xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision c505c4429840c353a86d4eb53b5e2bfc0092264e)
1 /*	$NetBSD: vfs_syscalls.c,v 1.407 2010/06/30 15:44:54 pooka 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 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.407 2010/06/30 15:44:54 pooka Exp $");
70 
71 #ifdef _KERNEL_OPT
72 #include "opt_fileassoc.h"
73 #include "veriexec.h"
74 #endif
75 
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/namei.h>
79 #include <sys/filedesc.h>
80 #include <sys/kernel.h>
81 #include <sys/file.h>
82 #include <sys/stat.h>
83 #include <sys/vnode.h>
84 #include <sys/mount.h>
85 #include <sys/proc.h>
86 #include <sys/uio.h>
87 #include <sys/kmem.h>
88 #include <sys/dirent.h>
89 #include <sys/sysctl.h>
90 #include <sys/syscallargs.h>
91 #include <sys/vfs_syscalls.h>
92 #include <sys/ktrace.h>
93 #ifdef FILEASSOC
94 #include <sys/fileassoc.h>
95 #endif /* FILEASSOC */
96 #include <sys/verified_exec.h>
97 #include <sys/kauth.h>
98 #include <sys/atomic.h>
99 #include <sys/module.h>
100 #include <sys/buf.h>
101 
102 #include <miscfs/genfs/genfs.h>
103 #include <miscfs/syncfs/syncfs.h>
104 #include <miscfs/specfs/specdev.h>
105 
106 #include <nfs/rpcv2.h>
107 #include <nfs/nfsproto.h>
108 #include <nfs/nfs.h>
109 #include <nfs/nfs_var.h>
110 
111 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
112 
113 static int change_flags(struct vnode *, u_long, struct lwp *);
114 static int change_mode(struct vnode *, int, struct lwp *l);
115 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
116 
117 void checkdirs(struct vnode *);
118 
119 /*
120  * Virtual File System System Calls
121  */
122 
123 /*
124  * Mount a file system.
125  */
126 
127 /*
128  * This table is used to maintain compatibility with 4.3BSD
129  * and NetBSD 0.9 mount syscalls - and possibly other systems.
130  * Note, the order is important!
131  *
132  * Do not modify this table. It should only contain filesystems
133  * supported by NetBSD 0.9 and 4.3BSD.
134  */
135 const char * const mountcompatnames[] = {
136 	NULL,		/* 0 = MOUNT_NONE */
137 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
138 	MOUNT_NFS,	/* 2 */
139 	MOUNT_MFS,	/* 3 */
140 	MOUNT_MSDOS,	/* 4 */
141 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
142 	MOUNT_FDESC,	/* 6 */
143 	MOUNT_KERNFS,	/* 7 */
144 	NULL,		/* 8 = MOUNT_DEVFS */
145 	MOUNT_AFS,	/* 9 */
146 };
147 const int nmountcompatnames = sizeof(mountcompatnames) /
148     sizeof(mountcompatnames[0]);
149 
150 static int
151 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
152     void *data, size_t *data_len)
153 {
154 	struct mount *mp;
155 	int error = 0, saved_flags;
156 
157 	mp = vp->v_mount;
158 	saved_flags = mp->mnt_flag;
159 
160 	/* We can operate only on VV_ROOT nodes. */
161 	if ((vp->v_vflag & VV_ROOT) == 0) {
162 		error = EINVAL;
163 		goto out;
164 	}
165 
166 	/*
167 	 * We only allow the filesystem to be reloaded if it
168 	 * is currently mounted read-only.  Additionally, we
169 	 * prevent read-write to read-only downgrades.
170 	 */
171 	if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
172 	    (mp->mnt_flag & MNT_RDONLY) == 0) {
173 		error = EOPNOTSUPP;	/* Needs translation */
174 		goto out;
175 	}
176 
177 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
178 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
179 	if (error)
180 		goto out;
181 
182 	if (vfs_busy(mp, NULL)) {
183 		error = EPERM;
184 		goto out;
185 	}
186 
187 	mutex_enter(&mp->mnt_updating);
188 
189 	mp->mnt_flag &= ~MNT_OP_FLAGS;
190 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
191 
192 	/*
193 	 * Set the mount level flags.
194 	 */
195 	if (flags & MNT_RDONLY)
196 		mp->mnt_flag |= MNT_RDONLY;
197 	else if (mp->mnt_flag & MNT_RDONLY)
198 		mp->mnt_iflag |= IMNT_WANTRDWR;
199 	mp->mnt_flag &=
200 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
201 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
202 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
203 	    MNT_LOG);
204 	mp->mnt_flag |= flags &
205 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
206 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
207 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
208 	    MNT_LOG | MNT_IGNORE);
209 
210 	error = VFS_MOUNT(mp, path, data, data_len);
211 
212 	if (error && data != NULL) {
213 		int error2;
214 
215 		/*
216 		 * Update failed; let's try and see if it was an
217 		 * export request.  For compat with 3.0 and earlier.
218 		 */
219 		error2 = vfs_hooks_reexport(mp, path, data);
220 
221 		/*
222 		 * Only update error code if the export request was
223 		 * understood but some problem occurred while
224 		 * processing it.
225 		 */
226 		if (error2 != EJUSTRETURN)
227 			error = error2;
228 	}
229 
230 	if (mp->mnt_iflag & IMNT_WANTRDWR)
231 		mp->mnt_flag &= ~MNT_RDONLY;
232 	if (error)
233 		mp->mnt_flag = saved_flags;
234 	mp->mnt_flag &= ~MNT_OP_FLAGS;
235 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
236 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
237 		if (mp->mnt_syncer == NULL)
238 			error = vfs_allocate_syncvnode(mp);
239 	} else {
240 		if (mp->mnt_syncer != NULL)
241 			vfs_deallocate_syncvnode(mp);
242 	}
243 	mutex_exit(&mp->mnt_updating);
244 	vfs_unbusy(mp, false, NULL);
245 
246  out:
247 	return (error);
248 }
249 
250 static int
251 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
252 {
253 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
254 	int error;
255 
256 	/* Copy file-system type from userspace.  */
257 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
258 	if (error) {
259 		/*
260 		 * Historically, filesystem types were identified by numbers.
261 		 * If we get an integer for the filesystem type instead of a
262 		 * string, we check to see if it matches one of the historic
263 		 * filesystem types.
264 		 */
265 		u_long fsindex = (u_long)fstype;
266 		if (fsindex >= nmountcompatnames ||
267 		    mountcompatnames[fsindex] == NULL)
268 			return ENODEV;
269 		strlcpy(fstypename, mountcompatnames[fsindex],
270 		    sizeof(fstypename));
271 	}
272 
273 	/* Accept `ufs' as an alias for `ffs', for compatibility. */
274 	if (strcmp(fstypename, "ufs") == 0)
275 		fstypename[0] = 'f';
276 
277 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
278 		return 0;
279 
280 	/* If we can autoload a vfs module, try again */
281 	mutex_enter(&module_lock);
282 	(void)module_autoload(fstypename, MODULE_CLASS_VFS);
283 	mutex_exit(&module_lock);
284 
285 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
286 		return 0;
287 
288 	return ENODEV;
289 }
290 
291 static int
292 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
293     const char *path, int flags, void *data, size_t *data_len)
294 {
295 	struct mount *mp;
296 	struct vnode *vp = *vpp;
297 	struct vattr va;
298 	struct nameidata nd;
299 	int error;
300 
301 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
302 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
303 	if (error) {
304 		vfs_delref(vfsops);
305 		return error;
306 	}
307 
308 	/* Can't make a non-dir a mount-point (from here anyway). */
309 	if (vp->v_type != VDIR) {
310 		vfs_delref(vfsops);
311 		return ENOTDIR;
312 	}
313 
314 	/*
315 	 * If the user is not root, ensure that they own the directory
316 	 * onto which we are attempting to mount.
317 	 */
318 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
319 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
320 	    (error = kauth_authorize_generic(l->l_cred,
321 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
322 		vfs_delref(vfsops);
323 		return error;
324 	}
325 
326 	if (flags & MNT_EXPORTED) {
327 		vfs_delref(vfsops);
328 		return EINVAL;
329 	}
330 
331 	if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) {
332 		vfs_delref(vfsops);
333 		return ENOMEM;
334 	}
335 
336 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
337 
338 	/*
339 	 * The underlying file system may refuse the mount for
340 	 * various reasons.  Allow the user to force it to happen.
341 	 *
342 	 * Set the mount level flags.
343 	 */
344 	mp->mnt_flag = flags &
345 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
346 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
347 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
348 	    MNT_LOG | MNT_IGNORE | MNT_RDONLY);
349 
350 	mutex_enter(&mp->mnt_updating);
351 	error = VFS_MOUNT(mp, path, data, data_len);
352 	mp->mnt_flag &= ~MNT_OP_FLAGS;
353 
354 	if (error != 0)
355 		goto err_unmounted;
356 
357 	/*
358 	 * Validate and prepare the mount point.
359 	 */
360 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT,
361 	    UIO_USERSPACE, path);
362 	error = namei(&nd);
363 	if (error != 0) {
364 		goto err_mounted;
365 	}
366 	if (nd.ni_vp != vp) {
367 		vput(nd.ni_vp);
368 		error = EINVAL;
369 		goto err_mounted;
370 	}
371 	if (vp->v_mountedhere != NULL) {
372 		vput(nd.ni_vp);
373 		error = EBUSY;
374 		goto err_mounted;
375 	}
376 	error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0);
377 	if (error != 0) {
378 		vput(nd.ni_vp);
379 		goto err_mounted;
380 	}
381 
382 	/*
383 	 * Put the new filesystem on the mount list after root.
384 	 */
385 	cache_purge(vp);
386 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
387 
388 	mutex_enter(&mountlist_lock);
389 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
390 	mutex_exit(&mountlist_lock);
391 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
392 		error = vfs_allocate_syncvnode(mp);
393 	if (error == 0)
394 		vp->v_mountedhere = mp;
395 	vput(nd.ni_vp);
396 	if (error != 0)
397 		goto err_onmountlist;
398 
399 	checkdirs(vp);
400 	mutex_exit(&mp->mnt_updating);
401 
402 	/* Hold an additional reference to the mount across VFS_START(). */
403 	vfs_unbusy(mp, true, NULL);
404 	(void) VFS_STATVFS(mp, &mp->mnt_stat);
405 	error = VFS_START(mp, 0);
406 	if (error)
407 		vrele(vp);
408 	/* Drop reference held for VFS_START(). */
409 	vfs_destroy(mp);
410 	*vpp = NULL;
411 	return error;
412 
413 err_onmountlist:
414 	mutex_enter(&mountlist_lock);
415 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
416 	mp->mnt_iflag |= IMNT_GONE;
417 	mutex_exit(&mountlist_lock);
418 
419 err_mounted:
420 	if (VFS_UNMOUNT(mp, MNT_FORCE) != 0)
421 		panic("Unmounting fresh file system failed");
422 
423 err_unmounted:
424 	vp->v_mountedhere = NULL;
425 	mutex_exit(&mp->mnt_updating);
426 	vfs_unbusy(mp, false, NULL);
427 	vfs_destroy(mp);
428 
429 	return error;
430 }
431 
432 static int
433 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
434     void *data, size_t *data_len)
435 {
436 	struct mount *mp;
437 	int error;
438 
439 	/* If MNT_GETARGS is specified, it should be the only flag. */
440 	if (flags & ~MNT_GETARGS)
441 		return EINVAL;
442 
443 	mp = vp->v_mount;
444 
445 	/* XXX: probably some notion of "can see" here if we want isolation. */
446 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
447 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
448 	if (error)
449 		return error;
450 
451 	if ((vp->v_vflag & VV_ROOT) == 0)
452 		return EINVAL;
453 
454 	if (vfs_busy(mp, NULL))
455 		return EPERM;
456 
457 	mutex_enter(&mp->mnt_updating);
458 	mp->mnt_flag &= ~MNT_OP_FLAGS;
459 	mp->mnt_flag |= MNT_GETARGS;
460 	error = VFS_MOUNT(mp, path, data, data_len);
461 	mp->mnt_flag &= ~MNT_OP_FLAGS;
462 	mutex_exit(&mp->mnt_updating);
463 
464 	vfs_unbusy(mp, false, NULL);
465 	return (error);
466 }
467 
468 int
469 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
470 {
471 	/* {
472 		syscallarg(const char *) type;
473 		syscallarg(const char *) path;
474 		syscallarg(int) flags;
475 		syscallarg(void *) data;
476 		syscallarg(size_t) data_len;
477 	} */
478 
479 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
480 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
481 	    SCARG(uap, data_len), retval);
482 }
483 
484 int
485 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
486     const char *path, int flags, void *data, enum uio_seg data_seg,
487     size_t data_len, register_t *retval)
488 {
489 	struct vnode *vp;
490 	void *data_buf = data;
491 	bool vfsopsrele = false;
492 	int error;
493 
494 	/* XXX: The calling convention of this routine is totally bizarre */
495 	if (vfsops)
496 		vfsopsrele = true;
497 
498 	/*
499 	 * Get vnode to be covered
500 	 */
501 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
502 	if (error != 0) {
503 		vp = NULL;
504 		goto done;
505 	}
506 
507 	if (vfsops == NULL) {
508 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
509 			vfsops = vp->v_mount->mnt_op;
510 		} else {
511 			/* 'type' is userspace */
512 			error = mount_get_vfsops(type, &vfsops);
513 			if (error != 0)
514 				goto done;
515 			vfsopsrele = true;
516 		}
517 	}
518 
519 	if (data != NULL && data_seg == UIO_USERSPACE) {
520 		if (data_len == 0) {
521 			/* No length supplied, use default for filesystem */
522 			data_len = vfsops->vfs_min_mount_data;
523 			if (data_len > VFS_MAX_MOUNT_DATA) {
524 				error = EINVAL;
525 				goto done;
526 			}
527 			/*
528 			 * Hopefully a longer buffer won't make copyin() fail.
529 			 * For compatibility with 3.0 and earlier.
530 			 */
531 			if (flags & MNT_UPDATE
532 			    && data_len < sizeof (struct mnt_export_args30))
533 				data_len = sizeof (struct mnt_export_args30);
534 		}
535 		data_buf = kmem_alloc(data_len, KM_SLEEP);
536 
537 		/* NFS needs the buffer even for mnt_getargs .... */
538 		error = copyin(data, data_buf, data_len);
539 		if (error != 0)
540 			goto done;
541 	}
542 
543 	if (flags & MNT_GETARGS) {
544 		if (data_len == 0) {
545 			error = EINVAL;
546 			goto done;
547 		}
548 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
549 		if (error != 0)
550 			goto done;
551 		if (data_seg == UIO_USERSPACE)
552 			error = copyout(data_buf, data, data_len);
553 		*retval = data_len;
554 	} else if (flags & MNT_UPDATE) {
555 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
556 	} else {
557 		/* Locking is handled internally in mount_domount(). */
558 		KASSERT(vfsopsrele == true);
559 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
560 		    &data_len);
561 		vfsopsrele = false;
562 	}
563 
564     done:
565 	if (vfsopsrele)
566 		vfs_delref(vfsops);
567     	if (vp != NULL) {
568 	    	vrele(vp);
569 	}
570 	if (data_buf != data)
571 		kmem_free(data_buf, data_len);
572 	return (error);
573 }
574 
575 /*
576  * Scan all active processes to see if any of them have a current
577  * or root directory onto which the new filesystem has just been
578  * mounted. If so, replace them with the new mount point.
579  */
580 void
581 checkdirs(struct vnode *olddp)
582 {
583 	struct cwdinfo *cwdi;
584 	struct vnode *newdp, *rele1, *rele2;
585 	struct proc *p;
586 	bool retry;
587 
588 	if (olddp->v_usecount == 1)
589 		return;
590 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
591 		panic("mount: lost mount");
592 
593 	do {
594 		retry = false;
595 		mutex_enter(proc_lock);
596 		PROCLIST_FOREACH(p, &allproc) {
597 			if ((cwdi = p->p_cwdi) == NULL)
598 				continue;
599 			/*
600 			 * Can't change to the old directory any more,
601 			 * so even if we see a stale value it's not a
602 			 * problem.
603 			 */
604 			if (cwdi->cwdi_cdir != olddp &&
605 			    cwdi->cwdi_rdir != olddp)
606 			    	continue;
607 			retry = true;
608 			rele1 = NULL;
609 			rele2 = NULL;
610 			atomic_inc_uint(&cwdi->cwdi_refcnt);
611 			mutex_exit(proc_lock);
612 			rw_enter(&cwdi->cwdi_lock, RW_WRITER);
613 			if (cwdi->cwdi_cdir == olddp) {
614 				rele1 = cwdi->cwdi_cdir;
615 				vref(newdp);
616 				cwdi->cwdi_cdir = newdp;
617 			}
618 			if (cwdi->cwdi_rdir == olddp) {
619 				rele2 = cwdi->cwdi_rdir;
620 				vref(newdp);
621 				cwdi->cwdi_rdir = newdp;
622 			}
623 			rw_exit(&cwdi->cwdi_lock);
624 			cwdfree(cwdi);
625 			if (rele1 != NULL)
626 				vrele(rele1);
627 			if (rele2 != NULL)
628 				vrele(rele2);
629 			mutex_enter(proc_lock);
630 			break;
631 		}
632 		mutex_exit(proc_lock);
633 	} while (retry);
634 
635 	if (rootvnode == olddp) {
636 		vrele(rootvnode);
637 		vref(newdp);
638 		rootvnode = newdp;
639 	}
640 	vput(newdp);
641 }
642 
643 /*
644  * Unmount a file system.
645  *
646  * Note: unmount takes a path to the vnode mounted on as argument,
647  * not special file (as before).
648  */
649 /* ARGSUSED */
650 int
651 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
652 {
653 	/* {
654 		syscallarg(const char *) path;
655 		syscallarg(int) flags;
656 	} */
657 	struct vnode *vp;
658 	struct mount *mp;
659 	int error;
660 	struct nameidata nd;
661 
662 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
663 	    SCARG(uap, path));
664 	if ((error = namei(&nd)) != 0)
665 		return (error);
666 	vp = nd.ni_vp;
667 	mp = vp->v_mount;
668 	atomic_inc_uint(&mp->mnt_refcnt);
669 	VOP_UNLOCK(vp);
670 
671 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
672 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
673 	if (error) {
674 		vrele(vp);
675 		vfs_destroy(mp);
676 		return (error);
677 	}
678 
679 	/*
680 	 * Don't allow unmounting the root file system.
681 	 */
682 	if (mp->mnt_flag & MNT_ROOTFS) {
683 		vrele(vp);
684 		vfs_destroy(mp);
685 		return (EINVAL);
686 	}
687 
688 	/*
689 	 * Must be the root of the filesystem
690 	 */
691 	if ((vp->v_vflag & VV_ROOT) == 0) {
692 		vrele(vp);
693 		vfs_destroy(mp);
694 		return (EINVAL);
695 	}
696 
697 	vrele(vp);
698 	error = dounmount(mp, SCARG(uap, flags), l);
699 	vfs_destroy(mp);
700 	return error;
701 }
702 
703 /*
704  * Do the actual file system unmount.  File system is assumed to have
705  * been locked by the caller.
706  *
707  * => Caller hold reference to the mount, explicitly for dounmount().
708  */
709 int
710 dounmount(struct mount *mp, int flags, struct lwp *l)
711 {
712 	struct vnode *coveredvp;
713 	int error;
714 	int async;
715 	int used_syncer;
716 
717 #if NVERIEXEC > 0
718 	error = veriexec_unmountchk(mp);
719 	if (error)
720 		return (error);
721 #endif /* NVERIEXEC > 0 */
722 
723 	/*
724 	 * XXX Freeze syncer.  Must do this before locking the
725 	 * mount point.  See dounmount() for details.
726 	 */
727 	mutex_enter(&syncer_mutex);
728 	rw_enter(&mp->mnt_unmounting, RW_WRITER);
729 	if ((mp->mnt_iflag & IMNT_GONE) != 0) {
730 		rw_exit(&mp->mnt_unmounting);
731 		mutex_exit(&syncer_mutex);
732 		return ENOENT;
733 	}
734 
735 	used_syncer = (mp->mnt_syncer != NULL);
736 
737 	/*
738 	 * XXX Syncer must be frozen when we get here.  This should really
739 	 * be done on a per-mountpoint basis, but the syncer doesn't work
740 	 * like that.
741 	 *
742 	 * The caller of dounmount() must acquire syncer_mutex because
743 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
744 	 * order, and we must preserve that order to avoid deadlock.
745 	 *
746 	 * So, if the file system did not use the syncer, now is
747 	 * the time to release the syncer_mutex.
748 	 */
749 	if (used_syncer == 0)
750 		mutex_exit(&syncer_mutex);
751 
752 	mp->mnt_iflag |= IMNT_UNMOUNT;
753 	async = mp->mnt_flag & MNT_ASYNC;
754 	mp->mnt_flag &= ~MNT_ASYNC;
755 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
756 	if (mp->mnt_syncer != NULL)
757 		vfs_deallocate_syncvnode(mp);
758 	error = 0;
759 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
760 		error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
761 	}
762 	vfs_scrubvnlist(mp);
763 	if (error == 0 || (flags & MNT_FORCE))
764 		error = VFS_UNMOUNT(mp, flags);
765 	if (error) {
766 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
767 			(void) vfs_allocate_syncvnode(mp);
768 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
769 		mp->mnt_flag |= async;
770 		rw_exit(&mp->mnt_unmounting);
771 		if (used_syncer)
772 			mutex_exit(&syncer_mutex);
773 		return (error);
774 	}
775 	vfs_scrubvnlist(mp);
776 	mutex_enter(&mountlist_lock);
777 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
778 		coveredvp->v_mountedhere = NULL;
779 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
780 	mp->mnt_iflag |= IMNT_GONE;
781 	mutex_exit(&mountlist_lock);
782 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
783 		panic("unmount: dangling vnode");
784 	if (used_syncer)
785 		mutex_exit(&syncer_mutex);
786 	vfs_hooks_unmount(mp);
787 	rw_exit(&mp->mnt_unmounting);
788 	vfs_destroy(mp);	/* reference from mount() */
789 	if (coveredvp != NULLVP)
790 		vrele(coveredvp);
791 	return (0);
792 }
793 
794 /*
795  * Sync each mounted filesystem.
796  */
797 #ifdef DEBUG
798 int syncprt = 0;
799 struct ctldebug debug0 = { "syncprt", &syncprt };
800 #endif
801 
802 /* ARGSUSED */
803 int
804 sys_sync(struct lwp *l, const void *v, register_t *retval)
805 {
806 	struct mount *mp, *nmp;
807 	int asyncflag;
808 
809 	if (l == NULL)
810 		l = &lwp0;
811 
812 	mutex_enter(&mountlist_lock);
813 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
814 	     mp = nmp) {
815 		if (vfs_busy(mp, &nmp)) {
816 			continue;
817 		}
818 		mutex_enter(&mp->mnt_updating);
819 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
820 			asyncflag = mp->mnt_flag & MNT_ASYNC;
821 			mp->mnt_flag &= ~MNT_ASYNC;
822 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
823 			if (asyncflag)
824 				 mp->mnt_flag |= MNT_ASYNC;
825 		}
826 		mutex_exit(&mp->mnt_updating);
827 		vfs_unbusy(mp, false, &nmp);
828 	}
829 	mutex_exit(&mountlist_lock);
830 #ifdef DEBUG
831 	if (syncprt)
832 		vfs_bufstats();
833 #endif /* DEBUG */
834 	return (0);
835 }
836 
837 /*
838  * Change filesystem quotas.
839  */
840 /* ARGSUSED */
841 int
842 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
843 {
844 	/* {
845 		syscallarg(const char *) path;
846 		syscallarg(int) cmd;
847 		syscallarg(int) uid;
848 		syscallarg(void *) arg;
849 	} */
850 	struct mount *mp;
851 	int error;
852 	struct vnode *vp;
853 
854 	error = namei_simple_user(SCARG(uap, path),
855 				NSM_FOLLOW_TRYEMULROOT, &vp);
856 	if (error != 0)
857 		return (error);
858 	mp = vp->v_mount;
859 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
860 	    SCARG(uap, arg));
861 	vrele(vp);
862 	return (error);
863 }
864 
865 int
866 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
867     int root)
868 {
869 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
870 	int error = 0;
871 
872 	/*
873 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
874 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
875 	 * overrides MNT_NOWAIT.
876 	 */
877 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
878 	    (flags != MNT_WAIT && flags != 0)) {
879 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
880 		goto done;
881 	}
882 
883 	/* Get the filesystem stats now */
884 	memset(sp, 0, sizeof(*sp));
885 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
886 		return error;
887 	}
888 
889 	if (cwdi->cwdi_rdir == NULL)
890 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
891 done:
892 	if (cwdi->cwdi_rdir != NULL) {
893 		size_t len;
894 		char *bp;
895 		char c;
896 		char *path = PNBUF_GET();
897 
898 		bp = path + MAXPATHLEN;
899 		*--bp = '\0';
900 		rw_enter(&cwdi->cwdi_lock, RW_READER);
901 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
902 		    MAXPATHLEN / 2, 0, l);
903 		rw_exit(&cwdi->cwdi_lock);
904 		if (error) {
905 			PNBUF_PUT(path);
906 			return error;
907 		}
908 		len = strlen(bp);
909 		if (len != 1) {
910 			/*
911 			 * for mount points that are below our root, we can see
912 			 * them, so we fix up the pathname and return them. The
913 			 * rest we cannot see, so we don't allow viewing the
914 			 * data.
915 			 */
916 			if (strncmp(bp, sp->f_mntonname, len) == 0 &&
917 			    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
918 				(void)strlcpy(sp->f_mntonname,
919 				    c == '\0' ? "/" : &sp->f_mntonname[len],
920 				    sizeof(sp->f_mntonname));
921 			} else {
922 				if (root)
923 					(void)strlcpy(sp->f_mntonname, "/",
924 					    sizeof(sp->f_mntonname));
925 				else
926 					error = EPERM;
927 			}
928 		}
929 		PNBUF_PUT(path);
930 	}
931 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
932 	return error;
933 }
934 
935 /*
936  * Get filesystem statistics by path.
937  */
938 int
939 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
940 {
941 	struct mount *mp;
942 	int error;
943 	struct vnode *vp;
944 
945 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
946 	if (error != 0)
947 		return error;
948 	mp = vp->v_mount;
949 	error = dostatvfs(mp, sb, l, flags, 1);
950 	vrele(vp);
951 	return error;
952 }
953 
954 /* ARGSUSED */
955 int
956 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
957 {
958 	/* {
959 		syscallarg(const char *) path;
960 		syscallarg(struct statvfs *) buf;
961 		syscallarg(int) flags;
962 	} */
963 	struct statvfs *sb;
964 	int error;
965 
966 	sb = STATVFSBUF_GET();
967 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
968 	if (error == 0)
969 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
970 	STATVFSBUF_PUT(sb);
971 	return error;
972 }
973 
974 /*
975  * Get filesystem statistics by fd.
976  */
977 int
978 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
979 {
980 	file_t *fp;
981 	struct mount *mp;
982 	int error;
983 
984 	/* fd_getvnode() will use the descriptor for us */
985 	if ((error = fd_getvnode(fd, &fp)) != 0)
986 		return (error);
987 	mp = ((struct vnode *)fp->f_data)->v_mount;
988 	error = dostatvfs(mp, sb, curlwp, flags, 1);
989 	fd_putfile(fd);
990 	return error;
991 }
992 
993 /* ARGSUSED */
994 int
995 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
996 {
997 	/* {
998 		syscallarg(int) fd;
999 		syscallarg(struct statvfs *) buf;
1000 		syscallarg(int) flags;
1001 	} */
1002 	struct statvfs *sb;
1003 	int error;
1004 
1005 	sb = STATVFSBUF_GET();
1006 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1007 	if (error == 0)
1008 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1009 	STATVFSBUF_PUT(sb);
1010 	return error;
1011 }
1012 
1013 
1014 /*
1015  * Get statistics on all filesystems.
1016  */
1017 int
1018 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1019     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1020     register_t *retval)
1021 {
1022 	int root = 0;
1023 	struct proc *p = l->l_proc;
1024 	struct mount *mp, *nmp;
1025 	struct statvfs *sb;
1026 	size_t count, maxcount;
1027 	int error = 0;
1028 
1029 	sb = STATVFSBUF_GET();
1030 	maxcount = bufsize / entry_sz;
1031 	mutex_enter(&mountlist_lock);
1032 	count = 0;
1033 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1034 	     mp = nmp) {
1035 		if (vfs_busy(mp, &nmp)) {
1036 			continue;
1037 		}
1038 		if (sfsp && count < maxcount) {
1039 			error = dostatvfs(mp, sb, l, flags, 0);
1040 			if (error) {
1041 				vfs_unbusy(mp, false, &nmp);
1042 				error = 0;
1043 				continue;
1044 			}
1045 			error = copyfn(sb, sfsp, entry_sz);
1046 			if (error) {
1047 				vfs_unbusy(mp, false, NULL);
1048 				goto out;
1049 			}
1050 			sfsp = (char *)sfsp + entry_sz;
1051 			root |= strcmp(sb->f_mntonname, "/") == 0;
1052 		}
1053 		count++;
1054 		vfs_unbusy(mp, false, &nmp);
1055 	}
1056 	mutex_exit(&mountlist_lock);
1057 
1058 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
1059 		/*
1060 		 * fake a root entry
1061 		 */
1062 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1063 		    sb, l, flags, 1);
1064 		if (error != 0)
1065 			goto out;
1066 		if (sfsp) {
1067 			error = copyfn(sb, sfsp, entry_sz);
1068 			if (error != 0)
1069 				goto out;
1070 		}
1071 		count++;
1072 	}
1073 	if (sfsp && count > maxcount)
1074 		*retval = maxcount;
1075 	else
1076 		*retval = count;
1077 out:
1078 	STATVFSBUF_PUT(sb);
1079 	return error;
1080 }
1081 
1082 int
1083 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1084 {
1085 	/* {
1086 		syscallarg(struct statvfs *) buf;
1087 		syscallarg(size_t) bufsize;
1088 		syscallarg(int) flags;
1089 	} */
1090 
1091 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1092 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1093 }
1094 
1095 /*
1096  * Change current working directory to a given file descriptor.
1097  */
1098 /* ARGSUSED */
1099 int
1100 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1101 {
1102 	/* {
1103 		syscallarg(int) fd;
1104 	} */
1105 	struct proc *p = l->l_proc;
1106 	struct cwdinfo *cwdi;
1107 	struct vnode *vp, *tdp;
1108 	struct mount *mp;
1109 	file_t *fp;
1110 	int error, fd;
1111 
1112 	/* fd_getvnode() will use the descriptor for us */
1113 	fd = SCARG(uap, fd);
1114 	if ((error = fd_getvnode(fd, &fp)) != 0)
1115 		return (error);
1116 	vp = fp->f_data;
1117 
1118 	vref(vp);
1119 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1120 	if (vp->v_type != VDIR)
1121 		error = ENOTDIR;
1122 	else
1123 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1124 	if (error) {
1125 		vput(vp);
1126 		goto out;
1127 	}
1128 	while ((mp = vp->v_mountedhere) != NULL) {
1129 		error = vfs_busy(mp, NULL);
1130 		vput(vp);
1131 		if (error != 0)
1132 			goto out;
1133 		error = VFS_ROOT(mp, &tdp);
1134 		vfs_unbusy(mp, false, NULL);
1135 		if (error)
1136 			goto out;
1137 		vp = tdp;
1138 	}
1139 	VOP_UNLOCK(vp);
1140 
1141 	/*
1142 	 * Disallow changing to a directory not under the process's
1143 	 * current root directory (if there is one).
1144 	 */
1145 	cwdi = p->p_cwdi;
1146 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1147 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1148 		vrele(vp);
1149 		error = EPERM;	/* operation not permitted */
1150 	} else {
1151 		vrele(cwdi->cwdi_cdir);
1152 		cwdi->cwdi_cdir = vp;
1153 	}
1154 	rw_exit(&cwdi->cwdi_lock);
1155 
1156  out:
1157 	fd_putfile(fd);
1158 	return (error);
1159 }
1160 
1161 /*
1162  * Change this process's notion of the root directory to a given file
1163  * descriptor.
1164  */
1165 int
1166 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1167 {
1168 	struct proc *p = l->l_proc;
1169 	struct vnode	*vp;
1170 	file_t	*fp;
1171 	int		 error, fd = SCARG(uap, fd);
1172 
1173 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1174  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1175 		return error;
1176 	/* fd_getvnode() will use the descriptor for us */
1177 	if ((error = fd_getvnode(fd, &fp)) != 0)
1178 		return error;
1179 	vp = fp->f_data;
1180 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1181 	if (vp->v_type != VDIR)
1182 		error = ENOTDIR;
1183 	else
1184 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1185 	VOP_UNLOCK(vp);
1186 	if (error)
1187 		goto out;
1188 	vref(vp);
1189 
1190 	change_root(p->p_cwdi, vp, l);
1191 
1192  out:
1193 	fd_putfile(fd);
1194 	return (error);
1195 }
1196 
1197 /*
1198  * Change current working directory (``.'').
1199  */
1200 /* ARGSUSED */
1201 int
1202 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1203 {
1204 	/* {
1205 		syscallarg(const char *) path;
1206 	} */
1207 	struct proc *p = l->l_proc;
1208 	struct cwdinfo *cwdi;
1209 	int error;
1210 	struct vnode *vp;
1211 
1212 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1213 				  &vp, l)) != 0)
1214 		return (error);
1215 	cwdi = p->p_cwdi;
1216 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1217 	vrele(cwdi->cwdi_cdir);
1218 	cwdi->cwdi_cdir = vp;
1219 	rw_exit(&cwdi->cwdi_lock);
1220 	return (0);
1221 }
1222 
1223 /*
1224  * Change notion of root (``/'') directory.
1225  */
1226 /* ARGSUSED */
1227 int
1228 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1229 {
1230 	/* {
1231 		syscallarg(const char *) path;
1232 	} */
1233 	struct proc *p = l->l_proc;
1234 	int error;
1235 	struct vnode *vp;
1236 
1237 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1238 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1239 		return (error);
1240 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1241 				  &vp, l)) != 0)
1242 		return (error);
1243 
1244 	change_root(p->p_cwdi, vp, l);
1245 
1246 	return (0);
1247 }
1248 
1249 /*
1250  * Common routine for chroot and fchroot.
1251  * NB: callers need to properly authorize the change root operation.
1252  */
1253 void
1254 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1255 {
1256 
1257 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1258 	if (cwdi->cwdi_rdir != NULL)
1259 		vrele(cwdi->cwdi_rdir);
1260 	cwdi->cwdi_rdir = vp;
1261 
1262 	/*
1263 	 * Prevent escaping from chroot by putting the root under
1264 	 * the working directory.  Silently chdir to / if we aren't
1265 	 * already there.
1266 	 */
1267 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1268 		/*
1269 		 * XXX would be more failsafe to change directory to a
1270 		 * deadfs node here instead
1271 		 */
1272 		vrele(cwdi->cwdi_cdir);
1273 		vref(vp);
1274 		cwdi->cwdi_cdir = vp;
1275 	}
1276 	rw_exit(&cwdi->cwdi_lock);
1277 }
1278 
1279 /*
1280  * Common routine for chroot and chdir.
1281  */
1282 int
1283 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1284 {
1285 	struct nameidata nd;
1286 	int error;
1287 
1288 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, where,
1289 	    path);
1290 	if ((error = namei(&nd)) != 0)
1291 		return (error);
1292 	*vpp = nd.ni_vp;
1293 	if ((*vpp)->v_type != VDIR)
1294 		error = ENOTDIR;
1295 	else
1296 		error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1297 
1298 	if (error)
1299 		vput(*vpp);
1300 	else
1301 		VOP_UNLOCK(*vpp);
1302 	return (error);
1303 }
1304 
1305 /*
1306  * Check permissions, allocate an open file structure,
1307  * and call the device open routine if any.
1308  */
1309 int
1310 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1311 {
1312 	/* {
1313 		syscallarg(const char *) path;
1314 		syscallarg(int) flags;
1315 		syscallarg(int) mode;
1316 	} */
1317 	struct proc *p = l->l_proc;
1318 	struct cwdinfo *cwdi = p->p_cwdi;
1319 	file_t *fp;
1320 	struct vnode *vp;
1321 	int flags, cmode;
1322 	int type, indx, error;
1323 	struct flock lf;
1324 	struct nameidata nd;
1325 
1326 	flags = FFLAGS(SCARG(uap, flags));
1327 	if ((flags & (FREAD | FWRITE)) == 0)
1328 		return (EINVAL);
1329 	if ((error = fd_allocfile(&fp, &indx)) != 0)
1330 		return (error);
1331 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1332 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1333 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1334 	    SCARG(uap, path));
1335 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1336 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1337 		fd_abort(p, fp, indx);
1338 		if ((error == EDUPFD || error == EMOVEFD) &&
1339 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1340 		    (error =
1341 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1342 			*retval = indx;
1343 			return (0);
1344 		}
1345 		if (error == ERESTART)
1346 			error = EINTR;
1347 		return (error);
1348 	}
1349 
1350 	l->l_dupfd = 0;
1351 	vp = nd.ni_vp;
1352 	fp->f_flag = flags & FMASK;
1353 	fp->f_type = DTYPE_VNODE;
1354 	fp->f_ops = &vnops;
1355 	fp->f_data = vp;
1356 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1357 		lf.l_whence = SEEK_SET;
1358 		lf.l_start = 0;
1359 		lf.l_len = 0;
1360 		if (flags & O_EXLOCK)
1361 			lf.l_type = F_WRLCK;
1362 		else
1363 			lf.l_type = F_RDLCK;
1364 		type = F_FLOCK;
1365 		if ((flags & FNONBLOCK) == 0)
1366 			type |= F_WAIT;
1367 		VOP_UNLOCK(vp);
1368 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1369 		if (error) {
1370 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1371 			fd_abort(p, fp, indx);
1372 			return (error);
1373 		}
1374 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1375 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1376 	}
1377 	VOP_UNLOCK(vp);
1378 	*retval = indx;
1379 	fd_affix(p, fp, indx);
1380 	return (0);
1381 }
1382 
1383 static void
1384 vfs__fhfree(fhandle_t *fhp)
1385 {
1386 	size_t fhsize;
1387 
1388 	if (fhp == NULL) {
1389 		return;
1390 	}
1391 	fhsize = FHANDLE_SIZE(fhp);
1392 	kmem_free(fhp, fhsize);
1393 }
1394 
1395 /*
1396  * vfs_composefh: compose a filehandle.
1397  */
1398 
1399 int
1400 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1401 {
1402 	struct mount *mp;
1403 	struct fid *fidp;
1404 	int error;
1405 	size_t needfhsize;
1406 	size_t fidsize;
1407 
1408 	mp = vp->v_mount;
1409 	fidp = NULL;
1410 	if (*fh_size < FHANDLE_SIZE_MIN) {
1411 		fidsize = 0;
1412 	} else {
1413 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1414 		if (fhp != NULL) {
1415 			memset(fhp, 0, *fh_size);
1416 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1417 			fidp = &fhp->fh_fid;
1418 		}
1419 	}
1420 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1421 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1422 	if (error == 0 && *fh_size < needfhsize) {
1423 		error = E2BIG;
1424 	}
1425 	*fh_size = needfhsize;
1426 	return error;
1427 }
1428 
1429 int
1430 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1431 {
1432 	struct mount *mp;
1433 	fhandle_t *fhp;
1434 	size_t fhsize;
1435 	size_t fidsize;
1436 	int error;
1437 
1438 	*fhpp = NULL;
1439 	mp = vp->v_mount;
1440 	fidsize = 0;
1441 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1442 	KASSERT(error != 0);
1443 	if (error != E2BIG) {
1444 		goto out;
1445 	}
1446 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1447 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1448 	if (fhp == NULL) {
1449 		error = ENOMEM;
1450 		goto out;
1451 	}
1452 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1453 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1454 	if (error == 0) {
1455 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1456 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1457 		*fhpp = fhp;
1458 	} else {
1459 		kmem_free(fhp, fhsize);
1460 	}
1461 out:
1462 	return error;
1463 }
1464 
1465 void
1466 vfs_composefh_free(fhandle_t *fhp)
1467 {
1468 
1469 	vfs__fhfree(fhp);
1470 }
1471 
1472 /*
1473  * vfs_fhtovp: lookup a vnode by a filehandle.
1474  */
1475 
1476 int
1477 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1478 {
1479 	struct mount *mp;
1480 	int error;
1481 
1482 	*vpp = NULL;
1483 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1484 	if (mp == NULL) {
1485 		error = ESTALE;
1486 		goto out;
1487 	}
1488 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1489 		error = EOPNOTSUPP;
1490 		goto out;
1491 	}
1492 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1493 out:
1494 	return error;
1495 }
1496 
1497 /*
1498  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1499  * the needed size.
1500  */
1501 
1502 int
1503 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1504 {
1505 	fhandle_t *fhp;
1506 	int error;
1507 
1508 	*fhpp = NULL;
1509 	if (fhsize > FHANDLE_SIZE_MAX) {
1510 		return EINVAL;
1511 	}
1512 	if (fhsize < FHANDLE_SIZE_MIN) {
1513 		return EINVAL;
1514 	}
1515 again:
1516 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1517 	if (fhp == NULL) {
1518 		return ENOMEM;
1519 	}
1520 	error = copyin(ufhp, fhp, fhsize);
1521 	if (error == 0) {
1522 		/* XXX this check shouldn't be here */
1523 		if (FHANDLE_SIZE(fhp) == fhsize) {
1524 			*fhpp = fhp;
1525 			return 0;
1526 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1527 			/*
1528 			 * a kludge for nfsv2 padded handles.
1529 			 */
1530 			size_t sz;
1531 
1532 			sz = FHANDLE_SIZE(fhp);
1533 			kmem_free(fhp, fhsize);
1534 			fhsize = sz;
1535 			goto again;
1536 		} else {
1537 			/*
1538 			 * userland told us wrong size.
1539 			 */
1540 		    	error = EINVAL;
1541 		}
1542 	}
1543 	kmem_free(fhp, fhsize);
1544 	return error;
1545 }
1546 
1547 void
1548 vfs_copyinfh_free(fhandle_t *fhp)
1549 {
1550 
1551 	vfs__fhfree(fhp);
1552 }
1553 
1554 /*
1555  * Get file handle system call
1556  */
1557 int
1558 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1559 {
1560 	/* {
1561 		syscallarg(char *) fname;
1562 		syscallarg(fhandle_t *) fhp;
1563 		syscallarg(size_t *) fh_size;
1564 	} */
1565 	struct vnode *vp;
1566 	fhandle_t *fh;
1567 	int error;
1568 	struct nameidata nd;
1569 	size_t sz;
1570 	size_t usz;
1571 
1572 	/*
1573 	 * Must be super user
1574 	 */
1575 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1576 	    0, NULL, NULL, NULL);
1577 	if (error)
1578 		return (error);
1579 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1580 	    SCARG(uap, fname));
1581 	error = namei(&nd);
1582 	if (error)
1583 		return (error);
1584 	vp = nd.ni_vp;
1585 	error = vfs_composefh_alloc(vp, &fh);
1586 	vput(vp);
1587 	if (error != 0) {
1588 		goto out;
1589 	}
1590 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1591 	if (error != 0) {
1592 		goto out;
1593 	}
1594 	sz = FHANDLE_SIZE(fh);
1595 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1596 	if (error != 0) {
1597 		goto out;
1598 	}
1599 	if (usz >= sz) {
1600 		error = copyout(fh, SCARG(uap, fhp), sz);
1601 	} else {
1602 		error = E2BIG;
1603 	}
1604 out:
1605 	vfs_composefh_free(fh);
1606 	return (error);
1607 }
1608 
1609 /*
1610  * Open a file given a file handle.
1611  *
1612  * Check permissions, allocate an open file structure,
1613  * and call the device open routine if any.
1614  */
1615 
1616 int
1617 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1618     register_t *retval)
1619 {
1620 	file_t *fp;
1621 	struct vnode *vp = NULL;
1622 	kauth_cred_t cred = l->l_cred;
1623 	file_t *nfp;
1624 	int type, indx, error=0;
1625 	struct flock lf;
1626 	struct vattr va;
1627 	fhandle_t *fh;
1628 	int flags;
1629 	proc_t *p;
1630 
1631 	p = curproc;
1632 
1633 	/*
1634 	 * Must be super user
1635 	 */
1636 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1637 	    0, NULL, NULL, NULL)))
1638 		return (error);
1639 
1640 	flags = FFLAGS(oflags);
1641 	if ((flags & (FREAD | FWRITE)) == 0)
1642 		return (EINVAL);
1643 	if ((flags & O_CREAT))
1644 		return (EINVAL);
1645 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1646 		return (error);
1647 	fp = nfp;
1648 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1649 	if (error != 0) {
1650 		goto bad;
1651 	}
1652 	error = vfs_fhtovp(fh, &vp);
1653 	if (error != 0) {
1654 		goto bad;
1655 	}
1656 
1657 	/* Now do an effective vn_open */
1658 
1659 	if (vp->v_type == VSOCK) {
1660 		error = EOPNOTSUPP;
1661 		goto bad;
1662 	}
1663 	error = vn_openchk(vp, cred, flags);
1664 	if (error != 0)
1665 		goto bad;
1666 	if (flags & O_TRUNC) {
1667 		VOP_UNLOCK(vp);			/* XXX */
1668 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1669 		vattr_null(&va);
1670 		va.va_size = 0;
1671 		error = VOP_SETATTR(vp, &va, cred);
1672 		if (error)
1673 			goto bad;
1674 	}
1675 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1676 		goto bad;
1677 	if (flags & FWRITE) {
1678 		mutex_enter(&vp->v_interlock);
1679 		vp->v_writecount++;
1680 		mutex_exit(&vp->v_interlock);
1681 	}
1682 
1683 	/* done with modified vn_open, now finish what sys_open does. */
1684 
1685 	fp->f_flag = flags & FMASK;
1686 	fp->f_type = DTYPE_VNODE;
1687 	fp->f_ops = &vnops;
1688 	fp->f_data = vp;
1689 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1690 		lf.l_whence = SEEK_SET;
1691 		lf.l_start = 0;
1692 		lf.l_len = 0;
1693 		if (flags & O_EXLOCK)
1694 			lf.l_type = F_WRLCK;
1695 		else
1696 			lf.l_type = F_RDLCK;
1697 		type = F_FLOCK;
1698 		if ((flags & FNONBLOCK) == 0)
1699 			type |= F_WAIT;
1700 		VOP_UNLOCK(vp);
1701 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1702 		if (error) {
1703 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1704 			fd_abort(p, fp, indx);
1705 			return (error);
1706 		}
1707 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1708 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1709 	}
1710 	VOP_UNLOCK(vp);
1711 	*retval = indx;
1712 	fd_affix(p, fp, indx);
1713 	vfs_copyinfh_free(fh);
1714 	return (0);
1715 
1716 bad:
1717 	fd_abort(p, fp, indx);
1718 	if (vp != NULL)
1719 		vput(vp);
1720 	vfs_copyinfh_free(fh);
1721 	return (error);
1722 }
1723 
1724 int
1725 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1726 {
1727 	/* {
1728 		syscallarg(const void *) fhp;
1729 		syscallarg(size_t) fh_size;
1730 		syscallarg(int) flags;
1731 	} */
1732 
1733 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1734 	    SCARG(uap, flags), retval);
1735 }
1736 
1737 int
1738 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1739 {
1740 	int error;
1741 	fhandle_t *fh;
1742 	struct vnode *vp;
1743 
1744 	/*
1745 	 * Must be super user
1746 	 */
1747 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1748 	    0, NULL, NULL, NULL)))
1749 		return (error);
1750 
1751 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1752 	if (error != 0)
1753 		return error;
1754 
1755 	error = vfs_fhtovp(fh, &vp);
1756 	vfs_copyinfh_free(fh);
1757 	if (error != 0)
1758 		return error;
1759 
1760 	error = vn_stat(vp, sb);
1761 	vput(vp);
1762 	return error;
1763 }
1764 
1765 
1766 /* ARGSUSED */
1767 int
1768 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1769 {
1770 	/* {
1771 		syscallarg(const void *) fhp;
1772 		syscallarg(size_t) fh_size;
1773 		syscallarg(struct stat *) sb;
1774 	} */
1775 	struct stat sb;
1776 	int error;
1777 
1778 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1779 	if (error)
1780 		return error;
1781 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1782 }
1783 
1784 int
1785 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1786     int flags)
1787 {
1788 	fhandle_t *fh;
1789 	struct mount *mp;
1790 	struct vnode *vp;
1791 	int error;
1792 
1793 	/*
1794 	 * Must be super user
1795 	 */
1796 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1797 	    0, NULL, NULL, NULL)))
1798 		return error;
1799 
1800 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1801 	if (error != 0)
1802 		return error;
1803 
1804 	error = vfs_fhtovp(fh, &vp);
1805 	vfs_copyinfh_free(fh);
1806 	if (error != 0)
1807 		return error;
1808 
1809 	mp = vp->v_mount;
1810 	error = dostatvfs(mp, sb, l, flags, 1);
1811 	vput(vp);
1812 	return error;
1813 }
1814 
1815 /* ARGSUSED */
1816 int
1817 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1818 {
1819 	/* {
1820 		syscallarg(const void *) fhp;
1821 		syscallarg(size_t) fh_size;
1822 		syscallarg(struct statvfs *) buf;
1823 		syscallarg(int)	flags;
1824 	} */
1825 	struct statvfs *sb = STATVFSBUF_GET();
1826 	int error;
1827 
1828 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1829 	    SCARG(uap, flags));
1830 	if (error == 0)
1831 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1832 	STATVFSBUF_PUT(sb);
1833 	return error;
1834 }
1835 
1836 /*
1837  * Create a special file.
1838  */
1839 /* ARGSUSED */
1840 int
1841 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
1842     register_t *retval)
1843 {
1844 	/* {
1845 		syscallarg(const char *) path;
1846 		syscallarg(mode_t) mode;
1847 		syscallarg(dev_t) dev;
1848 	} */
1849 	return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
1850 	    SCARG(uap, dev), retval, UIO_USERSPACE);
1851 }
1852 
1853 int
1854 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
1855     register_t *retval, enum uio_seg seg)
1856 {
1857 	struct proc *p = l->l_proc;
1858 	struct vnode *vp;
1859 	struct vattr vattr;
1860 	int error, optype;
1861 	struct nameidata nd;
1862 	char *path;
1863 	const char *cpath;
1864 
1865 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1866 	    0, NULL, NULL, NULL)) != 0)
1867 		return (error);
1868 
1869 	optype = VOP_MKNOD_DESCOFFSET;
1870 
1871 	VERIEXEC_PATH_GET(pathname, seg, cpath, path);
1872 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
1873 
1874 	if ((error = namei(&nd)) != 0)
1875 		goto out;
1876 	vp = nd.ni_vp;
1877 	if (vp != NULL)
1878 		error = EEXIST;
1879 	else {
1880 		vattr_null(&vattr);
1881 		/* We will read cwdi->cwdi_cmask unlocked. */
1882 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1883 		vattr.va_rdev = dev;
1884 
1885 		switch (mode & S_IFMT) {
1886 		case S_IFMT:	/* used by badsect to flag bad sectors */
1887 			vattr.va_type = VBAD;
1888 			break;
1889 		case S_IFCHR:
1890 			vattr.va_type = VCHR;
1891 			break;
1892 		case S_IFBLK:
1893 			vattr.va_type = VBLK;
1894 			break;
1895 		case S_IFWHT:
1896 			optype = VOP_WHITEOUT_DESCOFFSET;
1897 			break;
1898 		case S_IFREG:
1899 #if NVERIEXEC > 0
1900 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1901 			    O_CREAT);
1902 #endif /* NVERIEXEC > 0 */
1903 			vattr.va_type = VREG;
1904 			vattr.va_rdev = VNOVAL;
1905 			optype = VOP_CREATE_DESCOFFSET;
1906 			break;
1907 		default:
1908 			error = EINVAL;
1909 			break;
1910 		}
1911 	}
1912 	if (!error) {
1913 		switch (optype) {
1914 		case VOP_WHITEOUT_DESCOFFSET:
1915 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1916 			if (error)
1917 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1918 			vput(nd.ni_dvp);
1919 			break;
1920 
1921 		case VOP_MKNOD_DESCOFFSET:
1922 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1923 						&nd.ni_cnd, &vattr);
1924 			if (error == 0)
1925 				vput(nd.ni_vp);
1926 			break;
1927 
1928 		case VOP_CREATE_DESCOFFSET:
1929 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1930 						&nd.ni_cnd, &vattr);
1931 			if (error == 0)
1932 				vput(nd.ni_vp);
1933 			break;
1934 		}
1935 	} else {
1936 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1937 		if (nd.ni_dvp == vp)
1938 			vrele(nd.ni_dvp);
1939 		else
1940 			vput(nd.ni_dvp);
1941 		if (vp)
1942 			vrele(vp);
1943 	}
1944 out:
1945 	VERIEXEC_PATH_PUT(path);
1946 	return (error);
1947 }
1948 
1949 /*
1950  * Create a named pipe.
1951  */
1952 /* ARGSUSED */
1953 int
1954 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1955 {
1956 	/* {
1957 		syscallarg(const char *) path;
1958 		syscallarg(int) mode;
1959 	} */
1960 	struct proc *p = l->l_proc;
1961 	struct vattr vattr;
1962 	int error;
1963 	struct nameidata nd;
1964 
1965 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1966 	    SCARG(uap, path));
1967 	if ((error = namei(&nd)) != 0)
1968 		return (error);
1969 	if (nd.ni_vp != NULL) {
1970 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1971 		if (nd.ni_dvp == nd.ni_vp)
1972 			vrele(nd.ni_dvp);
1973 		else
1974 			vput(nd.ni_dvp);
1975 		vrele(nd.ni_vp);
1976 		return (EEXIST);
1977 	}
1978 	vattr_null(&vattr);
1979 	vattr.va_type = VFIFO;
1980 	/* We will read cwdi->cwdi_cmask unlocked. */
1981 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1982 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1983 	if (error == 0)
1984 		vput(nd.ni_vp);
1985 	return (error);
1986 }
1987 
1988 /*
1989  * Make a hard file link.
1990  */
1991 /* ARGSUSED */
1992 int
1993 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
1994 {
1995 	/* {
1996 		syscallarg(const char *) path;
1997 		syscallarg(const char *) link;
1998 	} */
1999 	struct vnode *vp;
2000 	struct nameidata nd;
2001 	int error;
2002 
2003 	error = namei_simple_user(SCARG(uap, path),
2004 				NSM_FOLLOW_TRYEMULROOT, &vp);
2005 	if (error != 0)
2006 		return (error);
2007 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2008 	    SCARG(uap, link));
2009 	if ((error = namei(&nd)) != 0)
2010 		goto out;
2011 	if (nd.ni_vp) {
2012 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2013 		if (nd.ni_dvp == nd.ni_vp)
2014 			vrele(nd.ni_dvp);
2015 		else
2016 			vput(nd.ni_dvp);
2017 		vrele(nd.ni_vp);
2018 		error = EEXIST;
2019 		goto out;
2020 	}
2021 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2022 out:
2023 	vrele(vp);
2024 	return (error);
2025 }
2026 
2027 int
2028 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg)
2029 {
2030 	struct proc *p = curproc;
2031 	struct vattr vattr;
2032 	char *path;
2033 	int error;
2034 	struct nameidata nd;
2035 
2036 	path = PNBUF_GET();
2037 	if (seg == UIO_USERSPACE) {
2038 		if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0)
2039 			goto out;
2040 	} else {
2041 		KASSERT(strlen(patharg) < MAXPATHLEN);
2042 		strcpy(path, patharg);
2043 	}
2044 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, link);
2045 	if ((error = namei(&nd)) != 0)
2046 		goto out;
2047 	if (nd.ni_vp) {
2048 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2049 		if (nd.ni_dvp == nd.ni_vp)
2050 			vrele(nd.ni_dvp);
2051 		else
2052 			vput(nd.ni_dvp);
2053 		vrele(nd.ni_vp);
2054 		error = EEXIST;
2055 		goto out;
2056 	}
2057 	vattr_null(&vattr);
2058 	vattr.va_type = VLNK;
2059 	/* We will read cwdi->cwdi_cmask unlocked. */
2060 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2061 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2062 	if (error == 0)
2063 		vput(nd.ni_vp);
2064 out:
2065 	PNBUF_PUT(path);
2066 	return (error);
2067 }
2068 
2069 /*
2070  * Make a symbolic link.
2071  */
2072 /* ARGSUSED */
2073 int
2074 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2075 {
2076 	/* {
2077 		syscallarg(const char *) path;
2078 		syscallarg(const char *) link;
2079 	} */
2080 
2081 	return do_sys_symlink(SCARG(uap, path), SCARG(uap, link),
2082 	    UIO_USERSPACE);
2083 }
2084 
2085 /*
2086  * Delete a whiteout from the filesystem.
2087  */
2088 /* ARGSUSED */
2089 int
2090 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2091 {
2092 	/* {
2093 		syscallarg(const char *) path;
2094 	} */
2095 	int error;
2096 	struct nameidata nd;
2097 
2098 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
2099 	    UIO_USERSPACE, SCARG(uap, path));
2100 	error = namei(&nd);
2101 	if (error)
2102 		return (error);
2103 
2104 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2105 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2106 		if (nd.ni_dvp == nd.ni_vp)
2107 			vrele(nd.ni_dvp);
2108 		else
2109 			vput(nd.ni_dvp);
2110 		if (nd.ni_vp)
2111 			vrele(nd.ni_vp);
2112 		return (EEXIST);
2113 	}
2114 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2115 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2116 	vput(nd.ni_dvp);
2117 	return (error);
2118 }
2119 
2120 /*
2121  * Delete a name from the filesystem.
2122  */
2123 /* ARGSUSED */
2124 int
2125 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2126 {
2127 	/* {
2128 		syscallarg(const char *) path;
2129 	} */
2130 
2131 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2132 }
2133 
2134 int
2135 do_sys_unlink(const char *arg, enum uio_seg seg)
2136 {
2137 	struct vnode *vp;
2138 	int error;
2139 	struct nameidata nd;
2140 	char *path;
2141 	const char *cpath;
2142 
2143 	VERIEXEC_PATH_GET(arg, seg, cpath, path);
2144 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
2145 
2146 	if ((error = namei(&nd)) != 0)
2147 		goto out;
2148 	vp = nd.ni_vp;
2149 
2150 	/*
2151 	 * The root of a mounted filesystem cannot be deleted.
2152 	 */
2153 	if (vp->v_vflag & VV_ROOT) {
2154 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2155 		if (nd.ni_dvp == vp)
2156 			vrele(nd.ni_dvp);
2157 		else
2158 			vput(nd.ni_dvp);
2159 		vput(vp);
2160 		error = EBUSY;
2161 		goto out;
2162 	}
2163 
2164 #if NVERIEXEC > 0
2165 	/* Handle remove requests for veriexec entries. */
2166 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
2167 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2168 		if (nd.ni_dvp == vp)
2169 			vrele(nd.ni_dvp);
2170 		else
2171 			vput(nd.ni_dvp);
2172 		vput(vp);
2173 		goto out;
2174 	}
2175 #endif /* NVERIEXEC > 0 */
2176 
2177 #ifdef FILEASSOC
2178 	(void)fileassoc_file_delete(vp);
2179 #endif /* FILEASSOC */
2180 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2181 out:
2182 	VERIEXEC_PATH_PUT(path);
2183 	return (error);
2184 }
2185 
2186 /*
2187  * Reposition read/write file offset.
2188  */
2189 int
2190 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2191 {
2192 	/* {
2193 		syscallarg(int) fd;
2194 		syscallarg(int) pad;
2195 		syscallarg(off_t) offset;
2196 		syscallarg(int) whence;
2197 	} */
2198 	kauth_cred_t cred = l->l_cred;
2199 	file_t *fp;
2200 	struct vnode *vp;
2201 	struct vattr vattr;
2202 	off_t newoff;
2203 	int error, fd;
2204 
2205 	fd = SCARG(uap, fd);
2206 
2207 	if ((fp = fd_getfile(fd)) == NULL)
2208 		return (EBADF);
2209 
2210 	vp = fp->f_data;
2211 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2212 		error = ESPIPE;
2213 		goto out;
2214 	}
2215 
2216 	switch (SCARG(uap, whence)) {
2217 	case SEEK_CUR:
2218 		newoff = fp->f_offset + SCARG(uap, offset);
2219 		break;
2220 	case SEEK_END:
2221 		error = VOP_GETATTR(vp, &vattr, cred);
2222 		if (error) {
2223 			goto out;
2224 		}
2225 		newoff = SCARG(uap, offset) + vattr.va_size;
2226 		break;
2227 	case SEEK_SET:
2228 		newoff = SCARG(uap, offset);
2229 		break;
2230 	default:
2231 		error = EINVAL;
2232 		goto out;
2233 	}
2234 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2235 		*(off_t *)retval = fp->f_offset = newoff;
2236 	}
2237  out:
2238  	fd_putfile(fd);
2239 	return (error);
2240 }
2241 
2242 /*
2243  * Positional read system call.
2244  */
2245 int
2246 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2247 {
2248 	/* {
2249 		syscallarg(int) fd;
2250 		syscallarg(void *) buf;
2251 		syscallarg(size_t) nbyte;
2252 		syscallarg(off_t) offset;
2253 	} */
2254 	file_t *fp;
2255 	struct vnode *vp;
2256 	off_t offset;
2257 	int error, fd = SCARG(uap, fd);
2258 
2259 	if ((fp = fd_getfile(fd)) == NULL)
2260 		return (EBADF);
2261 
2262 	if ((fp->f_flag & FREAD) == 0) {
2263 		fd_putfile(fd);
2264 		return (EBADF);
2265 	}
2266 
2267 	vp = fp->f_data;
2268 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2269 		error = ESPIPE;
2270 		goto out;
2271 	}
2272 
2273 	offset = SCARG(uap, offset);
2274 
2275 	/*
2276 	 * XXX This works because no file systems actually
2277 	 * XXX take any action on the seek operation.
2278 	 */
2279 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2280 		goto out;
2281 
2282 	/* dofileread() will unuse the descriptor for us */
2283 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2284 	    &offset, 0, retval));
2285 
2286  out:
2287 	fd_putfile(fd);
2288 	return (error);
2289 }
2290 
2291 /*
2292  * Positional scatter read system call.
2293  */
2294 int
2295 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2296 {
2297 	/* {
2298 		syscallarg(int) fd;
2299 		syscallarg(const struct iovec *) iovp;
2300 		syscallarg(int) iovcnt;
2301 		syscallarg(off_t) offset;
2302 	} */
2303 	off_t offset = SCARG(uap, offset);
2304 
2305 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2306 	    SCARG(uap, iovcnt), &offset, 0, retval);
2307 }
2308 
2309 /*
2310  * Positional write system call.
2311  */
2312 int
2313 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2314 {
2315 	/* {
2316 		syscallarg(int) fd;
2317 		syscallarg(const void *) buf;
2318 		syscallarg(size_t) nbyte;
2319 		syscallarg(off_t) offset;
2320 	} */
2321 	file_t *fp;
2322 	struct vnode *vp;
2323 	off_t offset;
2324 	int error, fd = SCARG(uap, fd);
2325 
2326 	if ((fp = fd_getfile(fd)) == NULL)
2327 		return (EBADF);
2328 
2329 	if ((fp->f_flag & FWRITE) == 0) {
2330 		fd_putfile(fd);
2331 		return (EBADF);
2332 	}
2333 
2334 	vp = fp->f_data;
2335 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2336 		error = ESPIPE;
2337 		goto out;
2338 	}
2339 
2340 	offset = SCARG(uap, offset);
2341 
2342 	/*
2343 	 * XXX This works because no file systems actually
2344 	 * XXX take any action on the seek operation.
2345 	 */
2346 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2347 		goto out;
2348 
2349 	/* dofilewrite() will unuse the descriptor for us */
2350 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2351 	    &offset, 0, retval));
2352 
2353  out:
2354 	fd_putfile(fd);
2355 	return (error);
2356 }
2357 
2358 /*
2359  * Positional gather write system call.
2360  */
2361 int
2362 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2363 {
2364 	/* {
2365 		syscallarg(int) fd;
2366 		syscallarg(const struct iovec *) iovp;
2367 		syscallarg(int) iovcnt;
2368 		syscallarg(off_t) offset;
2369 	} */
2370 	off_t offset = SCARG(uap, offset);
2371 
2372 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2373 	    SCARG(uap, iovcnt), &offset, 0, retval);
2374 }
2375 
2376 /*
2377  * Check access permissions.
2378  */
2379 int
2380 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2381 {
2382 	/* {
2383 		syscallarg(const char *) path;
2384 		syscallarg(int) flags;
2385 	} */
2386 	kauth_cred_t cred;
2387 	struct vnode *vp;
2388 	int error, flags;
2389 	struct nameidata nd;
2390 
2391 	cred = kauth_cred_dup(l->l_cred);
2392 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2393 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2394 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2395 	    SCARG(uap, path));
2396 	/* Override default credentials */
2397 	nd.ni_cnd.cn_cred = cred;
2398 	if ((error = namei(&nd)) != 0)
2399 		goto out;
2400 	vp = nd.ni_vp;
2401 
2402 	/* Flags == 0 means only check for existence. */
2403 	if (SCARG(uap, flags)) {
2404 		flags = 0;
2405 		if (SCARG(uap, flags) & R_OK)
2406 			flags |= VREAD;
2407 		if (SCARG(uap, flags) & W_OK)
2408 			flags |= VWRITE;
2409 		if (SCARG(uap, flags) & X_OK)
2410 			flags |= VEXEC;
2411 
2412 		error = VOP_ACCESS(vp, flags, cred);
2413 		if (!error && (flags & VWRITE))
2414 			error = vn_writechk(vp);
2415 	}
2416 	vput(vp);
2417 out:
2418 	kauth_cred_free(cred);
2419 	return (error);
2420 }
2421 
2422 /*
2423  * Common code for all sys_stat functions, including compat versions.
2424  */
2425 int
2426 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
2427 {
2428 	int error;
2429 	struct nameidata nd;
2430 
2431 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
2432 	    UIO_USERSPACE, path);
2433 	error = namei(&nd);
2434 	if (error != 0)
2435 		return error;
2436 	error = vn_stat(nd.ni_vp, sb);
2437 	vput(nd.ni_vp);
2438 	return error;
2439 }
2440 
2441 /*
2442  * Get file status; this version follows links.
2443  */
2444 /* ARGSUSED */
2445 int
2446 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
2447 {
2448 	/* {
2449 		syscallarg(const char *) path;
2450 		syscallarg(struct stat *) ub;
2451 	} */
2452 	struct stat sb;
2453 	int error;
2454 
2455 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2456 	if (error)
2457 		return error;
2458 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2459 }
2460 
2461 /*
2462  * Get file status; this version does not follow links.
2463  */
2464 /* ARGSUSED */
2465 int
2466 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
2467 {
2468 	/* {
2469 		syscallarg(const char *) path;
2470 		syscallarg(struct stat *) ub;
2471 	} */
2472 	struct stat sb;
2473 	int error;
2474 
2475 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2476 	if (error)
2477 		return error;
2478 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2479 }
2480 
2481 /*
2482  * Get configurable pathname variables.
2483  */
2484 /* ARGSUSED */
2485 int
2486 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2487 {
2488 	/* {
2489 		syscallarg(const char *) path;
2490 		syscallarg(int) name;
2491 	} */
2492 	int error;
2493 	struct nameidata nd;
2494 
2495 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2496 	    SCARG(uap, path));
2497 	if ((error = namei(&nd)) != 0)
2498 		return (error);
2499 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2500 	vput(nd.ni_vp);
2501 	return (error);
2502 }
2503 
2504 /*
2505  * Return target name of a symbolic link.
2506  */
2507 /* ARGSUSED */
2508 int
2509 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2510 {
2511 	/* {
2512 		syscallarg(const char *) path;
2513 		syscallarg(char *) buf;
2514 		syscallarg(size_t) count;
2515 	} */
2516 	struct vnode *vp;
2517 	struct iovec aiov;
2518 	struct uio auio;
2519 	int error;
2520 	struct nameidata nd;
2521 
2522 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2523 	    SCARG(uap, path));
2524 	if ((error = namei(&nd)) != 0)
2525 		return (error);
2526 	vp = nd.ni_vp;
2527 	if (vp->v_type != VLNK)
2528 		error = EINVAL;
2529 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2530 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2531 		aiov.iov_base = SCARG(uap, buf);
2532 		aiov.iov_len = SCARG(uap, count);
2533 		auio.uio_iov = &aiov;
2534 		auio.uio_iovcnt = 1;
2535 		auio.uio_offset = 0;
2536 		auio.uio_rw = UIO_READ;
2537 		KASSERT(l == curlwp);
2538 		auio.uio_vmspace = l->l_proc->p_vmspace;
2539 		auio.uio_resid = SCARG(uap, count);
2540 		error = VOP_READLINK(vp, &auio, l->l_cred);
2541 	}
2542 	vput(vp);
2543 	*retval = SCARG(uap, count) - auio.uio_resid;
2544 	return (error);
2545 }
2546 
2547 /*
2548  * Change flags of a file given a path name.
2549  */
2550 /* ARGSUSED */
2551 int
2552 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2553 {
2554 	/* {
2555 		syscallarg(const char *) path;
2556 		syscallarg(u_long) flags;
2557 	} */
2558 	struct vnode *vp;
2559 	int error;
2560 
2561 	error = namei_simple_user(SCARG(uap, path),
2562 				NSM_FOLLOW_TRYEMULROOT, &vp);
2563 	if (error != 0)
2564 		return (error);
2565 	error = change_flags(vp, SCARG(uap, flags), l);
2566 	vput(vp);
2567 	return (error);
2568 }
2569 
2570 /*
2571  * Change flags of a file given a file descriptor.
2572  */
2573 /* ARGSUSED */
2574 int
2575 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
2576 {
2577 	/* {
2578 		syscallarg(int) fd;
2579 		syscallarg(u_long) flags;
2580 	} */
2581 	struct vnode *vp;
2582 	file_t *fp;
2583 	int error;
2584 
2585 	/* fd_getvnode() will use the descriptor for us */
2586 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2587 		return (error);
2588 	vp = fp->f_data;
2589 	error = change_flags(vp, SCARG(uap, flags), l);
2590 	VOP_UNLOCK(vp);
2591 	fd_putfile(SCARG(uap, fd));
2592 	return (error);
2593 }
2594 
2595 /*
2596  * Change flags of a file given a path name; this version does
2597  * not follow links.
2598  */
2599 int
2600 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
2601 {
2602 	/* {
2603 		syscallarg(const char *) path;
2604 		syscallarg(u_long) flags;
2605 	} */
2606 	struct vnode *vp;
2607 	int error;
2608 
2609 	error = namei_simple_user(SCARG(uap, path),
2610 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2611 	if (error != 0)
2612 		return (error);
2613 	error = change_flags(vp, SCARG(uap, flags), l);
2614 	vput(vp);
2615 	return (error);
2616 }
2617 
2618 /*
2619  * Common routine to change flags of a file.
2620  */
2621 int
2622 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2623 {
2624 	struct vattr vattr;
2625 	int error;
2626 
2627 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2628 	/*
2629 	 * Non-superusers cannot change the flags on devices, even if they
2630 	 * own them.
2631 	 */
2632 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2633 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2634 			goto out;
2635 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2636 			error = EINVAL;
2637 			goto out;
2638 		}
2639 	}
2640 	vattr_null(&vattr);
2641 	vattr.va_flags = flags;
2642 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2643 out:
2644 	return (error);
2645 }
2646 
2647 /*
2648  * Change mode of a file given path name; this version follows links.
2649  */
2650 /* ARGSUSED */
2651 int
2652 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
2653 {
2654 	/* {
2655 		syscallarg(const char *) path;
2656 		syscallarg(int) mode;
2657 	} */
2658 	int error;
2659 	struct vnode *vp;
2660 
2661 	error = namei_simple_user(SCARG(uap, path),
2662 				NSM_FOLLOW_TRYEMULROOT, &vp);
2663 	if (error != 0)
2664 		return (error);
2665 
2666 	error = change_mode(vp, SCARG(uap, mode), l);
2667 
2668 	vrele(vp);
2669 	return (error);
2670 }
2671 
2672 /*
2673  * Change mode of a file given a file descriptor.
2674  */
2675 /* ARGSUSED */
2676 int
2677 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
2678 {
2679 	/* {
2680 		syscallarg(int) fd;
2681 		syscallarg(int) mode;
2682 	} */
2683 	file_t *fp;
2684 	int error;
2685 
2686 	/* fd_getvnode() will use the descriptor for us */
2687 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2688 		return (error);
2689 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
2690 	fd_putfile(SCARG(uap, fd));
2691 	return (error);
2692 }
2693 
2694 /*
2695  * Change mode of a file given path name; this version does not follow links.
2696  */
2697 /* ARGSUSED */
2698 int
2699 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
2700 {
2701 	/* {
2702 		syscallarg(const char *) path;
2703 		syscallarg(int) mode;
2704 	} */
2705 	int error;
2706 	struct vnode *vp;
2707 
2708 	error = namei_simple_user(SCARG(uap, path),
2709 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2710 	if (error != 0)
2711 		return (error);
2712 
2713 	error = change_mode(vp, SCARG(uap, mode), l);
2714 
2715 	vrele(vp);
2716 	return (error);
2717 }
2718 
2719 /*
2720  * Common routine to set mode given a vnode.
2721  */
2722 static int
2723 change_mode(struct vnode *vp, int mode, struct lwp *l)
2724 {
2725 	struct vattr vattr;
2726 	int error;
2727 
2728 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2729 	vattr_null(&vattr);
2730 	vattr.va_mode = mode & ALLPERMS;
2731 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2732 	VOP_UNLOCK(vp);
2733 	return (error);
2734 }
2735 
2736 /*
2737  * Set ownership given a path name; this version follows links.
2738  */
2739 /* ARGSUSED */
2740 int
2741 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
2742 {
2743 	/* {
2744 		syscallarg(const char *) path;
2745 		syscallarg(uid_t) uid;
2746 		syscallarg(gid_t) gid;
2747 	} */
2748 	int error;
2749 	struct vnode *vp;
2750 
2751 	error = namei_simple_user(SCARG(uap, path),
2752 				NSM_FOLLOW_TRYEMULROOT, &vp);
2753 	if (error != 0)
2754 		return (error);
2755 
2756 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2757 
2758 	vrele(vp);
2759 	return (error);
2760 }
2761 
2762 /*
2763  * Set ownership given a path name; this version follows links.
2764  * Provides POSIX semantics.
2765  */
2766 /* ARGSUSED */
2767 int
2768 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
2769 {
2770 	/* {
2771 		syscallarg(const char *) path;
2772 		syscallarg(uid_t) uid;
2773 		syscallarg(gid_t) gid;
2774 	} */
2775 	int error;
2776 	struct vnode *vp;
2777 
2778 	error = namei_simple_user(SCARG(uap, path),
2779 				NSM_FOLLOW_TRYEMULROOT, &vp);
2780 	if (error != 0)
2781 		return (error);
2782 
2783 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2784 
2785 	vrele(vp);
2786 	return (error);
2787 }
2788 
2789 /*
2790  * Set ownership given a file descriptor.
2791  */
2792 /* ARGSUSED */
2793 int
2794 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
2795 {
2796 	/* {
2797 		syscallarg(int) fd;
2798 		syscallarg(uid_t) uid;
2799 		syscallarg(gid_t) gid;
2800 	} */
2801 	int error;
2802 	file_t *fp;
2803 
2804 	/* fd_getvnode() will use the descriptor for us */
2805 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2806 		return (error);
2807 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2808 	    l, 0);
2809 	fd_putfile(SCARG(uap, fd));
2810 	return (error);
2811 }
2812 
2813 /*
2814  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2815  */
2816 /* ARGSUSED */
2817 int
2818 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
2819 {
2820 	/* {
2821 		syscallarg(int) fd;
2822 		syscallarg(uid_t) uid;
2823 		syscallarg(gid_t) gid;
2824 	} */
2825 	int error;
2826 	file_t *fp;
2827 
2828 	/* fd_getvnode() will use the descriptor for us */
2829 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2830 		return (error);
2831 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2832 	    l, 1);
2833 	fd_putfile(SCARG(uap, fd));
2834 	return (error);
2835 }
2836 
2837 /*
2838  * Set ownership given a path name; this version does not follow links.
2839  */
2840 /* ARGSUSED */
2841 int
2842 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
2843 {
2844 	/* {
2845 		syscallarg(const char *) path;
2846 		syscallarg(uid_t) uid;
2847 		syscallarg(gid_t) gid;
2848 	} */
2849 	int error;
2850 	struct vnode *vp;
2851 
2852 	error = namei_simple_user(SCARG(uap, path),
2853 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2854 	if (error != 0)
2855 		return (error);
2856 
2857 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2858 
2859 	vrele(vp);
2860 	return (error);
2861 }
2862 
2863 /*
2864  * Set ownership given a path name; this version does not follow links.
2865  * Provides POSIX/XPG semantics.
2866  */
2867 /* ARGSUSED */
2868 int
2869 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
2870 {
2871 	/* {
2872 		syscallarg(const char *) path;
2873 		syscallarg(uid_t) uid;
2874 		syscallarg(gid_t) gid;
2875 	} */
2876 	int error;
2877 	struct vnode *vp;
2878 
2879 	error = namei_simple_user(SCARG(uap, path),
2880 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2881 	if (error != 0)
2882 		return (error);
2883 
2884 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2885 
2886 	vrele(vp);
2887 	return (error);
2888 }
2889 
2890 /*
2891  * Common routine to set ownership given a vnode.
2892  */
2893 static int
2894 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2895     int posix_semantics)
2896 {
2897 	struct vattr vattr;
2898 	mode_t newmode;
2899 	int error;
2900 
2901 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2902 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2903 		goto out;
2904 
2905 #define CHANGED(x) ((int)(x) != -1)
2906 	newmode = vattr.va_mode;
2907 	if (posix_semantics) {
2908 		/*
2909 		 * POSIX/XPG semantics: if the caller is not the super-user,
2910 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2911 		 * the XPG consider the behaviour for calls by the super-user
2912 		 * implementation-defined; we leave the set-user-id and set-
2913 		 * group-id settings intact in that case.
2914 		 */
2915 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2916 				      NULL) != 0)
2917 			newmode &= ~(S_ISUID | S_ISGID);
2918 	} else {
2919 		/*
2920 		 * NetBSD semantics: when changing owner and/or group,
2921 		 * clear the respective bit(s).
2922 		 */
2923 		if (CHANGED(uid))
2924 			newmode &= ~S_ISUID;
2925 		if (CHANGED(gid))
2926 			newmode &= ~S_ISGID;
2927 	}
2928 	/* Update va_mode iff altered. */
2929 	if (vattr.va_mode == newmode)
2930 		newmode = VNOVAL;
2931 
2932 	vattr_null(&vattr);
2933 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2934 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2935 	vattr.va_mode = newmode;
2936 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2937 #undef CHANGED
2938 
2939 out:
2940 	VOP_UNLOCK(vp);
2941 	return (error);
2942 }
2943 
2944 /*
2945  * Set the access and modification times given a path name; this
2946  * version follows links.
2947  */
2948 /* ARGSUSED */
2949 int
2950 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
2951     register_t *retval)
2952 {
2953 	/* {
2954 		syscallarg(const char *) path;
2955 		syscallarg(const struct timeval *) tptr;
2956 	} */
2957 
2958 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
2959 	    SCARG(uap, tptr), UIO_USERSPACE);
2960 }
2961 
2962 /*
2963  * Set the access and modification times given a file descriptor.
2964  */
2965 /* ARGSUSED */
2966 int
2967 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
2968     register_t *retval)
2969 {
2970 	/* {
2971 		syscallarg(int) fd;
2972 		syscallarg(const struct timeval *) tptr;
2973 	} */
2974 	int error;
2975 	file_t *fp;
2976 
2977 	/* fd_getvnode() will use the descriptor for us */
2978 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2979 		return (error);
2980 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
2981 	    UIO_USERSPACE);
2982 	fd_putfile(SCARG(uap, fd));
2983 	return (error);
2984 }
2985 
2986 /*
2987  * Set the access and modification times given a path name; this
2988  * version does not follow links.
2989  */
2990 int
2991 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
2992     register_t *retval)
2993 {
2994 	/* {
2995 		syscallarg(const char *) path;
2996 		syscallarg(const struct timeval *) tptr;
2997 	} */
2998 
2999 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
3000 	    SCARG(uap, tptr), UIO_USERSPACE);
3001 }
3002 
3003 /*
3004  * Common routine to set access and modification times given a vnode.
3005  */
3006 int
3007 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3008     const struct timeval *tptr, enum uio_seg seg)
3009 {
3010 	struct vattr vattr;
3011 	int error, dorele = 0;
3012 	namei_simple_flags_t sflags;
3013 
3014 	bool vanull, setbirthtime;
3015 	struct timespec ts[2];
3016 
3017 	/*
3018 	 * I have checked all callers and they pass either FOLLOW,
3019 	 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
3020 	 * is 0. More to the point, they don't pass anything else.
3021 	 * Let's keep it that way at least until the namei interfaces
3022 	 * are fully sanitized.
3023 	 */
3024 	KASSERT(flag == NOFOLLOW || flag == FOLLOW);
3025 	sflags = (flag == FOLLOW) ?
3026 		NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3027 
3028 	if (tptr == NULL) {
3029 		vanull = true;
3030 		nanotime(&ts[0]);
3031 		ts[1] = ts[0];
3032 	} else {
3033 		struct timeval tv[2];
3034 
3035 		vanull = false;
3036 		if (seg != UIO_SYSSPACE) {
3037 			error = copyin(tptr, tv, sizeof (tv));
3038 			if (error != 0)
3039 				return error;
3040 			tptr = tv;
3041 		}
3042 		TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3043 		TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3044 	}
3045 
3046 	if (vp == NULL) {
3047 		/* note: SEG describes TPTR, not PATH; PATH is always user */
3048 		error = namei_simple_user(path, sflags, &vp);
3049 		if (error != 0)
3050 			return error;
3051 		dorele = 1;
3052 	}
3053 
3054 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3055 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3056 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
3057 	vattr_null(&vattr);
3058 	vattr.va_atime = ts[0];
3059 	vattr.va_mtime = ts[1];
3060 	if (setbirthtime)
3061 		vattr.va_birthtime = ts[1];
3062 	if (vanull)
3063 		vattr.va_vaflags |= VA_UTIMES_NULL;
3064 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3065 	VOP_UNLOCK(vp);
3066 
3067 	if (dorele != 0)
3068 		vrele(vp);
3069 
3070 	return error;
3071 }
3072 
3073 /*
3074  * Truncate a file given its path name.
3075  */
3076 /* ARGSUSED */
3077 int
3078 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3079 {
3080 	/* {
3081 		syscallarg(const char *) path;
3082 		syscallarg(int) pad;
3083 		syscallarg(off_t) length;
3084 	} */
3085 	struct vnode *vp;
3086 	struct vattr vattr;
3087 	int error;
3088 
3089 	error = namei_simple_user(SCARG(uap, path),
3090 				NSM_FOLLOW_TRYEMULROOT, &vp);
3091 	if (error != 0)
3092 		return (error);
3093 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3094 	if (vp->v_type == VDIR)
3095 		error = EISDIR;
3096 	else if ((error = vn_writechk(vp)) == 0 &&
3097 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3098 		vattr_null(&vattr);
3099 		vattr.va_size = SCARG(uap, length);
3100 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3101 	}
3102 	vput(vp);
3103 	return (error);
3104 }
3105 
3106 /*
3107  * Truncate a file given a file descriptor.
3108  */
3109 /* ARGSUSED */
3110 int
3111 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3112 {
3113 	/* {
3114 		syscallarg(int) fd;
3115 		syscallarg(int) pad;
3116 		syscallarg(off_t) length;
3117 	} */
3118 	struct vattr vattr;
3119 	struct vnode *vp;
3120 	file_t *fp;
3121 	int error;
3122 
3123 	/* fd_getvnode() will use the descriptor for us */
3124 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3125 		return (error);
3126 	if ((fp->f_flag & FWRITE) == 0) {
3127 		error = EINVAL;
3128 		goto out;
3129 	}
3130 	vp = fp->f_data;
3131 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3132 	if (vp->v_type == VDIR)
3133 		error = EISDIR;
3134 	else if ((error = vn_writechk(vp)) == 0) {
3135 		vattr_null(&vattr);
3136 		vattr.va_size = SCARG(uap, length);
3137 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3138 	}
3139 	VOP_UNLOCK(vp);
3140  out:
3141 	fd_putfile(SCARG(uap, fd));
3142 	return (error);
3143 }
3144 
3145 /*
3146  * Sync an open file.
3147  */
3148 /* ARGSUSED */
3149 int
3150 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3151 {
3152 	/* {
3153 		syscallarg(int) fd;
3154 	} */
3155 	struct vnode *vp;
3156 	file_t *fp;
3157 	int error;
3158 
3159 	/* fd_getvnode() will use the descriptor for us */
3160 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3161 		return (error);
3162 	vp = fp->f_data;
3163 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3164 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3165 	VOP_UNLOCK(vp);
3166 	fd_putfile(SCARG(uap, fd));
3167 	return (error);
3168 }
3169 
3170 /*
3171  * Sync a range of file data.  API modeled after that found in AIX.
3172  *
3173  * FDATASYNC indicates that we need only save enough metadata to be able
3174  * to re-read the written data.  Note we duplicate AIX's requirement that
3175  * the file be open for writing.
3176  */
3177 /* ARGSUSED */
3178 int
3179 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3180 {
3181 	/* {
3182 		syscallarg(int) fd;
3183 		syscallarg(int) flags;
3184 		syscallarg(off_t) start;
3185 		syscallarg(off_t) length;
3186 	} */
3187 	struct vnode *vp;
3188 	file_t *fp;
3189 	int flags, nflags;
3190 	off_t s, e, len;
3191 	int error;
3192 
3193 	/* fd_getvnode() will use the descriptor for us */
3194 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3195 		return (error);
3196 
3197 	if ((fp->f_flag & FWRITE) == 0) {
3198 		error = EBADF;
3199 		goto out;
3200 	}
3201 
3202 	flags = SCARG(uap, flags);
3203 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3204 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3205 		error = EINVAL;
3206 		goto out;
3207 	}
3208 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3209 	if (flags & FDATASYNC)
3210 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3211 	else
3212 		nflags = FSYNC_WAIT;
3213 	if (flags & FDISKSYNC)
3214 		nflags |= FSYNC_CACHE;
3215 
3216 	len = SCARG(uap, length);
3217 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3218 	if (len) {
3219 		s = SCARG(uap, start);
3220 		e = s + len;
3221 		if (e < s) {
3222 			error = EINVAL;
3223 			goto out;
3224 		}
3225 	} else {
3226 		e = 0;
3227 		s = 0;
3228 	}
3229 
3230 	vp = fp->f_data;
3231 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3232 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3233 	VOP_UNLOCK(vp);
3234 out:
3235 	fd_putfile(SCARG(uap, fd));
3236 	return (error);
3237 }
3238 
3239 /*
3240  * Sync the data of an open file.
3241  */
3242 /* ARGSUSED */
3243 int
3244 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3245 {
3246 	/* {
3247 		syscallarg(int) fd;
3248 	} */
3249 	struct vnode *vp;
3250 	file_t *fp;
3251 	int error;
3252 
3253 	/* fd_getvnode() will use the descriptor for us */
3254 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3255 		return (error);
3256 	if ((fp->f_flag & FWRITE) == 0) {
3257 		fd_putfile(SCARG(uap, fd));
3258 		return (EBADF);
3259 	}
3260 	vp = fp->f_data;
3261 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3262 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3263 	VOP_UNLOCK(vp);
3264 	fd_putfile(SCARG(uap, fd));
3265 	return (error);
3266 }
3267 
3268 /*
3269  * Rename files, (standard) BSD semantics frontend.
3270  */
3271 /* ARGSUSED */
3272 int
3273 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3274 {
3275 	/* {
3276 		syscallarg(const char *) from;
3277 		syscallarg(const char *) to;
3278 	} */
3279 
3280 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3281 }
3282 
3283 /*
3284  * Rename files, POSIX semantics frontend.
3285  */
3286 /* ARGSUSED */
3287 int
3288 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3289 {
3290 	/* {
3291 		syscallarg(const char *) from;
3292 		syscallarg(const char *) to;
3293 	} */
3294 
3295 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3296 }
3297 
3298 /*
3299  * Rename files.  Source and destination must either both be directories,
3300  * or both not be directories.  If target is a directory, it must be empty.
3301  * If `from' and `to' refer to the same object, the value of the `retain'
3302  * argument is used to determine whether `from' will be
3303  *
3304  * (retain == 0)	deleted unless `from' and `to' refer to the same
3305  *			object in the file system's name space (BSD).
3306  * (retain == 1)	always retained (POSIX).
3307  */
3308 int
3309 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3310 {
3311 	struct vnode *tvp, *fvp, *tdvp;
3312 	struct nameidata fromnd, tond;
3313 	struct mount *fs;
3314 	struct lwp *l = curlwp;
3315 	struct proc *p;
3316 	uint32_t saveflag;
3317 	int error;
3318 
3319 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT | INRENAME,
3320 	    seg, from);
3321 	if ((error = namei(&fromnd)) != 0)
3322 		return (error);
3323 	if (fromnd.ni_dvp != fromnd.ni_vp)
3324 		VOP_UNLOCK(fromnd.ni_dvp);
3325 	fvp = fromnd.ni_vp;
3326 
3327 	fs = fvp->v_mount;
3328 	error = VFS_RENAMELOCK_ENTER(fs);
3329 	if (error) {
3330 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3331 		vrele(fromnd.ni_dvp);
3332 		vrele(fvp);
3333 		goto out1;
3334 	}
3335 
3336 	/*
3337 	 * close, partially, yet another race - ideally we should only
3338 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3339 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3340 	 * that with namei as it stands.
3341 	 *
3342 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3343 	 * under us. The real fix is to get the locks in the right
3344 	 * order and do the lookups in the right places, but that's a
3345 	 * major rototill.
3346 	 *
3347 	 * Preserve the SAVESTART in cn_flags, because who knows what
3348 	 * might happen if we don't.
3349 	 *
3350 	 * Note: this logic (as well as this whole function) is cloned
3351 	 * in nfs_serv.c. Proceed accordingly.
3352 	 */
3353 	vrele(fvp);
3354 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3355 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3356 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3357 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3358 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3359 		error = EINVAL;
3360 		VFS_RENAMELOCK_EXIT(fs);
3361 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3362 		vrele(fromnd.ni_dvp);
3363 		goto out1;
3364 	}
3365 	saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
3366 	fromnd.ni_cnd.cn_flags &= ~SAVESTART;
3367 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3368 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
3369 	fromnd.ni_cnd.cn_flags |= saveflag;
3370 	if (error) {
3371 		VOP_UNLOCK(fromnd.ni_dvp);
3372 		VFS_RENAMELOCK_EXIT(fs);
3373 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3374 		vrele(fromnd.ni_dvp);
3375 		goto out1;
3376 	}
3377 	VOP_UNLOCK(fromnd.ni_vp);
3378 	if (fromnd.ni_dvp != fromnd.ni_vp)
3379 		VOP_UNLOCK(fromnd.ni_dvp);
3380 	fvp = fromnd.ni_vp;
3381 
3382 	NDINIT(&tond, RENAME,
3383 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3384 	      | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0),
3385 	    seg, to);
3386 	if ((error = namei(&tond)) != 0) {
3387 		VFS_RENAMELOCK_EXIT(fs);
3388 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3389 		vrele(fromnd.ni_dvp);
3390 		vrele(fvp);
3391 		goto out1;
3392 	}
3393 	tdvp = tond.ni_dvp;
3394 	tvp = tond.ni_vp;
3395 
3396 	if (tvp != NULL) {
3397 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3398 			error = ENOTDIR;
3399 			goto out;
3400 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3401 			error = EISDIR;
3402 			goto out;
3403 		}
3404 	}
3405 
3406 	if (fvp == tdvp)
3407 		error = EINVAL;
3408 
3409 	/*
3410 	 * Source and destination refer to the same object.
3411 	 */
3412 	if (fvp == tvp) {
3413 		if (retain)
3414 			error = -1;
3415 		else if (fromnd.ni_dvp == tdvp &&
3416 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3417 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3418 		          tond.ni_cnd.cn_nameptr,
3419 		          fromnd.ni_cnd.cn_namelen))
3420 		error = -1;
3421 	}
3422 
3423 #if NVERIEXEC > 0
3424 	if (!error) {
3425 		char *f1, *f2;
3426 		size_t f1_len;
3427 		size_t f2_len;
3428 
3429 		f1_len = fromnd.ni_cnd.cn_namelen + 1;
3430 		f1 = kmem_alloc(f1_len, KM_SLEEP);
3431 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
3432 
3433 		f2_len = tond.ni_cnd.cn_namelen + 1;
3434 		f2 = kmem_alloc(f2_len, KM_SLEEP);
3435 		strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
3436 
3437 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
3438 
3439 		kmem_free(f1, f1_len);
3440 		kmem_free(f2, f2_len);
3441 	}
3442 #endif /* NVERIEXEC > 0 */
3443 
3444 out:
3445 	p = l->l_proc;
3446 	if (!error) {
3447 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3448 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3449 		VFS_RENAMELOCK_EXIT(fs);
3450 	} else {
3451 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3452 		if (tdvp == tvp)
3453 			vrele(tdvp);
3454 		else
3455 			vput(tdvp);
3456 		if (tvp)
3457 			vput(tvp);
3458 		VFS_RENAMELOCK_EXIT(fs);
3459 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3460 		vrele(fromnd.ni_dvp);
3461 		vrele(fvp);
3462 	}
3463 	vrele(tond.ni_startdir);
3464 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3465 out1:
3466 	if (fromnd.ni_startdir)
3467 		vrele(fromnd.ni_startdir);
3468 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3469 	return (error == -1 ? 0 : error);
3470 }
3471 
3472 /*
3473  * Make a directory file.
3474  */
3475 /* ARGSUSED */
3476 int
3477 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
3478 {
3479 	/* {
3480 		syscallarg(const char *) path;
3481 		syscallarg(int) mode;
3482 	} */
3483 
3484 	return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE);
3485 }
3486 
3487 int
3488 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
3489 {
3490 	struct proc *p = curlwp->l_proc;
3491 	struct vnode *vp;
3492 	struct vattr vattr;
3493 	int error;
3494 	struct nameidata nd;
3495 
3496 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT,
3497 	    seg, path);
3498 	if ((error = namei(&nd)) != 0)
3499 		return (error);
3500 	vp = nd.ni_vp;
3501 	if (vp != NULL) {
3502 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3503 		if (nd.ni_dvp == vp)
3504 			vrele(nd.ni_dvp);
3505 		else
3506 			vput(nd.ni_dvp);
3507 		vrele(vp);
3508 		return (EEXIST);
3509 	}
3510 	vattr_null(&vattr);
3511 	vattr.va_type = VDIR;
3512 	/* We will read cwdi->cwdi_cmask unlocked. */
3513 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3514 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3515 	if (!error)
3516 		vput(nd.ni_vp);
3517 	return (error);
3518 }
3519 
3520 /*
3521  * Remove a directory file.
3522  */
3523 /* ARGSUSED */
3524 int
3525 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
3526 {
3527 	/* {
3528 		syscallarg(const char *) path;
3529 	} */
3530 	struct vnode *vp;
3531 	int error;
3532 	struct nameidata nd;
3533 
3534 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
3535 	    SCARG(uap, path));
3536 	if ((error = namei(&nd)) != 0)
3537 		return (error);
3538 	vp = nd.ni_vp;
3539 	if (vp->v_type != VDIR) {
3540 		error = ENOTDIR;
3541 		goto out;
3542 	}
3543 	/*
3544 	 * No rmdir "." please.
3545 	 */
3546 	if (nd.ni_dvp == vp) {
3547 		error = EINVAL;
3548 		goto out;
3549 	}
3550 	/*
3551 	 * The root of a mounted filesystem cannot be deleted.
3552 	 */
3553 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
3554 		error = EBUSY;
3555 		goto out;
3556 	}
3557 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3558 	return (error);
3559 
3560 out:
3561 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3562 	if (nd.ni_dvp == vp)
3563 		vrele(nd.ni_dvp);
3564 	else
3565 		vput(nd.ni_dvp);
3566 	vput(vp);
3567 	return (error);
3568 }
3569 
3570 /*
3571  * Read a block of directory entries in a file system independent format.
3572  */
3573 int
3574 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
3575 {
3576 	/* {
3577 		syscallarg(int) fd;
3578 		syscallarg(char *) buf;
3579 		syscallarg(size_t) count;
3580 	} */
3581 	file_t *fp;
3582 	int error, done;
3583 
3584 	/* fd_getvnode() will use the descriptor for us */
3585 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3586 		return (error);
3587 	if ((fp->f_flag & FREAD) == 0) {
3588 		error = EBADF;
3589 		goto out;
3590 	}
3591 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3592 			SCARG(uap, count), &done, l, 0, 0);
3593 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3594 	*retval = done;
3595  out:
3596 	fd_putfile(SCARG(uap, fd));
3597 	return (error);
3598 }
3599 
3600 /*
3601  * Set the mode mask for creation of filesystem nodes.
3602  */
3603 int
3604 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
3605 {
3606 	/* {
3607 		syscallarg(mode_t) newmask;
3608 	} */
3609 	struct proc *p = l->l_proc;
3610 	struct cwdinfo *cwdi;
3611 
3612 	/*
3613 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3614 	 * important is that we serialize changes to the mask.  The
3615 	 * rw_exit() will issue a write memory barrier on our behalf,
3616 	 * and force the changes out to other CPUs (as it must use an
3617 	 * atomic operation, draining the local CPU's store buffers).
3618 	 */
3619 	cwdi = p->p_cwdi;
3620 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3621 	*retval = cwdi->cwdi_cmask;
3622 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3623 	rw_exit(&cwdi->cwdi_lock);
3624 
3625 	return (0);
3626 }
3627 
3628 int
3629 dorevoke(struct vnode *vp, kauth_cred_t cred)
3630 {
3631 	struct vattr vattr;
3632 	int error;
3633 
3634 	if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
3635 		return error;
3636 	if (kauth_cred_geteuid(cred) == vattr.va_uid ||
3637 	    (error = kauth_authorize_generic(cred,
3638 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
3639 		VOP_REVOKE(vp, REVOKEALL);
3640 	return (error);
3641 }
3642 
3643 /*
3644  * Void all references to file by ripping underlying filesystem
3645  * away from vnode.
3646  */
3647 /* ARGSUSED */
3648 int
3649 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
3650 {
3651 	/* {
3652 		syscallarg(const char *) path;
3653 	} */
3654 	struct vnode *vp;
3655 	int error;
3656 
3657 	error = namei_simple_user(SCARG(uap, path),
3658 				NSM_FOLLOW_TRYEMULROOT, &vp);
3659 	if (error != 0)
3660 		return (error);
3661 	error = dorevoke(vp, l->l_cred);
3662 	vrele(vp);
3663 	return (error);
3664 }
3665