newdp);
393 			cwdi->cwdi_rdir = newdp;
394 		}
395 	}
396 	proclist_unlock_read();
397 	if (rootvnode == olddp) {
398 		vrele(rootvnode);
399 		VREF(newdp);
400 		rootvnode = newdp;
401 	}
402 	vput(newdp);
403 }
404 
405 /*
406  * Unmount a file system.
407  *
408  * Note: unmount takes a path to the vnode mounted on as argument,
409  * not special file (as before).
410  */
411 /* ARGSUSED */
412 int
413 sys_unmount(p, v, retval)
414 	struct proc *p;
415 	void *v;
416 	register_t *retval;
417 {
418 	struct sys_unmount_args /* {
419 		syscallarg(const char *) path;
420 		syscallarg(int) flags;
421 	} */ *uap = v;
422 	struct vnode *vp;
423 	struct mount *mp;
424 	int error;
425 	struct nameidata nd;
426 
427 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
428 	    SCARG(uap, path), p);
429 	if ((error = namei(&nd)) != 0)
430 		return (error);
431 	vp = nd.ni_vp;
432 	mp = vp->v_mount;
433 
434 	/*
435 	 * Only root, or the user that did the original mount is
436 	 * permitted to unmount this filesystem.
437 	 */
438 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
439 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
440 		vput(vp);
441 		return (error);
442 	}
443 
444 	/*
445 	 * Don't allow unmounting the root file system.
446 	 */
447 	if (mp->mnt_flag & MNT_ROOTFS) {
448 		vput(vp);
449 		return (EINVAL);
450 	}
451 
452 	/*
453 	 * Must be the root of the filesystem
454 	 */
455 	if ((vp->v_flag & VROOT) == 0) {
456 		vput(vp);
457 		return (EINVAL);
458 	}
459 	vput(vp);
460 
461 	if (vfs_busy(mp, 0, 0))
462 		return (EBUSY);
463 
464 	return (dounmount(mp, SCARG(uap, flags), p));
465 }
466 
467 /*
468  * Do the actual file system unmount. File system is assumed to have been
469  * marked busy by the caller.
470  */
471 int
472 dounmount(mp, flags, p)
473 	struct mount *mp;
474 	int flags;
475 	struct proc *p;
476 {
477 	struct vnode *coveredvp;
478 	int error;
479 	int async;
480 
481 	simple_lock(&mountlist_slock);
482 	vfs_unbusy(mp);
483 	/*
484 	 * XXX Freeze syncer. This should really be done on a mountpoint
485 	 * basis, but especially the softdep code possibly called from
486 	 * the syncer doesn't exactly work on a per-mountpoint basis,
487 	 * so the softdep code would become a maze of vfs_busy calls.
488 	 */
489 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
490 
491 	mp->mnt_flag |= MNT_UNMOUNT;
492 	mp->mnt_unmounter = p;
493 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
494 	if (mp->mnt_flag & MNT_EXPUBLIC)
495 		vfs_setpublicfs(NULL, NULL, NULL);
496 	async = mp->mnt_flag & MNT_ASYNC;
497 	mp->mnt_flag &= ~MNT_ASYNC;
498 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
499 	if (mp->mnt_syncer != NULL)
500 		vfs_deallocate_syncvnode(mp);
501 	if (((mp->mnt_flag & MNT_RDONLY) ||
502 	    (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
503 	    (flags & MNT_FORCE))
504 		error = VFS_UNMOUNT(mp, flags, p);
505 	simple_lock(&mountlist_slock);
506 	if (error) {
507 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
508 			(void) vfs_allocate_syncvnode(mp);
509 		mp->mnt_flag &= ~MNT_UNMOUNT;
510 		mp->mnt_unmounter = NULL;
511 		mp->mnt_flag |= async;
512 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
513 		    &mountlist_slock);
514 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
515 		while (mp->mnt_wcnt > 0) {
516 			wakeup((caddr_t)mp);
517 			tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
518 		}
519 		return (error);
520 	}
521 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
522 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
523 		coveredvp->v_mountedhere = NULL;
524 		vrele(coveredvp);
525 	}
526 	mp->mnt_op->vfs_refcount--;
527 	if (mp->mnt_vnodelist.lh_first != NULL)
528 		panic("unmount: dangling vnode");
529 	mp->mnt_flag |= MNT_GONE;
530 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
531 	lockmgr(&syncer_lock, LK_RELEASE, NULL);
532 	while(mp->mnt_wcnt > 0) {
533 		wakeup((caddr_t)mp);
534 		tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
535 	}
536 	free((caddr_t)mp, M_MOUNT);
537 	return (0);
538 }
539 
540 /*
541  * Sync each mounted filesystem.
542  */
543 #ifdef DEBUG
544 int syncprt = 0;
545 struct ctldebug debug0 = { "syncprt", &syncprt };
546 #endif
547 
548 /* ARGSUSED */
549 int
550 sys_sync(p, v, retval)
551 	struct proc *p;
552 	void *v;
553 	register_t *retval;
554 {
555 	struct mount *mp, *nmp;
556 	int asyncflag;
557 
558 	simple_lock(&mountlist_slock);
559 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
560 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
561 			nmp = mp->mnt_list.cqe_prev;
562 			continue;
563 		}
564 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
565 			asyncflag = mp->mnt_flag & MNT_ASYNC;
566 			mp->mnt_flag &= ~MNT_ASYNC;
567 			uvm_vnp_sync(mp);
568 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
569 			if (asyncflag)
570 				 mp->mnt_flag |= MNT_ASYNC;
571 		}
572 		simple_lock(&mountlist_slock);
573 		nmp = mp->mnt_list.cqe_prev;
574 		vfs_unbusy(mp);
575 
576 	}
577 	simple_unlock(&mountlist_slock);
578 #ifdef DEBUG
579 	if (syncprt)
580 		vfs_bufstats();
581 #endif /* DEBUG */
582 	return (0);
583 }
584 
585 /*
586  * Change filesystem quotas.
587  */
588 /* ARGSUSED */
589 int
590 sys_quotactl(p, v, retval)
591 	struct proc *p;
592 	void *v;
593 	register_t *retval;
594 {
595 	struct sys_quotactl_args /* {
596 		syscallarg(const char *) path;
597 		syscallarg(int) cmd;
598 		syscallarg(int) uid;
599 		syscallarg(caddr_t) arg;
600 	} */ *uap = v;
601 	struct mount *mp;
602 	int error;
603 	struct nameidata nd;
604 
605 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
606 	if ((error = namei(&nd)) != 0)
607 		return (error);
608 	mp = nd.ni_vp->v_mount;
609 	vrele(nd.ni_vp);
610 	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
611 	    SCARG(uap, arg), p));
612 }
613 
614 /*
615  * Get filesystem statistics.
616  */
617 /* ARGSUSED */
618 int
619 sys_statfs(p, v, retval)
620 	struct proc *p;
621 	void *v;
622 	register_t *retval;
623 {
624 	struct sys_statfs_args /* {
625 		syscallarg(const char *) path;
626 		syscallarg(struct statfs *) buf;
627 	} */ *uap = v;
628 	struct mount *mp;
629 	struct statfs *sp;
630 	int error;
631 	struct nameidata nd;
632 
633 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
634 	if ((error = namei(&nd)) != 0)
635 		return (error);
636 	mp = nd.ni_vp->v_mount;
637 	sp = &mp->mnt_stat;
638 	vrele(nd.ni_vp);
639 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
640 		return (error);
641 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
642 	sp->f_oflags = sp->f_flags & 0xffff;
643 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
644 }
645 
646 /*
647  * Get filesystem statistics.
648  */
649 /* ARGSUSED */
650 int
651 sys_fstatfs(p, v, retval)
652 	struct proc *p;
653 	void *v;
654 	register_t *retval;
655 {
656 	struct sys_fstatfs_args /* {
657 		syscallarg(int) fd;
658 		syscallarg(struct statfs *) buf;
659 	} */ *uap = v;
660 	struct file *fp;
661 	struct mount *mp;
662 	struct statfs *sp;
663 	int error;
664 
665 	/* getvnode() will use the descriptor for us */
666 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
667 		return (error);
668 	mp = ((struct vnode *)fp->f_data)->v_mount;
669 	sp = &mp->mnt_stat;
670 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
671 		goto out;
672 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
673 	sp->f_oflags = sp->f_flags & 0xffff;
674 	error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
675  out:
676 	FILE_UNUSE(fp, p);
677 	return (error);
678 }
679 
680 /*
681  * Get statistics on all filesystems.
682  */
683 int
684 sys_getfsstat(p, v, retval)
685 	struct proc *p;
686 	void *v;
687 	register_t *retval;
688 {
689 	struct sys_getfsstat_args /* {
690 		syscallarg(struct statfs *) buf;
691 		syscallarg(long) bufsize;
692 		syscallarg(int) flags;
693 	} */ *uap = v;
694 	struct mount *mp, *nmp;
695 	struct statfs *sp;
696 	caddr_t sfsp;
697 	long count, maxcount, error;
698 
699 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
700 	sfsp = (caddr_t)SCARG(uap, buf);
701 	simple_lock(&mountlist_slock);
702 	count = 0;
703 	for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
704 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
705 			nmp = mp->mnt_list.cqe_next;
706 			continue;
707 		}
708 		if (sfsp && count < maxcount) {
709 			sp = &mp->mnt_stat;
710 			/*
711 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
712 			 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
713 			 * overrides MNT_NOWAIT.
714 			 */
715 			if (SCARG(uap, flags) != MNT_NOWAIT &&
716 			    SCARG(uap, flags) != MNT_LAZY &&
717 			    (SCARG(uap, flags) == MNT_WAIT ||
718 			     SCARG(uap, flags) == 0) &&
719 			    (error = VFS_STATFS(mp, sp, p)) != 0) {
720 				simple_lock(&mountlist_slock);
721 				nmp = mp->mnt_list.cqe_next;
722 				vfs_unbusy(mp);
723 				continue;
724 			}
725 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
726 			sp->f_oflags = sp->f_flags & 0xffff;
727 			error = copyout(sp, sfsp, sizeof(*sp));
728 			if (error) {
729 				vfs_unbusy(mp);
730 				return (error);
731 			}
732 			sfsp += sizeof(*sp);
733 		}
734 		count++;
735 		simple_lock(&mountlist_slock);
736 		nmp = mp->mnt_list.cqe_next;
737 		vfs_unbusy(mp);
738 	}
739 	simple_unlock(&mountlist_slock);
740 	if (sfsp && count > maxcount)
741 		*retval = maxcount;
742 	else
743 		*retval = count;
744 	return (0);
745 }
746 
747 /*
748  * Change current working directory to a given file descriptor.
749  */
750 /* ARGSUSED */
751 int
752 sys_fchdir(p, v, retval)
753 	struct proc *p;
754 	void *v;
755 	register_t *retval;
756 {
757 	struct sys_fchdir_args /* {
758 		syscallarg(int) fd;
759 	} */ *uap = v;
760 	struct filedesc *fdp = p->p_fd;
761 	struct cwdinfo *cwdi = p->p_cwdi;
762 	struct vnode *vp, *tdp;
763 	struct mount *mp;
764 	struct file *fp;
765 	int error;
766 
767 	/* getvnode() will use the descriptor for us */
768 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
769 		return (error);
770 	vp = (struct vnode *)fp->f_data;
771 
772 	VREF(vp);
773 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
774 	if (vp->v_type != VDIR)
775 		error = ENOTDIR;
776 	else
777 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
778 	while (!error && (mp = vp->v_mountedhere) != NULL) {
779 		if (vfs_busy(mp, 0, 0))
780 			continue;
781 		error = VFS_ROOT(mp, &tdp);
782 		vfs_unbusy(mp);
783 		if (error)
784 			break;
785 		vput(vp);
786 		vp = tdp;
787 	}
788 	if (error) {
789 		vput(vp);
790 		goto out;
791 	}
792 	VOP_UNLOCK(vp, 0);
793 
794 	/*
795 	 * Disallow changing to a directory not under the process's
796 	 * current root directory (if there is one).
797 	 */
798 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
799 		vrele(vp);
800 		error = EPERM;	/* operation not permitted */
801 		goto out;
802 	}
803 
804 	vrele(cwdi->cwdi_cdir);
805 	cwdi->cwdi_cdir = vp;
806  out:
807 	FILE_UNUSE(fp, p);
808 	return (error);
809 }
810 
811 /*
812  * Change this process's notion of the root directory to a given file descriptor.
813  */
814 
815 int
816 sys_fchroot(p, v, retval)
817 	struct proc *p;
818 	void *v;
819 	register_t *retval;
820 {
821 	struct sys_fchroot_args *uap = v;
822 	struct filedesc *fdp = p->p_fd;
823 	struct cwdinfo *cwdi = p->p_cwdi;
824 	struct vnode	*vp;
825 	struct file	*fp;
826 	int		 error;
827 
828 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
829 		return error;
830 	/* getvnode() will use the descriptor for us */
831 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
832 		return error;
833 	vp = (struct vnode *) fp->f_data;
834 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
835 	if (vp->v_type != VDIR)
836 		error = ENOTDIR;
837 	else
838 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
839 	VOP_UNLOCK(vp, 0);
840 	if (error)
841 		goto out;
842 	VREF(vp);
843 
844 	/*
845 	 * Prevent escaping from chroot by putting the root under
846 	 * the working directory.  Silently chdir to / if we aren't
847 	 * already there.
848 	 */
849 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
850 		/*
851 		 * XXX would be more failsafe to change directory to a
852 		 * deadfs node here instead
853 		 */
854 		vrele(cwdi->cwdi_cdir);
855 		VREF(vp);
856 		cwdi->cwdi_cdir = vp;
857 	}
858 
859 	if (cwdi->cwdi_rdir != NULL)
860 		vrele(cwdi->cwdi_rdir);
861 	cwdi->cwdi_rdir = vp;
862  out:
863 	FILE_UNUSE(fp, p);
864 	return (error);
865 }
866 
867 
868 
869 /*
870  * Change current working directory (``.'').
871  */
872 /* ARGSUSED */
873 int
874 sys_chdir(p, v, retval)
875 	struct proc *p;
876 	void *v;
877 	register_t *retval;
878 {
879 	struct sys_chdir_args /* {
880 		syscallarg(const char *) path;
881 	} */ *uap = v;
882 	struct cwdinfo *cwdi = p->p_cwdi;
883 	int error;
884 	struct nameidata nd;
885 
886 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
887 	    SCARG(uap, path), p);
888 	if ((error = change_dir(&nd, p)) != 0)
889 		return (error);
890 	vrele(cwdi->cwdi_cdir);
891 	cwdi->cwdi_cdir = nd.ni_vp;
892 	return (0);
893 }
894 
895 /*
896  * Change notion of root (``/'') directory.
897  */
898 /* ARGSUSED */
899 int
900 sys_chroot(p, v, retval)
901 	struct proc *p;
902 	void *v;
903 	register_t *retval;
904 {
905 	struct sys_chroot_args /* {
906 		syscallarg(const char *) path;
907 	} */ *uap = v;
908 	struct cwdinfo *cwdi = p->p_cwdi;
909 	struct vnode *vp;
910 	int error;
911 	struct nameidata nd;
912 
913 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
914 		return (error);
915 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
916 	    SCARG(uap, path), p);
917 	if ((error = change_dir(&nd, p)) != 0)
918 		return (error);
919 	if (cwdi->cwdi_rdir != NULL)
920 		vrele(cwdi->cwdi_rdir);
921 	vp = nd.ni_vp;
922 	cwdi->cwdi_rdir = vp;
923 
924 	/*
925 	 * Prevent escaping from chroot by putting the root under
926 	 * the working directory.  Silently chdir to / if we aren't
927 	 * already there.
928 	 */
929 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
930 		/*
931 		 * XXX would be more failsafe to change directory to a
932 		 * deadfs node here instead
933 		 */
934 		vrele(cwdi->cwdi_cdir);
935 		VREF(vp);
936 		cwdi->cwdi_cdir = vp;
937 	}
938 
939 	return (0);
940 }
941 
942 /*
943  * Common routine for chroot and chdir.
944  */
945 static int
946 change_dir(ndp, p)
947 	struct nameidata *ndp;
948 	struct proc *p;
949 {
950 	struct vnode *vp;
951 	int error;
952 
953 	if ((error = namei(ndp)) != 0)
954 		return (error);
955 	vp = ndp->ni_vp;
956 	if (vp->v_type != VDIR)
957 		error = ENOTDIR;
958 	else
959 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
960 
961 	if (error)
962 		vput(vp);
963 	else
964 		VOP_UNLOCK(vp, 0);
965 	return (error);
966 }
967 
968 /*
969  * Check permissions, allocate an open file structure,
970  * and call the device open routine if any.
971  */
972 int
973 sys_open(p, v, retval)
974 	struct proc *p;
975 	void *v;
976 	register_t *retval;
977 {
978 	struct sys_open_args /* {
979 		syscallarg(const char *) path;
980 		syscallarg(int) flags;
981 		syscallarg(int) mode;
982 	} */ *uap = v;
983 	struct cwdinfo *cwdi = p->p_cwdi;
984 	struct filedesc *fdp = p->p_fd;
985 	struct file *fp;
986 	struct vnode *vp;
987 	int flags, cmode;
988 	int type, indx, error;
989 	struct flock lf;
990 	struct nameidata nd;
991 
992 	flags = FFLAGS(SCARG(uap, flags));
993 	if ((flags & (FREAD | FWRITE)) == 0)
994 		return (EINVAL);
995 	/* falloc() will use the file descriptor for us */
996 	if ((error = falloc(p, &fp, &indx)) != 0)
997 		return (error);
998 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
999 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1000 	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
1001 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1002 		FILE_UNUSE(fp, p);
1003 		ffree(fp);
1004 		if ((error == ENODEV || error == ENXIO) &&
1005 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
1006 		    (error =
1007 			dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1008 			*retval = indx;
1009 			return (0);
1010 		}
1011 		if (error == ERESTART)
1012 			error = EINTR;
1013 		fdremove(fdp, indx);
1014 		return (error);
1015 	}
1016 	p->p_dupfd = 0;
1017 	vp = nd.ni_vp;
1018 	fp->f_flag = flags & FMASK;
1019 	fp->f_type = DTYPE_VNODE;
1020 	fp->f_ops = &vnops;
1021 	fp->f_data = (caddr_t)vp;
1022 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1023 		lf.l_whence = SEEK_SET;
1024 		lf.l_start = 0;
1025 		lf.l_len = 0;
1026 		if (flags & O_EXLOCK)
1027 			lf.l_type = F_WRLCK;
1028 		else
1029 			lf.l_type = F_RDLCK;
1030 		type = F_FLOCK;
1031 		if ((flags & FNONBLOCK) == 0)
1032 			type |= F_WAIT;
1033 		VOP_UNLOCK(vp, 0);
1034 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1035 		if (error) {
1036 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1037 			FILE_UNUSE(fp, p);
1038 			ffree(fp);
1039 			fdremove(fdp, indx);
1040 			return (error);
1041 		}
1042 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1043 		fp->f_flag |= FHASLOCK;
1044 	}
1045 	VOP_UNLOCK(vp, 0);
1046 	*retval = indx;
1047 	FILE_UNUSE(fp, p);
1048 	return (0);
1049 }
1050 
1051 /*
1052  * Get file handle system call
1053  */
1054 int
1055 sys_getfh(p, v, retval)
1056 	struct proc *p;
1057 	void *v;
1058 	register_t *retval;
1059 {
1060 	struct sys_getfh_args /* {
1061 		syscallarg(char *) fname;
1062 		syscallarg(fhandle_t *) fhp;
1063 	} */ *uap = v;
1064 	struct vnode *vp;
1065 	fhandle_t fh;
1066 	int error;
1067 	struct nameidata nd;
1068 
1069 	/*
1070 	 * Must be super user
1071 	 */
1072 	error = suser(p->p_ucred, &p->p_acflag);
1073 	if (error)
1074 		return (error);
1075 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1076 	    SCARG(uap, fname), p);
1077 	error = namei(&nd);
1078 	if (error)
1079 		return (error);
1080 	vp = nd.ni_vp;
1081 	memset((caddr_t)&fh, 0, sizeof(fh));
1082 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1083 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1084 	vput(vp);
1085 	if (error)
1086 		return (error);
1087 	error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1088 	return (error);
1089 }
1090 
1091 /*
1092  * Open a file given a file handle.
1093  *
1094  * Check permissions, allocate an open file structure,
1095  * and call the device open routine if any.
1096  */
1097 int
1098 sys_fhopen(p, v, retval)
1099 	struct proc *p;
1100 	void *v;
1101 	register_t *retval;
1102 {
1103 	struct sys_fhopen_args /* {
1104 		syscallarg(const fhandle_t *) fhp;
1105 		syscallarg(int) flags;
1106 	} */ *uap = v;
1107 	struct filedesc *fdp = p->p_fd;
1108 	struct file *fp;
1109 	struct vnode *vp = NULL;
1110 	struct mount *mp;
1111 	struct ucred *cred = p->p_ucred;
1112 	int flags;
1113 	struct file *nfp;
1114 	int type, indx, error=0;
1115 	struct flock lf;
1116 	struct vattr va;
1117 	fhandle_t fh;
1118 
1119 	/*
1120 	 * Must be super user
1121 	 */
1122 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1123 		return (error);
1124 
1125 	flags = FFLAGS(SCARG(uap, flags));
1126 	if ((flags & (FREAD | FWRITE)) == 0)
1127 		return (EINVAL);
1128 	if ((flags & O_CREAT))
1129 		return (EINVAL);
1130 	/* falloc() will use the file descriptor for us */
1131 	if ((error = falloc(p, &nfp, &indx)) != 0)
1132 		return (error);
1133 	fp = nfp;
1134 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1135 		goto bad;
1136 
1137 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1138 		error = ESTALE;
1139 		goto bad;
1140 	}
1141 
1142 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1143 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1144 		goto bad;
1145 	}
1146 
1147 	/* Now do an effective vn_open */
1148 
1149 	if (vp->v_type == VSOCK) {
1150 		error = EOPNOTSUPP;
1151 		goto bad;
1152 	}
1153 	if (flags & FREAD) {
1154 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1155 			goto bad;
1156 	}
1157 	if (flags & (FWRITE | O_TRUNC)) {
1158 		if (vp->v_type == VDIR) {
1159 			error = EISDIR;
1160 			goto bad;
1161 		}
1162 		if ((error = vn_writechk(vp)) != 0 ||
1163 		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1164 			goto bad;
1165 	}
1166 	if (flags & O_TRUNC) {
1167 		VOP_UNLOCK(vp, 0);			/* XXX */
1168 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1169 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1170 		VATTR_NULL(&va);
1171 		va.va_size = 0;
1172 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1173 			goto bad;
1174 	}
1175 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1176 		goto bad;
1177 	if (flags & FWRITE)
1178 		vp->v_writecount++;
1179 
1180 	/* done with modified vn_open, now finish what sys_open does. */
1181 
1182 	fp->f_flag = flags & FMASK;
1183 	fp->f_type = DTYPE_VNODE;
1184 	fp->f_ops = &vnops;
1185 	fp->f_data = (caddr_t)vp;
1186 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1187 		lf.l_whence = SEEK_SET;
1188 		lf.l_start = 0;
1189 		lf.l_len = 0;
1190 		if (flags & O_EXLOCK)
1191 			lf.l_type = F_WRLCK;
1192 		else
1193 			lf.l_type = F_RDLCK;
1194 		type = F_FLOCK;
1195 		if ((flags & FNONBLOCK) == 0)
1196 			type |= F_WAIT;
1197 		VOP_UNLOCK(vp, 0);
1198 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1199 		if (error) {
1200 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1201 			FILE_UNUSE(fp, p);
1202 			ffree(fp);
1203 			fdremove(fdp, indx);
1204 			return (error);
1205 		}
1206 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1207 		fp->f_flag |= FHASLOCK;
1208 	}
1209 	VOP_UNLOCK(vp, 0);
1210 	*retval = indx;
1211 	FILE_UNUSE(fp, p);
1212 	return (0);
1213 
1214 bad:
1215 	FILE_UNUSE(fp, p);
1216 	ffree(fp);
1217 	fdremove(fdp, indx);
1218 	if (vp != NULL)
1219 		vput(vp);
1220 	return (error);
1221 }
1222 
1223 /* ARGSUSED */
1224 int
1225 sys_fhstat(p, v, retval)
1226 	struct proc *p;
1227 	void *v;
1228 	register_t *retval;
1229 {
1230 	struct sys_fhstat_args /* {
1231 		syscallarg(const fhandle_t *) fhp;
1232 		syscallarg(struct stat *) sb;
1233 	} */ *uap = v;
1234 	struct stat sb;
1235 	int error;
1236 	fhandle_t fh;
1237 	struct mount *mp;
1238 	struct vnode *vp;
1239 
1240 	/*
1241 	 * Must be super user
1242 	 */
1243 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1244 		return (error);
1245 
1246 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1247 		return (error);
1248 
1249 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1250 		return (ESTALE);
1251 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1252 		return (error);
1253 	error = vn_stat(vp, &sb, p);
1254 	vput(vp);
1255 	if (error)
1256 		return (error);
1257 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1258 	return (error);
1259 }
1260 
1261 /* ARGSUSED */
1262 int
1263 sys_fhstatfs(p, v, retval)
1264 	struct proc *p;
1265 	void *v;
1266 	register_t *retval;
1267 {
1268 	struct sys_fhstatfs_args /*
1269 		syscallarg(const fhandle_t *) fhp;
1270 		syscallarg(struct statfs *) buf;
1271 	} */ *uap = v;
1272 	struct statfs sp;
1273 	fhandle_t fh;
1274 	struct mount *mp;
1275 	struct vnode *vp;
1276 	int error;
1277 
1278 	/*
1279 	 * Must be super user
1280 	 */
1281 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1282 		return (error);
1283 
1284 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1285 		return (error);
1286 
1287 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1288 		return (ESTALE);
1289 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1290 		return (error);
1291 	mp = vp->v_mount;
1292 	vput(vp);
1293 	if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1294 		return (error);
1295 	sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1296 	sp.f_oflags = sp.f_flags & 0xffff;
1297 	return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1298 }
1299 
1300 /*
1301  * Create a special file.
1302  */
1303 /* ARGSUSED */
1304 int
1305 sys_mknod(p, v, retval)
1306 	struct proc *p;
1307 	void *v;
1308 	register_t *retval;
1309 {
1310 	struct sys_mknod_args /* {
1311 		syscallarg(const char *) path;
1312 		syscallarg(int) mode;
1313 		syscallarg(int) dev;
1314 	} */ *uap = v;
1315 	struct vnode *vp;
1316 	struct vattr vattr;
1317 	int error;
1318 	int whiteout = 0;
1319 	struct nameidata nd;
1320 
1321 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1322 		return (error);
1323 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1324 	if ((error = namei(&nd)) != 0)
1325 		return (error);
1326 	vp = nd.ni_vp;
1327 	if (vp != NULL)
1328 		error = EEXIST;
1329 	else {
1330 		VATTR_NULL(&vattr);
1331 		vattr.va_mode =
1332 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1333 		vattr.va_rdev = SCARG(uap, dev);
1334 		whiteout = 0;
1335 
1336 		switch (SCARG(uap, mode) & S_IFMT) {
1337 		case S_IFMT:	/* used by badsect to flag bad sectors */
1338 			vattr.va_type = VBAD;
1339 			break;
1340 		case S_IFCHR:
1341 			vattr.va_type = VCHR;
1342 			break;
1343 		case S_IFBLK:
1344 			vattr.va_type = VBLK;
1345 			break;
1346 		case S_IFWHT:
1347 			whiteout = 1;
1348 			break;
1349 		default:
1350 			error = EINVAL;
1351 			break;
1352 		}
1353 	}
1354 	if (!error) {
1355 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1356 		if (whiteout) {
1357 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1358 			if (error)
1359 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1360 			vput(nd.ni_dvp);
1361 		} else {
1362 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1363 						&nd.ni_cnd, &vattr);
1364 		}
1365 	} else {
1366 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1367 		if (nd.ni_dvp == vp)
1368 			vrele(nd.ni_dvp);
1369 		else
1370 			vput(nd.ni_dvp);
1371 		if (vp)
1372 			vrele(vp);
1373 	}
1374 	return (error);
1375 }
1376 
1377 /*
1378  * Create a named pipe.
1379  */
1380 /* ARGSUSED */
1381 int
1382 sys_mkfifo(p, v, retval)
1383 	struct proc *p;
1384 	void *v;
1385 	register_t *retval;
1386 {
1387 	struct sys_mkfifo_args /* {
1388 		syscallarg(const char *) path;
1389 		syscallarg(int) mode;
1390 	} */ *uap = v;
1391 	struct vattr vattr;
1392 	int error;
1393 	struct nameidata nd;
1394 
1395 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1396 	if ((error = namei(&nd)) != 0)
1397 		return (error);
1398 	if (nd.ni_vp != NULL) {
1399 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1400 		if (nd.ni_dvp == nd.ni_vp)
1401 			vrele(nd.ni_dvp);
1402 		else
1403 			vput(nd.ni_dvp);
1404 		vrele(nd.ni_vp);
1405 		return (EEXIST);
1406 	}
1407 	VATTR_NULL(&vattr);
1408 	vattr.va_type = VFIFO;
1409 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1410 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1411 	return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
1412 }
1413 
1414 /*
1415  * Make a hard file link.
1416  */
1417 /* ARGSUSED */
1418 int
1419 sys_link(p, v, retval)
1420 	struct proc *p;
1421 	void *v;
1422 	register_t *retval;
1423 {
1424 	struct sys_link_args /* {
1425 		syscallarg(const char *) path;
1426 		syscallarg(const char *) link;
1427 	} */ *uap = v;
1428 	struct vnode *vp;
1429 	struct nameidata nd;
1430 	int error;
1431 
1432 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1433 	if ((error = namei(&nd)) != 0)
1434 		return (error);
1435 	vp = nd.ni_vp;
1436 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1437 	if ((error = namei(&nd)) != 0)
1438 		goto out;
1439 	if (nd.ni_vp) {
1440 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1441 		if (nd.ni_dvp == nd.ni_vp)
1442 			vrele(nd.ni_dvp);
1443 		else
1444 			vput(nd.ni_dvp);
1445 		vrele(nd.ni_vp);
1446 		error = EEXIST;
1447 		goto out;
1448 	}
1449 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1450 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1451 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1452 out:
1453 	vrele(vp);
1454 	return (error);
1455 }
1456 
1457 /*
1458  * Make a symbolic link.
1459  */
1460 /* ARGSUSED */
1461 int
1462 sys_symlink(p, v, retval)
1463 	struct proc *p;
1464 	void *v;
1465 	register_t *retval;
1466 {
1467 	struct sys_symlink_args /* {
1468 		syscallarg(const char *) path;
1469 		syscallarg(const char *) link;
1470 	} */ *uap = v;
1471 	struct vattr vattr;
1472 	char *path;
1473 	int error;
1474 	struct nameidata nd;
1475 
1476 	MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1477 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1478 	if (error)
1479 		goto out;
1480 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1481 	if ((error = namei(&nd)) != 0)
1482 		goto out;
1483 	if (nd.ni_vp) {
1484 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1485 		if (nd.ni_dvp == nd.ni_vp)
1486 			vrele(nd.ni_dvp);
1487 		else
1488 			vput(nd.ni_dvp);
1489 		vrele(nd.ni_vp);
1490 		error = EEXIST;
1491 		goto out;
1492 	}
1493 	VATTR_NULL(&vattr);
1494 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1495 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1496 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1497 out:
1498 	FREE(path, M_NAMEI);
1499 	return (error);
1500 }
1501 
1502 /*
1503  * Delete a whiteout from the filesystem.
1504  */
1505 /* ARGSUSED */
1506 int
1507 sys_undelete(p, v, retval)
1508 	struct proc *p;
1509 	void *v;
1510 	register_t *retval;
1511 {
1512 	struct sys_undelete_args /* {
1513 		syscallarg(const char *) path;
1514 	} */ *uap = v;
1515 	int error;
1516 	struct nameidata nd;
1517 
1518 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1519 	    SCARG(uap, path), p);
1520 	error = namei(&nd);
1521 	if (error)
1522 		return (error);
1523 
1524 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1525 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1526 		if (nd.ni_dvp == nd.ni_vp)
1527 			vrele(nd.ni_dvp);
1528 		else
1529 			vput(nd.ni_dvp);
1530 		if (nd.ni_vp)
1531 			vrele(nd.ni_vp);
1532 		return (EEXIST);
1533 	}
1534 
1535 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1536 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1537 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1538 	vput(nd.ni_dvp);
1539 	return (error);
1540 }
1541 
1542 /*
1543  * Delete a name from the filesystem.
1544  */
1545 /* ARGSUSED */
1546 int
1547 sys_unlink(p, v, retval)
1548 	struct proc *p;
1549 	void *v;
1550 	register_t *retval;
1551 {
1552 	struct sys_unlink_args /* {
1553 		syscallarg(const char *) path;
1554 	} */ *uap = v;
1555 	struct vnode *vp;
1556 	int error;
1557 	struct nameidata nd;
1558 
1559 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1560 	    SCARG(uap, path), p);
1561 	if ((error = namei(&nd)) != 0)
1562 		return (error);
1563 	vp = nd.ni_vp;
1564 
1565 	/*
1566 	 * The root of a mounted filesystem cannot be deleted.
1567 	 */
1568 	if (vp->v_flag & VROOT) {
1569 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1570 		if (nd.ni_dvp == vp)
1571 			vrele(nd.ni_dvp);
1572 		else
1573 			vput(nd.ni_dvp);
1574 		vput(vp);
1575 		error = EBUSY;
1576 		goto out;
1577 	}
1578 
1579 	(void)uvm_vnp_uncache(vp);
1580 
1581 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1582 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1583 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1584 out:
1585 	return (error);
1586 }
1587 
1588 /*
1589  * Reposition read/write file offset.
1590  */
1591 int
1592 sys_lseek(p, v, retval)
1593 	struct proc *p;
1594 	void *v;
1595 	register_t *retval;
1596 {
1597 	struct sys_lseek_args /* {
1598 		syscallarg(int) fd;
1599 		syscallarg(int) pad;
1600 		syscallarg(off_t) offset;
1601 		syscallarg(int) whence;
1602 	} */ *uap = v;
1603 	struct ucred *cred = p->p_ucred;
1604 	struct filedesc *fdp = p->p_fd;
1605 	struct file *fp;
1606 	struct vnode *vp;
1607 	struct vattr vattr;
1608 	off_t newoff;
1609 	int error;
1610 
1611 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1612 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL ||
1613 	    (fp->f_iflags & FIF_WANTCLOSE) != 0)
1614 		return (EBADF);
1615 
1616 	FILE_USE(fp);
1617 
1618 	vp = (struct vnode *)fp->f_data;
1619 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1620 		error = ESPIPE;
1621 		goto out;
1622 	}
1623 
1624 	switch (SCARG(uap, whence)) {
1625 	case SEEK_CUR:
1626 		newoff = fp->f_offset + SCARG(uap, offset);
1627 		break;
1628 	case SEEK_END:
1629 		error = VOP_GETATTR(vp, &vattr, cred, p);
1630 		if (error)
1631 			goto out;
1632 		newoff = SCARG(uap, offset) + vattr.va_size;
1633 		break;
1634 	case SEEK_SET:
1635 		newoff = SCARG(uap, offset);
1636 		break;
1637 	default:
1638 		error = EINVAL;
1639 		goto out;
1640 	}
1641 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1642 		goto out;
1643 
1644 	*(off_t *)retval = fp->f_offset = newoff;
1645  out:
1646 	FILE_UNUSE(fp, p);
1647 	return (error);
1648 }
1649 
1650 /*
1651  * Positional read system call.
1652  */
1653 int
1654 sys_pread(p, v, retval)
1655 	struct proc *p;
1656 	void *v;
1657 	register_t *retval;
1658 {
1659 	struct sys_pread_args /* {
1660 		syscallarg(int) fd;
1661 		syscallarg(void *) buf;
1662 		syscallarg(size_t) nbyte;
1663 		syscallarg(off_t) offset;
1664 	} */ *uap = v;
1665 	struct filedesc *fdp = p->p_fd;
1666 	struct file *fp;
1667 	struct vnode *vp;
1668 	off_t offset;
1669 	int error, fd = SCARG(uap, fd);
1670 
1671 	if ((u_int)fd >= fdp->fd_nfiles ||
1672 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
1673 	    (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1674 	    (fp->f_flag & FREAD) == 0)
1675 		return (EBADF);
1676 
1677 	FILE_USE(fp);
1678 
1679 	vp = (struct vnode *)fp->f_data;
1680 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1681 		error = ESPIPE;
1682 		goto out;
1683 	}
1684 
1685 	offset = SCARG(uap, offset);
1686 
1687 	/*
1688 	 * XXX This works because no file systems actually
1689 	 * XXX take any action on the seek operation.
1690 	 */
1691 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1692 		goto out;
1693 
1694 	/* dofileread() will unuse the descriptor for us */
1695 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1696 	    &offset, 0, retval));
1697 
1698  out:
1699 	FILE_UNUSE(fp, p);
1700 	return (error);
1701 }
1702 
1703 /*
1704  * Positional scatter read system call.
1705  */
1706 int
1707 sys_preadv(p, v, retval)
1708 	struct proc *p;
1709 	void *v;
1710 	register_t *retval;
1711 {
1712 	struct sys_preadv_args /* {
1713 		syscallarg(int) fd;
1714 		syscallarg(const struct iovec *) iovp;
1715 		syscallarg(int) iovcnt;
1716 		syscallarg(off_t) offset;
1717 	} */ *uap = v;
1718 	struct filedesc *fdp = p->p_fd;
1719 	struct file *fp;
1720 	struct vnode *vp;
1721 	off_t offset;
1722 	int error, fd = SCARG(uap, fd);
1723 
1724 	if ((u_int)fd >= fdp->fd_nfiles ||
1725 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
1726 	    (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1727 	    (fp->f_flag & FREAD) == 0)
1728 		return (EBADF);
1729 
1730 	FILE_USE(fp);
1731 
1732 	vp = (struct vnode *)fp->f_data;
1733 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1734 		error = ESPIPE;
1735 		goto out;
1736 	}
1737 
1738 	offset = SCARG(uap, offset);
1739 
1740 	/*
1741 	 * XXX This works because no file systems actually
1742 	 * XXX take any action on the seek operation.
1743 	 */
1744 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1745 		goto out;
1746 
1747 	/* dofilereadv() will unuse the descriptor for us */
1748 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1749 	    &offset, 0, retval));
1750 
1751  out:
1752 	FILE_UNUSE(fp, p);
1753 	return (error);
1754 }
1755 
1756 /*
1757  * Positional write system call.
1758  */
1759 int
1760 sys_pwrite(p, v, retval)
1761 	struct proc *p;
1762 	void *v;
1763 	register_t *retval;
1764 {
1765 	struct sys_pwrite_args /* {
1766 		syscallarg(int) fd;
1767 		syscallarg(const void *) buf;
1768 		syscallarg(size_t) nbyte;
1769 		syscallarg(off_t) offset;
1770 	} */ *uap = v;
1771 	struct filedesc *fdp = p->p_fd;
1772 	struct file *fp;
1773 	struct vnode *vp;
1774 	off_t offset;
1775 	int error, fd = SCARG(uap, fd);
1776 
1777 	if ((u_int)fd >= fdp->fd_nfiles ||
1778 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
1779 	    (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1780 	    (fp->f_flag & FWRITE) == 0)
1781 		return (EBADF);
1782 
1783 	FILE_USE(fp);
1784 
1785 	vp = (struct vnode *)fp->f_data;
1786 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1787 		error = ESPIPE;
1788 		goto out;
1789 	}
1790 
1791 	offset = SCARG(uap, offset);
1792 
1793 	/*
1794 	 * XXX This works because no file systems actually
1795 	 * XXX take any action on the seek operation.
1796 	 */
1797 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1798 		goto out;
1799 
1800 	/* dofilewrite() will unuse the descriptor for us */
1801 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1802 	    &offset, 0, retval));
1803 
1804  out:
1805 	FILE_UNUSE(fp, p);
1806 	return (error);
1807 }
1808 
1809 /*
1810  * Positional gather write system call.
1811  */
1812 int
1813 sys_pwritev(p, v, retval)
1814 	struct proc *p;
1815 	void *v;
1816 	register_t *retval;
1817 {
1818 	struct sys_pwritev_args /* {
1819 		syscallarg(int) fd;
1820 		syscallarg(const struct iovec *) iovp;
1821 		syscallarg(int) iovcnt;
1822 		syscallarg(off_t) offset;
1823 	} */ *uap = v;
1824 	struct filedesc *fdp = p->p_fd;
1825 	struct file *fp;
1826 	struct vnode *vp;
1827 	off_t offset;
1828 	int error, fd = SCARG(uap, fd);
1829 
1830 	if ((u_int)fd >= fdp->fd_nfiles ||
1831 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
1832 	    (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1833 	    (fp->f_flag & FWRITE) == 0)
1834 		return (EBADF);
1835 
1836 	FILE_USE(fp);
1837 
1838 	vp = (struct vnode *)fp->f_data;
1839 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1840 		error = ESPIPE;
1841 		goto out;
1842 	}
1843 
1844 	offset = SCARG(uap, offset);
1845 
1846 	/*
1847 	 * XXX This works because no file systems actually
1848 	 * XXX take any action on the seek operation.
1849 	 */
1850 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1851 		goto out;
1852 
1853 	/* dofilewritev() will unuse the descriptor for us */
1854 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1855 	    &offset, 0, retval));
1856 
1857  out:
1858 	FILE_UNUSE(fp, p);
1859 	return (error);
1860 }
1861 
1862 /*
1863  * Check access permissions.
1864  */
1865 int
1866 sys_access(p, v, retval)
1867 	struct proc *p;
1868 	void *v;
1869 	register_t *retval;
1870 {
1871 	struct sys_access_args /* {
1872 		syscallarg(const char *) path;
1873 		syscallarg(int) flags;
1874 	} */ *uap = v;
1875 	struct ucred *cred = p->p_ucred;
1876 	struct vnode *vp;
1877 	int error, flags, t_gid, t_uid;
1878 	struct nameidata nd;
1879 
1880 	t_uid = cred->cr_uid;
1881 	t_gid = cred->cr_gid;
1882 	cred->cr_uid = p->p_cred->p_ruid;
1883 	cred->cr_gid = p->p_cred->p_rgid;
1884 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1885 	    SCARG(uap, path), p);
1886 	if ((error = namei(&nd)) != 0)
1887 		goto out1;
1888 	vp = nd.ni_vp;
1889 
1890 	/* Flags == 0 means only check for existence. */
1891 	if (SCARG(uap, flags)) {
1892 		flags = 0;
1893 		if (SCARG(uap, flags) & R_OK)
1894 			flags |= VREAD;
1895 		if (SCARG(uap, flags) & W_OK)
1896 			flags |= VWRITE;
1897 		if (SCARG(uap, flags) & X_OK)
1898 			flags |= VEXEC;
1899 
1900 		error = VOP_ACCESS(vp, flags, cred, p);
1901 		if (!error && (flags & VWRITE))
1902 			error = vn_writechk(vp);
1903 	}
1904 	vput(vp);
1905 out1:
1906 	cred->cr_uid = t_uid;
1907 	cred->cr_gid = t_gid;
1908 	return (error);
1909 }
1910 
1911 /*
1912  * Get file status; this version follows links.
1913  */
1914 /* ARGSUSED */
1915 int
1916 sys___stat13(p, v, retval)
1917 	struct proc *p;
1918 	void *v;
1919 	register_t *retval;
1920 {
1921 	struct sys___stat13_args /* {
1922 		syscallarg(const char *) path;
1923 		syscallarg(struct stat *) ub;
1924 	} */ *uap = v;
1925 	struct stat sb;
1926 	int error;
1927 	struct nameidata nd;
1928 
1929 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1930 	    SCARG(uap, path), p);
1931 	if ((error = namei(&nd)) != 0)
1932 		return (error);
1933 	error = vn_stat(nd.ni_vp, &sb, p);
1934 	vput(nd.ni_vp);
1935 	if (error)
1936 		return (error);
1937 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1938 	return (error);
1939 }
1940 
1941 /*
1942  * Get file status; this version does not follow links.
1943  */
1944 /* ARGSUSED */
1945 int
1946 sys___lstat13(p, v, retval)
1947 	struct proc *p;
1948 	void *v;
1949 	register_t *retval;
1950 {
1951 	struct sys___lstat13_args /* {
1952 		syscallarg(const char *) path;
1953 		syscallarg(struct stat *) ub;
1954 	} */ *uap = v;
1955 	struct stat sb;
1956 	int error;
1957 	struct nameidata nd;
1958 
1959 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1960 	    SCARG(uap, path), p);
1961 	if ((error = namei(&nd)) != 0)
1962 		return (error);
1963 	error = vn_stat(nd.ni_vp, &sb, p);
1964 	vput(nd.ni_vp);
1965 	if (error)
1966 		return (error);
1967 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1968 	return (error);
1969 }
1970 
1971 /*
1972  * Get configurable pathname variables.
1973  */
1974 /* ARGSUSED */
1975 int
1976 sys_pathconf(p, v, retval)
1977 	struct proc *p;
1978 	void *v;
1979 	register_t *retval;
1980 {
1981 	struct sys_pathconf_args /* {
1982 		syscallarg(const char *) path;
1983 		syscallarg(int) name;
1984 	} */ *uap = v;
1985 	int error;
1986 	struct nameidata nd;
1987 
1988 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1989 	    SCARG(uap, path), p);
1990 	if ((error = namei(&nd)) != 0)
1991 		return (error);
1992 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1993 	vput(nd.ni_vp);
1994 	return (error);
1995 }
1996 
1997 /*
1998  * Return target name of a symbolic link.
1999  */
2000 /* ARGSUSED */
2001 int
2002 sys_readlink(p, v, retval)
2003 	struct proc *p;
2004 	void *v;
2005 	register_t *retval;
2006 {
2007 	struct sys_readlink_args /* {
2008 		syscallarg(const char *) path;
2009 		syscallarg(char *) buf;
2010 		syscallarg(size_t) count;
2011 	} */ *uap = v;
2012 	struct vnode *vp;
2013 	struct iovec aiov;
2014 	struct uio auio;
2015 	int error;
2016 	struct nameidata nd;
2017 
2018 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2019 	    SCARG(uap, path), p);
2020 	if ((error = namei(&nd)) != 0)
2021 		return (error);
2022 	vp = nd.ni_vp;
2023 	if (vp->v_type != VLNK)
2024 		error = EINVAL;
2025 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2026 	    (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2027 		aiov.iov_base = SCARG(uap, buf);
2028 		aiov.iov_len = SCARG(uap, count);
2029 		auio.uio_iov = &aiov;
2030 		auio.uio_iovcnt = 1;
2031 		auio.uio_offset = 0;
2032 		auio.uio_rw = UIO_READ;
2033 		auio.uio_segflg = UIO_USERSPACE;
2034 		auio.uio_procp = p;
2035 		auio.uio_resid = SCARG(uap, count);
2036 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2037 	}
2038 	vput(vp);
2039 	*retval = SCARG(uap, count) - auio.uio_resid;
2040 	return (error);
2041 }
2042 
2043 /*
2044  * Change flags of a file given a path name.
2045  */
2046 /* ARGSUSED */
2047 int
2048 sys_chflags(p, v, retval)
2049 	struct proc *p;
2050 	void *v;
2051 	register_t *retval;
2052 {
2053 	struct sys_chflags_args /* {
2054 		syscallarg(const char *) path;
2055 		syscallarg(u_long) flags;
2056 	} */ *uap = v;
2057 	struct vnode *vp;
2058 	struct vattr vattr;
2059 	int error;
2060 	struct nameidata nd;
2061 
2062 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2063 	if ((error = namei(&nd)) != 0)
2064 		return (error);
2065 	vp = nd.ni_vp;
2066 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2067 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2068 	/* Non-superusers cannot change the flags on devices, even if they
2069 	   own them. */
2070 	if (suser(p->p_ucred, &p->p_acflag)) {
2071 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2072 			goto out;
2073 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2074 			error = EINVAL;
2075 			goto out;
2076 		}
2077 	}
2078 	VATTR_NULL(&vattr);
2079 	vattr.va_flags = SCARG(uap, flags);
2080 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2081 out:
2082 	vput(vp);
2083 	return (error);
2084 }
2085 
2086 /*
2087  * Change flags of a file given a file descriptor.
2088  */
2089 /* ARGSUSED */
2090 int
2091 sys_fchflags(p, v, retval)
2092 	struct proc *p;
2093 	void *v;
2094 	register_t *retval;
2095 {
2096 	struct sys_fchflags_args /* {
2097 		syscallarg(int) fd;
2098 		syscallarg(u_long) flags;
2099 	} */ *uap = v;
2100 	struct vattr vattr;
2101 	struct vnode *vp;
2102 	struct file *fp;
2103 	int error;
2104 
2105 	/* getvnode() will use the descriptor for us */
2106 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2107 		return (error);
2108 	vp = (struct vnode *)fp->f_data;
2109 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2110 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2111 	/* Non-superusers cannot change the flags on devices, even if they
2112 	   own them. */
2113 	if (suser(p->p_ucred, &p->p_acflag)) {
2114 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
2115 		    != 0)
2116 			goto out;
2117 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2118 			error = EINVAL;
2119 			goto out;
2120 		}
2121 	}
2122 	VATTR_NULL(&vattr);
2123 	vattr.va_flags = SCARG(uap, flags);
2124 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2125 out:
2126 	VOP_UNLOCK(vp, 0);
2127 	FILE_UNUSE(fp, p);
2128 	return (error);
2129 }
2130 
2131 /*
2132  * Change flags of a file given a file descriptor; this version does
2133  * not follow links.
2134  */
2135 int
2136 sys_lchflags(p, v, retval)
2137 	struct proc *p;
2138 	void *v;
2139 	register_t *retval;
2140 {
2141 	register struct sys_chflags_args /* {
2142 		syscallarg(const char *) path;
2143 		syscallarg(u_long) flags;
2144 	} */ *uap = v;
2145 	register struct vnode *vp;
2146 	struct vattr vattr;
2147 	int error;
2148 	struct nameidata nd;
2149 
2150 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2151 	if ((error = namei(&nd)) != 0)
2152 		return (error);
2153 	vp = nd.ni_vp;
2154 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2155 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2156 	/* Non-superusers cannot change the flags on devices, even if they
2157 	   own them. */
2158 	if (suser(p->p_ucred, &p->p_acflag)) {
2159 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2160 			goto out;
2161 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2162 			error = EINVAL;
2163 			goto out;
2164 		}
2165 	}
2166 	VATTR_NULL(&vattr);
2167 	vattr.va_flags = SCARG(uap, flags);
2168 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2169 out:
2170 	vput(vp);
2171 	return (error);
2172 }
2173 
2174 /*
2175  * Change mode of a file given path name; this version follows links.
2176  */
2177 /* ARGSUSED */
2178 int
2179 sys_chmod(p, v, retval)
2180 	struct proc *p;
2181 	void *v;
2182 	register_t *retval;
2183 {
2184 	struct sys_chmod_args /* {
2185 		syscallarg(const char *) path;
2186 		syscallarg(int) mode;
2187 	} */ *uap = v;
2188 	int error;
2189 	struct nameidata nd;
2190 
2191 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2192 	if ((error = namei(&nd)) != 0)
2193 		return (error);
2194 
2195 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2196 
2197 	vrele(nd.ni_vp);
2198 	return (error);
2199 }
2200 
2201 /*
2202  * Change mode of a file given a file descriptor.
2203  */
2204 /* ARGSUSED */
2205 int
2206 sys_fchmod(p, v, retval)
2207 	struct proc *p;
2208 	void *v;
2209 	register_t *retval;
2210 {
2211 	struct sys_fchmod_args /* {
2212 		syscallarg(int) fd;
2213 		syscallarg(int) mode;
2214 	} */ *uap = v;
2215 	struct file *fp;
2216 	int error;
2217 
2218 	/* getvnode() will use the descriptor for us */
2219 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2220 		return (error);
2221 
2222 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2223 	FILE_UNUSE(fp, p);
2224 	return (error);
2225 }
2226 
2227 /*
2228  * Change mode of a file given path name; this version does not follow links.
2229  */
2230 /* ARGSUSED */
2231 int
2232 sys_lchmod(p, v, retval)
2233 	struct proc *p;
2234 	void *v;
2235 	register_t *retval;
2236 {
2237 	struct sys_lchmod_args /* {
2238 		syscallarg(const char *) path;
2239 		syscallarg(int) mode;
2240 	} */ *uap = v;
2241 	int error;
2242 	struct nameidata nd;
2243 
2244 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2245 	if ((error = namei(&nd)) != 0)
2246 		return (error);
2247 
2248 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2249 
2250 	vrele(nd.ni_vp);
2251 	return (error);
2252 }
2253 
2254 /*
2255  * Common routine to set mode given a vnode.
2256  */
2257 static int
2258 change_mode(vp, mode, p)
2259 	struct vnode *vp;
2260 	int mode;
2261 	struct proc *p;
2262 {
2263 	struct vattr vattr;
2264 	int error;
2265 
2266 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2267 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2268 	VATTR_NULL(&vattr);
2269 	vattr.va_mode = mode & ALLPERMS;
2270 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2271 	VOP_UNLOCK(vp, 0);
2272 	return (error);
2273 }
2274 
2275 /*
2276  * Set ownership given a path name; this version follows links.
2277  */
2278 /* ARGSUSED */
2279 int
2280 sys_chown(p, v, retval)
2281 	struct proc *p;
2282 	void *v;
2283 	register_t *retval;
2284 {
2285 	struct sys_chown_args /* {
2286 		syscallarg(const char *) path;
2287 		syscallarg(uid_t) uid;
2288 		syscallarg(gid_t) gid;
2289 	} */ *uap = v;
2290 	int error;
2291 	struct nameidata nd;
2292 
2293 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2294 	if ((error = namei(&nd)) != 0)
2295 		return (error);
2296 
2297 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2298 
2299 	vrele(nd.ni_vp);
2300 	return (error);
2301 }
2302 
2303 /*
2304  * Set ownership given a path name; this version follows links.
2305  * Provides POSIX semantics.
2306  */
2307 /* ARGSUSED */
2308 int
2309 sys___posix_chown(p, v, retval)
2310 	struct proc *p;
2311 	void *v;
2312 	register_t *retval;
2313 {
2314 	struct sys_chown_args /* {
2315 		syscallarg(const char *) path;
2316 		syscallarg(uid_t) uid;
2317 		syscallarg(gid_t) gid;
2318 	} */ *uap = v;
2319 	int error;
2320 	struct nameidata nd;
2321 
2322 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2323 	if ((error = namei(&nd)) != 0)
2324 		return (error);
2325 
2326 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2327 
2328 	vrele(nd.ni_vp);
2329 	return (error);
2330 }
2331 
2332 /*
2333  * Set ownership given a file descriptor.
2334  */
2335 /* ARGSUSED */
2336 int
2337 sys_fchown(p, v, retval)
2338 	struct proc *p;
2339 	void *v;
2340 	register_t *retval;
2341 {
2342 	struct sys_fchown_args /* {
2343 		syscallarg(int) fd;
2344 		syscallarg(uid_t) uid;
2345 		syscallarg(gid_t) gid;
2346 	} */ *uap = v;
2347 	int error;
2348 	struct file *fp;
2349 
2350 	/* getvnode() will use the descriptor for us */
2351 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2352 		return (error);
2353 
2354 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2355 	    SCARG(uap, gid), p, 0);
2356 	FILE_UNUSE(fp, p);
2357 	return (error);
2358 }
2359 
2360 /*
2361  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2362  */
2363 /* ARGSUSED */
2364 int
2365 sys___posix_fchown(p, v, retval)
2366 	struct proc *p;
2367 	void *v;
2368 	register_t *retval;
2369 {
2370 	struct sys_fchown_args /* {
2371 		syscallarg(int) fd;
2372 		syscallarg(uid_t) uid;
2373 		syscallarg(gid_t) gid;
2374 	} */ *uap = v;
2375 	int error;
2376 	struct file *fp;
2377 
2378 	/* getvnode() will use the descriptor for us */
2379 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2380 		return (error);
2381 
2382 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2383 	    SCARG(uap, gid), p, 1);
2384 	FILE_UNUSE(fp, p);
2385 	return (error);
2386 }
2387 
2388 /*
2389  * Set ownership given a path name; this version does not follow links.
2390  */
2391 /* ARGSUSED */
2392 int
2393 sys_lchown(p, v, retval)
2394 	struct proc *p;
2395 	void *v;
2396 	register_t *retval;
2397 {
2398 	struct sys_lchown_args /* {
2399 		syscallarg(const char *) path;
2400 		syscallarg(uid_t) uid;
2401 		syscallarg(gid_t) gid;
2402 	} */ *uap = v;
2403 	int error;
2404 	struct nameidata nd;
2405 
2406 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2407 	if ((error = namei(&nd)) != 0)
2408 		return (error);
2409 
2410 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2411 
2412 	vrele(nd.ni_vp);
2413 	return (error);
2414 }
2415 
2416 /*
2417  * Set ownership given a path name; this version does not follow links.
2418  * Provides POSIX/XPG semantics.
2419  */
2420 /* ARGSUSED */
2421 int
2422 sys___posix_lchown(p, v, retval)
2423 	struct proc *p;
2424 	void *v;
2425 	register_t *retval;
2426 {
2427 	struct sys_lchown_args /* {
2428 		syscallarg(const char *) path;
2429 		syscallarg(uid_t) uid;
2430 		syscallarg(gid_t) gid;
2431 	} */ *uap = v;
2432 	int error;
2433 	struct nameidata nd;
2434 
2435 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2436 	if ((error = namei(&nd)) != 0)
2437 		return (error);
2438 
2439 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2440 
2441 	vrele(nd.ni_vp);
2442 	return (error);
2443 }
2444 
2445 /*
2446  * Common routine to set ownership given a vnode.
2447  */
2448 static int
2449 change_owner(vp, uid, gid, p, posix_semantics)
2450 	struct vnode *vp;
2451 	uid_t uid;
2452 	gid_t gid;
2453 	struct proc *p;
2454 	int posix_semantics;
2455 {
2456 	struct vattr vattr;
2457 	mode_t newmode;
2458 	int error;
2459 
2460 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2461 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2462 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2463 		goto out;
2464 
2465 #define CHANGED(x) ((x) != -1)
2466 	newmode = vattr.va_mode;
2467 	if (posix_semantics) {
2468 		/*
2469 		 * POSIX/XPG semantics: if the caller is not the super-user,
2470 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2471 		 * the XPG consider the behaviour for calls by the super-user
2472 		 * implementation-defined; we leave the set-user-id and set-
2473 		 * group-id settings intact in that case.
2474 		 */
2475 		if (suser(p->p_ucred, NULL) != 0)
2476 			newmode &= ~(S_ISUID | S_ISGID);
2477 	} else {
2478 		/*
2479 		 * NetBSD semantics: when changing owner and/or group,
2480 		 * clear the respective bit(s).
2481 		 */
2482 		if (CHANGED(uid))
2483 			newmode &= ~S_ISUID;
2484 		if (CHANGED(gid))
2485 			newmode &= ~S_ISGID;
2486 	}
2487 	/* Update va_mode iff altered. */
2488 	if (vattr.va_mode == newmode)
2489 		newmode = VNOVAL;
2490 
2491 	VATTR_NULL(&vattr);
2492 	vattr.va_uid = CHANGED(uid) ? uid : VNOVAL;
2493 	vattr.va_gid = CHANGED(gid) ? gid : VNOVAL;
2494 	vattr.va_mode = newmode;
2495 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2496 #undef CHANGED
2497 
2498 out:
2499 	VOP_UNLOCK(vp, 0);
2500 	return (error);
2501 }
2502 
2503 /*
2504  * Set the access and modification times given a path name; this
2505  * version follows links.
2506  */
2507 /* ARGSUSED */
2508 int
2509 sys_utimes(p, v, retval)
2510 	struct proc *p;
2511 	void *v;
2512 	register_t *retval;
2513 {
2514 	struct sys_utimes_args /* {
2515 		syscallarg(const char *) path;
2516 		syscallarg(const struct timeval *) tptr;
2517 	} */ *uap = v;
2518 	int error;
2519 	struct nameidata nd;
2520 
2521 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2522 	if ((error = namei(&nd)) != 0)
2523 		return (error);
2524 
2525 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2526 
2527 	vrele(nd.ni_vp);
2528 	return (error);
2529 }
2530 
2531 /*
2532  * Set the access and modification times given a file descriptor.
2533  */
2534 /* ARGSUSED */
2535 int
2536 sys_futimes(p, v, retval)
2537 	struct proc *p;
2538 	void *v;
2539 	register_t *retval;
2540 {
2541 	struct sys_futimes_args /* {
2542 		syscallarg(int) fd;
2543 		syscallarg(const struct timeval *) tptr;
2544 	} */ *uap = v;
2545 	int error;
2546 	struct file *fp;
2547 
2548 	/* getvnode() will use the descriptor for us */
2549 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2550 		return (error);
2551 
2552 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2553 	FILE_UNUSE(fp, p);
2554 	return (error);
2555 }
2556 
2557 /*
2558  * Set the access and modification times given a path name; this
2559  * version does not follow links.
2560  */
2561 /* ARGSUSED */
2562 int
2563 sys_lutimes(p, v, retval)
2564 	struct proc *p;
2565 	void *v;
2566 	register_t *retval;
2567 {
2568 	struct sys_lutimes_args /* {
2569 		syscallarg(const char *) path;
2570 		syscallarg(const struct timeval *) tptr;
2571 	} */ *uap = v;
2572 	int error;
2573 	struct nameidata nd;
2574 
2575 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2576 	if ((error = namei(&nd)) != 0)
2577 		return (error);
2578 
2579 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2580 
2581 	vrele(nd.ni_vp);
2582 	return (error);
2583 }
2584 
2585 /*
2586  * Common routine to set access and modification times given a vnode.
2587  */
2588 static int
2589 change_utimes(vp, tptr, p)
2590 	struct vnode *vp;
2591 	const struct timeval *tptr;
2592 	struct proc *p;
2593 {
2594 	struct timeval tv[2];
2595 	struct vattr vattr;
2596 	int error;
2597 
2598 	VATTR_NULL(&vattr);
2599 	if (tptr == NULL) {
2600 		microtime(&tv[0]);
2601 		tv[1] = tv[0];
2602 		vattr.va_vaflags |= VA_UTIMES_NULL;
2603 	} else {
2604 		error = copyin(tptr, tv, sizeof(tv));
2605 		if (error)
2606 			return (error);
2607 	}
2608 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2609 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2610 	vattr.va_atime.tv_sec = tv[0].tv_sec;
2611 	vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2612 	vattr.va_mtime.tv_sec = tv[1].tv_sec;
2613 	vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2614 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2615 	VOP_UNLOCK(vp, 0);
2616 	return (error);
2617 }
2618 
2619 /*
2620  * Truncate a file given its path name.
2621  */
2622 /* ARGSUSED */
2623 int
2624 sys_truncate(p, v, retval)
2625 	struct proc *p;
2626 	void *v;
2627 	register_t *retval;
2628 {
2629 	struct sys_truncate_args /* {
2630 		syscallarg(const char *) path;
2631 		syscallarg(int) pad;
2632 		syscallarg(off_t) length;
2633 	} */ *uap = v;
2634 	struct vnode *vp;
2635 	struct vattr vattr;
2636 	int error;
2637 	struct nameidata nd;
2638 
2639 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2640 	if ((error = namei(&nd)) != 0)
2641 		return (error);
2642 	vp = nd.ni_vp;
2643 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2644 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2645 	if (vp->v_type == VDIR)
2646 		error = EISDIR;
2647 	else if ((error = vn_writechk(vp)) == 0 &&
2648 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2649 		VATTR_NULL(&vattr);
2650 		vattr.va_size = SCARG(uap, length);
2651 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2652 	}
2653 	vput(vp);
2654 	return (error);
2655 }
2656 
2657 /*
2658  * Truncate a file given a file descriptor.
2659  */
2660 /* ARGSUSED */
2661 int
2662 sys_ftruncate(p, v, retval)
2663 	struct proc *p;
2664 	void *v;
2665 	register_t *retval;
2666 {
2667 	struct sys_ftruncate_args /* {
2668 		syscallarg(int) fd;
2669 		syscallarg(int) pad;
2670 		syscallarg(off_t) length;
2671 	} */ *uap = v;
2672 	struct vattr vattr;
2673 	struct vnode *vp;
2674 	struct file *fp;
2675 	int error;
2676 
2677 	/* getvnode() will use the descriptor for us */
2678 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2679 		return (error);
2680 	if ((fp->f_flag & FWRITE) == 0) {
2681 		error = EINVAL;
2682 		goto out;
2683 	}
2684 	vp = (struct vnode *)fp->f_data;
2685 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2686 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2687 	if (vp->v_type == VDIR)
2688 		error = EISDIR;
2689 	else if ((error = vn_writechk(vp)) == 0) {
2690 		VATTR_NULL(&vattr);
2691 		vattr.va_size = SCARG(uap, length);
2692 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2693 	}
2694 	VOP_UNLOCK(vp, 0);
2695  out:
2696 	FILE_UNUSE(fp, p);
2697 	return (error);
2698 }
2699 
2700 /*
2701  * Sync an open file.
2702  */
2703 /* ARGSUSED */
2704 int
2705 sys_fsync(p, v, retval)
2706 	struct proc *p;
2707 	void *v;
2708 	register_t *retval;
2709 {
2710 	struct sys_fsync_args /* {
2711 		syscallarg(int) fd;
2712 	} */ *uap = v;
2713 	struct vnode *vp;
2714 	struct file *fp;
2715 	int error;
2716 
2717 	/* getvnode() will use the descriptor for us */
2718 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2719 		return (error);
2720 	vp = (struct vnode *)fp->f_data;
2721 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2722 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, p);
2723 	if (error == 0 && bioops.io_fsync != NULL &&
2724 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2725 		(*bioops.io_fsync)(vp);
2726 	VOP_UNLOCK(vp, 0);
2727 	FILE_UNUSE(fp, p);
2728 	return (error);
2729 }
2730 
2731 /*
2732  * Sync the data of an open file.
2733  */
2734 /* ARGSUSED */
2735 int
2736 sys_fdatasync(p, v, retval)
2737 	struct proc *p;
2738 	void *v;
2739 	register_t *retval;
2740 {
2741 	struct sys_fdatasync_args /* {
2742 		syscallarg(int) fd;
2743 	} */ *uap = v;
2744 	struct vnode *vp;
2745 	struct file *fp;
2746 	int error;
2747 
2748 	/* getvnode() will use the descriptor for us */
2749 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2750 		return (error);
2751 	vp = (struct vnode *)fp->f_data;
2752 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2753 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, p);
2754 	VOP_UNLOCK(vp, 0);
2755 	FILE_UNUSE(fp, p);
2756 	return (error);
2757 }
2758 
2759 /*
2760  * Rename files, (standard) BSD semantics frontend.
2761  */
2762 /* ARGSUSED */
2763 int
2764 sys_rename(p, v, retval)
2765 	struct proc *p;
2766 	void *v;
2767 	register_t *retval;
2768 {
2769 	struct sys_rename_args /* {
2770 		syscallarg(const char *) from;
2771 		syscallarg(const char *) to;
2772 	} */ *uap = v;
2773 
2774 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2775 }
2776 
2777 /*
2778  * Rename files, POSIX semantics frontend.
2779  */
2780 /* ARGSUSED */
2781 int
2782 sys___posix_rename(p, v, retval)
2783 	struct proc *p;
2784 	void *v;
2785 	register_t *retval;
2786 {
2787 	struct sys___posix_rename_args /* {
2788 		syscallarg(const char *) from;
2789 		syscallarg(const char *) to;
2790 	} */ *uap = v;
2791 
2792 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2793 }
2794 
2795 /*
2796  * Rename files.  Source and destination must either both be directories,
2797  * or both not be directories.  If target is a directory, it must be empty.
2798  * If `from' and `to' refer to the same object, the value of the `retain'
2799  * argument is used to determine whether `from' will be
2800  *
2801  * (retain == 0)	deleted unless `from' and `to' refer to the same
2802  *			object in the file system's name space (BSD).
2803  * (retain == 1)	always retained (POSIX).
2804  */
2805 static int
2806 rename_files(from, to, p, retain)
2807 	const char *from, *to;
2808 	struct proc *p;
2809 	int retain;
2810 {
2811 	struct vnode *tvp, *fvp, *tdvp;
2812 	struct nameidata fromnd, tond;
2813 	int error;
2814 
2815 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2816 	    from, p);
2817 	if ((error = namei(&fromnd)) != 0)
2818 		return (error);
2819 	fvp = fromnd.ni_vp;
2820 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2821 	    UIO_USERSPACE, to, p);
2822 	if ((error = namei(&tond)) != 0) {
2823 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2824 		vrele(fromnd.ni_dvp);
2825 		vrele(fvp);
2826 		goto out1;
2827 	}
2828 	tdvp = tond.ni_dvp;
2829 	tvp = tond.ni_vp;
2830 
2831 	if (tvp != NULL) {
2832 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2833 			error = ENOTDIR;
2834 			goto out;
2835 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2836 			error = EISDIR;
2837 			goto out;
2838 		}
2839 	}
2840 
2841 	if (fvp == tdvp)
2842 		error = EINVAL;
2843 
2844 	/*
2845 	 * Source and destination refer to the same object.
2846 	 */
2847 	if (fvp == tvp) {
2848 		if (retain)
2849 			error = -1;
2850 		else if (fromnd.ni_dvp == tdvp &&
2851 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2852 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
2853 		          tond.ni_cnd.cn_nameptr,
2854 		          fromnd.ni_cnd.cn_namelen))
2855 		error = -1;
2856 	}
2857 
2858 out:
2859 	if (!error) {
2860 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2861 		if (fromnd.ni_dvp != tdvp)
2862 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2863 		if (tvp) {
2864 			(void)uvm_vnp_uncache(tvp);
2865 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2866 		}
2867 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2868 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2869 	} else {
2870 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2871 		if (tdvp == tvp)
2872 			vrele(tdvp);
2873 		else
2874 			vput(tdvp);
2875 		if (tvp)
2876 			vput(tvp);
2877 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2878 		vrele(fromnd.ni_dvp);
2879 		vrele(fvp);
2880 	}
2881 	vrele(tond.ni_startdir);
2882 	FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
2883 out1:
2884 	if (fromnd.ni_startdir)
2885 		vrele(fromnd.ni_startdir);
2886 	FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
2887 	return (error == -1 ? 0 : error);
2888 }
2889 
2890 /*
2891  * Make a directory file.
2892  */
2893 /* ARGSUSED */
2894 int
2895 sys_mkdir(p, v, retval)
2896 	struct proc *p;
2897 	void *v;
2898 	register_t *retval;
2899 {
2900 	struct sys_mkdir_args /* {
2901 		syscallarg(const char *) path;
2902 		syscallarg(int) mode;
2903 	} */ *uap = v;
2904 	struct vnode *vp;
2905 	struct vattr vattr;
2906 	int error;
2907 	struct nameidata nd;
2908 
2909 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2910 	if ((error = namei(&nd)) != 0)
2911 		return (error);
2912 	vp = nd.ni_vp;
2913 	if (vp != NULL) {
2914 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2915 		if (nd.ni_dvp == vp)
2916 			vrele(nd.ni_dvp);
2917 		else
2918 			vput(nd.ni_dvp);
2919 		vrele(vp);
2920 		return (EEXIST);
2921 	}
2922 	VATTR_NULL(&vattr);
2923 	vattr.va_type = VDIR;
2924 	vattr.va_mode =
2925 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2926 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2927 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2928 	if (!error)
2929 		vput(nd.ni_vp);
2930 	return (error);
2931 }
2932 
2933 /*
2934  * Remove a directory file.
2935  */
2936 /* ARGSUSED */
2937 int
2938 sys_rmdir(p, v, retval)
2939 	struct proc *p;
2940 	void *v;
2941 	register_t *retval;
2942 {
2943 	struct sys_rmdir_args /* {
2944 		syscallarg(const char *) path;
2945 	} */ *uap = v;
2946 	struct vnode *vp;
2947 	int error;
2948 	struct nameidata nd;
2949 
2950 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2951 	    SCARG(uap, path), p);
2952 	if ((error = namei(&nd)) != 0)
2953 		return (error);
2954 	vp = nd.ni_vp;
2955 	if (vp->v_type != VDIR) {
2956 		error = ENOTDIR;
2957 		goto out;
2958 	}
2959 	/*
2960 	 * No rmdir "." please.
2961 	 */
2962 	if (nd.ni_dvp == vp) {
2963 		error = EINVAL;
2964 		goto out;
2965 	}
2966 	/*
2967 	 * The root of a mounted filesystem cannot be deleted.
2968 	 */
2969 	if (vp->v_flag & VROOT)
2970 		error = EBUSY;
2971 out:
2972 	if (!error) {
2973 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2974 		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2975 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2976 	} else {
2977 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2978 		if (nd.ni_dvp == vp)
2979 			vrele(nd.ni_dvp);
2980 		else
2981 			vput(nd.ni_dvp);
2982 		vput(vp);
2983 	}
2984 	return (error);
2985 }
2986 
2987 /*
2988  * Read a block of directory entries in a file system independent format.
2989  */
2990 int
2991 sys_getdents(p, v, retval)
2992 	struct proc *p;
2993 	void *v;
2994 	register_t *retval;
2995 {
2996 	struct sys_getdents_args /* {
2997 		syscallarg(int) fd;
2998 		syscallarg(char *) buf;
2999 		syscallarg(size_t) count;
3000 	} */ *uap = v;
3001 	struct file *fp;
3002 	int error, done;
3003 
3004 	/* getvnode() will use the descriptor for us */
3005 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3006 		return (error);
3007 	if ((fp->f_flag & FREAD) == 0) {
3008 		error = EBADF;
3009 		goto out;
3010 	}
3011 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3012 			SCARG(uap, count), &done, p, 0, 0);
3013 	*retval = done;
3014  out:
3015 	FILE_UNUSE(fp, p);
3016 	return (error);
3017 }
3018 
3019 /*
3020  * Set the mode mask for creation of filesystem nodes.
3021  */
3022 int
3023 sys_umask(p, v, retval)
3024 	struct proc *p;
3025 	void *v;
3026 	register_t *retval;
3027 {
3028 	struct sys_umask_args /* {
3029 		syscallarg(mode_t) newmask;
3030 	} */ *uap = v;
3031 	struct cwdinfo *cwdi;
3032 
3033 	cwdi = p->p_cwdi;
3034 	*retval = cwdi->cwdi_cmask;
3035 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3036 	return (0);
3037 }
3038 
3039 /*
3040  * Void all references to file by ripping underlying filesystem
3041  * away from vnode.
3042  */
3043 /* ARGSUSED */
3044 int
3045 sys_revoke(p, v, retval)
3046 	struct proc *p;
3047 	void *v;
3048 	register_t *retval;
3049 {
3050 	struct sys_revoke_args /* {
3051 		syscallarg(const char *) path;
3052 	} */ *uap = v;
3053 	struct vnode *vp;
3054 	struct vattr vattr;
3055 	int error;
3056 	struct nameidata nd;
3057 
3058 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3059 	if ((error = namei(&nd)) != 0)
3060 		return (error);
3061 	vp = nd.ni_vp;
3062 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3063 		goto out;
3064 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3065 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3066 		goto out;
3067 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3068 		VOP_REVOKE(vp, REVOKEALL);
3069 out:
3070 	vrele(vp);
3071 	return (error);
3072 }
3073 
3074 /*
3075  * Convert a user file descriptor to a kernel file entry.
3076  */
3077 int
3078 getvnode(fdp, fd, fpp)
3079 	struct filedesc *fdp;
3080 	int fd;
3081 	struct file **fpp;
3082 {
3083 	struct vnode *vp;
3084 	struct file *fp;
3085 
3086 	if ((u_int)fd >= fdp->fd_nfiles ||
3087 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
3088 	    (fp->f_iflags & FIF_WANTCLOSE) != 0)
3089 		return (EBADF);
3090 
3091 	FILE_USE(fp);
3092 
3093 	if (fp->f_type != DTYPE_VNODE) {
3094 		FILE_UNUSE(fp, NULL);
3095 		return (EINVAL);
3096 	}
3097 
3098 	vp = (struct vnode *)fp->f_data;
3099 	if (vp->v_type == VBAD) {
3100 		FILE_UNUSE(fp, NULL);
3101 		return (EBADF);
3102 	}
3103 
3104 	*fpp = fp;
3105 	return (0);
3106 }
3107