xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision b757af438b42b93f8c6571f026d8b8ef3eaf5fc9)
1 /*	$NetBSD: vfs_syscalls.c,v 1.450 2012/03/13 18:40:56 elad 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 1.450 2012/03/13 18:40:56 elad 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/quota.h>
98 #include <sys/quotactl.h>
99 #include <sys/ktrace.h>
100 #ifdef FILEASSOC
101 #include <sys/fileassoc.h>
102 #endif /* FILEASSOC */
103 #include <sys/extattr.h>
104 #include <sys/verified_exec.h>
105 #include <sys/kauth.h>
106 #include <sys/atomic.h>
107 #include <sys/module.h>
108 #include <sys/buf.h>
109 
110 #include <miscfs/genfs/genfs.h>
111 #include <miscfs/syncfs/syncfs.h>
112 #include <miscfs/specfs/specdev.h>
113 
114 #include <nfs/rpcv2.h>
115 #include <nfs/nfsproto.h>
116 #include <nfs/nfs.h>
117 #include <nfs/nfs_var.h>
118 
119 static int change_flags(struct vnode *, u_long, struct lwp *);
120 static int change_mode(struct vnode *, int, struct lwp *l);
121 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
122 static int do_open(lwp_t *, struct pathbuf *, int, int, int *);
123 
124 /*
125  * This table is used to maintain compatibility with 4.3BSD
126  * and NetBSD 0.9 mount syscalls - and possibly other systems.
127  * Note, the order is important!
128  *
129  * Do not modify this table. It should only contain filesystems
130  * supported by NetBSD 0.9 and 4.3BSD.
131  */
132 const char * const mountcompatnames[] = {
133 	NULL,		/* 0 = MOUNT_NONE */
134 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
135 	MOUNT_NFS,	/* 2 */
136 	MOUNT_MFS,	/* 3 */
137 	MOUNT_MSDOS,	/* 4 */
138 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
139 	MOUNT_FDESC,	/* 6 */
140 	MOUNT_KERNFS,	/* 7 */
141 	NULL,		/* 8 = MOUNT_DEVFS */
142 	MOUNT_AFS,	/* 9 */
143 };
144 
145 const int nmountcompatnames = __arraycount(mountcompatnames);
146 
147 static int
148 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags)
149 {
150 	int error;
151 
152 	fp->f_flag = flags & FMASK;
153 	fp->f_type = DTYPE_VNODE;
154 	fp->f_ops = &vnops;
155 	fp->f_data = vp;
156 
157 	if (flags & (O_EXLOCK | O_SHLOCK)) {
158 		struct flock lf;
159 		int type;
160 
161 		lf.l_whence = SEEK_SET;
162 		lf.l_start = 0;
163 		lf.l_len = 0;
164 		if (flags & O_EXLOCK)
165 			lf.l_type = F_WRLCK;
166 		else
167 			lf.l_type = F_RDLCK;
168 		type = F_FLOCK;
169 		if ((flags & FNONBLOCK) == 0)
170 			type |= F_WAIT;
171 		VOP_UNLOCK(vp);
172 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
173 		if (error) {
174 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
175 			fd_abort(l->l_proc, fp, indx);
176 			return error;
177 		}
178 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
179 		atomic_or_uint(&fp->f_flag, FHASLOCK);
180 	}
181 	if (flags & O_CLOEXEC)
182 		fd_set_exclose(l, indx, true);
183 	return 0;
184 }
185 
186 static int
187 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
188     void *data, size_t *data_len)
189 {
190 	struct mount *mp;
191 	int error = 0, saved_flags;
192 
193 	mp = vp->v_mount;
194 	saved_flags = mp->mnt_flag;
195 
196 	/* We can operate only on VV_ROOT nodes. */
197 	if ((vp->v_vflag & VV_ROOT) == 0) {
198 		error = EINVAL;
199 		goto out;
200 	}
201 
202 	/*
203 	 * We only allow the filesystem to be reloaded if it
204 	 * is currently mounted read-only.  Additionally, we
205 	 * prevent read-write to read-only downgrades.
206 	 */
207 	if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
208 	    (mp->mnt_flag & MNT_RDONLY) == 0 &&
209 	    (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) {
210 		error = EOPNOTSUPP;	/* Needs translation */
211 		goto out;
212 	}
213 
214 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
215 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
216 	if (error)
217 		goto out;
218 
219 	if (vfs_busy(mp, NULL)) {
220 		error = EPERM;
221 		goto out;
222 	}
223 
224 	mutex_enter(&mp->mnt_updating);
225 
226 	mp->mnt_flag &= ~MNT_OP_FLAGS;
227 	mp->mnt_flag |= flags & MNT_OP_FLAGS;
228 
229 	/*
230 	 * Set the mount level flags.
231 	 */
232 	if (flags & MNT_RDONLY)
233 		mp->mnt_flag |= MNT_RDONLY;
234 	else if (mp->mnt_flag & MNT_RDONLY)
235 		mp->mnt_iflag |= IMNT_WANTRDWR;
236 	mp->mnt_flag &= ~MNT_BASIC_FLAGS;
237 	mp->mnt_flag |= flags & MNT_BASIC_FLAGS;
238 	error = VFS_MOUNT(mp, path, data, data_len);
239 
240 	if (error && data != NULL) {
241 		int error2;
242 
243 		/*
244 		 * Update failed; let's try and see if it was an
245 		 * export request.  For compat with 3.0 and earlier.
246 		 */
247 		error2 = vfs_hooks_reexport(mp, path, data);
248 
249 		/*
250 		 * Only update error code if the export request was
251 		 * understood but some problem occurred while
252 		 * processing it.
253 		 */
254 		if (error2 != EJUSTRETURN)
255 			error = error2;
256 	}
257 
258 	if (mp->mnt_iflag & IMNT_WANTRDWR)
259 		mp->mnt_flag &= ~MNT_RDONLY;
260 	if (error)
261 		mp->mnt_flag = saved_flags;
262 	mp->mnt_flag &= ~MNT_OP_FLAGS;
263 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
264 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
265 		if (mp->mnt_syncer == NULL)
266 			error = vfs_allocate_syncvnode(mp);
267 	} else {
268 		if (mp->mnt_syncer != NULL)
269 			vfs_deallocate_syncvnode(mp);
270 	}
271 	mutex_exit(&mp->mnt_updating);
272 	vfs_unbusy(mp, false, NULL);
273 
274 	if ((error == 0) && !(saved_flags & MNT_EXTATTR) &&
275 	    (flags & MNT_EXTATTR)) {
276 		if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_START,
277 				   NULL, 0, NULL) != 0) {
278 			printf("%s: failed to start extattr, error = %d",
279 			       vp->v_mount->mnt_stat.f_mntonname, error);
280 			mp->mnt_flag &= ~MNT_EXTATTR;
281 		}
282 	}
283 
284 	if ((error == 0) && (saved_flags & MNT_EXTATTR) &&
285 	    !(flags & MNT_EXTATTR)) {
286 		if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_STOP,
287 				   NULL, 0, NULL) != 0) {
288 			printf("%s: failed to stop extattr, error = %d",
289 			       vp->v_mount->mnt_stat.f_mntonname, error);
290 			mp->mnt_flag |= MNT_RDONLY;
291 		}
292 	}
293  out:
294 	return (error);
295 }
296 
297 static int
298 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
299 {
300 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
301 	int error;
302 
303 	/* Copy file-system type from userspace.  */
304 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
305 	if (error) {
306 		/*
307 		 * Historically, filesystem types were identified by numbers.
308 		 * If we get an integer for the filesystem type instead of a
309 		 * string, we check to see if it matches one of the historic
310 		 * filesystem types.
311 		 */
312 		u_long fsindex = (u_long)fstype;
313 		if (fsindex >= nmountcompatnames ||
314 		    mountcompatnames[fsindex] == NULL)
315 			return ENODEV;
316 		strlcpy(fstypename, mountcompatnames[fsindex],
317 		    sizeof(fstypename));
318 	}
319 
320 	/* Accept `ufs' as an alias for `ffs', for compatibility. */
321 	if (strcmp(fstypename, "ufs") == 0)
322 		fstypename[0] = 'f';
323 
324 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
325 		return 0;
326 
327 	/* If we can autoload a vfs module, try again */
328 	(void)module_autoload(fstypename, MODULE_CLASS_VFS);
329 
330 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
331 		return 0;
332 
333 	return ENODEV;
334 }
335 
336 static int
337 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
338     void *data, size_t *data_len)
339 {
340 	struct mount *mp;
341 	int error;
342 
343 	/* If MNT_GETARGS is specified, it should be the only flag. */
344 	if (flags & ~MNT_GETARGS)
345 		return EINVAL;
346 
347 	mp = vp->v_mount;
348 
349 	/* XXX: probably some notion of "can see" here if we want isolation. */
350 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
351 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
352 	if (error)
353 		return error;
354 
355 	if ((vp->v_vflag & VV_ROOT) == 0)
356 		return EINVAL;
357 
358 	if (vfs_busy(mp, NULL))
359 		return EPERM;
360 
361 	mutex_enter(&mp->mnt_updating);
362 	mp->mnt_flag &= ~MNT_OP_FLAGS;
363 	mp->mnt_flag |= MNT_GETARGS;
364 	error = VFS_MOUNT(mp, path, data, data_len);
365 	mp->mnt_flag &= ~MNT_OP_FLAGS;
366 	mutex_exit(&mp->mnt_updating);
367 
368 	vfs_unbusy(mp, false, NULL);
369 	return (error);
370 }
371 
372 int
373 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
374 {
375 	/* {
376 		syscallarg(const char *) type;
377 		syscallarg(const char *) path;
378 		syscallarg(int) flags;
379 		syscallarg(void *) data;
380 		syscallarg(size_t) data_len;
381 	} */
382 
383 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
384 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
385 	    SCARG(uap, data_len), retval);
386 }
387 
388 int
389 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
390     const char *path, int flags, void *data, enum uio_seg data_seg,
391     size_t data_len, register_t *retval)
392 {
393 	struct vnode *vp;
394 	void *data_buf = data;
395 	bool vfsopsrele = false;
396 	int error;
397 
398 	/* XXX: The calling convention of this routine is totally bizarre */
399 	if (vfsops)
400 		vfsopsrele = true;
401 
402 	/*
403 	 * Get vnode to be covered
404 	 */
405 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
406 	if (error != 0) {
407 		vp = NULL;
408 		goto done;
409 	}
410 
411 	if (vfsops == NULL) {
412 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
413 			vfsops = vp->v_mount->mnt_op;
414 		} else {
415 			/* 'type' is userspace */
416 			error = mount_get_vfsops(type, &vfsops);
417 			if (error != 0)
418 				goto done;
419 			vfsopsrele = true;
420 		}
421 	}
422 
423 	if (data != NULL && data_seg == UIO_USERSPACE) {
424 		if (data_len == 0) {
425 			/* No length supplied, use default for filesystem */
426 			data_len = vfsops->vfs_min_mount_data;
427 			if (data_len > VFS_MAX_MOUNT_DATA) {
428 				error = EINVAL;
429 				goto done;
430 			}
431 			/*
432 			 * Hopefully a longer buffer won't make copyin() fail.
433 			 * For compatibility with 3.0 and earlier.
434 			 */
435 			if (flags & MNT_UPDATE
436 			    && data_len < sizeof (struct mnt_export_args30))
437 				data_len = sizeof (struct mnt_export_args30);
438 		}
439 		data_buf = kmem_alloc(data_len, KM_SLEEP);
440 
441 		/* NFS needs the buffer even for mnt_getargs .... */
442 		error = copyin(data, data_buf, data_len);
443 		if (error != 0)
444 			goto done;
445 	}
446 
447 	if (flags & MNT_GETARGS) {
448 		if (data_len == 0) {
449 			error = EINVAL;
450 			goto done;
451 		}
452 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
453 		if (error != 0)
454 			goto done;
455 		if (data_seg == UIO_USERSPACE)
456 			error = copyout(data_buf, data, data_len);
457 		*retval = data_len;
458 	} else if (flags & MNT_UPDATE) {
459 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
460 	} else {
461 		/* Locking is handled internally in mount_domount(). */
462 		KASSERT(vfsopsrele == true);
463 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
464 		    &data_len);
465 		vfsopsrele = false;
466 
467 		if ((error == 0) && (flags & MNT_EXTATTR)) {
468 			if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_START,
469 					   NULL, 0, NULL) != 0)
470 				printf("%s: failed to start extattr",
471 				       vp->v_mount->mnt_stat.f_mntonname);
472 				/* XXX remove flag */
473 		}
474 	}
475 
476     done:
477 	if (vfsopsrele)
478 		vfs_delref(vfsops);
479     	if (vp != NULL) {
480 	    	vrele(vp);
481 	}
482 	if (data_buf != data)
483 		kmem_free(data_buf, data_len);
484 	return (error);
485 }
486 
487 /*
488  * Unmount a file system.
489  *
490  * Note: unmount takes a path to the vnode mounted on as argument,
491  * not special file (as before).
492  */
493 /* ARGSUSED */
494 int
495 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
496 {
497 	/* {
498 		syscallarg(const char *) path;
499 		syscallarg(int) flags;
500 	} */
501 	struct vnode *vp;
502 	struct mount *mp;
503 	int error;
504 	struct pathbuf *pb;
505 	struct nameidata nd;
506 
507 	error = pathbuf_copyin(SCARG(uap, path), &pb);
508 	if (error) {
509 		return error;
510 	}
511 
512 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
513 	if ((error = namei(&nd)) != 0) {
514 		pathbuf_destroy(pb);
515 		return error;
516 	}
517 	vp = nd.ni_vp;
518 	pathbuf_destroy(pb);
519 
520 	mp = vp->v_mount;
521 	atomic_inc_uint(&mp->mnt_refcnt);
522 	VOP_UNLOCK(vp);
523 
524 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
525 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
526 	if (error) {
527 		vrele(vp);
528 		vfs_destroy(mp);
529 		return (error);
530 	}
531 
532 	/*
533 	 * Don't allow unmounting the root file system.
534 	 */
535 	if (mp->mnt_flag & MNT_ROOTFS) {
536 		vrele(vp);
537 		vfs_destroy(mp);
538 		return (EINVAL);
539 	}
540 
541 	/*
542 	 * Must be the root of the filesystem
543 	 */
544 	if ((vp->v_vflag & VV_ROOT) == 0) {
545 		vrele(vp);
546 		vfs_destroy(mp);
547 		return (EINVAL);
548 	}
549 
550 	vrele(vp);
551 	error = dounmount(mp, SCARG(uap, flags), l);
552 	vfs_destroy(mp);
553 	return error;
554 }
555 
556 /*
557  * Sync each mounted filesystem.
558  */
559 #ifdef DEBUG
560 int syncprt = 0;
561 struct ctldebug debug0 = { "syncprt", &syncprt };
562 #endif
563 
564 void
565 do_sys_sync(struct lwp *l)
566 {
567 	struct mount *mp, *nmp;
568 	int asyncflag;
569 
570 	mutex_enter(&mountlist_lock);
571 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
572 	     mp = nmp) {
573 		if (vfs_busy(mp, &nmp)) {
574 			continue;
575 		}
576 		mutex_enter(&mp->mnt_updating);
577 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
578 			asyncflag = mp->mnt_flag & MNT_ASYNC;
579 			mp->mnt_flag &= ~MNT_ASYNC;
580 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
581 			if (asyncflag)
582 				 mp->mnt_flag |= MNT_ASYNC;
583 		}
584 		mutex_exit(&mp->mnt_updating);
585 		vfs_unbusy(mp, false, &nmp);
586 	}
587 	mutex_exit(&mountlist_lock);
588 #ifdef DEBUG
589 	if (syncprt)
590 		vfs_bufstats();
591 #endif /* DEBUG */
592 }
593 
594 /* ARGSUSED */
595 int
596 sys_sync(struct lwp *l, const void *v, register_t *retval)
597 {
598 	do_sys_sync(l);
599 	return (0);
600 }
601 
602 
603 /*
604  * Access or change filesystem quotas.
605  *
606  * (this is really 14 different calls bundled into one)
607  */
608 
609 static int
610 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u)
611 {
612 	struct quotastat info_k;
613 	int error;
614 
615 	/* ensure any padding bytes are cleared */
616 	memset(&info_k, 0, sizeof(info_k));
617 
618 	error = vfs_quotactl_stat(mp, &info_k);
619 	if (error) {
620 		return error;
621 	}
622 
623 	return copyout(&info_k, info_u, sizeof(info_k));
624 }
625 
626 static int
627 do_sys_quotactl_idtypestat(struct mount *mp, int idtype,
628     struct quotaidtypestat *info_u)
629 {
630 	struct quotaidtypestat info_k;
631 	int error;
632 
633 	/* ensure any padding bytes are cleared */
634 	memset(&info_k, 0, sizeof(info_k));
635 
636 	error = vfs_quotactl_idtypestat(mp, idtype, &info_k);
637 	if (error) {
638 		return error;
639 	}
640 
641 	return copyout(&info_k, info_u, sizeof(info_k));
642 }
643 
644 static int
645 do_sys_quotactl_objtypestat(struct mount *mp, int objtype,
646     struct quotaobjtypestat *info_u)
647 {
648 	struct quotaobjtypestat info_k;
649 	int error;
650 
651 	/* ensure any padding bytes are cleared */
652 	memset(&info_k, 0, sizeof(info_k));
653 
654 	error = vfs_quotactl_objtypestat(mp, objtype, &info_k);
655 	if (error) {
656 		return error;
657 	}
658 
659 	return copyout(&info_k, info_u, sizeof(info_k));
660 }
661 
662 static int
663 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u,
664     struct quotaval *val_u)
665 {
666 	struct quotakey key_k;
667 	struct quotaval val_k;
668 	int error;
669 
670 	/* ensure any padding bytes are cleared */
671 	memset(&val_k, 0, sizeof(val_k));
672 
673 	error = copyin(key_u, &key_k, sizeof(key_k));
674 	if (error) {
675 		return error;
676 	}
677 
678 	error = vfs_quotactl_get(mp, &key_k, &val_k);
679 	if (error) {
680 		return error;
681 	}
682 
683 	return copyout(&val_k, val_u, sizeof(val_k));
684 }
685 
686 static int
687 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u,
688     const struct quotaval *val_u)
689 {
690 	struct quotakey key_k;
691 	struct quotaval val_k;
692 	int error;
693 
694 	error = copyin(key_u, &key_k, sizeof(key_k));
695 	if (error) {
696 		return error;
697 	}
698 
699 	error = copyin(val_u, &val_k, sizeof(val_k));
700 	if (error) {
701 		return error;
702 	}
703 
704 	return vfs_quotactl_put(mp, &key_k, &val_k);
705 }
706 
707 static int
708 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u)
709 {
710 	struct quotakey key_k;
711 	int error;
712 
713 	error = copyin(key_u, &key_k, sizeof(key_k));
714 	if (error) {
715 		return error;
716 	}
717 
718 	return vfs_quotactl_delete(mp, &key_k);
719 }
720 
721 static int
722 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u)
723 {
724 	struct quotakcursor cursor_k;
725 	int error;
726 
727 	/* ensure any padding bytes are cleared */
728 	memset(&cursor_k, 0, sizeof(cursor_k));
729 
730 	error = vfs_quotactl_cursoropen(mp, &cursor_k);
731 	if (error) {
732 		return error;
733 	}
734 
735 	return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
736 }
737 
738 static int
739 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u)
740 {
741 	struct quotakcursor cursor_k;
742 	int error;
743 
744 	error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
745 	if (error) {
746 		return error;
747 	}
748 
749 	return vfs_quotactl_cursorclose(mp, &cursor_k);
750 }
751 
752 static int
753 do_sys_quotactl_cursorskipidtype(struct mount *mp,
754     struct quotakcursor *cursor_u, int idtype)
755 {
756 	struct quotakcursor cursor_k;
757 	int error;
758 
759 	error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
760 	if (error) {
761 		return error;
762 	}
763 
764 	error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype);
765 	if (error) {
766 		return error;
767 	}
768 
769 	return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
770 }
771 
772 static int
773 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u,
774     struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum,
775     unsigned *ret_u)
776 {
777 #define CGET_STACK_MAX 8
778 	struct quotakcursor cursor_k;
779 	struct quotakey stackkeys[CGET_STACK_MAX];
780 	struct quotaval stackvals[CGET_STACK_MAX];
781 	struct quotakey *keys_k;
782 	struct quotaval *vals_k;
783 	unsigned ret_k;
784 	int error;
785 
786 	if (maxnum > 128) {
787 		maxnum = 128;
788 	}
789 
790 	error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
791 	if (error) {
792 		return error;
793 	}
794 
795 	if (maxnum <= CGET_STACK_MAX) {
796 		keys_k = stackkeys;
797 		vals_k = stackvals;
798 		/* ensure any padding bytes are cleared */
799 		memset(keys_k, 0, maxnum * sizeof(keys_k[0]));
800 		memset(vals_k, 0, maxnum * sizeof(vals_k[0]));
801 	} else {
802 		keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP);
803 		vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP);
804 	}
805 
806 	error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum,
807 				       &ret_k);
808 	if (error) {
809 		goto fail;
810 	}
811 
812 	error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0]));
813 	if (error) {
814 		goto fail;
815 	}
816 
817 	error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0]));
818 	if (error) {
819 		goto fail;
820 	}
821 
822 	error = copyout(&ret_k, ret_u, sizeof(ret_k));
823 	if (error) {
824 		goto fail;
825 	}
826 
827 	/* do last to maximize the chance of being able to recover a failure */
828 	error = copyout(&cursor_k, cursor_u, sizeof(cursor_k));
829 
830 fail:
831 	if (keys_k != stackkeys) {
832 		kmem_free(keys_k, maxnum * sizeof(keys_k[0]));
833 	}
834 	if (vals_k != stackvals) {
835 		kmem_free(vals_k, maxnum * sizeof(vals_k[0]));
836 	}
837 	return error;
838 }
839 
840 static int
841 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u,
842     int *ret_u)
843 {
844 	struct quotakcursor cursor_k;
845 	int ret_k;
846 	int error;
847 
848 	error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
849 	if (error) {
850 		return error;
851 	}
852 
853 	error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k);
854 	if (error) {
855 		return error;
856 	}
857 
858 	error = copyout(&ret_k, ret_u, sizeof(ret_k));
859 	if (error) {
860 		return error;
861 	}
862 
863 	return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
864 }
865 
866 static int
867 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u)
868 {
869 	struct quotakcursor cursor_k;
870 	int error;
871 
872 	error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
873 	if (error) {
874 		return error;
875 	}
876 
877 	error = vfs_quotactl_cursorrewind(mp, &cursor_k);
878 	if (error) {
879 		return error;
880 	}
881 
882 	return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
883 }
884 
885 static int
886 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u)
887 {
888 	char *path_k;
889 	int error;
890 
891 	/* XXX this should probably be a struct pathbuf */
892 	path_k = PNBUF_GET();
893 	error = copyin(path_u, path_k, PATH_MAX);
894 	if (error) {
895 		PNBUF_PUT(path_k);
896 		return error;
897 	}
898 
899 	error = vfs_quotactl_quotaon(mp, idtype, path_k);
900 
901 	PNBUF_PUT(path_k);
902 	return error;
903 }
904 
905 static int
906 do_sys_quotactl_quotaoff(struct mount *mp, int idtype)
907 {
908 	return vfs_quotactl_quotaoff(mp, idtype);
909 }
910 
911 int
912 do_sys_quotactl(const char *path_u, const struct quotactl_args *args)
913 {
914 	struct mount *mp;
915 	struct vnode *vp;
916 	int error;
917 
918 	error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp);
919 	if (error != 0)
920 		return (error);
921 	mp = vp->v_mount;
922 
923 	switch (args->qc_op) {
924 	    case QUOTACTL_STAT:
925 		error = do_sys_quotactl_stat(mp, args->u.stat.qc_info);
926 		break;
927 	    case QUOTACTL_IDTYPESTAT:
928 		error = do_sys_quotactl_idtypestat(mp,
929 				args->u.idtypestat.qc_idtype,
930 				args->u.idtypestat.qc_info);
931 		break;
932 	    case QUOTACTL_OBJTYPESTAT:
933 		error = do_sys_quotactl_objtypestat(mp,
934 				args->u.objtypestat.qc_objtype,
935 				args->u.objtypestat.qc_info);
936 		break;
937 	    case QUOTACTL_GET:
938 		error = do_sys_quotactl_get(mp,
939 				args->u.get.qc_key,
940 				args->u.get.qc_val);
941 		break;
942 	    case QUOTACTL_PUT:
943 		error = do_sys_quotactl_put(mp,
944 				args->u.put.qc_key,
945 				args->u.put.qc_val);
946 		break;
947 	    case QUOTACTL_DELETE:
948 		error = do_sys_quotactl_delete(mp, args->u.delete.qc_key);
949 		break;
950 	    case QUOTACTL_CURSOROPEN:
951 		error = do_sys_quotactl_cursoropen(mp,
952 				args->u.cursoropen.qc_cursor);
953 		break;
954 	    case QUOTACTL_CURSORCLOSE:
955 		error = do_sys_quotactl_cursorclose(mp,
956 				args->u.cursorclose.qc_cursor);
957 		break;
958 	    case QUOTACTL_CURSORSKIPIDTYPE:
959 		error = do_sys_quotactl_cursorskipidtype(mp,
960 				args->u.cursorskipidtype.qc_cursor,
961 				args->u.cursorskipidtype.qc_idtype);
962 		break;
963 	    case QUOTACTL_CURSORGET:
964 		error = do_sys_quotactl_cursorget(mp,
965 				args->u.cursorget.qc_cursor,
966 				args->u.cursorget.qc_keys,
967 				args->u.cursorget.qc_vals,
968 				args->u.cursorget.qc_maxnum,
969 				args->u.cursorget.qc_ret);
970 		break;
971 	    case QUOTACTL_CURSORATEND:
972 		error = do_sys_quotactl_cursoratend(mp,
973 				args->u.cursoratend.qc_cursor,
974 				args->u.cursoratend.qc_ret);
975 		break;
976 	    case QUOTACTL_CURSORREWIND:
977 		error = do_sys_quotactl_cursorrewind(mp,
978 				args->u.cursorrewind.qc_cursor);
979 		break;
980 	    case QUOTACTL_QUOTAON:
981 		error = do_sys_quotactl_quotaon(mp,
982 				args->u.quotaon.qc_idtype,
983 				args->u.quotaon.qc_quotafile);
984 		break;
985 	    case QUOTACTL_QUOTAOFF:
986 		error = do_sys_quotactl_quotaoff(mp,
987 				args->u.quotaoff.qc_idtype);
988 		break;
989 	    default:
990 		error = EINVAL;
991 		break;
992 	}
993 
994 	vrele(vp);
995 	return error;
996 }
997 
998 /* ARGSUSED */
999 int
1000 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap,
1001     register_t *retval)
1002 {
1003 	/* {
1004 		syscallarg(const char *) path;
1005 		syscallarg(struct quotactl_args *) args;
1006 	} */
1007 	struct quotactl_args args;
1008 	int error;
1009 
1010 	error = copyin(SCARG(uap, args), &args, sizeof(args));
1011 	if (error) {
1012 		return error;
1013 	}
1014 
1015 	return do_sys_quotactl(SCARG(uap, path), &args);
1016 }
1017 
1018 int
1019 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
1020     int root)
1021 {
1022 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
1023 	int error = 0;
1024 
1025 	/*
1026 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1027 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
1028 	 * overrides MNT_NOWAIT.
1029 	 */
1030 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
1031 	    (flags != MNT_WAIT && flags != 0)) {
1032 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
1033 		goto done;
1034 	}
1035 
1036 	/* Get the filesystem stats now */
1037 	memset(sp, 0, sizeof(*sp));
1038 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
1039 		return error;
1040 	}
1041 
1042 	if (cwdi->cwdi_rdir == NULL)
1043 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
1044 done:
1045 	if (cwdi->cwdi_rdir != NULL) {
1046 		size_t len;
1047 		char *bp;
1048 		char c;
1049 		char *path = PNBUF_GET();
1050 
1051 		bp = path + MAXPATHLEN;
1052 		*--bp = '\0';
1053 		rw_enter(&cwdi->cwdi_lock, RW_READER);
1054 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
1055 		    MAXPATHLEN / 2, 0, l);
1056 		rw_exit(&cwdi->cwdi_lock);
1057 		if (error) {
1058 			PNBUF_PUT(path);
1059 			return error;
1060 		}
1061 		len = strlen(bp);
1062 		if (len != 1) {
1063 			/*
1064 			 * for mount points that are below our root, we can see
1065 			 * them, so we fix up the pathname and return them. The
1066 			 * rest we cannot see, so we don't allow viewing the
1067 			 * data.
1068 			 */
1069 			if (strncmp(bp, sp->f_mntonname, len) == 0 &&
1070 			    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
1071 				(void)strlcpy(sp->f_mntonname,
1072 				    c == '\0' ? "/" : &sp->f_mntonname[len],
1073 				    sizeof(sp->f_mntonname));
1074 			} else {
1075 				if (root)
1076 					(void)strlcpy(sp->f_mntonname, "/",
1077 					    sizeof(sp->f_mntonname));
1078 				else
1079 					error = EPERM;
1080 			}
1081 		}
1082 		PNBUF_PUT(path);
1083 	}
1084 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
1085 	return error;
1086 }
1087 
1088 /*
1089  * Get filesystem statistics by path.
1090  */
1091 int
1092 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
1093 {
1094 	struct mount *mp;
1095 	int error;
1096 	struct vnode *vp;
1097 
1098 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
1099 	if (error != 0)
1100 		return error;
1101 	mp = vp->v_mount;
1102 	error = dostatvfs(mp, sb, l, flags, 1);
1103 	vrele(vp);
1104 	return error;
1105 }
1106 
1107 /* ARGSUSED */
1108 int
1109 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
1110 {
1111 	/* {
1112 		syscallarg(const char *) path;
1113 		syscallarg(struct statvfs *) buf;
1114 		syscallarg(int) flags;
1115 	} */
1116 	struct statvfs *sb;
1117 	int error;
1118 
1119 	sb = STATVFSBUF_GET();
1120 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
1121 	if (error == 0)
1122 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1123 	STATVFSBUF_PUT(sb);
1124 	return error;
1125 }
1126 
1127 /*
1128  * Get filesystem statistics by fd.
1129  */
1130 int
1131 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
1132 {
1133 	file_t *fp;
1134 	struct mount *mp;
1135 	int error;
1136 
1137 	/* fd_getvnode() will use the descriptor for us */
1138 	if ((error = fd_getvnode(fd, &fp)) != 0)
1139 		return (error);
1140 	mp = ((struct vnode *)fp->f_data)->v_mount;
1141 	error = dostatvfs(mp, sb, curlwp, flags, 1);
1142 	fd_putfile(fd);
1143 	return error;
1144 }
1145 
1146 /* ARGSUSED */
1147 int
1148 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
1149 {
1150 	/* {
1151 		syscallarg(int) fd;
1152 		syscallarg(struct statvfs *) buf;
1153 		syscallarg(int) flags;
1154 	} */
1155 	struct statvfs *sb;
1156 	int error;
1157 
1158 	sb = STATVFSBUF_GET();
1159 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1160 	if (error == 0)
1161 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1162 	STATVFSBUF_PUT(sb);
1163 	return error;
1164 }
1165 
1166 
1167 /*
1168  * Get statistics on all filesystems.
1169  */
1170 int
1171 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1172     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1173     register_t *retval)
1174 {
1175 	int root = 0;
1176 	struct proc *p = l->l_proc;
1177 	struct mount *mp, *nmp;
1178 	struct statvfs *sb;
1179 	size_t count, maxcount;
1180 	int error = 0;
1181 
1182 	sb = STATVFSBUF_GET();
1183 	maxcount = bufsize / entry_sz;
1184 	mutex_enter(&mountlist_lock);
1185 	count = 0;
1186 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1187 	     mp = nmp) {
1188 		if (vfs_busy(mp, &nmp)) {
1189 			continue;
1190 		}
1191 		if (sfsp && count < maxcount) {
1192 			error = dostatvfs(mp, sb, l, flags, 0);
1193 			if (error) {
1194 				vfs_unbusy(mp, false, &nmp);
1195 				error = 0;
1196 				continue;
1197 			}
1198 			error = copyfn(sb, sfsp, entry_sz);
1199 			if (error) {
1200 				vfs_unbusy(mp, false, NULL);
1201 				goto out;
1202 			}
1203 			sfsp = (char *)sfsp + entry_sz;
1204 			root |= strcmp(sb->f_mntonname, "/") == 0;
1205 		}
1206 		count++;
1207 		vfs_unbusy(mp, false, &nmp);
1208 	}
1209 	mutex_exit(&mountlist_lock);
1210 
1211 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
1212 		/*
1213 		 * fake a root entry
1214 		 */
1215 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1216 		    sb, l, flags, 1);
1217 		if (error != 0)
1218 			goto out;
1219 		if (sfsp) {
1220 			error = copyfn(sb, sfsp, entry_sz);
1221 			if (error != 0)
1222 				goto out;
1223 		}
1224 		count++;
1225 	}
1226 	if (sfsp && count > maxcount)
1227 		*retval = maxcount;
1228 	else
1229 		*retval = count;
1230 out:
1231 	STATVFSBUF_PUT(sb);
1232 	return error;
1233 }
1234 
1235 int
1236 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1237 {
1238 	/* {
1239 		syscallarg(struct statvfs *) buf;
1240 		syscallarg(size_t) bufsize;
1241 		syscallarg(int) flags;
1242 	} */
1243 
1244 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1245 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1246 }
1247 
1248 /*
1249  * Change current working directory to a given file descriptor.
1250  */
1251 /* ARGSUSED */
1252 int
1253 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1254 {
1255 	/* {
1256 		syscallarg(int) fd;
1257 	} */
1258 	struct proc *p = l->l_proc;
1259 	struct cwdinfo *cwdi;
1260 	struct vnode *vp, *tdp;
1261 	struct mount *mp;
1262 	file_t *fp;
1263 	int error, fd;
1264 
1265 	/* fd_getvnode() will use the descriptor for us */
1266 	fd = SCARG(uap, fd);
1267 	if ((error = fd_getvnode(fd, &fp)) != 0)
1268 		return (error);
1269 	vp = fp->f_data;
1270 
1271 	vref(vp);
1272 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1273 	if (vp->v_type != VDIR)
1274 		error = ENOTDIR;
1275 	else
1276 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1277 	if (error) {
1278 		vput(vp);
1279 		goto out;
1280 	}
1281 	while ((mp = vp->v_mountedhere) != NULL) {
1282 		error = vfs_busy(mp, NULL);
1283 		vput(vp);
1284 		if (error != 0)
1285 			goto out;
1286 		error = VFS_ROOT(mp, &tdp);
1287 		vfs_unbusy(mp, false, NULL);
1288 		if (error)
1289 			goto out;
1290 		vp = tdp;
1291 	}
1292 	VOP_UNLOCK(vp);
1293 
1294 	/*
1295 	 * Disallow changing to a directory not under the process's
1296 	 * current root directory (if there is one).
1297 	 */
1298 	cwdi = p->p_cwdi;
1299 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1300 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1301 		vrele(vp);
1302 		error = EPERM;	/* operation not permitted */
1303 	} else {
1304 		vrele(cwdi->cwdi_cdir);
1305 		cwdi->cwdi_cdir = vp;
1306 	}
1307 	rw_exit(&cwdi->cwdi_lock);
1308 
1309  out:
1310 	fd_putfile(fd);
1311 	return (error);
1312 }
1313 
1314 /*
1315  * Change this process's notion of the root directory to a given file
1316  * descriptor.
1317  */
1318 int
1319 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1320 {
1321 	struct proc *p = l->l_proc;
1322 	struct vnode	*vp;
1323 	file_t	*fp;
1324 	int		 error, fd = SCARG(uap, fd);
1325 
1326 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1327  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1328 		return error;
1329 	/* fd_getvnode() will use the descriptor for us */
1330 	if ((error = fd_getvnode(fd, &fp)) != 0)
1331 		return error;
1332 	vp = fp->f_data;
1333 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1334 	if (vp->v_type != VDIR)
1335 		error = ENOTDIR;
1336 	else
1337 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1338 	VOP_UNLOCK(vp);
1339 	if (error)
1340 		goto out;
1341 	vref(vp);
1342 
1343 	change_root(p->p_cwdi, vp, l);
1344 
1345  out:
1346 	fd_putfile(fd);
1347 	return (error);
1348 }
1349 
1350 /*
1351  * Change current working directory (``.'').
1352  */
1353 /* ARGSUSED */
1354 int
1355 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1356 {
1357 	/* {
1358 		syscallarg(const char *) path;
1359 	} */
1360 	struct proc *p = l->l_proc;
1361 	struct cwdinfo *cwdi;
1362 	int error;
1363 	struct vnode *vp;
1364 
1365 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1366 				  &vp, l)) != 0)
1367 		return (error);
1368 	cwdi = p->p_cwdi;
1369 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1370 	vrele(cwdi->cwdi_cdir);
1371 	cwdi->cwdi_cdir = vp;
1372 	rw_exit(&cwdi->cwdi_lock);
1373 	return (0);
1374 }
1375 
1376 /*
1377  * Change notion of root (``/'') directory.
1378  */
1379 /* ARGSUSED */
1380 int
1381 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1382 {
1383 	/* {
1384 		syscallarg(const char *) path;
1385 	} */
1386 	struct proc *p = l->l_proc;
1387 	int error;
1388 	struct vnode *vp;
1389 
1390 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1391 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1392 		return (error);
1393 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1394 				  &vp, l)) != 0)
1395 		return (error);
1396 
1397 	change_root(p->p_cwdi, vp, l);
1398 
1399 	return (0);
1400 }
1401 
1402 /*
1403  * Common routine for chroot and fchroot.
1404  * NB: callers need to properly authorize the change root operation.
1405  */
1406 void
1407 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1408 {
1409 
1410 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1411 	if (cwdi->cwdi_rdir != NULL)
1412 		vrele(cwdi->cwdi_rdir);
1413 	cwdi->cwdi_rdir = vp;
1414 
1415 	/*
1416 	 * Prevent escaping from chroot by putting the root under
1417 	 * the working directory.  Silently chdir to / if we aren't
1418 	 * already there.
1419 	 */
1420 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1421 		/*
1422 		 * XXX would be more failsafe to change directory to a
1423 		 * deadfs node here instead
1424 		 */
1425 		vrele(cwdi->cwdi_cdir);
1426 		vref(vp);
1427 		cwdi->cwdi_cdir = vp;
1428 	}
1429 	rw_exit(&cwdi->cwdi_lock);
1430 }
1431 
1432 /*
1433  * Common routine for chroot and chdir.
1434  * XXX "where" should be enum uio_seg
1435  */
1436 int
1437 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1438 {
1439 	struct pathbuf *pb;
1440 	struct nameidata nd;
1441 	int error;
1442 
1443 	error = pathbuf_maybe_copyin(path, where, &pb);
1444 	if (error) {
1445 		return error;
1446 	}
1447 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1448 	if ((error = namei(&nd)) != 0) {
1449 		pathbuf_destroy(pb);
1450 		return error;
1451 	}
1452 	*vpp = nd.ni_vp;
1453 	pathbuf_destroy(pb);
1454 
1455 	if ((*vpp)->v_type != VDIR)
1456 		error = ENOTDIR;
1457 	else
1458 		error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1459 
1460 	if (error)
1461 		vput(*vpp);
1462 	else
1463 		VOP_UNLOCK(*vpp);
1464 	return (error);
1465 }
1466 
1467 /*
1468  * Internals of sys_open - path has already been converted into a pathbuf
1469  * (so we can easily reuse this function from other parts of the kernel,
1470  * like posix_spawn post-processing).
1471  */
1472 static int
1473 do_open(lwp_t *l, struct pathbuf *pb, int open_flags, int open_mode, int *fd)
1474 {
1475 	struct proc *p = l->l_proc;
1476 	struct cwdinfo *cwdi = p->p_cwdi;
1477 	file_t *fp;
1478 	struct vnode *vp;
1479 	int flags, cmode;
1480 	int indx, error;
1481 	struct nameidata nd;
1482 
1483 	flags = FFLAGS(open_flags);
1484 	if ((flags & (FREAD | FWRITE)) == 0)
1485 		return EINVAL;
1486 
1487 	if ((error = fd_allocfile(&fp, &indx)) != 0) {
1488 		pathbuf_destroy(pb);
1489 		return error;
1490 	}
1491 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1492 	cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1493 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb);
1494 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1495 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1496 		fd_abort(p, fp, indx);
1497 		if ((error == EDUPFD || error == EMOVEFD) &&
1498 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1499 		    (error =
1500 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1501 			*fd = indx;
1502 			pathbuf_destroy(pb);
1503 			return (0);
1504 		}
1505 		if (error == ERESTART)
1506 			error = EINTR;
1507 		pathbuf_destroy(pb);
1508 		return error;
1509 	}
1510 
1511 	l->l_dupfd = 0;
1512 	vp = nd.ni_vp;
1513 	pathbuf_destroy(pb);
1514 
1515 	if ((error = open_setfp(l, fp, vp, indx, flags)))
1516 		return error;
1517 
1518 	VOP_UNLOCK(vp);
1519 	*fd = indx;
1520 	fd_affix(p, fp, indx);
1521 	return 0;
1522 }
1523 
1524 int
1525 fd_open(const char *path, int open_flags, int open_mode, int *fd)
1526 {
1527 	struct pathbuf *pb;
1528 	int oflags;
1529 
1530 	oflags = FFLAGS(open_flags);
1531 	if ((oflags & (FREAD | FWRITE)) == 0)
1532 		return EINVAL;
1533 
1534 	pb = pathbuf_create(path);
1535 	if (pb == NULL)
1536 		return ENOMEM;
1537 
1538 	return do_open(curlwp, pb, open_flags, open_mode, fd);
1539 }
1540 
1541 /*
1542  * Check permissions, allocate an open file structure,
1543  * and call the device open routine if any.
1544  */
1545 int
1546 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1547 {
1548 	/* {
1549 		syscallarg(const char *) path;
1550 		syscallarg(int) flags;
1551 		syscallarg(int) mode;
1552 	} */
1553 	struct pathbuf *pb;
1554 	int result, flags, error;
1555 
1556 	flags = FFLAGS(SCARG(uap, flags));
1557 	if ((flags & (FREAD | FWRITE)) == 0)
1558 		return (EINVAL);
1559 
1560 	error = pathbuf_copyin(SCARG(uap, path), &pb);
1561 	if (error)
1562 		return error;
1563 
1564 	error = do_open(l, pb, SCARG(uap, flags), SCARG(uap, mode), &result);
1565 	if (error)
1566 		return error;
1567 
1568 	*retval = result;
1569 	return 0;
1570 }
1571 
1572 int
1573 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval)
1574 {
1575 	/* {
1576 		syscallarg(int) fd;
1577 		syscallarg(const char *) path;
1578 		syscallarg(int) flags;
1579 		syscallarg(int) mode;
1580 	} */
1581 
1582 	return ENOSYS;
1583 }
1584 
1585 static void
1586 vfs__fhfree(fhandle_t *fhp)
1587 {
1588 	size_t fhsize;
1589 
1590 	if (fhp == NULL) {
1591 		return;
1592 	}
1593 	fhsize = FHANDLE_SIZE(fhp);
1594 	kmem_free(fhp, fhsize);
1595 }
1596 
1597 /*
1598  * vfs_composefh: compose a filehandle.
1599  */
1600 
1601 int
1602 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1603 {
1604 	struct mount *mp;
1605 	struct fid *fidp;
1606 	int error;
1607 	size_t needfhsize;
1608 	size_t fidsize;
1609 
1610 	mp = vp->v_mount;
1611 	fidp = NULL;
1612 	if (*fh_size < FHANDLE_SIZE_MIN) {
1613 		fidsize = 0;
1614 	} else {
1615 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1616 		if (fhp != NULL) {
1617 			memset(fhp, 0, *fh_size);
1618 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1619 			fidp = &fhp->fh_fid;
1620 		}
1621 	}
1622 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1623 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1624 	if (error == 0 && *fh_size < needfhsize) {
1625 		error = E2BIG;
1626 	}
1627 	*fh_size = needfhsize;
1628 	return error;
1629 }
1630 
1631 int
1632 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1633 {
1634 	struct mount *mp;
1635 	fhandle_t *fhp;
1636 	size_t fhsize;
1637 	size_t fidsize;
1638 	int error;
1639 
1640 	*fhpp = NULL;
1641 	mp = vp->v_mount;
1642 	fidsize = 0;
1643 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1644 	KASSERT(error != 0);
1645 	if (error != E2BIG) {
1646 		goto out;
1647 	}
1648 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1649 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1650 	if (fhp == NULL) {
1651 		error = ENOMEM;
1652 		goto out;
1653 	}
1654 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1655 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1656 	if (error == 0) {
1657 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1658 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1659 		*fhpp = fhp;
1660 	} else {
1661 		kmem_free(fhp, fhsize);
1662 	}
1663 out:
1664 	return error;
1665 }
1666 
1667 void
1668 vfs_composefh_free(fhandle_t *fhp)
1669 {
1670 
1671 	vfs__fhfree(fhp);
1672 }
1673 
1674 /*
1675  * vfs_fhtovp: lookup a vnode by a filehandle.
1676  */
1677 
1678 int
1679 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1680 {
1681 	struct mount *mp;
1682 	int error;
1683 
1684 	*vpp = NULL;
1685 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1686 	if (mp == NULL) {
1687 		error = ESTALE;
1688 		goto out;
1689 	}
1690 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1691 		error = EOPNOTSUPP;
1692 		goto out;
1693 	}
1694 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1695 out:
1696 	return error;
1697 }
1698 
1699 /*
1700  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1701  * the needed size.
1702  */
1703 
1704 int
1705 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1706 {
1707 	fhandle_t *fhp;
1708 	int error;
1709 
1710 	*fhpp = NULL;
1711 	if (fhsize > FHANDLE_SIZE_MAX) {
1712 		return EINVAL;
1713 	}
1714 	if (fhsize < FHANDLE_SIZE_MIN) {
1715 		return EINVAL;
1716 	}
1717 again:
1718 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1719 	if (fhp == NULL) {
1720 		return ENOMEM;
1721 	}
1722 	error = copyin(ufhp, fhp, fhsize);
1723 	if (error == 0) {
1724 		/* XXX this check shouldn't be here */
1725 		if (FHANDLE_SIZE(fhp) == fhsize) {
1726 			*fhpp = fhp;
1727 			return 0;
1728 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1729 			/*
1730 			 * a kludge for nfsv2 padded handles.
1731 			 */
1732 			size_t sz;
1733 
1734 			sz = FHANDLE_SIZE(fhp);
1735 			kmem_free(fhp, fhsize);
1736 			fhsize = sz;
1737 			goto again;
1738 		} else {
1739 			/*
1740 			 * userland told us wrong size.
1741 			 */
1742 		    	error = EINVAL;
1743 		}
1744 	}
1745 	kmem_free(fhp, fhsize);
1746 	return error;
1747 }
1748 
1749 void
1750 vfs_copyinfh_free(fhandle_t *fhp)
1751 {
1752 
1753 	vfs__fhfree(fhp);
1754 }
1755 
1756 /*
1757  * Get file handle system call
1758  */
1759 int
1760 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1761 {
1762 	/* {
1763 		syscallarg(char *) fname;
1764 		syscallarg(fhandle_t *) fhp;
1765 		syscallarg(size_t *) fh_size;
1766 	} */
1767 	struct vnode *vp;
1768 	fhandle_t *fh;
1769 	int error;
1770 	struct pathbuf *pb;
1771 	struct nameidata nd;
1772 	size_t sz;
1773 	size_t usz;
1774 
1775 	/*
1776 	 * Must be super user
1777 	 */
1778 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1779 	    0, NULL, NULL, NULL);
1780 	if (error)
1781 		return (error);
1782 
1783 	error = pathbuf_copyin(SCARG(uap, fname), &pb);
1784 	if (error) {
1785 		return error;
1786 	}
1787 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1788 	error = namei(&nd);
1789 	if (error) {
1790 		pathbuf_destroy(pb);
1791 		return error;
1792 	}
1793 	vp = nd.ni_vp;
1794 	pathbuf_destroy(pb);
1795 
1796 	error = vfs_composefh_alloc(vp, &fh);
1797 	vput(vp);
1798 	if (error != 0) {
1799 		goto out;
1800 	}
1801 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1802 	if (error != 0) {
1803 		goto out;
1804 	}
1805 	sz = FHANDLE_SIZE(fh);
1806 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1807 	if (error != 0) {
1808 		goto out;
1809 	}
1810 	if (usz >= sz) {
1811 		error = copyout(fh, SCARG(uap, fhp), sz);
1812 	} else {
1813 		error = E2BIG;
1814 	}
1815 out:
1816 	vfs_composefh_free(fh);
1817 	return (error);
1818 }
1819 
1820 /*
1821  * Open a file given a file handle.
1822  *
1823  * Check permissions, allocate an open file structure,
1824  * and call the device open routine if any.
1825  */
1826 
1827 int
1828 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1829     register_t *retval)
1830 {
1831 	file_t *fp;
1832 	struct vnode *vp = NULL;
1833 	kauth_cred_t cred = l->l_cred;
1834 	file_t *nfp;
1835 	int indx, error = 0;
1836 	struct vattr va;
1837 	fhandle_t *fh;
1838 	int flags;
1839 	proc_t *p;
1840 
1841 	p = curproc;
1842 
1843 	/*
1844 	 * Must be super user
1845 	 */
1846 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1847 	    0, NULL, NULL, NULL)))
1848 		return (error);
1849 
1850 	flags = FFLAGS(oflags);
1851 	if ((flags & (FREAD | FWRITE)) == 0)
1852 		return (EINVAL);
1853 	if ((flags & O_CREAT))
1854 		return (EINVAL);
1855 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1856 		return (error);
1857 	fp = nfp;
1858 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1859 	if (error != 0) {
1860 		goto bad;
1861 	}
1862 	error = vfs_fhtovp(fh, &vp);
1863 	if (error != 0) {
1864 		goto bad;
1865 	}
1866 
1867 	/* Now do an effective vn_open */
1868 
1869 	if (vp->v_type == VSOCK) {
1870 		error = EOPNOTSUPP;
1871 		goto bad;
1872 	}
1873 	error = vn_openchk(vp, cred, flags);
1874 	if (error != 0)
1875 		goto bad;
1876 	if (flags & O_TRUNC) {
1877 		VOP_UNLOCK(vp);			/* XXX */
1878 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1879 		vattr_null(&va);
1880 		va.va_size = 0;
1881 		error = VOP_SETATTR(vp, &va, cred);
1882 		if (error)
1883 			goto bad;
1884 	}
1885 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1886 		goto bad;
1887 	if (flags & FWRITE) {
1888 		mutex_enter(vp->v_interlock);
1889 		vp->v_writecount++;
1890 		mutex_exit(vp->v_interlock);
1891 	}
1892 
1893 	/* done with modified vn_open, now finish what sys_open does. */
1894 	if ((error = open_setfp(l, fp, vp, indx, flags)))
1895 		return error;
1896 
1897 	VOP_UNLOCK(vp);
1898 	*retval = indx;
1899 	fd_affix(p, fp, indx);
1900 	vfs_copyinfh_free(fh);
1901 	return (0);
1902 
1903 bad:
1904 	fd_abort(p, fp, indx);
1905 	if (vp != NULL)
1906 		vput(vp);
1907 	vfs_copyinfh_free(fh);
1908 	return (error);
1909 }
1910 
1911 int
1912 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1913 {
1914 	/* {
1915 		syscallarg(const void *) fhp;
1916 		syscallarg(size_t) fh_size;
1917 		syscallarg(int) flags;
1918 	} */
1919 
1920 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1921 	    SCARG(uap, flags), retval);
1922 }
1923 
1924 int
1925 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1926 {
1927 	int error;
1928 	fhandle_t *fh;
1929 	struct vnode *vp;
1930 
1931 	/*
1932 	 * Must be super user
1933 	 */
1934 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1935 	    0, NULL, NULL, NULL)))
1936 		return (error);
1937 
1938 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1939 	if (error != 0)
1940 		return error;
1941 
1942 	error = vfs_fhtovp(fh, &vp);
1943 	vfs_copyinfh_free(fh);
1944 	if (error != 0)
1945 		return error;
1946 
1947 	error = vn_stat(vp, sb);
1948 	vput(vp);
1949 	return error;
1950 }
1951 
1952 
1953 /* ARGSUSED */
1954 int
1955 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1956 {
1957 	/* {
1958 		syscallarg(const void *) fhp;
1959 		syscallarg(size_t) fh_size;
1960 		syscallarg(struct stat *) sb;
1961 	} */
1962 	struct stat sb;
1963 	int error;
1964 
1965 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1966 	if (error)
1967 		return error;
1968 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1969 }
1970 
1971 int
1972 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1973     int flags)
1974 {
1975 	fhandle_t *fh;
1976 	struct mount *mp;
1977 	struct vnode *vp;
1978 	int error;
1979 
1980 	/*
1981 	 * Must be super user
1982 	 */
1983 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1984 	    0, NULL, NULL, NULL)))
1985 		return error;
1986 
1987 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1988 	if (error != 0)
1989 		return error;
1990 
1991 	error = vfs_fhtovp(fh, &vp);
1992 	vfs_copyinfh_free(fh);
1993 	if (error != 0)
1994 		return error;
1995 
1996 	mp = vp->v_mount;
1997 	error = dostatvfs(mp, sb, l, flags, 1);
1998 	vput(vp);
1999 	return error;
2000 }
2001 
2002 /* ARGSUSED */
2003 int
2004 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
2005 {
2006 	/* {
2007 		syscallarg(const void *) fhp;
2008 		syscallarg(size_t) fh_size;
2009 		syscallarg(struct statvfs *) buf;
2010 		syscallarg(int)	flags;
2011 	} */
2012 	struct statvfs *sb = STATVFSBUF_GET();
2013 	int error;
2014 
2015 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
2016 	    SCARG(uap, flags));
2017 	if (error == 0)
2018 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
2019 	STATVFSBUF_PUT(sb);
2020 	return error;
2021 }
2022 
2023 /*
2024  * Create a special file.
2025  */
2026 /* ARGSUSED */
2027 int
2028 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
2029     register_t *retval)
2030 {
2031 	/* {
2032 		syscallarg(const char *) path;
2033 		syscallarg(mode_t) mode;
2034 		syscallarg(dev_t) dev;
2035 	} */
2036 	return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
2037 	    SCARG(uap, dev), retval, UIO_USERSPACE);
2038 }
2039 
2040 int
2041 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap,
2042     register_t *retval)
2043 {
2044 	/* {
2045 		syscallarg(int) fd;
2046 		syscallarg(const char *) path;
2047 		syscallarg(mode_t) mode;
2048 		syscallarg(uint32_t) dev;
2049 	} */
2050 
2051 	return ENOSYS;
2052 }
2053 
2054 int
2055 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
2056     register_t *retval, enum uio_seg seg)
2057 {
2058 	struct proc *p = l->l_proc;
2059 	struct vnode *vp;
2060 	struct vattr vattr;
2061 	int error, optype;
2062 	struct pathbuf *pb;
2063 	struct nameidata nd;
2064 	const char *pathstring;
2065 
2066 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
2067 	    0, NULL, NULL, NULL)) != 0)
2068 		return (error);
2069 
2070 	optype = VOP_MKNOD_DESCOFFSET;
2071 
2072 	error = pathbuf_maybe_copyin(pathname, seg, &pb);
2073 	if (error) {
2074 		return error;
2075 	}
2076 	pathstring = pathbuf_stringcopy_get(pb);
2077 	if (pathstring == NULL) {
2078 		pathbuf_destroy(pb);
2079 		return ENOMEM;
2080 	}
2081 
2082 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
2083 	if ((error = namei(&nd)) != 0)
2084 		goto out;
2085 	vp = nd.ni_vp;
2086 
2087 	if (vp != NULL)
2088 		error = EEXIST;
2089 	else {
2090 		vattr_null(&vattr);
2091 		/* We will read cwdi->cwdi_cmask unlocked. */
2092 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
2093 		vattr.va_rdev = dev;
2094 
2095 		switch (mode & S_IFMT) {
2096 		case S_IFMT:	/* used by badsect to flag bad sectors */
2097 			vattr.va_type = VBAD;
2098 			break;
2099 		case S_IFCHR:
2100 			vattr.va_type = VCHR;
2101 			break;
2102 		case S_IFBLK:
2103 			vattr.va_type = VBLK;
2104 			break;
2105 		case S_IFWHT:
2106 			optype = VOP_WHITEOUT_DESCOFFSET;
2107 			break;
2108 		case S_IFREG:
2109 #if NVERIEXEC > 0
2110 			error = veriexec_openchk(l, nd.ni_vp, pathstring,
2111 			    O_CREAT);
2112 #endif /* NVERIEXEC > 0 */
2113 			vattr.va_type = VREG;
2114 			vattr.va_rdev = VNOVAL;
2115 			optype = VOP_CREATE_DESCOFFSET;
2116 			break;
2117 		default:
2118 			error = EINVAL;
2119 			break;
2120 		}
2121 	}
2122 	if (error == 0 && optype == VOP_MKNOD_DESCOFFSET
2123 	    && vattr.va_rdev == VNOVAL)
2124 		error = EINVAL;
2125 	if (!error) {
2126 		switch (optype) {
2127 		case VOP_WHITEOUT_DESCOFFSET:
2128 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
2129 			if (error)
2130 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2131 			vput(nd.ni_dvp);
2132 			break;
2133 
2134 		case VOP_MKNOD_DESCOFFSET:
2135 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
2136 						&nd.ni_cnd, &vattr);
2137 			if (error == 0)
2138 				vput(nd.ni_vp);
2139 			break;
2140 
2141 		case VOP_CREATE_DESCOFFSET:
2142 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
2143 						&nd.ni_cnd, &vattr);
2144 			if (error == 0)
2145 				vput(nd.ni_vp);
2146 			break;
2147 		}
2148 	} else {
2149 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2150 		if (nd.ni_dvp == vp)
2151 			vrele(nd.ni_dvp);
2152 		else
2153 			vput(nd.ni_dvp);
2154 		if (vp)
2155 			vrele(vp);
2156 	}
2157 out:
2158 	pathbuf_stringcopy_put(pb, pathstring);
2159 	pathbuf_destroy(pb);
2160 	return (error);
2161 }
2162 
2163 /*
2164  * Create a named pipe.
2165  */
2166 /* ARGSUSED */
2167 int
2168 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
2169 {
2170 	/* {
2171 		syscallarg(const char *) path;
2172 		syscallarg(int) mode;
2173 	} */
2174 	struct proc *p = l->l_proc;
2175 	struct vattr vattr;
2176 	int error;
2177 	struct pathbuf *pb;
2178 	struct nameidata nd;
2179 
2180 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2181 	if (error) {
2182 		return error;
2183 	}
2184 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
2185 	if ((error = namei(&nd)) != 0) {
2186 		pathbuf_destroy(pb);
2187 		return error;
2188 	}
2189 	if (nd.ni_vp != NULL) {
2190 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2191 		if (nd.ni_dvp == nd.ni_vp)
2192 			vrele(nd.ni_dvp);
2193 		else
2194 			vput(nd.ni_dvp);
2195 		vrele(nd.ni_vp);
2196 		pathbuf_destroy(pb);
2197 		return (EEXIST);
2198 	}
2199 	vattr_null(&vattr);
2200 	vattr.va_type = VFIFO;
2201 	/* We will read cwdi->cwdi_cmask unlocked. */
2202 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
2203 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2204 	if (error == 0)
2205 		vput(nd.ni_vp);
2206 	pathbuf_destroy(pb);
2207 	return (error);
2208 }
2209 
2210 int
2211 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap,
2212     register_t *retval)
2213 {
2214 	/* {
2215 		syscallarg(int) fd;
2216 		syscallarg(const char *) path;
2217 		syscallarg(int) mode;
2218 	} */
2219 
2220 	return ENOSYS;
2221 }
2222 /*
2223  * Make a hard file link.
2224  */
2225 /* ARGSUSED */
2226 static int
2227 do_sys_link(struct lwp *l, const char *path, const char *link,
2228 	    int follow, register_t *retval)
2229 {
2230 	struct vnode *vp;
2231 	struct pathbuf *linkpb;
2232 	struct nameidata nd;
2233 	namei_simple_flags_t namei_simple_flags;
2234 	int error;
2235 
2236 	if (follow)
2237 		namei_simple_flags = NSM_FOLLOW_TRYEMULROOT;
2238 	else
2239 		namei_simple_flags =  NSM_NOFOLLOW_TRYEMULROOT;
2240 
2241 	error = namei_simple_user(path, namei_simple_flags, &vp);
2242 	if (error != 0)
2243 		return (error);
2244 	error = pathbuf_copyin(link, &linkpb);
2245 	if (error) {
2246 		goto out1;
2247 	}
2248 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
2249 	if ((error = namei(&nd)) != 0)
2250 		goto out2;
2251 	if (nd.ni_vp) {
2252 		error = EEXIST;
2253 		goto abortop;
2254 	}
2255 	/* Prevent hard links on directories. */
2256 	if (vp->v_type == VDIR) {
2257 		error = EPERM;
2258 		goto abortop;
2259 	}
2260 	/* Prevent cross-mount operation. */
2261 	if (nd.ni_dvp->v_mount != vp->v_mount) {
2262 		error = EXDEV;
2263 		goto abortop;
2264 	}
2265 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2266 out2:
2267 	pathbuf_destroy(linkpb);
2268 out1:
2269 	vrele(vp);
2270 	return (error);
2271 abortop:
2272 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2273 	if (nd.ni_dvp == nd.ni_vp)
2274 		vrele(nd.ni_dvp);
2275 	else
2276 		vput(nd.ni_dvp);
2277 	if (nd.ni_vp != NULL)
2278 		vrele(nd.ni_vp);
2279 	goto out2;
2280 }
2281 
2282 int
2283 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
2284 {
2285 	/* {
2286 		syscallarg(const char *) path;
2287 		syscallarg(const char *) link;
2288 	} */
2289 	const char *path = SCARG(uap, path);
2290 	const char *link = SCARG(uap, link);
2291 
2292 	return do_sys_link(l, path, link, 1, retval);
2293 }
2294 
2295 int
2296 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap,
2297     register_t *retval)
2298 {
2299 	/* {
2300 		syscallarg(int) fd1;
2301 		syscallarg(const char *) name1;
2302 		syscallarg(int) fd2;
2303 		syscallarg(const char *) name2;
2304 		syscallarg(int) flags;
2305 	} */
2306 	const char *name1 = SCARG(uap, name1);
2307 	const char *name2 = SCARG(uap, name2);
2308 	int follow;
2309 
2310 	/*
2311 	 * Specified fd1 and fd2 are not yet implemented
2312 	 */
2313 	if ((SCARG(uap, fd1) != AT_FDCWD) || (SCARG(uap, fd2) != AT_FDCWD))
2314 		return ENOSYS;
2315 
2316 	follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW;
2317 
2318 	return do_sys_link(l, name1, name2, follow, retval);
2319 }
2320 
2321 
2322 int
2323 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg)
2324 {
2325 	struct proc *p = curproc;
2326 	struct vattr vattr;
2327 	char *path;
2328 	int error;
2329 	struct pathbuf *linkpb;
2330 	struct nameidata nd;
2331 
2332 	path = PNBUF_GET();
2333 	if (seg == UIO_USERSPACE) {
2334 		if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0)
2335 			goto out1;
2336 		if ((error = pathbuf_copyin(link, &linkpb)) != 0)
2337 			goto out1;
2338 	} else {
2339 		KASSERT(strlen(patharg) < MAXPATHLEN);
2340 		strcpy(path, patharg);
2341 		linkpb = pathbuf_create(link);
2342 		if (linkpb == NULL) {
2343 			error = ENOMEM;
2344 			goto out1;
2345 		}
2346 	}
2347 	ktrkuser("symlink-target", path, strlen(path));
2348 
2349 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
2350 	if ((error = namei(&nd)) != 0)
2351 		goto out2;
2352 	if (nd.ni_vp) {
2353 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2354 		if (nd.ni_dvp == nd.ni_vp)
2355 			vrele(nd.ni_dvp);
2356 		else
2357 			vput(nd.ni_dvp);
2358 		vrele(nd.ni_vp);
2359 		error = EEXIST;
2360 		goto out2;
2361 	}
2362 	vattr_null(&vattr);
2363 	vattr.va_type = VLNK;
2364 	/* We will read cwdi->cwdi_cmask unlocked. */
2365 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2366 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2367 	if (error == 0)
2368 		vput(nd.ni_vp);
2369 out2:
2370 	pathbuf_destroy(linkpb);
2371 out1:
2372 	PNBUF_PUT(path);
2373 	return (error);
2374 }
2375 
2376 /*
2377  * Make a symbolic link.
2378  */
2379 /* ARGSUSED */
2380 int
2381 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2382 {
2383 	/* {
2384 		syscallarg(const char *) path;
2385 		syscallarg(const char *) link;
2386 	} */
2387 
2388 	return do_sys_symlink(SCARG(uap, path), SCARG(uap, link),
2389 	    UIO_USERSPACE);
2390 }
2391 
2392 int
2393 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap,
2394     register_t *retval)
2395 {
2396 	/* {
2397 		syscallarg(int) fd;
2398 		syscallarg(const char *) path;
2399 		syscallarg(const char *) link;
2400 	} */
2401 
2402 	return ENOSYS;
2403 }
2404 
2405 /*
2406  * Delete a whiteout from the filesystem.
2407  */
2408 /* ARGSUSED */
2409 int
2410 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2411 {
2412 	/* {
2413 		syscallarg(const char *) path;
2414 	} */
2415 	int error;
2416 	struct pathbuf *pb;
2417 	struct nameidata nd;
2418 
2419 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2420 	if (error) {
2421 		return error;
2422 	}
2423 
2424 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb);
2425 	error = namei(&nd);
2426 	if (error) {
2427 		pathbuf_destroy(pb);
2428 		return (error);
2429 	}
2430 
2431 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2432 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2433 		if (nd.ni_dvp == nd.ni_vp)
2434 			vrele(nd.ni_dvp);
2435 		else
2436 			vput(nd.ni_dvp);
2437 		if (nd.ni_vp)
2438 			vrele(nd.ni_vp);
2439 		pathbuf_destroy(pb);
2440 		return (EEXIST);
2441 	}
2442 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2443 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2444 	vput(nd.ni_dvp);
2445 	pathbuf_destroy(pb);
2446 	return (error);
2447 }
2448 
2449 /*
2450  * Delete a name from the filesystem.
2451  */
2452 /* ARGSUSED */
2453 int
2454 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2455 {
2456 	/* {
2457 		syscallarg(const char *) path;
2458 	} */
2459 
2460 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2461 }
2462 
2463 int
2464 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap,
2465     register_t *retval)
2466 {
2467 	/* {
2468 		syscallarg(int) fd;
2469 		syscallarg(const char *) path;
2470 	} */
2471 
2472 	return ENOSYS;
2473 }
2474 
2475 int
2476 do_sys_unlink(const char *arg, enum uio_seg seg)
2477 {
2478 	struct vnode *vp;
2479 	int error;
2480 	struct pathbuf *pb;
2481 	struct nameidata nd;
2482 	const char *pathstring;
2483 
2484 	error = pathbuf_maybe_copyin(arg, seg, &pb);
2485 	if (error) {
2486 		return error;
2487 	}
2488 	pathstring = pathbuf_stringcopy_get(pb);
2489 	if (pathstring == NULL) {
2490 		pathbuf_destroy(pb);
2491 		return ENOMEM;
2492 	}
2493 
2494 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
2495 	if ((error = namei(&nd)) != 0)
2496 		goto out;
2497 	vp = nd.ni_vp;
2498 
2499 	/*
2500 	 * The root of a mounted filesystem cannot be deleted.
2501 	 */
2502 	if (vp->v_vflag & VV_ROOT) {
2503 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2504 		if (nd.ni_dvp == vp)
2505 			vrele(nd.ni_dvp);
2506 		else
2507 			vput(nd.ni_dvp);
2508 		vput(vp);
2509 		error = EBUSY;
2510 		goto out;
2511 	}
2512 
2513 #if NVERIEXEC > 0
2514 	/* Handle remove requests for veriexec entries. */
2515 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) {
2516 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2517 		if (nd.ni_dvp == vp)
2518 			vrele(nd.ni_dvp);
2519 		else
2520 			vput(nd.ni_dvp);
2521 		vput(vp);
2522 		goto out;
2523 	}
2524 #endif /* NVERIEXEC > 0 */
2525 
2526 #ifdef FILEASSOC
2527 	(void)fileassoc_file_delete(vp);
2528 #endif /* FILEASSOC */
2529 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2530 out:
2531 	pathbuf_stringcopy_put(pb, pathstring);
2532 	pathbuf_destroy(pb);
2533 	return (error);
2534 }
2535 
2536 /*
2537  * Reposition read/write file offset.
2538  */
2539 int
2540 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2541 {
2542 	/* {
2543 		syscallarg(int) fd;
2544 		syscallarg(int) pad;
2545 		syscallarg(off_t) offset;
2546 		syscallarg(int) whence;
2547 	} */
2548 	kauth_cred_t cred = l->l_cred;
2549 	file_t *fp;
2550 	struct vnode *vp;
2551 	struct vattr vattr;
2552 	off_t newoff;
2553 	int error, fd;
2554 
2555 	fd = SCARG(uap, fd);
2556 
2557 	if ((fp = fd_getfile(fd)) == NULL)
2558 		return (EBADF);
2559 
2560 	vp = fp->f_data;
2561 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2562 		error = ESPIPE;
2563 		goto out;
2564 	}
2565 
2566 	switch (SCARG(uap, whence)) {
2567 	case SEEK_CUR:
2568 		newoff = fp->f_offset + SCARG(uap, offset);
2569 		break;
2570 	case SEEK_END:
2571 		vn_lock(vp, LK_SHARED | LK_RETRY);
2572 		error = VOP_GETATTR(vp, &vattr, cred);
2573 		VOP_UNLOCK(vp);
2574 		if (error) {
2575 			goto out;
2576 		}
2577 		newoff = SCARG(uap, offset) + vattr.va_size;
2578 		break;
2579 	case SEEK_SET:
2580 		newoff = SCARG(uap, offset);
2581 		break;
2582 	default:
2583 		error = EINVAL;
2584 		goto out;
2585 	}
2586 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2587 		*(off_t *)retval = fp->f_offset = newoff;
2588 	}
2589  out:
2590  	fd_putfile(fd);
2591 	return (error);
2592 }
2593 
2594 /*
2595  * Positional read system call.
2596  */
2597 int
2598 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2599 {
2600 	/* {
2601 		syscallarg(int) fd;
2602 		syscallarg(void *) buf;
2603 		syscallarg(size_t) nbyte;
2604 		syscallarg(off_t) offset;
2605 	} */
2606 	file_t *fp;
2607 	struct vnode *vp;
2608 	off_t offset;
2609 	int error, fd = SCARG(uap, fd);
2610 
2611 	if ((fp = fd_getfile(fd)) == NULL)
2612 		return (EBADF);
2613 
2614 	if ((fp->f_flag & FREAD) == 0) {
2615 		fd_putfile(fd);
2616 		return (EBADF);
2617 	}
2618 
2619 	vp = fp->f_data;
2620 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2621 		error = ESPIPE;
2622 		goto out;
2623 	}
2624 
2625 	offset = SCARG(uap, offset);
2626 
2627 	/*
2628 	 * XXX This works because no file systems actually
2629 	 * XXX take any action on the seek operation.
2630 	 */
2631 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2632 		goto out;
2633 
2634 	/* dofileread() will unuse the descriptor for us */
2635 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2636 	    &offset, 0, retval));
2637 
2638  out:
2639 	fd_putfile(fd);
2640 	return (error);
2641 }
2642 
2643 /*
2644  * Positional scatter read system call.
2645  */
2646 int
2647 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2648 {
2649 	/* {
2650 		syscallarg(int) fd;
2651 		syscallarg(const struct iovec *) iovp;
2652 		syscallarg(int) iovcnt;
2653 		syscallarg(off_t) offset;
2654 	} */
2655 	off_t offset = SCARG(uap, offset);
2656 
2657 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2658 	    SCARG(uap, iovcnt), &offset, 0, retval);
2659 }
2660 
2661 /*
2662  * Positional write system call.
2663  */
2664 int
2665 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2666 {
2667 	/* {
2668 		syscallarg(int) fd;
2669 		syscallarg(const void *) buf;
2670 		syscallarg(size_t) nbyte;
2671 		syscallarg(off_t) offset;
2672 	} */
2673 	file_t *fp;
2674 	struct vnode *vp;
2675 	off_t offset;
2676 	int error, fd = SCARG(uap, fd);
2677 
2678 	if ((fp = fd_getfile(fd)) == NULL)
2679 		return (EBADF);
2680 
2681 	if ((fp->f_flag & FWRITE) == 0) {
2682 		fd_putfile(fd);
2683 		return (EBADF);
2684 	}
2685 
2686 	vp = fp->f_data;
2687 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2688 		error = ESPIPE;
2689 		goto out;
2690 	}
2691 
2692 	offset = SCARG(uap, offset);
2693 
2694 	/*
2695 	 * XXX This works because no file systems actually
2696 	 * XXX take any action on the seek operation.
2697 	 */
2698 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2699 		goto out;
2700 
2701 	/* dofilewrite() will unuse the descriptor for us */
2702 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2703 	    &offset, 0, retval));
2704 
2705  out:
2706 	fd_putfile(fd);
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Positional gather write system call.
2712  */
2713 int
2714 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2715 {
2716 	/* {
2717 		syscallarg(int) fd;
2718 		syscallarg(const struct iovec *) iovp;
2719 		syscallarg(int) iovcnt;
2720 		syscallarg(off_t) offset;
2721 	} */
2722 	off_t offset = SCARG(uap, offset);
2723 
2724 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2725 	    SCARG(uap, iovcnt), &offset, 0, retval);
2726 }
2727 
2728 /*
2729  * Check access permissions.
2730  */
2731 int
2732 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2733 {
2734 	/* {
2735 		syscallarg(const char *) path;
2736 		syscallarg(int) flags;
2737 	} */
2738 	kauth_cred_t cred;
2739 	struct vnode *vp;
2740 	int error, flags;
2741 	struct pathbuf *pb;
2742 	struct nameidata nd;
2743 
2744 	CTASSERT(F_OK == 0);
2745 	if ((SCARG(uap, flags) & ~(R_OK | W_OK | X_OK)) != 0) {
2746 		/* nonsense flags */
2747 		return EINVAL;
2748 	}
2749 
2750 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2751 	if (error) {
2752 		return error;
2753 	}
2754 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2755 
2756 	/* Override default credentials */
2757 	cred = kauth_cred_dup(l->l_cred);
2758 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2759 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2760 	nd.ni_cnd.cn_cred = cred;
2761 
2762 	if ((error = namei(&nd)) != 0) {
2763 		pathbuf_destroy(pb);
2764 		goto out;
2765 	}
2766 	vp = nd.ni_vp;
2767 	pathbuf_destroy(pb);
2768 
2769 	/* Flags == 0 means only check for existence. */
2770 	if (SCARG(uap, flags)) {
2771 		flags = 0;
2772 		if (SCARG(uap, flags) & R_OK)
2773 			flags |= VREAD;
2774 		if (SCARG(uap, flags) & W_OK)
2775 			flags |= VWRITE;
2776 		if (SCARG(uap, flags) & X_OK)
2777 			flags |= VEXEC;
2778 
2779 		error = VOP_ACCESS(vp, flags, cred);
2780 		if (!error && (flags & VWRITE))
2781 			error = vn_writechk(vp);
2782 	}
2783 	vput(vp);
2784 out:
2785 	kauth_cred_free(cred);
2786 	return (error);
2787 }
2788 
2789 int
2790 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap,
2791     register_t *retval)
2792 {
2793 	/* {
2794 		syscallarg(int) fd;
2795 		syscallarg(const char *) path;
2796 		syscallarg(int) amode;
2797 		syscallarg(int) flag;
2798 	} */
2799 
2800 	return ENOSYS;
2801 }
2802 
2803 /*
2804  * Common code for all sys_stat functions, including compat versions.
2805  */
2806 int
2807 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb)
2808 {
2809 	int error;
2810 	struct pathbuf *pb;
2811 	struct nameidata nd;
2812 
2813 	error = pathbuf_copyin(userpath, &pb);
2814 	if (error) {
2815 		return error;
2816 	}
2817 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb);
2818 	error = namei(&nd);
2819 	if (error != 0) {
2820 		pathbuf_destroy(pb);
2821 		return error;
2822 	}
2823 	error = vn_stat(nd.ni_vp, sb);
2824 	vput(nd.ni_vp);
2825 	pathbuf_destroy(pb);
2826 	return error;
2827 }
2828 
2829 /*
2830  * Get file status; this version follows links.
2831  */
2832 /* ARGSUSED */
2833 int
2834 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
2835 {
2836 	/* {
2837 		syscallarg(const char *) path;
2838 		syscallarg(struct stat *) ub;
2839 	} */
2840 	struct stat sb;
2841 	int error;
2842 
2843 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2844 	if (error)
2845 		return error;
2846 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2847 }
2848 
2849 /*
2850  * Get file status; this version does not follow links.
2851  */
2852 /* ARGSUSED */
2853 int
2854 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
2855 {
2856 	/* {
2857 		syscallarg(const char *) path;
2858 		syscallarg(struct stat *) ub;
2859 	} */
2860 	struct stat sb;
2861 	int error;
2862 
2863 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2864 	if (error)
2865 		return error;
2866 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2867 }
2868 
2869 int
2870 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap,
2871     register_t *retval)
2872 {
2873 	/* {
2874 		syscallarg(int) fd;
2875 		syscallarg(const char *) path;
2876 		syscallarg(struct stat *) ub;
2877 		syscallarg(int) flag;
2878 	} */
2879 
2880 	return ENOSYS;
2881 }
2882 /*
2883  * Get configurable pathname variables.
2884  */
2885 /* ARGSUSED */
2886 int
2887 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2888 {
2889 	/* {
2890 		syscallarg(const char *) path;
2891 		syscallarg(int) name;
2892 	} */
2893 	int error;
2894 	struct pathbuf *pb;
2895 	struct nameidata nd;
2896 
2897 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2898 	if (error) {
2899 		return error;
2900 	}
2901 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2902 	if ((error = namei(&nd)) != 0) {
2903 		pathbuf_destroy(pb);
2904 		return (error);
2905 	}
2906 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2907 	vput(nd.ni_vp);
2908 	pathbuf_destroy(pb);
2909 	return (error);
2910 }
2911 
2912 /*
2913  * Return target name of a symbolic link.
2914  */
2915 /* ARGSUSED */
2916 int
2917 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2918 {
2919 	/* {
2920 		syscallarg(const char *) path;
2921 		syscallarg(char *) buf;
2922 		syscallarg(size_t) count;
2923 	} */
2924 	struct vnode *vp;
2925 	struct iovec aiov;
2926 	struct uio auio;
2927 	int error;
2928 	struct pathbuf *pb;
2929 	struct nameidata nd;
2930 
2931 	error = pathbuf_copyin(SCARG(uap, path), &pb);
2932 	if (error) {
2933 		return error;
2934 	}
2935 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2936 	if ((error = namei(&nd)) != 0) {
2937 		pathbuf_destroy(pb);
2938 		return error;
2939 	}
2940 	vp = nd.ni_vp;
2941 	pathbuf_destroy(pb);
2942 	if (vp->v_type != VLNK)
2943 		error = EINVAL;
2944 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2945 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2946 		aiov.iov_base = SCARG(uap, buf);
2947 		aiov.iov_len = SCARG(uap, count);
2948 		auio.uio_iov = &aiov;
2949 		auio.uio_iovcnt = 1;
2950 		auio.uio_offset = 0;
2951 		auio.uio_rw = UIO_READ;
2952 		KASSERT(l == curlwp);
2953 		auio.uio_vmspace = l->l_proc->p_vmspace;
2954 		auio.uio_resid = SCARG(uap, count);
2955 		error = VOP_READLINK(vp, &auio, l->l_cred);
2956 	}
2957 	vput(vp);
2958 	*retval = SCARG(uap, count) - auio.uio_resid;
2959 	return (error);
2960 }
2961 
2962 int
2963 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap,
2964     register_t *retval)
2965 {
2966 	/* {
2967 		syscallarg(int) fd;
2968 		syscallarg(const char *) path;
2969 		syscallarg(char *) buf;
2970 		syscallarg(size_t) count;
2971 	} */
2972 
2973 	return ENOSYS;
2974 }
2975 
2976 /*
2977  * Change flags of a file given a path name.
2978  */
2979 /* ARGSUSED */
2980 int
2981 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2982 {
2983 	/* {
2984 		syscallarg(const char *) path;
2985 		syscallarg(u_long) flags;
2986 	} */
2987 	struct vnode *vp;
2988 	int error;
2989 
2990 	error = namei_simple_user(SCARG(uap, path),
2991 				NSM_FOLLOW_TRYEMULROOT, &vp);
2992 	if (error != 0)
2993 		return (error);
2994 	error = change_flags(vp, SCARG(uap, flags), l);
2995 	vput(vp);
2996 	return (error);
2997 }
2998 
2999 /*
3000  * Change flags of a file given a file descriptor.
3001  */
3002 /* ARGSUSED */
3003 int
3004 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
3005 {
3006 	/* {
3007 		syscallarg(int) fd;
3008 		syscallarg(u_long) flags;
3009 	} */
3010 	struct vnode *vp;
3011 	file_t *fp;
3012 	int error;
3013 
3014 	/* fd_getvnode() will use the descriptor for us */
3015 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3016 		return (error);
3017 	vp = fp->f_data;
3018 	error = change_flags(vp, SCARG(uap, flags), l);
3019 	VOP_UNLOCK(vp);
3020 	fd_putfile(SCARG(uap, fd));
3021 	return (error);
3022 }
3023 
3024 /*
3025  * Change flags of a file given a path name; this version does
3026  * not follow links.
3027  */
3028 int
3029 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
3030 {
3031 	/* {
3032 		syscallarg(const char *) path;
3033 		syscallarg(u_long) flags;
3034 	} */
3035 	struct vnode *vp;
3036 	int error;
3037 
3038 	error = namei_simple_user(SCARG(uap, path),
3039 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
3040 	if (error != 0)
3041 		return (error);
3042 	error = change_flags(vp, SCARG(uap, flags), l);
3043 	vput(vp);
3044 	return (error);
3045 }
3046 
3047 /*
3048  * Common routine to change flags of a file.
3049  */
3050 int
3051 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
3052 {
3053 	struct vattr vattr;
3054 	int error;
3055 
3056 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3057 
3058 	vattr_null(&vattr);
3059 	vattr.va_flags = flags;
3060 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3061 
3062 	return (error);
3063 }
3064 
3065 /*
3066  * Change mode of a file given path name; this version follows links.
3067  */
3068 /* ARGSUSED */
3069 int
3070 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
3071 {
3072 	/* {
3073 		syscallarg(const char *) path;
3074 		syscallarg(int) mode;
3075 	} */
3076 	int error;
3077 	struct vnode *vp;
3078 
3079 	error = namei_simple_user(SCARG(uap, path),
3080 				NSM_FOLLOW_TRYEMULROOT, &vp);
3081 	if (error != 0)
3082 		return (error);
3083 
3084 	error = change_mode(vp, SCARG(uap, mode), l);
3085 
3086 	vrele(vp);
3087 	return (error);
3088 }
3089 
3090 /*
3091  * Change mode of a file given a file descriptor.
3092  */
3093 /* ARGSUSED */
3094 int
3095 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
3096 {
3097 	/* {
3098 		syscallarg(int) fd;
3099 		syscallarg(int) mode;
3100 	} */
3101 	file_t *fp;
3102 	int error;
3103 
3104 	/* fd_getvnode() will use the descriptor for us */
3105 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3106 		return (error);
3107 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
3108 	fd_putfile(SCARG(uap, fd));
3109 	return (error);
3110 }
3111 
3112 int
3113 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap,
3114     register_t *retval)
3115 {
3116 	/* {
3117 		syscallarg(int) fd;
3118 		syscallarg(const char *) path;
3119 		syscallarg(int) mode;
3120 		syscallarg(int) flag;
3121 	} */
3122 
3123 	return ENOSYS;
3124 }
3125 
3126 /*
3127  * Change mode of a file given path name; this version does not follow links.
3128  */
3129 /* ARGSUSED */
3130 int
3131 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
3132 {
3133 	/* {
3134 		syscallarg(const char *) path;
3135 		syscallarg(int) mode;
3136 	} */
3137 	int error;
3138 	struct vnode *vp;
3139 
3140 	error = namei_simple_user(SCARG(uap, path),
3141 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
3142 	if (error != 0)
3143 		return (error);
3144 
3145 	error = change_mode(vp, SCARG(uap, mode), l);
3146 
3147 	vrele(vp);
3148 	return (error);
3149 }
3150 
3151 /*
3152  * Common routine to set mode given a vnode.
3153  */
3154 static int
3155 change_mode(struct vnode *vp, int mode, struct lwp *l)
3156 {
3157 	struct vattr vattr;
3158 	int error;
3159 
3160 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3161 	vattr_null(&vattr);
3162 	vattr.va_mode = mode & ALLPERMS;
3163 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3164 	VOP_UNLOCK(vp);
3165 	return (error);
3166 }
3167 
3168 /*
3169  * Set ownership given a path name; this version follows links.
3170  */
3171 /* ARGSUSED */
3172 int
3173 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
3174 {
3175 	/* {
3176 		syscallarg(const char *) path;
3177 		syscallarg(uid_t) uid;
3178 		syscallarg(gid_t) gid;
3179 	} */
3180 	int error;
3181 	struct vnode *vp;
3182 
3183 	error = namei_simple_user(SCARG(uap, path),
3184 				NSM_FOLLOW_TRYEMULROOT, &vp);
3185 	if (error != 0)
3186 		return (error);
3187 
3188 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
3189 
3190 	vrele(vp);
3191 	return (error);
3192 }
3193 
3194 /*
3195  * Set ownership given a path name; this version follows links.
3196  * Provides POSIX semantics.
3197  */
3198 /* ARGSUSED */
3199 int
3200 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
3201 {
3202 	/* {
3203 		syscallarg(const char *) path;
3204 		syscallarg(uid_t) uid;
3205 		syscallarg(gid_t) gid;
3206 	} */
3207 	int error;
3208 	struct vnode *vp;
3209 
3210 	error = namei_simple_user(SCARG(uap, path),
3211 				NSM_FOLLOW_TRYEMULROOT, &vp);
3212 	if (error != 0)
3213 		return (error);
3214 
3215 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
3216 
3217 	vrele(vp);
3218 	return (error);
3219 }
3220 
3221 /*
3222  * Set ownership given a file descriptor.
3223  */
3224 /* ARGSUSED */
3225 int
3226 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
3227 {
3228 	/* {
3229 		syscallarg(int) fd;
3230 		syscallarg(uid_t) uid;
3231 		syscallarg(gid_t) gid;
3232 	} */
3233 	int error;
3234 	file_t *fp;
3235 
3236 	/* fd_getvnode() will use the descriptor for us */
3237 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3238 		return (error);
3239 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
3240 	    l, 0);
3241 	fd_putfile(SCARG(uap, fd));
3242 	return (error);
3243 }
3244 
3245 int
3246 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap,
3247     register_t *retval)
3248 {
3249 	/* {
3250 		syscallarg(int) fd;
3251 		syscallarg(const char *) path;
3252 		syscallarg(uid_t) uid;
3253 		syscallarg(gid_t) gid;
3254 		syscallarg(int) flag;
3255 	} */
3256 
3257 	return ENOSYS;
3258 }
3259 
3260 /*
3261  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
3262  */
3263 /* ARGSUSED */
3264 int
3265 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
3266 {
3267 	/* {
3268 		syscallarg(int) fd;
3269 		syscallarg(uid_t) uid;
3270 		syscallarg(gid_t) gid;
3271 	} */
3272 	int error;
3273 	file_t *fp;
3274 
3275 	/* fd_getvnode() will use the descriptor for us */
3276 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3277 		return (error);
3278 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
3279 	    l, 1);
3280 	fd_putfile(SCARG(uap, fd));
3281 	return (error);
3282 }
3283 
3284 /*
3285  * Set ownership given a path name; this version does not follow links.
3286  */
3287 /* ARGSUSED */
3288 int
3289 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
3290 {
3291 	/* {
3292 		syscallarg(const char *) path;
3293 		syscallarg(uid_t) uid;
3294 		syscallarg(gid_t) gid;
3295 	} */
3296 	int error;
3297 	struct vnode *vp;
3298 
3299 	error = namei_simple_user(SCARG(uap, path),
3300 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
3301 	if (error != 0)
3302 		return (error);
3303 
3304 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
3305 
3306 	vrele(vp);
3307 	return (error);
3308 }
3309 
3310 /*
3311  * Set ownership given a path name; this version does not follow links.
3312  * Provides POSIX/XPG semantics.
3313  */
3314 /* ARGSUSED */
3315 int
3316 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
3317 {
3318 	/* {
3319 		syscallarg(const char *) path;
3320 		syscallarg(uid_t) uid;
3321 		syscallarg(gid_t) gid;
3322 	} */
3323 	int error;
3324 	struct vnode *vp;
3325 
3326 	error = namei_simple_user(SCARG(uap, path),
3327 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
3328 	if (error != 0)
3329 		return (error);
3330 
3331 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
3332 
3333 	vrele(vp);
3334 	return (error);
3335 }
3336 
3337 /*
3338  * Common routine to set ownership given a vnode.
3339  */
3340 static int
3341 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
3342     int posix_semantics)
3343 {
3344 	struct vattr vattr;
3345 	mode_t newmode;
3346 	int error;
3347 
3348 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3349 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
3350 		goto out;
3351 
3352 #define CHANGED(x) ((int)(x) != -1)
3353 	newmode = vattr.va_mode;
3354 	if (posix_semantics) {
3355 		/*
3356 		 * POSIX/XPG semantics: if the caller is not the super-user,
3357 		 * clear set-user-id and set-group-id bits.  Both POSIX and
3358 		 * the XPG consider the behaviour for calls by the super-user
3359 		 * implementation-defined; we leave the set-user-id and set-
3360 		 * group-id settings intact in that case.
3361 		 */
3362 		if (vattr.va_mode & S_ISUID) {
3363 			error = kauth_authorize_vnode(l->l_cred,
3364 			    KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM);
3365 			if (error)
3366 				newmode &= ~S_ISUID;
3367 		}
3368 		if (vattr.va_mode & S_ISGID) {
3369 			error = kauth_authorize_vnode(l->l_cred,
3370 			    KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM);
3371 			if (error)
3372 				newmode &= ~S_ISGID;
3373 		}
3374 	} else {
3375 		/*
3376 		 * NetBSD semantics: when changing owner and/or group,
3377 		 * clear the respective bit(s).
3378 		 */
3379 		if (CHANGED(uid))
3380 			newmode &= ~S_ISUID;
3381 		if (CHANGED(gid))
3382 			newmode &= ~S_ISGID;
3383 	}
3384 	/* Update va_mode iff altered. */
3385 	if (vattr.va_mode == newmode)
3386 		newmode = VNOVAL;
3387 
3388 	vattr_null(&vattr);
3389 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
3390 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
3391 	vattr.va_mode = newmode;
3392 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3393 #undef CHANGED
3394 
3395 out:
3396 	VOP_UNLOCK(vp);
3397 	return (error);
3398 }
3399 
3400 /*
3401  * Set the access and modification times given a path name; this
3402  * version follows links.
3403  */
3404 /* ARGSUSED */
3405 int
3406 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
3407     register_t *retval)
3408 {
3409 	/* {
3410 		syscallarg(const char *) path;
3411 		syscallarg(const struct timeval *) tptr;
3412 	} */
3413 
3414 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
3415 	    SCARG(uap, tptr), UIO_USERSPACE);
3416 }
3417 
3418 /*
3419  * Set the access and modification times given a file descriptor.
3420  */
3421 /* ARGSUSED */
3422 int
3423 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
3424     register_t *retval)
3425 {
3426 	/* {
3427 		syscallarg(int) fd;
3428 		syscallarg(const struct timeval *) tptr;
3429 	} */
3430 	int error;
3431 	file_t *fp;
3432 
3433 	/* fd_getvnode() will use the descriptor for us */
3434 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3435 		return (error);
3436 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
3437 	    UIO_USERSPACE);
3438 	fd_putfile(SCARG(uap, fd));
3439 	return (error);
3440 }
3441 
3442 int
3443 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap,
3444     register_t *retval)
3445 {
3446 	/* {
3447 		syscallarg(int) fd;
3448 		syscallarg(const struct timespec *) tptr;
3449 	} */
3450 	int error;
3451 	file_t *fp;
3452 
3453 	/* fd_getvnode() will use the descriptor for us */
3454 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3455 		return (error);
3456 	error = do_sys_utimens(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
3457 	    UIO_USERSPACE);
3458 	fd_putfile(SCARG(uap, fd));
3459 	return (error);
3460 }
3461 
3462 /*
3463  * Set the access and modification times given a path name; this
3464  * version does not follow links.
3465  */
3466 int
3467 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
3468     register_t *retval)
3469 {
3470 	/* {
3471 		syscallarg(const char *) path;
3472 		syscallarg(const struct timeval *) tptr;
3473 	} */
3474 
3475 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
3476 	    SCARG(uap, tptr), UIO_USERSPACE);
3477 }
3478 
3479 int
3480 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap,
3481     register_t *retval)
3482 {
3483 	/* {
3484 		syscallarg(int) fd;
3485 		syscallarg(const char *) path;
3486 		syscallarg(const struct timespec *) tptr;
3487 		syscallarg(int) flag;
3488 	} */
3489 	int follow;
3490 	const struct timespec *tptr;
3491 
3492 	/*
3493 	 * Specified fd is not yet implemented
3494 	 */
3495 	if (SCARG(uap, fd) != AT_FDCWD)
3496 		return ENOSYS;
3497 
3498 	tptr = SCARG(uap, tptr);
3499 	follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3500 
3501 	return do_sys_utimens(l, NULL, SCARG(uap, path), follow,
3502 	    tptr, UIO_USERSPACE);
3503 }
3504 
3505 /*
3506  * Common routine to set access and modification times given a vnode.
3507  */
3508 int
3509 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag,
3510     const struct timespec *tptr, enum uio_seg seg)
3511 {
3512 	struct vattr vattr;
3513 	int error, dorele = 0;
3514 	namei_simple_flags_t sflags;
3515 
3516 	bool vanull, setbirthtime;
3517 	struct timespec ts[2];
3518 
3519 	/*
3520 	 * I have checked all callers and they pass either FOLLOW,
3521 	 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
3522 	 * is 0. More to the point, they don't pass anything else.
3523 	 * Let's keep it that way at least until the namei interfaces
3524 	 * are fully sanitized.
3525 	 */
3526 	KASSERT(flag == NOFOLLOW || flag == FOLLOW);
3527 	sflags = (flag == FOLLOW) ?
3528 		NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3529 
3530 	if (tptr == NULL) {
3531 		vanull = true;
3532 		nanotime(&ts[0]);
3533 		ts[1] = ts[0];
3534 	} else {
3535 		vanull = false;
3536 		if (seg != UIO_SYSSPACE) {
3537 			error = copyin(tptr, ts, sizeof (ts));
3538 			if (error != 0)
3539 				return error;
3540 		} else {
3541 			ts[0] = tptr[0];
3542 			ts[1] = tptr[1];
3543 		}
3544 	}
3545 
3546 	if (ts[0].tv_nsec == UTIME_NOW) {
3547 		nanotime(&ts[0]);
3548 		if (ts[1].tv_nsec == UTIME_NOW) {
3549 			vanull = true;
3550 			ts[1] = ts[0];
3551 		}
3552 	} else if (ts[1].tv_nsec == UTIME_NOW)
3553 		nanotime(&ts[1]);
3554 
3555 	if (vp == NULL) {
3556 		/* note: SEG describes TPTR, not PATH; PATH is always user */
3557 		error = namei_simple_user(path, sflags, &vp);
3558 		if (error != 0)
3559 			return error;
3560 		dorele = 1;
3561 	}
3562 
3563 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3564 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3565 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
3566 	vattr_null(&vattr);
3567 
3568 	if (ts[0].tv_nsec != UTIME_OMIT)
3569 		vattr.va_atime = ts[0];
3570 
3571 	if (ts[1].tv_nsec != UTIME_OMIT) {
3572 		vattr.va_mtime = ts[1];
3573 		if (setbirthtime)
3574 			vattr.va_birthtime = ts[1];
3575 	}
3576 
3577 	if (vanull)
3578 		vattr.va_vaflags |= VA_UTIMES_NULL;
3579 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3580 	VOP_UNLOCK(vp);
3581 
3582 	if (dorele != 0)
3583 		vrele(vp);
3584 
3585 	return error;
3586 }
3587 
3588 int
3589 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3590     const struct timeval *tptr, enum uio_seg seg)
3591 {
3592 	struct timespec ts[2];
3593 	struct timespec *tsptr = NULL;
3594 	int error;
3595 
3596 	if (tptr != NULL) {
3597 		struct timeval tv[2];
3598 
3599 		if (seg != UIO_SYSSPACE) {
3600 			error = copyin(tptr, tv, sizeof (tv));
3601 			if (error != 0)
3602 				return error;
3603 			tptr = tv;
3604 		}
3605 
3606 		if ((tv[0].tv_usec == UTIME_NOW) ||
3607 		    (tv[0].tv_usec == UTIME_OMIT))
3608 			ts[0].tv_nsec = tv[0].tv_usec;
3609 		else
3610 			TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3611 
3612 		if ((tv[1].tv_usec == UTIME_NOW) ||
3613 		    (tv[1].tv_usec == UTIME_OMIT))
3614 			ts[1].tv_nsec = tv[1].tv_usec;
3615 		else
3616 			TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3617 
3618 		tsptr = &ts[0];
3619 	}
3620 
3621 	return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE);
3622 }
3623 
3624 /*
3625  * Truncate a file given its path name.
3626  */
3627 /* ARGSUSED */
3628 int
3629 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3630 {
3631 	/* {
3632 		syscallarg(const char *) path;
3633 		syscallarg(int) pad;
3634 		syscallarg(off_t) length;
3635 	} */
3636 	struct vnode *vp;
3637 	struct vattr vattr;
3638 	int error;
3639 
3640 	error = namei_simple_user(SCARG(uap, path),
3641 				NSM_FOLLOW_TRYEMULROOT, &vp);
3642 	if (error != 0)
3643 		return (error);
3644 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3645 	if (vp->v_type == VDIR)
3646 		error = EISDIR;
3647 	else if ((error = vn_writechk(vp)) == 0 &&
3648 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3649 		vattr_null(&vattr);
3650 		vattr.va_size = SCARG(uap, length);
3651 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3652 	}
3653 	vput(vp);
3654 	return (error);
3655 }
3656 
3657 /*
3658  * Truncate a file given a file descriptor.
3659  */
3660 /* ARGSUSED */
3661 int
3662 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3663 {
3664 	/* {
3665 		syscallarg(int) fd;
3666 		syscallarg(int) pad;
3667 		syscallarg(off_t) length;
3668 	} */
3669 	struct vattr vattr;
3670 	struct vnode *vp;
3671 	file_t *fp;
3672 	int error;
3673 
3674 	/* fd_getvnode() will use the descriptor for us */
3675 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3676 		return (error);
3677 	if ((fp->f_flag & FWRITE) == 0) {
3678 		error = EINVAL;
3679 		goto out;
3680 	}
3681 	vp = fp->f_data;
3682 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3683 	if (vp->v_type == VDIR)
3684 		error = EISDIR;
3685 	else if ((error = vn_writechk(vp)) == 0) {
3686 		vattr_null(&vattr);
3687 		vattr.va_size = SCARG(uap, length);
3688 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3689 	}
3690 	VOP_UNLOCK(vp);
3691  out:
3692 	fd_putfile(SCARG(uap, fd));
3693 	return (error);
3694 }
3695 
3696 /*
3697  * Sync an open file.
3698  */
3699 /* ARGSUSED */
3700 int
3701 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3702 {
3703 	/* {
3704 		syscallarg(int) fd;
3705 	} */
3706 	struct vnode *vp;
3707 	file_t *fp;
3708 	int error;
3709 
3710 	/* fd_getvnode() will use the descriptor for us */
3711 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3712 		return (error);
3713 	vp = fp->f_data;
3714 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3715 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3716 	VOP_UNLOCK(vp);
3717 	fd_putfile(SCARG(uap, fd));
3718 	return (error);
3719 }
3720 
3721 /*
3722  * Sync a range of file data.  API modeled after that found in AIX.
3723  *
3724  * FDATASYNC indicates that we need only save enough metadata to be able
3725  * to re-read the written data.  Note we duplicate AIX's requirement that
3726  * the file be open for writing.
3727  */
3728 /* ARGSUSED */
3729 int
3730 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3731 {
3732 	/* {
3733 		syscallarg(int) fd;
3734 		syscallarg(int) flags;
3735 		syscallarg(off_t) start;
3736 		syscallarg(off_t) length;
3737 	} */
3738 	struct vnode *vp;
3739 	file_t *fp;
3740 	int flags, nflags;
3741 	off_t s, e, len;
3742 	int error;
3743 
3744 	/* fd_getvnode() will use the descriptor for us */
3745 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3746 		return (error);
3747 
3748 	if ((fp->f_flag & FWRITE) == 0) {
3749 		error = EBADF;
3750 		goto out;
3751 	}
3752 
3753 	flags = SCARG(uap, flags);
3754 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3755 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3756 		error = EINVAL;
3757 		goto out;
3758 	}
3759 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3760 	if (flags & FDATASYNC)
3761 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3762 	else
3763 		nflags = FSYNC_WAIT;
3764 	if (flags & FDISKSYNC)
3765 		nflags |= FSYNC_CACHE;
3766 
3767 	len = SCARG(uap, length);
3768 	/* If length == 0, we do the whole file, and s = e = 0 will do that */
3769 	if (len) {
3770 		s = SCARG(uap, start);
3771 		e = s + len;
3772 		if (e < s) {
3773 			error = EINVAL;
3774 			goto out;
3775 		}
3776 	} else {
3777 		e = 0;
3778 		s = 0;
3779 	}
3780 
3781 	vp = fp->f_data;
3782 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3783 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3784 	VOP_UNLOCK(vp);
3785 out:
3786 	fd_putfile(SCARG(uap, fd));
3787 	return (error);
3788 }
3789 
3790 /*
3791  * Sync the data of an open file.
3792  */
3793 /* ARGSUSED */
3794 int
3795 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3796 {
3797 	/* {
3798 		syscallarg(int) fd;
3799 	} */
3800 	struct vnode *vp;
3801 	file_t *fp;
3802 	int error;
3803 
3804 	/* fd_getvnode() will use the descriptor for us */
3805 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3806 		return (error);
3807 	if ((fp->f_flag & FWRITE) == 0) {
3808 		fd_putfile(SCARG(uap, fd));
3809 		return (EBADF);
3810 	}
3811 	vp = fp->f_data;
3812 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3813 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3814 	VOP_UNLOCK(vp);
3815 	fd_putfile(SCARG(uap, fd));
3816 	return (error);
3817 }
3818 
3819 /*
3820  * Rename files, (standard) BSD semantics frontend.
3821  */
3822 /* ARGSUSED */
3823 int
3824 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3825 {
3826 	/* {
3827 		syscallarg(const char *) from;
3828 		syscallarg(const char *) to;
3829 	} */
3830 
3831 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3832 }
3833 
3834 int
3835 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap,
3836     register_t *retval)
3837 {
3838 	/* {
3839 		syscallarg(int) fromfd;
3840 		syscallarg(const char *) from;
3841 		syscallarg(int) tofd;
3842 		syscallarg(const char *) to;
3843 	} */
3844 
3845 	return ENOSYS;
3846 }
3847 
3848 /*
3849  * Rename files, POSIX semantics frontend.
3850  */
3851 /* ARGSUSED */
3852 int
3853 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3854 {
3855 	/* {
3856 		syscallarg(const char *) from;
3857 		syscallarg(const char *) to;
3858 	} */
3859 
3860 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3861 }
3862 
3863 /*
3864  * Rename files.  Source and destination must either both be directories,
3865  * or both not be directories.  If target is a directory, it must be empty.
3866  * If `from' and `to' refer to the same object, the value of the `retain'
3867  * argument is used to determine whether `from' will be
3868  *
3869  * (retain == 0)	deleted unless `from' and `to' refer to the same
3870  *			object in the file system's name space (BSD).
3871  * (retain == 1)	always retained (POSIX).
3872  */
3873 int
3874 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3875 {
3876 	struct vnode *tvp, *fvp, *tdvp;
3877 	struct pathbuf *frompb, *topb;
3878 	struct nameidata fromnd, tond;
3879 	struct mount *fs;
3880 	int error;
3881 
3882 	error = pathbuf_maybe_copyin(from, seg, &frompb);
3883 	if (error) {
3884 		return error;
3885 	}
3886 	error = pathbuf_maybe_copyin(to, seg, &topb);
3887 	if (error) {
3888 		pathbuf_destroy(frompb);
3889 		return error;
3890 	}
3891 
3892 	NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME,
3893 	    frompb);
3894 	if ((error = namei(&fromnd)) != 0) {
3895 		pathbuf_destroy(frompb);
3896 		pathbuf_destroy(topb);
3897 		return (error);
3898 	}
3899 	if (fromnd.ni_dvp != fromnd.ni_vp)
3900 		VOP_UNLOCK(fromnd.ni_dvp);
3901 	fvp = fromnd.ni_vp;
3902 
3903 	fs = fvp->v_mount;
3904 	error = VFS_RENAMELOCK_ENTER(fs);
3905 	if (error) {
3906 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3907 		vrele(fromnd.ni_dvp);
3908 		vrele(fvp);
3909 		goto out1;
3910 	}
3911 
3912 	/*
3913 	 * close, partially, yet another race - ideally we should only
3914 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3915 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3916 	 * that with namei as it stands.
3917 	 *
3918 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3919 	 * under us. The real fix is to get the locks in the right
3920 	 * order and do the lookups in the right places, but that's a
3921 	 * major rototill.
3922 	 *
3923 	 * Note: this logic (as well as this whole function) is cloned
3924 	 * in nfs_serv.c. Proceed accordingly.
3925 	 */
3926 	vrele(fvp);
3927 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3928 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3929 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3930 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3931 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3932 		error = EINVAL;
3933 		VFS_RENAMELOCK_EXIT(fs);
3934 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3935 		vrele(fromnd.ni_dvp);
3936 		goto out1;
3937 	}
3938 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3939 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0);
3940 	if (error) {
3941 		VOP_UNLOCK(fromnd.ni_dvp);
3942 		VFS_RENAMELOCK_EXIT(fs);
3943 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3944 		vrele(fromnd.ni_dvp);
3945 		goto out1;
3946 	}
3947 	VOP_UNLOCK(fromnd.ni_vp);
3948 	if (fromnd.ni_dvp != fromnd.ni_vp)
3949 		VOP_UNLOCK(fromnd.ni_dvp);
3950 	fvp = fromnd.ni_vp;
3951 
3952 	NDINIT(&tond, RENAME,
3953 	    LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT
3954 	      | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0),
3955 	    topb);
3956 	if ((error = namei(&tond)) != 0) {
3957 		VFS_RENAMELOCK_EXIT(fs);
3958 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3959 		vrele(fromnd.ni_dvp);
3960 		vrele(fvp);
3961 		goto out1;
3962 	}
3963 	tdvp = tond.ni_dvp;
3964 	tvp = tond.ni_vp;
3965 
3966 	if (tvp != NULL) {
3967 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3968 			error = ENOTDIR;
3969 			goto out;
3970 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3971 			error = EISDIR;
3972 			goto out;
3973 		}
3974 	}
3975 
3976 	if (fvp == tdvp)
3977 		error = EINVAL;
3978 
3979 	/*
3980 	 * Source and destination refer to the same object.
3981 	 */
3982 	if (fvp == tvp) {
3983 		if (retain)
3984 			error = -1;
3985 		else if (fromnd.ni_dvp == tdvp &&
3986 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3987 		    !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
3988 		          fromnd.ni_cnd.cn_namelen))
3989 			error = -1;
3990 	}
3991 	/*
3992 	 * Prevent cross-mount operation.
3993 	 */
3994 	if (error == 0) {
3995 		if (tond.ni_dvp->v_mount != fromnd.ni_dvp->v_mount) {
3996 			error = EXDEV;
3997 		}
3998 	}
3999 #if NVERIEXEC > 0
4000 	if (!error) {
4001 		char *f1, *f2;
4002 		size_t f1_len;
4003 		size_t f2_len;
4004 
4005 		f1_len = fromnd.ni_cnd.cn_namelen + 1;
4006 		f1 = kmem_alloc(f1_len, KM_SLEEP);
4007 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
4008 
4009 		f2_len = tond.ni_cnd.cn_namelen + 1;
4010 		f2 = kmem_alloc(f2_len, KM_SLEEP);
4011 		strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
4012 
4013 		error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2);
4014 
4015 		kmem_free(f1, f1_len);
4016 		kmem_free(f2, f2_len);
4017 	}
4018 #endif /* NVERIEXEC > 0 */
4019 
4020 out:
4021 	if (!error) {
4022 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
4023 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
4024 		VFS_RENAMELOCK_EXIT(fs);
4025 	} else {
4026 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
4027 		if (tdvp == tvp)
4028 			vrele(tdvp);
4029 		else
4030 			vput(tdvp);
4031 		if (tvp)
4032 			vput(tvp);
4033 		VFS_RENAMELOCK_EXIT(fs);
4034 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
4035 		vrele(fromnd.ni_dvp);
4036 		vrele(fvp);
4037 	}
4038 out1:
4039 	pathbuf_destroy(frompb);
4040 	pathbuf_destroy(topb);
4041 	return (error == -1 ? 0 : error);
4042 }
4043 
4044 /*
4045  * Make a directory file.
4046  */
4047 /* ARGSUSED */
4048 int
4049 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
4050 {
4051 	/* {
4052 		syscallarg(const char *) path;
4053 		syscallarg(int) mode;
4054 	} */
4055 
4056 	return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE);
4057 }
4058 
4059 int
4060 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap,
4061     register_t *retval)
4062 {
4063 	/* {
4064 		syscallarg(int) fd;
4065 		syscallarg(const char *) path;
4066 		syscallarg(int) mode;
4067 	} */
4068 
4069 	return ENOSYS;
4070 }
4071 
4072 
4073 int
4074 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
4075 {
4076 	struct proc *p = curlwp->l_proc;
4077 	struct vnode *vp;
4078 	struct vattr vattr;
4079 	int error;
4080 	struct pathbuf *pb;
4081 	struct nameidata nd;
4082 
4083 	/* XXX bollocks, should pass in a pathbuf */
4084 	error = pathbuf_maybe_copyin(path, seg, &pb);
4085 	if (error) {
4086 		return error;
4087 	}
4088 
4089 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb);
4090 	if ((error = namei(&nd)) != 0) {
4091 		pathbuf_destroy(pb);
4092 		return (error);
4093 	}
4094 	vp = nd.ni_vp;
4095 	if (vp != NULL) {
4096 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
4097 		if (nd.ni_dvp == vp)
4098 			vrele(nd.ni_dvp);
4099 		else
4100 			vput(nd.ni_dvp);
4101 		vrele(vp);
4102 		pathbuf_destroy(pb);
4103 		return (EEXIST);
4104 	}
4105 	vattr_null(&vattr);
4106 	vattr.va_type = VDIR;
4107 	/* We will read cwdi->cwdi_cmask unlocked. */
4108 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
4109 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
4110 	if (!error)
4111 		vput(nd.ni_vp);
4112 	pathbuf_destroy(pb);
4113 	return (error);
4114 }
4115 
4116 /*
4117  * Remove a directory file.
4118  */
4119 /* ARGSUSED */
4120 int
4121 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
4122 {
4123 	/* {
4124 		syscallarg(const char *) path;
4125 	} */
4126 	struct vnode *vp;
4127 	int error;
4128 	struct pathbuf *pb;
4129 	struct nameidata nd;
4130 
4131 	error = pathbuf_copyin(SCARG(uap, path), &pb);
4132 	if (error) {
4133 		return error;
4134 	}
4135 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
4136 	if ((error = namei(&nd)) != 0) {
4137 		pathbuf_destroy(pb);
4138 		return error;
4139 	}
4140 	vp = nd.ni_vp;
4141 	if (vp->v_type != VDIR) {
4142 		error = ENOTDIR;
4143 		goto out;
4144 	}
4145 	/*
4146 	 * No rmdir "." please.
4147 	 */
4148 	if (nd.ni_dvp == vp) {
4149 		error = EINVAL;
4150 		goto out;
4151 	}
4152 	/*
4153 	 * The root of a mounted filesystem cannot be deleted.
4154 	 */
4155 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
4156 		error = EBUSY;
4157 		goto out;
4158 	}
4159 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
4160 	pathbuf_destroy(pb);
4161 	return (error);
4162 
4163 out:
4164 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
4165 	if (nd.ni_dvp == vp)
4166 		vrele(nd.ni_dvp);
4167 	else
4168 		vput(nd.ni_dvp);
4169 	vput(vp);
4170 	pathbuf_destroy(pb);
4171 	return (error);
4172 }
4173 
4174 /*
4175  * Read a block of directory entries in a file system independent format.
4176  */
4177 int
4178 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
4179 {
4180 	/* {
4181 		syscallarg(int) fd;
4182 		syscallarg(char *) buf;
4183 		syscallarg(size_t) count;
4184 	} */
4185 	file_t *fp;
4186 	int error, done;
4187 
4188 	/* fd_getvnode() will use the descriptor for us */
4189 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
4190 		return (error);
4191 	if ((fp->f_flag & FREAD) == 0) {
4192 		error = EBADF;
4193 		goto out;
4194 	}
4195 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
4196 			SCARG(uap, count), &done, l, 0, 0);
4197 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
4198 	*retval = done;
4199  out:
4200 	fd_putfile(SCARG(uap, fd));
4201 	return (error);
4202 }
4203 
4204 /*
4205  * Set the mode mask for creation of filesystem nodes.
4206  */
4207 int
4208 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
4209 {
4210 	/* {
4211 		syscallarg(mode_t) newmask;
4212 	} */
4213 	struct proc *p = l->l_proc;
4214 	struct cwdinfo *cwdi;
4215 
4216 	/*
4217 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
4218 	 * important is that we serialize changes to the mask.  The
4219 	 * rw_exit() will issue a write memory barrier on our behalf,
4220 	 * and force the changes out to other CPUs (as it must use an
4221 	 * atomic operation, draining the local CPU's store buffers).
4222 	 */
4223 	cwdi = p->p_cwdi;
4224 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
4225 	*retval = cwdi->cwdi_cmask;
4226 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
4227 	rw_exit(&cwdi->cwdi_lock);
4228 
4229 	return (0);
4230 }
4231 
4232 int
4233 dorevoke(struct vnode *vp, kauth_cred_t cred)
4234 {
4235 	struct vattr vattr;
4236 	int error, fs_decision;
4237 
4238 	vn_lock(vp, LK_SHARED | LK_RETRY);
4239 	error = VOP_GETATTR(vp, &vattr, cred);
4240 	VOP_UNLOCK(vp);
4241 	if (error != 0)
4242 		return error;
4243 	fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM;
4244 	error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL,
4245 	    fs_decision);
4246 	if (!error)
4247 		VOP_REVOKE(vp, REVOKEALL);
4248 	return (error);
4249 }
4250 
4251 /*
4252  * Void all references to file by ripping underlying filesystem
4253  * away from vnode.
4254  */
4255 /* ARGSUSED */
4256 int
4257 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
4258 {
4259 	/* {
4260 		syscallarg(const char *) path;
4261 	} */
4262 	struct vnode *vp;
4263 	int error;
4264 
4265 	error = namei_simple_user(SCARG(uap, path),
4266 				NSM_FOLLOW_TRYEMULROOT, &vp);
4267 	if (error != 0)
4268 		return (error);
4269 	error = dorevoke(vp, l->l_cred);
4270 	vrele(vp);
4271 	return (error);
4272 }
4273