xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 8a5e2a50be13e77dd4df5daf258ddceeeeb47ce6)
1 /*	$NetBSD: vfs_syscalls.c,v 1.226 2005/07/16 22:47:18 christos 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 1.226 2005/07/16 22:47:18 christos Exp $");
41 
42 #include "opt_compat_netbsd.h"
43 #include "opt_compat_43.h"
44 #include "opt_ktrace.h"
45 #include "opt_verified_exec.h"
46 #include "fss.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/namei.h>
51 #include <sys/filedesc.h>
52 #include <sys/kernel.h>
53 #include <sys/file.h>
54 #include <sys/stat.h>
55 #include <sys/vnode.h>
56 #include <sys/mount.h>
57 #include <sys/proc.h>
58 #include <sys/uio.h>
59 #include <sys/malloc.h>
60 #include <sys/dirent.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 #ifdef VERIFIED_EXEC
68 #include <sys/verified_exec.h>
69 #endif
70 
71 #include <miscfs/genfs/genfs.h>
72 #include <miscfs/syncfs/syncfs.h>
73 
74 #if NFSS > 0
75 #include <dev/fssvar.h>
76 #endif
77 
78 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
79 
80 static int change_dir(struct nameidata *, struct proc *);
81 static int change_flags(struct vnode *, u_long, struct proc *);
82 static int change_mode(struct vnode *, int, struct proc *p);
83 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int);
84 static int change_utimes(struct vnode *vp, const struct timeval *,
85 	       struct proc *p);
86 static int rename_files(const char *, const char *, struct proc *, int);
87 
88 void checkdirs(struct vnode *);
89 
90 int dovfsusermount = 0;
91 
92 /*
93  * Virtual File System System Calls
94  */
95 
96 /*
97  * Mount a file system.
98  */
99 
100 #if defined(COMPAT_09) || defined(COMPAT_43)
101 /*
102  * This table is used to maintain compatibility with 4.3BSD
103  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
104  *
105  * Do not modify this table. It should only contain filesystems
106  * supported by NetBSD 0.9 and 4.3BSD.
107  */
108 const char * const mountcompatnames[] = {
109 	NULL,		/* 0 = MOUNT_NONE */
110 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
111 	MOUNT_NFS,	/* 2 */
112 	MOUNT_MFS,	/* 3 */
113 	MOUNT_MSDOS,	/* 4 */
114 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
115 	MOUNT_FDESC,	/* 6 */
116 	MOUNT_KERNFS,	/* 7 */
117 	NULL,		/* 8 = MOUNT_DEVFS */
118 	MOUNT_AFS,	/* 9 */
119 };
120 const int nmountcompatnames = sizeof(mountcompatnames) /
121     sizeof(mountcompatnames[0]);
122 #endif /* COMPAT_09 || COMPAT_43 */
123 
124 /* ARGSUSED */
125 int
126 sys_mount(struct lwp *l, void *v, 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 	/*
144 	 * if MNT_GETARGS is specified, it should be only flag.
145 	 */
146 
147 	if ((SCARG(uap, flags) & MNT_GETARGS) != 0 &&
148 	    (SCARG(uap, flags) & ~MNT_GETARGS) != 0) {
149 		return EINVAL;
150 	}
151 
152 	if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
153 	    (error = suser(p->p_ucred, &p->p_acflag)))
154 		return (error);
155 	/*
156 	 * Get vnode to be covered
157 	 */
158 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
159 	    SCARG(uap, path), p);
160 	if ((error = namei(&nd)) != 0)
161 		return (error);
162 	vp = nd.ni_vp;
163 	/*
164 	 * A lookup in VFS_MOUNT might result in an attempt to
165 	 * lock this vnode again, so make the lock recursive.
166 	 */
167 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
168 	if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
169 		if ((vp->v_flag & VROOT) == 0) {
170 			vput(vp);
171 			return (EINVAL);
172 		}
173 		mp = vp->v_mount;
174 		flag = mp->mnt_flag;
175 		vfs = mp->mnt_op;
176 		/*
177 		 * We only allow the filesystem to be reloaded if it
178 		 * is currently mounted read-only.
179 		 */
180 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
181 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
182 			vput(vp);
183 			return (EOPNOTSUPP);	/* Needs translation */
184 		}
185 		/*
186 		 * In "highly secure" mode, don't let the caller do anything
187 		 * but downgrade a filesystem from read-write to read-only.
188 		 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
189 		 */
190 		if (securelevel >= 2 &&
191 		    SCARG(uap, flags) != MNT_GETARGS &&
192 		    SCARG(uap, flags) !=
193 		    (mp->mnt_flag | MNT_RDONLY |
194 		     MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
195 			vput(vp);
196 			return (EPERM);
197 		}
198 		mp->mnt_flag |= SCARG(uap, flags) &
199 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
200 		/*
201 		 * Only root, or the user that did the original mount is
202 		 * permitted to update it.
203 		 */
204 		if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
205 		    mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
206 		    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
207 			vput(vp);
208 			return (error);
209 		}
210 		/*
211 		 * Do not allow NFS export by non-root users. For non-root
212 		 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
213 		 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
214 		 */
215 		if (p->p_ucred->cr_uid != 0) {
216 			if (SCARG(uap, flags) & MNT_EXPORTED) {
217 				vput(vp);
218 				return (EPERM);
219 			}
220 			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
221 			if (flag & MNT_NOEXEC)
222 				SCARG(uap, flags) |= MNT_NOEXEC;
223 		}
224 		if (vfs_busy(mp, LK_NOWAIT, 0)) {
225 			vput(vp);
226 			return (EPERM);
227 		}
228 		goto update;
229 	} else {
230 		if (securelevel >= 2) {
231 			vput(vp);
232 			return (EPERM);
233 		}
234 	}
235 	/*
236 	 * If the user is not root, ensure that they own the directory
237 	 * onto which we are attempting to mount.
238 	 */
239 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
240 	    (va.va_uid != p->p_ucred->cr_uid &&
241 		(error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
242 		vput(vp);
243 		return (error);
244 	}
245 	/*
246 	 * Do not allow NFS export by non-root users. For non-root users,
247 	 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
248 	 * mount point is already MNT_NOEXEC.
249 	 */
250 	if (p->p_ucred->cr_uid != 0) {
251 		if (SCARG(uap, flags) & MNT_EXPORTED) {
252 			vput(vp);
253 			return (EPERM);
254 		}
255 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
256 		if (vp->v_mount->mnt_flag & MNT_NOEXEC)
257 			SCARG(uap, flags) |= MNT_NOEXEC;
258 	}
259 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) {
260 		vput(vp);
261 		return (error);
262 	}
263 	if (vp->v_type != VDIR) {
264 		vput(vp);
265 		return (ENOTDIR);
266 	}
267 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
268 	if (error) {
269 #if defined(COMPAT_09) || defined(COMPAT_43)
270 		/*
271 		 * Historically filesystem types were identified by number.
272 		 * If we get an integer for the filesystem type instead of a
273 		 * string, we check to see if it matches one of the historic
274 		 * filesystem types.
275 		 */
276 		u_long fsindex = (u_long)SCARG(uap, type);
277 		if (fsindex >= nmountcompatnames ||
278 		    mountcompatnames[fsindex] == NULL) {
279 			vput(vp);
280 			return (ENODEV);
281 		}
282 		strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
283 #else
284 		vput(vp);
285 		return (error);
286 #endif
287 	}
288 #ifdef	COMPAT_10
289 	/* Accept `ufs' as an alias for `ffs'. */
290 	if (!strncmp(fstypename, "ufs", MFSNAMELEN))
291 		strncpy(fstypename, "ffs", MFSNAMELEN);
292 #endif
293 	if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
294 		vput(vp);
295 		return (ENODEV);
296 	}
297 	if (vp->v_mountedhere != NULL) {
298 		vput(vp);
299 		return (EBUSY);
300 	}
301 
302 	/*
303 	 * Allocate and initialize the file system.
304 	 */
305 	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
306 		M_MOUNT, M_WAITOK);
307 	memset((char *)mp, 0, (u_long)sizeof(struct mount));
308 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
309 	simple_lock_init(&mp->mnt_slock);
310 	(void)vfs_busy(mp, LK_NOWAIT, 0);
311 	mp->mnt_op = vfs;
312 	vfs->vfs_refcount++;
313 	mp->mnt_vnodecovered = vp;
314 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
315 	mp->mnt_unmounter = NULL;
316 	mp->mnt_leaf = mp;
317 
318 	/*
319 	 * The underlying file system may refuse the mount for
320 	 * various reasons.  Allow the user to force it to happen.
321 	 */
322 	mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE;
323  update:
324 	if ((SCARG(uap, flags) & MNT_GETARGS) == 0) {
325 		/*
326 		 * Set the mount level flags.
327 		 */
328 		if (SCARG(uap, flags) & MNT_RDONLY)
329 			mp->mnt_flag |= MNT_RDONLY;
330 		else if (mp->mnt_flag & MNT_RDONLY)
331 			mp->mnt_iflag |= IMNT_WANTRDWR;
332 		mp->mnt_flag &=
333 		  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
334 		    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
335 		    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
336 		    MNT_MAGICLINKS);
337 		mp->mnt_flag |= SCARG(uap, flags) &
338 		   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
339 		    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
340 		    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
341 		    MNT_IGNORE | MNT_MAGICLINKS);
342 	}
343 	/*
344 	 * Mount the filesystem.
345 	 */
346 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
347 	if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
348 		if (mp->mnt_iflag & IMNT_WANTRDWR)
349 			mp->mnt_flag &= ~MNT_RDONLY;
350 		if (error)
351 			mp->mnt_flag = flag;
352 		mp->mnt_flag &=~
353 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
354 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
355 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
356 			if (mp->mnt_syncer == NULL)
357 				error = vfs_allocate_syncvnode(mp);
358 		} else {
359 			if (mp->mnt_syncer != NULL)
360 				vfs_deallocate_syncvnode(mp);
361 		}
362 		vfs_unbusy(mp);
363 		VOP_UNLOCK(vp, 0);
364 		vrele(vp);
365 		return (error);
366 	}
367 	/*
368 	 * Put the new filesystem on the mount list after root.
369 	 */
370 	cache_purge(vp);
371 	if (!error) {
372 		mp->mnt_flag &=~
373 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
374 		mp->mnt_iflag &=~ IMNT_WANTRDWR;
375 		vp->v_mountedhere = mp;
376 		simple_lock(&mountlist_slock);
377 		CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
378 		simple_unlock(&mountlist_slock);
379 		checkdirs(vp);
380 		VOP_UNLOCK(vp, 0);
381 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
382 			error = vfs_allocate_syncvnode(mp);
383 		vfs_unbusy(mp);
384 		(void) VFS_STATVFS(mp, &mp->mnt_stat, p);
385 		if ((error = VFS_START(mp, 0, p)))
386 			vrele(vp);
387 	} else {
388 		vp->v_mountedhere = (struct mount *)0;
389 		vfs->vfs_refcount--;
390 		vfs_unbusy(mp);
391 		free(mp, M_MOUNT);
392 		vput(vp);
393 	}
394 	return (error);
395 }
396 
397 /*
398  * Scan all active processes to see if any of them have a current
399  * or root directory onto which the new filesystem has just been
400  * mounted. If so, replace them with the new mount point.
401  */
402 void
403 checkdirs(struct vnode *olddp)
404 {
405 	struct cwdinfo *cwdi;
406 	struct vnode *newdp;
407 	struct proc *p;
408 
409 	if (olddp->v_usecount == 1)
410 		return;
411 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
412 		panic("mount: lost mount");
413 	proclist_lock_read();
414 	PROCLIST_FOREACH(p, &allproc) {
415 		cwdi = p->p_cwdi;
416 		if (!cwdi)
417 			continue;
418 		if (cwdi->cwdi_cdir == olddp) {
419 			vrele(cwdi->cwdi_cdir);
420 			VREF(newdp);
421 			cwdi->cwdi_cdir = newdp;
422 		}
423 		if (cwdi->cwdi_rdir == olddp) {
424 			vrele(cwdi->cwdi_rdir);
425 			VREF(newdp);
426 			cwdi->cwdi_rdir = newdp;
427 		}
428 	}
429 	proclist_unlock_read();
430 	if (rootvnode == olddp) {
431 		vrele(rootvnode);
432 		VREF(newdp);
433 		rootvnode = newdp;
434 	}
435 	vput(newdp);
436 }
437 
438 /*
439  * Unmount a file system.
440  *
441  * Note: unmount takes a path to the vnode mounted on as argument,
442  * not special file (as before).
443  */
444 /* ARGSUSED */
445 int
446 sys_unmount(struct lwp *l, void *v, register_t *retval)
447 {
448 	struct sys_unmount_args /* {
449 		syscallarg(const char *) path;
450 		syscallarg(int) flags;
451 	} */ *uap = v;
452 	struct proc *p = l->l_proc;
453 	struct vnode *vp;
454 	struct mount *mp;
455 	int error;
456 	struct nameidata nd;
457 
458 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
459 	    SCARG(uap, path), p);
460 	if ((error = namei(&nd)) != 0)
461 		return (error);
462 	vp = nd.ni_vp;
463 	mp = vp->v_mount;
464 
465 	/*
466 	 * Only root, or the user that did the original mount is
467 	 * permitted to unmount this filesystem.
468 	 */
469 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
470 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
471 		vput(vp);
472 		return (error);
473 	}
474 
475 	/*
476 	 * Don't allow unmounting the root file system.
477 	 */
478 	if (mp->mnt_flag & MNT_ROOTFS) {
479 		vput(vp);
480 		return (EINVAL);
481 	}
482 
483 	/*
484 	 * Must be the root of the filesystem
485 	 */
486 	if ((vp->v_flag & VROOT) == 0) {
487 		vput(vp);
488 		return (EINVAL);
489 	}
490 	vput(vp);
491 
492 	/*
493 	 * XXX Freeze syncer.  Must do this before locking the
494 	 * mount point.  See dounmount() for details.
495 	 */
496 	lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
497 
498 	if (vfs_busy(mp, 0, 0)) {
499 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
500 		return (EBUSY);
501 	}
502 
503 	return (dounmount(mp, SCARG(uap, flags), p));
504 }
505 
506 /*
507  * Do the actual file system unmount. File system is assumed to have been
508  * marked busy by the caller.
509  */
510 int
511 dounmount(struct mount *mp, int flags, struct proc *p)
512 {
513 	struct vnode *coveredvp;
514 	int error;
515 	int async;
516 	int used_syncer;
517 
518 	simple_lock(&mountlist_slock);
519 	vfs_unbusy(mp);
520 	used_syncer = (mp->mnt_syncer != NULL);
521 
522 	/*
523 	 * XXX Syncer must be frozen when we get here.  This should really
524 	 * be done on a per-mountpoint basis, but especially the softdep
525 	 * code possibly called from the syncer doens't exactly work on a
526 	 * per-mountpoint basis, so the softdep code would become a maze
527 	 * of vfs_busy() calls.
528 	 *
529 	 * The caller of dounmount() must acquire syncer_lock because
530 	 * the syncer itself acquires locks in syncer_lock -> vfs_busy
531 	 * order, and we must preserve that order to avoid deadlock.
532 	 *
533 	 * So, if the file system did not use the syncer, now is
534 	 * the time to release the syncer_lock.
535 	 */
536 	if (used_syncer == 0)
537 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
538 
539 	mp->mnt_iflag |= IMNT_UNMOUNT;
540 	mp->mnt_unmounter = p;
541 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
542 	vn_start_write(NULL, &mp, V_WAIT);
543 
544 	if (mp->mnt_flag & MNT_EXPUBLIC)
545 		vfs_setpublicfs(NULL, NULL, NULL);
546 	async = mp->mnt_flag & MNT_ASYNC;
547 	mp->mnt_flag &= ~MNT_ASYNC;
548 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
549 	if (mp->mnt_syncer != NULL)
550 		vfs_deallocate_syncvnode(mp);
551 	error = 0;
552 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
553 #if NFSS > 0
554 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
555 #endif
556 		if (error == 0)
557 			error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
558 	}
559 	if (error == 0 || (flags & MNT_FORCE))
560 		error = VFS_UNMOUNT(mp, flags, p);
561 	vn_finished_write(mp, 0);
562 	simple_lock(&mountlist_slock);
563 	if (error) {
564 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
565 			(void) vfs_allocate_syncvnode(mp);
566 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
567 		mp->mnt_unmounter = NULL;
568 		mp->mnt_flag |= async;
569 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
570 		    &mountlist_slock);
571 		if (used_syncer)
572 			lockmgr(&syncer_lock, LK_RELEASE, NULL);
573 		simple_lock(&mp->mnt_slock);
574 		while (mp->mnt_wcnt > 0) {
575 			wakeup(mp);
576 			ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
577 				0, &mp->mnt_slock);
578 		}
579 		simple_unlock(&mp->mnt_slock);
580 		return (error);
581 	}
582 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
583 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
584 		coveredvp->v_mountedhere = NULL;
585 		vrele(coveredvp);
586 	}
587 	mp->mnt_op->vfs_refcount--;
588 	if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
589 		panic("unmount: dangling vnode");
590 	mp->mnt_iflag |= IMNT_GONE;
591 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
592 	if (used_syncer)
593 		lockmgr(&syncer_lock, LK_RELEASE, NULL);
594 	simple_lock(&mp->mnt_slock);
595 	while (mp->mnt_wcnt > 0) {
596 		wakeup(mp);
597 		ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
598 	}
599 	simple_unlock(&mp->mnt_slock);
600 	free(mp, M_MOUNT);
601 	return (0);
602 }
603 
604 /*
605  * Sync each mounted filesystem.
606  */
607 #ifdef DEBUG
608 int syncprt = 0;
609 struct ctldebug debug0 = { "syncprt", &syncprt };
610 #endif
611 
612 /* ARGSUSED */
613 int
614 sys_sync(struct lwp *l, void *v, register_t *retval)
615 {
616 	struct mount *mp, *nmp;
617 	int asyncflag;
618 	struct proc *p = l == NULL ? &proc0 : l->l_proc;
619 
620 	simple_lock(&mountlist_slock);
621 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
622 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
623 			nmp = mp->mnt_list.cqe_prev;
624 			continue;
625 		}
626 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
627 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
628 			asyncflag = mp->mnt_flag & MNT_ASYNC;
629 			mp->mnt_flag &= ~MNT_ASYNC;
630 			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
631 			if (asyncflag)
632 				 mp->mnt_flag |= MNT_ASYNC;
633 			vn_finished_write(mp, 0);
634 		}
635 		simple_lock(&mountlist_slock);
636 		nmp = mp->mnt_list.cqe_prev;
637 		vfs_unbusy(mp);
638 
639 	}
640 	simple_unlock(&mountlist_slock);
641 #ifdef DEBUG
642 	if (syncprt)
643 		vfs_bufstats();
644 #endif /* DEBUG */
645 	return (0);
646 }
647 
648 /*
649  * Change filesystem quotas.
650  */
651 /* ARGSUSED */
652 int
653 sys_quotactl(struct lwp *l, void *v, register_t *retval)
654 {
655 	struct sys_quotactl_args /* {
656 		syscallarg(const char *) path;
657 		syscallarg(int) cmd;
658 		syscallarg(int) uid;
659 		syscallarg(caddr_t) arg;
660 	} */ *uap = v;
661 	struct proc *p = l->l_proc;
662 	struct mount *mp;
663 	int error;
664 	struct nameidata nd;
665 
666 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
667 	if ((error = namei(&nd)) != 0)
668 		return (error);
669 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
670 	vrele(nd.ni_vp);
671 	if (error)
672 		return (error);
673 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
674 	    SCARG(uap, arg), p);
675 	vn_finished_write(mp, 0);
676 	return (error);
677 }
678 
679 int
680 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags,
681     int root)
682 {
683 	struct cwdinfo *cwdi = p->p_cwdi;
684 	int error = 0;
685 
686 	/*
687 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
688 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
689 	 * overrides MNT_NOWAIT.
690 	 */
691 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
692 	    (flags != MNT_WAIT && flags != 0)) {
693 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
694 		goto done;
695 	}
696 
697 	/* Get the filesystem stats now */
698 	memset(sp, 0, sizeof(*sp));
699 	if ((error = VFS_STATVFS(mp, sp, p)) != 0) {
700 		return error;
701 	}
702 
703 	if (cwdi->cwdi_rdir == NULL)
704 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
705 done:
706 	if (cwdi->cwdi_rdir != NULL) {
707 		size_t len;
708 		char *bp;
709 		char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
710 		if (!path)
711 			return ENOMEM;
712 
713 		bp = path + MAXPATHLEN;
714 		*--bp = '\0';
715 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
716 		    MAXPATHLEN / 2, 0, p);
717 		if (error) {
718 			free(path, M_TEMP);
719 			return error;
720 		}
721 		len = strlen(bp);
722 		/*
723 		 * for mount points that are below our root, we can see
724 		 * them, so we fix up the pathname and return them. The
725 		 * rest we cannot see, so we don't allow viewing the
726 		 * data.
727 		 */
728 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
729 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
730 			    sizeof(sp->f_mntonname));
731 			if (sp->f_mntonname[0] == '\0')
732 				(void)strlcpy(sp->f_mntonname, "/",
733 				    sizeof(sp->f_mntonname));
734 		} else {
735 			if (root)
736 				(void)strlcpy(sp->f_mntonname, "/",
737 				    sizeof(sp->f_mntonname));
738 			else
739 				error = EPERM;
740 		}
741 		free(path, M_TEMP);
742 	}
743 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
744 	return error;
745 }
746 
747 /*
748  * Get filesystem statistics.
749  */
750 /* ARGSUSED */
751 int
752 sys_statvfs1(struct lwp *l, void *v, register_t *retval)
753 {
754 	struct sys_statvfs1_args /* {
755 		syscallarg(const char *) path;
756 		syscallarg(struct statvfs *) buf;
757 		syscallarg(int) flags;
758 	} */ *uap = v;
759 	struct proc *p = l->l_proc;
760 	struct mount *mp;
761 	struct statvfs sbuf;
762 	int error;
763 	struct nameidata nd;
764 
765 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
766 	if ((error = namei(&nd)) != 0)
767 		return error;
768 	mp = nd.ni_vp->v_mount;
769 	vrele(nd.ni_vp);
770 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
771 		return error;
772 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
773 }
774 
775 /*
776  * Get filesystem statistics.
777  */
778 /* ARGSUSED */
779 int
780 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval)
781 {
782 	struct sys_fstatvfs1_args /* {
783 		syscallarg(int) fd;
784 		syscallarg(struct statvfs *) buf;
785 		syscallarg(int) flags;
786 	} */ *uap = v;
787 	struct proc *p = l->l_proc;
788 	struct file *fp;
789 	struct mount *mp;
790 	struct statvfs sbuf;
791 	int error;
792 
793 	/* getvnode() will use the descriptor for us */
794 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
795 		return (error);
796 	mp = ((struct vnode *)fp->f_data)->v_mount;
797 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
798 		goto out;
799 	error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
800  out:
801 	FILE_UNUSE(fp, p);
802 	return error;
803 }
804 
805 
806 /*
807  * Get statistics on all filesystems.
808  */
809 int
810 sys_getvfsstat(struct lwp *l, void *v, register_t *retval)
811 {
812 	struct sys_getvfsstat_args /* {
813 		syscallarg(struct statvfs *) buf;
814 		syscallarg(size_t) bufsize;
815 		syscallarg(int) flags;
816 	} */ *uap = v;
817 	int root = 0;
818 	struct proc *p = l->l_proc;
819 	struct mount *mp, *nmp;
820 	struct statvfs sbuf;
821 	struct statvfs *sfsp;
822 	size_t count, maxcount;
823 	int error = 0;
824 
825 	maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs);
826 	sfsp = SCARG(uap, buf);
827 	simple_lock(&mountlist_slock);
828 	count = 0;
829 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
830 	     mp = nmp) {
831 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
832 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
833 			continue;
834 		}
835 		if (sfsp && count < maxcount) {
836 			error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0);
837 			if (error) {
838 				simple_lock(&mountlist_slock);
839 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
840 				vfs_unbusy(mp);
841 				continue;
842 			}
843 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
844 			if (error) {
845 				vfs_unbusy(mp);
846 				return (error);
847 			}
848 			sfsp++;
849 			root |= strcmp(sbuf.f_mntonname, "/") == 0;
850 		}
851 		count++;
852 		simple_lock(&mountlist_slock);
853 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
854 		vfs_unbusy(mp);
855 	}
856 	simple_unlock(&mountlist_slock);
857 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
858 		/*
859 		 * fake a root entry
860 		 */
861 		if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p,
862 		    SCARG(uap, flags), 1)) != 0)
863 			return error;
864 		if (sfsp)
865 			error = copyout(&sbuf, sfsp, sizeof(*sfsp));
866 		count++;
867 	}
868 	if (sfsp && count > maxcount)
869 		*retval = maxcount;
870 	else
871 		*retval = count;
872 	return error;
873 }
874 
875 /*
876  * Change current working directory to a given file descriptor.
877  */
878 /* ARGSUSED */
879 int
880 sys_fchdir(struct lwp *l, void *v, register_t *retval)
881 {
882 	struct sys_fchdir_args /* {
883 		syscallarg(int) fd;
884 	} */ *uap = v;
885 	struct proc *p = l->l_proc;
886 	struct filedesc *fdp = p->p_fd;
887 	struct cwdinfo *cwdi = p->p_cwdi;
888 	struct vnode *vp, *tdp;
889 	struct mount *mp;
890 	struct file *fp;
891 	int error;
892 
893 	/* getvnode() will use the descriptor for us */
894 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
895 		return (error);
896 	vp = (struct vnode *)fp->f_data;
897 
898 	VREF(vp);
899 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
900 	if (vp->v_type != VDIR)
901 		error = ENOTDIR;
902 	else
903 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
904 	while (!error && (mp = vp->v_mountedhere) != NULL) {
905 		if (vfs_busy(mp, 0, 0))
906 			continue;
907 		error = VFS_ROOT(mp, &tdp);
908 		vfs_unbusy(mp);
909 		if (error)
910 			break;
911 		vput(vp);
912 		vp = tdp;
913 	}
914 	if (error) {
915 		vput(vp);
916 		goto out;
917 	}
918 	VOP_UNLOCK(vp, 0);
919 
920 	/*
921 	 * Disallow changing to a directory not under the process's
922 	 * current root directory (if there is one).
923 	 */
924 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
925 		vrele(vp);
926 		error = EPERM;	/* operation not permitted */
927 		goto out;
928 	}
929 
930 	vrele(cwdi->cwdi_cdir);
931 	cwdi->cwdi_cdir = vp;
932  out:
933 	FILE_UNUSE(fp, p);
934 	return (error);
935 }
936 
937 /*
938  * Change this process's notion of the root directory to a given file
939  * descriptor.
940  */
941 int
942 sys_fchroot(struct lwp *l, void *v, register_t *retval)
943 {
944 	struct sys_fchroot_args *uap = v;
945 	struct proc *p = l->l_proc;
946 	struct filedesc *fdp = p->p_fd;
947 	struct cwdinfo *cwdi = p->p_cwdi;
948 	struct vnode	*vp;
949 	struct file	*fp;
950 	int		 error;
951 
952 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
953 		return error;
954 	/* getvnode() will use the descriptor for us */
955 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
956 		return error;
957 	vp = (struct vnode *) fp->f_data;
958 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
959 	if (vp->v_type != VDIR)
960 		error = ENOTDIR;
961 	else
962 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
963 	VOP_UNLOCK(vp, 0);
964 	if (error)
965 		goto out;
966 	VREF(vp);
967 
968 	/*
969 	 * Prevent escaping from chroot by putting the root under
970 	 * the working directory.  Silently chdir to / if we aren't
971 	 * already there.
972 	 */
973 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
974 		/*
975 		 * XXX would be more failsafe to change directory to a
976 		 * deadfs node here instead
977 		 */
978 		vrele(cwdi->cwdi_cdir);
979 		VREF(vp);
980 		cwdi->cwdi_cdir = vp;
981 	}
982 
983 	if (cwdi->cwdi_rdir != NULL)
984 		vrele(cwdi->cwdi_rdir);
985 	cwdi->cwdi_rdir = vp;
986  out:
987 	FILE_UNUSE(fp, p);
988 	return (error);
989 }
990 
991 /*
992  * Change current working directory (``.'').
993  */
994 /* ARGSUSED */
995 int
996 sys_chdir(struct lwp *l, void *v, register_t *retval)
997 {
998 	struct sys_chdir_args /* {
999 		syscallarg(const char *) path;
1000 	} */ *uap = v;
1001 	struct proc *p = l->l_proc;
1002 	struct cwdinfo *cwdi = p->p_cwdi;
1003 	int error;
1004 	struct nameidata nd;
1005 
1006 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1007 	    SCARG(uap, path), p);
1008 	if ((error = change_dir(&nd, p)) != 0)
1009 		return (error);
1010 	vrele(cwdi->cwdi_cdir);
1011 	cwdi->cwdi_cdir = nd.ni_vp;
1012 	return (0);
1013 }
1014 
1015 /*
1016  * Change notion of root (``/'') directory.
1017  */
1018 /* ARGSUSED */
1019 int
1020 sys_chroot(struct lwp *l, void *v, register_t *retval)
1021 {
1022 	struct sys_chroot_args /* {
1023 		syscallarg(const char *) path;
1024 	} */ *uap = v;
1025 	struct proc *p = l->l_proc;
1026 	struct cwdinfo *cwdi = p->p_cwdi;
1027 	struct vnode *vp;
1028 	int error;
1029 	struct nameidata nd;
1030 
1031 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1032 		return (error);
1033 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1034 	    SCARG(uap, path), p);
1035 	if ((error = change_dir(&nd, p)) != 0)
1036 		return (error);
1037 	if (cwdi->cwdi_rdir != NULL)
1038 		vrele(cwdi->cwdi_rdir);
1039 	vp = nd.ni_vp;
1040 	cwdi->cwdi_rdir = vp;
1041 
1042 	/*
1043 	 * Prevent escaping from chroot by putting the root under
1044 	 * the working directory.  Silently chdir to / if we aren't
1045 	 * already there.
1046 	 */
1047 	if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1048 		/*
1049 		 * XXX would be more failsafe to change directory to a
1050 		 * deadfs node here instead
1051 		 */
1052 		vrele(cwdi->cwdi_cdir);
1053 		VREF(vp);
1054 		cwdi->cwdi_cdir = vp;
1055 	}
1056 
1057 	return (0);
1058 }
1059 
1060 /*
1061  * Common routine for chroot and chdir.
1062  */
1063 static int
1064 change_dir(struct nameidata *ndp, struct proc *p)
1065 {
1066 	struct vnode *vp;
1067 	int error;
1068 
1069 	if ((error = namei(ndp)) != 0)
1070 		return (error);
1071 	vp = ndp->ni_vp;
1072 	if (vp->v_type != VDIR)
1073 		error = ENOTDIR;
1074 	else
1075 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
1076 
1077 	if (error)
1078 		vput(vp);
1079 	else
1080 		VOP_UNLOCK(vp, 0);
1081 	return (error);
1082 }
1083 
1084 /*
1085  * Check permissions, allocate an open file structure,
1086  * and call the device open routine if any.
1087  */
1088 int
1089 sys_open(struct lwp *l, void *v, register_t *retval)
1090 {
1091 	struct sys_open_args /* {
1092 		syscallarg(const char *) path;
1093 		syscallarg(int) flags;
1094 		syscallarg(int) mode;
1095 	} */ *uap = v;
1096 	struct proc *p = l->l_proc;
1097 	struct cwdinfo *cwdi = p->p_cwdi;
1098 	struct filedesc *fdp = p->p_fd;
1099 	struct file *fp;
1100 	struct vnode *vp;
1101 	int flags, cmode;
1102 	int type, indx, error;
1103 	struct flock lf;
1104 	struct nameidata nd;
1105 
1106 	flags = FFLAGS(SCARG(uap, flags));
1107 	if ((flags & (FREAD | FWRITE)) == 0)
1108 		return (EINVAL);
1109 	/* falloc() will use the file descriptor for us */
1110 	if ((error = falloc(p, &fp, &indx)) != 0)
1111 		return (error);
1112 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1113 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1114 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1115 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1116 		FILE_UNUSE(fp, p);
1117 		fdp->fd_ofiles[indx] = NULL;
1118 		ffree(fp);
1119 		if ((error == EDUPFD || error == EMOVEFD) &&
1120 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1121 		    (error =
1122 			dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) {
1123 			*retval = indx;
1124 			return (0);
1125 		}
1126 		if (error == ERESTART)
1127 			error = EINTR;
1128 		fdremove(fdp, indx);
1129 		return (error);
1130 	}
1131 	l->l_dupfd = 0;
1132 	vp = nd.ni_vp;
1133 	fp->f_flag = flags & FMASK;
1134 	fp->f_type = DTYPE_VNODE;
1135 	fp->f_ops = &vnops;
1136 	fp->f_data = vp;
1137 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1138 		lf.l_whence = SEEK_SET;
1139 		lf.l_start = 0;
1140 		lf.l_len = 0;
1141 		if (flags & O_EXLOCK)
1142 			lf.l_type = F_WRLCK;
1143 		else
1144 			lf.l_type = F_RDLCK;
1145 		type = F_FLOCK;
1146 		if ((flags & FNONBLOCK) == 0)
1147 			type |= F_WAIT;
1148 		VOP_UNLOCK(vp, 0);
1149 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1150 		if (error) {
1151 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1152 			FILE_UNUSE(fp, p);
1153 			ffree(fp);
1154 			fdremove(fdp, indx);
1155 			return (error);
1156 		}
1157 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1158 		fp->f_flag |= FHASLOCK;
1159 	}
1160 	VOP_UNLOCK(vp, 0);
1161 	*retval = indx;
1162 	FILE_SET_MATURE(fp);
1163 	FILE_UNUSE(fp, p);
1164 	return (0);
1165 }
1166 
1167 /*
1168  * Get file handle system call
1169  */
1170 int
1171 sys_getfh(struct lwp *l, void *v, register_t *retval)
1172 {
1173 	struct sys_getfh_args /* {
1174 		syscallarg(char *) fname;
1175 		syscallarg(fhandle_t *) fhp;
1176 	} */ *uap = v;
1177 	struct proc *p = l->l_proc;
1178 	struct vnode *vp;
1179 	fhandle_t fh;
1180 	int error;
1181 	struct nameidata nd;
1182 
1183 	/*
1184 	 * Must be super user
1185 	 */
1186 	error = suser(p->p_ucred, &p->p_acflag);
1187 	if (error)
1188 		return (error);
1189 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1190 	    SCARG(uap, fname), p);
1191 	error = namei(&nd);
1192 	if (error)
1193 		return (error);
1194 	vp = nd.ni_vp;
1195 	memset(&fh, 0, sizeof(fh));
1196 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1197 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1198 	vput(vp);
1199 	if (error)
1200 		return (error);
1201 	error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1202 	return (error);
1203 }
1204 
1205 /*
1206  * Open a file given a file handle.
1207  *
1208  * Check permissions, allocate an open file structure,
1209  * and call the device open routine if any.
1210  */
1211 int
1212 sys_fhopen(struct lwp *l, void *v, register_t *retval)
1213 {
1214 	struct sys_fhopen_args /* {
1215 		syscallarg(const fhandle_t *) fhp;
1216 		syscallarg(int) flags;
1217 	} */ *uap = v;
1218 	struct proc *p = l->l_proc;
1219 	struct filedesc *fdp = p->p_fd;
1220 	struct file *fp;
1221 	struct vnode *vp = NULL;
1222 	struct mount *mp;
1223 	struct ucred *cred = p->p_ucred;
1224 	int flags;
1225 	struct file *nfp;
1226 	int type, indx, error=0;
1227 	struct flock lf;
1228 	struct vattr va;
1229 	fhandle_t fh;
1230 
1231 	/*
1232 	 * Must be super user
1233 	 */
1234 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1235 		return (error);
1236 
1237 	flags = FFLAGS(SCARG(uap, flags));
1238 	if ((flags & (FREAD | FWRITE)) == 0)
1239 		return (EINVAL);
1240 	if ((flags & O_CREAT))
1241 		return (EINVAL);
1242 	/* falloc() will use the file descriptor for us */
1243 	if ((error = falloc(p, &nfp, &indx)) != 0)
1244 		return (error);
1245 	fp = nfp;
1246 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1247 		goto bad;
1248 
1249 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1250 		error = ESTALE;
1251 		goto bad;
1252 	}
1253 
1254 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1255 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1256 		goto bad;
1257 	}
1258 
1259 	/* Now do an effective vn_open */
1260 
1261 	if (vp->v_type == VSOCK) {
1262 		error = EOPNOTSUPP;
1263 		goto bad;
1264 	}
1265 	if (flags & FREAD) {
1266 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1267 			goto bad;
1268 	}
1269 	if (flags & (FWRITE | O_TRUNC)) {
1270 		if (vp->v_type == VDIR) {
1271 			error = EISDIR;
1272 			goto bad;
1273 		}
1274 		if ((error = vn_writechk(vp)) != 0 ||
1275 		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1276 			goto bad;
1277 	}
1278 	if (flags & O_TRUNC) {
1279 		if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
1280 			goto bad;
1281 		VOP_UNLOCK(vp, 0);			/* XXX */
1282 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1283 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1284 		VATTR_NULL(&va);
1285 		va.va_size = 0;
1286 		error = VOP_SETATTR(vp, &va, cred, p);
1287 		vn_finished_write(mp, 0);
1288 		if (error)
1289 			goto bad;
1290 	}
1291 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1292 		goto bad;
1293 	if (vp->v_type == VREG &&
1294 	    uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1295 		error = EIO;
1296 		goto bad;
1297 	}
1298 	if (flags & FWRITE)
1299 		vp->v_writecount++;
1300 
1301 	/* done with modified vn_open, now finish what sys_open does. */
1302 
1303 	fp->f_flag = flags & FMASK;
1304 	fp->f_type = DTYPE_VNODE;
1305 	fp->f_ops = &vnops;
1306 	fp->f_data = vp;
1307 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1308 		lf.l_whence = SEEK_SET;
1309 		lf.l_start = 0;
1310 		lf.l_len = 0;
1311 		if (flags & O_EXLOCK)
1312 			lf.l_type = F_WRLCK;
1313 		else
1314 			lf.l_type = F_RDLCK;
1315 		type = F_FLOCK;
1316 		if ((flags & FNONBLOCK) == 0)
1317 			type |= F_WAIT;
1318 		VOP_UNLOCK(vp, 0);
1319 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1320 		if (error) {
1321 			(void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1322 			FILE_UNUSE(fp, p);
1323 			ffree(fp);
1324 			fdremove(fdp, indx);
1325 			return (error);
1326 		}
1327 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1328 		fp->f_flag |= FHASLOCK;
1329 	}
1330 	VOP_UNLOCK(vp, 0);
1331 	*retval = indx;
1332 	FILE_SET_MATURE(fp);
1333 	FILE_UNUSE(fp, p);
1334 	return (0);
1335 
1336 bad:
1337 	FILE_UNUSE(fp, p);
1338 	ffree(fp);
1339 	fdremove(fdp, indx);
1340 	if (vp != NULL)
1341 		vput(vp);
1342 	return (error);
1343 }
1344 
1345 /* ARGSUSED */
1346 int
1347 sys_fhstat(struct lwp *l, void *v, register_t *retval)
1348 {
1349 	struct sys_fhstat_args /* {
1350 		syscallarg(const fhandle_t *) fhp;
1351 		syscallarg(struct stat *) sb;
1352 	} */ *uap = v;
1353 	struct proc *p = l->l_proc;
1354 	struct stat sb;
1355 	int error;
1356 	fhandle_t fh;
1357 	struct mount *mp;
1358 	struct vnode *vp;
1359 
1360 	/*
1361 	 * Must be super user
1362 	 */
1363 	if ((error = suser(p->p_ucred, &p->p_acflag)))
1364 		return (error);
1365 
1366 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1367 		return (error);
1368 
1369 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1370 		return (ESTALE);
1371 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1372 		return (error);
1373 	error = vn_stat(vp, &sb, p);
1374 	vput(vp);
1375 	if (error)
1376 		return (error);
1377 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1378 	return (error);
1379 }
1380 
1381 /* ARGSUSED */
1382 int
1383 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval)
1384 {
1385 	struct sys_fhstatvfs1_args /* {
1386 		syscallarg(const fhandle_t *) fhp;
1387 		syscallarg(struct statvfs *) buf;
1388 		syscallarg(int)	flags;
1389 	} */ *uap = v;
1390 	struct proc *p = l->l_proc;
1391 	struct statvfs sbuf;
1392 	fhandle_t fh;
1393 	struct mount *mp;
1394 	struct vnode *vp;
1395 	int error;
1396 
1397 	/*
1398 	 * Must be super user
1399 	 */
1400 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1401 		return error;
1402 
1403 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1404 		return error;
1405 
1406 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1407 		return ESTALE;
1408 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1409 		return error;
1410 
1411 	mp = vp->v_mount;
1412 	if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) {
1413 		vput(vp);
1414 		return error;
1415 	}
1416 	vput(vp);
1417 	return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
1418 }
1419 
1420 /*
1421  * Create a special file.
1422  */
1423 /* ARGSUSED */
1424 int
1425 sys_mknod(struct lwp *l, void *v, register_t *retval)
1426 {
1427 	struct sys_mknod_args /* {
1428 		syscallarg(const char *) path;
1429 		syscallarg(int) mode;
1430 		syscallarg(int) dev;
1431 	} */ *uap = v;
1432 	struct proc *p = l->l_proc;
1433 	struct vnode *vp;
1434 	struct mount *mp;
1435 	struct vattr vattr;
1436 	int error;
1437 	int whiteout = 0;
1438 	struct nameidata nd;
1439 
1440 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1441 		return (error);
1442 restart:
1443 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1444 	if ((error = namei(&nd)) != 0)
1445 		return (error);
1446 	vp = nd.ni_vp;
1447 	if (vp != NULL)
1448 		error = EEXIST;
1449 	else {
1450 		VATTR_NULL(&vattr);
1451 		vattr.va_mode =
1452 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1453 		vattr.va_rdev = SCARG(uap, dev);
1454 		whiteout = 0;
1455 
1456 		switch (SCARG(uap, mode) & S_IFMT) {
1457 		case S_IFMT:	/* used by badsect to flag bad sectors */
1458 			vattr.va_type = VBAD;
1459 			break;
1460 		case S_IFCHR:
1461 			vattr.va_type = VCHR;
1462 			break;
1463 		case S_IFBLK:
1464 			vattr.va_type = VBLK;
1465 			break;
1466 		case S_IFWHT:
1467 			whiteout = 1;
1468 			break;
1469 		default:
1470 			error = EINVAL;
1471 			break;
1472 		}
1473 	}
1474 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1475 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1476 		if (nd.ni_dvp == vp)
1477 			vrele(nd.ni_dvp);
1478 		else
1479 			vput(nd.ni_dvp);
1480 		if (vp)
1481 			vrele(vp);
1482 		if ((error = vn_start_write(NULL, &mp,
1483 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1484 			return (error);
1485 		goto restart;
1486 	}
1487 	if (!error) {
1488 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1489 		if (whiteout) {
1490 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1491 			if (error)
1492 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1493 			vput(nd.ni_dvp);
1494 		} else {
1495 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1496 						&nd.ni_cnd, &vattr);
1497 			if (error == 0)
1498 				vput(nd.ni_vp);
1499 		}
1500 	} else {
1501 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1502 		if (nd.ni_dvp == vp)
1503 			vrele(nd.ni_dvp);
1504 		else
1505 			vput(nd.ni_dvp);
1506 		if (vp)
1507 			vrele(vp);
1508 	}
1509 	vn_finished_write(mp, 0);
1510 	return (error);
1511 }
1512 
1513 /*
1514  * Create a named pipe.
1515  */
1516 /* ARGSUSED */
1517 int
1518 sys_mkfifo(struct lwp *l, void *v, register_t *retval)
1519 {
1520 	struct sys_mkfifo_args /* {
1521 		syscallarg(const char *) path;
1522 		syscallarg(int) mode;
1523 	} */ *uap = v;
1524 	struct proc *p = l->l_proc;
1525 	struct mount *mp;
1526 	struct vattr vattr;
1527 	int error;
1528 	struct nameidata nd;
1529 
1530 restart:
1531 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1532 	if ((error = namei(&nd)) != 0)
1533 		return (error);
1534 	if (nd.ni_vp != NULL) {
1535 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1536 		if (nd.ni_dvp == nd.ni_vp)
1537 			vrele(nd.ni_dvp);
1538 		else
1539 			vput(nd.ni_dvp);
1540 		vrele(nd.ni_vp);
1541 		return (EEXIST);
1542 	}
1543 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1544 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1545 		if (nd.ni_dvp == nd.ni_vp)
1546 			vrele(nd.ni_dvp);
1547 		else
1548 			vput(nd.ni_dvp);
1549 		if (nd.ni_vp)
1550 			vrele(nd.ni_vp);
1551 		if ((error = vn_start_write(NULL, &mp,
1552 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1553 			return (error);
1554 		goto restart;
1555 	}
1556 	VATTR_NULL(&vattr);
1557 	vattr.va_type = VFIFO;
1558 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1559 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1560 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1561 	if (error == 0)
1562 		vput(nd.ni_vp);
1563 	vn_finished_write(mp, 0);
1564 	return (error);
1565 }
1566 
1567 /*
1568  * Make a hard file link.
1569  */
1570 /* ARGSUSED */
1571 int
1572 sys_link(struct lwp *l, void *v, register_t *retval)
1573 {
1574 	struct sys_link_args /* {
1575 		syscallarg(const char *) path;
1576 		syscallarg(const char *) link;
1577 	} */ *uap = v;
1578 	struct proc *p = l->l_proc;
1579 	struct vnode *vp;
1580 	struct mount *mp;
1581 	struct nameidata nd;
1582 	int error;
1583 
1584 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1585 	if ((error = namei(&nd)) != 0)
1586 		return (error);
1587 	vp = nd.ni_vp;
1588 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
1589 		vrele(vp);
1590 		return (error);
1591 	}
1592 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1593 	if ((error = namei(&nd)) != 0)
1594 		goto out;
1595 	if (nd.ni_vp) {
1596 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1597 		if (nd.ni_dvp == nd.ni_vp)
1598 			vrele(nd.ni_dvp);
1599 		else
1600 			vput(nd.ni_dvp);
1601 		vrele(nd.ni_vp);
1602 		error = EEXIST;
1603 		goto out;
1604 	}
1605 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1606 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1607 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1608 out:
1609 	vrele(vp);
1610 	vn_finished_write(mp, 0);
1611 	return (error);
1612 }
1613 
1614 /*
1615  * Make a symbolic link.
1616  */
1617 /* ARGSUSED */
1618 int
1619 sys_symlink(struct lwp *l, void *v, register_t *retval)
1620 {
1621 	struct sys_symlink_args /* {
1622 		syscallarg(const char *) path;
1623 		syscallarg(const char *) link;
1624 	} */ *uap = v;
1625 	struct proc *p = l->l_proc;
1626 	struct mount *mp;
1627 	struct vattr vattr;
1628 	char *path;
1629 	int error;
1630 	struct nameidata nd;
1631 
1632 	path = PNBUF_GET();
1633 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1634 	if (error)
1635 		goto out;
1636 restart:
1637 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1638 	if ((error = namei(&nd)) != 0)
1639 		goto out;
1640 	if (nd.ni_vp) {
1641 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1642 		if (nd.ni_dvp == nd.ni_vp)
1643 			vrele(nd.ni_dvp);
1644 		else
1645 			vput(nd.ni_dvp);
1646 		vrele(nd.ni_vp);
1647 		error = EEXIST;
1648 		goto out;
1649 	}
1650 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1651 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1652 		if (nd.ni_dvp == nd.ni_vp)
1653 			vrele(nd.ni_dvp);
1654 		else
1655 			vput(nd.ni_dvp);
1656 		if ((error = vn_start_write(NULL, &mp,
1657 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1658 			return (error);
1659 		goto restart;
1660 	}
1661 	VATTR_NULL(&vattr);
1662 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1663 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1664 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1665 	if (error == 0)
1666 		vput(nd.ni_vp);
1667 	vn_finished_write(mp, 0);
1668 out:
1669 	PNBUF_PUT(path);
1670 	return (error);
1671 }
1672 
1673 /*
1674  * Delete a whiteout from the filesystem.
1675  */
1676 /* ARGSUSED */
1677 int
1678 sys_undelete(struct lwp *l, void *v, register_t *retval)
1679 {
1680 	struct sys_undelete_args /* {
1681 		syscallarg(const char *) path;
1682 	} */ *uap = v;
1683 	struct proc *p = l->l_proc;
1684 	int error;
1685 	struct mount *mp;
1686 	struct nameidata nd;
1687 
1688 restart:
1689 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1690 	    SCARG(uap, path), p);
1691 	error = namei(&nd);
1692 	if (error)
1693 		return (error);
1694 
1695 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1696 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1697 		if (nd.ni_dvp == nd.ni_vp)
1698 			vrele(nd.ni_dvp);
1699 		else
1700 			vput(nd.ni_dvp);
1701 		if (nd.ni_vp)
1702 			vrele(nd.ni_vp);
1703 		return (EEXIST);
1704 	}
1705 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1706 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1707 		if (nd.ni_dvp == nd.ni_vp)
1708 			vrele(nd.ni_dvp);
1709 		else
1710 			vput(nd.ni_dvp);
1711 		if ((error = vn_start_write(NULL, &mp,
1712 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1713 			return (error);
1714 		goto restart;
1715 	}
1716 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1717 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1718 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1719 	vput(nd.ni_dvp);
1720 	vn_finished_write(mp, 0);
1721 	return (error);
1722 }
1723 
1724 /*
1725  * Delete a name from the filesystem.
1726  */
1727 /* ARGSUSED */
1728 int
1729 sys_unlink(struct lwp *l, void *v, register_t *retval)
1730 {
1731 	struct sys_unlink_args /* {
1732 		syscallarg(const char *) path;
1733 	} */ *uap = v;
1734 	struct proc *p = l->l_proc;
1735 	struct mount *mp;
1736 	struct vnode *vp;
1737 	int error;
1738 	struct nameidata nd;
1739 
1740 restart:
1741 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1742 	    SCARG(uap, path), p);
1743 	if ((error = namei(&nd)) != 0)
1744 		return (error);
1745 	vp = nd.ni_vp;
1746 
1747 	/*
1748 	 * The root of a mounted filesystem cannot be deleted.
1749 	 */
1750 	if (vp->v_flag & VROOT) {
1751 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1752 		if (nd.ni_dvp == vp)
1753 			vrele(nd.ni_dvp);
1754 		else
1755 			vput(nd.ni_dvp);
1756 		vput(vp);
1757 		error = EBUSY;
1758 		goto out;
1759 	}
1760 
1761 #ifdef VERIFIED_EXEC
1762 	/* Handle remove requests for veriexec entries. */
1763 	if ((error = veriexec_removechk(p, vp, nd.ni_dirp)) != 0) {
1764 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1765 		if (nd.ni_dvp == vp)
1766 			vrele(nd.ni_dvp);
1767 		else
1768 			vput(nd.ni_dvp);
1769 		vput(vp);
1770 		goto out;
1771 	}
1772 #endif
1773 
1774 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1775 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1776 		if (nd.ni_dvp == vp)
1777 			vrele(nd.ni_dvp);
1778 		else
1779 			vput(nd.ni_dvp);
1780 		vput(vp);
1781 		if ((error = vn_start_write(NULL, &mp,
1782 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1783 			return (error);
1784 		goto restart;
1785 	}
1786 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1787 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1788 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1789 	vn_finished_write(mp, 0);
1790 out:
1791 	return (error);
1792 }
1793 
1794 /*
1795  * Reposition read/write file offset.
1796  */
1797 int
1798 sys_lseek(struct lwp *l, void *v, register_t *retval)
1799 {
1800 	struct sys_lseek_args /* {
1801 		syscallarg(int) fd;
1802 		syscallarg(int) pad;
1803 		syscallarg(off_t) offset;
1804 		syscallarg(int) whence;
1805 	} */ *uap = v;
1806 	struct proc *p = l->l_proc;
1807 	struct ucred *cred = p->p_ucred;
1808 	struct filedesc *fdp = p->p_fd;
1809 	struct file *fp;
1810 	struct vnode *vp;
1811 	struct vattr vattr;
1812 	off_t newoff;
1813 	int error;
1814 
1815 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1816 		return (EBADF);
1817 
1818 	FILE_USE(fp);
1819 
1820 	vp = (struct vnode *)fp->f_data;
1821 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1822 		error = ESPIPE;
1823 		goto out;
1824 	}
1825 
1826 	switch (SCARG(uap, whence)) {
1827 	case SEEK_CUR:
1828 		newoff = fp->f_offset + SCARG(uap, offset);
1829 		break;
1830 	case SEEK_END:
1831 		error = VOP_GETATTR(vp, &vattr, cred, p);
1832 		if (error)
1833 			goto out;
1834 		newoff = SCARG(uap, offset) + vattr.va_size;
1835 		break;
1836 	case SEEK_SET:
1837 		newoff = SCARG(uap, offset);
1838 		break;
1839 	default:
1840 		error = EINVAL;
1841 		goto out;
1842 	}
1843 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1844 		goto out;
1845 
1846 	*(off_t *)retval = fp->f_offset = newoff;
1847  out:
1848 	FILE_UNUSE(fp, p);
1849 	return (error);
1850 }
1851 
1852 /*
1853  * Positional read system call.
1854  */
1855 int
1856 sys_pread(struct lwp *l, void *v, register_t *retval)
1857 {
1858 	struct sys_pread_args /* {
1859 		syscallarg(int) fd;
1860 		syscallarg(void *) buf;
1861 		syscallarg(size_t) nbyte;
1862 		syscallarg(off_t) offset;
1863 	} */ *uap = v;
1864 	struct proc *p = l->l_proc;
1865 	struct filedesc *fdp = p->p_fd;
1866 	struct file *fp;
1867 	struct vnode *vp;
1868 	off_t offset;
1869 	int error, fd = SCARG(uap, fd);
1870 
1871 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1872 		return (EBADF);
1873 
1874 	if ((fp->f_flag & FREAD) == 0) {
1875 		simple_unlock(&fp->f_slock);
1876 		return (EBADF);
1877 	}
1878 
1879 	FILE_USE(fp);
1880 
1881 	vp = (struct vnode *)fp->f_data;
1882 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1883 		error = ESPIPE;
1884 		goto out;
1885 	}
1886 
1887 	offset = SCARG(uap, offset);
1888 
1889 	/*
1890 	 * XXX This works because no file systems actually
1891 	 * XXX take any action on the seek operation.
1892 	 */
1893 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1894 		goto out;
1895 
1896 	/* dofileread() will unuse the descriptor for us */
1897 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1898 	    &offset, 0, retval));
1899 
1900  out:
1901 	FILE_UNUSE(fp, p);
1902 	return (error);
1903 }
1904 
1905 /*
1906  * Positional scatter read system call.
1907  */
1908 int
1909 sys_preadv(struct lwp *l, void *v, register_t *retval)
1910 {
1911 	struct sys_preadv_args /* {
1912 		syscallarg(int) fd;
1913 		syscallarg(const struct iovec *) iovp;
1914 		syscallarg(int) iovcnt;
1915 		syscallarg(off_t) offset;
1916 	} */ *uap = v;
1917 	struct proc *p = l->l_proc;
1918 	struct filedesc *fdp = p->p_fd;
1919 	struct file *fp;
1920 	struct vnode *vp;
1921 	off_t offset;
1922 	int error, fd = SCARG(uap, fd);
1923 
1924 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1925 		return (EBADF);
1926 
1927 	if ((fp->f_flag & FREAD) == 0) {
1928 		simple_unlock(&fp->f_slock);
1929 		return (EBADF);
1930 	}
1931 
1932 	FILE_USE(fp);
1933 
1934 	vp = (struct vnode *)fp->f_data;
1935 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1936 		error = ESPIPE;
1937 		goto out;
1938 	}
1939 
1940 	offset = SCARG(uap, offset);
1941 
1942 	/*
1943 	 * XXX This works because no file systems actually
1944 	 * XXX take any action on the seek operation.
1945 	 */
1946 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1947 		goto out;
1948 
1949 	/* dofilereadv() will unuse the descriptor for us */
1950 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1951 	    &offset, 0, retval));
1952 
1953  out:
1954 	FILE_UNUSE(fp, p);
1955 	return (error);
1956 }
1957 
1958 /*
1959  * Positional write system call.
1960  */
1961 int
1962 sys_pwrite(struct lwp *l, void *v, register_t *retval)
1963 {
1964 	struct sys_pwrite_args /* {
1965 		syscallarg(int) fd;
1966 		syscallarg(const void *) buf;
1967 		syscallarg(size_t) nbyte;
1968 		syscallarg(off_t) offset;
1969 	} */ *uap = v;
1970 	struct proc *p = l->l_proc;
1971 	struct filedesc *fdp = p->p_fd;
1972 	struct file *fp;
1973 	struct vnode *vp;
1974 	off_t offset;
1975 	int error, fd = SCARG(uap, fd);
1976 
1977 	if ((fp = fd_getfile(fdp, fd)) == NULL)
1978 		return (EBADF);
1979 
1980 	if ((fp->f_flag & FWRITE) == 0) {
1981 		simple_unlock(&fp->f_slock);
1982 		return (EBADF);
1983 	}
1984 
1985 	FILE_USE(fp);
1986 
1987 	vp = (struct vnode *)fp->f_data;
1988 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1989 		error = ESPIPE;
1990 		goto out;
1991 	}
1992 
1993 	offset = SCARG(uap, offset);
1994 
1995 	/*
1996 	 * XXX This works because no file systems actually
1997 	 * XXX take any action on the seek operation.
1998 	 */
1999 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2000 		goto out;
2001 
2002 	/* dofilewrite() will unuse the descriptor for us */
2003 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2004 	    &offset, 0, retval));
2005 
2006  out:
2007 	FILE_UNUSE(fp, p);
2008 	return (error);
2009 }
2010 
2011 /*
2012  * Positional gather write system call.
2013  */
2014 int
2015 sys_pwritev(struct lwp *l, void *v, register_t *retval)
2016 {
2017 	struct sys_pwritev_args /* {
2018 		syscallarg(int) fd;
2019 		syscallarg(const struct iovec *) iovp;
2020 		syscallarg(int) iovcnt;
2021 		syscallarg(off_t) offset;
2022 	} */ *uap = v;
2023 	struct proc *p = l->l_proc;
2024 	struct filedesc *fdp = p->p_fd;
2025 	struct file *fp;
2026 	struct vnode *vp;
2027 	off_t offset;
2028 	int error, fd = SCARG(uap, fd);
2029 
2030 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2031 		return (EBADF);
2032 
2033 	if ((fp->f_flag & FWRITE) == 0) {
2034 		simple_unlock(&fp->f_slock);
2035 		return (EBADF);
2036 	}
2037 
2038 	FILE_USE(fp);
2039 
2040 	vp = (struct vnode *)fp->f_data;
2041 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2042 		error = ESPIPE;
2043 		goto out;
2044 	}
2045 
2046 	offset = SCARG(uap, offset);
2047 
2048 	/*
2049 	 * XXX This works because no file systems actually
2050 	 * XXX take any action on the seek operation.
2051 	 */
2052 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2053 		goto out;
2054 
2055 	/* dofilewritev() will unuse the descriptor for us */
2056 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2057 	    &offset, 0, retval));
2058 
2059  out:
2060 	FILE_UNUSE(fp, p);
2061 	return (error);
2062 }
2063 
2064 /*
2065  * Check access permissions.
2066  */
2067 int
2068 sys_access(struct lwp *l, void *v, register_t *retval)
2069 {
2070 	struct sys_access_args /* {
2071 		syscallarg(const char *) path;
2072 		syscallarg(int) flags;
2073 	} */ *uap = v;
2074 	struct proc *p = l->l_proc;
2075 	struct ucred *cred;
2076 	struct vnode *vp;
2077 	int error, flags;
2078 	struct nameidata nd;
2079 
2080 	cred = crdup(p->p_ucred);
2081 	cred->cr_uid = p->p_cred->p_ruid;
2082 	cred->cr_gid = p->p_cred->p_rgid;
2083 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2084 	    SCARG(uap, path), p);
2085 	/* Override default credentials */
2086 	nd.ni_cnd.cn_cred = cred;
2087 	if ((error = namei(&nd)) != 0)
2088 		goto out;
2089 	vp = nd.ni_vp;
2090 
2091 	/* Flags == 0 means only check for existence. */
2092 	if (SCARG(uap, flags)) {
2093 		flags = 0;
2094 		if (SCARG(uap, flags) & R_OK)
2095 			flags |= VREAD;
2096 		if (SCARG(uap, flags) & W_OK)
2097 			flags |= VWRITE;
2098 		if (SCARG(uap, flags) & X_OK)
2099 			flags |= VEXEC;
2100 
2101 		error = VOP_ACCESS(vp, flags, cred, p);
2102 		if (!error && (flags & VWRITE))
2103 			error = vn_writechk(vp);
2104 	}
2105 	vput(vp);
2106 out:
2107 	crfree(cred);
2108 	return (error);
2109 }
2110 
2111 /*
2112  * Get file status; this version follows links.
2113  */
2114 /* ARGSUSED */
2115 int
2116 sys___stat13(struct lwp *l, void *v, register_t *retval)
2117 {
2118 	struct sys___stat13_args /* {
2119 		syscallarg(const char *) path;
2120 		syscallarg(struct stat *) ub;
2121 	} */ *uap = v;
2122 	struct proc *p = l->l_proc;
2123 	struct stat sb;
2124 	int error;
2125 	struct nameidata nd;
2126 
2127 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2128 	    SCARG(uap, path), p);
2129 	if ((error = namei(&nd)) != 0)
2130 		return (error);
2131 	error = vn_stat(nd.ni_vp, &sb, p);
2132 	vput(nd.ni_vp);
2133 	if (error)
2134 		return (error);
2135 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2136 	return (error);
2137 }
2138 
2139 /*
2140  * Get file status; this version does not follow links.
2141  */
2142 /* ARGSUSED */
2143 int
2144 sys___lstat13(struct lwp *l, void *v, register_t *retval)
2145 {
2146 	struct sys___lstat13_args /* {
2147 		syscallarg(const char *) path;
2148 		syscallarg(struct stat *) ub;
2149 	} */ *uap = v;
2150 	struct proc *p = l->l_proc;
2151 	struct stat sb;
2152 	int error;
2153 	struct nameidata nd;
2154 
2155 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2156 	    SCARG(uap, path), p);
2157 	if ((error = namei(&nd)) != 0)
2158 		return (error);
2159 	error = vn_stat(nd.ni_vp, &sb, p);
2160 	vput(nd.ni_vp);
2161 	if (error)
2162 		return (error);
2163 	error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2164 	return (error);
2165 }
2166 
2167 /*
2168  * Get configurable pathname variables.
2169  */
2170 /* ARGSUSED */
2171 int
2172 sys_pathconf(struct lwp *l, void *v, register_t *retval)
2173 {
2174 	struct sys_pathconf_args /* {
2175 		syscallarg(const char *) path;
2176 		syscallarg(int) name;
2177 	} */ *uap = v;
2178 	struct proc *p = l->l_proc;
2179 	int error;
2180 	struct nameidata nd;
2181 
2182 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2183 	    SCARG(uap, path), p);
2184 	if ((error = namei(&nd)) != 0)
2185 		return (error);
2186 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2187 	vput(nd.ni_vp);
2188 	return (error);
2189 }
2190 
2191 /*
2192  * Return target name of a symbolic link.
2193  */
2194 /* ARGSUSED */
2195 int
2196 sys_readlink(struct lwp *l, void *v, register_t *retval)
2197 {
2198 	struct sys_readlink_args /* {
2199 		syscallarg(const char *) path;
2200 		syscallarg(char *) buf;
2201 		syscallarg(size_t) count;
2202 	} */ *uap = v;
2203 	struct proc *p = l->l_proc;
2204 	struct vnode *vp;
2205 	struct iovec aiov;
2206 	struct uio auio;
2207 	int error;
2208 	struct nameidata nd;
2209 
2210 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2211 	    SCARG(uap, path), p);
2212 	if ((error = namei(&nd)) != 0)
2213 		return (error);
2214 	vp = nd.ni_vp;
2215 	if (vp->v_type != VLNK)
2216 		error = EINVAL;
2217 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2218 	    (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2219 		aiov.iov_base = SCARG(uap, buf);
2220 		aiov.iov_len = SCARG(uap, count);
2221 		auio.uio_iov = &aiov;
2222 		auio.uio_iovcnt = 1;
2223 		auio.uio_offset = 0;
2224 		auio.uio_rw = UIO_READ;
2225 		auio.uio_segflg = UIO_USERSPACE;
2226 		auio.uio_procp = p;
2227 		auio.uio_resid = SCARG(uap, count);
2228 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2229 	}
2230 	vput(vp);
2231 	*retval = SCARG(uap, count) - auio.uio_resid;
2232 	return (error);
2233 }
2234 
2235 /*
2236  * Change flags of a file given a path name.
2237  */
2238 /* ARGSUSED */
2239 int
2240 sys_chflags(struct lwp *l, void *v, register_t *retval)
2241 {
2242 	struct sys_chflags_args /* {
2243 		syscallarg(const char *) path;
2244 		syscallarg(u_long) flags;
2245 	} */ *uap = v;
2246 	struct proc *p = l->l_proc;
2247 	struct vnode *vp;
2248 	int error;
2249 	struct nameidata nd;
2250 
2251 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2252 	if ((error = namei(&nd)) != 0)
2253 		return (error);
2254 	vp = nd.ni_vp;
2255 	error = change_flags(vp, SCARG(uap, flags), p);
2256 	vput(vp);
2257 	return (error);
2258 }
2259 
2260 /*
2261  * Change flags of a file given a file descriptor.
2262  */
2263 /* ARGSUSED */
2264 int
2265 sys_fchflags(struct lwp *l, void *v, register_t *retval)
2266 {
2267 	struct sys_fchflags_args /* {
2268 		syscallarg(int) fd;
2269 		syscallarg(u_long) flags;
2270 	} */ *uap = v;
2271 	struct proc *p = l->l_proc;
2272 	struct vnode *vp;
2273 	struct file *fp;
2274 	int error;
2275 
2276 	/* getvnode() will use the descriptor for us */
2277 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2278 		return (error);
2279 	vp = (struct vnode *)fp->f_data;
2280 	error = change_flags(vp, SCARG(uap, flags), p);
2281 	VOP_UNLOCK(vp, 0);
2282 	FILE_UNUSE(fp, p);
2283 	return (error);
2284 }
2285 
2286 /*
2287  * Change flags of a file given a path name; this version does
2288  * not follow links.
2289  */
2290 int
2291 sys_lchflags(struct lwp *l, void *v, register_t *retval)
2292 {
2293 	struct sys_lchflags_args /* {
2294 		syscallarg(const char *) path;
2295 		syscallarg(u_long) flags;
2296 	} */ *uap = v;
2297 	struct proc *p = l->l_proc;
2298 	struct vnode *vp;
2299 	int error;
2300 	struct nameidata nd;
2301 
2302 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2303 	if ((error = namei(&nd)) != 0)
2304 		return (error);
2305 	vp = nd.ni_vp;
2306 	error = change_flags(vp, SCARG(uap, flags), p);
2307 	vput(vp);
2308 	return (error);
2309 }
2310 
2311 /*
2312  * Common routine to change flags of a file.
2313  */
2314 int
2315 change_flags(struct vnode *vp, u_long flags, struct proc *p)
2316 {
2317 	struct mount *mp;
2318 	struct vattr vattr;
2319 	int error;
2320 
2321 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2322 		return (error);
2323 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2324 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2325 	/*
2326 	 * Non-superusers cannot change the flags on devices, even if they
2327 	 * own them.
2328 	 */
2329 	if (suser(p->p_ucred, &p->p_acflag) != 0) {
2330 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2331 			goto out;
2332 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2333 			error = EINVAL;
2334 			goto out;
2335 		}
2336 	}
2337 	VATTR_NULL(&vattr);
2338 	vattr.va_flags = flags;
2339 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2340 out:
2341 	vn_finished_write(mp, 0);
2342 	return (error);
2343 }
2344 
2345 /*
2346  * Change mode of a file given path name; this version follows links.
2347  */
2348 /* ARGSUSED */
2349 int
2350 sys_chmod(struct lwp *l, void *v, register_t *retval)
2351 {
2352 	struct sys_chmod_args /* {
2353 		syscallarg(const char *) path;
2354 		syscallarg(int) mode;
2355 	} */ *uap = v;
2356 	struct proc *p = l->l_proc;
2357 	int error;
2358 	struct nameidata nd;
2359 
2360 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2361 	if ((error = namei(&nd)) != 0)
2362 		return (error);
2363 
2364 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2365 
2366 	vrele(nd.ni_vp);
2367 	return (error);
2368 }
2369 
2370 /*
2371  * Change mode of a file given a file descriptor.
2372  */
2373 /* ARGSUSED */
2374 int
2375 sys_fchmod(struct lwp *l, void *v, register_t *retval)
2376 {
2377 	struct sys_fchmod_args /* {
2378 		syscallarg(int) fd;
2379 		syscallarg(int) mode;
2380 	} */ *uap = v;
2381 	struct proc *p = l->l_proc;
2382 	struct file *fp;
2383 	int error;
2384 
2385 	/* getvnode() will use the descriptor for us */
2386 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2387 		return (error);
2388 
2389 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2390 	FILE_UNUSE(fp, p);
2391 	return (error);
2392 }
2393 
2394 /*
2395  * Change mode of a file given path name; this version does not follow links.
2396  */
2397 /* ARGSUSED */
2398 int
2399 sys_lchmod(struct lwp *l, void *v, register_t *retval)
2400 {
2401 	struct sys_lchmod_args /* {
2402 		syscallarg(const char *) path;
2403 		syscallarg(int) mode;
2404 	} */ *uap = v;
2405 	struct proc *p = l->l_proc;
2406 	int error;
2407 	struct nameidata nd;
2408 
2409 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2410 	if ((error = namei(&nd)) != 0)
2411 		return (error);
2412 
2413 	error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2414 
2415 	vrele(nd.ni_vp);
2416 	return (error);
2417 }
2418 
2419 /*
2420  * Common routine to set mode given a vnode.
2421  */
2422 static int
2423 change_mode(struct vnode *vp, int mode, struct proc *p)
2424 {
2425 	struct mount *mp;
2426 	struct vattr vattr;
2427 	int error;
2428 
2429 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2430 		return (error);
2431 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2432 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2433 	VATTR_NULL(&vattr);
2434 	vattr.va_mode = mode & ALLPERMS;
2435 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2436 	VOP_UNLOCK(vp, 0);
2437 	vn_finished_write(mp, 0);
2438 	return (error);
2439 }
2440 
2441 /*
2442  * Set ownership given a path name; this version follows links.
2443  */
2444 /* ARGSUSED */
2445 int
2446 sys_chown(struct lwp *l, void *v, register_t *retval)
2447 {
2448 	struct sys_chown_args /* {
2449 		syscallarg(const char *) path;
2450 		syscallarg(uid_t) uid;
2451 		syscallarg(gid_t) gid;
2452 	} */ *uap = v;
2453 	struct proc *p = l->l_proc;
2454 	int error;
2455 	struct nameidata nd;
2456 
2457 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2458 	if ((error = namei(&nd)) != 0)
2459 		return (error);
2460 
2461 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2462 
2463 	vrele(nd.ni_vp);
2464 	return (error);
2465 }
2466 
2467 /*
2468  * Set ownership given a path name; this version follows links.
2469  * Provides POSIX semantics.
2470  */
2471 /* ARGSUSED */
2472 int
2473 sys___posix_chown(struct lwp *l, void *v, register_t *retval)
2474 {
2475 	struct sys_chown_args /* {
2476 		syscallarg(const char *) path;
2477 		syscallarg(uid_t) uid;
2478 		syscallarg(gid_t) gid;
2479 	} */ *uap = v;
2480 	struct proc *p = l->l_proc;
2481 	int error;
2482 	struct nameidata nd;
2483 
2484 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2485 	if ((error = namei(&nd)) != 0)
2486 		return (error);
2487 
2488 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2489 
2490 	vrele(nd.ni_vp);
2491 	return (error);
2492 }
2493 
2494 /*
2495  * Set ownership given a file descriptor.
2496  */
2497 /* ARGSUSED */
2498 int
2499 sys_fchown(struct lwp *l, void *v, register_t *retval)
2500 {
2501 	struct sys_fchown_args /* {
2502 		syscallarg(int) fd;
2503 		syscallarg(uid_t) uid;
2504 		syscallarg(gid_t) gid;
2505 	} */ *uap = v;
2506 	struct proc *p = l->l_proc;
2507 	int error;
2508 	struct file *fp;
2509 
2510 	/* getvnode() will use the descriptor for us */
2511 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2512 		return (error);
2513 
2514 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2515 	    SCARG(uap, gid), p, 0);
2516 	FILE_UNUSE(fp, p);
2517 	return (error);
2518 }
2519 
2520 /*
2521  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2522  */
2523 /* ARGSUSED */
2524 int
2525 sys___posix_fchown(struct lwp *l, void *v, register_t *retval)
2526 {
2527 	struct sys_fchown_args /* {
2528 		syscallarg(int) fd;
2529 		syscallarg(uid_t) uid;
2530 		syscallarg(gid_t) gid;
2531 	} */ *uap = v;
2532 	struct proc *p = l->l_proc;
2533 	int error;
2534 	struct file *fp;
2535 
2536 	/* getvnode() will use the descriptor for us */
2537 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2538 		return (error);
2539 
2540 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2541 	    SCARG(uap, gid), p, 1);
2542 	FILE_UNUSE(fp, p);
2543 	return (error);
2544 }
2545 
2546 /*
2547  * Set ownership given a path name; this version does not follow links.
2548  */
2549 /* ARGSUSED */
2550 int
2551 sys_lchown(struct lwp *l, void *v, register_t *retval)
2552 {
2553 	struct sys_lchown_args /* {
2554 		syscallarg(const char *) path;
2555 		syscallarg(uid_t) uid;
2556 		syscallarg(gid_t) gid;
2557 	} */ *uap = v;
2558 	struct proc *p = l->l_proc;
2559 	int error;
2560 	struct nameidata nd;
2561 
2562 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2563 	if ((error = namei(&nd)) != 0)
2564 		return (error);
2565 
2566 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2567 
2568 	vrele(nd.ni_vp);
2569 	return (error);
2570 }
2571 
2572 /*
2573  * Set ownership given a path name; this version does not follow links.
2574  * Provides POSIX/XPG semantics.
2575  */
2576 /* ARGSUSED */
2577 int
2578 sys___posix_lchown(struct lwp *l, void *v, register_t *retval)
2579 {
2580 	struct sys_lchown_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, NOFOLLOW, 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  * Common routine to set ownership given a vnode.
2601  */
2602 static int
2603 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct proc *p,
2604     int posix_semantics)
2605 {
2606 	struct mount *mp;
2607 	struct vattr vattr;
2608 	mode_t newmode;
2609 	int error;
2610 
2611 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2612 		return (error);
2613 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2614 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2615 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2616 		goto out;
2617 
2618 #define CHANGED(x) ((int)(x) != -1)
2619 	newmode = vattr.va_mode;
2620 	if (posix_semantics) {
2621 		/*
2622 		 * POSIX/XPG semantics: if the caller is not the super-user,
2623 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2624 		 * the XPG consider the behaviour for calls by the super-user
2625 		 * implementation-defined; we leave the set-user-id and set-
2626 		 * group-id settings intact in that case.
2627 		 */
2628 		if (suser(p->p_ucred, NULL) != 0)
2629 			newmode &= ~(S_ISUID | S_ISGID);
2630 	} else {
2631 		/*
2632 		 * NetBSD semantics: when changing owner and/or group,
2633 		 * clear the respective bit(s).
2634 		 */
2635 		if (CHANGED(uid))
2636 			newmode &= ~S_ISUID;
2637 		if (CHANGED(gid))
2638 			newmode &= ~S_ISGID;
2639 	}
2640 	/* Update va_mode iff altered. */
2641 	if (vattr.va_mode == newmode)
2642 		newmode = VNOVAL;
2643 
2644 	VATTR_NULL(&vattr);
2645 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2646 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2647 	vattr.va_mode = newmode;
2648 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2649 #undef CHANGED
2650 
2651 out:
2652 	VOP_UNLOCK(vp, 0);
2653 	vn_finished_write(mp, 0);
2654 	return (error);
2655 }
2656 
2657 /*
2658  * Set the access and modification times given a path name; this
2659  * version follows links.
2660  */
2661 /* ARGSUSED */
2662 int
2663 sys_utimes(struct lwp *l, void *v, register_t *retval)
2664 {
2665 	struct sys_utimes_args /* {
2666 		syscallarg(const char *) path;
2667 		syscallarg(const struct timeval *) tptr;
2668 	} */ *uap = v;
2669 	struct proc *p = l->l_proc;
2670 	int error;
2671 	struct nameidata nd;
2672 
2673 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2674 	if ((error = namei(&nd)) != 0)
2675 		return (error);
2676 
2677 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2678 
2679 	vrele(nd.ni_vp);
2680 	return (error);
2681 }
2682 
2683 /*
2684  * Set the access and modification times given a file descriptor.
2685  */
2686 /* ARGSUSED */
2687 int
2688 sys_futimes(struct lwp *l, void *v, register_t *retval)
2689 {
2690 	struct sys_futimes_args /* {
2691 		syscallarg(int) fd;
2692 		syscallarg(const struct timeval *) tptr;
2693 	} */ *uap = v;
2694 	struct proc *p = l->l_proc;
2695 	int error;
2696 	struct file *fp;
2697 
2698 	/* getvnode() will use the descriptor for us */
2699 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2700 		return (error);
2701 
2702 	error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2703 	FILE_UNUSE(fp, p);
2704 	return (error);
2705 }
2706 
2707 /*
2708  * Set the access and modification times given a path name; this
2709  * version does not follow links.
2710  */
2711 /* ARGSUSED */
2712 int
2713 sys_lutimes(struct lwp *l, void *v, register_t *retval)
2714 {
2715 	struct sys_lutimes_args /* {
2716 		syscallarg(const char *) path;
2717 		syscallarg(const struct timeval *) tptr;
2718 	} */ *uap = v;
2719 	struct proc *p = l->l_proc;
2720 	int error;
2721 	struct nameidata nd;
2722 
2723 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2724 	if ((error = namei(&nd)) != 0)
2725 		return (error);
2726 
2727 	error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2728 
2729 	vrele(nd.ni_vp);
2730 	return (error);
2731 }
2732 
2733 /*
2734  * Common routine to set access and modification times given a vnode.
2735  */
2736 static int
2737 change_utimes(struct vnode *vp, const struct timeval *tptr, struct proc *p)
2738 {
2739 	struct timeval tv[2];
2740 	struct mount *mp;
2741 	struct vattr vattr;
2742 	int error;
2743 
2744 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2745 		return (error);
2746 	VATTR_NULL(&vattr);
2747 	if (tptr == NULL) {
2748 		microtime(&tv[0]);
2749 		tv[1] = tv[0];
2750 		vattr.va_vaflags |= VA_UTIMES_NULL;
2751 	} else {
2752 		error = copyin(tptr, tv, sizeof(tv));
2753 		if (error)
2754 			goto out;
2755 	}
2756 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2757 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2758 	vattr.va_atime.tv_sec = tv[0].tv_sec;
2759 	vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2760 	vattr.va_mtime.tv_sec = tv[1].tv_sec;
2761 	vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2762 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2763 	VOP_UNLOCK(vp, 0);
2764 out:
2765 	vn_finished_write(mp, 0);
2766 	return (error);
2767 }
2768 
2769 /*
2770  * Truncate a file given its path name.
2771  */
2772 /* ARGSUSED */
2773 int
2774 sys_truncate(struct lwp *l, void *v, register_t *retval)
2775 {
2776 	struct sys_truncate_args /* {
2777 		syscallarg(const char *) path;
2778 		syscallarg(int) pad;
2779 		syscallarg(off_t) length;
2780 	} */ *uap = v;
2781 	struct proc *p = l->l_proc;
2782 	struct vnode *vp;
2783 	struct mount *mp;
2784 	struct vattr vattr;
2785 	int error;
2786 	struct nameidata nd;
2787 
2788 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2789 	if ((error = namei(&nd)) != 0)
2790 		return (error);
2791 	vp = nd.ni_vp;
2792 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2793 		vrele(vp);
2794 		return (error);
2795 	}
2796 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2797 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2798 	if (vp->v_type == VDIR)
2799 		error = EISDIR;
2800 	else if ((error = vn_writechk(vp)) == 0 &&
2801 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2802 		VATTR_NULL(&vattr);
2803 		vattr.va_size = SCARG(uap, length);
2804 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2805 	}
2806 	vput(vp);
2807 	vn_finished_write(mp, 0);
2808 	return (error);
2809 }
2810 
2811 /*
2812  * Truncate a file given a file descriptor.
2813  */
2814 /* ARGSUSED */
2815 int
2816 sys_ftruncate(struct lwp *l, void *v, register_t *retval)
2817 {
2818 	struct sys_ftruncate_args /* {
2819 		syscallarg(int) fd;
2820 		syscallarg(int) pad;
2821 		syscallarg(off_t) length;
2822 	} */ *uap = v;
2823 	struct proc *p = l->l_proc;
2824 	struct mount *mp;
2825 	struct vattr vattr;
2826 	struct vnode *vp;
2827 	struct file *fp;
2828 	int error;
2829 
2830 	/* getvnode() will use the descriptor for us */
2831 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2832 		return (error);
2833 	if ((fp->f_flag & FWRITE) == 0) {
2834 		error = EINVAL;
2835 		goto out;
2836 	}
2837 	vp = (struct vnode *)fp->f_data;
2838 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2839 		FILE_UNUSE(fp, p);
2840 		return (error);
2841 	}
2842 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2843 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2844 	if (vp->v_type == VDIR)
2845 		error = EISDIR;
2846 	else if ((error = vn_writechk(vp)) == 0) {
2847 		VATTR_NULL(&vattr);
2848 		vattr.va_size = SCARG(uap, length);
2849 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2850 	}
2851 	VOP_UNLOCK(vp, 0);
2852 	vn_finished_write(mp, 0);
2853  out:
2854 	FILE_UNUSE(fp, p);
2855 	return (error);
2856 }
2857 
2858 /*
2859  * Sync an open file.
2860  */
2861 /* ARGSUSED */
2862 int
2863 sys_fsync(struct lwp *l, void *v, register_t *retval)
2864 {
2865 	struct sys_fsync_args /* {
2866 		syscallarg(int) fd;
2867 	} */ *uap = v;
2868 	struct proc *p = l->l_proc;
2869 	struct vnode *vp;
2870 	struct mount *mp;
2871 	struct file *fp;
2872 	int error;
2873 
2874 	/* getvnode() will use the descriptor for us */
2875 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2876 		return (error);
2877 	vp = (struct vnode *)fp->f_data;
2878 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2879 		FILE_UNUSE(fp, p);
2880 		return (error);
2881 	}
2882 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2883 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2884 	if (error == 0 && bioops.io_fsync != NULL &&
2885 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2886 		(*bioops.io_fsync)(vp, 0);
2887 	VOP_UNLOCK(vp, 0);
2888 	vn_finished_write(mp, 0);
2889 	FILE_UNUSE(fp, p);
2890 	return (error);
2891 }
2892 
2893 /*
2894  * Sync a range of file data.  API modeled after that found in AIX.
2895  *
2896  * FDATASYNC indicates that we need only save enough metadata to be able
2897  * to re-read the written data.  Note we duplicate AIX's requirement that
2898  * the file be open for writing.
2899  */
2900 /* ARGSUSED */
2901 int
2902 sys_fsync_range(struct lwp *l, void *v, register_t *retval)
2903 {
2904 	struct sys_fsync_range_args /* {
2905 		syscallarg(int) fd;
2906 		syscallarg(int) flags;
2907 		syscallarg(off_t) start;
2908 		syscallarg(off_t) length;
2909 	} */ *uap = v;
2910 	struct proc *p = l->l_proc;
2911 	struct vnode *vp;
2912 	struct file *fp;
2913 	int flags, nflags;
2914 	off_t s, e, len;
2915 	int error;
2916 
2917 	/* getvnode() will use the descriptor for us */
2918 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2919 		return (error);
2920 
2921 	if ((fp->f_flag & FWRITE) == 0) {
2922 		FILE_UNUSE(fp, p);
2923 		return (EBADF);
2924 	}
2925 
2926 	flags = SCARG(uap, flags);
2927 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
2928 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
2929 		return (EINVAL);
2930 	}
2931 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
2932 	if (flags & FDATASYNC)
2933 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
2934 	else
2935 		nflags = FSYNC_WAIT;
2936 	if (flags & FDISKSYNC)
2937 		nflags |= FSYNC_CACHE;
2938 
2939 	len = SCARG(uap, length);
2940 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
2941 	if (len) {
2942 		s = SCARG(uap, start);
2943 		e = s + len;
2944 		if (e < s) {
2945 			FILE_UNUSE(fp, p);
2946 			return (EINVAL);
2947 		}
2948 	} else {
2949 		e = 0;
2950 		s = 0;
2951 	}
2952 
2953 	vp = (struct vnode *)fp->f_data;
2954 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2955 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p);
2956 
2957 	if (error == 0 && bioops.io_fsync != NULL &&
2958 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2959 		(*bioops.io_fsync)(vp, nflags);
2960 
2961 	VOP_UNLOCK(vp, 0);
2962 	FILE_UNUSE(fp, p);
2963 	return (error);
2964 }
2965 
2966 /*
2967  * Sync the data of an open file.
2968  */
2969 /* ARGSUSED */
2970 int
2971 sys_fdatasync(struct lwp *l, void *v, register_t *retval)
2972 {
2973 	struct sys_fdatasync_args /* {
2974 		syscallarg(int) fd;
2975 	} */ *uap = v;
2976 	struct proc *p = l->l_proc;
2977 	struct vnode *vp;
2978 	struct file *fp;
2979 	int error;
2980 
2981 	/* getvnode() will use the descriptor for us */
2982 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2983 		return (error);
2984 	if ((fp->f_flag & FWRITE) == 0) {
2985 		FILE_UNUSE(fp, p);
2986 		return (EBADF);
2987 	}
2988 	vp = (struct vnode *)fp->f_data;
2989 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2990 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2991 	VOP_UNLOCK(vp, 0);
2992 	FILE_UNUSE(fp, p);
2993 	return (error);
2994 }
2995 
2996 /*
2997  * Rename files, (standard) BSD semantics frontend.
2998  */
2999 /* ARGSUSED */
3000 int
3001 sys_rename(struct lwp *l, void *v, register_t *retval)
3002 {
3003 	struct sys_rename_args /* {
3004 		syscallarg(const char *) from;
3005 		syscallarg(const char *) to;
3006 	} */ *uap = v;
3007 	struct proc *p = l->l_proc;
3008 
3009 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
3010 }
3011 
3012 /*
3013  * Rename files, POSIX semantics frontend.
3014  */
3015 /* ARGSUSED */
3016 int
3017 sys___posix_rename(struct lwp *l, void *v, register_t *retval)
3018 {
3019 	struct sys___posix_rename_args /* {
3020 		syscallarg(const char *) from;
3021 		syscallarg(const char *) to;
3022 	} */ *uap = v;
3023 	struct proc *p = l->l_proc;
3024 
3025 	return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
3026 }
3027 
3028 /*
3029  * Rename files.  Source and destination must either both be directories,
3030  * or both not be directories.  If target is a directory, it must be empty.
3031  * If `from' and `to' refer to the same object, the value of the `retain'
3032  * argument is used to determine whether `from' will be
3033  *
3034  * (retain == 0)	deleted unless `from' and `to' refer to the same
3035  *			object in the file system's name space (BSD).
3036  * (retain == 1)	always retained (POSIX).
3037  */
3038 static int
3039 rename_files(const char *from, const char *to, struct proc *p, int retain)
3040 {
3041 	struct mount *mp = NULL;
3042 	struct vnode *tvp, *fvp, *tdvp;
3043 	struct nameidata fromnd, tond;
3044 	int error;
3045 
3046 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
3047 	    from, p);
3048 	if ((error = namei(&fromnd)) != 0)
3049 		return (error);
3050 	fvp = fromnd.ni_vp;
3051 	error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH);
3052 	if (error != 0) {
3053 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3054 		vrele(fromnd.ni_dvp);
3055 		vrele(fvp);
3056 		if (fromnd.ni_startdir)
3057 			vrele(fromnd.ni_startdir);
3058 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3059 		return (error);
3060 	}
3061 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3062 	    (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p);
3063 	if ((error = namei(&tond)) != 0) {
3064 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3065 		vrele(fromnd.ni_dvp);
3066 		vrele(fvp);
3067 		goto out1;
3068 	}
3069 	tdvp = tond.ni_dvp;
3070 	tvp = tond.ni_vp;
3071 
3072 	if (tvp != NULL) {
3073 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3074 			error = ENOTDIR;
3075 			goto out;
3076 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3077 			error = EISDIR;
3078 			goto out;
3079 		}
3080 	}
3081 
3082 	if (fvp == tdvp)
3083 		error = EINVAL;
3084 
3085 	/*
3086 	 * Source and destination refer to the same object.
3087 	 */
3088 	if (fvp == tvp) {
3089 		if (retain)
3090 			error = -1;
3091 		else if (fromnd.ni_dvp == tdvp &&
3092 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3093 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3094 		          tond.ni_cnd.cn_nameptr,
3095 		          fromnd.ni_cnd.cn_namelen))
3096 		error = -1;
3097 	}
3098 
3099 out:
3100 	if (!error) {
3101 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
3102 		if (fromnd.ni_dvp != tdvp)
3103 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3104 		if (tvp) {
3105 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
3106 		}
3107 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3108 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3109 	} else {
3110 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3111 		if (tdvp == tvp)
3112 			vrele(tdvp);
3113 		else
3114 			vput(tdvp);
3115 		if (tvp)
3116 			vput(tvp);
3117 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3118 		vrele(fromnd.ni_dvp);
3119 		vrele(fvp);
3120 	}
3121 	vrele(tond.ni_startdir);
3122 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3123 out1:
3124 	vn_finished_write(mp, 0);
3125 	if (fromnd.ni_startdir)
3126 		vrele(fromnd.ni_startdir);
3127 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3128 	return (error == -1 ? 0 : error);
3129 }
3130 
3131 /*
3132  * Make a directory file.
3133  */
3134 /* ARGSUSED */
3135 int
3136 sys_mkdir(struct lwp *l, void *v, register_t *retval)
3137 {
3138 	struct sys_mkdir_args /* {
3139 		syscallarg(const char *) path;
3140 		syscallarg(int) mode;
3141 	} */ *uap = v;
3142 	struct proc *p = l->l_proc;
3143 	struct mount *mp;
3144 	struct vnode *vp;
3145 	struct vattr vattr;
3146 	int error;
3147 	struct nameidata nd;
3148 
3149 restart:
3150 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE,
3151 	    SCARG(uap, path), p);
3152 	if ((error = namei(&nd)) != 0)
3153 		return (error);
3154 	vp = nd.ni_vp;
3155 	if (vp != NULL) {
3156 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3157 		if (nd.ni_dvp == vp)
3158 			vrele(nd.ni_dvp);
3159 		else
3160 			vput(nd.ni_dvp);
3161 		vrele(vp);
3162 		return (EEXIST);
3163 	}
3164 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3165 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3166 		if (nd.ni_dvp == vp)
3167 			vrele(nd.ni_dvp);
3168 		else
3169 			vput(nd.ni_dvp);
3170 		if ((error = vn_start_write(NULL, &mp,
3171 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3172 			return (error);
3173 		goto restart;
3174 	}
3175 	VATTR_NULL(&vattr);
3176 	vattr.va_type = VDIR;
3177 	vattr.va_mode =
3178 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3179 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3180 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3181 	if (!error)
3182 		vput(nd.ni_vp);
3183 	vn_finished_write(mp, 0);
3184 	return (error);
3185 }
3186 
3187 /*
3188  * Remove a directory file.
3189  */
3190 /* ARGSUSED */
3191 int
3192 sys_rmdir(struct lwp *l, void *v, register_t *retval)
3193 {
3194 	struct sys_rmdir_args /* {
3195 		syscallarg(const char *) path;
3196 	} */ *uap = v;
3197 	struct proc *p = l->l_proc;
3198 	struct mount *mp;
3199 	struct vnode *vp;
3200 	int error;
3201 	struct nameidata nd;
3202 
3203 restart:
3204 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3205 	    SCARG(uap, path), p);
3206 	if ((error = namei(&nd)) != 0)
3207 		return (error);
3208 	vp = nd.ni_vp;
3209 	if (vp->v_type != VDIR) {
3210 		error = ENOTDIR;
3211 		goto out;
3212 	}
3213 	/*
3214 	 * No rmdir "." please.
3215 	 */
3216 	if (nd.ni_dvp == vp) {
3217 		error = EINVAL;
3218 		goto out;
3219 	}
3220 	/*
3221 	 * The root of a mounted filesystem cannot be deleted.
3222 	 */
3223 	if (vp->v_flag & VROOT) {
3224 		error = EBUSY;
3225 		goto out;
3226 	}
3227 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3228 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3229 		if (nd.ni_dvp == vp)
3230 			vrele(nd.ni_dvp);
3231 		else
3232 			vput(nd.ni_dvp);
3233 		vput(vp);
3234 		if ((error = vn_start_write(NULL, &mp,
3235 		    V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3236 			return (error);
3237 		goto restart;
3238 	}
3239 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3240 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3241 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3242 	vn_finished_write(mp, 0);
3243 	return (error);
3244 
3245 out:
3246 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3247 	if (nd.ni_dvp == vp)
3248 		vrele(nd.ni_dvp);
3249 	else
3250 		vput(nd.ni_dvp);
3251 	vput(vp);
3252 	return (error);
3253 }
3254 
3255 /*
3256  * Read a block of directory entries in a file system independent format.
3257  */
3258 int
3259 sys_getdents(struct lwp *l, void *v, register_t *retval)
3260 {
3261 	struct sys_getdents_args /* {
3262 		syscallarg(int) fd;
3263 		syscallarg(char *) buf;
3264 		syscallarg(size_t) count;
3265 	} */ *uap = v;
3266 	struct proc *p = l->l_proc;
3267 	struct file *fp;
3268 	int error, done;
3269 
3270 	/* getvnode() will use the descriptor for us */
3271 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3272 		return (error);
3273 	if ((fp->f_flag & FREAD) == 0) {
3274 		error = EBADF;
3275 		goto out;
3276 	}
3277 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3278 			SCARG(uap, count), &done, p, 0, 0);
3279 #ifdef KTRACE
3280 	if (!error && KTRPOINT(p, KTR_GENIO)) {
3281 		struct iovec iov;
3282 		iov.iov_base = SCARG(uap, buf);
3283 		iov.iov_len = done;
3284 		ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0);
3285 	}
3286 #endif
3287 	*retval = done;
3288  out:
3289 	FILE_UNUSE(fp, p);
3290 	return (error);
3291 }
3292 
3293 /*
3294  * Set the mode mask for creation of filesystem nodes.
3295  */
3296 int
3297 sys_umask(struct lwp *l, void *v, register_t *retval)
3298 {
3299 	struct sys_umask_args /* {
3300 		syscallarg(mode_t) newmask;
3301 	} */ *uap = v;
3302 	struct proc *p = l->l_proc;
3303 	struct cwdinfo *cwdi;
3304 
3305 	cwdi = p->p_cwdi;
3306 	*retval = cwdi->cwdi_cmask;
3307 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3308 	return (0);
3309 }
3310 
3311 /*
3312  * Void all references to file by ripping underlying filesystem
3313  * away from vnode.
3314  */
3315 /* ARGSUSED */
3316 int
3317 sys_revoke(struct lwp *l, void *v, register_t *retval)
3318 {
3319 	struct sys_revoke_args /* {
3320 		syscallarg(const char *) path;
3321 	} */ *uap = v;
3322 	struct proc *p = l->l_proc;
3323 	struct mount *mp;
3324 	struct vnode *vp;
3325 	struct vattr vattr;
3326 	int error;
3327 	struct nameidata nd;
3328 
3329 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3330 	if ((error = namei(&nd)) != 0)
3331 		return (error);
3332 	vp = nd.ni_vp;
3333 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3334 		goto out;
3335 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3336 	    (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3337 		goto out;
3338 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3339 		goto out;
3340 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3341 		VOP_REVOKE(vp, REVOKEALL);
3342 	vn_finished_write(mp, 0);
3343 out:
3344 	vrele(vp);
3345 	return (error);
3346 }
3347 
3348 /*
3349  * Convert a user file descriptor to a kernel file entry.
3350  */
3351 int
3352 getvnode(struct filedesc *fdp, int fd, struct file **fpp)
3353 {
3354 	struct vnode *vp;
3355 	struct file *fp;
3356 
3357 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3358 		return (EBADF);
3359 
3360 	FILE_USE(fp);
3361 
3362 	if (fp->f_type != DTYPE_VNODE) {
3363 		FILE_UNUSE(fp, NULL);
3364 		return (EINVAL);
3365 	}
3366 
3367 	vp = (struct vnode *)fp->f_data;
3368 	if (vp->v_type == VBAD) {
3369 		FILE_UNUSE(fp, NULL);
3370 		return (EBADF);
3371 	}
3372 
3373 	*fpp = fp;
3374 	return (0);
3375 }
3376