xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 1c9b56c830954ccf3b57004ac65562e3d6afacf6)
1 /*	$NetBSD: vfs_syscalls.c,v 1.216 2005/01/25 23:55:20 wrstuden Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.216 2005/01/25 23:55:20 wrstuden Exp $");
41 
42 #include "opt_compat_netbsd.h"
43 #include "opt_compat_43.h"
44 #include "opt_ktrace.h"
45 #include "fss.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/namei.h>
50 #include <sys/filedesc.h>
51 #include <sys/kernel.h>
52 #include <sys/file.h>
53 #include <sys/stat.h>
54 #include <sys/vnode.h>
55 #include <sys/mount.h>
56 #include <sys/proc.h>
57 #include <sys/uio.h>
58 #include <sys/malloc.h>
59 #include <sys/dirent.h>
60 #include <sys/extattr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sa.h>
63 #include <sys/syscallargs.h>
64 #ifdef KTRACE
65 #include <sys/ktrace.h>
66 #endif
67 
68 #include <miscfs/genfs/genfs.h>
69 #include <miscfs/syncfs/syncfs.h>
70 
71 #if NFSS > 0
72 #include <dev/fssvar.h>
73 #endif
74 
75 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
76 
77 static int change_dir(struct nameidata *, struct proc *);
78 static int change_flags(struct vnode *, u_long, struct proc *);
79 static int change_mode(struct vnode *, int, struct proc *p);
80 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int);
81 static int change_utimes(struct vnode *vp, const struct timeval *,
82 	       struct proc *p);
83 static int rename_files(const char *, const char *, struct proc *, int);
84 
85 void checkdirs(struct vnode *);
86 
87 int dovfsusermount = 0;
88 
89 /*
90  * Virtual File System System Calls
91  */
92 
93 /*
94  * Mount a file system.
95  */
96 
97 #if defined(COMPAT_09) || defined(COMPAT_43)
98 /*
99  * This table is used to maintain compatibility with 4.3BSD
100  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
101  *
102  * Do not modify this table. It should only contain filesystems
103  * supported by NetBSD 0.9 and 4.3BSD.
104  */
105 const char * const mountcompatnames[] = {
106 	NULL,		/* 0 = MOUNT_NONE */
107 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
108 	MOUNT_NFS,	/* 2 */
109 	MOUNT_MFS,	/* 3 */
110 	MOUNT_MSDOS,	/* 4 */
111 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
112 	MOUNT_FDESC,	/* 6 */
113 	MOUNT_KERNFS,	/* 7 */
114 	NULL,		/* 8 = MOUNT_DEVFS */
115 	MOUNT_AFS,	/* 9 */
116 };
117 const int nmountcompatnames = sizeof(mountcompatnames) /
118     sizeof(mountcompatnames[0]);
119 #endif /* COMPAT_09 || COMPAT_43 */
120 
121 /* ARGSUSED */
122 int
123 sys_mount(l, v, retval)
124 	struct lwp *l;
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 proc *p = l->l_proc;
135 	struct vnode *vp;
136 	struct mount *mp;
137 	int error, flag = 0;
138 	char fstypename[MFSNAMELEN];
139 	struct vattr va;
140 	struct nameidata nd;
141 	struct vfsops *vfs;
142 
143 	if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
144 	    (error = suser(p->p_ucred, &p->p_acflag)))
145 		return (error);
146 	/*
147 	 * Get vnode to be covered
148 	 */
149 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
150 	    SCARG(uap, path), p);
151 	if ((error = namei(&nd)) != 0)
152 		return (error);
153 	vp = nd.ni_vp;
154 	/*
155 	 * A lookup in VFS_MOUNT might result in an attempt to
156 	 * lock this vnode again, so make the lock recursive.
157 	 */
158 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
159 	if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
160 		if ((vp->v_flag & VROOT) == 0) {
161 			vput(vp);
162 			return (EINVAL);
163 		}
164 		mp = vp->v_mount;
165 		flag = mp->mnt_flag;
166 		vfs = mp->mnt_op;
167 		/*
168 		 * We only allow the filesystem to be reloaded if it
169 		 * is currently mounted read-only.
170 		 */
171 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
172 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
173 			vput(vp);
174 			return (EOPNOTSUPP);	/* Needs translation */
175 		}
176 		/*
177 		 * In "highly secure" mode, don't let the caller do anything
178 		 * but downgrade a filesystem from read-write to read-only.
179 		 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
180 		 */
181 		if (securelevel >= 2 &&
182 		    SCARG(uap, flags) != MNT_GETARGS &&
183 		    SCARG(uap, flags) !=
184 		    (mp->mnt_flag | MNT_RDONLY |
185 		     MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
186 			vput(vp);
187 			return (EPERM);
188 		}
189 		mp->mnt_flag |= SCARG(uap, flags) &
190 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
191 		/*
192 		 * Only root, or the user that did the original mount is
193 		 * permitted to update it.
194 		 */
195 		if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
196 		    mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
197 		    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
198 			vput(vp);
199 			return (error);
200 		}
201 		/*
202 		 * Do not allow NFS export by non-root users. For non-root
203 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
204 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
205 		 */
206 		if (p->p_ucred->cr_uid != 0) {
207 			if (SCARG(uap, flags) & MNT_EXPORTED) {
208 				vput(vp);
209 				return (EPERM);
210 			}
211 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
212 			if (flag & MNT_NOEXEC)
213 				SCARG(uap, flags) |= MNT_NOEXEC;
214 		}
215 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
216 			vput(vp);
217 			return (EPERM);
218 		}
219 		goto update;
220 	} else {
221 		if (securelevel >= 2) {
222 			vput(vp);
223 			return (EPERM);
224 		}
225 	}
226 	/*
227 	 * If the user is not root, ensure that they own the directory
228 	 * onto which we are attempting to mount.
229 	 */
230 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
231 	    (va.va_uid != p->p_ucred->cr_uid &&
232 		(error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
233 		vput(vp);
234 		return (error);
235 	}
236 	/*
237 	 * Do not allow NFS export by non-root users. For non-root users,
238 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
239 	 * mount point is already MNT_NOEXEC.
240 	 */
241 	if (p->p_ucred->cr_uid != 0) {
242 		if (SCARG(uap, flags) & MNT_EXPORTED) {
243 			vput(vp);
244 			return (EPERM);
245 		}
246 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
247 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
248 			SCARG(uap, flags) |= MNT_NOEXEC;
249 	}
250 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) {
251 		vput(vp);
252 		return (error);
253 	}
254 	if (vp->v_type != VDIR) {
255 		vput(vp);
256 		return (ENOTDIR);
257 	}
258 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
259 	if (error) {
260 #if defined(COMPAT_09) || defined(COMPAT_43)
261 		/*
262 		 * Historically filesystem types were identified by number.
263 		 * If we get an integer for the filesystem type instead of a
264 		 * string, we check to see if it matches one of the historic
265 		 * filesystem types.
266 		 */
267 		u_long fsindex = (u_long)SCARG(uap, type);
268 		if (fsindex >= nmountcompatnames ||
269 		    mountcompatnames[fsindex] == NULL) {
270 			vput(vp);
271 			return (ENODEV);
272 		}
273 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
274 #else
275 		vput(vp);
276 		return (error);
277 #endif
278 	}
279 #ifdef	COMPAT_10
280 	/* Accept `ufs' as an alias for `ffs'. */
281 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
282 		strncpy(fstypename, "ffs", MFSNAMELEN);
283 #endif
284 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
285 		vput(vp);
286 		return (ENODEV);
287 	}
288 	if (vp->v_mountedhere != NULL) {
289 		vput(vp);
290 		return (EBUSY);
291 	}
292 
293 	/*
294 	 * Allocate and initialize the file system.
295 	 */
296 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
297 		M_MOUNT, M_WAITOK);
298 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
299 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
300 	simple_lock_init(&mp->mnt_slock);
301 	(void)vfs_busy(mp, LK_NOWAIT, 0);
302 	mp->mnt_op = vfs;
303 	vfs->vfs_refcount++;
304 	mp->mnt_vnodecovered = vp;
305 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
306 	mp->mnt_unmounter = NULL;
307 	mp->mnt_leaf = mp;
308 
309 	/*
310 	 * The underlying file system may refuse the mount for
311 	 * various reasons.  Allow the user to force it to happen.
312 	 */
313 	mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE;
314  update:
315 	/*
316 	 * Set the mount level flags.
317 	 */
318 	if (SCARG(uap, flags) & MNT_RDONLY)
319 		mp->mnt_flag |= MNT_RDONLY;
320 	else if (mp->mnt_flag & MNT_RDONLY)
321 		mp->mnt_iflag |= IMNT_WANTRDWR;
322 	mp->mnt_flag &=
323 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
324 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
325 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
326 	mp->mnt_flag |= SCARG(uap, flags) &
327 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
328 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
329 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
330 	    MNT_IGNORE);
331 	/*
332 	 * Mount the filesystem.
333 	 */
334 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
335 	if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
336 		if (mp->mnt_iflag & IMNT_WANTRDWR)
337 			mp->mnt_flag &= ~MNT_RDONLY;
338 		if (error || (mp->mnt_flag & MNT_GETARGS))
339 			mp->mnt_flag = flag;
340 		mp->mnt_flag &=~
341 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
342 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
343 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
344 			if (mp->mnt_syncer == NULL)
345 				error = vfs_allocate_syncvnode(mp);
346 		} else {
347 			if (mp->mnt_syncer != NULL)
348 				vfs_deallocate_syncvnode(mp);
349 		}
350 		vfs_unbusy(mp);
351 		VOP_UNLOCK(vp, 0);
352 		vrele(vp);
353 		return (error);
354 	}
355 	/*
356 	 * Put the new filesystem on the mount list after root.
357 	 */
358 	cache_purge(vp);
359 	if (!error) {
360 		mp->mnt_flag &=~
361 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
362 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
363 		vp->v_mountedhere = mp;
364 		simple_lock(&mountlist_slock);
365 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
366 		simple_unlock(&mountlist_slock);
367 		checkdirs(vp);
368 		VOP_UNLOCK(vp, 0);
369 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
370 			error = vfs_allocate_syncvnode(mp);
371 		vfs_unbusy(mp);
372 		(void) VFS_STATVFS(mp, &mp->mnt_stat, p);
373 		if ((error = VFS_START(mp, 0, p)))
374 			vrele(vp);
375 	} else {
376 		vp->v_mountedhere = (struct mount *)0;
377 		vfs->vfs_refcount--;
378 		vfs_unbusy(mp);
379 		free(mp, M_MOUNT);
380 		vput(vp);
381 	}
382 	return (error);
383 }
384 
385 /*
386  * Scan all active processes to see if any of them have a current
387  * or root directory onto which the new filesystem has just been
388  * mounted. If so, replace them with the new mount point.
389  */
390 void
391 checkdirs(olddp)
392 	struct vnode *olddp;
393 {
394 	struct cwdinfo *cwdi;
395 	struct vnode *newdp;
396 	struct proc *p;
397 
398 	if (olddp->v_usecount == 1)
399 		return;
400 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
401 		panic("mount: lost mount");
402 	proclist_lock_read();
403 	PROCLIST_FOREACH(p, &allproc) {
404 		cwdi = p->p_cwdi;
405 		if (!cwdi)
406 			continue;
407 		if (cwdi->cwdi_cdir == olddp) {
408 			vrele(cwdi->cwdi_cdir);
409 			VREF(newdp);
410 			cwdi->cwdi_cdir = newdp;
411 		}
412 		if (cwdi->cwdi_rdir == olddp) {
413 			vrele(cwdi->cwdi_rdir);
414 			VREF(newdp);
415 			cwdi->cwdi_rdir = newdp;
416 		}
417 	}
418 	proclist_unlock_read();
419 	if (rootvnode == olddp) {
420 		vrele(rootvnode);
421 		VREF(newdp);
422 		rootvnode = newdp;
423 	}
424 	vput(newdp);
425 }
426 
427 /*
428  * Unmount a file system.
429  *
430  * Note: unmount takes a path to the vnode mounted on as argument,
431  * not special file (as before).
432  */
433 /* ARGSUSED */
434 int
435 sys_unmount(l, v, retval)
436 	struct lwp *l;
437 	void *v;
438 	register_t *retval;
439 {
440 	struct sys_unmount_args /* {
441 		syscallarg(const char *) path;
442 		syscallarg(int) flags;
443 	} */ *uap = v;
444 	struct proc *p = l->l_proc;
445 	struct vnode *vp;
446 	struct mount *mp;
447 	int error;
448 	struct nameidata nd;
449 
450 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
451 	    SCARG(uap, path), p);
452 	if ((error = namei(&nd)) != 0)
453 		return (error);
454 	vp = nd.ni_vp;
455 	mp = vp->v_mount;
456 
457 	/*
458 	 * Only root, or the user that did the original mount is
459 	 * permitted to unmount this filesystem.
460 	 */
461 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
462 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
463 		vput(vp);
464 		return (error);
465 	}
466 
467 	/*
468 	 * Don't allow unmounting the root file system.
469 	 */
470 	if (mp->mnt_flag & MNT_ROOTFS) {
471 		vput(vp);
472 		return (EINVAL);
473 	}
474 
475 	/*
476 	 * Must be the root of the filesystem
477 	 */
478 	if ((vp->v_flag & VROOT) == 0) {
479 		vput(vp);
480 		return (EINVAL);
481 	}
482 	vput(vp);
483 
484 	/*
485 	 * XXX Freeze syncer.  Must do this before locking the
486 	 * mount point.  See dounmount() for details.
487 	 */
488 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
489 
490 	if (vfs_busy(mp, 0, 0)) {
491 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
492 		return (EBUSY);
493 	}
494 
495 	return (dounmount(mp, SCARG(uap, flags), p));
496 }
497 
498 /*
499  * Do the actual file system unmount. File system is assumed to have been
500  * marked busy by the caller.
501  */
502 int
503 dounmount(mp, flags, p)
504 	struct mount *mp;
505 	int flags;
506 	struct proc *p;
507 {
508 	struct vnode *coveredvp;
509 	int error;
510 	int async;
511 	int used_syncer;
512 
513 	simple_lock(&mountlist_slock);
514 	vfs_unbusy(mp);
515 	used_syncer = (mp->mnt_syncer != NULL);
516 
517 	/*
518 	 * XXX Syncer must be frozen when we get here.  This should really
519 	 * be done on a per-mountpoint basis, but especially the softdep
520 	 * code possibly called from the syncer doens't exactly work on a
521 	 * per-mountpoint basis, so the softdep code would become a maze
522 	 * of vfs_busy() calls.
523 	 *
524 	 * The caller of dounmount() must acquire syncer_lock because
525 	 * the syncer itself acquires locks in syncer_lock -> vfs_busy
526 	 * order, and we must preserve that order to avoid deadlock.
527 	 *
528 	 * So, if the file system did not use the syncer, now is
529 	 * the time to release the syncer_lock.
530 	 */
531 	if (used_syncer == 0)
532 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
533 
534 	mp->mnt_iflag |= IMNT_UNMOUNT;
535 	mp->mnt_unmounter = p;
536 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
537 	vn_start_write(NULL, &mp, V_WAIT);
538 
539 	if (mp->mnt_flag & MNT_EXPUBLIC)
540 		vfs_setpublicfs(NULL, NULL, NULL);
541 	async = mp->mnt_flag & MNT_ASYNC;
542 	mp->mnt_flag &= ~MNT_ASYNC;
543 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
544 	if (mp->mnt_syncer != NULL)
545 		vfs_deallocate_syncvnode(mp);
546 	error = 0;
547 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
548 #if NFSS > 0
549 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
550 #endif
551 		if (error == 0)
552 			error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
553 	}
554 	if (error == 0 || (flags & MNT_FORCE))
555 		error = VFS_UNMOUNT(mp, flags, p);
556 	vn_finished_write(mp, 0);
557 	simple_lock(&mountlist_slock);
558 	if (error) {
559 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
560 			(void) vfs_allocate_syncvnode(mp);
561 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
562 		mp->mnt_unmounter = NULL;
563 		mp->mnt_flag |= async;
564 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
565 		    &mountlist_slock);
566 		if (used_syncer)
567 			lockmgr(&syncer_lock, LK_RELEASE, NULL);
568 		simple_lock(&mp->mnt_slock);
569 		while (mp->mnt_wcnt > 0) {
570 			wakeup(mp);
571 			ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
572 				0, &mp->mnt_slock);
573 		}
574 		simple_unlock(&mp->mnt_slock);
575 		return (error);
576 	}
577 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
578 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
579 		coveredvp->v_mountedhere = NULL;
580 		vrele(coveredvp);
581 	}
582 	mp->mnt_op->vfs_refcount--;
583 	if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
584 		panic("unmount: dangling vnode");
585 	mp->mnt_iflag |= IMNT_GONE;
586 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
587 	if (used_syncer)
588 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
589 	simple_lock(&mp->mnt_slock);
590 	while (mp->mnt_wcnt > 0) {
591 		wakeup(mp);
592 		ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
593 	}
594 	simple_unlock(&mp->mnt_slock);
595 	free(mp, M_MOUNT);
596 	return (0);
597 }
598 
599 /*
600  * Sync each mounted filesystem.
601  */
602 #ifdef DEBUG
603 int syncprt = 0;
604 struct ctldebug debug0 = { "syncprt", &syncprt };
605 #endif
606 
607 /* ARGSUSED */
608 int
609 sys_sync(l, v, retval)
610 	struct lwp *l;
611 	void *v;
612 	register_t *retval;
613 {
614 	struct mount *mp, *nmp;
615 	int asyncflag;
616 	struct proc *p = l == NULL ? &proc0 : l->l_proc;
617 
618 	simple_lock(&mountlist_slock);
619 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
620 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
621 			nmp = mp->mnt_list.cqe_prev;
622 			continue;
623 		}
624 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
625 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
626 			asyncflag = mp->mnt_flag & MNT_ASYNC;
627 			mp->mnt_flag &= ~MNT_ASYNC;
628 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
629 			if (asyncflag)
630 				 mp->mnt_flag |= MNT_ASYNC;
631 			vn_finished_write(mp, 0);
632 		}
633 		simple_lock(&mountlist_slock);
634 		nmp = mp->mnt_list.cqe_prev;
635 		vfs_unbusy(mp);
636 
637 	}
638 	simple_unlock(&mountlist_slock);
639 #ifdef DEBUG
640 	if (syncprt)
641 		vfs_bufstats();
642 #endif /* DEBUG */
643 	return (0);
644 }
645 
646 /*
647  * Change filesystem quotas.
648  */
649 /* ARGSUSED */
650 int
651 sys_quotactl(l, v, retval)
652 	struct lwp *l;
653 	void *v;
654 	register_t *retval;
655 {
656 	struct sys_quotactl_args /* {
657 		syscallarg(const char *) path;
658 		syscallarg(int) cmd;
659 		syscallarg(int) uid;
660 		syscallarg(caddr_t) arg;
661 	} */ *uap = v;
662 	struct proc *p = l->l_proc;
663 	struct mount *mp;
664 	int error;
665 	struct nameidata nd;
666 
667 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
668 	if ((error = namei(&nd)) != 0)
669 		return (error);
670 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
671 	vrele(nd.ni_vp);
672 	if (error)
673 		return (error);
674 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
675 	    SCARG(uap, arg), p);
676 	vn_finished_write(mp, 0);
677 	return (error);
678 }
679 
680 int
681 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags,
682     int root)
683 {
684 	struct cwdinfo *cwdi = p->p_cwdi;
685 	int error = 0;
686 
687 	/*
688 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
689 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
690 	 * overrides MNT_NOWAIT.
691 	 */
692 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
693 	    (flags != MNT_WAIT && flags != 0)) {
694 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
695 		goto done;
696 	}
697 
698 	/* Get the filesystem stats now */
699 	memset(sp, 0, sizeof(*sp));
700 	if ((error = VFS_STATVFS(mp, sp, p)) != 0) {
701 		return error;
702 	}
703 
704 	if (cwdi->cwdi_rdir == NULL)
705 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
706 done:
707 	if (cwdi->cwdi_rdir != NULL) {
708 		size_t len;
709 		char *bp;
710 		char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
711 		if (!path)
712 			return ENOMEM;
713 
714 		bp = path + MAXPATHLEN;
715 		*--bp = '\0';
716 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
717 		    MAXPATHLEN / 2, 0, p);
718 		if (error) {
719 			free(path, M_TEMP);
720 			return error;
721 		}
722 		len = strlen(bp);
723 		/*
724 		 * for mount points that are below our root, we can see
725 		 * them, so we fix up the pathname and return them. The
726 		 * rest we cannot see, so we don't allow viewing the
727 		 * data.
728 		 */
729 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
730 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
731 			    sizeof(sp->f_mntonname));
732 			if (sp->f_mntonname[0] == '\0')
733 				(void)strlcpy(sp->f_mntonname, "/",
734 				    sizeof(sp->f_mntonname));
735 		} else {
736 			if (root)
737 				(void)strlcpy(sp->f_mntonname, "/",
738 				    sizeof(sp->f_mntonname));
739 			else
740 				error = EPERM;
741 		}
742 		free(path, M_TEMP);
743 	}
744 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
745 	return error;
746 }
747 
748 /*
749  * Get filesystem statistics.
750  */
751 /* ARGSUSED */
752 int
753 sys_statvfs1(l, v, retval)
754 	struct lwp *l;
755 	void *v;
756 	register_t *retval;
757 {
758 	struct sys_statvfs1_args /* {
759 		syscallarg(const char *) path;
760 		syscallarg(struct statvfs *) buf;
761 		syscallarg(int) flags;
762 	} */ *uap = v;
763 	struct proc *p = l->l_proc;
764 	struct mount *mp;
765 	struct statvfs sbuf;
766 	int error;
767 	struct nameidata nd;
768 
769 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
770 	if ((error = namei(&nd)) != 0)
771 		return error;
772 	mp = nd.ni_vp->v_mount;
773 	vrele(nd.ni_vp);
774 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
775 		return error;
776 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
777 }
778 
779 /*
780  * Get filesystem statistics.
781  */
782 /* ARGSUSED */
783 int
784 sys_fstatvfs1(l, v, retval)
785 	struct lwp *l;
786 	void *v;
787 	register_t *retval;
788 {
789 	struct sys_fstatvfs1_args /* {
790 		syscallarg(int) fd;
791 		syscallarg(struct statvfs *) buf;
792 		syscallarg(int) flags;
793 	} */ *uap = v;
794 	struct proc *p = l->l_proc;
795 	struct file *fp;
796 	struct mount *mp;
797 	struct statvfs sbuf;
798 	int error;
799 
800 	/* getvnode() will use the descriptor for us */
801 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
802 		return (error);
803 	mp = ((struct vnode *)fp->f_data)->v_mount;
804 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
805 		goto out;
806 	error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
807  out:
808 	FILE_UNUSE(fp, p);
809 	return error;
810 }
811 
812 
813 /*
814  * Get statistics on all filesystems.
815  */
816 int
817 sys_getvfsstat(l, v, retval)
818 	struct lwp *l;
819 	void *v;
820 	register_t *retval;
821 {
822 	struct sys_getvfsstat_args /* {
823 		syscallarg(struct statvfs *) buf;
824 		syscallarg(size_t) bufsize;
825 		syscallarg(int) flags;
826 	} */ *uap = v;
827 	int root = 0;
828 	struct proc *p = l->l_proc;
829 	struct mount *mp, *nmp;
830 	struct statvfs sbuf;
831 	struct statvfs *sfsp;
832 	size_t count, maxcount;
833 	int error = 0;
834 
835 	maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs);
836 	sfsp = SCARG(uap, buf);
837 	simple_lock(&mountlist_slock);
838 	count = 0;
839 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
840 	     mp = nmp) {
841 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
842 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
843 			continue;
844 		}
845 		if (sfsp && count < maxcount) {
846 			error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0);
847 			if (error) {
848 				simple_lock(&mountlist_slock);
849 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
850 				vfs_unbusy(mp);
851 				continue;
852 			}
853 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
854 			if (error) {
855 				vfs_unbusy(mp);
856 				return (error);
857 			}
858 			sfsp++;
859 			root |= strcmp(sbuf.f_mntonname, "/") == 0;
860 		}
861 		count++;
862 		simple_lock(&mountlist_slock);
863 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
864 		vfs_unbusy(mp);
865 	}
866 	simple_unlock(&mountlist_slock);
867 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
868 		/*
869 		 * fake a root entry
870 		 */
871 		if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p,
872 		    SCARG(uap, flags), 1)) != 0)
873 			return error;
874 		if (sfsp)
875 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
876 		count++;
877 	}
878 	if (sfsp && count > maxcount)
879 		*retval = maxcount;
880 	else
881 		*retval = count;
882 	return error;
883 }
884 
885 /*
886  * Change current working directory to a given file descriptor.
887  */
888 /* ARGSUSED */
889 int
890 sys_fchdir(l, v, retval)
891 	struct lwp *l;
892 	void *v;
893 	register_t *retval;
894 {
895 	struct sys_fchdir_args /* {
896 		syscallarg(int) fd;
897 	} */ *uap = v;
898 	struct proc *p = l->l_proc;
899 	struct filedesc *fdp = p->p_fd;
900 	struct cwdinfo *cwdi = p->p_cwdi;
901 	struct vnode *vp, *tdp;
902 	struct mount *mp;
903 	struct file *fp;
904 	int error;
905 
906 	/* getvnode() will use the descriptor for us */
907 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
908 		return (error);
909 	vp = (struct vnode *)fp->f_data;
910 
911 	VREF(vp);
912 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
913 	if (vp->v_type != VDIR)
914 		error = ENOTDIR;
915 	else
916 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
917 	while (!error && (mp = vp->v_mountedhere) != NULL) {
918 		if (vfs_busy(mp, 0, 0))
919 			continue;
920 		error = VFS_ROOT(mp, &tdp);
921 		vfs_unbusy(mp);
922 		if (error)
923 			break;
924 		vput(vp);
925 		vp = tdp;
926 	}
927 	if (error) {
928 		vput(vp);
929 		goto out;
930 	}
931 	VOP_UNLOCK(vp, 0);
932 
933 	/*
934 	 * Disallow changing to a directory not under the process's
935 	 * current root directory (if there is one).
936 	 */
937 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
938 		vrele(vp);
939 		error = EPERM;	/* operation not permitted */
940 		goto out;
941 	}
942 
943 	vrele(cwdi->cwdi_cdir);
944 	cwdi->cwdi_cdir = vp;
945  out:
946 	FILE_UNUSE(fp, p);
947 	return (error);
948 }
949 
950 /*
951  * Change this process's notion of the root directory to a given file descriptor.
952  */
953 
954 int
955 sys_fchroot(l, v, retval)
956 	struct lwp *l;
957 	void *v;
958 	register_t *retval;
959 {
960 	struct sys_fchroot_args *uap = v;
961 	struct proc *p = l->l_proc;
962 	struct filedesc *fdp = p->p_fd;
963 	struct cwdinfo *cwdi = p->p_cwdi;
964 	struct vnode	*vp;
965 	struct file	*fp;
966 	int		 error;
967 
968 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
969 		return error;
970 	/* getvnode() will use the descriptor for us */
971 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
972 		return error;
973 	vp = (struct vnode *) fp->f_data;
974 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
975 	if (vp->v_type != VDIR)
976 		error = ENOTDIR;
977 	else
978 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
979 	VOP_UNLOCK(vp, 0);
980 	if (error)
981 		goto out;
982 	VREF(vp);
983 
984 	/*
985 	 * Prevent escaping from chroot by putting the root under
986 	 * the working directory.  Silently chdir to / if we aren't
987 	 * already there.
988 	 */
989 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
990 		/*
991 		 * XXX would be more failsafe to change directory to a
992 		 * deadfs node here instead
993 		 */
994 		vrele(cwdi->cwdi_cdir);
995 		VREF(vp);
996 		cwdi->cwdi_cdir = vp;
997 	}
998 
999 	if (cwdi->cwdi_rdir != NULL)
1000 		vrele(cwdi->cwdi_rdir);
1001 	cwdi->cwdi_rdir = vp;
1002  out:
1003 	FILE_UNUSE(fp, p);
1004 	return (error);
1005 }
1006 
1007 
1008 
1009 /*
1010  * Change current working directory (``.'').
1011  */
1012 /* ARGSUSED */
1013 int
1014 sys_chdir(l, v, retval)
1015 	struct lwp *l;
1016 	void *v;
1017 	register_t *retval;
1018 {
1019 	struct sys_chdir_args /* {
1020 		syscallarg(const char *) path;
1021 	} */ *uap = v;
1022 	struct proc *p = l->l_proc;
1023 	struct cwdinfo *cwdi = p->p_cwdi;
1024 	int error;
1025 	struct nameidata nd;
1026 
1027 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1028 	    SCARG(uap, path), p);
1029 	if ((error = change_dir(&nd, p)) != 0)
1030 		return (error);
1031 	vrele(cwdi->cwdi_cdir);
1032 	cwdi->cwdi_cdir = nd.ni_vp;
1033 	return (0);
1034 }
1035 
1036 /*
1037  * Change notion of root (``/'') directory.
1038  */
1039 /* ARGSUSED */
1040 int
1041 sys_chroot(l, v, retval)
1042 	struct lwp *l;
1043 	void *v;
1044 	register_t *retval;
1045 {
1046 	struct sys_chroot_args /* {
1047 		syscallarg(const char *) path;
1048 	} */ *uap = v;
1049 	struct proc *p = l->l_proc;
1050 	struct cwdinfo *cwdi = p->p_cwdi;
1051 	struct vnode *vp;
1052 	int error;
1053 	struct nameidata nd;
1054 
1055 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1056 		return (error);
1057 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1058 	    SCARG(uap, path), p);
1059 	if ((error = change_dir(&nd, p)) != 0)
1060 		return (error);
1061 	if (cwdi->cwdi_rdir != NULL)
1062 		vrele(cwdi->cwdi_rdir);
1063 	vp = nd.ni_vp;
1064 	cwdi->cwdi_rdir = vp;
1065 
1066 	/*
1067 	 * Prevent escaping from chroot by putting the root under
1068 	 * the working directory.  Silently chdir to / if we aren't
1069 	 * already there.
1070 	 */
1071 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1072 		/*
1073 		 * XXX would be more failsafe to change directory to a
1074 		 * deadfs node here instead
1075 		 */
1076 		vrele(cwdi->cwdi_cdir);
1077 		VREF(vp);
1078 		cwdi->cwdi_cdir = vp;
1079 	}
1080 
1081 	return (0);
1082 }
1083 
1084 /*
1085  * Common routine for chroot and chdir.
1086  */
1087 static int
1088 change_dir(ndp, p)
1089 	struct nameidata *ndp;
1090 	struct proc *p;
1091 {
1092 	struct vnode *vp;
1093 	int error;
1094 
1095 	if ((error = namei(ndp)) != 0)
1096 		return (error);
1097 	vp = ndp->ni_vp;
1098 	if (vp->v_type != VDIR)
1099 		error = ENOTDIR;
1100 	else
1101 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
1102 
1103 	if (error)
1104 		vput(vp);
1105 	else
1106 		VOP_UNLOCK(vp, 0);
1107 	return (error);
1108 }
1109 
1110 /*
1111  * Check permissions, allocate an open file structure,
1112  * and call the device open routine if any.
1113  */
1114 int
1115 sys_open(l, v, retval)
1116 	struct lwp *l;
1117 	void *v;
1118 	register_t *retval;
1119 {
1120 	struct sys_open_args /* {
1121 		syscallarg(const char *) path;
1122 		syscallarg(int) flags;
1123 		syscallarg(int) mode;
1124 	} */ *uap = v;
1125 	struct proc *p = l->l_proc;
1126 	struct cwdinfo *cwdi = p->p_cwdi;
1127 	struct filedesc *fdp = p->p_fd;
1128 	struct file *fp;
1129 	struct vnode *vp;
1130 	int flags, cmode;
1131 	int type, indx, error;
1132 	struct flock lf;
1133 	struct nameidata nd;
1134 
1135 	flags = FFLAGS(SCARG(uap, flags));
1136 	if ((flags & (FREAD | FWRITE)) == 0)
1137 		return (EINVAL);
1138 	/* falloc() will use the file descriptor for us */
1139 	if ((error = falloc(p, &fp, &indx)) != 0)
1140 		return (error);
1141 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1142 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1143 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1144 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1145 		FILE_UNUSE(fp, p);
1146 		fdp->fd_ofiles[indx] = NULL;
1147 		ffree(fp);
1148 		if ((error == EDUPFD || error == EMOVEFD) &&
1149 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1150 		    (error =
1151 			dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) {
1152 			*retval = indx;
1153 			return (0);
1154 		}
1155 		if (error == ERESTART)
1156 			error = EINTR;
1157 		fdremove(fdp, indx);
1158 		return (error);
1159 	}
1160 	l->l_dupfd = 0;
1161 	vp = nd.ni_vp;
1162 	fp->f_flag = flags & FMASK;
1163 	fp->f_type = DTYPE_VNODE;
1164 	fp->f_ops = &vnops;
1165 	fp->f_data = vp;
1166 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1167 		lf.l_whence = SEEK_SET;
1168 		lf.l_start = 0;
1169 		lf.l_len = 0;
1170 		if (flags & O_EXLOCK)
1171 			lf.l_type = F_WRLCK;
1172 		else
1173 			lf.l_type = F_RDLCK;
1174 		type = F_FLOCK;
1175 		if ((flags & FNONBLOCK) == 0)
1176 			type |= F_WAIT;
1177 		VOP_UNLOCK(vp, 0);
1178 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1179 		if (error) {
1180 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1181 			FILE_UNUSE(fp, p);
1182 			ffree(fp);
1183 			fdremove(fdp, indx);
1184 			return (error);
1185 		}
1186 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1187 		fp->f_flag |= FHASLOCK;
1188 	}
1189 	VOP_UNLOCK(vp, 0);
1190 	*retval = indx;
1191 	FILE_SET_MATURE(fp);
1192 	FILE_UNUSE(fp, p);
1193 	return (0);
1194 }
1195 
1196 /*
1197  * Get file handle system call
1198  */
1199 int
1200 sys_getfh(l, v, retval)
1201 	struct lwp *l;
1202 	void *v;
1203 	register_t *retval;
1204 {
1205 	struct sys_getfh_args /* {
1206 		syscallarg(char *) fname;
1207 		syscallarg(fhandle_t *) fhp;
1208 	} */ *uap = v;
1209 	struct proc *p = l->l_proc;
1210 	struct vnode *vp;
1211 	fhandle_t fh;
1212 	int error;
1213 	struct nameidata nd;
1214 
1215 	/*
1216 	 * Must be super user
1217 	 */
1218 	error = suser(p->p_ucred, &p->p_acflag);
1219 	if (error)
1220 		return (error);
1221 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1222 	    SCARG(uap, fname), p);
1223 	error = namei(&nd);
1224 	if (error)
1225 		return (error);
1226 	vp = nd.ni_vp;
1227 	memset(&fh, 0, sizeof(fh));
1228 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1229 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1230 	vput(vp);
1231 	if (error)
1232 		return (error);
1233 	error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1234 	return (error);
1235 }
1236 
1237 /*
1238  * Open a file given a file handle.
1239  *
1240  * Check permissions, allocate an open file structure,
1241  * and call the device open routine if any.
1242  */
1243 int
1244 sys_fhopen(l, v, retval)
1245 	struct lwp *l;
1246 	void *v;
1247 	register_t *retval;
1248 {
1249 	struct sys_fhopen_args /* {
1250 		syscallarg(const fhandle_t *) fhp;
1251 		syscallarg(int) flags;
1252 	} */ *uap = v;
1253 	struct proc *p = l->l_proc;
1254 	struct filedesc *fdp = p->p_fd;
1255 	struct file *fp;
1256 	struct vnode *vp = NULL;
1257 	struct mount *mp;
1258 	struct ucred *cred = p->p_ucred;
1259 	int flags;
1260 	struct file *nfp;
1261 	int type, indx, error=0;
1262 	struct flock lf;
1263 	struct vattr va;
1264 	fhandle_t fh;
1265 
1266 	/*
1267 	 * Must be super user
1268 	 */
1269 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1270 		return (error);
1271 
1272 	flags = FFLAGS(SCARG(uap, flags));
1273 	if ((flags & (FREAD | FWRITE)) == 0)
1274 		return (EINVAL);
1275 	if ((flags & O_CREAT))
1276 		return (EINVAL);
1277 	/* falloc() will use the file descriptor for us */
1278 	if ((error = falloc(p, &nfp, &indx)) != 0)
1279 		return (error);
1280 	fp = nfp;
1281 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1282 		goto bad;
1283 
1284 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1285 		error = ESTALE;
1286 		goto bad;
1287 	}
1288 
1289 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1290 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1291 		goto bad;
1292 	}
1293 
1294 	/* Now do an effective vn_open */
1295 
1296 	if (vp->v_type == VSOCK) {
1297 		error = EOPNOTSUPP;
1298 		goto bad;
1299 	}
1300 	if (flags & FREAD) {
1301 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1302 			goto bad;
1303 	}
1304 	if (flags & (FWRITE | O_TRUNC)) {
1305 		if (vp->v_type == VDIR) {
1306 			error = EISDIR;
1307 			goto bad;
1308 		}
1309 		if ((error = vn_writechk(vp)) != 0 ||
1310 		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1311 			goto bad;
1312 	}
1313 	if (flags & O_TRUNC) {
1314 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
1315 			goto bad;
1316 		VOP_UNLOCK(vp, 0);			/* XXX */
1317 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1318 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1319 		VATTR_NULL(&va);
1320 		va.va_size = 0;
1321 		error = VOP_SETATTR(vp, &va, cred, p);
1322 		vn_finished_write(mp, 0);
1323 		if (error)
1324 			goto bad;
1325 	}
1326 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1327 		goto bad;
1328 	if (vp->v_type == VREG &&
1329 	    uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1330 		error = EIO;
1331 		goto bad;
1332 	}
1333 	if (flags & FWRITE)
1334 		vp->v_writecount++;
1335 
1336 	/* done with modified vn_open, now finish what sys_open does. */
1337 
1338 	fp->f_flag = flags & FMASK;
1339 	fp->f_type = DTYPE_VNODE;
1340 	fp->f_ops = &vnops;
1341 	fp->f_data = vp;
1342 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1343 		lf.l_whence = SEEK_SET;
1344 		lf.l_start = 0;
1345 		lf.l_len = 0;
1346 		if (flags & O_EXLOCK)
1347 			lf.l_type = F_WRLCK;
1348 		else
1349 			lf.l_type = F_RDLCK;
1350 		type = F_FLOCK;
1351 		if ((flags & FNONBLOCK) == 0)
1352 			type |= F_WAIT;
1353 		VOP_UNLOCK(vp, 0);
1354 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1355 		if (error) {
1356 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1357 			FILE_UNUSE(fp, p);
1358 			ffree(fp);
1359 			fdremove(fdp, indx);
1360 			return (error);
1361 		}
1362 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1363 		fp->f_flag |= FHASLOCK;
1364 	}
1365 	VOP_UNLOCK(vp, 0);
1366 	*retval = indx;
1367 	FILE_SET_MATURE(fp);
1368 	FILE_UNUSE(fp, p);
1369 	return (0);
1370 
1371 bad:
1372 	FILE_UNUSE(fp, p);
1373 	ffree(fp);
1374 	fdremove(fdp, indx);
1375 	if (vp != NULL)
1376 		vput(vp);
1377 	return (error);
1378 }
1379 
1380 /* ARGSUSED */
1381 int
1382 sys_fhstat(l, v, retval)
1383 	struct lwp *l;
1384 	void *v;
1385 	register_t *retval;
1386 {
1387 	struct sys_fhstat_args /* {
1388 		syscallarg(const fhandle_t *) fhp;
1389 		syscallarg(struct stat *) sb;
1390 	} */ *uap = v;
1391 	struct proc *p = l->l_proc;
1392 	struct stat sb;
1393 	int error;
1394 	fhandle_t fh;
1395 	struct mount *mp;
1396 	struct vnode *vp;
1397 
1398 	/*
1399 	 * Must be super user
1400 	 */
1401 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1402 		return (error);
1403 
1404 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1405 		return (error);
1406 
1407 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1408 		return (ESTALE);
1409 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1410 		return (error);
1411 	error = vn_stat(vp, &sb, p);
1412 	vput(vp);
1413 	if (error)
1414 		return (error);
1415 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1416 	return (error);
1417 }
1418 
1419 /* ARGSUSED */
1420 int
1421 sys_fhstatvfs1(l, v, retval)
1422 	struct lwp *l;
1423 	void *v;
1424 	register_t *retval;
1425 {
1426 	struct sys_fhstatvfs1_args /*
1427 		syscallarg(const fhandle_t *) fhp;
1428 		syscallarg(struct statvfs *) buf;
1429 		syscallarg(int)	flags;
1430 	} */ *uap = v;
1431 	struct proc *p = l->l_proc;
1432 	struct statvfs sbuf;
1433 	fhandle_t fh;
1434 	struct mount *mp;
1435 	struct vnode *vp;
1436 	int error;
1437 
1438 	/*
1439 	 * Must be super user
1440 	 */
1441 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1442 		return error;
1443 
1444 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1445 		return error;
1446 
1447 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1448 		return ESTALE;
1449 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1450 		return error;
1451 
1452 	mp = vp->v_mount;
1453 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) {
1454 		vput(vp);
1455 		return error;
1456 	}
1457 	vput(vp);
1458 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
1459 }
1460 
1461 /*
1462  * Create a special file.
1463  */
1464 /* ARGSUSED */
1465 int
1466 sys_mknod(l, v, retval)
1467 	struct lwp *l;
1468 	void *v;
1469 	register_t *retval;
1470 {
1471 	struct sys_mknod_args /* {
1472 		syscallarg(const char *) path;
1473 		syscallarg(int) mode;
1474 		syscallarg(int) dev;
1475 	} */ *uap = v;
1476 	struct proc *p = l->l_proc;
1477 	struct vnode *vp;
1478 	struct mount *mp;
1479 	struct vattr vattr;
1480 	int error;
1481 	int whiteout = 0;
1482 	struct nameidata nd;
1483 
1484 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1485 		return (error);
1486 restart:
1487 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1488 	if ((error = namei(&nd)) != 0)
1489 		return (error);
1490 	vp = nd.ni_vp;
1491 	if (vp != NULL)
1492 		error = EEXIST;
1493 	else {
1494 		VATTR_NULL(&vattr);
1495 		vattr.va_mode =
1496 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1497 		vattr.va_rdev = SCARG(uap, dev);
1498 		whiteout = 0;
1499 
1500 		switch (SCARG(uap, mode) & S_IFMT) {
1501 		case S_IFMT:	/* used by badsect to flag bad sectors */
1502 			vattr.va_type = VBAD;
1503 			break;
1504 		case S_IFCHR:
1505 			vattr.va_type = VCHR;
1506 			break;
1507 		case S_IFBLK:
1508 			vattr.va_type = VBLK;
1509 			break;
1510 		case S_IFWHT:
1511 			whiteout = 1;
1512 			break;
1513 		default:
1514 			error = EINVAL;
1515 			break;
1516 		}
1517 	}
1518 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1519 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1520 		if (nd.ni_dvp == vp)
1521 			vrele(nd.ni_dvp);
1522 		else
1523 			vput(nd.ni_dvp);
1524 		if (vp)
1525 			vrele(vp);
1526 		if ((error = vn_start_write(NULL, &mp,
1527 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1528 			return (error);
1529 		goto restart;
1530 	}
1531 	if (!error) {
1532 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1533 		if (whiteout) {
1534 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1535 			if (error)
1536 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1537 			vput(nd.ni_dvp);
1538 		} else {
1539 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1540 						&nd.ni_cnd, &vattr);
1541 			if (error == 0)
1542 				vput(nd.ni_vp);
1543 		}
1544 	} else {
1545 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1546 		if (nd.ni_dvp == vp)
1547 			vrele(nd.ni_dvp);
1548 		else
1549 			vput(nd.ni_dvp);
1550 		if (vp)
1551 			vrele(vp);
1552 	}
1553 	vn_finished_write(mp, 0);
1554 	return (error);
1555 }
1556 
1557 /*
1558  * Create a named pipe.
1559  */
1560 /* ARGSUSED */
1561 int
1562 sys_mkfifo(l, v, retval)
1563 	struct lwp *l;
1564 	void *v;
1565 	register_t *retval;
1566 {
1567 	struct sys_mkfifo_args /* {
1568 		syscallarg(const char *) path;
1569 		syscallarg(int) mode;
1570 	} */ *uap = v;
1571 	struct proc *p = l->l_proc;
1572 	struct mount *mp;
1573 	struct vattr vattr;
1574 	int error;
1575 	struct nameidata nd;
1576 
1577 restart:
1578 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1579 	if ((error = namei(&nd)) != 0)
1580 		return (error);
1581 	if (nd.ni_vp != NULL) {
1582 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1583 		if (nd.ni_dvp == nd.ni_vp)
1584 			vrele(nd.ni_dvp);
1585 		else
1586 			vput(nd.ni_dvp);
1587 		vrele(nd.ni_vp);
1588 		return (EEXIST);
1589 	}
1590 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1591 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1592 		if (nd.ni_dvp == nd.ni_vp)
1593 			vrele(nd.ni_dvp);
1594 		else
1595 			vput(nd.ni_dvp);
1596 		if (nd.ni_vp)
1597 			vrele(nd.ni_vp);
1598 		if ((error = vn_start_write(NULL, &mp,
1599 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1600 			return (error);
1601 		goto restart;
1602 	}
1603 	VATTR_NULL(&vattr);
1604 	vattr.va_type = VFIFO;
1605 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1606 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1607 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1608 	if (error == 0)
1609 		vput(nd.ni_vp);
1610 	vn_finished_write(mp, 0);
1611 	return (error);
1612 }
1613 
1614 /*
1615  * Make a hard file link.
1616  */
1617 /* ARGSUSED */
1618 int
1619 sys_link(l, v, retval)
1620 	struct lwp *l;
1621 	void *v;
1622 	register_t *retval;
1623 {
1624 	struct sys_link_args /* {
1625 		syscallarg(const char *) path;
1626 		syscallarg(const char *) link;
1627 	} */ *uap = v;
1628 	struct proc *p = l->l_proc;
1629 	struct vnode *vp;
1630 	struct mount *mp;
1631 	struct nameidata nd;
1632 	int error;
1633 
1634 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1635 	if ((error = namei(&nd)) != 0)
1636 		return (error);
1637 	vp = nd.ni_vp;
1638 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
1639 		vrele(vp);
1640 		return (error);
1641 	}
1642 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1643 	if ((error = namei(&nd)) != 0)
1644 		goto out;
1645 	if (nd.ni_vp) {
1646 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1647 		if (nd.ni_dvp == nd.ni_vp)
1648 			vrele(nd.ni_dvp);
1649 		else
1650 			vput(nd.ni_dvp);
1651 		vrele(nd.ni_vp);
1652 		error = EEXIST;
1653 		goto out;
1654 	}
1655 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1656 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1657 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1658 out:
1659 	vrele(vp);
1660 	vn_finished_write(mp, 0);
1661 	return (error);
1662 }
1663 
1664 /*
1665  * Make a symbolic link.
1666  */
1667 /* ARGSUSED */
1668 int
1669 sys_symlink(l, v, retval)
1670 	struct lwp *l;
1671 	void *v;
1672 	register_t *retval;
1673 {
1674 	struct sys_symlink_args /* {
1675 		syscallarg(const char *) path;
1676 		syscallarg(const char *) link;
1677 	} */ *uap = v;
1678 	struct proc *p = l->l_proc;
1679 	struct mount *mp;
1680 	struct vattr vattr;
1681 	char *path;
1682 	int error;
1683 	struct nameidata nd;
1684 
1685 	path = PNBUF_GET();
1686 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1687 	if (error)
1688 		goto out;
1689 restart:
1690 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1691 	if ((error = namei(&nd)) != 0)
1692 		goto out;
1693 	if (nd.ni_vp) {
1694 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1695 		if (nd.ni_dvp == nd.ni_vp)
1696 			vrele(nd.ni_dvp);
1697 		else
1698 			vput(nd.ni_dvp);
1699 		vrele(nd.ni_vp);
1700 		error = EEXIST;
1701 		goto out;
1702 	}
1703 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1704 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1705 		if (nd.ni_dvp == nd.ni_vp)
1706 			vrele(nd.ni_dvp);
1707 		else
1708 			vput(nd.ni_dvp);
1709 		if ((error = vn_start_write(NULL, &mp,
1710 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1711 			return (error);
1712 		goto restart;
1713 	}
1714 	VATTR_NULL(&vattr);
1715 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1716 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1717 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1718 	if (error == 0)
1719 		vput(nd.ni_vp);
1720 	vn_finished_write(mp, 0);
1721 out:
1722 	PNBUF_PUT(path);
1723 	return (error);
1724 }
1725 
1726 /*
1727  * Delete a whiteout from the filesystem.
1728  */
1729 /* ARGSUSED */
1730 int
1731 sys_undelete(l, v, retval)
1732 	struct lwp *l;
1733 	void *v;
1734 	register_t *retval;
1735 {
1736 	struct sys_undelete_args /* {
1737 		syscallarg(const char *) path;
1738 	} */ *uap = v;
1739 	struct proc *p = l->l_proc;
1740 	int error;
1741 	struct mount *mp;
1742 	struct nameidata nd;
1743 
1744 restart:
1745 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1746 	    SCARG(uap, path), p);
1747 	error = namei(&nd);
1748 	if (error)
1749 		return (error);
1750 
1751 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1752 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1753 		if (nd.ni_dvp == nd.ni_vp)
1754 			vrele(nd.ni_dvp);
1755 		else
1756 			vput(nd.ni_dvp);
1757 		if (nd.ni_vp)
1758 			vrele(nd.ni_vp);
1759 		return (EEXIST);
1760 	}
1761 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1762 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1763 		if (nd.ni_dvp == nd.ni_vp)
1764 			vrele(nd.ni_dvp);
1765 		else
1766 			vput(nd.ni_dvp);
1767 		if ((error = vn_start_write(NULL, &mp,
1768 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1769 			return (error);
1770 		goto restart;
1771 	}
1772 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1773 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1774 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1775 	vput(nd.ni_dvp);
1776 	vn_finished_write(mp, 0);
1777 	return (error);
1778 }
1779 
1780 /*
1781  * Delete a name from the filesystem.
1782  */
1783 /* ARGSUSED */
1784 int
1785 sys_unlink(l, v, retval)
1786 	struct lwp *l;
1787 	void *v;
1788 	register_t *retval;
1789 {
1790 	struct sys_unlink_args /* {
1791 		syscallarg(const char *) path;
1792 	} */ *uap = v;
1793 	struct proc *p = l->l_proc;
1794 	struct mount *mp;
1795 	struct vnode *vp;
1796 	int error;
1797 	struct nameidata nd;
1798 
1799 restart:
1800 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1801 	    SCARG(uap, path), p);
1802 	if ((error = namei(&nd)) != 0)
1803 		return (error);
1804 	vp = nd.ni_vp;
1805 
1806 	/*
1807 	 * The root of a mounted filesystem cannot be deleted.
1808 	 */
1809 	if (vp->v_flag & VROOT) {
1810 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1811 		if (nd.ni_dvp == vp)
1812 			vrele(nd.ni_dvp);
1813 		else
1814 			vput(nd.ni_dvp);
1815 		vput(vp);
1816 		error = EBUSY;
1817 		goto out;
1818 	}
1819 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1820 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1821 		if (nd.ni_dvp == vp)
1822 			vrele(nd.ni_dvp);
1823 		else
1824 			vput(nd.ni_dvp);
1825 		vput(vp);
1826 		if ((error = vn_start_write(NULL, &mp,
1827 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1828 			return (error);
1829 		goto restart;
1830 	}
1831 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1832 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1833 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1834 	vn_finished_write(mp, 0);
1835 out:
1836 	return (error);
1837 }
1838 
1839 /*
1840  * Reposition read/write file offset.
1841  */
1842 int
1843 sys_lseek(l, v, retval)
1844 	struct lwp *l;
1845 	void *v;
1846 	register_t *retval;
1847 {
1848 	struct sys_lseek_args /* {
1849 		syscallarg(int) fd;
1850 		syscallarg(int) pad;
1851 		syscallarg(off_t) offset;
1852 		syscallarg(int) whence;
1853 	} */ *uap = v;
1854 	struct proc *p = l->l_proc;
1855 	struct ucred *cred = p->p_ucred;
1856 	struct filedesc *fdp = p->p_fd;
1857 	struct file *fp;
1858 	struct vnode *vp;
1859 	struct vattr vattr;
1860 	off_t newoff;
1861 	int error;
1862 
1863 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1864 		return (EBADF);
1865 
1866 	FILE_USE(fp);
1867 
1868 	vp = (struct vnode *)fp->f_data;
1869 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1870 		error = ESPIPE;
1871 		goto out;
1872 	}
1873 
1874 	switch (SCARG(uap, whence)) {
1875 	case SEEK_CUR:
1876 		newoff = fp->f_offset + SCARG(uap, offset);
1877 		break;
1878 	case SEEK_END:
1879 		error = VOP_GETATTR(vp, &vattr, cred, p);
1880 		if (error)
1881 			goto out;
1882 		newoff = SCARG(uap, offset) + vattr.va_size;
1883 		break;
1884 	case SEEK_SET:
1885 		newoff = SCARG(uap, offset);
1886 		break;
1887 	default:
1888 		error = EINVAL;
1889 		goto out;
1890 	}
1891 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1892 		goto out;
1893 
1894 	*(off_t *)retval = fp->f_offset = newoff;
1895  out:
1896 	FILE_UNUSE(fp, p);
1897 	return (error);
1898 }
1899 
1900 /*
1901  * Positional read system call.
1902  */
1903 int
1904 sys_pread(l, v, retval)
1905 	struct lwp *l;
1906 	void *v;
1907 	register_t *retval;
1908 {
1909 	struct sys_pread_args /* {
1910 		syscallarg(int) fd;
1911 		syscallarg(void *) buf;
1912 		syscallarg(size_t) nbyte;
1913 		syscallarg(off_t) offset;
1914 	} */ *uap = v;
1915 	struct proc *p = l->l_proc;
1916 	struct filedesc *fdp = p->p_fd;
1917 	struct file *fp;
1918 	struct vnode *vp;
1919 	off_t offset;
1920 	int error, fd = SCARG(uap, fd);
1921 
1922 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1923 		return (EBADF);
1924 
1925 	if ((fp->f_flag & FREAD) == 0) {
1926 		simple_unlock(&fp->f_slock);
1927 		return (EBADF);
1928 	}
1929 
1930 	FILE_USE(fp);
1931 
1932 	vp = (struct vnode *)fp->f_data;
1933 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1934 		error = ESPIPE;
1935 		goto out;
1936 	}
1937 
1938 	offset = SCARG(uap, offset);
1939 
1940 	/*
1941 	 * XXX This works because no file systems actually
1942 	 * XXX take any action on the seek operation.
1943 	 */
1944 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1945 		goto out;
1946 
1947 	/* dofileread() will unuse the descriptor for us */
1948 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1949 	    &offset, 0, retval));
1950 
1951  out:
1952 	FILE_UNUSE(fp, p);
1953 	return (error);
1954 }
1955 
1956 /*
1957  * Positional scatter read system call.
1958  */
1959 int
1960 sys_preadv(l, v, retval)
1961 	struct lwp *l;
1962 	void *v;
1963 	register_t *retval;
1964 {
1965 	struct sys_preadv_args /* {
1966 		syscallarg(int) fd;
1967 		syscallarg(const struct iovec *) iovp;
1968 		syscallarg(int) iovcnt;
1969 		syscallarg(off_t) offset;
1970 	} */ *uap = v;
1971 	struct proc *p = l->l_proc;
1972 	struct filedesc *fdp = p->p_fd;
1973 	struct file *fp;
1974 	struct vnode *vp;
1975 	off_t offset;
1976 	int error, fd = SCARG(uap, fd);
1977 
1978 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1979 		return (EBADF);
1980 
1981 	if ((fp->f_flag & FREAD) == 0) {
1982 		simple_unlock(&fp->f_slock);
1983 		return (EBADF);
1984 	}
1985 
1986 	FILE_USE(fp);
1987 
1988 	vp = (struct vnode *)fp->f_data;
1989 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1990 		error = ESPIPE;
1991 		goto out;
1992 	}
1993 
1994 	offset = SCARG(uap, offset);
1995 
1996 	/*
1997 	 * XXX This works because no file systems actually
1998 	 * XXX take any action on the seek operation.
1999 	 */
2000 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2001 		goto out;
2002 
2003 	/* dofilereadv() will unuse the descriptor for us */
2004 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2005 	    &offset, 0, retval));
2006 
2007  out:
2008 	FILE_UNUSE(fp, p);
2009 	return (error);
2010 }
2011 
2012 /*
2013  * Positional write system call.
2014  */
2015 int
2016 sys_pwrite(l, v, retval)
2017 	struct lwp *l;
2018 	void *v;
2019 	register_t *retval;
2020 {
2021 	struct sys_pwrite_args /* {
2022 		syscallarg(int) fd;
2023 		syscallarg(const void *) buf;
2024 		syscallarg(size_t) nbyte;
2025 		syscallarg(off_t) offset;
2026 	} */ *uap = v;
2027 	struct proc *p = l->l_proc;
2028 	struct filedesc *fdp = p->p_fd;
2029 	struct file *fp;
2030 	struct vnode *vp;
2031 	off_t offset;
2032 	int error, fd = SCARG(uap, fd);
2033 
2034 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2035 		return (EBADF);
2036 
2037 	if ((fp->f_flag & FWRITE) == 0) {
2038 		simple_unlock(&fp->f_slock);
2039 		return (EBADF);
2040 	}
2041 
2042 	FILE_USE(fp);
2043 
2044 	vp = (struct vnode *)fp->f_data;
2045 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2046 		error = ESPIPE;
2047 		goto out;
2048 	}
2049 
2050 	offset = SCARG(uap, offset);
2051 
2052 	/*
2053 	 * XXX This works because no file systems actually
2054 	 * XXX take any action on the seek operation.
2055 	 */
2056 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2057 		goto out;
2058 
2059 	/* dofilewrite() will unuse the descriptor for us */
2060 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2061 	    &offset, 0, retval));
2062 
2063  out:
2064 	FILE_UNUSE(fp, p);
2065 	return (error);
2066 }
2067 
2068 /*
2069  * Positional gather write system call.
2070  */
2071 int
2072 sys_pwritev(l, v, retval)
2073 	struct lwp *l;
2074 	void *v;
2075 	register_t *retval;
2076 {
2077 	struct sys_pwritev_args /* {
2078 		syscallarg(int) fd;
2079 		syscallarg(const struct iovec *) iovp;
2080 		syscallarg(int) iovcnt;
2081 		syscallarg(off_t) offset;
2082 	} */ *uap = v;
2083 	struct proc *p = l->l_proc;
2084 	struct filedesc *fdp = p->p_fd;
2085 	struct file *fp;
2086 	struct vnode *vp;
2087 	off_t offset;
2088 	int error, fd = SCARG(uap, fd);
2089 
2090 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2091 		return (EBADF);
2092 
2093 	if ((fp->f_flag & FWRITE) == 0) {
2094 		simple_unlock(&fp->f_slock);
2095 		return (EBADF);
2096 	}
2097 
2098 	FILE_USE(fp);
2099 
2100 	vp = (struct vnode *)fp->f_data;
2101 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2102 		error = ESPIPE;
2103 		goto out;
2104 	}
2105 
2106 	offset = SCARG(uap, offset);
2107 
2108 	/*
2109 	 * XXX This works because no file systems actually
2110 	 * XXX take any action on the seek operation.
2111 	 */
2112 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2113 		goto out;
2114 
2115 	/* dofilewritev() will unuse the descriptor for us */
2116 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2117 	    &offset, 0, retval));
2118 
2119  out:
2120 	FILE_UNUSE(fp, p);
2121 	return (error);
2122 }
2123 
2124 /*
2125  * Check access permissions.
2126  */
2127 int
2128 sys_access(l, v, retval)
2129 	struct lwp *l;
2130 	void *v;
2131 	register_t *retval;
2132 {
2133 	struct sys_access_args /* {
2134 		syscallarg(const char *) path;
2135 		syscallarg(int) flags;
2136 	} */ *uap = v;
2137 	struct proc *p = l->l_proc;
2138 	struct ucred *cred;
2139 	struct vnode *vp;
2140 	int error, flags;
2141 	struct nameidata nd;
2142 
2143 	cred = crdup(p->p_ucred);
2144 	cred->cr_uid = p->p_cred->p_ruid;
2145 	cred->cr_gid = p->p_cred->p_rgid;
2146 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2147 	    SCARG(uap, path), p);
2148 	/* Override default credentials */
2149 	nd.ni_cnd.cn_cred = cred;
2150 	if ((error = namei(&nd)) != 0)
2151 		goto out;
2152 	vp = nd.ni_vp;
2153 
2154 	/* Flags == 0 means only check for existence. */
2155 	if (SCARG(uap, flags)) {
2156 		flags = 0;
2157 		if (SCARG(uap, flags) & R_OK)
2158 			flags |= VREAD;
2159 		if (SCARG(uap, flags) & W_OK)
2160 			flags |= VWRITE;
2161 		if (SCARG(uap, flags) & X_OK)
2162 			flags |= VEXEC;
2163 
2164 		error = VOP_ACCESS(vp, flags, cred, p);
2165 		if (!error && (flags & VWRITE))
2166 			error = vn_writechk(vp);
2167 	}
2168 	vput(vp);
2169 out:
2170 	crfree(cred);
2171 	return (error);
2172 }
2173 
2174 /*
2175  * Get file status; this version follows links.
2176  */
2177 /* ARGSUSED */
2178 int
2179 sys___stat13(l, v, retval)
2180 	struct lwp *l;
2181 	void *v;
2182 	register_t *retval;
2183 {
2184 	struct sys___stat13_args /* {
2185 		syscallarg(const char *) path;
2186 		syscallarg(struct stat *) ub;
2187 	} */ *uap = v;
2188 	struct proc *p = l->l_proc;
2189 	struct stat sb;
2190 	int error;
2191 	struct nameidata nd;
2192 
2193 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2194 	    SCARG(uap, path), p);
2195 	if ((error = namei(&nd)) != 0)
2196 		return (error);
2197 	error = vn_stat(nd.ni_vp, &sb, p);
2198 	vput(nd.ni_vp);
2199 	if (error)
2200 		return (error);
2201 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2202 	return (error);
2203 }
2204 
2205 /*
2206  * Get file status; this version does not follow links.
2207  */
2208 /* ARGSUSED */
2209 int
2210 sys___lstat13(l, v, retval)
2211 	struct lwp *l;
2212 	void *v;
2213 	register_t *retval;
2214 {
2215 	struct sys___lstat13_args /* {
2216 		syscallarg(const char *) path;
2217 		syscallarg(struct stat *) ub;
2218 	} */ *uap = v;
2219 	struct proc *p = l->l_proc;
2220 	struct stat sb;
2221 	int error;
2222 	struct nameidata nd;
2223 
2224 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2225 	    SCARG(uap, path), p);
2226 	if ((error = namei(&nd)) != 0)
2227 		return (error);
2228 	error = vn_stat(nd.ni_vp, &sb, p);
2229 	vput(nd.ni_vp);
2230 	if (error)
2231 		return (error);
2232 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2233 	return (error);
2234 }
2235 
2236 /*
2237  * Get configurable pathname variables.
2238  */
2239 /* ARGSUSED */
2240 int
2241 sys_pathconf(l, v, retval)
2242 	struct lwp *l;
2243 	void *v;
2244 	register_t *retval;
2245 {
2246 	struct sys_pathconf_args /* {
2247 		syscallarg(const char *) path;
2248 		syscallarg(int) name;
2249 	} */ *uap = v;
2250 	struct proc *p = l->l_proc;
2251 	int error;
2252 	struct nameidata nd;
2253 
2254 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2255 	    SCARG(uap, path), p);
2256 	if ((error = namei(&nd)) != 0)
2257 		return (error);
2258 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2259 	vput(nd.ni_vp);
2260 	return (error);
2261 }
2262 
2263 /*
2264  * Return target name of a symbolic link.
2265  */
2266 /* ARGSUSED */
2267 int
2268 sys_readlink(l, v, retval)
2269 	struct lwp *l;
2270 	void *v;
2271 	register_t *retval;
2272 {
2273 	struct sys_readlink_args /* {
2274 		syscallarg(const char *) path;
2275 		syscallarg(char *) buf;
2276 		syscallarg(size_t) count;
2277 	} */ *uap = v;
2278 	struct proc *p = l->l_proc;
2279 	struct vnode *vp;
2280 	struct iovec aiov;
2281 	struct uio auio;
2282 	int error;
2283 	struct nameidata nd;
2284 
2285 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2286 	    SCARG(uap, path), p);
2287 	if ((error = namei(&nd)) != 0)
2288 		return (error);
2289 	vp = nd.ni_vp;
2290 	if (vp->v_type != VLNK)
2291 		error = EINVAL;
2292 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2293 	    (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2294 		aiov.iov_base = SCARG(uap, buf);
2295 		aiov.iov_len = SCARG(uap, count);
2296 		auio.uio_iov = &aiov;
2297 		auio.uio_iovcnt = 1;
2298 		auio.uio_offset = 0;
2299 		auio.uio_rw = UIO_READ;
2300 		auio.uio_segflg = UIO_USERSPACE;
2301 		auio.uio_procp = p;
2302 		auio.uio_resid = SCARG(uap, count);
2303 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2304 	}
2305 	vput(vp);
2306 	*retval = SCARG(uap, count) - auio.uio_resid;
2307 	return (error);
2308 }
2309 
2310 /*
2311  * Change flags of a file given a path name.
2312  */
2313 /* ARGSUSED */
2314 int
2315 sys_chflags(l, v, retval)
2316 	struct lwp *l;
2317 	void *v;
2318 	register_t *retval;
2319 {
2320 	struct sys_chflags_args /* {
2321 		syscallarg(const char *) path;
2322 		syscallarg(u_long) flags;
2323 	} */ *uap = v;
2324 	struct proc *p = l->l_proc;
2325 	struct vnode *vp;
2326 	int error;
2327 	struct nameidata nd;
2328 
2329 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2330 	if ((error = namei(&nd)) != 0)
2331 		return (error);
2332 	vp = nd.ni_vp;
2333 	error = change_flags(vp, SCARG(uap, flags), p);
2334 	vput(vp);
2335 	return (error);
2336 }
2337 
2338 /*
2339  * Change flags of a file given a file descriptor.
2340  */
2341 /* ARGSUSED */
2342 int
2343 sys_fchflags(l, v, retval)
2344 	struct lwp *l;
2345 	void *v;
2346 	register_t *retval;
2347 {
2348 	struct sys_fchflags_args /* {
2349 		syscallarg(int) fd;
2350 		syscallarg(u_long) flags;
2351 	} */ *uap = v;
2352 	struct proc *p = l->l_proc;
2353 	struct vnode *vp;
2354 	struct file *fp;
2355 	int error;
2356 
2357 	/* getvnode() will use the descriptor for us */
2358 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2359 		return (error);
2360 	vp = (struct vnode *)fp->f_data;
2361 	error = change_flags(vp, SCARG(uap, flags), p);
2362 	VOP_UNLOCK(vp, 0);
2363 	FILE_UNUSE(fp, p);
2364 	return (error);
2365 }
2366 
2367 /*
2368  * Change flags of a file given a path name; this version does
2369  * not follow links.
2370  */
2371 int
2372 sys_lchflags(l, v, retval)
2373 	struct lwp *l;
2374 	void *v;
2375 	register_t *retval;
2376 {
2377 	struct sys_lchflags_args /* {
2378 		syscallarg(const char *) path;
2379 		syscallarg(u_long) flags;
2380 	} */ *uap = v;
2381 	struct proc *p = l->l_proc;
2382 	struct vnode *vp;
2383 	int error;
2384 	struct nameidata nd;
2385 
2386 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2387 	if ((error = namei(&nd)) != 0)
2388 		return (error);
2389 	vp = nd.ni_vp;
2390 	error = change_flags(vp, SCARG(uap, flags), p);
2391 	vput(vp);
2392 	return (error);
2393 }
2394 
2395 /*
2396  * Common routine to change flags of a file.
2397  */
2398 int
2399 change_flags(vp, flags, p)
2400 	struct vnode *vp;
2401 	u_long flags;
2402 	struct proc *p;
2403 {
2404 	struct mount *mp;
2405 	struct vattr vattr;
2406 	int error;
2407 
2408 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2409 		return (error);
2410 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2411 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2412 	/*
2413 	 * Non-superusers cannot change the flags on devices, even if they
2414 	 * own them.
2415 	 */
2416 	if (suser(p->p_ucred, &p->p_acflag) != 0) {
2417 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2418 			goto out;
2419 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2420 			error = EINVAL;
2421 			goto out;
2422 		}
2423 	}
2424 	VATTR_NULL(&vattr);
2425 	vattr.va_flags = flags;
2426 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2427 out:
2428 	vn_finished_write(mp, 0);
2429 	return (error);
2430 }
2431 
2432 /*
2433  * Change mode of a file given path name; this version follows links.
2434  */
2435 /* ARGSUSED */
2436 int
2437 sys_chmod(l, v, retval)
2438 	struct lwp *l;
2439 	void *v;
2440 	register_t *retval;
2441 {
2442 	struct sys_chmod_args /* {
2443 		syscallarg(const char *) path;
2444 		syscallarg(int) mode;
2445 	} */ *uap = v;
2446 	struct proc *p = l->l_proc;
2447 	int error;
2448 	struct nameidata nd;
2449 
2450 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2451 	if ((error = namei(&nd)) != 0)
2452 		return (error);
2453 
2454 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2455 
2456 	vrele(nd.ni_vp);
2457 	return (error);
2458 }
2459 
2460 /*
2461  * Change mode of a file given a file descriptor.
2462  */
2463 /* ARGSUSED */
2464 int
2465 sys_fchmod(l, v, retval)
2466 	struct lwp *l;
2467 	void *v;
2468 	register_t *retval;
2469 {
2470 	struct sys_fchmod_args /* {
2471 		syscallarg(int) fd;
2472 		syscallarg(int) mode;
2473 	} */ *uap = v;
2474 	struct proc *p = l->l_proc;
2475 	struct file *fp;
2476 	int error;
2477 
2478 	/* getvnode() will use the descriptor for us */
2479 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2480 		return (error);
2481 
2482 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2483 	FILE_UNUSE(fp, p);
2484 	return (error);
2485 }
2486 
2487 /*
2488  * Change mode of a file given path name; this version does not follow links.
2489  */
2490 /* ARGSUSED */
2491 int
2492 sys_lchmod(l, v, retval)
2493 	struct lwp *l;
2494 	void *v;
2495 	register_t *retval;
2496 {
2497 	struct sys_lchmod_args /* {
2498 		syscallarg(const char *) path;
2499 		syscallarg(int) mode;
2500 	} */ *uap = v;
2501 	struct proc *p = l->l_proc;
2502 	int error;
2503 	struct nameidata nd;
2504 
2505 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2506 	if ((error = namei(&nd)) != 0)
2507 		return (error);
2508 
2509 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2510 
2511 	vrele(nd.ni_vp);
2512 	return (error);
2513 }
2514 
2515 /*
2516  * Common routine to set mode given a vnode.
2517  */
2518 static int
2519 change_mode(vp, mode, p)
2520 	struct vnode *vp;
2521 	int mode;
2522 	struct proc *p;
2523 {
2524 	struct mount *mp;
2525 	struct vattr vattr;
2526 	int error;
2527 
2528 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2529 		return (error);
2530 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2531 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2532 	VATTR_NULL(&vattr);
2533 	vattr.va_mode = mode & ALLPERMS;
2534 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2535 	VOP_UNLOCK(vp, 0);
2536 	vn_finished_write(mp, 0);
2537 	return (error);
2538 }
2539 
2540 /*
2541  * Set ownership given a path name; this version follows links.
2542  */
2543 /* ARGSUSED */
2544 int
2545 sys_chown(l, v, retval)
2546 	struct lwp *l;
2547 	void *v;
2548 	register_t *retval;
2549 {
2550 	struct sys_chown_args /* {
2551 		syscallarg(const char *) path;
2552 		syscallarg(uid_t) uid;
2553 		syscallarg(gid_t) gid;
2554 	} */ *uap = v;
2555 	struct proc *p = l->l_proc;
2556 	int error;
2557 	struct nameidata nd;
2558 
2559 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2560 	if ((error = namei(&nd)) != 0)
2561 		return (error);
2562 
2563 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2564 
2565 	vrele(nd.ni_vp);
2566 	return (error);
2567 }
2568 
2569 /*
2570  * Set ownership given a path name; this version follows links.
2571  * Provides POSIX semantics.
2572  */
2573 /* ARGSUSED */
2574 int
2575 sys___posix_chown(l, v, retval)
2576 	struct lwp *l;
2577 	void *v;
2578 	register_t *retval;
2579 {
2580 	struct sys_chown_args /* {
2581 		syscallarg(const char *) path;
2582 		syscallarg(uid_t) uid;
2583 		syscallarg(gid_t) gid;
2584 	} */ *uap = v;
2585 	struct proc *p = l->l_proc;
2586 	int error;
2587 	struct nameidata nd;
2588 
2589 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2590 	if ((error = namei(&nd)) != 0)
2591 		return (error);
2592 
2593 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2594 
2595 	vrele(nd.ni_vp);
2596 	return (error);
2597 }
2598 
2599 /*
2600  * Set ownership given a file descriptor.
2601  */
2602 /* ARGSUSED */
2603 int
2604 sys_fchown(l, v, retval)
2605 	struct lwp *l;
2606 	void *v;
2607 	register_t *retval;
2608 {
2609 	struct sys_fchown_args /* {
2610 		syscallarg(int) fd;
2611 		syscallarg(uid_t) uid;
2612 		syscallarg(gid_t) gid;
2613 	} */ *uap = v;
2614 	struct proc *p = l->l_proc;
2615 	int error;
2616 	struct file *fp;
2617 
2618 	/* getvnode() will use the descriptor for us */
2619 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2620 		return (error);
2621 
2622 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2623 	    SCARG(uap, gid), p, 0);
2624 	FILE_UNUSE(fp, p);
2625 	return (error);
2626 }
2627 
2628 /*
2629  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2630  */
2631 /* ARGSUSED */
2632 int
2633 sys___posix_fchown(l, v, retval)
2634 	struct lwp *l;
2635 	void *v;
2636 	register_t *retval;
2637 {
2638 	struct sys_fchown_args /* {
2639 		syscallarg(int) fd;
2640 		syscallarg(uid_t) uid;
2641 		syscallarg(gid_t) gid;
2642 	} */ *uap = v;
2643 	struct proc *p = l->l_proc;
2644 	int error;
2645 	struct file *fp;
2646 
2647 	/* getvnode() will use the descriptor for us */
2648 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2649 		return (error);
2650 
2651 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2652 	    SCARG(uap, gid), p, 1);
2653 	FILE_UNUSE(fp, p);
2654 	return (error);
2655 }
2656 
2657 /*
2658  * Set ownership given a path name; this version does not follow links.
2659  */
2660 /* ARGSUSED */
2661 int
2662 sys_lchown(l, v, retval)
2663 	struct lwp *l;
2664 	void *v;
2665 	register_t *retval;
2666 {
2667 	struct sys_lchown_args /* {
2668 		syscallarg(const char *) path;
2669 		syscallarg(uid_t) uid;
2670 		syscallarg(gid_t) gid;
2671 	} */ *uap = v;
2672 	struct proc *p = l->l_proc;
2673 	int error;
2674 	struct nameidata nd;
2675 
2676 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2677 	if ((error = namei(&nd)) != 0)
2678 		return (error);
2679 
2680 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2681 
2682 	vrele(nd.ni_vp);
2683 	return (error);
2684 }
2685 
2686 /*
2687  * Set ownership given a path name; this version does not follow links.
2688  * Provides POSIX/XPG semantics.
2689  */
2690 /* ARGSUSED */
2691 int
2692 sys___posix_lchown(l, v, retval)
2693 	struct lwp *l;
2694 	void *v;
2695 	register_t *retval;
2696 {
2697 	struct sys_lchown_args /* {
2698 		syscallarg(const char *) path;
2699 		syscallarg(uid_t) uid;
2700 		syscallarg(gid_t) gid;
2701 	} */ *uap = v;
2702 	struct proc *p = l->l_proc;
2703 	int error;
2704 	struct nameidata nd;
2705 
2706 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2707 	if ((error = namei(&nd)) != 0)
2708 		return (error);
2709 
2710 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2711 
2712 	vrele(nd.ni_vp);
2713 	return (error);
2714 }
2715 
2716 /*
2717  * Common routine to set ownership given a vnode.
2718  */
2719 static int
2720 change_owner(vp, uid, gid, p, posix_semantics)
2721 	struct vnode *vp;
2722 	uid_t uid;
2723 	gid_t gid;
2724 	struct proc *p;
2725 	int posix_semantics;
2726 {
2727 	struct mount *mp;
2728 	struct vattr vattr;
2729 	mode_t newmode;
2730 	int error;
2731 
2732 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2733 		return (error);
2734 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2735 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2736 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2737 		goto out;
2738 
2739 #define CHANGED(x) ((int)(x) != -1)
2740 	newmode = vattr.va_mode;
2741 	if (posix_semantics) {
2742 		/*
2743 		 * POSIX/XPG semantics: if the caller is not the super-user,
2744 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2745 		 * the XPG consider the behaviour for calls by the super-user
2746 		 * implementation-defined; we leave the set-user-id and set-
2747 		 * group-id settings intact in that case.
2748 		 */
2749 		if (suser(p->p_ucred, NULL) != 0)
2750 			newmode &= ~(S_ISUID | S_ISGID);
2751 	} else {
2752 		/*
2753 		 * NetBSD semantics: when changing owner and/or group,
2754 		 * clear the respective bit(s).
2755 		 */
2756 		if (CHANGED(uid))
2757 			newmode &= ~S_ISUID;
2758 		if (CHANGED(gid))
2759 			newmode &= ~S_ISGID;
2760 	}
2761 	/* Update va_mode iff altered. */
2762 	if (vattr.va_mode == newmode)
2763 		newmode = VNOVAL;
2764 
2765 	VATTR_NULL(&vattr);
2766 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2767 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2768 	vattr.va_mode = newmode;
2769 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2770 #undef CHANGED
2771 
2772 out:
2773 	VOP_UNLOCK(vp, 0);
2774 	vn_finished_write(mp, 0);
2775 	return (error);
2776 }
2777 
2778 /*
2779  * Set the access and modification times given a path name; this
2780  * version follows links.
2781  */
2782 /* ARGSUSED */
2783 int
2784 sys_utimes(l, v, retval)
2785 	struct lwp *l;
2786 	void *v;
2787 	register_t *retval;
2788 {
2789 	struct sys_utimes_args /* {
2790 		syscallarg(const char *) path;
2791 		syscallarg(const struct timeval *) tptr;
2792 	} */ *uap = v;
2793 	struct proc *p = l->l_proc;
2794 	int error;
2795 	struct nameidata nd;
2796 
2797 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2798 	if ((error = namei(&nd)) != 0)
2799 		return (error);
2800 
2801 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2802 
2803 	vrele(nd.ni_vp);
2804 	return (error);
2805 }
2806 
2807 /*
2808  * Set the access and modification times given a file descriptor.
2809  */
2810 /* ARGSUSED */
2811 int
2812 sys_futimes(l, v, retval)
2813 	struct lwp *l;
2814 	void *v;
2815 	register_t *retval;
2816 {
2817 	struct sys_futimes_args /* {
2818 		syscallarg(int) fd;
2819 		syscallarg(const struct timeval *) tptr;
2820 	} */ *uap = v;
2821 	struct proc *p = l->l_proc;
2822 	int error;
2823 	struct file *fp;
2824 
2825 	/* getvnode() will use the descriptor for us */
2826 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2827 		return (error);
2828 
2829 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2830 	FILE_UNUSE(fp, p);
2831 	return (error);
2832 }
2833 
2834 /*
2835  * Set the access and modification times given a path name; this
2836  * version does not follow links.
2837  */
2838 /* ARGSUSED */
2839 int
2840 sys_lutimes(l, v, retval)
2841 	struct lwp *l;
2842 	void *v;
2843 	register_t *retval;
2844 {
2845 	struct sys_lutimes_args /* {
2846 		syscallarg(const char *) path;
2847 		syscallarg(const struct timeval *) tptr;
2848 	} */ *uap = v;
2849 	struct proc *p = l->l_proc;
2850 	int error;
2851 	struct nameidata nd;
2852 
2853 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2854 	if ((error = namei(&nd)) != 0)
2855 		return (error);
2856 
2857 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2858 
2859 	vrele(nd.ni_vp);
2860 	return (error);
2861 }
2862 
2863 /*
2864  * Common routine to set access and modification times given a vnode.
2865  */
2866 static int
2867 change_utimes(vp, tptr, p)
2868 	struct vnode *vp;
2869 	const struct timeval *tptr;
2870 	struct proc *p;
2871 {
2872 	struct timeval tv[2];
2873 	struct mount *mp;
2874 	struct vattr vattr;
2875 	int error;
2876 
2877 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2878 		return (error);
2879 	VATTR_NULL(&vattr);
2880 	if (tptr == NULL) {
2881 		microtime(&tv[0]);
2882 		tv[1] = tv[0];
2883 		vattr.va_vaflags |= VA_UTIMES_NULL;
2884 	} else {
2885 		error = copyin(tptr, tv, sizeof(tv));
2886 		if (error)
2887 			goto out;
2888 	}
2889 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2890 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2891 	vattr.va_atime.tv_sec = tv[0].tv_sec;
2892 	vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2893 	vattr.va_mtime.tv_sec = tv[1].tv_sec;
2894 	vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2895 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2896 	VOP_UNLOCK(vp, 0);
2897 out:
2898 	vn_finished_write(mp, 0);
2899 	return (error);
2900 }
2901 
2902 /*
2903  * Truncate a file given its path name.
2904  */
2905 /* ARGSUSED */
2906 int
2907 sys_truncate(l, v, retval)
2908 	struct lwp *l;
2909 	void *v;
2910 	register_t *retval;
2911 {
2912 	struct sys_truncate_args /* {
2913 		syscallarg(const char *) path;
2914 		syscallarg(int) pad;
2915 		syscallarg(off_t) length;
2916 	} */ *uap = v;
2917 	struct proc *p = l->l_proc;
2918 	struct vnode *vp;
2919 	struct mount *mp;
2920 	struct vattr vattr;
2921 	int error;
2922 	struct nameidata nd;
2923 
2924 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2925 	if ((error = namei(&nd)) != 0)
2926 		return (error);
2927 	vp = nd.ni_vp;
2928 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2929 		vrele(vp);
2930 		return (error);
2931 	}
2932 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2933 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2934 	if (vp->v_type == VDIR)
2935 		error = EISDIR;
2936 	else if ((error = vn_writechk(vp)) == 0 &&
2937 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2938 		VATTR_NULL(&vattr);
2939 		vattr.va_size = SCARG(uap, length);
2940 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2941 	}
2942 	vput(vp);
2943 	vn_finished_write(mp, 0);
2944 	return (error);
2945 }
2946 
2947 /*
2948  * Truncate a file given a file descriptor.
2949  */
2950 /* ARGSUSED */
2951 int
2952 sys_ftruncate(l, v, retval)
2953 	struct lwp *l;
2954 	void *v;
2955 	register_t *retval;
2956 {
2957 	struct sys_ftruncate_args /* {
2958 		syscallarg(int) fd;
2959 		syscallarg(int) pad;
2960 		syscallarg(off_t) length;
2961 	} */ *uap = v;
2962 	struct proc *p = l->l_proc;
2963 	struct mount *mp;
2964 	struct vattr vattr;
2965 	struct vnode *vp;
2966 	struct file *fp;
2967 	int error;
2968 
2969 	/* getvnode() will use the descriptor for us */
2970 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2971 		return (error);
2972 	if ((fp->f_flag & FWRITE) == 0) {
2973 		error = EINVAL;
2974 		goto out;
2975 	}
2976 	vp = (struct vnode *)fp->f_data;
2977 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2978 		FILE_UNUSE(fp, p);
2979 		return (error);
2980 	}
2981 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2982 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2983 	if (vp->v_type == VDIR)
2984 		error = EISDIR;
2985 	else if ((error = vn_writechk(vp)) == 0) {
2986 		VATTR_NULL(&vattr);
2987 		vattr.va_size = SCARG(uap, length);
2988 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2989 	}
2990 	VOP_UNLOCK(vp, 0);
2991 	vn_finished_write(mp, 0);
2992  out:
2993 	FILE_UNUSE(fp, p);
2994 	return (error);
2995 }
2996 
2997 /*
2998  * Sync an open file.
2999  */
3000 /* ARGSUSED */
3001 int
3002 sys_fsync(l, v, retval)
3003 	struct lwp *l;
3004 	void *v;
3005 	register_t *retval;
3006 {
3007 	struct sys_fsync_args /* {
3008 		syscallarg(int) fd;
3009 	} */ *uap = v;
3010 	struct proc *p = l->l_proc;
3011 	struct vnode *vp;
3012 	struct mount *mp;
3013 	struct file *fp;
3014 	int error;
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 	vp = (struct vnode *)fp->f_data;
3020 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3021 		FILE_UNUSE(fp, p);
3022 		return (error);
3023 	}
3024 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3025 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
3026 	if (error == 0 && bioops.io_fsync != NULL &&
3027 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3028 		(*bioops.io_fsync)(vp, 0);
3029 	VOP_UNLOCK(vp, 0);
3030 	vn_finished_write(mp, 0);
3031 	FILE_UNUSE(fp, p);
3032 	return (error);
3033 }
3034 
3035 /*
3036  * Sync a range of file data.  API modeled after that found in AIX.
3037  *
3038  * FDATASYNC indicates that we need only save enough metadata to be able
3039  * to re-read the written data.  Note we duplicate AIX's requirement that
3040  * the file be open for writing.
3041  */
3042 /* ARGSUSED */
3043 int
3044 sys_fsync_range(l, v, retval)
3045 	struct lwp *l;
3046 	void *v;
3047 	register_t *retval;
3048 {
3049 	struct sys_fsync_range_args /* {
3050 		syscallarg(int) fd;
3051 		syscallarg(int) flags;
3052 		syscallarg(off_t) start;
3053 		syscallarg(int) length;
3054 	} */ *uap = v;
3055 	struct proc *p = l->l_proc;
3056 	struct vnode *vp;
3057 	struct file *fp;
3058 	int flags, nflags;
3059 	off_t s, e, len;
3060 	int error;
3061 
3062 	/* getvnode() will use the descriptor for us */
3063 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3064 		return (error);
3065 
3066 	if ((fp->f_flag & FWRITE) == 0) {
3067 		FILE_UNUSE(fp, p);
3068 		return (EBADF);
3069 	}
3070 
3071 	flags = SCARG(uap, flags);
3072 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3073 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3074 		return (EINVAL);
3075 	}
3076 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3077 	if (flags & FDATASYNC)
3078 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3079 	else
3080 		nflags = FSYNC_WAIT;
3081 	if (flags & FDISKSYNC)
3082 		nflags |= FSYNC_CACHE;
3083 
3084 	len = SCARG(uap, length);
3085 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3086 	if (len) {
3087 		s = SCARG(uap, start);
3088 		e = s + len;
3089 		if (e < s) {
3090 			FILE_UNUSE(fp, p);
3091 			return (EINVAL);
3092 		}
3093 	} else {
3094 		e = 0;
3095 		s = 0;
3096 	}
3097 
3098 	vp = (struct vnode *)fp->f_data;
3099 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3100 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p);
3101 
3102 	if (error == 0 && bioops.io_fsync != NULL &&
3103 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3104 		(*bioops.io_fsync)(vp, nflags);
3105 
3106 	VOP_UNLOCK(vp, 0);
3107 	FILE_UNUSE(fp, p);
3108 	return (error);
3109 }
3110 
3111 /*
3112  * Sync the data of an open file.
3113  */
3114 /* ARGSUSED */
3115 int
3116 sys_fdatasync(l, v, retval)
3117 	struct lwp *l;
3118 	void *v;
3119 	register_t *retval;
3120 {
3121 	struct sys_fdatasync_args /* {
3122 		syscallarg(int) fd;
3123 	} */ *uap = v;
3124 	struct proc *p = l->l_proc;
3125 	struct vnode *vp;
3126 	struct file *fp;
3127 	int error;
3128 
3129 	/* getvnode() will use the descriptor for us */
3130 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3131 		return (error);
3132 	if ((fp->f_flag & FWRITE) == 0) {
3133 		FILE_UNUSE(fp, p);
3134 		return (EBADF);
3135 	}
3136 	vp = (struct vnode *)fp->f_data;
3137 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3138 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
3139 	VOP_UNLOCK(vp, 0);
3140 	FILE_UNUSE(fp, p);
3141 	return (error);
3142 }
3143 
3144 /*
3145  * Rename files, (standard) BSD semantics frontend.
3146  */
3147 /* ARGSUSED */
3148 int
3149 sys_rename(l, v, retval)
3150 	struct lwp *l;
3151 	void *v;
3152 	register_t *retval;
3153 {
3154 	struct sys_rename_args /* {
3155 		syscallarg(const char *) from;
3156 		syscallarg(const char *) to;
3157 	} */ *uap = v;
3158 	struct proc *p = l->l_proc;
3159 
3160 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
3161 }
3162 
3163 /*
3164  * Rename files, POSIX semantics frontend.
3165  */
3166 /* ARGSUSED */
3167 int
3168 sys___posix_rename(l, v, retval)
3169 	struct lwp *l;
3170 	void *v;
3171 	register_t *retval;
3172 {
3173 	struct sys___posix_rename_args /* {
3174 		syscallarg(const char *) from;
3175 		syscallarg(const char *) to;
3176 	} */ *uap = v;
3177 	struct proc *p = l->l_proc;
3178 
3179 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
3180 }
3181 
3182 /*
3183  * Rename files.  Source and destination must either both be directories,
3184  * or both not be directories.  If target is a directory, it must be empty.
3185  * If `from' and `to' refer to the same object, the value of the `retain'
3186  * argument is used to determine whether `from' will be
3187  *
3188  * (retain == 0)	deleted unless `from' and `to' refer to the same
3189  *			object in the file system's name space (BSD).
3190  * (retain == 1)	always retained (POSIX).
3191  */
3192 static int
3193 rename_files(from, to, p, retain)
3194 	const char *from, *to;
3195 	struct proc *p;
3196 	int retain;
3197 {
3198 	struct mount *mp = NULL;
3199 	struct vnode *tvp, *fvp, *tdvp;
3200 	struct nameidata fromnd, tond;
3201 	int error;
3202 
3203 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
3204 	    from, p);
3205 	if ((error = namei(&fromnd)) != 0)
3206 		return (error);
3207 	fvp = fromnd.ni_vp;
3208 	error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH);
3209 	if (error != 0) {
3210 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3211 		vrele(fromnd.ni_dvp);
3212 		vrele(fvp);
3213 		if (fromnd.ni_startdir)
3214 			vrele(fromnd.ni_startdir);
3215 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3216 		return (error);
3217 	}
3218 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3219 	    (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p);
3220 	if ((error = namei(&tond)) != 0) {
3221 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3222 		vrele(fromnd.ni_dvp);
3223 		vrele(fvp);
3224 		goto out1;
3225 	}
3226 	tdvp = tond.ni_dvp;
3227 	tvp = tond.ni_vp;
3228 
3229 	if (tvp != NULL) {
3230 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3231 			error = ENOTDIR;
3232 			goto out;
3233 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3234 			error = EISDIR;
3235 			goto out;
3236 		}
3237 	}
3238 
3239 	if (fvp == tdvp)
3240 		error = EINVAL;
3241 
3242 	/*
3243 	 * Source and destination refer to the same object.
3244 	 */
3245 	if (fvp == tvp) {
3246 		if (retain)
3247 			error = -1;
3248 		else if (fromnd.ni_dvp == tdvp &&
3249 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3250 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3251 		          tond.ni_cnd.cn_nameptr,
3252 		          fromnd.ni_cnd.cn_namelen))
3253 		error = -1;
3254 	}
3255 
3256 out:
3257 	if (!error) {
3258 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
3259 		if (fromnd.ni_dvp != tdvp)
3260 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3261 		if (tvp) {
3262 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
3263 		}
3264 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3265 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3266 	} else {
3267 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3268 		if (tdvp == tvp)
3269 			vrele(tdvp);
3270 		else
3271 			vput(tdvp);
3272 		if (tvp)
3273 			vput(tvp);
3274 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3275 		vrele(fromnd.ni_dvp);
3276 		vrele(fvp);
3277 	}
3278 	vrele(tond.ni_startdir);
3279 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3280 out1:
3281 	vn_finished_write(mp, 0);
3282 	if (fromnd.ni_startdir)
3283 		vrele(fromnd.ni_startdir);
3284 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3285 	return (error == -1 ? 0 : error);
3286 }
3287 
3288 /*
3289  * Make a directory file.
3290  */
3291 /* ARGSUSED */
3292 int
3293 sys_mkdir(l, v, retval)
3294 	struct lwp *l;
3295 	void *v;
3296 	register_t *retval;
3297 {
3298 	struct sys_mkdir_args /* {
3299 		syscallarg(const char *) path;
3300 		syscallarg(int) mode;
3301 	} */ *uap = v;
3302 	struct proc *p = l->l_proc;
3303 	struct mount *mp;
3304 	struct vnode *vp;
3305 	struct vattr vattr;
3306 	int error;
3307 	struct nameidata nd;
3308 
3309 restart:
3310 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE,
3311 	    SCARG(uap, path), p);
3312 	if ((error = namei(&nd)) != 0)
3313 		return (error);
3314 	vp = nd.ni_vp;
3315 	if (vp != NULL) {
3316 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3317 		if (nd.ni_dvp == vp)
3318 			vrele(nd.ni_dvp);
3319 		else
3320 			vput(nd.ni_dvp);
3321 		vrele(vp);
3322 		return (EEXIST);
3323 	}
3324 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3325 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3326 		if (nd.ni_dvp == vp)
3327 			vrele(nd.ni_dvp);
3328 		else
3329 			vput(nd.ni_dvp);
3330 		if ((error = vn_start_write(NULL, &mp,
3331 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3332 			return (error);
3333 		goto restart;
3334 	}
3335 	VATTR_NULL(&vattr);
3336 	vattr.va_type = VDIR;
3337 	vattr.va_mode =
3338 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3339 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3340 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3341 	if (!error)
3342 		vput(nd.ni_vp);
3343 	vn_finished_write(mp, 0);
3344 	return (error);
3345 }
3346 
3347 /*
3348  * Remove a directory file.
3349  */
3350 /* ARGSUSED */
3351 int
3352 sys_rmdir(l, v, retval)
3353 	struct lwp *l;
3354 	void *v;
3355 	register_t *retval;
3356 {
3357 	struct sys_rmdir_args /* {
3358 		syscallarg(const char *) path;
3359 	} */ *uap = v;
3360 	struct proc *p = l->l_proc;
3361 	struct mount *mp;
3362 	struct vnode *vp;
3363 	int error;
3364 	struct nameidata nd;
3365 
3366 restart:
3367 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3368 	    SCARG(uap, path), p);
3369 	if ((error = namei(&nd)) != 0)
3370 		return (error);
3371 	vp = nd.ni_vp;
3372 	if (vp->v_type != VDIR) {
3373 		error = ENOTDIR;
3374 		goto out;
3375 	}
3376 	/*
3377 	 * No rmdir "." please.
3378 	 */
3379 	if (nd.ni_dvp == vp) {
3380 		error = EINVAL;
3381 		goto out;
3382 	}
3383 	/*
3384 	 * The root of a mounted filesystem cannot be deleted.
3385 	 */
3386 	if (vp->v_flag & VROOT) {
3387 		error = EBUSY;
3388 		goto out;
3389 	}
3390 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3391 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3392 		if (nd.ni_dvp == vp)
3393 			vrele(nd.ni_dvp);
3394 		else
3395 			vput(nd.ni_dvp);
3396 		vput(vp);
3397 		if ((error = vn_start_write(NULL, &mp,
3398 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3399 			return (error);
3400 		goto restart;
3401 	}
3402 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3403 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3404 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3405 	vn_finished_write(mp, 0);
3406 	return (error);
3407 
3408 out:
3409 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3410 	if (nd.ni_dvp == vp)
3411 		vrele(nd.ni_dvp);
3412 	else
3413 		vput(nd.ni_dvp);
3414 	vput(vp);
3415 	return (error);
3416 }
3417 
3418 /*
3419  * Read a block of directory entries in a file system independent format.
3420  */
3421 int
3422 sys_getdents(l, v, retval)
3423 	struct lwp *l;
3424 	void *v;
3425 	register_t *retval;
3426 {
3427 	struct sys_getdents_args /* {
3428 		syscallarg(int) fd;
3429 		syscallarg(char *) buf;
3430 		syscallarg(size_t) count;
3431 	} */ *uap = v;
3432 	struct proc *p = l->l_proc;
3433 	struct file *fp;
3434 	int error, done;
3435 
3436 	/* getvnode() will use the descriptor for us */
3437 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3438 		return (error);
3439 	if ((fp->f_flag & FREAD) == 0) {
3440 		error = EBADF;
3441 		goto out;
3442 	}
3443 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3444 			SCARG(uap, count), &done, p, 0, 0);
3445 #ifdef KTRACE
3446 	if (!error && KTRPOINT(p, KTR_GENIO)) {
3447 		struct iovec iov;
3448 		iov.iov_base = SCARG(uap, buf);
3449 		iov.iov_len = done;
3450 		ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0);
3451 	}
3452 #endif
3453 	*retval = done;
3454  out:
3455 	FILE_UNUSE(fp, p);
3456 	return (error);
3457 }
3458 
3459 /*
3460  * Set the mode mask for creation of filesystem nodes.
3461  */
3462 int
3463 sys_umask(l, v, retval)
3464 	struct lwp *l;
3465 	void *v;
3466 	register_t *retval;
3467 {
3468 	struct sys_umask_args /* {
3469 		syscallarg(mode_t) newmask;
3470 	} */ *uap = v;
3471 	struct proc *p = l->l_proc;
3472 	struct cwdinfo *cwdi;
3473 
3474 	cwdi = p->p_cwdi;
3475 	*retval = cwdi->cwdi_cmask;
3476 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3477 	return (0);
3478 }
3479 
3480 /*
3481  * Void all references to file by ripping underlying filesystem
3482  * away from vnode.
3483  */
3484 /* ARGSUSED */
3485 int
3486 sys_revoke(l, v, retval)
3487 	struct lwp *l;
3488 	void *v;
3489 	register_t *retval;
3490 {
3491 	struct sys_revoke_args /* {
3492 		syscallarg(const char *) path;
3493 	} */ *uap = v;
3494 	struct proc *p = l->l_proc;
3495 	struct mount *mp;
3496 	struct vnode *vp;
3497 	struct vattr vattr;
3498 	int error;
3499 	struct nameidata nd;
3500 
3501 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3502 	if ((error = namei(&nd)) != 0)
3503 		return (error);
3504 	vp = nd.ni_vp;
3505 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3506 		goto out;
3507 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3508 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3509 		goto out;
3510 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3511 		goto out;
3512 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3513 		VOP_REVOKE(vp, REVOKEALL);
3514 	vn_finished_write(mp, 0);
3515 out:
3516 	vrele(vp);
3517 	return (error);
3518 }
3519 
3520 /*
3521  * Convert a user file descriptor to a kernel file entry.
3522  */
3523 int
3524 getvnode(fdp, fd, fpp)
3525 	struct filedesc *fdp;
3526 	int fd;
3527 	struct file **fpp;
3528 {
3529 	struct vnode *vp;
3530 	struct file *fp;
3531 
3532 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3533 		return (EBADF);
3534 
3535 	FILE_USE(fp);
3536 
3537 	if (fp->f_type != DTYPE_VNODE) {
3538 		FILE_UNUSE(fp, NULL);
3539 		return (EINVAL);
3540 	}
3541 
3542 	vp = (struct vnode *)fp->f_data;
3543 	if (vp->v_type == VBAD) {
3544 		FILE_UNUSE(fp, NULL);
3545 		return (EBADF);
3546 	}
3547 
3548 	*fpp = fp;
3549 	return (0);
3550 }
3551 
3552 /*
3553  * Push extended attribute configuration information into the VFS.
3554  *
3555  * NOTE: Not all file systems that support extended attributes will
3556  * require the use of this system call.
3557  */
3558 int
3559 sys_extattrctl(struct lwp *l, void *v, register_t *retval)
3560 {
3561 	struct sys_extattrctl_args /* {
3562 		syscallarg(const char *) path;
3563 		syscallarg(int) cmd;
3564 		syscallarg(const char *) filename;
3565 		syscallarg(int) attrnamespace;
3566 		syscallarg(const char *) attrname;
3567 	} */ *uap = v;
3568 	struct proc *p = l->l_proc;
3569 	struct vnode *vp;
3570 	struct nameidata nd;
3571 	struct mount *mp;
3572 	char attrname[EXTATTR_MAXNAMELEN];
3573 	int error;
3574 
3575 	if (SCARG(uap, attrname) != NULL) {
3576 		error = copyinstr(SCARG(uap, attrname), attrname,
3577 		    sizeof(attrname), NULL);
3578 		if (error)
3579 			return (error);
3580 	}
3581 
3582 	vp = NULL;
3583 	if (SCARG(uap, filename) != NULL) {
3584 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3585 		    SCARG(uap, filename), p);
3586 		error = namei(&nd);
3587 		if (error)
3588 			return (error);
3589 		vp = nd.ni_vp;
3590 	}
3591 
3592 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3593 	error = namei(&nd);
3594 	if (error) {
3595 		if (vp != NULL)
3596 			vput(vp);
3597 		return (error);
3598 	}
3599 
3600 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
3601 	if (error) {
3602 		if (vp != NULL)
3603 			vput(vp);
3604 		return (error);
3605 	}
3606 
3607 	error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), vp,
3608 	    SCARG(uap, attrnamespace),
3609 	    SCARG(uap, attrname) != NULL ? attrname : NULL, p);
3610 
3611 	vn_finished_write(mp, 0);
3612 
3613 	if (vp != NULL)
3614 		vrele(vp);
3615 
3616 	return (error);
3617 }
3618 
3619 /*
3620  * Set a named extended attribute on a file or directory.
3621  */
3622 static int
3623 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3624     const void *data, size_t nbytes, struct proc *p, register_t *retval)
3625 {
3626 	struct mount *mp;
3627 	struct uio auio;
3628 	struct iovec aiov;
3629 	ssize_t cnt;
3630 	int error;
3631 
3632 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3633 	if (error)
3634 		return (error);
3635 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3636 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3637 
3638 	aiov.iov_base = (caddr_t) data;		/* XXX kills const */
3639 	aiov.iov_len = nbytes;
3640 	auio.uio_iov = &aiov;
3641 	auio.uio_iovcnt = 1;
3642 	auio.uio_offset = 0;
3643 	if (nbytes > INT_MAX) {
3644 		error = EINVAL;
3645 		goto done;
3646 	}
3647 	auio.uio_resid = nbytes;
3648 	auio.uio_rw = UIO_WRITE;
3649 	auio.uio_segflg = UIO_USERSPACE;
3650 	auio.uio_procp = p;
3651 	cnt = nbytes;
3652 
3653 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
3654 	    p->p_ucred, p);
3655 	cnt -= auio.uio_resid;
3656 	retval[0] = cnt;
3657 
3658  done:
3659 	VOP_UNLOCK(vp, 0);
3660 	vn_finished_write(mp, 0);
3661 	return (error);
3662 }
3663 
3664 int
3665 sys_extattr_set_fd(struct lwp *l, void *v, register_t *retval)
3666 {
3667 	struct sys_extattr_set_fd_args /* {
3668 		syscallarg(int) fd;
3669 		syscallarg(int) attrnamespace;
3670 		syscallarg(const char *) attrname;
3671 		syscallarg(const void *) data;
3672 		syscallarg(size_t) nbytes;
3673 	} */ *uap = v;
3674 	struct proc *p = l->l_proc;
3675 	struct file *fp;
3676 	struct vnode *vp;
3677 	char attrname[EXTATTR_MAXNAMELEN];
3678 	int error;
3679 
3680 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3681 	    NULL);
3682 	if (error)
3683 		return (error);
3684 
3685 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3686 	if (error)
3687 		return (error);
3688 	vp = (struct vnode *) fp->f_data;
3689 
3690 	error = extattr_set_vp(vp, SCARG(uap, attrnamespace), attrname,
3691 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3692 
3693 	FILE_UNUSE(fp, p);
3694 	return (error);
3695 }
3696 
3697 int
3698 sys_extattr_set_file(struct lwp *l, void *v, register_t *retval)
3699 {
3700 	struct sys_extattr_set_file_args /* {
3701 		syscallarg(const char *) path;
3702 		syscallarg(int) attrnamespace;
3703 		syscallarg(const char *) attrname;
3704 		syscallarg(const void *) data;
3705 		syscallarg(size_t) nbytes;
3706 	} */ *uap = v;
3707 	struct proc *p = l->l_proc;
3708 	struct nameidata nd;
3709 	char attrname[EXTATTR_MAXNAMELEN];
3710 	int error;
3711 
3712 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3713 	    NULL);
3714 	if (error)
3715 		return (error);
3716 
3717 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3718 	error = namei(&nd);
3719 	if (error)
3720 		return (error);
3721 
3722 	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3723 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3724 
3725 	vrele(nd.ni_vp);
3726 	return (error);
3727 }
3728 
3729 int
3730 sys_extattr_set_link(struct lwp *l, void *v, register_t *retval)
3731 {
3732 	struct sys_extattr_set_link_args /* {
3733 		syscallarg(const char *) path;
3734 		syscallarg(int) attrnamespace;
3735 		syscallarg(const char *) attrname;
3736 		syscallarg(const void *) data;
3737 		syscallarg(size_t) nbytes;
3738 	} */ *uap = v;
3739 	struct proc *p = l->l_proc;
3740 	struct nameidata nd;
3741 	char attrname[EXTATTR_MAXNAMELEN];
3742 	int error;
3743 
3744 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3745 	    NULL);
3746 	if (error)
3747 		return (error);
3748 
3749 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3750 	error = namei(&nd);
3751 	if (error)
3752 		return (error);
3753 
3754 	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3755 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3756 
3757 	vrele(nd.ni_vp);
3758 	return (error);
3759 }
3760 
3761 /*
3762  * Get a named extended attribute on a file or directory.
3763  */
3764 static int
3765 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3766     void *data, size_t nbytes, struct proc *p, register_t *retval)
3767 {
3768 	struct uio auio, *auiop;
3769 	struct iovec aiov;
3770 	ssize_t cnt;
3771 	size_t size, *sizep;
3772 	int error;
3773 
3774 	VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
3775 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3776 
3777 	/*
3778 	 * Slightly unusual semantics: if the user provides a NULL data
3779 	 * pointer, they don't want to receive the data, just the maximum
3780 	 * read length.
3781 	 */
3782 	auiop = NULL;
3783 	sizep = NULL;
3784 	cnt = 0;
3785 	if (data != NULL) {
3786 		aiov.iov_base = data;
3787 		aiov.iov_len = nbytes;
3788 		auio.uio_iov = &aiov;
3789 		auio.uio_offset = 0;
3790 		if (nbytes > INT_MAX) {
3791 			error = EINVAL;
3792 			goto done;
3793 		}
3794 		auio.uio_resid = nbytes;
3795 		auio.uio_rw = UIO_READ;
3796 		auio.uio_segflg = UIO_USERSPACE;
3797 		auio.uio_procp = p;
3798 		auiop = &auio;
3799 		cnt = nbytes;
3800 	} else
3801 		sizep = &size;
3802 
3803 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, auiop, sizep,
3804 	    p->p_ucred, p);
3805 
3806 	if (auiop != NULL) {
3807 		cnt -= auio.uio_resid;
3808 		retval[0] = cnt;
3809 	} else
3810 		retval[0] = size;
3811 
3812  done:
3813 	VOP_UNLOCK(vp, 0);
3814 	return (error);
3815 }
3816 
3817 int
3818 sys_extattr_get_fd(struct lwp *l, void *v, register_t *retval)
3819 {
3820 	struct sys_extattr_get_fd_args /* {
3821 		syscallarg(int) fd;
3822 		syscallarg(int) attrnamespace;
3823 		syscallarg(const char *) attrname;
3824 		syscallarg(void *) data;
3825 		syscallarg(size_t) nbytes;
3826 	} */ *uap = v;
3827 	struct proc *p = l->l_proc;
3828 	struct file *fp;
3829 	struct vnode *vp;
3830 	char attrname[EXTATTR_MAXNAMELEN];
3831 	int error;
3832 
3833 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3834 	    NULL);
3835 	if (error)
3836 		return (error);
3837 
3838 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3839 	if (error)
3840 		return (error);
3841 	vp = (struct vnode *) fp->f_data;
3842 
3843 	error = extattr_get_vp(vp, SCARG(uap, attrnamespace), attrname,
3844 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3845 
3846 	FILE_UNUSE(fp, p);
3847 	return (error);
3848 }
3849 
3850 int
3851 sys_extattr_get_file(struct lwp *l, void *v, register_t *retval)
3852 {
3853 	struct sys_extattr_get_file_args /* {
3854 		syscallarg(const char *) path;
3855 		syscallarg(int) attrnamespace;
3856 		syscallarg(const char *) attrname;
3857 		syscallarg(void *) data;
3858 		syscallarg(size_t) nbytes;
3859 	} */ *uap = v;
3860 	struct proc *p = l->l_proc;
3861 	struct nameidata nd;
3862 	char attrname[EXTATTR_MAXNAMELEN];
3863 	int error;
3864 
3865 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3866 	    NULL);
3867 	if (error)
3868 		return (error);
3869 
3870 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3871 	error = namei(&nd);
3872 	if (error)
3873 		return (error);
3874 
3875 	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3876 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3877 
3878 	vrele(nd.ni_vp);
3879 	return (error);
3880 }
3881 
3882 int
3883 sys_extattr_get_link(struct lwp *l, void *v, register_t *retval)
3884 {
3885 	struct sys_extattr_get_link_args /* {
3886 		syscallarg(const char *) path;
3887 		syscallarg(int) attrnamespace;
3888 		syscallarg(const char *) attrname;
3889 		syscallarg(void *) data;
3890 		syscallarg(size_t) nbytes;
3891 	} */ *uap = v;
3892 	struct proc *p = l->l_proc;
3893 	struct nameidata nd;
3894 	char attrname[EXTATTR_MAXNAMELEN];
3895 	int error;
3896 
3897 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3898 	    NULL);
3899 	if (error)
3900 		return (error);
3901 
3902 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3903 	error = namei(&nd);
3904 	if (error)
3905 		return (error);
3906 
3907 	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3908 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3909 
3910 	vrele(nd.ni_vp);
3911 	return (error);
3912 }
3913 
3914 /*
3915  * Delete a named extended attribute on a file or directory.
3916  */
3917 static int
3918 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3919     struct proc *p)
3920 {
3921 	struct mount *mp;
3922 	int error;
3923 
3924 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3925 	if (error)
3926 		return (error);
3927 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3928 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3929 
3930 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, p->p_ucred, p);
3931 	if (error == EOPNOTSUPP)
3932 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
3933 		    p->p_ucred, p);
3934 
3935 	VOP_UNLOCK(vp, 0);
3936 	vn_finished_write(mp, 0);
3937 	return (error);
3938 }
3939 
3940 int
3941 sys_extattr_delete_fd(struct lwp *l, void *v, register_t *retval)
3942 {
3943 	struct sys_extattr_delete_fd_args /* {
3944 		syscallarg(int) fd;
3945 		syscallarg(int) attrnamespace;
3946 		syscallarg(const char *) attrname;
3947 	} */ *uap = v;
3948 	struct proc *p = l->l_proc;
3949 	struct file *fp;
3950 	struct vnode *vp;
3951 	char attrname[EXTATTR_MAXNAMELEN];
3952 	int error;
3953 
3954 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3955 	    NULL);
3956 	if (error)
3957 		return (error);
3958 
3959 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3960 	if (error)
3961 		return (error);
3962 	vp = (struct vnode *) fp->f_data;
3963 
3964 	error = extattr_delete_vp(vp, SCARG(uap, attrnamespace), attrname, p);
3965 
3966 	FILE_UNUSE(fp, p);
3967 	return (error);
3968 }
3969 
3970 int
3971 sys_extattr_delete_file(struct lwp *l, void *v, register_t *retval)
3972 {
3973 	struct sys_extattr_delete_file_args /* {
3974 		syscallarg(const char *) path;
3975 		syscallarg(int) attrnamespace;
3976 		syscallarg(const char *) attrname;
3977 	} */ *uap = v;
3978 	struct proc *p = l->l_proc;
3979 	struct nameidata nd;
3980 	char attrname[EXTATTR_MAXNAMELEN];
3981 	int error;
3982 
3983 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3984 	    NULL);
3985 	if (error)
3986 		return (error);
3987 
3988 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3989 	error = namei(&nd);
3990 	if (error)
3991 		return (error);
3992 
3993 	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3994 	    p);
3995 
3996 	vrele(nd.ni_vp);
3997 	return (error);
3998 }
3999 
4000 int
4001 sys_extattr_delete_link(struct lwp *l, void *v, register_t *retval)
4002 {
4003 	struct sys_extattr_delete_link_args /* {
4004 		syscallarg(const char *) path;
4005 		syscallarg(int) attrnamespace;
4006 		syscallarg(const char *) attrname;
4007 	} */ *uap = v;
4008 	struct proc *p = l->l_proc;
4009 	struct nameidata nd;
4010 	char attrname[EXTATTR_MAXNAMELEN];
4011 	int error;
4012 
4013 	error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
4014 	    NULL);
4015 	if (error)
4016 		return (error);
4017 
4018 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4019 	error = namei(&nd);
4020 	if (error)
4021 		return (error);
4022 
4023 	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4024 	    p);
4025 
4026 	vrele(nd.ni_vp);
4027 	return (error);
4028 }
4029 
4030 /*
4031  * Retrieve a list of extended attributes on a file or directory.
4032  */
4033 static int
4034 extattr_list_vp(struct vnode *vp, int attrnamespace, void *data, size_t nbytes,
4035     struct proc *p, register_t *retval)
4036 {
4037 	struct uio auio, *auiop;
4038 	size_t size, *sizep;
4039 	struct iovec aiov;
4040 	ssize_t cnt;
4041 	int error;
4042 
4043 	VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
4044 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4045 
4046 	auiop = NULL;
4047 	sizep = NULL;
4048 	cnt = 0;
4049 	if (data != NULL) {
4050 		aiov.iov_base = data;
4051 		aiov.iov_len = nbytes;
4052 		auio.uio_iov = &aiov;
4053 		auio.uio_offset = 0;
4054 		if (nbytes > INT_MAX) {
4055 			error = EINVAL;
4056 			goto done;
4057 		}
4058 		auio.uio_resid = nbytes;
4059 		auio.uio_rw = UIO_READ;
4060 		auio.uio_segflg = UIO_USERSPACE;
4061 		auio.uio_procp = p;
4062 		auiop = &auio;
4063 		cnt = nbytes;
4064 	} else
4065 		sizep = &size;
4066 
4067 	error = VOP_LISTEXTATTR(vp, attrnamespace, auiop, sizep,
4068 	    p->p_ucred, p);
4069 
4070 	if (auiop != NULL) {
4071 		cnt -= auio.uio_resid;
4072 		retval[0] = cnt;
4073 	} else
4074 		retval[0] = size;
4075 
4076  done:
4077 	VOP_UNLOCK(vp, 0);
4078 	return (error);
4079 }
4080 
4081 int
4082 sys_extattr_list_fd(struct lwp *l, void *v, register_t *retval)
4083 {
4084 	struct sys_extattr_list_fd_args /* {
4085 		syscallarg(int) fd;
4086 		syscallarg(int) attrnamespace;
4087 		syscallarg(void *) data;
4088 		syscallarg(size_t) nbytes;
4089 	} */ *uap = v;
4090 	struct proc *p = l->l_proc;
4091 	struct file *fp;
4092 	struct vnode *vp;
4093 	int error;
4094 
4095 	error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
4096 	if (error)
4097 		return (error);
4098 	vp = (struct vnode *) fp->f_data;
4099 
4100 	error = extattr_list_vp(vp, SCARG(uap, attrnamespace),
4101 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4102 
4103 	FILE_UNUSE(fp, p);
4104 	return (error);
4105 }
4106 
4107 int
4108 sys_extattr_list_file(struct lwp *l, void *v, register_t *retval)
4109 {
4110 	struct sys_extattr_list_file_args /* {
4111 		syscallarg(const char *) path;
4112 		syscallarg(int) attrnamespace;
4113 		syscallarg(void *) data;
4114 		syscallarg(size_t) nbytes;
4115 	} */ *uap = v;
4116 	struct proc *p = l->l_proc;
4117 	struct nameidata nd;
4118 	int error;
4119 
4120 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4121 	error = namei(&nd);
4122 	if (error)
4123 		return (error);
4124 
4125 	error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4126 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4127 
4128 	vrele(nd.ni_vp);
4129 	return (error);
4130 }
4131 
4132 int
4133 sys_extattr_list_link(struct lwp *l, void *v, register_t *retval)
4134 {
4135 	struct sys_extattr_list_link_args /* {
4136 		syscallarg(const char *) path;
4137 		syscallarg(int) attrnamespace;
4138 		syscallarg(void *) data;
4139 		syscallarg(size_t) nbytes;
4140 	} */ *uap = v;
4141 	struct proc *p = l->l_proc;
4142 	struct nameidata nd;
4143 	int error;
4144 
4145 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4146 	error = namei(&nd);
4147 	if (error)
4148 		return (error);
4149 
4150 	error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4151 	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4152 
4153 	vrele(nd.ni_vp);
4154 	return (error);
4155 }
4156