xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision a5847cc334d9a7029f6352b847e9e8d71a0f9e0c)
1 /*	$NetBSD: vfs_syscalls.c,v 1.440 2011/10/14 09:23:31 hannken Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1989, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  * (c) UNIX System Laboratories, Inc.
36  * All or some portions of this file are derived from material licensed
37  * to the University of California by American Telephone and Telegraph
38  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
39  * the permission of UNIX System Laboratories, Inc.
40  *
41  * Redistribution and use in source and binary forms, with or without
42  * modification, are permitted provided that the following conditions
43  * are met:
44  * 1. Redistributions of source code must retain the above copyright
45  *    notice, this list of conditions and the following disclaimer.
46  * 2. Redistributions in binary form must reproduce the above copyright
47  *    notice, this list of conditions and the following disclaimer in the
48  *    documentation and/or other materials provided with the distribution.
49  * 3. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
66  */
67 
68 /*
69  * Virtual File System System Calls
70  */
71 
72 #include <sys/cdefs.h>
73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.440 2011/10/14 09:23:31 hannken Exp $");
74 
75 #ifdef _KERNEL_OPT
76 #include "opt_fileassoc.h"
77 #include "veriexec.h"
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/namei.h>
83 #include <sys/filedesc.h>
84 #include <sys/kernel.h>
85 #include <sys/file.h>
86 #include <sys/fcntl.h>
87 #include <sys/stat.h>
88 #include <sys/vnode.h>
89 #include <sys/mount.h>
90 #include <sys/proc.h>
91 #include <sys/uio.h>
92 #include <sys/kmem.h>
93 #include <sys/dirent.h>
94 #include <sys/sysctl.h>
95 #include <sys/syscallargs.h>
96 #include <sys/vfs_syscalls.h>
97 #include <sys/ktrace.h>
98 #ifdef FILEASSOC
99 #include <sys/fileassoc.h>
100 #endif /* FILEASSOC */
101 #include <sys/extattr.h>
102 #include <sys/verified_exec.h>
103 #include <sys/kauth.h>
104 #include <sys/atomic.h>
105 #include <sys/module.h>
106 #include <sys/buf.h>
107 
108 #include <miscfs/genfs/genfs.h>
109 #include <miscfs/syncfs/syncfs.h>
110 #include <miscfs/specfs/specdev.h>
111 
112 #include <nfs/rpcv2.h>
113 #include <nfs/nfsproto.h>
114 #include <nfs/nfs.h>
115 #include <nfs/nfs_var.h>
116 
117 static int change_flags(struct vnode *, u_long, struct lwp *);
118 static int change_mode(struct vnode *, int, struct lwp *l);
119 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
120 
121 /*
122  * This table is used to maintain compatibility with 4.3BSD
123  * and NetBSD 0.9 mount syscalls - and possibly other systems.
124  * Note, the order is important!
125  *
126  * Do not modify this table. It should only contain filesystems
127  * supported by NetBSD 0.9 and 4.3BSD.
128  */
129 const char * const mountcompatnames[] = {
130 	NULL,		/* 0 = MOUNT_NONE */
131 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
132 	MOUNT_NFS,	/* 2 */
133 	MOUNT_MFS,	/* 3 */
134 	MOUNT_MSDOS,	/* 4 */
135 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
136 	MOUNT_FDESC,	/* 6 */
137 	MOUNT_KERNFS,	/* 7 */
138 	NULL,		/* 8 = MOUNT_DEVFS */
139 	MOUNT_AFS,	/* 9 */
140 };
141 
142 const int nmountcompatnames = __arraycount(mountcompatnames);
143 
144 static int
145 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags)
146 {
147 	int error;
148 
149 	fp->f_flag = flags & FMASK;
150 	fp->f_type = DTYPE_VNODE;
151 	fp->f_ops = &vnops;
152 	fp->f_data = vp;
153 
154 	if (flags & (O_EXLOCK | O_SHLOCK)) {
155 		struct flock lf;
156 		int type;
157 
158 		lf.l_whence = SEEK_SET;
159 		lf.l_start = 0;
160 		lf.l_len = 0;
161 		if (flags & O_EXLOCK)
162 			lf.l_type = F_WRLCK;
163 		else
164 			lf.l_type = F_RDLCK;
165 		type = F_FLOCK;
166 		if ((flags & FNONBLOCK) == 0)
167 			type |= F_WAIT;
168 		VOP_UNLOCK(vp);
169 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
170 		if (error) {
171 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
172 			fd_abort(l->l_proc, fp, indx);
173 			return error;
174 		}
175 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
176 		atomic_or_uint(&fp->f_flag, FHASLOCK);
177 	}
178 	if (flags & O_CLOEXEC)
179 		fd_set_exclose(l, indx, true);
180 	return 0;
181 }
182 
183 static int
184 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
185     void *data, size_t *data_len)
186 {
187 	struct mount *mp;
188 	int error = 0, saved_flags;
189 
190 	mp = vp->v_mount;
191 	saved_flags = mp->mnt_flag;
192 
193 	/* We can operate only on VV_ROOT nodes. */
194 	if ((vp->v_vflag & VV_ROOT) == 0) {
195 		error = EINVAL;
196 		goto out;
197 	}
198 
199 	/*
200 	 * We only allow the filesystem to be reloaded if it
201 	 * is currently mounted read-only.  Additionally, we
202 	 * prevent read-write to read-only downgrades.
203 	 */
204 	if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
205 	    (mp->mnt_flag & MNT_RDONLY) == 0 &&
206 	    (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) {
207 		error = EOPNOTSUPP;	/* Needs translation */
208 		goto out;
209 	}
210 
211 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
212 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
213 	if (error)
214 		goto out;
215 
216 	if (vfs_busy(mp, NULL)) {
217 		error = EPERM;
218 		goto out;
219 	}
220 
221 	mutex_enter(&mp->mnt_updating);
222 
223 	mp->mnt_flag &= ~MNT_OP_FLAGS;
224 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
225 
226 	/*
227 	 * Set the mount level flags.
228 	 */
229 	if (flags & MNT_RDONLY)
230 		mp->mnt_flag |= MNT_RDONLY;
231 	else if (mp->mnt_flag & MNT_RDONLY)
232 		mp->mnt_iflag |= IMNT_WANTRDWR;
233 	mp->mnt_flag &=
234 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
235 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
236 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
237 	    MNT_LOG | MNT_EXTATTR);
238 	mp->mnt_flag |= flags &
239 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
240 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
241 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
242 	    MNT_LOG | MNT_EXTATTR | MNT_IGNORE);
243 
244 	error = VFS_MOUNT(mp, path, data, data_len);
245 
246 	if (error && data != NULL) {
247 		int error2;
248 
249 		/*
250 		 * Update failed; let's try and see if it was an
251 		 * export request.  For compat with 3.0 and earlier.
252 		 */
253 		error2 = vfs_hooks_reexport(mp, path, data);
254 
255 		/*
256 		 * Only update error code if the export request was
257 		 * understood but some problem occurred while
258 		 * processing it.
259 		 */
260 		if (error2 != EJUSTRETURN)
261 			error = error2;
262 	}
263 
264 	if (mp->mnt_iflag & IMNT_WANTRDWR)
265 		mp->mnt_flag &= ~MNT_RDONLY;
266 	if (error)
267 		mp->mnt_flag = saved_flags;
268 	mp->mnt_flag &= ~MNT_OP_FLAGS;
269 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
270 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
271 		if (mp->mnt_syncer == NULL)
272 			error = vfs_allocate_syncvnode(mp);
273 	} else {
274 		if (mp->mnt_syncer != NULL)
275 			vfs_deallocate_syncvnode(mp);
276 	}
277 	mutex_exit(&mp->mnt_updating);
278 	vfs_unbusy(mp, false, NULL);
279 
280 	if ((error == 0) && !(saved_flags & MNT_EXTATTR) &&
281 	    (flags & MNT_EXTATTR)) {
282 		if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_START,
283 				   NULL, 0, NULL) != 0) {
284 			printf("%s: failed to start extattr, error = %d",
285 			       vp->v_mount->mnt_stat.f_mntonname, error);
286 			mp->mnt_flag &= ~MNT_EXTATTR;
287 		}
288 	}
289 
290 	if ((error == 0) && (saved_flags & MNT_EXTATTR) &&
291 	    !(flags & MNT_EXTATTR)) {
292 		if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_STOP,
293 				   NULL, 0, NULL) != 0) {
294 			printf("%s: failed to stop extattr, error = %d",
295 			       vp->v_mount->mnt_stat.f_mntonname, error);
296 			mp->mnt_flag |= MNT_RDONLY;
297 		}
298 	}
299  out:
300 	return (error);
301 }
302 
303 static int
304 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
305 {
306 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
307 	int error;
308 
309 	/* Copy file-system type from userspace.  */
310 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
311 	if (error) {
312 		/*
313 		 * Historically, filesystem types were identified by numbers.
314 		 * If we get an integer for the filesystem type instead of a
315 		 * string, we check to see if it matches one of the historic
316 		 * filesystem types.
317 		 */
318 		u_long fsindex = (u_long)fstype;
319 		if (fsindex >= nmountcompatnames ||
320 		    mountcompatnames[fsindex] == NULL)
321 			return ENODEV;
322 		strlcpy(fstypename, mountcompatnames[fsindex],
323 		    sizeof(fstypename));
324 	}
325 
326 	/* Accept `ufs' as an alias for `ffs', for compatibility. */
327 	if (strcmp(fstypename, "ufs") == 0)
328 		fstypename[0] = 'f';
329 
330 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
331 		return 0;
332 
333 	/* If we can autoload a vfs module, try again */
334 	(void)module_autoload(fstypename, MODULE_CLASS_VFS);
335 
336 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
337 		return 0;
338 
339 	return ENODEV;
340 }
341 
342 static int
343 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
344     void *data, size_t *data_len)
345 {
346 	struct mount *mp;
347 	int error;
348 
349 	/* If MNT_GETARGS is specified, it should be the only flag. */
350 	if (flags & ~MNT_GETARGS)
351 		return EINVAL;
352 
353 	mp = vp->v_mount;
354 
355 	/* XXX: probably some notion of "can see" here if we want isolation. */
356 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
357 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
358 	if (error)
359 		return error;
360 
361 	if ((vp->v_vflag & VV_ROOT) == 0)
362 		return EINVAL;
363 
364 	if (vfs_busy(mp, NULL))
365 		return EPERM;
366 
367 	mutex_enter(&mp->mnt_updating);
368 	mp->mnt_flag &= ~MNT_OP_FLAGS;
369 	mp->mnt_flag |= MNT_GETARGS;
370 	error = VFS_MOUNT(mp, path, data, data_len);
371 	mp->mnt_flag &= ~MNT_OP_FLAGS;
372 	mutex_exit(&mp->mnt_updating);
373 
374 	vfs_unbusy(mp, false, NULL);
375 	return (error);
376 }
377 
378 int
379 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
380 {
381 	/* {
382 		syscallarg(const char *) type;
383 		syscallarg(const char *) path;
384 		syscallarg(int) flags;
385 		syscallarg(void *) data;
386 		syscallarg(size_t) data_len;
387 	} */
388 
389 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
390 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
391 	    SCARG(uap, data_len), retval);
392 }
393 
394 int
395 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
396     const char *path, int flags, void *data, enum uio_seg data_seg,
397     size_t data_len, register_t *retval)
398 {
399 	struct vnode *vp;
400 	void *data_buf = data;
401 	bool vfsopsrele = false;
402 	int error;
403 
404 	/* XXX: The calling convention of this routine is totally bizarre */
405 	if (vfsops)
406 		vfsopsrele = true;
407 
408 	/*
409 	 * Get vnode to be covered
410 	 */
411 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
412 	if (error != 0) {
413 		vp = NULL;
414 		goto done;
415 	}
416 
417 	if (vfsops == NULL) {
418 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
419 			vfsops = vp->v_mount->mnt_op;
420 		} else {
421 			/* 'type' is userspace */
422 			error = mount_get_vfsops(type, &vfsops);
423 			if (error != 0)
424 				goto done;
425 			vfsopsrele = true;
426 		}
427 	}
428 
429 	if (data != NULL && data_seg == UIO_USERSPACE) {
430 		if (data_len == 0) {
431 			/* No length supplied, use default for filesystem */
432 			data_len = vfsops->vfs_min_mount_data;
433 			if (data_len > VFS_MAX_MOUNT_DATA) {
434 				error = EINVAL;
435 				goto done;
436 			}
437 			/*
438 			 * Hopefully a longer buffer won't make copyin() fail.
439 			 * For compatibility with 3.0 and earlier.
440 			 */
441 			if (flags & MNT_UPDATE
442 			    && data_len < sizeof (struct mnt_export_args30))
443 				data_len = sizeof (struct mnt_export_args30);
444 		}
445 		data_buf = kmem_alloc(data_len, KM_SLEEP);
446 
447 		/* NFS needs the buffer even for mnt_getargs .... */
448 		error = copyin(data, data_buf, data_len);
449 		if (error != 0)
450 			goto done;
451 	}
452 
453 	if (flags & MNT_GETARGS) {
454 		if (data_len == 0) {
455 			error = EINVAL;
456 			goto done;
457 		}
458 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
459 		if (error != 0)
460 			goto done;
461 		if (data_seg == UIO_USERSPACE)
462 			error = copyout(data_buf, data, data_len);
463 		*retval = data_len;
464 	} else if (flags & MNT_UPDATE) {
465 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
466 	} else {
467 		/* Locking is handled internally in mount_domount(). */
468 		KASSERT(vfsopsrele == true);
469 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
470 		    &data_len);
471 		vfsopsrele = false;
472 
473 		if ((error == 0) && (flags & MNT_EXTATTR)) {
474 			if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_START,
475 					   NULL, 0, NULL) != 0)
476 				printf("%s: failed to start extattr",
477 				       vp->v_mount->mnt_stat.f_mntonname);
478 				/* XXX remove flag */
479 		}
480 	}
481 
482     done:
483 	if (vfsopsrele)
484 		vfs_delref(vfsops);
485     	if (vp != NULL) {
486 	    	vrele(vp);
487 	}
488 	if (data_buf != data)
489 		kmem_free(data_buf, data_len);
490 	return (error);
491 }
492 
493 /*
494  * Unmount a file system.
495  *
496  * Note: unmount takes a path to the vnode mounted on as argument,
497  * not special file (as before).
498  */
499 /* ARGSUSED */
500 int
501 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
502 {
503 	/* {
504 		syscallarg(const char *) path;
505 		syscallarg(int) flags;
506 	} */
507 	struct vnode *vp;
508 	struct mount *mp;
509 	int error;
510 	struct pathbuf *pb;
511 	struct nameidata nd;
512 
513 	error = pathbuf_copyin(SCARG(uap, path), &pb);
514 	if (error) {
515 		return error;
516 	}
517 
518 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
519 	if ((error = namei(&nd)) != 0) {
520 		pathbuf_destroy(pb);
521 		return error;
522 	}
523 	vp = nd.ni_vp;
524 	pathbuf_destroy(pb);
525 
526 	mp = vp->v_mount;
527 	atomic_inc_uint(&mp->mnt_refcnt);
528 	VOP_UNLOCK(vp);
529 
530 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
531 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
532 	if (error) {
533 		vrele(vp);
534 		vfs_destroy(mp);
535 		return (error);
536 	}
537 
538 	/*
539 	 * Don't allow unmounting the root file system.
540 	 */
541 	if (mp->mnt_flag & MNT_ROOTFS) {
542 		vrele(vp);
543 		vfs_destroy(mp);
544 		return (EINVAL);
545 	}
546 
547 	/*
548 	 * Must be the root of the filesystem
549 	 */
550 	if ((vp->v_vflag & VV_ROOT) == 0) {
551 		vrele(vp);
552 		vfs_destroy(mp);
553 		return (EINVAL);
554 	}
555 
556 	vrele(vp);
557 	error = dounmount(mp, SCARG(uap, flags), l);
558 	vfs_destroy(mp);
559 	return error;
560 }
561 
562 /*
563  * Sync each mounted filesystem.
564  */
565 #ifdef DEBUG
566 int syncprt = 0;
567 struct ctldebug debug0 = { "syncprt", &syncprt };
568 #endif
569 
570 void
571 do_sys_sync(struct lwp *l)
572 {
573 	struct mount *mp, *nmp;
574 	int asyncflag;
575 
576 	mutex_enter(&mountlist_lock);
577 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
578 	     mp = nmp) {
579 		if (vfs_busy(mp, &nmp)) {
580 			continue;
581 		}
582 		mutex_enter(&mp->mnt_updating);
583 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
584 			asyncflag = mp->mnt_flag & MNT_ASYNC;
585 			mp->mnt_flag &= ~MNT_ASYNC;
586 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
587 			if (asyncflag)
588 				 mp->mnt_flag |= MNT_ASYNC;
589 		}
590 		mutex_exit(&mp->mnt_updating);
591 		vfs_unbusy(mp, false, &nmp);
592 	}
593 	mutex_exit(&mountlist_lock);
594 #ifdef DEBUG
595 	if (syncprt)
596 		vfs_bufstats();
597 #endif /* DEBUG */
598 }
599 
600 /* ARGSUSED */
601 int
602 sys_sync(struct lwp *l, const void *v, register_t *retval)
603 {
604 	do_sys_sync(l);
605 	return (0);
606 }
607 
608 
609 /*
610  * Change filesystem quotas.
611  */
612 /* ARGSUSED */
613 int
614 sys___quotactl50(struct lwp *l, const struct sys___quotactl50_args *uap,
615     register_t *retval)
616 {
617 	/* {
618 		syscallarg(const char *) path;
619 		syscallarg(struct plistref *) pref;
620 	} */
621 	struct mount *mp;
622 	int error;
623 	struct vnode *vp;
624 	prop_dictionary_t dict;
625 	struct plistref pref;
626 
627 	error = namei_simple_user(SCARG(uap, path),
628 				NSM_FOLLOW_TRYEMULROOT, &vp);
629 	if (error != 0)
630 		return (error);
631 	mp = vp->v_mount;
632 	error = copyin(SCARG(uap, pref), &pref, sizeof(pref));
633 	if (error)
634 		return error;
635 	error = prop_dictionary_copyin(&pref, &dict);
636 	if (error)
637 		return error;
638 	error = VFS_QUOTACTL(mp, dict);
639 	vrele(vp);
640 	if (!error)
641 		error = prop_dictionary_copyout(&pref, dict);
642 	if (!error)
643 		error = copyout(&pref, SCARG(uap, pref), sizeof(pref));
644 	prop_object_release(dict);
645 	return (error);
646 }
647 
648 int
649 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
650     int root)
651 {
652 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
653 	int error = 0;
654 
655 	/*
656 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
657 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
658 	 * overrides MNT_NOWAIT.
659 	 */
660 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
661 	    (flags != MNT_WAIT && flags != 0)) {
662 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
663 		goto done;
664 	}
665 
666 	/* Get the filesystem stats now */
667 	memset(sp, 0, sizeof(*sp));
668 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
669 		return error;
670 	}
671 
672 	if (cwdi->cwdi_rdir == NULL)
673 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
674 done:
675 	if (cwdi->cwdi_rdir != NULL) {
676 		size_t len;
677 		char *bp;
678 		char c;
679 		char *path = PNBUF_GET();
680 
681 		bp = path + MAXPATHLEN;
682 		*--bp = '\0';
683 		rw_enter(&cwdi->cwdi_lock, RW_READER);
684 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
685 		    MAXPATHLEN / 2, 0, l);
686 		rw_exit(&cwdi->cwdi_lock);
687 		if (error) {
688 			PNBUF_PUT(path);
689 			return error;
690 		}
691 		len = strlen(bp);
692 		if (len != 1) {
693 			/*
694 			 * for mount points that are below our root, we can see
695 			 * them, so we fix up the pathname and return them. The
696 			 * rest we cannot see, so we don't allow viewing the
697 			 * data.
698 			 */
699 			if (strncmp(bp, sp->f_mntonname, len) == 0 &&
700 			    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
701 				(void)strlcpy(sp->f_mntonname,
702 				    c == '\0' ? "/" : &sp->f_mntonname[len],
703 				    sizeof(sp->f_mntonname));
704 			} else {
705 				if (root)
706 					(void)strlcpy(sp->f_mntonname, "/",
707 					    sizeof(sp->f_mntonname));
708 				else
709 					error = EPERM;
710 			}
711 		}
712 		PNBUF_PUT(path);
713 	}
714 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
715 	return error;
716 }
717 
718 /*
719  * Get filesystem statistics by path.
720  */
721 int
722 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
723 {
724 	struct mount *mp;
725 	int error;
726 	struct vnode *vp;
727 
728 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
729 	if (error != 0)
730 		return error;
731 	mp = vp->v_mount;
732 	error = dostatvfs(mp, sb, l, flags, 1);
733 	vrele(vp);
734 	return error;
735 }
736 
737 /* ARGSUSED */
738 int
739 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
740 {
741 	/* {
742 		syscallarg(const char *) path;
743 		syscallarg(struct statvfs *) buf;
744 		syscallarg(int) flags;
745 	} */
746 	struct statvfs *sb;
747 	int error;
748 
749 	sb = STATVFSBUF_GET();
750 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
751 	if (error == 0)
752 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
753 	STATVFSBUF_PUT(sb);
754 	return error;
755 }
756 
757 /*
758  * Get filesystem statistics by fd.
759  */
760 int
761 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
762 {
763 	file_t *fp;
764 	struct mount *mp;
765 	int error;
766 
767 	/* fd_getvnode() will use the descriptor for us */
768 	if ((error = fd_getvnode(fd, &fp)) != 0)
769 		return (error);
770 	mp = ((struct vnode *)fp->f_data)->v_mount;
771 	error = dostatvfs(mp, sb, curlwp, flags, 1);
772 	fd_putfile(fd);
773 	return error;
774 }
775 
776 /* ARGSUSED */
777 int
778 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
779 {
780 	/* {
781 		syscallarg(int) fd;
782 		syscallarg(struct statvfs *) buf;
783 		syscallarg(int) flags;
784 	} */
785 	struct statvfs *sb;
786 	int error;
787 
788 	sb = STATVFSBUF_GET();
789 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
790 	if (error == 0)
791 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
792 	STATVFSBUF_PUT(sb);
793 	return error;
794 }
795 
796 
797 /*
798  * Get statistics on all filesystems.
799  */
800 int
801 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
802     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
803     register_t *retval)
804 {
805 	int root = 0;
806 	struct proc *p = l->l_proc;
807 	struct mount *mp, *nmp;
808 	struct statvfs *sb;
809 	size_t count, maxcount;
810 	int error = 0;
811 
812 	sb = STATVFSBUF_GET();
813 	maxcount = bufsize / entry_sz;
814 	mutex_enter(&mountlist_lock);
815 	count = 0;
816 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
817 	     mp = nmp) {
818 		if (vfs_busy(mp, &nmp)) {
819 			continue;
820 		}
821 		if (sfsp && count < maxcount) {
822 			error = dostatvfs(mp, sb, l, flags, 0);
823 			if (error) {
824 				vfs_unbusy(mp, false, &nmp);
825 				error = 0;
826 				continue;
827 			}
828 			error = copyfn(sb, sfsp, entry_sz);
829 			if (error) {
830 				vfs_unbusy(mp, false, NULL);
831 				goto out;
832 			}
833 			sfsp = (char *)sfsp + entry_sz;
834 			root |= strcmp(sb->f_mntonname, "/") == 0;
835 		}
836 		count++;
837 		vfs_unbusy(mp, false, &nmp);
838 	}
839 	mutex_exit(&mountlist_lock);
840 
841 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
842 		/*
843 		 * fake a root entry
844 		 */
845 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
846 		    sb, l, flags, 1);
847 		if (error != 0)
848 			goto out;
849 		if (sfsp) {
850 			error = copyfn(sb, sfsp, entry_sz);
851 			if (error != 0)
852 				goto out;
853 		}
854 		count++;
855 	}
856 	if (sfsp && count > maxcount)
857 		*retval = maxcount;
858 	else
859 		*retval = count;
860 out:
861 	STATVFSBUF_PUT(sb);
862 	return error;
863 }
864 
865 int
866 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
867 {
868 	/* {
869 		syscallarg(struct statvfs *) buf;
870 		syscallarg(size_t) bufsize;
871 		syscallarg(int) flags;
872 	} */
873 
874 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
875 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
876 }
877 
878 /*
879  * Change current working directory to a given file descriptor.
880  */
881 /* ARGSUSED */
882 int
883 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
884 {
885 	/* {
886 		syscallarg(int) fd;
887 	} */
888 	struct proc *p = l->l_proc;
889 	struct cwdinfo *cwdi;
890 	struct vnode *vp, *tdp;
891 	struct mount *mp;
892 	file_t *fp;
893 	int error, fd;
894 
895 	/* fd_getvnode() will use the descriptor for us */
896 	fd = SCARG(uap, fd);
897 	if ((error = fd_getvnode(fd, &fp)) != 0)
898 		return (error);
899 	vp = fp->f_data;
900 
901 	vref(vp);
902 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
903 	if (vp->v_type != VDIR)
904 		error = ENOTDIR;
905 	else
906 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
907 	if (error) {
908 		vput(vp);
909 		goto out;
910 	}
911 	while ((mp = vp->v_mountedhere) != NULL) {
912 		error = vfs_busy(mp, NULL);
913 		vput(vp);
914 		if (error != 0)
915 			goto out;
916 		error = VFS_ROOT(mp, &tdp);
917 		vfs_unbusy(mp, false, NULL);
918 		if (error)
919 			goto out;
920 		vp = tdp;
921 	}
922 	VOP_UNLOCK(vp);
923 
924 	/*
925 	 * Disallow changing to a directory not under the process's
926 	 * current root directory (if there is one).
927 	 */
928 	cwdi = p->p_cwdi;
929 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
930 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
931 		vrele(vp);
932 		error = EPERM;	/* operation not permitted */
933 	} else {
934 		vrele(cwdi->cwdi_cdir);
935 		cwdi->cwdi_cdir = vp;
936 	}
937 	rw_exit(&cwdi->cwdi_lock);
938 
939  out:
940 	fd_putfile(fd);
941 	return (error);
942 }
943 
944 /*
945  * Change this process's notion of the root directory to a given file
946  * descriptor.
947  */
948 int
949 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
950 {
951 	struct proc *p = l->l_proc;
952 	struct vnode	*vp;
953 	file_t	*fp;
954 	int		 error, fd = SCARG(uap, fd);
955 
956 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
957  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
958 		return error;
959 	/* fd_getvnode() will use the descriptor for us */
960 	if ((error = fd_getvnode(fd, &fp)) != 0)
961 		return error;
962 	vp = fp->f_data;
963 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
964 	if (vp->v_type != VDIR)
965 		error = ENOTDIR;
966 	else
967 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
968 	VOP_UNLOCK(vp);
969 	if (error)
970 		goto out;
971 	vref(vp);
972 
973 	change_root(p->p_cwdi, vp, l);
974 
975  out:
976 	fd_putfile(fd);
977 	return (error);
978 }
979 
980 /*
981  * Change current working directory (``.'').
982  */
983 /* ARGSUSED */
984 int
985 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
986 {
987 	/* {
988 		syscallarg(const char *) path;
989 	} */
990 	struct proc *p = l->l_proc;
991 	struct cwdinfo *cwdi;
992 	int error;
993 	struct vnode *vp;
994 
995 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
996 				  &vp, l)) != 0)
997 		return (error);
998 	cwdi = p->p_cwdi;
999 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1000 	vrele(cwdi->cwdi_cdir);
1001 	cwdi->cwdi_cdir = vp;
1002 	rw_exit(&cwdi->cwdi_lock);
1003 	return (0);
1004 }
1005 
1006 /*
1007  * Change notion of root (``/'') directory.
1008  */
1009 /* ARGSUSED */
1010 int
1011 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1012 {
1013 	/* {
1014 		syscallarg(const char *) path;
1015 	} */
1016 	struct proc *p = l->l_proc;
1017 	int error;
1018 	struct vnode *vp;
1019 
1020 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1021 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1022 		return (error);
1023 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1024 				  &vp, l)) != 0)
1025 		return (error);
1026 
1027 	change_root(p->p_cwdi, vp, l);
1028 
1029 	return (0);
1030 }
1031 
1032 /*
1033  * Common routine for chroot and fchroot.
1034  * NB: callers need to properly authorize the change root operation.
1035  */
1036 void
1037 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1038 {
1039 
1040 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1041 	if (cwdi->cwdi_rdir != NULL)
1042 		vrele(cwdi->cwdi_rdir);
1043 	cwdi->cwdi_rdir = vp;
1044 
1045 	/*
1046 	 * Prevent escaping from chroot by putting the root under
1047 	 * the working directory.  Silently chdir to / if we aren't
1048 	 * already there.
1049 	 */
1050 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1051 		/*
1052 		 * XXX would be more failsafe to change directory to a
1053 		 * deadfs node here instead
1054 		 */
1055 		vrele(cwdi->cwdi_cdir);
1056 		vref(vp);
1057 		cwdi->cwdi_cdir = vp;
1058 	}
1059 	rw_exit(&cwdi->cwdi_lock);
1060 }
1061 
1062 /*
1063  * Common routine for chroot and chdir.
1064  * XXX "where" should be enum uio_seg
1065  */
1066 int
1067 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1068 {
1069 	struct pathbuf *pb;
1070 	struct nameidata nd;
1071 	int error;
1072 
1073 	error = pathbuf_maybe_copyin(path, where, &pb);
1074 	if (error) {
1075 		return error;
1076 	}
1077 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1078 	if ((error = namei(&nd)) != 0) {
1079 		pathbuf_destroy(pb);
1080 		return error;
1081 	}
1082 	*vpp = nd.ni_vp;
1083 	pathbuf_destroy(pb);
1084 
1085 	if ((*vpp)->v_type != VDIR)
1086 		error = ENOTDIR;
1087 	else
1088 		error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1089 
1090 	if (error)
1091 		vput(*vpp);
1092 	else
1093 		VOP_UNLOCK(*vpp);
1094 	return (error);
1095 }
1096 
1097 /*
1098  * Check permissions, allocate an open file structure,
1099  * and call the device open routine if any.
1100  */
1101 int
1102 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1103 {
1104 	/* {
1105 		syscallarg(const char *) path;
1106 		syscallarg(int) flags;
1107 		syscallarg(int) mode;
1108 	} */
1109 	struct proc *p = l->l_proc;
1110 	struct cwdinfo *cwdi = p->p_cwdi;
1111 	file_t *fp;
1112 	struct vnode *vp;
1113 	int flags, cmode;
1114 	int indx, error;
1115 	struct pathbuf *pb;
1116 	struct nameidata nd;
1117 
1118 	flags = FFLAGS(SCARG(uap, flags));
1119 	if ((flags & (FREAD | FWRITE)) == 0)
1120 		return (EINVAL);
1121 
1122 	error = pathbuf_copyin(SCARG(uap, path), &pb);
1123 	if (error) {
1124 		return error;
1125 	}
1126 
1127 	if ((error = fd_allocfile(&fp, &indx)) != 0) {
1128 		pathbuf_destroy(pb);
1129 		return error;
1130 	}
1131 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1132 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1133 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb);
1134 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1135 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1136 		fd_abort(p, fp, indx);
1137 		if ((error == EDUPFD || error == EMOVEFD) &&
1138 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1139 		    (error =
1140 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1141 			*retval = indx;
1142 			pathbuf_destroy(pb);
1143 			return (0);
1144 		}
1145 		if (error == ERESTART)
1146 			error = EINTR;
1147 		pathbuf_destroy(pb);
1148 		return (error);
1149 	}
1150 
1151 	l->l_dupfd = 0;
1152 	vp = nd.ni_vp;
1153 	pathbuf_destroy(pb);
1154 
1155 	if ((error = open_setfp(l, fp, vp, indx, flags)))
1156 		return error;
1157 
1158 	VOP_UNLOCK(vp);
1159 	*retval = indx;
1160 	fd_affix(p, fp, indx);
1161 	return (0);
1162 }
1163 
1164 int
1165 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval)
1166 {
1167 	/* {
1168 		syscallarg(int) fd;
1169 		syscallarg(const char *) path;
1170 		syscallarg(int) flags;
1171 		syscallarg(int) mode;
1172 	} */
1173 
1174 	return ENOSYS;
1175 }
1176 
1177 static void
1178 vfs__fhfree(fhandle_t *fhp)
1179 {
1180 	size_t fhsize;
1181 
1182 	if (fhp == NULL) {
1183 		return;
1184 	}
1185 	fhsize = FHANDLE_SIZE(fhp);
1186 	kmem_free(fhp, fhsize);
1187 }
1188 
1189 /*
1190  * vfs_composefh: compose a filehandle.
1191  */
1192 
1193 int
1194 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1195 {
1196 	struct mount *mp;
1197 	struct fid *fidp;
1198 	int error;
1199 	size_t needfhsize;
1200 	size_t fidsize;
1201 
1202 	mp = vp->v_mount;
1203 	fidp = NULL;
1204 	if (*fh_size < FHANDLE_SIZE_MIN) {
1205 		fidsize = 0;
1206 	} else {
1207 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1208 		if (fhp != NULL) {
1209 			memset(fhp, 0, *fh_size);
1210 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1211 			fidp = &fhp->fh_fid;
1212 		}
1213 	}
1214 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1215 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1216 	if (error == 0 && *fh_size < needfhsize) {
1217 		error = E2BIG;
1218 	}
1219 	*fh_size = needfhsize;
1220 	return error;
1221 }
1222 
1223 int
1224 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1225 {
1226 	struct mount *mp;
1227 	fhandle_t *fhp;
1228 	size_t fhsize;
1229 	size_t fidsize;
1230 	int error;
1231 
1232 	*fhpp = NULL;
1233 	mp = vp->v_mount;
1234 	fidsize = 0;
1235 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1236 	KASSERT(error != 0);
1237 	if (error != E2BIG) {
1238 		goto out;
1239 	}
1240 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1241 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1242 	if (fhp == NULL) {
1243 		error = ENOMEM;
1244 		goto out;
1245 	}
1246 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1247 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1248 	if (error == 0) {
1249 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1250 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1251 		*fhpp = fhp;
1252 	} else {
1253 		kmem_free(fhp, fhsize);
1254 	}
1255 out:
1256 	return error;
1257 }
1258 
1259 void
1260 vfs_composefh_free(fhandle_t *fhp)
1261 {
1262 
1263 	vfs__fhfree(fhp);
1264 }
1265 
1266 /*
1267  * vfs_fhtovp: lookup a vnode by a filehandle.
1268  */
1269 
1270 int
1271 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1272 {
1273 	struct mount *mp;
1274 	int error;
1275 
1276 	*vpp = NULL;
1277 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1278 	if (mp == NULL) {
1279 		error = ESTALE;
1280 		goto out;
1281 	}
1282 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1283 		error = EOPNOTSUPP;
1284 		goto out;
1285 	}
1286 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1287 out:
1288 	return error;
1289 }
1290 
1291 /*
1292  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1293  * the needed size.
1294  */
1295 
1296 int
1297 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1298 {
1299 	fhandle_t *fhp;
1300 	int error;
1301 
1302 	*fhpp = NULL;
1303 	if (fhsize > FHANDLE_SIZE_MAX) {
1304 		return EINVAL;
1305 	}
1306 	if (fhsize < FHANDLE_SIZE_MIN) {
1307 		return EINVAL;
1308 	}
1309 again:
1310 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1311 	if (fhp == NULL) {
1312 		return ENOMEM;
1313 	}
1314 	error = copyin(ufhp, fhp, fhsize);
1315 	if (error == 0) {
1316 		/* XXX this check shouldn't be here */
1317 		if (FHANDLE_SIZE(fhp) == fhsize) {
1318 			*fhpp = fhp;
1319 			return 0;
1320 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1321 			/*
1322 			 * a kludge for nfsv2 padded handles.
1323 			 */
1324 			size_t sz;
1325 
1326 			sz = FHANDLE_SIZE(fhp);
1327 			kmem_free(fhp, fhsize);
1328 			fhsize = sz;
1329 			goto again;
1330 		} else {
1331 			/*
1332 			 * userland told us wrong size.
1333 			 */
1334 		    	error = EINVAL;
1335 		}
1336 	}
1337 	kmem_free(fhp, fhsize);
1338 	return error;
1339 }
1340 
1341 void
1342 vfs_copyinfh_free(fhandle_t *fhp)
1343 {
1344 
1345 	vfs__fhfree(fhp);
1346 }
1347 
1348 /*
1349  * Get file handle system call
1350  */
1351 int
1352 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1353 {
1354 	/* {
1355 		syscallarg(char *) fname;
1356 		syscallarg(fhandle_t *) fhp;
1357 		syscallarg(size_t *) fh_size;
1358 	} */
1359 	struct vnode *vp;
1360 	fhandle_t *fh;
1361 	int error;
1362 	struct pathbuf *pb;
1363 	struct nameidata nd;
1364 	size_t sz;
1365 	size_t usz;
1366 
1367 	/*
1368 	 * Must be super user
1369 	 */
1370 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1371 	    0, NULL, NULL, NULL);
1372 	if (error)
1373 		return (error);
1374 
1375 	error = pathbuf_copyin(SCARG(uap, fname), &pb);
1376 	if (error) {
1377 		return error;
1378 	}
1379 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1380 	error = namei(&nd);
1381 	if (error) {
1382 		pathbuf_destroy(pb);
1383 		return error;
1384 	}
1385 	vp = nd.ni_vp;
1386 	pathbuf_destroy(pb);
1387 
1388 	error = vfs_composefh_alloc(vp, &fh);
1389 	vput(vp);
1390 	if (error != 0) {
1391 		goto out;
1392 	}
1393 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1394 	if (error != 0) {
1395 		goto out;
1396 	}
1397 	sz = FHANDLE_SIZE(fh);
1398 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1399 	if (error != 0) {
1400 		goto out;
1401 	}
1402 	if (usz >= sz) {
1403 		error = copyout(fh, SCARG(uap, fhp), sz);
1404 	} else {
1405 		error = E2BIG;
1406 	}
1407 out:
1408 	vfs_composefh_free(fh);
1409 	return (error);
1410 }
1411 
1412 /*
1413  * Open a file given a file handle.
1414  *
1415  * Check permissions, allocate an open file structure,
1416  * and call the device open routine if any.
1417  */
1418 
1419 int
1420 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1421     register_t *retval)
1422 {
1423 	file_t *fp;
1424 	struct vnode *vp = NULL;
1425 	kauth_cred_t cred = l->l_cred;
1426 	file_t *nfp;
1427 	int indx, error = 0;
1428 	struct vattr va;
1429 	fhandle_t *fh;
1430 	int flags;
1431 	proc_t *p;
1432 
1433 	p = curproc;
1434 
1435 	/*
1436 	 * Must be super user
1437 	 */
1438 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1439 	    0, NULL, NULL, NULL)))
1440 		return (error);
1441 
1442 	flags = FFLAGS(oflags);
1443 	if ((flags & (FREAD | FWRITE)) == 0)
1444 		return (EINVAL);
1445 	if ((flags & O_CREAT))
1446 		return (EINVAL);
1447 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1448 		return (error);
1449 	fp = nfp;
1450 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1451 	if (error != 0) {
1452 		goto bad;
1453 	}
1454 	error = vfs_fhtovp(fh, &vp);
1455 	if (error != 0) {
1456 		goto bad;
1457 	}
1458 
1459 	/* Now do an effective vn_open */
1460 
1461 	if (vp->v_type == VSOCK) {
1462 		error = EOPNOTSUPP;
1463 		goto bad;
1464 	}
1465 	error = vn_openchk(vp, cred, flags);
1466 	if (error != 0)
1467 		goto bad;
1468 	if (flags & O_TRUNC) {
1469 		VOP_UNLOCK(vp);			/* XXX */
1470 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1471 		vattr_null(&va);
1472 		va.va_size = 0;
1473 		error = VOP_SETATTR(vp, &va, cred);
1474 		if (error)
1475 			goto bad;
1476 	}
1477 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1478 		goto bad;
1479 	if (flags & FWRITE) {
1480 		mutex_enter(vp->v_interlock);
1481 		vp->v_writecount++;
1482 		mutex_exit(vp->v_interlock);
1483 	}
1484 
1485 	/* done with modified vn_open, now finish what sys_open does. */
1486 	if ((error = open_setfp(l, fp, vp, indx, flags)))
1487 		return error;
1488 
1489 	VOP_UNLOCK(vp);
1490 	*retval = indx;
1491 	fd_affix(p, fp, indx);
1492 	vfs_copyinfh_free(fh);
1493 	return (0);
1494 
1495 bad:
1496 	fd_abort(p, fp, indx);
1497 	if (vp != NULL)
1498 		vput(vp);
1499 	vfs_copyinfh_free(fh);
1500 	return (error);
1501 }
1502 
1503 int
1504 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1505 {
1506 	/* {
1507 		syscallarg(const void *) fhp;
1508 		syscallarg(size_t) fh_size;
1509 		syscallarg(int) flags;
1510 	} */
1511 
1512 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1513 	    SCARG(uap, flags), retval);
1514 }
1515 
1516 int
1517 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1518 {
1519 	int error;
1520 	fhandle_t *fh;
1521 	struct vnode *vp;
1522 
1523 	/*
1524 	 * Must be super user
1525 	 */
1526 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1527 	    0, NULL, NULL, NULL)))
1528 		return (error);
1529 
1530 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1531 	if (error != 0)
1532 		return error;
1533 
1534 	error = vfs_fhtovp(fh, &vp);
1535 	vfs_copyinfh_free(fh);
1536 	if (error != 0)
1537 		return error;
1538 
1539 	error = vn_stat(vp, sb);
1540 	vput(vp);
1541 	return error;
1542 }
1543 
1544 
1545 /* ARGSUSED */
1546 int
1547 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1548 {
1549 	/* {
1550 		syscallarg(const void *) fhp;
1551 		syscallarg(size_t) fh_size;
1552 		syscallarg(struct stat *) sb;
1553 	} */
1554 	struct stat sb;
1555 	int error;
1556 
1557 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1558 	if (error)
1559 		return error;
1560 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1561 }
1562 
1563 int
1564 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1565     int flags)
1566 {
1567 	fhandle_t *fh;
1568 	struct mount *mp;
1569 	struct vnode *vp;
1570 	int error;
1571 
1572 	/*
1573 	 * Must be super user
1574 	 */
1575 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1576 	    0, NULL, NULL, NULL)))
1577 		return error;
1578 
1579 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1580 	if (error != 0)
1581 		return error;
1582 
1583 	error = vfs_fhtovp(fh, &vp);
1584 	vfs_copyinfh_free(fh);
1585 	if (error != 0)
1586 		return error;
1587 
1588 	mp = vp->v_mount;
1589 	error = dostatvfs(mp, sb, l, flags, 1);
1590 	vput(vp);
1591 	return error;
1592 }
1593 
1594 /* ARGSUSED */
1595 int
1596 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1597 {
1598 	/* {
1599 		syscallarg(const void *) fhp;
1600 		syscallarg(size_t) fh_size;
1601 		syscallarg(struct statvfs *) buf;
1602 		syscallarg(int)	flags;
1603 	} */
1604 	struct statvfs *sb = STATVFSBUF_GET();
1605 	int error;
1606 
1607 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1608 	    SCARG(uap, flags));
1609 	if (error == 0)
1610 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1611 	STATVFSBUF_PUT(sb);
1612 	return error;
1613 }
1614 
1615 /*
1616  * Create a special file.
1617  */
1618 /* ARGSUSED */
1619 int
1620 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
1621     register_t *retval)
1622 {
1623 	/* {
1624 		syscallarg(const char *) path;
1625 		syscallarg(mode_t) mode;
1626 		syscallarg(dev_t) dev;
1627 	} */
1628 	return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
1629 	    SCARG(uap, dev), retval, UIO_USERSPACE);
1630 }
1631 
1632 int
1633 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap,
1634     register_t *retval)
1635 {
1636 	/* {
1637 		syscallarg(int) fd;
1638 		syscallarg(const char *) path;
1639 		syscallarg(mode_t) mode;
1640 		syscallarg(uint32_t) dev;
1641 	} */
1642 
1643 	return ENOSYS;
1644 }
1645 
1646 int
1647 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
1648     register_t *retval, enum uio_seg seg)
1649 {
1650 	struct proc *p = l->l_proc;
1651 	struct vnode *vp;
1652 	struct vattr vattr;
1653 	int error, optype;
1654 	struct pathbuf *pb;
1655 	struct nameidata nd;
1656 	const char *pathstring;
1657 
1658 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1659 	    0, NULL, NULL, NULL)) != 0)
1660 		return (error);
1661 
1662 	optype = VOP_MKNOD_DESCOFFSET;
1663 
1664 	error = pathbuf_maybe_copyin(pathname, seg, &pb);
1665 	if (error) {
1666 		return error;
1667 	}
1668 	pathstring = pathbuf_stringcopy_get(pb);
1669 	if (pathstring == NULL) {
1670 		pathbuf_destroy(pb);
1671 		return ENOMEM;
1672 	}
1673 
1674 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
1675 	if ((error = namei(&nd)) != 0)
1676 		goto out;
1677 	vp = nd.ni_vp;
1678 
1679 	if (vp != NULL)
1680 		error = EEXIST;
1681 	else {
1682 		vattr_null(&vattr);
1683 		/* We will read cwdi->cwdi_cmask unlocked. */
1684 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1685 		vattr.va_rdev = dev;
1686 
1687 		switch (mode & S_IFMT) {
1688 		case S_IFMT:	/* used by badsect to flag bad sectors */
1689 			vattr.va_type = VBAD;
1690 			break;
1691 		case S_IFCHR:
1692 			vattr.va_type = VCHR;
1693 			break;
1694 		case S_IFBLK:
1695 			vattr.va_type = VBLK;
1696 			break;
1697 		case S_IFWHT:
1698 			optype = VOP_WHITEOUT_DESCOFFSET;
1699 			break;
1700 		case S_IFREG:
1701 #if NVERIEXEC > 0
1702 			error = veriexec_openchk(l, nd.ni_vp, pathstring,
1703 			    O_CREAT);
1704 #endif /* NVERIEXEC > 0 */
1705 			vattr.va_type = VREG;
1706 			vattr.va_rdev = VNOVAL;
1707 			optype = VOP_CREATE_DESCOFFSET;
1708 			break;
1709 		default:
1710 			error = EINVAL;
1711 			break;
1712 		}
1713 	}
1714 	if (error == 0 && optype == VOP_MKNOD_DESCOFFSET
1715 	    && vattr.va_rdev == VNOVAL)
1716 		error = EINVAL;
1717 	if (!error) {
1718 		switch (optype) {
1719 		case VOP_WHITEOUT_DESCOFFSET:
1720 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1721 			if (error)
1722 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1723 			vput(nd.ni_dvp);
1724 			break;
1725 
1726 		case VOP_MKNOD_DESCOFFSET:
1727 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1728 						&nd.ni_cnd, &vattr);
1729 			if (error == 0)
1730 				vput(nd.ni_vp);
1731 			break;
1732 
1733 		case VOP_CREATE_DESCOFFSET:
1734 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1735 						&nd.ni_cnd, &vattr);
1736 			if (error == 0)
1737 				vput(nd.ni_vp);
1738 			break;
1739 		}
1740 	} else {
1741 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1742 		if (nd.ni_dvp == vp)
1743 			vrele(nd.ni_dvp);
1744 		else
1745 			vput(nd.ni_dvp);
1746 		if (vp)
1747 			vrele(vp);
1748 	}
1749 out:
1750 	pathbuf_stringcopy_put(pb, pathstring);
1751 	pathbuf_destroy(pb);
1752 	return (error);
1753 }
1754 
1755 /*
1756  * Create a named pipe.
1757  */
1758 /* ARGSUSED */
1759 int
1760 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1761 {
1762 	/* {
1763 		syscallarg(const char *) path;
1764 		syscallarg(int) mode;
1765 	} */
1766 	struct proc *p = l->l_proc;
1767 	struct vattr vattr;
1768 	int error;
1769 	struct pathbuf *pb;
1770 	struct nameidata nd;
1771 
1772 	error = pathbuf_copyin(SCARG(uap, path), &pb);
1773 	if (error) {
1774 		return error;
1775 	}
1776 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
1777 	if ((error = namei(&nd)) != 0) {
1778 		pathbuf_destroy(pb);
1779 		return error;
1780 	}
1781 	if (nd.ni_vp != NULL) {
1782 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1783 		if (nd.ni_dvp == nd.ni_vp)
1784 			vrele(nd.ni_dvp);
1785 		else
1786 			vput(nd.ni_dvp);
1787 		vrele(nd.ni_vp);
1788 		pathbuf_destroy(pb);
1789 		return (EEXIST);
1790 	}
1791 	vattr_null(&vattr);
1792 	vattr.va_type = VFIFO;
1793 	/* We will read cwdi->cwdi_cmask unlocked. */
1794 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1795 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1796 	if (error == 0)
1797 		vput(nd.ni_vp);
1798 	pathbuf_destroy(pb);
1799 	return (error);
1800 }
1801 
1802 int
1803 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap,
1804     register_t *retval)
1805 {
1806 	/* {
1807 		syscallarg(int) fd;
1808 		syscallarg(const char *) path;
1809 		syscallarg(int) mode;
1810 	} */
1811 
1812 	return ENOSYS;
1813 }
1814 /*
1815  * Make a hard file link.
1816  */
1817 /* ARGSUSED */
1818 static int
1819 do_sys_link(struct lwp *l, const char *path, const char *link,
1820 	    int follow, register_t *retval)
1821 {
1822 	struct vnode *vp;
1823 	struct pathbuf *linkpb;
1824 	struct nameidata nd;
1825 	namei_simple_flags_t namei_simple_flags;
1826 	int error;
1827 
1828 	if (follow)
1829 		namei_simple_flags = NSM_FOLLOW_TRYEMULROOT;
1830 	else
1831 		namei_simple_flags =  NSM_NOFOLLOW_TRYEMULROOT;
1832 
1833 	error = namei_simple_user(path, namei_simple_flags, &vp);
1834 	if (error != 0)
1835 		return (error);
1836 	error = pathbuf_copyin(link, &linkpb);
1837 	if (error) {
1838 		goto out1;
1839 	}
1840 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
1841 	if ((error = namei(&nd)) != 0)
1842 		goto out2;
1843 	if (nd.ni_vp) {
1844 		error = EEXIST;
1845 		goto abortop;
1846 	}
1847 	/* Prevent hard links on directories. */
1848 	if (vp->v_type == VDIR) {
1849 		error = EPERM;
1850 		goto abortop;
1851 	}
1852 	/* Prevent cross-mount operation. */
1853 	if (nd.ni_dvp->v_mount != vp->v_mount) {
1854 		error = EXDEV;
1855 		goto abortop;
1856 	}
1857 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1858 out2:
1859 	pathbuf_destroy(linkpb);
1860 out1:
1861 	vrele(vp);
1862 	return (error);
1863 abortop:
1864 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1865 	if (nd.ni_dvp == nd.ni_vp)
1866 		vrele(nd.ni_dvp);
1867 	else
1868 		vput(nd.ni_dvp);
1869 	if (nd.ni_vp != NULL)
1870 		vrele(nd.ni_vp);
1871 	goto out2;
1872 }
1873 
1874 int
1875 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
1876 {
1877 	/* {
1878 		syscallarg(const char *) path;
1879 		syscallarg(const char *) link;
1880 	} */
1881 	const char *path = SCARG(uap, path);
1882 	const char *link = SCARG(uap, link);
1883 
1884 	return do_sys_link(l, path, link, 1, retval);
1885 }
1886 
1887 int
1888 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap,
1889     register_t *retval)
1890 {
1891 	/* {
1892 		syscallarg(int) fd1;
1893 		syscallarg(const char *) name1;
1894 		syscallarg(int) fd2;
1895 		syscallarg(const char *) name2;
1896 		syscallarg(int) flags;
1897 	} */
1898 	const char *name1 = SCARG(uap, name1);
1899 	const char *name2 = SCARG(uap, name2);
1900 	int follow;
1901 
1902 	/*
1903 	 * Specified fd1 and fd2 are not yet implemented
1904 	 */
1905 	if ((SCARG(uap, fd1) != AT_FDCWD) || (SCARG(uap, fd2) != AT_FDCWD))
1906 		return ENOSYS;
1907 
1908 	follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW;
1909 
1910 	return do_sys_link(l, name1, name2, follow, retval);
1911 }
1912 
1913 
1914 int
1915 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg)
1916 {
1917 	struct proc *p = curproc;
1918 	struct vattr vattr;
1919 	char *path;
1920 	int error;
1921 	struct pathbuf *linkpb;
1922 	struct nameidata nd;
1923 
1924 	path = PNBUF_GET();
1925 	if (seg == UIO_USERSPACE) {
1926 		if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0)
1927 			goto out1;
1928 		if ((error = pathbuf_copyin(link, &linkpb)) != 0)
1929 			goto out1;
1930 	} else {
1931 		KASSERT(strlen(patharg) < MAXPATHLEN);
1932 		strcpy(path, patharg);
1933 		linkpb = pathbuf_create(link);
1934 		if (linkpb == NULL) {
1935 			error = ENOMEM;
1936 			goto out1;
1937 		}
1938 	}
1939 	ktrkuser("symlink-target", path, strlen(path));
1940 
1941 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
1942 	if ((error = namei(&nd)) != 0)
1943 		goto out2;
1944 	if (nd.ni_vp) {
1945 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1946 		if (nd.ni_dvp == nd.ni_vp)
1947 			vrele(nd.ni_dvp);
1948 		else
1949 			vput(nd.ni_dvp);
1950 		vrele(nd.ni_vp);
1951 		error = EEXIST;
1952 		goto out2;
1953 	}
1954 	vattr_null(&vattr);
1955 	vattr.va_type = VLNK;
1956 	/* We will read cwdi->cwdi_cmask unlocked. */
1957 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1958 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1959 	if (error == 0)
1960 		vput(nd.ni_vp);
1961 out2:
1962 	pathbuf_destroy(linkpb);
1963 out1:
1964 	PNBUF_PUT(path);
1965 	return (error);
1966 }
1967 
1968 /*
1969  * Make a symbolic link.
1970  */
1971 /* ARGSUSED */
1972 int
1973 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
1974 {
1975 	/* {
1976 		syscallarg(const char *) path;
1977 		syscallarg(const char *) link;
1978 	} */
1979 
1980 	return do_sys_symlink(SCARG(uap, path), SCARG(uap, link),
1981 	    UIO_USERSPACE);
1982 }
1983 
1984 int
1985 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap,
1986     register_t *retval)
1987 {
1988 	/* {
1989 		syscallarg(int) fd;
1990 		syscallarg(const char *) path;
1991 		syscallarg(const char *) link;
1992 	} */
1993 
1994 	return ENOSYS;
1995 }
1996 
1997 /*
1998  * Delete a whiteout from the filesystem.
1999  */
2000 /* ARGSUSED */
2001 int
2002 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2003 {
2004 	/* {
2005 		syscallarg(const char *) path;
2006 	} */
2007 	int error;
2008 	struct pathbuf *pb;
2009 	struct nameidata nd;
2010 
2011 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2012 	if (error) {
2013 		return error;
2014 	}
2015 
2016 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb);
2017 	error = namei(&nd);
2018 	if (error) {
2019 		pathbuf_destroy(pb);
2020 		return (error);
2021 	}
2022 
2023 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2024 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2025 		if (nd.ni_dvp == nd.ni_vp)
2026 			vrele(nd.ni_dvp);
2027 		else
2028 			vput(nd.ni_dvp);
2029 		if (nd.ni_vp)
2030 			vrele(nd.ni_vp);
2031 		pathbuf_destroy(pb);
2032 		return (EEXIST);
2033 	}
2034 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2035 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2036 	vput(nd.ni_dvp);
2037 	pathbuf_destroy(pb);
2038 	return (error);
2039 }
2040 
2041 /*
2042  * Delete a name from the filesystem.
2043  */
2044 /* ARGSUSED */
2045 int
2046 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2047 {
2048 	/* {
2049 		syscallarg(const char *) path;
2050 	} */
2051 
2052 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2053 }
2054 
2055 int
2056 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap,
2057     register_t *retval)
2058 {
2059 	/* {
2060 		syscallarg(int) fd;
2061 		syscallarg(const char *) path;
2062 	} */
2063 
2064 	return ENOSYS;
2065 }
2066 
2067 int
2068 do_sys_unlink(const char *arg, enum uio_seg seg)
2069 {
2070 	struct vnode *vp;
2071 	int error;
2072 	struct pathbuf *pb;
2073 	struct nameidata nd;
2074 	const char *pathstring;
2075 
2076 	error = pathbuf_maybe_copyin(arg, seg, &pb);
2077 	if (error) {
2078 		return error;
2079 	}
2080 	pathstring = pathbuf_stringcopy_get(pb);
2081 	if (pathstring == NULL) {
2082 		pathbuf_destroy(pb);
2083 		return ENOMEM;
2084 	}
2085 
2086 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
2087 	if ((error = namei(&nd)) != 0)
2088 		goto out;
2089 	vp = nd.ni_vp;
2090 
2091 	/*
2092 	 * The root of a mounted filesystem cannot be deleted.
2093 	 */
2094 	if (vp->v_vflag & VV_ROOT) {
2095 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2096 		if (nd.ni_dvp == vp)
2097 			vrele(nd.ni_dvp);
2098 		else
2099 			vput(nd.ni_dvp);
2100 		vput(vp);
2101 		error = EBUSY;
2102 		goto out;
2103 	}
2104 
2105 #if NVERIEXEC > 0
2106 	/* Handle remove requests for veriexec entries. */
2107 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) {
2108 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2109 		if (nd.ni_dvp == vp)
2110 			vrele(nd.ni_dvp);
2111 		else
2112 			vput(nd.ni_dvp);
2113 		vput(vp);
2114 		goto out;
2115 	}
2116 #endif /* NVERIEXEC > 0 */
2117 
2118 #ifdef FILEASSOC
2119 	(void)fileassoc_file_delete(vp);
2120 #endif /* FILEASSOC */
2121 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2122 out:
2123 	pathbuf_stringcopy_put(pb, pathstring);
2124 	pathbuf_destroy(pb);
2125 	return (error);
2126 }
2127 
2128 /*
2129  * Reposition read/write file offset.
2130  */
2131 int
2132 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2133 {
2134 	/* {
2135 		syscallarg(int) fd;
2136 		syscallarg(int) pad;
2137 		syscallarg(off_t) offset;
2138 		syscallarg(int) whence;
2139 	} */
2140 	kauth_cred_t cred = l->l_cred;
2141 	file_t *fp;
2142 	struct vnode *vp;
2143 	struct vattr vattr;
2144 	off_t newoff;
2145 	int error, fd;
2146 
2147 	fd = SCARG(uap, fd);
2148 
2149 	if ((fp = fd_getfile(fd)) == NULL)
2150 		return (EBADF);
2151 
2152 	vp = fp->f_data;
2153 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2154 		error = ESPIPE;
2155 		goto out;
2156 	}
2157 
2158 	switch (SCARG(uap, whence)) {
2159 	case SEEK_CUR:
2160 		newoff = fp->f_offset + SCARG(uap, offset);
2161 		break;
2162 	case SEEK_END:
2163 		vn_lock(vp, LK_SHARED | LK_RETRY);
2164 		error = VOP_GETATTR(vp, &vattr, cred);
2165 		VOP_UNLOCK(vp);
2166 		if (error) {
2167 			goto out;
2168 		}
2169 		newoff = SCARG(uap, offset) + vattr.va_size;
2170 		break;
2171 	case SEEK_SET:
2172 		newoff = SCARG(uap, offset);
2173 		break;
2174 	default:
2175 		error = EINVAL;
2176 		goto out;
2177 	}
2178 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2179 		*(off_t *)retval = fp->f_offset = newoff;
2180 	}
2181  out:
2182  	fd_putfile(fd);
2183 	return (error);
2184 }
2185 
2186 /*
2187  * Positional read system call.
2188  */
2189 int
2190 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2191 {
2192 	/* {
2193 		syscallarg(int) fd;
2194 		syscallarg(void *) buf;
2195 		syscallarg(size_t) nbyte;
2196 		syscallarg(off_t) offset;
2197 	} */
2198 	file_t *fp;
2199 	struct vnode *vp;
2200 	off_t offset;
2201 	int error, fd = SCARG(uap, fd);
2202 
2203 	if ((fp = fd_getfile(fd)) == NULL)
2204 		return (EBADF);
2205 
2206 	if ((fp->f_flag & FREAD) == 0) {
2207 		fd_putfile(fd);
2208 		return (EBADF);
2209 	}
2210 
2211 	vp = fp->f_data;
2212 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2213 		error = ESPIPE;
2214 		goto out;
2215 	}
2216 
2217 	offset = SCARG(uap, offset);
2218 
2219 	/*
2220 	 * XXX This works because no file systems actually
2221 	 * XXX take any action on the seek operation.
2222 	 */
2223 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2224 		goto out;
2225 
2226 	/* dofileread() will unuse the descriptor for us */
2227 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2228 	    &offset, 0, retval));
2229 
2230  out:
2231 	fd_putfile(fd);
2232 	return (error);
2233 }
2234 
2235 /*
2236  * Positional scatter read system call.
2237  */
2238 int
2239 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2240 {
2241 	/* {
2242 		syscallarg(int) fd;
2243 		syscallarg(const struct iovec *) iovp;
2244 		syscallarg(int) iovcnt;
2245 		syscallarg(off_t) offset;
2246 	} */
2247 	off_t offset = SCARG(uap, offset);
2248 
2249 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2250 	    SCARG(uap, iovcnt), &offset, 0, retval);
2251 }
2252 
2253 /*
2254  * Positional write system call.
2255  */
2256 int
2257 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2258 {
2259 	/* {
2260 		syscallarg(int) fd;
2261 		syscallarg(const void *) buf;
2262 		syscallarg(size_t) nbyte;
2263 		syscallarg(off_t) offset;
2264 	} */
2265 	file_t *fp;
2266 	struct vnode *vp;
2267 	off_t offset;
2268 	int error, fd = SCARG(uap, fd);
2269 
2270 	if ((fp = fd_getfile(fd)) == NULL)
2271 		return (EBADF);
2272 
2273 	if ((fp->f_flag & FWRITE) == 0) {
2274 		fd_putfile(fd);
2275 		return (EBADF);
2276 	}
2277 
2278 	vp = fp->f_data;
2279 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2280 		error = ESPIPE;
2281 		goto out;
2282 	}
2283 
2284 	offset = SCARG(uap, offset);
2285 
2286 	/*
2287 	 * XXX This works because no file systems actually
2288 	 * XXX take any action on the seek operation.
2289 	 */
2290 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2291 		goto out;
2292 
2293 	/* dofilewrite() will unuse the descriptor for us */
2294 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2295 	    &offset, 0, retval));
2296 
2297  out:
2298 	fd_putfile(fd);
2299 	return (error);
2300 }
2301 
2302 /*
2303  * Positional gather write system call.
2304  */
2305 int
2306 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2307 {
2308 	/* {
2309 		syscallarg(int) fd;
2310 		syscallarg(const struct iovec *) iovp;
2311 		syscallarg(int) iovcnt;
2312 		syscallarg(off_t) offset;
2313 	} */
2314 	off_t offset = SCARG(uap, offset);
2315 
2316 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2317 	    SCARG(uap, iovcnt), &offset, 0, retval);
2318 }
2319 
2320 /*
2321  * Check access permissions.
2322  */
2323 int
2324 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2325 {
2326 	/* {
2327 		syscallarg(const char *) path;
2328 		syscallarg(int) flags;
2329 	} */
2330 	kauth_cred_t cred;
2331 	struct vnode *vp;
2332 	int error, flags;
2333 	struct pathbuf *pb;
2334 	struct nameidata nd;
2335 
2336 	CTASSERT(F_OK == 0);
2337 	if ((SCARG(uap, flags) & ~(R_OK | W_OK | X_OK)) != 0) {
2338 		/* nonsense flags */
2339 		return EINVAL;
2340 	}
2341 
2342 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2343 	if (error) {
2344 		return error;
2345 	}
2346 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2347 
2348 	/* Override default credentials */
2349 	cred = kauth_cred_dup(l->l_cred);
2350 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2351 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2352 	nd.ni_cnd.cn_cred = cred;
2353 
2354 	if ((error = namei(&nd)) != 0) {
2355 		pathbuf_destroy(pb);
2356 		goto out;
2357 	}
2358 	vp = nd.ni_vp;
2359 	pathbuf_destroy(pb);
2360 
2361 	/* Flags == 0 means only check for existence. */
2362 	if (SCARG(uap, flags)) {
2363 		flags = 0;
2364 		if (SCARG(uap, flags) & R_OK)
2365 			flags |= VREAD;
2366 		if (SCARG(uap, flags) & W_OK)
2367 			flags |= VWRITE;
2368 		if (SCARG(uap, flags) & X_OK)
2369 			flags |= VEXEC;
2370 
2371 		error = VOP_ACCESS(vp, flags, cred);
2372 		if (!error && (flags & VWRITE))
2373 			error = vn_writechk(vp);
2374 	}
2375 	vput(vp);
2376 out:
2377 	kauth_cred_free(cred);
2378 	return (error);
2379 }
2380 
2381 int
2382 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap,
2383     register_t *retval)
2384 {
2385 	/* {
2386 		syscallarg(int) fd;
2387 		syscallarg(const char *) path;
2388 		syscallarg(int) amode;
2389 		syscallarg(int) flag;
2390 	} */
2391 
2392 	return ENOSYS;
2393 }
2394 
2395 /*
2396  * Common code for all sys_stat functions, including compat versions.
2397  */
2398 int
2399 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb)
2400 {
2401 	int error;
2402 	struct pathbuf *pb;
2403 	struct nameidata nd;
2404 
2405 	error = pathbuf_copyin(userpath, &pb);
2406 	if (error) {
2407 		return error;
2408 	}
2409 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb);
2410 	error = namei(&nd);
2411 	if (error != 0) {
2412 		pathbuf_destroy(pb);
2413 		return error;
2414 	}
2415 	error = vn_stat(nd.ni_vp, sb);
2416 	vput(nd.ni_vp);
2417 	pathbuf_destroy(pb);
2418 	return error;
2419 }
2420 
2421 /*
2422  * Get file status; this version follows links.
2423  */
2424 /* ARGSUSED */
2425 int
2426 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
2427 {
2428 	/* {
2429 		syscallarg(const char *) path;
2430 		syscallarg(struct stat *) ub;
2431 	} */
2432 	struct stat sb;
2433 	int error;
2434 
2435 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2436 	if (error)
2437 		return error;
2438 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2439 }
2440 
2441 /*
2442  * Get file status; this version does not follow links.
2443  */
2444 /* ARGSUSED */
2445 int
2446 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
2447 {
2448 	/* {
2449 		syscallarg(const char *) path;
2450 		syscallarg(struct stat *) ub;
2451 	} */
2452 	struct stat sb;
2453 	int error;
2454 
2455 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2456 	if (error)
2457 		return error;
2458 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2459 }
2460 
2461 int
2462 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap,
2463     register_t *retval)
2464 {
2465 	/* {
2466 		syscallarg(int) fd;
2467 		syscallarg(const char *) path;
2468 		syscallarg(struct stat *) ub;
2469 		syscallarg(int) flag;
2470 	} */
2471 
2472 	return ENOSYS;
2473 }
2474 /*
2475  * Get configurable pathname variables.
2476  */
2477 /* ARGSUSED */
2478 int
2479 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2480 {
2481 	/* {
2482 		syscallarg(const char *) path;
2483 		syscallarg(int) name;
2484 	} */
2485 	int error;
2486 	struct pathbuf *pb;
2487 	struct nameidata nd;
2488 
2489 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2490 	if (error) {
2491 		return error;
2492 	}
2493 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2494 	if ((error = namei(&nd)) != 0) {
2495 		pathbuf_destroy(pb);
2496 		return (error);
2497 	}
2498 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2499 	vput(nd.ni_vp);
2500 	pathbuf_destroy(pb);
2501 	return (error);
2502 }
2503 
2504 /*
2505  * Return target name of a symbolic link.
2506  */
2507 /* ARGSUSED */
2508 int
2509 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2510 {
2511 	/* {
2512 		syscallarg(const char *) path;
2513 		syscallarg(char *) buf;
2514 		syscallarg(size_t) count;
2515 	} */
2516 	struct vnode *vp;
2517 	struct iovec aiov;
2518 	struct uio auio;
2519 	int error;
2520 	struct pathbuf *pb;
2521 	struct nameidata nd;
2522 
2523 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2524 	if (error) {
2525 		return error;
2526 	}
2527 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2528 	if ((error = namei(&nd)) != 0) {
2529 		pathbuf_destroy(pb);
2530 		return error;
2531 	}
2532 	vp = nd.ni_vp;
2533 	pathbuf_destroy(pb);
2534 	if (vp->v_type != VLNK)
2535 		error = EINVAL;
2536 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2537 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2538 		aiov.iov_base = SCARG(uap, buf);
2539 		aiov.iov_len = SCARG(uap, count);
2540 		auio.uio_iov = &aiov;
2541 		auio.uio_iovcnt = 1;
2542 		auio.uio_offset = 0;
2543 		auio.uio_rw = UIO_READ;
2544 		KASSERT(l == curlwp);
2545 		auio.uio_vmspace = l->l_proc->p_vmspace;
2546 		auio.uio_resid = SCARG(uap, count);
2547 		error = VOP_READLINK(vp, &auio, l->l_cred);
2548 	}
2549 	vput(vp);
2550 	*retval = SCARG(uap, count) - auio.uio_resid;
2551 	return (error);
2552 }
2553 
2554 int
2555 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap,
2556     register_t *retval)
2557 {
2558 	/* {
2559 		syscallarg(int) fd;
2560 		syscallarg(const char *) path;
2561 		syscallarg(char *) buf;
2562 		syscallarg(size_t) count;
2563 	} */
2564 
2565 	return ENOSYS;
2566 }
2567 
2568 /*
2569  * Change flags of a file given a path name.
2570  */
2571 /* ARGSUSED */
2572 int
2573 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2574 {
2575 	/* {
2576 		syscallarg(const char *) path;
2577 		syscallarg(u_long) flags;
2578 	} */
2579 	struct vnode *vp;
2580 	int error;
2581 
2582 	error = namei_simple_user(SCARG(uap, path),
2583 				NSM_FOLLOW_TRYEMULROOT, &vp);
2584 	if (error != 0)
2585 		return (error);
2586 	error = change_flags(vp, SCARG(uap, flags), l);
2587 	vput(vp);
2588 	return (error);
2589 }
2590 
2591 /*
2592  * Change flags of a file given a file descriptor.
2593  */
2594 /* ARGSUSED */
2595 int
2596 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
2597 {
2598 	/* {
2599 		syscallarg(int) fd;
2600 		syscallarg(u_long) flags;
2601 	} */
2602 	struct vnode *vp;
2603 	file_t *fp;
2604 	int error;
2605 
2606 	/* fd_getvnode() will use the descriptor for us */
2607 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2608 		return (error);
2609 	vp = fp->f_data;
2610 	error = change_flags(vp, SCARG(uap, flags), l);
2611 	VOP_UNLOCK(vp);
2612 	fd_putfile(SCARG(uap, fd));
2613 	return (error);
2614 }
2615 
2616 /*
2617  * Change flags of a file given a path name; this version does
2618  * not follow links.
2619  */
2620 int
2621 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
2622 {
2623 	/* {
2624 		syscallarg(const char *) path;
2625 		syscallarg(u_long) flags;
2626 	} */
2627 	struct vnode *vp;
2628 	int error;
2629 
2630 	error = namei_simple_user(SCARG(uap, path),
2631 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2632 	if (error != 0)
2633 		return (error);
2634 	error = change_flags(vp, SCARG(uap, flags), l);
2635 	vput(vp);
2636 	return (error);
2637 }
2638 
2639 /*
2640  * Common routine to change flags of a file.
2641  */
2642 int
2643 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2644 {
2645 	struct vattr vattr;
2646 	int error;
2647 
2648 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2649 	/*
2650 	 * Non-superusers cannot change the flags on devices, even if they
2651 	 * own them.
2652 	 */
2653 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2654 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2655 			goto out;
2656 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2657 			error = EINVAL;
2658 			goto out;
2659 		}
2660 	}
2661 	vattr_null(&vattr);
2662 	vattr.va_flags = flags;
2663 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2664 out:
2665 	return (error);
2666 }
2667 
2668 /*
2669  * Change mode of a file given path name; this version follows links.
2670  */
2671 /* ARGSUSED */
2672 int
2673 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
2674 {
2675 	/* {
2676 		syscallarg(const char *) path;
2677 		syscallarg(int) mode;
2678 	} */
2679 	int error;
2680 	struct vnode *vp;
2681 
2682 	error = namei_simple_user(SCARG(uap, path),
2683 				NSM_FOLLOW_TRYEMULROOT, &vp);
2684 	if (error != 0)
2685 		return (error);
2686 
2687 	error = change_mode(vp, SCARG(uap, mode), l);
2688 
2689 	vrele(vp);
2690 	return (error);
2691 }
2692 
2693 /*
2694  * Change mode of a file given a file descriptor.
2695  */
2696 /* ARGSUSED */
2697 int
2698 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
2699 {
2700 	/* {
2701 		syscallarg(int) fd;
2702 		syscallarg(int) mode;
2703 	} */
2704 	file_t *fp;
2705 	int error;
2706 
2707 	/* fd_getvnode() will use the descriptor for us */
2708 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2709 		return (error);
2710 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
2711 	fd_putfile(SCARG(uap, fd));
2712 	return (error);
2713 }
2714 
2715 int
2716 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap,
2717     register_t *retval)
2718 {
2719 	/* {
2720 		syscallarg(int) fd;
2721 		syscallarg(const char *) path;
2722 		syscallarg(int) mode;
2723 		syscallarg(int) flag;
2724 	} */
2725 
2726 	return ENOSYS;
2727 }
2728 
2729 /*
2730  * Change mode of a file given path name; this version does not follow links.
2731  */
2732 /* ARGSUSED */
2733 int
2734 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
2735 {
2736 	/* {
2737 		syscallarg(const char *) path;
2738 		syscallarg(int) mode;
2739 	} */
2740 	int error;
2741 	struct vnode *vp;
2742 
2743 	error = namei_simple_user(SCARG(uap, path),
2744 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2745 	if (error != 0)
2746 		return (error);
2747 
2748 	error = change_mode(vp, SCARG(uap, mode), l);
2749 
2750 	vrele(vp);
2751 	return (error);
2752 }
2753 
2754 /*
2755  * Common routine to set mode given a vnode.
2756  */
2757 static int
2758 change_mode(struct vnode *vp, int mode, struct lwp *l)
2759 {
2760 	struct vattr vattr;
2761 	int error;
2762 
2763 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2764 	vattr_null(&vattr);
2765 	vattr.va_mode = mode & ALLPERMS;
2766 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2767 	VOP_UNLOCK(vp);
2768 	return (error);
2769 }
2770 
2771 /*
2772  * Set ownership given a path name; this version follows links.
2773  */
2774 /* ARGSUSED */
2775 int
2776 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
2777 {
2778 	/* {
2779 		syscallarg(const char *) path;
2780 		syscallarg(uid_t) uid;
2781 		syscallarg(gid_t) gid;
2782 	} */
2783 	int error;
2784 	struct vnode *vp;
2785 
2786 	error = namei_simple_user(SCARG(uap, path),
2787 				NSM_FOLLOW_TRYEMULROOT, &vp);
2788 	if (error != 0)
2789 		return (error);
2790 
2791 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2792 
2793 	vrele(vp);
2794 	return (error);
2795 }
2796 
2797 /*
2798  * Set ownership given a path name; this version follows links.
2799  * Provides POSIX semantics.
2800  */
2801 /* ARGSUSED */
2802 int
2803 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
2804 {
2805 	/* {
2806 		syscallarg(const char *) path;
2807 		syscallarg(uid_t) uid;
2808 		syscallarg(gid_t) gid;
2809 	} */
2810 	int error;
2811 	struct vnode *vp;
2812 
2813 	error = namei_simple_user(SCARG(uap, path),
2814 				NSM_FOLLOW_TRYEMULROOT, &vp);
2815 	if (error != 0)
2816 		return (error);
2817 
2818 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2819 
2820 	vrele(vp);
2821 	return (error);
2822 }
2823 
2824 /*
2825  * Set ownership given a file descriptor.
2826  */
2827 /* ARGSUSED */
2828 int
2829 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
2830 {
2831 	/* {
2832 		syscallarg(int) fd;
2833 		syscallarg(uid_t) uid;
2834 		syscallarg(gid_t) gid;
2835 	} */
2836 	int error;
2837 	file_t *fp;
2838 
2839 	/* fd_getvnode() will use the descriptor for us */
2840 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2841 		return (error);
2842 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2843 	    l, 0);
2844 	fd_putfile(SCARG(uap, fd));
2845 	return (error);
2846 }
2847 
2848 int
2849 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap,
2850     register_t *retval)
2851 {
2852 	/* {
2853 		syscallarg(int) fd;
2854 		syscallarg(const char *) path;
2855 		syscallarg(uid_t) uid;
2856 		syscallarg(gid_t) gid;
2857 		syscallarg(int) flag;
2858 	} */
2859 
2860 	return ENOSYS;
2861 }
2862 
2863 /*
2864  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2865  */
2866 /* ARGSUSED */
2867 int
2868 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
2869 {
2870 	/* {
2871 		syscallarg(int) fd;
2872 		syscallarg(uid_t) uid;
2873 		syscallarg(gid_t) gid;
2874 	} */
2875 	int error;
2876 	file_t *fp;
2877 
2878 	/* fd_getvnode() will use the descriptor for us */
2879 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2880 		return (error);
2881 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2882 	    l, 1);
2883 	fd_putfile(SCARG(uap, fd));
2884 	return (error);
2885 }
2886 
2887 /*
2888  * Set ownership given a path name; this version does not follow links.
2889  */
2890 /* ARGSUSED */
2891 int
2892 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
2893 {
2894 	/* {
2895 		syscallarg(const char *) path;
2896 		syscallarg(uid_t) uid;
2897 		syscallarg(gid_t) gid;
2898 	} */
2899 	int error;
2900 	struct vnode *vp;
2901 
2902 	error = namei_simple_user(SCARG(uap, path),
2903 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2904 	if (error != 0)
2905 		return (error);
2906 
2907 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2908 
2909 	vrele(vp);
2910 	return (error);
2911 }
2912 
2913 /*
2914  * Set ownership given a path name; this version does not follow links.
2915  * Provides POSIX/XPG semantics.
2916  */
2917 /* ARGSUSED */
2918 int
2919 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
2920 {
2921 	/* {
2922 		syscallarg(const char *) path;
2923 		syscallarg(uid_t) uid;
2924 		syscallarg(gid_t) gid;
2925 	} */
2926 	int error;
2927 	struct vnode *vp;
2928 
2929 	error = namei_simple_user(SCARG(uap, path),
2930 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2931 	if (error != 0)
2932 		return (error);
2933 
2934 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2935 
2936 	vrele(vp);
2937 	return (error);
2938 }
2939 
2940 /*
2941  * Common routine to set ownership given a vnode.
2942  */
2943 static int
2944 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2945     int posix_semantics)
2946 {
2947 	struct vattr vattr;
2948 	mode_t newmode;
2949 	int error;
2950 
2951 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2952 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2953 		goto out;
2954 
2955 #define CHANGED(x) ((int)(x) != -1)
2956 	newmode = vattr.va_mode;
2957 	if (posix_semantics) {
2958 		/*
2959 		 * POSIX/XPG semantics: if the caller is not the super-user,
2960 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2961 		 * the XPG consider the behaviour for calls by the super-user
2962 		 * implementation-defined; we leave the set-user-id and set-
2963 		 * group-id settings intact in that case.
2964 		 */
2965 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2966 				      NULL) != 0)
2967 			newmode &= ~(S_ISUID | S_ISGID);
2968 	} else {
2969 		/*
2970 		 * NetBSD semantics: when changing owner and/or group,
2971 		 * clear the respective bit(s).
2972 		 */
2973 		if (CHANGED(uid))
2974 			newmode &= ~S_ISUID;
2975 		if (CHANGED(gid))
2976 			newmode &= ~S_ISGID;
2977 	}
2978 	/* Update va_mode iff altered. */
2979 	if (vattr.va_mode == newmode)
2980 		newmode = VNOVAL;
2981 
2982 	vattr_null(&vattr);
2983 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2984 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2985 	vattr.va_mode = newmode;
2986 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2987 #undef CHANGED
2988 
2989 out:
2990 	VOP_UNLOCK(vp);
2991 	return (error);
2992 }
2993 
2994 /*
2995  * Set the access and modification times given a path name; this
2996  * version follows links.
2997  */
2998 /* ARGSUSED */
2999 int
3000 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
3001     register_t *retval)
3002 {
3003 	/* {
3004 		syscallarg(const char *) path;
3005 		syscallarg(const struct timeval *) tptr;
3006 	} */
3007 
3008 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
3009 	    SCARG(uap, tptr), UIO_USERSPACE);
3010 }
3011 
3012 /*
3013  * Set the access and modification times given a file descriptor.
3014  */
3015 /* ARGSUSED */
3016 int
3017 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
3018     register_t *retval)
3019 {
3020 	/* {
3021 		syscallarg(int) fd;
3022 		syscallarg(const struct timeval *) tptr;
3023 	} */
3024 	int error;
3025 	file_t *fp;
3026 
3027 	/* fd_getvnode() will use the descriptor for us */
3028 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3029 		return (error);
3030 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
3031 	    UIO_USERSPACE);
3032 	fd_putfile(SCARG(uap, fd));
3033 	return (error);
3034 }
3035 
3036 int
3037 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap,
3038     register_t *retval)
3039 {
3040 	/* {
3041 		syscallarg(int) fd;
3042 		syscallarg(const struct timespec *) tptr;
3043 	} */
3044 	int error;
3045 	file_t *fp;
3046 
3047 	/* fd_getvnode() will use the descriptor for us */
3048 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3049 		return (error);
3050 	error = do_sys_utimens(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
3051 	    UIO_USERSPACE);
3052 	fd_putfile(SCARG(uap, fd));
3053 	return (error);
3054 }
3055 
3056 /*
3057  * Set the access and modification times given a path name; this
3058  * version does not follow links.
3059  */
3060 int
3061 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
3062     register_t *retval)
3063 {
3064 	/* {
3065 		syscallarg(const char *) path;
3066 		syscallarg(const struct timeval *) tptr;
3067 	} */
3068 
3069 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
3070 	    SCARG(uap, tptr), UIO_USERSPACE);
3071 }
3072 
3073 int
3074 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap,
3075     register_t *retval)
3076 {
3077 	/* {
3078 		syscallarg(int) fd;
3079 		syscallarg(const char *) path;
3080 		syscallarg(const struct timespec *) tptr;
3081 		syscallarg(int) flag;
3082 	} */
3083 	int follow;
3084 	const struct timespec *tptr;
3085 
3086 	/*
3087 	 * Specified fd is not yet implemented
3088 	 */
3089 	if (SCARG(uap, fd) != AT_FDCWD)
3090 		return ENOSYS;
3091 
3092 	tptr = SCARG(uap, tptr);
3093 	follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3094 
3095 	return do_sys_utimens(l, NULL, SCARG(uap, path), follow,
3096 	    tptr, UIO_USERSPACE);
3097 }
3098 
3099 /*
3100  * Common routine to set access and modification times given a vnode.
3101  */
3102 int
3103 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag,
3104     const struct timespec *tptr, enum uio_seg seg)
3105 {
3106 	struct vattr vattr;
3107 	int error, dorele = 0;
3108 	namei_simple_flags_t sflags;
3109 
3110 	bool vanull, setbirthtime;
3111 	struct timespec ts[2];
3112 
3113 	/*
3114 	 * I have checked all callers and they pass either FOLLOW,
3115 	 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
3116 	 * is 0. More to the point, they don't pass anything else.
3117 	 * Let's keep it that way at least until the namei interfaces
3118 	 * are fully sanitized.
3119 	 */
3120 	KASSERT(flag == NOFOLLOW || flag == FOLLOW);
3121 	sflags = (flag == FOLLOW) ?
3122 		NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3123 
3124 	if (tptr == NULL) {
3125 		vanull = true;
3126 		nanotime(&ts[0]);
3127 		ts[1] = ts[0];
3128 	} else {
3129 		vanull = false;
3130 		if (seg != UIO_SYSSPACE) {
3131 			error = copyin(tptr, ts, sizeof (ts));
3132 			if (error != 0)
3133 				return error;
3134 		} else {
3135 			ts[0] = tptr[0];
3136 			ts[1] = tptr[1];
3137 		}
3138 	}
3139 
3140 	if (ts[0].tv_nsec == UTIME_NOW) {
3141 		nanotime(&ts[0]);
3142 		if (ts[1].tv_nsec == UTIME_NOW) {
3143 			vanull = true;
3144 			ts[1] = ts[0];
3145 		}
3146 	} else if (ts[1].tv_nsec == UTIME_NOW)
3147 		nanotime(&ts[1]);
3148 
3149 	if (vp == NULL) {
3150 		/* note: SEG describes TPTR, not PATH; PATH is always user */
3151 		error = namei_simple_user(path, sflags, &vp);
3152 		if (error != 0)
3153 			return error;
3154 		dorele = 1;
3155 	}
3156 
3157 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3158 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3159 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
3160 	vattr_null(&vattr);
3161 
3162 	if (ts[0].tv_nsec != UTIME_OMIT)
3163 		vattr.va_atime = ts[0];
3164 
3165 	if (ts[1].tv_nsec != UTIME_OMIT) {
3166 		vattr.va_mtime = ts[1];
3167 		if (setbirthtime)
3168 			vattr.va_birthtime = ts[1];
3169 	}
3170 
3171 	if (vanull)
3172 		vattr.va_vaflags |= VA_UTIMES_NULL;
3173 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3174 	VOP_UNLOCK(vp);
3175 
3176 	if (dorele != 0)
3177 		vrele(vp);
3178 
3179 	return error;
3180 }
3181 
3182 int
3183 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3184     const struct timeval *tptr, enum uio_seg seg)
3185 {
3186 	struct timespec ts[2];
3187 	struct timespec *tsptr = NULL;
3188 	int error;
3189 
3190 	if (tptr != NULL) {
3191 		struct timeval tv[2];
3192 
3193 		if (seg != UIO_SYSSPACE) {
3194 			error = copyin(tptr, tv, sizeof (tv));
3195 			if (error != 0)
3196 				return error;
3197 			tptr = tv;
3198 		}
3199 
3200 		if ((tv[0].tv_usec == UTIME_NOW) ||
3201 		    (tv[0].tv_usec == UTIME_OMIT))
3202 			ts[0].tv_nsec = tv[0].tv_usec;
3203 		else
3204 			TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3205 
3206 		if ((tv[1].tv_usec == UTIME_NOW) ||
3207 		    (tv[1].tv_usec == UTIME_OMIT))
3208 			ts[1].tv_nsec = tv[1].tv_usec;
3209 		else
3210 			TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3211 
3212 		tsptr = &ts[0];
3213 	}
3214 
3215 	return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE);
3216 }
3217 
3218 /*
3219  * Truncate a file given its path name.
3220  */
3221 /* ARGSUSED */
3222 int
3223 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3224 {
3225 	/* {
3226 		syscallarg(const char *) path;
3227 		syscallarg(int) pad;
3228 		syscallarg(off_t) length;
3229 	} */
3230 	struct vnode *vp;
3231 	struct vattr vattr;
3232 	int error;
3233 
3234 	error = namei_simple_user(SCARG(uap, path),
3235 				NSM_FOLLOW_TRYEMULROOT, &vp);
3236 	if (error != 0)
3237 		return (error);
3238 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3239 	if (vp->v_type == VDIR)
3240 		error = EISDIR;
3241 	else if ((error = vn_writechk(vp)) == 0 &&
3242 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3243 		vattr_null(&vattr);
3244 		vattr.va_size = SCARG(uap, length);
3245 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3246 	}
3247 	vput(vp);
3248 	return (error);
3249 }
3250 
3251 /*
3252  * Truncate a file given a file descriptor.
3253  */
3254 /* ARGSUSED */
3255 int
3256 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3257 {
3258 	/* {
3259 		syscallarg(int) fd;
3260 		syscallarg(int) pad;
3261 		syscallarg(off_t) length;
3262 	} */
3263 	struct vattr vattr;
3264 	struct vnode *vp;
3265 	file_t *fp;
3266 	int error;
3267 
3268 	/* fd_getvnode() will use the descriptor for us */
3269 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3270 		return (error);
3271 	if ((fp->f_flag & FWRITE) == 0) {
3272 		error = EINVAL;
3273 		goto out;
3274 	}
3275 	vp = fp->f_data;
3276 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3277 	if (vp->v_type == VDIR)
3278 		error = EISDIR;
3279 	else if ((error = vn_writechk(vp)) == 0) {
3280 		vattr_null(&vattr);
3281 		vattr.va_size = SCARG(uap, length);
3282 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3283 	}
3284 	VOP_UNLOCK(vp);
3285  out:
3286 	fd_putfile(SCARG(uap, fd));
3287 	return (error);
3288 }
3289 
3290 /*
3291  * Sync an open file.
3292  */
3293 /* ARGSUSED */
3294 int
3295 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3296 {
3297 	/* {
3298 		syscallarg(int) fd;
3299 	} */
3300 	struct vnode *vp;
3301 	file_t *fp;
3302 	int error;
3303 
3304 	/* fd_getvnode() will use the descriptor for us */
3305 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3306 		return (error);
3307 	vp = fp->f_data;
3308 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3309 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3310 	VOP_UNLOCK(vp);
3311 	fd_putfile(SCARG(uap, fd));
3312 	return (error);
3313 }
3314 
3315 /*
3316  * Sync a range of file data.  API modeled after that found in AIX.
3317  *
3318  * FDATASYNC indicates that we need only save enough metadata to be able
3319  * to re-read the written data.  Note we duplicate AIX's requirement that
3320  * the file be open for writing.
3321  */
3322 /* ARGSUSED */
3323 int
3324 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3325 {
3326 	/* {
3327 		syscallarg(int) fd;
3328 		syscallarg(int) flags;
3329 		syscallarg(off_t) start;
3330 		syscallarg(off_t) length;
3331 	} */
3332 	struct vnode *vp;
3333 	file_t *fp;
3334 	int flags, nflags;
3335 	off_t s, e, len;
3336 	int error;
3337 
3338 	/* fd_getvnode() will use the descriptor for us */
3339 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3340 		return (error);
3341 
3342 	if ((fp->f_flag & FWRITE) == 0) {
3343 		error = EBADF;
3344 		goto out;
3345 	}
3346 
3347 	flags = SCARG(uap, flags);
3348 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3349 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3350 		error = EINVAL;
3351 		goto out;
3352 	}
3353 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3354 	if (flags & FDATASYNC)
3355 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3356 	else
3357 		nflags = FSYNC_WAIT;
3358 	if (flags & FDISKSYNC)
3359 		nflags |= FSYNC_CACHE;
3360 
3361 	len = SCARG(uap, length);
3362 	/* If length == 0, we do the whole file, and s = e = 0 will do that */
3363 	if (len) {
3364 		s = SCARG(uap, start);
3365 		e = s + len;
3366 		if (e < s) {
3367 			error = EINVAL;
3368 			goto out;
3369 		}
3370 	} else {
3371 		e = 0;
3372 		s = 0;
3373 	}
3374 
3375 	vp = fp->f_data;
3376 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3377 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3378 	VOP_UNLOCK(vp);
3379 out:
3380 	fd_putfile(SCARG(uap, fd));
3381 	return (error);
3382 }
3383 
3384 /*
3385  * Sync the data of an open file.
3386  */
3387 /* ARGSUSED */
3388 int
3389 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3390 {
3391 	/* {
3392 		syscallarg(int) fd;
3393 	} */
3394 	struct vnode *vp;
3395 	file_t *fp;
3396 	int error;
3397 
3398 	/* fd_getvnode() will use the descriptor for us */
3399 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3400 		return (error);
3401 	if ((fp->f_flag & FWRITE) == 0) {
3402 		fd_putfile(SCARG(uap, fd));
3403 		return (EBADF);
3404 	}
3405 	vp = fp->f_data;
3406 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3407 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3408 	VOP_UNLOCK(vp);
3409 	fd_putfile(SCARG(uap, fd));
3410 	return (error);
3411 }
3412 
3413 /*
3414  * Rename files, (standard) BSD semantics frontend.
3415  */
3416 /* ARGSUSED */
3417 int
3418 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3419 {
3420 	/* {
3421 		syscallarg(const char *) from;
3422 		syscallarg(const char *) to;
3423 	} */
3424 
3425 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3426 }
3427 
3428 int
3429 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap,
3430     register_t *retval)
3431 {
3432 	/* {
3433 		syscallarg(int) fromfd;
3434 		syscallarg(const char *) from;
3435 		syscallarg(int) tofd;
3436 		syscallarg(const char *) to;
3437 	} */
3438 
3439 	return ENOSYS;
3440 }
3441 
3442 /*
3443  * Rename files, POSIX semantics frontend.
3444  */
3445 /* ARGSUSED */
3446 int
3447 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3448 {
3449 	/* {
3450 		syscallarg(const char *) from;
3451 		syscallarg(const char *) to;
3452 	} */
3453 
3454 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3455 }
3456 
3457 /*
3458  * Rename files.  Source and destination must either both be directories,
3459  * or both not be directories.  If target is a directory, it must be empty.
3460  * If `from' and `to' refer to the same object, the value of the `retain'
3461  * argument is used to determine whether `from' will be
3462  *
3463  * (retain == 0)	deleted unless `from' and `to' refer to the same
3464  *			object in the file system's name space (BSD).
3465  * (retain == 1)	always retained (POSIX).
3466  */
3467 int
3468 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3469 {
3470 	struct vnode *tvp, *fvp, *tdvp;
3471 	struct pathbuf *frompb, *topb;
3472 	struct nameidata fromnd, tond;
3473 	struct mount *fs;
3474 	int error;
3475 
3476 	error = pathbuf_maybe_copyin(from, seg, &frompb);
3477 	if (error) {
3478 		return error;
3479 	}
3480 	error = pathbuf_maybe_copyin(to, seg, &topb);
3481 	if (error) {
3482 		pathbuf_destroy(frompb);
3483 		return error;
3484 	}
3485 
3486 	NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME,
3487 	    frompb);
3488 	if ((error = namei(&fromnd)) != 0) {
3489 		pathbuf_destroy(frompb);
3490 		pathbuf_destroy(topb);
3491 		return (error);
3492 	}
3493 	if (fromnd.ni_dvp != fromnd.ni_vp)
3494 		VOP_UNLOCK(fromnd.ni_dvp);
3495 	fvp = fromnd.ni_vp;
3496 
3497 	fs = fvp->v_mount;
3498 	error = VFS_RENAMELOCK_ENTER(fs);
3499 	if (error) {
3500 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3501 		vrele(fromnd.ni_dvp);
3502 		vrele(fvp);
3503 		goto out1;
3504 	}
3505 
3506 	/*
3507 	 * close, partially, yet another race - ideally we should only
3508 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3509 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3510 	 * that with namei as it stands.
3511 	 *
3512 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3513 	 * under us. The real fix is to get the locks in the right
3514 	 * order and do the lookups in the right places, but that's a
3515 	 * major rototill.
3516 	 *
3517 	 * Note: this logic (as well as this whole function) is cloned
3518 	 * in nfs_serv.c. Proceed accordingly.
3519 	 */
3520 	vrele(fvp);
3521 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3522 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3523 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3524 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3525 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3526 		error = EINVAL;
3527 		VFS_RENAMELOCK_EXIT(fs);
3528 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3529 		vrele(fromnd.ni_dvp);
3530 		goto out1;
3531 	}
3532 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3533 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0);
3534 	if (error) {
3535 		VOP_UNLOCK(fromnd.ni_dvp);
3536 		VFS_RENAMELOCK_EXIT(fs);
3537 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3538 		vrele(fromnd.ni_dvp);
3539 		goto out1;
3540 	}
3541 	VOP_UNLOCK(fromnd.ni_vp);
3542 	if (fromnd.ni_dvp != fromnd.ni_vp)
3543 		VOP_UNLOCK(fromnd.ni_dvp);
3544 	fvp = fromnd.ni_vp;
3545 
3546 	NDINIT(&tond, RENAME,
3547 	    LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT
3548 	      | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0),
3549 	    topb);
3550 	if ((error = namei(&tond)) != 0) {
3551 		VFS_RENAMELOCK_EXIT(fs);
3552 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3553 		vrele(fromnd.ni_dvp);
3554 		vrele(fvp);
3555 		goto out1;
3556 	}
3557 	tdvp = tond.ni_dvp;
3558 	tvp = tond.ni_vp;
3559 
3560 	if (tvp != NULL) {
3561 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3562 			error = ENOTDIR;
3563 			goto out;
3564 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3565 			error = EISDIR;
3566 			goto out;
3567 		}
3568 	}
3569 
3570 	if (fvp == tdvp)
3571 		error = EINVAL;
3572 
3573 	/*
3574 	 * Source and destination refer to the same object.
3575 	 */
3576 	if (fvp == tvp) {
3577 		if (retain)
3578 			error = -1;
3579 		else if (fromnd.ni_dvp == tdvp &&
3580 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3581 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3582 		          tond.ni_cnd.cn_nameptr,
3583 		          fromnd.ni_cnd.cn_namelen))
3584 		error = -1;
3585 	}
3586 	/*
3587 	 * Prevent cross-mount operation.
3588 	 */
3589 	if (error == 0) {
3590 		if (tond.ni_dvp->v_mount != fromnd.ni_dvp->v_mount) {
3591 			error = EXDEV;
3592 		}
3593 	}
3594 #if NVERIEXEC > 0
3595 	if (!error) {
3596 		char *f1, *f2;
3597 		size_t f1_len;
3598 		size_t f2_len;
3599 
3600 		f1_len = fromnd.ni_cnd.cn_namelen + 1;
3601 		f1 = kmem_alloc(f1_len, KM_SLEEP);
3602 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
3603 
3604 		f2_len = tond.ni_cnd.cn_namelen + 1;
3605 		f2 = kmem_alloc(f2_len, KM_SLEEP);
3606 		strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
3607 
3608 		error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2);
3609 
3610 		kmem_free(f1, f1_len);
3611 		kmem_free(f2, f2_len);
3612 	}
3613 #endif /* NVERIEXEC > 0 */
3614 
3615 out:
3616 	if (!error) {
3617 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3618 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3619 		VFS_RENAMELOCK_EXIT(fs);
3620 	} else {
3621 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3622 		if (tdvp == tvp)
3623 			vrele(tdvp);
3624 		else
3625 			vput(tdvp);
3626 		if (tvp)
3627 			vput(tvp);
3628 		VFS_RENAMELOCK_EXIT(fs);
3629 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3630 		vrele(fromnd.ni_dvp);
3631 		vrele(fvp);
3632 	}
3633 out1:
3634 	pathbuf_destroy(frompb);
3635 	pathbuf_destroy(topb);
3636 	return (error == -1 ? 0 : error);
3637 }
3638 
3639 /*
3640  * Make a directory file.
3641  */
3642 /* ARGSUSED */
3643 int
3644 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
3645 {
3646 	/* {
3647 		syscallarg(const char *) path;
3648 		syscallarg(int) mode;
3649 	} */
3650 
3651 	return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE);
3652 }
3653 
3654 int
3655 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap,
3656     register_t *retval)
3657 {
3658 	/* {
3659 		syscallarg(int) fd;
3660 		syscallarg(const char *) path;
3661 		syscallarg(int) mode;
3662 	} */
3663 
3664 	return ENOSYS;
3665 }
3666 
3667 
3668 int
3669 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
3670 {
3671 	struct proc *p = curlwp->l_proc;
3672 	struct vnode *vp;
3673 	struct vattr vattr;
3674 	int error;
3675 	struct pathbuf *pb;
3676 	struct nameidata nd;
3677 
3678 	/* XXX bollocks, should pass in a pathbuf */
3679 	error = pathbuf_maybe_copyin(path, seg, &pb);
3680 	if (error) {
3681 		return error;
3682 	}
3683 
3684 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb);
3685 	if ((error = namei(&nd)) != 0) {
3686 		pathbuf_destroy(pb);
3687 		return (error);
3688 	}
3689 	vp = nd.ni_vp;
3690 	if (vp != NULL) {
3691 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3692 		if (nd.ni_dvp == vp)
3693 			vrele(nd.ni_dvp);
3694 		else
3695 			vput(nd.ni_dvp);
3696 		vrele(vp);
3697 		pathbuf_destroy(pb);
3698 		return (EEXIST);
3699 	}
3700 	vattr_null(&vattr);
3701 	vattr.va_type = VDIR;
3702 	/* We will read cwdi->cwdi_cmask unlocked. */
3703 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3704 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3705 	if (!error)
3706 		vput(nd.ni_vp);
3707 	pathbuf_destroy(pb);
3708 	return (error);
3709 }
3710 
3711 /*
3712  * Remove a directory file.
3713  */
3714 /* ARGSUSED */
3715 int
3716 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
3717 {
3718 	/* {
3719 		syscallarg(const char *) path;
3720 	} */
3721 	struct vnode *vp;
3722 	int error;
3723 	struct pathbuf *pb;
3724 	struct nameidata nd;
3725 
3726 	error = pathbuf_copyin(SCARG(uap, path), &pb);
3727 	if (error) {
3728 		return error;
3729 	}
3730 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
3731 	if ((error = namei(&nd)) != 0) {
3732 		pathbuf_destroy(pb);
3733 		return error;
3734 	}
3735 	vp = nd.ni_vp;
3736 	if (vp->v_type != VDIR) {
3737 		error = ENOTDIR;
3738 		goto out;
3739 	}
3740 	/*
3741 	 * No rmdir "." please.
3742 	 */
3743 	if (nd.ni_dvp == vp) {
3744 		error = EINVAL;
3745 		goto out;
3746 	}
3747 	/*
3748 	 * The root of a mounted filesystem cannot be deleted.
3749 	 */
3750 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
3751 		error = EBUSY;
3752 		goto out;
3753 	}
3754 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3755 	pathbuf_destroy(pb);
3756 	return (error);
3757 
3758 out:
3759 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3760 	if (nd.ni_dvp == vp)
3761 		vrele(nd.ni_dvp);
3762 	else
3763 		vput(nd.ni_dvp);
3764 	vput(vp);
3765 	pathbuf_destroy(pb);
3766 	return (error);
3767 }
3768 
3769 /*
3770  * Read a block of directory entries in a file system independent format.
3771  */
3772 int
3773 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
3774 {
3775 	/* {
3776 		syscallarg(int) fd;
3777 		syscallarg(char *) buf;
3778 		syscallarg(size_t) count;
3779 	} */
3780 	file_t *fp;
3781 	int error, done;
3782 
3783 	/* fd_getvnode() will use the descriptor for us */
3784 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3785 		return (error);
3786 	if ((fp->f_flag & FREAD) == 0) {
3787 		error = EBADF;
3788 		goto out;
3789 	}
3790 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3791 			SCARG(uap, count), &done, l, 0, 0);
3792 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3793 	*retval = done;
3794  out:
3795 	fd_putfile(SCARG(uap, fd));
3796 	return (error);
3797 }
3798 
3799 /*
3800  * Set the mode mask for creation of filesystem nodes.
3801  */
3802 int
3803 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
3804 {
3805 	/* {
3806 		syscallarg(mode_t) newmask;
3807 	} */
3808 	struct proc *p = l->l_proc;
3809 	struct cwdinfo *cwdi;
3810 
3811 	/*
3812 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3813 	 * important is that we serialize changes to the mask.  The
3814 	 * rw_exit() will issue a write memory barrier on our behalf,
3815 	 * and force the changes out to other CPUs (as it must use an
3816 	 * atomic operation, draining the local CPU's store buffers).
3817 	 */
3818 	cwdi = p->p_cwdi;
3819 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3820 	*retval = cwdi->cwdi_cmask;
3821 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3822 	rw_exit(&cwdi->cwdi_lock);
3823 
3824 	return (0);
3825 }
3826 
3827 int
3828 dorevoke(struct vnode *vp, kauth_cred_t cred)
3829 {
3830 	struct vattr vattr;
3831 	int error;
3832 
3833 	vn_lock(vp, LK_SHARED | LK_RETRY);
3834 	error = VOP_GETATTR(vp, &vattr, cred);
3835 	VOP_UNLOCK(vp);
3836 	if (error != 0)
3837 		return error;
3838 	if (kauth_cred_geteuid(cred) == vattr.va_uid ||
3839 	    (error = kauth_authorize_generic(cred,
3840 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
3841 		VOP_REVOKE(vp, REVOKEALL);
3842 	return (error);
3843 }
3844 
3845 /*
3846  * Void all references to file by ripping underlying filesystem
3847  * away from vnode.
3848  */
3849 /* ARGSUSED */
3850 int
3851 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
3852 {
3853 	/* {
3854 		syscallarg(const char *) path;
3855 	} */
3856 	struct vnode *vp;
3857 	int error;
3858 
3859 	error = namei_simple_user(SCARG(uap, path),
3860 				NSM_FOLLOW_TRYEMULROOT, &vp);
3861 	if (error != 0)
3862 		return (error);
3863 	error = dorevoke(vp, l->l_cred);
3864 	vrele(vp);
3865 	return (error);
3866 }
3867