xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: vfs_syscalls.c,v 1.408 2010/08/21 13:19:39 pgoyette 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 1.408 2010/08/21 13:19:39 pgoyette 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 	(void)module_autoload(fstypename, MODULE_CLASS_VFS);
282 
283 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
284 		return 0;
285 
286 	return ENODEV;
287 }
288 
289 static int
290 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
291     const char *path, int flags, void *data, size_t *data_len)
292 {
293 	struct mount *mp;
294 	struct vnode *vp = *vpp;
295 	struct vattr va;
296 	struct nameidata nd;
297 	int error;
298 
299 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
300 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
301 	if (error) {
302 		vfs_delref(vfsops);
303 		return error;
304 	}
305 
306 	/* Can't make a non-dir a mount-point (from here anyway). */
307 	if (vp->v_type != VDIR) {
308 		vfs_delref(vfsops);
309 		return ENOTDIR;
310 	}
311 
312 	/*
313 	 * If the user is not root, ensure that they own the directory
314 	 * onto which we are attempting to mount.
315 	 */
316 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
317 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
318 	    (error = kauth_authorize_generic(l->l_cred,
319 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
320 		vfs_delref(vfsops);
321 		return error;
322 	}
323 
324 	if (flags & MNT_EXPORTED) {
325 		vfs_delref(vfsops);
326 		return EINVAL;
327 	}
328 
329 	if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) {
330 		vfs_delref(vfsops);
331 		return ENOMEM;
332 	}
333 
334 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
335 
336 	/*
337 	 * The underlying file system may refuse the mount for
338 	 * various reasons.  Allow the user to force it to happen.
339 	 *
340 	 * Set the mount level flags.
341 	 */
342 	mp->mnt_flag = flags &
343 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
344 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
345 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
346 	    MNT_LOG | MNT_IGNORE | MNT_RDONLY);
347 
348 	mutex_enter(&mp->mnt_updating);
349 	error = VFS_MOUNT(mp, path, data, data_len);
350 	mp->mnt_flag &= ~MNT_OP_FLAGS;
351 
352 	if (error != 0)
353 		goto err_unmounted;
354 
355 	/*
356 	 * Validate and prepare the mount point.
357 	 */
358 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT,
359 	    UIO_USERSPACE, path);
360 	error = namei(&nd);
361 	if (error != 0) {
362 		goto err_mounted;
363 	}
364 	if (nd.ni_vp != vp) {
365 		vput(nd.ni_vp);
366 		error = EINVAL;
367 		goto err_mounted;
368 	}
369 	if (vp->v_mountedhere != NULL) {
370 		vput(nd.ni_vp);
371 		error = EBUSY;
372 		goto err_mounted;
373 	}
374 	error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0);
375 	if (error != 0) {
376 		vput(nd.ni_vp);
377 		goto err_mounted;
378 	}
379 
380 	/*
381 	 * Put the new filesystem on the mount list after root.
382 	 */
383 	cache_purge(vp);
384 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
385 
386 	mutex_enter(&mountlist_lock);
387 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
388 	mutex_exit(&mountlist_lock);
389 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
390 		error = vfs_allocate_syncvnode(mp);
391 	if (error == 0)
392 		vp->v_mountedhere = mp;
393 	vput(nd.ni_vp);
394 	if (error != 0)
395 		goto err_onmountlist;
396 
397 	checkdirs(vp);
398 	mutex_exit(&mp->mnt_updating);
399 
400 	/* Hold an additional reference to the mount across VFS_START(). */
401 	vfs_unbusy(mp, true, NULL);
402 	(void) VFS_STATVFS(mp, &mp->mnt_stat);
403 	error = VFS_START(mp, 0);
404 	if (error)
405 		vrele(vp);
406 	/* Drop reference held for VFS_START(). */
407 	vfs_destroy(mp);
408 	*vpp = NULL;
409 	return error;
410 
411 err_onmountlist:
412 	mutex_enter(&mountlist_lock);
413 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
414 	mp->mnt_iflag |= IMNT_GONE;
415 	mutex_exit(&mountlist_lock);
416 
417 err_mounted:
418 	if (VFS_UNMOUNT(mp, MNT_FORCE) != 0)
419 		panic("Unmounting fresh file system failed");
420 
421 err_unmounted:
422 	vp->v_mountedhere = NULL;
423 	mutex_exit(&mp->mnt_updating);
424 	vfs_unbusy(mp, false, NULL);
425 	vfs_destroy(mp);
426 
427 	return error;
428 }
429 
430 static int
431 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
432     void *data, size_t *data_len)
433 {
434 	struct mount *mp;
435 	int error;
436 
437 	/* If MNT_GETARGS is specified, it should be the only flag. */
438 	if (flags & ~MNT_GETARGS)
439 		return EINVAL;
440 
441 	mp = vp->v_mount;
442 
443 	/* XXX: probably some notion of "can see" here if we want isolation. */
444 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
445 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
446 	if (error)
447 		return error;
448 
449 	if ((vp->v_vflag & VV_ROOT) == 0)
450 		return EINVAL;
451 
452 	if (vfs_busy(mp, NULL))
453 		return EPERM;
454 
455 	mutex_enter(&mp->mnt_updating);
456 	mp->mnt_flag &= ~MNT_OP_FLAGS;
457 	mp->mnt_flag |= MNT_GETARGS;
458 	error = VFS_MOUNT(mp, path, data, data_len);
459 	mp->mnt_flag &= ~MNT_OP_FLAGS;
460 	mutex_exit(&mp->mnt_updating);
461 
462 	vfs_unbusy(mp, false, NULL);
463 	return (error);
464 }
465 
466 int
467 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
468 {
469 	/* {
470 		syscallarg(const char *) type;
471 		syscallarg(const char *) path;
472 		syscallarg(int) flags;
473 		syscallarg(void *) data;
474 		syscallarg(size_t) data_len;
475 	} */
476 
477 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
478 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
479 	    SCARG(uap, data_len), retval);
480 }
481 
482 int
483 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
484     const char *path, int flags, void *data, enum uio_seg data_seg,
485     size_t data_len, register_t *retval)
486 {
487 	struct vnode *vp;
488 	void *data_buf = data;
489 	bool vfsopsrele = false;
490 	int error;
491 
492 	/* XXX: The calling convention of this routine is totally bizarre */
493 	if (vfsops)
494 		vfsopsrele = true;
495 
496 	/*
497 	 * Get vnode to be covered
498 	 */
499 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
500 	if (error != 0) {
501 		vp = NULL;
502 		goto done;
503 	}
504 
505 	if (vfsops == NULL) {
506 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
507 			vfsops = vp->v_mount->mnt_op;
508 		} else {
509 			/* 'type' is userspace */
510 			error = mount_get_vfsops(type, &vfsops);
511 			if (error != 0)
512 				goto done;
513 			vfsopsrele = true;
514 		}
515 	}
516 
517 	if (data != NULL && data_seg == UIO_USERSPACE) {
518 		if (data_len == 0) {
519 			/* No length supplied, use default for filesystem */
520 			data_len = vfsops->vfs_min_mount_data;
521 			if (data_len > VFS_MAX_MOUNT_DATA) {
522 				error = EINVAL;
523 				goto done;
524 			}
525 			/*
526 			 * Hopefully a longer buffer won't make copyin() fail.
527 			 * For compatibility with 3.0 and earlier.
528 			 */
529 			if (flags & MNT_UPDATE
530 			    && data_len < sizeof (struct mnt_export_args30))
531 				data_len = sizeof (struct mnt_export_args30);
532 		}
533 		data_buf = kmem_alloc(data_len, KM_SLEEP);
534 
535 		/* NFS needs the buffer even for mnt_getargs .... */
536 		error = copyin(data, data_buf, data_len);
537 		if (error != 0)
538 			goto done;
539 	}
540 
541 	if (flags & MNT_GETARGS) {
542 		if (data_len == 0) {
543 			error = EINVAL;
544 			goto done;
545 		}
546 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
547 		if (error != 0)
548 			goto done;
549 		if (data_seg == UIO_USERSPACE)
550 			error = copyout(data_buf, data, data_len);
551 		*retval = data_len;
552 	} else if (flags & MNT_UPDATE) {
553 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
554 	} else {
555 		/* Locking is handled internally in mount_domount(). */
556 		KASSERT(vfsopsrele == true);
557 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
558 		    &data_len);
559 		vfsopsrele = false;
560 	}
561 
562     done:
563 	if (vfsopsrele)
564 		vfs_delref(vfsops);
565     	if (vp != NULL) {
566 	    	vrele(vp);
567 	}
568 	if (data_buf != data)
569 		kmem_free(data_buf, data_len);
570 	return (error);
571 }
572 
573 /*
574  * Scan all active processes to see if any of them have a current
575  * or root directory onto which the new filesystem has just been
576  * mounted. If so, replace them with the new mount point.
577  */
578 void
579 checkdirs(struct vnode *olddp)
580 {
581 	struct cwdinfo *cwdi;
582 	struct vnode *newdp, *rele1, *rele2;
583 	struct proc *p;
584 	bool retry;
585 
586 	if (olddp->v_usecount == 1)
587 		return;
588 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
589 		panic("mount: lost mount");
590 
591 	do {
592 		retry = false;
593 		mutex_enter(proc_lock);
594 		PROCLIST_FOREACH(p, &allproc) {
595 			if ((cwdi = p->p_cwdi) == NULL)
596 				continue;
597 			/*
598 			 * Can't change to the old directory any more,
599 			 * so even if we see a stale value it's not a
600 			 * problem.
601 			 */
602 			if (cwdi->cwdi_cdir != olddp &&
603 			    cwdi->cwdi_rdir != olddp)
604 			    	continue;
605 			retry = true;
606 			rele1 = NULL;
607 			rele2 = NULL;
608 			atomic_inc_uint(&cwdi->cwdi_refcnt);
609 			mutex_exit(proc_lock);
610 			rw_enter(&cwdi->cwdi_lock, RW_WRITER);
611 			if (cwdi->cwdi_cdir == olddp) {
612 				rele1 = cwdi->cwdi_cdir;
613 				vref(newdp);
614 				cwdi->cwdi_cdir = newdp;
615 			}
616 			if (cwdi->cwdi_rdir == olddp) {
617 				rele2 = cwdi->cwdi_rdir;
618 				vref(newdp);
619 				cwdi->cwdi_rdir = newdp;
620 			}
621 			rw_exit(&cwdi->cwdi_lock);
622 			cwdfree(cwdi);
623 			if (rele1 != NULL)
624 				vrele(rele1);
625 			if (rele2 != NULL)
626 				vrele(rele2);
627 			mutex_enter(proc_lock);
628 			break;
629 		}
630 		mutex_exit(proc_lock);
631 	} while (retry);
632 
633 	if (rootvnode == olddp) {
634 		vrele(rootvnode);
635 		vref(newdp);
636 		rootvnode = newdp;
637 	}
638 	vput(newdp);
639 }
640 
641 /*
642  * Unmount a file system.
643  *
644  * Note: unmount takes a path to the vnode mounted on as argument,
645  * not special file (as before).
646  */
647 /* ARGSUSED */
648 int
649 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
650 {
651 	/* {
652 		syscallarg(const char *) path;
653 		syscallarg(int) flags;
654 	} */
655 	struct vnode *vp;
656 	struct mount *mp;
657 	int error;
658 	struct nameidata nd;
659 
660 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
661 	    SCARG(uap, path));
662 	if ((error = namei(&nd)) != 0)
663 		return (error);
664 	vp = nd.ni_vp;
665 	mp = vp->v_mount;
666 	atomic_inc_uint(&mp->mnt_refcnt);
667 	VOP_UNLOCK(vp);
668 
669 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
670 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
671 	if (error) {
672 		vrele(vp);
673 		vfs_destroy(mp);
674 		return (error);
675 	}
676 
677 	/*
678 	 * Don't allow unmounting the root file system.
679 	 */
680 	if (mp->mnt_flag & MNT_ROOTFS) {
681 		vrele(vp);
682 		vfs_destroy(mp);
683 		return (EINVAL);
684 	}
685 
686 	/*
687 	 * Must be the root of the filesystem
688 	 */
689 	if ((vp->v_vflag & VV_ROOT) == 0) {
690 		vrele(vp);
691 		vfs_destroy(mp);
692 		return (EINVAL);
693 	}
694 
695 	vrele(vp);
696 	error = dounmount(mp, SCARG(uap, flags), l);
697 	vfs_destroy(mp);
698 	return error;
699 }
700 
701 /*
702  * Do the actual file system unmount.  File system is assumed to have
703  * been locked by the caller.
704  *
705  * => Caller hold reference to the mount, explicitly for dounmount().
706  */
707 int
708 dounmount(struct mount *mp, int flags, struct lwp *l)
709 {
710 	struct vnode *coveredvp;
711 	int error;
712 	int async;
713 	int used_syncer;
714 
715 #if NVERIEXEC > 0
716 	error = veriexec_unmountchk(mp);
717 	if (error)
718 		return (error);
719 #endif /* NVERIEXEC > 0 */
720 
721 	/*
722 	 * XXX Freeze syncer.  Must do this before locking the
723 	 * mount point.  See dounmount() for details.
724 	 */
725 	mutex_enter(&syncer_mutex);
726 	rw_enter(&mp->mnt_unmounting, RW_WRITER);
727 	if ((mp->mnt_iflag & IMNT_GONE) != 0) {
728 		rw_exit(&mp->mnt_unmounting);
729 		mutex_exit(&syncer_mutex);
730 		return ENOENT;
731 	}
732 
733 	used_syncer = (mp->mnt_syncer != NULL);
734 
735 	/*
736 	 * XXX Syncer must be frozen when we get here.  This should really
737 	 * be done on a per-mountpoint basis, but the syncer doesn't work
738 	 * like that.
739 	 *
740 	 * The caller of dounmount() must acquire syncer_mutex because
741 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
742 	 * order, and we must preserve that order to avoid deadlock.
743 	 *
744 	 * So, if the file system did not use the syncer, now is
745 	 * the time to release the syncer_mutex.
746 	 */
747 	if (used_syncer == 0)
748 		mutex_exit(&syncer_mutex);
749 
750 	mp->mnt_iflag |= IMNT_UNMOUNT;
751 	async = mp->mnt_flag & MNT_ASYNC;
752 	mp->mnt_flag &= ~MNT_ASYNC;
753 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
754 	if (mp->mnt_syncer != NULL)
755 		vfs_deallocate_syncvnode(mp);
756 	error = 0;
757 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
758 		error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
759 	}
760 	vfs_scrubvnlist(mp);
761 	if (error == 0 || (flags & MNT_FORCE))
762 		error = VFS_UNMOUNT(mp, flags);
763 	if (error) {
764 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
765 			(void) vfs_allocate_syncvnode(mp);
766 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
767 		mp->mnt_flag |= async;
768 		rw_exit(&mp->mnt_unmounting);
769 		if (used_syncer)
770 			mutex_exit(&syncer_mutex);
771 		return (error);
772 	}
773 	vfs_scrubvnlist(mp);
774 	mutex_enter(&mountlist_lock);
775 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
776 		coveredvp->v_mountedhere = NULL;
777 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
778 	mp->mnt_iflag |= IMNT_GONE;
779 	mutex_exit(&mountlist_lock);
780 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
781 		panic("unmount: dangling vnode");
782 	if (used_syncer)
783 		mutex_exit(&syncer_mutex);
784 	vfs_hooks_unmount(mp);
785 	rw_exit(&mp->mnt_unmounting);
786 	vfs_destroy(mp);	/* reference from mount() */
787 	if (coveredvp != NULLVP)
788 		vrele(coveredvp);
789 	return (0);
790 }
791 
792 /*
793  * Sync each mounted filesystem.
794  */
795 #ifdef DEBUG
796 int syncprt = 0;
797 struct ctldebug debug0 = { "syncprt", &syncprt };
798 #endif
799 
800 /* ARGSUSED */
801 int
802 sys_sync(struct lwp *l, const void *v, register_t *retval)
803 {
804 	struct mount *mp, *nmp;
805 	int asyncflag;
806 
807 	if (l == NULL)
808 		l = &lwp0;
809 
810 	mutex_enter(&mountlist_lock);
811 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
812 	     mp = nmp) {
813 		if (vfs_busy(mp, &nmp)) {
814 			continue;
815 		}
816 		mutex_enter(&mp->mnt_updating);
817 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
818 			asyncflag = mp->mnt_flag & MNT_ASYNC;
819 			mp->mnt_flag &= ~MNT_ASYNC;
820 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
821 			if (asyncflag)
822 				 mp->mnt_flag |= MNT_ASYNC;
823 		}
824 		mutex_exit(&mp->mnt_updating);
825 		vfs_unbusy(mp, false, &nmp);
826 	}
827 	mutex_exit(&mountlist_lock);
828 #ifdef DEBUG
829 	if (syncprt)
830 		vfs_bufstats();
831 #endif /* DEBUG */
832 	return (0);
833 }
834 
835 /*
836  * Change filesystem quotas.
837  */
838 /* ARGSUSED */
839 int
840 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
841 {
842 	/* {
843 		syscallarg(const char *) path;
844 		syscallarg(int) cmd;
845 		syscallarg(int) uid;
846 		syscallarg(void *) arg;
847 	} */
848 	struct mount *mp;
849 	int error;
850 	struct vnode *vp;
851 
852 	error = namei_simple_user(SCARG(uap, path),
853 				NSM_FOLLOW_TRYEMULROOT, &vp);
854 	if (error != 0)
855 		return (error);
856 	mp = vp->v_mount;
857 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
858 	    SCARG(uap, arg));
859 	vrele(vp);
860 	return (error);
861 }
862 
863 int
864 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
865     int root)
866 {
867 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
868 	int error = 0;
869 
870 	/*
871 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
872 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
873 	 * overrides MNT_NOWAIT.
874 	 */
875 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
876 	    (flags != MNT_WAIT && flags != 0)) {
877 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
878 		goto done;
879 	}
880 
881 	/* Get the filesystem stats now */
882 	memset(sp, 0, sizeof(*sp));
883 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
884 		return error;
885 	}
886 
887 	if (cwdi->cwdi_rdir == NULL)
888 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
889 done:
890 	if (cwdi->cwdi_rdir != NULL) {
891 		size_t len;
892 		char *bp;
893 		char c;
894 		char *path = PNBUF_GET();
895 
896 		bp = path + MAXPATHLEN;
897 		*--bp = '\0';
898 		rw_enter(&cwdi->cwdi_lock, RW_READER);
899 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
900 		    MAXPATHLEN / 2, 0, l);
901 		rw_exit(&cwdi->cwdi_lock);
902 		if (error) {
903 			PNBUF_PUT(path);
904 			return error;
905 		}
906 		len = strlen(bp);
907 		if (len != 1) {
908 			/*
909 			 * for mount points that are below our root, we can see
910 			 * them, so we fix up the pathname and return them. The
911 			 * rest we cannot see, so we don't allow viewing the
912 			 * data.
913 			 */
914 			if (strncmp(bp, sp->f_mntonname, len) == 0 &&
915 			    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
916 				(void)strlcpy(sp->f_mntonname,
917 				    c == '\0' ? "/" : &sp->f_mntonname[len],
918 				    sizeof(sp->f_mntonname));
919 			} else {
920 				if (root)
921 					(void)strlcpy(sp->f_mntonname, "/",
922 					    sizeof(sp->f_mntonname));
923 				else
924 					error = EPERM;
925 			}
926 		}
927 		PNBUF_PUT(path);
928 	}
929 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
930 	return error;
931 }
932 
933 /*
934  * Get filesystem statistics by path.
935  */
936 int
937 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
938 {
939 	struct mount *mp;
940 	int error;
941 	struct vnode *vp;
942 
943 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
944 	if (error != 0)
945 		return error;
946 	mp = vp->v_mount;
947 	error = dostatvfs(mp, sb, l, flags, 1);
948 	vrele(vp);
949 	return error;
950 }
951 
952 /* ARGSUSED */
953 int
954 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
955 {
956 	/* {
957 		syscallarg(const char *) path;
958 		syscallarg(struct statvfs *) buf;
959 		syscallarg(int) flags;
960 	} */
961 	struct statvfs *sb;
962 	int error;
963 
964 	sb = STATVFSBUF_GET();
965 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
966 	if (error == 0)
967 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
968 	STATVFSBUF_PUT(sb);
969 	return error;
970 }
971 
972 /*
973  * Get filesystem statistics by fd.
974  */
975 int
976 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
977 {
978 	file_t *fp;
979 	struct mount *mp;
980 	int error;
981 
982 	/* fd_getvnode() will use the descriptor for us */
983 	if ((error = fd_getvnode(fd, &fp)) != 0)
984 		return (error);
985 	mp = ((struct vnode *)fp->f_data)->v_mount;
986 	error = dostatvfs(mp, sb, curlwp, flags, 1);
987 	fd_putfile(fd);
988 	return error;
989 }
990 
991 /* ARGSUSED */
992 int
993 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
994 {
995 	/* {
996 		syscallarg(int) fd;
997 		syscallarg(struct statvfs *) buf;
998 		syscallarg(int) flags;
999 	} */
1000 	struct statvfs *sb;
1001 	int error;
1002 
1003 	sb = STATVFSBUF_GET();
1004 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1005 	if (error == 0)
1006 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1007 	STATVFSBUF_PUT(sb);
1008 	return error;
1009 }
1010 
1011 
1012 /*
1013  * Get statistics on all filesystems.
1014  */
1015 int
1016 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1017     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1018     register_t *retval)
1019 {
1020 	int root = 0;
1021 	struct proc *p = l->l_proc;
1022 	struct mount *mp, *nmp;
1023 	struct statvfs *sb;
1024 	size_t count, maxcount;
1025 	int error = 0;
1026 
1027 	sb = STATVFSBUF_GET();
1028 	maxcount = bufsize / entry_sz;
1029 	mutex_enter(&mountlist_lock);
1030 	count = 0;
1031 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1032 	     mp = nmp) {
1033 		if (vfs_busy(mp, &nmp)) {
1034 			continue;
1035 		}
1036 		if (sfsp && count < maxcount) {
1037 			error = dostatvfs(mp, sb, l, flags, 0);
1038 			if (error) {
1039 				vfs_unbusy(mp, false, &nmp);
1040 				error = 0;
1041 				continue;
1042 			}
1043 			error = copyfn(sb, sfsp, entry_sz);
1044 			if (error) {
1045 				vfs_unbusy(mp, false, NULL);
1046 				goto out;
1047 			}
1048 			sfsp = (char *)sfsp + entry_sz;
1049 			root |= strcmp(sb->f_mntonname, "/") == 0;
1050 		}
1051 		count++;
1052 		vfs_unbusy(mp, false, &nmp);
1053 	}
1054 	mutex_exit(&mountlist_lock);
1055 
1056 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
1057 		/*
1058 		 * fake a root entry
1059 		 */
1060 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1061 		    sb, l, flags, 1);
1062 		if (error != 0)
1063 			goto out;
1064 		if (sfsp) {
1065 			error = copyfn(sb, sfsp, entry_sz);
1066 			if (error != 0)
1067 				goto out;
1068 		}
1069 		count++;
1070 	}
1071 	if (sfsp && count > maxcount)
1072 		*retval = maxcount;
1073 	else
1074 		*retval = count;
1075 out:
1076 	STATVFSBUF_PUT(sb);
1077 	return error;
1078 }
1079 
1080 int
1081 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1082 {
1083 	/* {
1084 		syscallarg(struct statvfs *) buf;
1085 		syscallarg(size_t) bufsize;
1086 		syscallarg(int) flags;
1087 	} */
1088 
1089 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1090 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1091 }
1092 
1093 /*
1094  * Change current working directory to a given file descriptor.
1095  */
1096 /* ARGSUSED */
1097 int
1098 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1099 {
1100 	/* {
1101 		syscallarg(int) fd;
1102 	} */
1103 	struct proc *p = l->l_proc;
1104 	struct cwdinfo *cwdi;
1105 	struct vnode *vp, *tdp;
1106 	struct mount *mp;
1107 	file_t *fp;
1108 	int error, fd;
1109 
1110 	/* fd_getvnode() will use the descriptor for us */
1111 	fd = SCARG(uap, fd);
1112 	if ((error = fd_getvnode(fd, &fp)) != 0)
1113 		return (error);
1114 	vp = fp->f_data;
1115 
1116 	vref(vp);
1117 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1118 	if (vp->v_type != VDIR)
1119 		error = ENOTDIR;
1120 	else
1121 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1122 	if (error) {
1123 		vput(vp);
1124 		goto out;
1125 	}
1126 	while ((mp = vp->v_mountedhere) != NULL) {
1127 		error = vfs_busy(mp, NULL);
1128 		vput(vp);
1129 		if (error != 0)
1130 			goto out;
1131 		error = VFS_ROOT(mp, &tdp);
1132 		vfs_unbusy(mp, false, NULL);
1133 		if (error)
1134 			goto out;
1135 		vp = tdp;
1136 	}
1137 	VOP_UNLOCK(vp);
1138 
1139 	/*
1140 	 * Disallow changing to a directory not under the process's
1141 	 * current root directory (if there is one).
1142 	 */
1143 	cwdi = p->p_cwdi;
1144 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1145 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1146 		vrele(vp);
1147 		error = EPERM;	/* operation not permitted */
1148 	} else {
1149 		vrele(cwdi->cwdi_cdir);
1150 		cwdi->cwdi_cdir = vp;
1151 	}
1152 	rw_exit(&cwdi->cwdi_lock);
1153 
1154  out:
1155 	fd_putfile(fd);
1156 	return (error);
1157 }
1158 
1159 /*
1160  * Change this process's notion of the root directory to a given file
1161  * descriptor.
1162  */
1163 int
1164 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1165 {
1166 	struct proc *p = l->l_proc;
1167 	struct vnode	*vp;
1168 	file_t	*fp;
1169 	int		 error, fd = SCARG(uap, fd);
1170 
1171 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1172  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1173 		return error;
1174 	/* fd_getvnode() will use the descriptor for us */
1175 	if ((error = fd_getvnode(fd, &fp)) != 0)
1176 		return error;
1177 	vp = fp->f_data;
1178 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1179 	if (vp->v_type != VDIR)
1180 		error = ENOTDIR;
1181 	else
1182 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1183 	VOP_UNLOCK(vp);
1184 	if (error)
1185 		goto out;
1186 	vref(vp);
1187 
1188 	change_root(p->p_cwdi, vp, l);
1189 
1190  out:
1191 	fd_putfile(fd);
1192 	return (error);
1193 }
1194 
1195 /*
1196  * Change current working directory (``.'').
1197  */
1198 /* ARGSUSED */
1199 int
1200 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1201 {
1202 	/* {
1203 		syscallarg(const char *) path;
1204 	} */
1205 	struct proc *p = l->l_proc;
1206 	struct cwdinfo *cwdi;
1207 	int error;
1208 	struct vnode *vp;
1209 
1210 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1211 				  &vp, l)) != 0)
1212 		return (error);
1213 	cwdi = p->p_cwdi;
1214 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1215 	vrele(cwdi->cwdi_cdir);
1216 	cwdi->cwdi_cdir = vp;
1217 	rw_exit(&cwdi->cwdi_lock);
1218 	return (0);
1219 }
1220 
1221 /*
1222  * Change notion of root (``/'') directory.
1223  */
1224 /* ARGSUSED */
1225 int
1226 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1227 {
1228 	/* {
1229 		syscallarg(const char *) path;
1230 	} */
1231 	struct proc *p = l->l_proc;
1232 	int error;
1233 	struct vnode *vp;
1234 
1235 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1236 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1237 		return (error);
1238 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1239 				  &vp, l)) != 0)
1240 		return (error);
1241 
1242 	change_root(p->p_cwdi, vp, l);
1243 
1244 	return (0);
1245 }
1246 
1247 /*
1248  * Common routine for chroot and fchroot.
1249  * NB: callers need to properly authorize the change root operation.
1250  */
1251 void
1252 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1253 {
1254 
1255 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1256 	if (cwdi->cwdi_rdir != NULL)
1257 		vrele(cwdi->cwdi_rdir);
1258 	cwdi->cwdi_rdir = vp;
1259 
1260 	/*
1261 	 * Prevent escaping from chroot by putting the root under
1262 	 * the working directory.  Silently chdir to / if we aren't
1263 	 * already there.
1264 	 */
1265 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1266 		/*
1267 		 * XXX would be more failsafe to change directory to a
1268 		 * deadfs node here instead
1269 		 */
1270 		vrele(cwdi->cwdi_cdir);
1271 		vref(vp);
1272 		cwdi->cwdi_cdir = vp;
1273 	}
1274 	rw_exit(&cwdi->cwdi_lock);
1275 }
1276 
1277 /*
1278  * Common routine for chroot and chdir.
1279  */
1280 int
1281 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1282 {
1283 	struct nameidata nd;
1284 	int error;
1285 
1286 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, where,
1287 	    path);
1288 	if ((error = namei(&nd)) != 0)
1289 		return (error);
1290 	*vpp = nd.ni_vp;
1291 	if ((*vpp)->v_type != VDIR)
1292 		error = ENOTDIR;
1293 	else
1294 		error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1295 
1296 	if (error)
1297 		vput(*vpp);
1298 	else
1299 		VOP_UNLOCK(*vpp);
1300 	return (error);
1301 }
1302 
1303 /*
1304  * Check permissions, allocate an open file structure,
1305  * and call the device open routine if any.
1306  */
1307 int
1308 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1309 {
1310 	/* {
1311 		syscallarg(const char *) path;
1312 		syscallarg(int) flags;
1313 		syscallarg(int) mode;
1314 	} */
1315 	struct proc *p = l->l_proc;
1316 	struct cwdinfo *cwdi = p->p_cwdi;
1317 	file_t *fp;
1318 	struct vnode *vp;
1319 	int flags, cmode;
1320 	int type, indx, error;
1321 	struct flock lf;
1322 	struct nameidata nd;
1323 
1324 	flags = FFLAGS(SCARG(uap, flags));
1325 	if ((flags & (FREAD | FWRITE)) == 0)
1326 		return (EINVAL);
1327 	if ((error = fd_allocfile(&fp, &indx)) != 0)
1328 		return (error);
1329 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1330 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1331 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1332 	    SCARG(uap, path));
1333 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1334 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1335 		fd_abort(p, fp, indx);
1336 		if ((error == EDUPFD || error == EMOVEFD) &&
1337 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1338 		    (error =
1339 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1340 			*retval = indx;
1341 			return (0);
1342 		}
1343 		if (error == ERESTART)
1344 			error = EINTR;
1345 		return (error);
1346 	}
1347 
1348 	l->l_dupfd = 0;
1349 	vp = nd.ni_vp;
1350 	fp->f_flag = flags & FMASK;
1351 	fp->f_type = DTYPE_VNODE;
1352 	fp->f_ops = &vnops;
1353 	fp->f_data = vp;
1354 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1355 		lf.l_whence = SEEK_SET;
1356 		lf.l_start = 0;
1357 		lf.l_len = 0;
1358 		if (flags & O_EXLOCK)
1359 			lf.l_type = F_WRLCK;
1360 		else
1361 			lf.l_type = F_RDLCK;
1362 		type = F_FLOCK;
1363 		if ((flags & FNONBLOCK) == 0)
1364 			type |= F_WAIT;
1365 		VOP_UNLOCK(vp);
1366 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1367 		if (error) {
1368 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1369 			fd_abort(p, fp, indx);
1370 			return (error);
1371 		}
1372 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1373 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1374 	}
1375 	VOP_UNLOCK(vp);
1376 	*retval = indx;
1377 	fd_affix(p, fp, indx);
1378 	return (0);
1379 }
1380 
1381 static void
1382 vfs__fhfree(fhandle_t *fhp)
1383 {
1384 	size_t fhsize;
1385 
1386 	if (fhp == NULL) {
1387 		return;
1388 	}
1389 	fhsize = FHANDLE_SIZE(fhp);
1390 	kmem_free(fhp, fhsize);
1391 }
1392 
1393 /*
1394  * vfs_composefh: compose a filehandle.
1395  */
1396 
1397 int
1398 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1399 {
1400 	struct mount *mp;
1401 	struct fid *fidp;
1402 	int error;
1403 	size_t needfhsize;
1404 	size_t fidsize;
1405 
1406 	mp = vp->v_mount;
1407 	fidp = NULL;
1408 	if (*fh_size < FHANDLE_SIZE_MIN) {
1409 		fidsize = 0;
1410 	} else {
1411 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1412 		if (fhp != NULL) {
1413 			memset(fhp, 0, *fh_size);
1414 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1415 			fidp = &fhp->fh_fid;
1416 		}
1417 	}
1418 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1419 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1420 	if (error == 0 && *fh_size < needfhsize) {
1421 		error = E2BIG;
1422 	}
1423 	*fh_size = needfhsize;
1424 	return error;
1425 }
1426 
1427 int
1428 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1429 {
1430 	struct mount *mp;
1431 	fhandle_t *fhp;
1432 	size_t fhsize;
1433 	size_t fidsize;
1434 	int error;
1435 
1436 	*fhpp = NULL;
1437 	mp = vp->v_mount;
1438 	fidsize = 0;
1439 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1440 	KASSERT(error != 0);
1441 	if (error != E2BIG) {
1442 		goto out;
1443 	}
1444 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1445 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1446 	if (fhp == NULL) {
1447 		error = ENOMEM;
1448 		goto out;
1449 	}
1450 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1451 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1452 	if (error == 0) {
1453 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1454 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1455 		*fhpp = fhp;
1456 	} else {
1457 		kmem_free(fhp, fhsize);
1458 	}
1459 out:
1460 	return error;
1461 }
1462 
1463 void
1464 vfs_composefh_free(fhandle_t *fhp)
1465 {
1466 
1467 	vfs__fhfree(fhp);
1468 }
1469 
1470 /*
1471  * vfs_fhtovp: lookup a vnode by a filehandle.
1472  */
1473 
1474 int
1475 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1476 {
1477 	struct mount *mp;
1478 	int error;
1479 
1480 	*vpp = NULL;
1481 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1482 	if (mp == NULL) {
1483 		error = ESTALE;
1484 		goto out;
1485 	}
1486 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1487 		error = EOPNOTSUPP;
1488 		goto out;
1489 	}
1490 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1491 out:
1492 	return error;
1493 }
1494 
1495 /*
1496  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1497  * the needed size.
1498  */
1499 
1500 int
1501 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1502 {
1503 	fhandle_t *fhp;
1504 	int error;
1505 
1506 	*fhpp = NULL;
1507 	if (fhsize > FHANDLE_SIZE_MAX) {
1508 		return EINVAL;
1509 	}
1510 	if (fhsize < FHANDLE_SIZE_MIN) {
1511 		return EINVAL;
1512 	}
1513 again:
1514 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1515 	if (fhp == NULL) {
1516 		return ENOMEM;
1517 	}
1518 	error = copyin(ufhp, fhp, fhsize);
1519 	if (error == 0) {
1520 		/* XXX this check shouldn't be here */
1521 		if (FHANDLE_SIZE(fhp) == fhsize) {
1522 			*fhpp = fhp;
1523 			return 0;
1524 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1525 			/*
1526 			 * a kludge for nfsv2 padded handles.
1527 			 */
1528 			size_t sz;
1529 
1530 			sz = FHANDLE_SIZE(fhp);
1531 			kmem_free(fhp, fhsize);
1532 			fhsize = sz;
1533 			goto again;
1534 		} else {
1535 			/*
1536 			 * userland told us wrong size.
1537 			 */
1538 		    	error = EINVAL;
1539 		}
1540 	}
1541 	kmem_free(fhp, fhsize);
1542 	return error;
1543 }
1544 
1545 void
1546 vfs_copyinfh_free(fhandle_t *fhp)
1547 {
1548 
1549 	vfs__fhfree(fhp);
1550 }
1551 
1552 /*
1553  * Get file handle system call
1554  */
1555 int
1556 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1557 {
1558 	/* {
1559 		syscallarg(char *) fname;
1560 		syscallarg(fhandle_t *) fhp;
1561 		syscallarg(size_t *) fh_size;
1562 	} */
1563 	struct vnode *vp;
1564 	fhandle_t *fh;
1565 	int error;
1566 	struct nameidata nd;
1567 	size_t sz;
1568 	size_t usz;
1569 
1570 	/*
1571 	 * Must be super user
1572 	 */
1573 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1574 	    0, NULL, NULL, NULL);
1575 	if (error)
1576 		return (error);
1577 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1578 	    SCARG(uap, fname));
1579 	error = namei(&nd);
1580 	if (error)
1581 		return (error);
1582 	vp = nd.ni_vp;
1583 	error = vfs_composefh_alloc(vp, &fh);
1584 	vput(vp);
1585 	if (error != 0) {
1586 		goto out;
1587 	}
1588 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1589 	if (error != 0) {
1590 		goto out;
1591 	}
1592 	sz = FHANDLE_SIZE(fh);
1593 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1594 	if (error != 0) {
1595 		goto out;
1596 	}
1597 	if (usz >= sz) {
1598 		error = copyout(fh, SCARG(uap, fhp), sz);
1599 	} else {
1600 		error = E2BIG;
1601 	}
1602 out:
1603 	vfs_composefh_free(fh);
1604 	return (error);
1605 }
1606 
1607 /*
1608  * Open a file given a file handle.
1609  *
1610  * Check permissions, allocate an open file structure,
1611  * and call the device open routine if any.
1612  */
1613 
1614 int
1615 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1616     register_t *retval)
1617 {
1618 	file_t *fp;
1619 	struct vnode *vp = NULL;
1620 	kauth_cred_t cred = l->l_cred;
1621 	file_t *nfp;
1622 	int type, indx, error=0;
1623 	struct flock lf;
1624 	struct vattr va;
1625 	fhandle_t *fh;
1626 	int flags;
1627 	proc_t *p;
1628 
1629 	p = curproc;
1630 
1631 	/*
1632 	 * Must be super user
1633 	 */
1634 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1635 	    0, NULL, NULL, NULL)))
1636 		return (error);
1637 
1638 	flags = FFLAGS(oflags);
1639 	if ((flags & (FREAD | FWRITE)) == 0)
1640 		return (EINVAL);
1641 	if ((flags & O_CREAT))
1642 		return (EINVAL);
1643 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1644 		return (error);
1645 	fp = nfp;
1646 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1647 	if (error != 0) {
1648 		goto bad;
1649 	}
1650 	error = vfs_fhtovp(fh, &vp);
1651 	if (error != 0) {
1652 		goto bad;
1653 	}
1654 
1655 	/* Now do an effective vn_open */
1656 
1657 	if (vp->v_type == VSOCK) {
1658 		error = EOPNOTSUPP;
1659 		goto bad;
1660 	}
1661 	error = vn_openchk(vp, cred, flags);
1662 	if (error != 0)
1663 		goto bad;
1664 	if (flags & O_TRUNC) {
1665 		VOP_UNLOCK(vp);			/* XXX */
1666 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1667 		vattr_null(&va);
1668 		va.va_size = 0;
1669 		error = VOP_SETATTR(vp, &va, cred);
1670 		if (error)
1671 			goto bad;
1672 	}
1673 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1674 		goto bad;
1675 	if (flags & FWRITE) {
1676 		mutex_enter(&vp->v_interlock);
1677 		vp->v_writecount++;
1678 		mutex_exit(&vp->v_interlock);
1679 	}
1680 
1681 	/* done with modified vn_open, now finish what sys_open does. */
1682 
1683 	fp->f_flag = flags & FMASK;
1684 	fp->f_type = DTYPE_VNODE;
1685 	fp->f_ops = &vnops;
1686 	fp->f_data = vp;
1687 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1688 		lf.l_whence = SEEK_SET;
1689 		lf.l_start = 0;
1690 		lf.l_len = 0;
1691 		if (flags & O_EXLOCK)
1692 			lf.l_type = F_WRLCK;
1693 		else
1694 			lf.l_type = F_RDLCK;
1695 		type = F_FLOCK;
1696 		if ((flags & FNONBLOCK) == 0)
1697 			type |= F_WAIT;
1698 		VOP_UNLOCK(vp);
1699 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1700 		if (error) {
1701 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1702 			fd_abort(p, fp, indx);
1703 			return (error);
1704 		}
1705 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1706 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1707 	}
1708 	VOP_UNLOCK(vp);
1709 	*retval = indx;
1710 	fd_affix(p, fp, indx);
1711 	vfs_copyinfh_free(fh);
1712 	return (0);
1713 
1714 bad:
1715 	fd_abort(p, fp, indx);
1716 	if (vp != NULL)
1717 		vput(vp);
1718 	vfs_copyinfh_free(fh);
1719 	return (error);
1720 }
1721 
1722 int
1723 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1724 {
1725 	/* {
1726 		syscallarg(const void *) fhp;
1727 		syscallarg(size_t) fh_size;
1728 		syscallarg(int) flags;
1729 	} */
1730 
1731 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1732 	    SCARG(uap, flags), retval);
1733 }
1734 
1735 int
1736 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1737 {
1738 	int error;
1739 	fhandle_t *fh;
1740 	struct vnode *vp;
1741 
1742 	/*
1743 	 * Must be super user
1744 	 */
1745 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1746 	    0, NULL, NULL, NULL)))
1747 		return (error);
1748 
1749 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1750 	if (error != 0)
1751 		return error;
1752 
1753 	error = vfs_fhtovp(fh, &vp);
1754 	vfs_copyinfh_free(fh);
1755 	if (error != 0)
1756 		return error;
1757 
1758 	error = vn_stat(vp, sb);
1759 	vput(vp);
1760 	return error;
1761 }
1762 
1763 
1764 /* ARGSUSED */
1765 int
1766 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1767 {
1768 	/* {
1769 		syscallarg(const void *) fhp;
1770 		syscallarg(size_t) fh_size;
1771 		syscallarg(struct stat *) sb;
1772 	} */
1773 	struct stat sb;
1774 	int error;
1775 
1776 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1777 	if (error)
1778 		return error;
1779 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1780 }
1781 
1782 int
1783 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1784     int flags)
1785 {
1786 	fhandle_t *fh;
1787 	struct mount *mp;
1788 	struct vnode *vp;
1789 	int error;
1790 
1791 	/*
1792 	 * Must be super user
1793 	 */
1794 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1795 	    0, NULL, NULL, NULL)))
1796 		return error;
1797 
1798 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1799 	if (error != 0)
1800 		return error;
1801 
1802 	error = vfs_fhtovp(fh, &vp);
1803 	vfs_copyinfh_free(fh);
1804 	if (error != 0)
1805 		return error;
1806 
1807 	mp = vp->v_mount;
1808 	error = dostatvfs(mp, sb, l, flags, 1);
1809 	vput(vp);
1810 	return error;
1811 }
1812 
1813 /* ARGSUSED */
1814 int
1815 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1816 {
1817 	/* {
1818 		syscallarg(const void *) fhp;
1819 		syscallarg(size_t) fh_size;
1820 		syscallarg(struct statvfs *) buf;
1821 		syscallarg(int)	flags;
1822 	} */
1823 	struct statvfs *sb = STATVFSBUF_GET();
1824 	int error;
1825 
1826 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1827 	    SCARG(uap, flags));
1828 	if (error == 0)
1829 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1830 	STATVFSBUF_PUT(sb);
1831 	return error;
1832 }
1833 
1834 /*
1835  * Create a special file.
1836  */
1837 /* ARGSUSED */
1838 int
1839 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
1840     register_t *retval)
1841 {
1842 	/* {
1843 		syscallarg(const char *) path;
1844 		syscallarg(mode_t) mode;
1845 		syscallarg(dev_t) dev;
1846 	} */
1847 	return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
1848 	    SCARG(uap, dev), retval, UIO_USERSPACE);
1849 }
1850 
1851 int
1852 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
1853     register_t *retval, enum uio_seg seg)
1854 {
1855 	struct proc *p = l->l_proc;
1856 	struct vnode *vp;
1857 	struct vattr vattr;
1858 	int error, optype;
1859 	struct nameidata nd;
1860 	char *path;
1861 	const char *cpath;
1862 
1863 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1864 	    0, NULL, NULL, NULL)) != 0)
1865 		return (error);
1866 
1867 	optype = VOP_MKNOD_DESCOFFSET;
1868 
1869 	VERIEXEC_PATH_GET(pathname, seg, cpath, path);
1870 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
1871 
1872 	if ((error = namei(&nd)) != 0)
1873 		goto out;
1874 	vp = nd.ni_vp;
1875 	if (vp != NULL)
1876 		error = EEXIST;
1877 	else {
1878 		vattr_null(&vattr);
1879 		/* We will read cwdi->cwdi_cmask unlocked. */
1880 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1881 		vattr.va_rdev = dev;
1882 
1883 		switch (mode & S_IFMT) {
1884 		case S_IFMT:	/* used by badsect to flag bad sectors */
1885 			vattr.va_type = VBAD;
1886 			break;
1887 		case S_IFCHR:
1888 			vattr.va_type = VCHR;
1889 			break;
1890 		case S_IFBLK:
1891 			vattr.va_type = VBLK;
1892 			break;
1893 		case S_IFWHT:
1894 			optype = VOP_WHITEOUT_DESCOFFSET;
1895 			break;
1896 		case S_IFREG:
1897 #if NVERIEXEC > 0
1898 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1899 			    O_CREAT);
1900 #endif /* NVERIEXEC > 0 */
1901 			vattr.va_type = VREG;
1902 			vattr.va_rdev = VNOVAL;
1903 			optype = VOP_CREATE_DESCOFFSET;
1904 			break;
1905 		default:
1906 			error = EINVAL;
1907 			break;
1908 		}
1909 	}
1910 	if (!error) {
1911 		switch (optype) {
1912 		case VOP_WHITEOUT_DESCOFFSET:
1913 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1914 			if (error)
1915 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1916 			vput(nd.ni_dvp);
1917 			break;
1918 
1919 		case VOP_MKNOD_DESCOFFSET:
1920 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1921 						&nd.ni_cnd, &vattr);
1922 			if (error == 0)
1923 				vput(nd.ni_vp);
1924 			break;
1925 
1926 		case VOP_CREATE_DESCOFFSET:
1927 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1928 						&nd.ni_cnd, &vattr);
1929 			if (error == 0)
1930 				vput(nd.ni_vp);
1931 			break;
1932 		}
1933 	} else {
1934 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1935 		if (nd.ni_dvp == vp)
1936 			vrele(nd.ni_dvp);
1937 		else
1938 			vput(nd.ni_dvp);
1939 		if (vp)
1940 			vrele(vp);
1941 	}
1942 out:
1943 	VERIEXEC_PATH_PUT(path);
1944 	return (error);
1945 }
1946 
1947 /*
1948  * Create a named pipe.
1949  */
1950 /* ARGSUSED */
1951 int
1952 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1953 {
1954 	/* {
1955 		syscallarg(const char *) path;
1956 		syscallarg(int) mode;
1957 	} */
1958 	struct proc *p = l->l_proc;
1959 	struct vattr vattr;
1960 	int error;
1961 	struct nameidata nd;
1962 
1963 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1964 	    SCARG(uap, path));
1965 	if ((error = namei(&nd)) != 0)
1966 		return (error);
1967 	if (nd.ni_vp != NULL) {
1968 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1969 		if (nd.ni_dvp == nd.ni_vp)
1970 			vrele(nd.ni_dvp);
1971 		else
1972 			vput(nd.ni_dvp);
1973 		vrele(nd.ni_vp);
1974 		return (EEXIST);
1975 	}
1976 	vattr_null(&vattr);
1977 	vattr.va_type = VFIFO;
1978 	/* We will read cwdi->cwdi_cmask unlocked. */
1979 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1980 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1981 	if (error == 0)
1982 		vput(nd.ni_vp);
1983 	return (error);
1984 }
1985 
1986 /*
1987  * Make a hard file link.
1988  */
1989 /* ARGSUSED */
1990 int
1991 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
1992 {
1993 	/* {
1994 		syscallarg(const char *) path;
1995 		syscallarg(const char *) link;
1996 	} */
1997 	struct vnode *vp;
1998 	struct nameidata nd;
1999 	int error;
2000 
2001 	error = namei_simple_user(SCARG(uap, path),
2002 				NSM_FOLLOW_TRYEMULROOT, &vp);
2003 	if (error != 0)
2004 		return (error);
2005 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2006 	    SCARG(uap, link));
2007 	if ((error = namei(&nd)) != 0)
2008 		goto out;
2009 	if (nd.ni_vp) {
2010 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2011 		if (nd.ni_dvp == nd.ni_vp)
2012 			vrele(nd.ni_dvp);
2013 		else
2014 			vput(nd.ni_dvp);
2015 		vrele(nd.ni_vp);
2016 		error = EEXIST;
2017 		goto out;
2018 	}
2019 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2020 out:
2021 	vrele(vp);
2022 	return (error);
2023 }
2024 
2025 int
2026 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg)
2027 {
2028 	struct proc *p = curproc;
2029 	struct vattr vattr;
2030 	char *path;
2031 	int error;
2032 	struct nameidata nd;
2033 
2034 	path = PNBUF_GET();
2035 	if (seg == UIO_USERSPACE) {
2036 		if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0)
2037 			goto out;
2038 	} else {
2039 		KASSERT(strlen(patharg) < MAXPATHLEN);
2040 		strcpy(path, patharg);
2041 	}
2042 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, link);
2043 	if ((error = namei(&nd)) != 0)
2044 		goto out;
2045 	if (nd.ni_vp) {
2046 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2047 		if (nd.ni_dvp == nd.ni_vp)
2048 			vrele(nd.ni_dvp);
2049 		else
2050 			vput(nd.ni_dvp);
2051 		vrele(nd.ni_vp);
2052 		error = EEXIST;
2053 		goto out;
2054 	}
2055 	vattr_null(&vattr);
2056 	vattr.va_type = VLNK;
2057 	/* We will read cwdi->cwdi_cmask unlocked. */
2058 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2059 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2060 	if (error == 0)
2061 		vput(nd.ni_vp);
2062 out:
2063 	PNBUF_PUT(path);
2064 	return (error);
2065 }
2066 
2067 /*
2068  * Make a symbolic link.
2069  */
2070 /* ARGSUSED */
2071 int
2072 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2073 {
2074 	/* {
2075 		syscallarg(const char *) path;
2076 		syscallarg(const char *) link;
2077 	} */
2078 
2079 	return do_sys_symlink(SCARG(uap, path), SCARG(uap, link),
2080 	    UIO_USERSPACE);
2081 }
2082 
2083 /*
2084  * Delete a whiteout from the filesystem.
2085  */
2086 /* ARGSUSED */
2087 int
2088 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2089 {
2090 	/* {
2091 		syscallarg(const char *) path;
2092 	} */
2093 	int error;
2094 	struct nameidata nd;
2095 
2096 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
2097 	    UIO_USERSPACE, SCARG(uap, path));
2098 	error = namei(&nd);
2099 	if (error)
2100 		return (error);
2101 
2102 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2103 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2104 		if (nd.ni_dvp == nd.ni_vp)
2105 			vrele(nd.ni_dvp);
2106 		else
2107 			vput(nd.ni_dvp);
2108 		if (nd.ni_vp)
2109 			vrele(nd.ni_vp);
2110 		return (EEXIST);
2111 	}
2112 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2113 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2114 	vput(nd.ni_dvp);
2115 	return (error);
2116 }
2117 
2118 /*
2119  * Delete a name from the filesystem.
2120  */
2121 /* ARGSUSED */
2122 int
2123 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2124 {
2125 	/* {
2126 		syscallarg(const char *) path;
2127 	} */
2128 
2129 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2130 }
2131 
2132 int
2133 do_sys_unlink(const char *arg, enum uio_seg seg)
2134 {
2135 	struct vnode *vp;
2136 	int error;
2137 	struct nameidata nd;
2138 	char *path;
2139 	const char *cpath;
2140 
2141 	VERIEXEC_PATH_GET(arg, seg, cpath, path);
2142 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
2143 
2144 	if ((error = namei(&nd)) != 0)
2145 		goto out;
2146 	vp = nd.ni_vp;
2147 
2148 	/*
2149 	 * The root of a mounted filesystem cannot be deleted.
2150 	 */
2151 	if (vp->v_vflag & VV_ROOT) {
2152 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2153 		if (nd.ni_dvp == vp)
2154 			vrele(nd.ni_dvp);
2155 		else
2156 			vput(nd.ni_dvp);
2157 		vput(vp);
2158 		error = EBUSY;
2159 		goto out;
2160 	}
2161 
2162 #if NVERIEXEC > 0
2163 	/* Handle remove requests for veriexec entries. */
2164 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
2165 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2166 		if (nd.ni_dvp == vp)
2167 			vrele(nd.ni_dvp);
2168 		else
2169 			vput(nd.ni_dvp);
2170 		vput(vp);
2171 		goto out;
2172 	}
2173 #endif /* NVERIEXEC > 0 */
2174 
2175 #ifdef FILEASSOC
2176 	(void)fileassoc_file_delete(vp);
2177 #endif /* FILEASSOC */
2178 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2179 out:
2180 	VERIEXEC_PATH_PUT(path);
2181 	return (error);
2182 }
2183 
2184 /*
2185  * Reposition read/write file offset.
2186  */
2187 int
2188 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2189 {
2190 	/* {
2191 		syscallarg(int) fd;
2192 		syscallarg(int) pad;
2193 		syscallarg(off_t) offset;
2194 		syscallarg(int) whence;
2195 	} */
2196 	kauth_cred_t cred = l->l_cred;
2197 	file_t *fp;
2198 	struct vnode *vp;
2199 	struct vattr vattr;
2200 	off_t newoff;
2201 	int error, fd;
2202 
2203 	fd = SCARG(uap, fd);
2204 
2205 	if ((fp = fd_getfile(fd)) == NULL)
2206 		return (EBADF);
2207 
2208 	vp = fp->f_data;
2209 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2210 		error = ESPIPE;
2211 		goto out;
2212 	}
2213 
2214 	switch (SCARG(uap, whence)) {
2215 	case SEEK_CUR:
2216 		newoff = fp->f_offset + SCARG(uap, offset);
2217 		break;
2218 	case SEEK_END:
2219 		error = VOP_GETATTR(vp, &vattr, cred);
2220 		if (error) {
2221 			goto out;
2222 		}
2223 		newoff = SCARG(uap, offset) + vattr.va_size;
2224 		break;
2225 	case SEEK_SET:
2226 		newoff = SCARG(uap, offset);
2227 		break;
2228 	default:
2229 		error = EINVAL;
2230 		goto out;
2231 	}
2232 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2233 		*(off_t *)retval = fp->f_offset = newoff;
2234 	}
2235  out:
2236  	fd_putfile(fd);
2237 	return (error);
2238 }
2239 
2240 /*
2241  * Positional read system call.
2242  */
2243 int
2244 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2245 {
2246 	/* {
2247 		syscallarg(int) fd;
2248 		syscallarg(void *) buf;
2249 		syscallarg(size_t) nbyte;
2250 		syscallarg(off_t) offset;
2251 	} */
2252 	file_t *fp;
2253 	struct vnode *vp;
2254 	off_t offset;
2255 	int error, fd = SCARG(uap, fd);
2256 
2257 	if ((fp = fd_getfile(fd)) == NULL)
2258 		return (EBADF);
2259 
2260 	if ((fp->f_flag & FREAD) == 0) {
2261 		fd_putfile(fd);
2262 		return (EBADF);
2263 	}
2264 
2265 	vp = fp->f_data;
2266 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2267 		error = ESPIPE;
2268 		goto out;
2269 	}
2270 
2271 	offset = SCARG(uap, offset);
2272 
2273 	/*
2274 	 * XXX This works because no file systems actually
2275 	 * XXX take any action on the seek operation.
2276 	 */
2277 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2278 		goto out;
2279 
2280 	/* dofileread() will unuse the descriptor for us */
2281 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2282 	    &offset, 0, retval));
2283 
2284  out:
2285 	fd_putfile(fd);
2286 	return (error);
2287 }
2288 
2289 /*
2290  * Positional scatter read system call.
2291  */
2292 int
2293 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2294 {
2295 	/* {
2296 		syscallarg(int) fd;
2297 		syscallarg(const struct iovec *) iovp;
2298 		syscallarg(int) iovcnt;
2299 		syscallarg(off_t) offset;
2300 	} */
2301 	off_t offset = SCARG(uap, offset);
2302 
2303 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2304 	    SCARG(uap, iovcnt), &offset, 0, retval);
2305 }
2306 
2307 /*
2308  * Positional write system call.
2309  */
2310 int
2311 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2312 {
2313 	/* {
2314 		syscallarg(int) fd;
2315 		syscallarg(const void *) buf;
2316 		syscallarg(size_t) nbyte;
2317 		syscallarg(off_t) offset;
2318 	} */
2319 	file_t *fp;
2320 	struct vnode *vp;
2321 	off_t offset;
2322 	int error, fd = SCARG(uap, fd);
2323 
2324 	if ((fp = fd_getfile(fd)) == NULL)
2325 		return (EBADF);
2326 
2327 	if ((fp->f_flag & FWRITE) == 0) {
2328 		fd_putfile(fd);
2329 		return (EBADF);
2330 	}
2331 
2332 	vp = fp->f_data;
2333 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2334 		error = ESPIPE;
2335 		goto out;
2336 	}
2337 
2338 	offset = SCARG(uap, offset);
2339 
2340 	/*
2341 	 * XXX This works because no file systems actually
2342 	 * XXX take any action on the seek operation.
2343 	 */
2344 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2345 		goto out;
2346 
2347 	/* dofilewrite() will unuse the descriptor for us */
2348 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2349 	    &offset, 0, retval));
2350 
2351  out:
2352 	fd_putfile(fd);
2353 	return (error);
2354 }
2355 
2356 /*
2357  * Positional gather write system call.
2358  */
2359 int
2360 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2361 {
2362 	/* {
2363 		syscallarg(int) fd;
2364 		syscallarg(const struct iovec *) iovp;
2365 		syscallarg(int) iovcnt;
2366 		syscallarg(off_t) offset;
2367 	} */
2368 	off_t offset = SCARG(uap, offset);
2369 
2370 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2371 	    SCARG(uap, iovcnt), &offset, 0, retval);
2372 }
2373 
2374 /*
2375  * Check access permissions.
2376  */
2377 int
2378 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2379 {
2380 	/* {
2381 		syscallarg(const char *) path;
2382 		syscallarg(int) flags;
2383 	} */
2384 	kauth_cred_t cred;
2385 	struct vnode *vp;
2386 	int error, flags;
2387 	struct nameidata nd;
2388 
2389 	cred = kauth_cred_dup(l->l_cred);
2390 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2391 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2392 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2393 	    SCARG(uap, path));
2394 	/* Override default credentials */
2395 	nd.ni_cnd.cn_cred = cred;
2396 	if ((error = namei(&nd)) != 0)
2397 		goto out;
2398 	vp = nd.ni_vp;
2399 
2400 	/* Flags == 0 means only check for existence. */
2401 	if (SCARG(uap, flags)) {
2402 		flags = 0;
2403 		if (SCARG(uap, flags) & R_OK)
2404 			flags |= VREAD;
2405 		if (SCARG(uap, flags) & W_OK)
2406 			flags |= VWRITE;
2407 		if (SCARG(uap, flags) & X_OK)
2408 			flags |= VEXEC;
2409 
2410 		error = VOP_ACCESS(vp, flags, cred);
2411 		if (!error && (flags & VWRITE))
2412 			error = vn_writechk(vp);
2413 	}
2414 	vput(vp);
2415 out:
2416 	kauth_cred_free(cred);
2417 	return (error);
2418 }
2419 
2420 /*
2421  * Common code for all sys_stat functions, including compat versions.
2422  */
2423 int
2424 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
2425 {
2426 	int error;
2427 	struct nameidata nd;
2428 
2429 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
2430 	    UIO_USERSPACE, path);
2431 	error = namei(&nd);
2432 	if (error != 0)
2433 		return error;
2434 	error = vn_stat(nd.ni_vp, sb);
2435 	vput(nd.ni_vp);
2436 	return error;
2437 }
2438 
2439 /*
2440  * Get file status; this version follows links.
2441  */
2442 /* ARGSUSED */
2443 int
2444 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
2445 {
2446 	/* {
2447 		syscallarg(const char *) path;
2448 		syscallarg(struct stat *) ub;
2449 	} */
2450 	struct stat sb;
2451 	int error;
2452 
2453 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2454 	if (error)
2455 		return error;
2456 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2457 }
2458 
2459 /*
2460  * Get file status; this version does not follow links.
2461  */
2462 /* ARGSUSED */
2463 int
2464 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
2465 {
2466 	/* {
2467 		syscallarg(const char *) path;
2468 		syscallarg(struct stat *) ub;
2469 	} */
2470 	struct stat sb;
2471 	int error;
2472 
2473 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2474 	if (error)
2475 		return error;
2476 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2477 }
2478 
2479 /*
2480  * Get configurable pathname variables.
2481  */
2482 /* ARGSUSED */
2483 int
2484 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2485 {
2486 	/* {
2487 		syscallarg(const char *) path;
2488 		syscallarg(int) name;
2489 	} */
2490 	int error;
2491 	struct nameidata nd;
2492 
2493 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2494 	    SCARG(uap, path));
2495 	if ((error = namei(&nd)) != 0)
2496 		return (error);
2497 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2498 	vput(nd.ni_vp);
2499 	return (error);
2500 }
2501 
2502 /*
2503  * Return target name of a symbolic link.
2504  */
2505 /* ARGSUSED */
2506 int
2507 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2508 {
2509 	/* {
2510 		syscallarg(const char *) path;
2511 		syscallarg(char *) buf;
2512 		syscallarg(size_t) count;
2513 	} */
2514 	struct vnode *vp;
2515 	struct iovec aiov;
2516 	struct uio auio;
2517 	int error;
2518 	struct nameidata nd;
2519 
2520 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2521 	    SCARG(uap, path));
2522 	if ((error = namei(&nd)) != 0)
2523 		return (error);
2524 	vp = nd.ni_vp;
2525 	if (vp->v_type != VLNK)
2526 		error = EINVAL;
2527 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2528 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2529 		aiov.iov_base = SCARG(uap, buf);
2530 		aiov.iov_len = SCARG(uap, count);
2531 		auio.uio_iov = &aiov;
2532 		auio.uio_iovcnt = 1;
2533 		auio.uio_offset = 0;
2534 		auio.uio_rw = UIO_READ;
2535 		KASSERT(l == curlwp);
2536 		auio.uio_vmspace = l->l_proc->p_vmspace;
2537 		auio.uio_resid = SCARG(uap, count);
2538 		error = VOP_READLINK(vp, &auio, l->l_cred);
2539 	}
2540 	vput(vp);
2541 	*retval = SCARG(uap, count) - auio.uio_resid;
2542 	return (error);
2543 }
2544 
2545 /*
2546  * Change flags of a file given a path name.
2547  */
2548 /* ARGSUSED */
2549 int
2550 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2551 {
2552 	/* {
2553 		syscallarg(const char *) path;
2554 		syscallarg(u_long) flags;
2555 	} */
2556 	struct vnode *vp;
2557 	int error;
2558 
2559 	error = namei_simple_user(SCARG(uap, path),
2560 				NSM_FOLLOW_TRYEMULROOT, &vp);
2561 	if (error != 0)
2562 		return (error);
2563 	error = change_flags(vp, SCARG(uap, flags), l);
2564 	vput(vp);
2565 	return (error);
2566 }
2567 
2568 /*
2569  * Change flags of a file given a file descriptor.
2570  */
2571 /* ARGSUSED */
2572 int
2573 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
2574 {
2575 	/* {
2576 		syscallarg(int) fd;
2577 		syscallarg(u_long) flags;
2578 	} */
2579 	struct vnode *vp;
2580 	file_t *fp;
2581 	int error;
2582 
2583 	/* fd_getvnode() will use the descriptor for us */
2584 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2585 		return (error);
2586 	vp = fp->f_data;
2587 	error = change_flags(vp, SCARG(uap, flags), l);
2588 	VOP_UNLOCK(vp);
2589 	fd_putfile(SCARG(uap, fd));
2590 	return (error);
2591 }
2592 
2593 /*
2594  * Change flags of a file given a path name; this version does
2595  * not follow links.
2596  */
2597 int
2598 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
2599 {
2600 	/* {
2601 		syscallarg(const char *) path;
2602 		syscallarg(u_long) flags;
2603 	} */
2604 	struct vnode *vp;
2605 	int error;
2606 
2607 	error = namei_simple_user(SCARG(uap, path),
2608 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2609 	if (error != 0)
2610 		return (error);
2611 	error = change_flags(vp, SCARG(uap, flags), l);
2612 	vput(vp);
2613 	return (error);
2614 }
2615 
2616 /*
2617  * Common routine to change flags of a file.
2618  */
2619 int
2620 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2621 {
2622 	struct vattr vattr;
2623 	int error;
2624 
2625 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2626 	/*
2627 	 * Non-superusers cannot change the flags on devices, even if they
2628 	 * own them.
2629 	 */
2630 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2631 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2632 			goto out;
2633 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2634 			error = EINVAL;
2635 			goto out;
2636 		}
2637 	}
2638 	vattr_null(&vattr);
2639 	vattr.va_flags = flags;
2640 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2641 out:
2642 	return (error);
2643 }
2644 
2645 /*
2646  * Change mode of a file given path name; this version follows links.
2647  */
2648 /* ARGSUSED */
2649 int
2650 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
2651 {
2652 	/* {
2653 		syscallarg(const char *) path;
2654 		syscallarg(int) mode;
2655 	} */
2656 	int error;
2657 	struct vnode *vp;
2658 
2659 	error = namei_simple_user(SCARG(uap, path),
2660 				NSM_FOLLOW_TRYEMULROOT, &vp);
2661 	if (error != 0)
2662 		return (error);
2663 
2664 	error = change_mode(vp, SCARG(uap, mode), l);
2665 
2666 	vrele(vp);
2667 	return (error);
2668 }
2669 
2670 /*
2671  * Change mode of a file given a file descriptor.
2672  */
2673 /* ARGSUSED */
2674 int
2675 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
2676 {
2677 	/* {
2678 		syscallarg(int) fd;
2679 		syscallarg(int) mode;
2680 	} */
2681 	file_t *fp;
2682 	int error;
2683 
2684 	/* fd_getvnode() will use the descriptor for us */
2685 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2686 		return (error);
2687 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
2688 	fd_putfile(SCARG(uap, fd));
2689 	return (error);
2690 }
2691 
2692 /*
2693  * Change mode of a file given path name; this version does not follow links.
2694  */
2695 /* ARGSUSED */
2696 int
2697 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
2698 {
2699 	/* {
2700 		syscallarg(const char *) path;
2701 		syscallarg(int) mode;
2702 	} */
2703 	int error;
2704 	struct vnode *vp;
2705 
2706 	error = namei_simple_user(SCARG(uap, path),
2707 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2708 	if (error != 0)
2709 		return (error);
2710 
2711 	error = change_mode(vp, SCARG(uap, mode), l);
2712 
2713 	vrele(vp);
2714 	return (error);
2715 }
2716 
2717 /*
2718  * Common routine to set mode given a vnode.
2719  */
2720 static int
2721 change_mode(struct vnode *vp, int mode, struct lwp *l)
2722 {
2723 	struct vattr vattr;
2724 	int error;
2725 
2726 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2727 	vattr_null(&vattr);
2728 	vattr.va_mode = mode & ALLPERMS;
2729 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2730 	VOP_UNLOCK(vp);
2731 	return (error);
2732 }
2733 
2734 /*
2735  * Set ownership given a path name; this version follows links.
2736  */
2737 /* ARGSUSED */
2738 int
2739 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
2740 {
2741 	/* {
2742 		syscallarg(const char *) path;
2743 		syscallarg(uid_t) uid;
2744 		syscallarg(gid_t) gid;
2745 	} */
2746 	int error;
2747 	struct vnode *vp;
2748 
2749 	error = namei_simple_user(SCARG(uap, path),
2750 				NSM_FOLLOW_TRYEMULROOT, &vp);
2751 	if (error != 0)
2752 		return (error);
2753 
2754 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2755 
2756 	vrele(vp);
2757 	return (error);
2758 }
2759 
2760 /*
2761  * Set ownership given a path name; this version follows links.
2762  * Provides POSIX semantics.
2763  */
2764 /* ARGSUSED */
2765 int
2766 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
2767 {
2768 	/* {
2769 		syscallarg(const char *) path;
2770 		syscallarg(uid_t) uid;
2771 		syscallarg(gid_t) gid;
2772 	} */
2773 	int error;
2774 	struct vnode *vp;
2775 
2776 	error = namei_simple_user(SCARG(uap, path),
2777 				NSM_FOLLOW_TRYEMULROOT, &vp);
2778 	if (error != 0)
2779 		return (error);
2780 
2781 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2782 
2783 	vrele(vp);
2784 	return (error);
2785 }
2786 
2787 /*
2788  * Set ownership given a file descriptor.
2789  */
2790 /* ARGSUSED */
2791 int
2792 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
2793 {
2794 	/* {
2795 		syscallarg(int) fd;
2796 		syscallarg(uid_t) uid;
2797 		syscallarg(gid_t) gid;
2798 	} */
2799 	int error;
2800 	file_t *fp;
2801 
2802 	/* fd_getvnode() will use the descriptor for us */
2803 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2804 		return (error);
2805 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2806 	    l, 0);
2807 	fd_putfile(SCARG(uap, fd));
2808 	return (error);
2809 }
2810 
2811 /*
2812  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2813  */
2814 /* ARGSUSED */
2815 int
2816 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
2817 {
2818 	/* {
2819 		syscallarg(int) fd;
2820 		syscallarg(uid_t) uid;
2821 		syscallarg(gid_t) gid;
2822 	} */
2823 	int error;
2824 	file_t *fp;
2825 
2826 	/* fd_getvnode() will use the descriptor for us */
2827 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2828 		return (error);
2829 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2830 	    l, 1);
2831 	fd_putfile(SCARG(uap, fd));
2832 	return (error);
2833 }
2834 
2835 /*
2836  * Set ownership given a path name; this version does not follow links.
2837  */
2838 /* ARGSUSED */
2839 int
2840 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
2841 {
2842 	/* {
2843 		syscallarg(const char *) path;
2844 		syscallarg(uid_t) uid;
2845 		syscallarg(gid_t) gid;
2846 	} */
2847 	int error;
2848 	struct vnode *vp;
2849 
2850 	error = namei_simple_user(SCARG(uap, path),
2851 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2852 	if (error != 0)
2853 		return (error);
2854 
2855 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2856 
2857 	vrele(vp);
2858 	return (error);
2859 }
2860 
2861 /*
2862  * Set ownership given a path name; this version does not follow links.
2863  * Provides POSIX/XPG semantics.
2864  */
2865 /* ARGSUSED */
2866 int
2867 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
2868 {
2869 	/* {
2870 		syscallarg(const char *) path;
2871 		syscallarg(uid_t) uid;
2872 		syscallarg(gid_t) gid;
2873 	} */
2874 	int error;
2875 	struct vnode *vp;
2876 
2877 	error = namei_simple_user(SCARG(uap, path),
2878 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2879 	if (error != 0)
2880 		return (error);
2881 
2882 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2883 
2884 	vrele(vp);
2885 	return (error);
2886 }
2887 
2888 /*
2889  * Common routine to set ownership given a vnode.
2890  */
2891 static int
2892 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2893     int posix_semantics)
2894 {
2895 	struct vattr vattr;
2896 	mode_t newmode;
2897 	int error;
2898 
2899 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2900 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2901 		goto out;
2902 
2903 #define CHANGED(x) ((int)(x) != -1)
2904 	newmode = vattr.va_mode;
2905 	if (posix_semantics) {
2906 		/*
2907 		 * POSIX/XPG semantics: if the caller is not the super-user,
2908 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2909 		 * the XPG consider the behaviour for calls by the super-user
2910 		 * implementation-defined; we leave the set-user-id and set-
2911 		 * group-id settings intact in that case.
2912 		 */
2913 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2914 				      NULL) != 0)
2915 			newmode &= ~(S_ISUID | S_ISGID);
2916 	} else {
2917 		/*
2918 		 * NetBSD semantics: when changing owner and/or group,
2919 		 * clear the respective bit(s).
2920 		 */
2921 		if (CHANGED(uid))
2922 			newmode &= ~S_ISUID;
2923 		if (CHANGED(gid))
2924 			newmode &= ~S_ISGID;
2925 	}
2926 	/* Update va_mode iff altered. */
2927 	if (vattr.va_mode == newmode)
2928 		newmode = VNOVAL;
2929 
2930 	vattr_null(&vattr);
2931 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2932 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2933 	vattr.va_mode = newmode;
2934 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2935 #undef CHANGED
2936 
2937 out:
2938 	VOP_UNLOCK(vp);
2939 	return (error);
2940 }
2941 
2942 /*
2943  * Set the access and modification times given a path name; this
2944  * version follows links.
2945  */
2946 /* ARGSUSED */
2947 int
2948 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
2949     register_t *retval)
2950 {
2951 	/* {
2952 		syscallarg(const char *) path;
2953 		syscallarg(const struct timeval *) tptr;
2954 	} */
2955 
2956 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
2957 	    SCARG(uap, tptr), UIO_USERSPACE);
2958 }
2959 
2960 /*
2961  * Set the access and modification times given a file descriptor.
2962  */
2963 /* ARGSUSED */
2964 int
2965 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
2966     register_t *retval)
2967 {
2968 	/* {
2969 		syscallarg(int) fd;
2970 		syscallarg(const struct timeval *) tptr;
2971 	} */
2972 	int error;
2973 	file_t *fp;
2974 
2975 	/* fd_getvnode() will use the descriptor for us */
2976 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2977 		return (error);
2978 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
2979 	    UIO_USERSPACE);
2980 	fd_putfile(SCARG(uap, fd));
2981 	return (error);
2982 }
2983 
2984 /*
2985  * Set the access and modification times given a path name; this
2986  * version does not follow links.
2987  */
2988 int
2989 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
2990     register_t *retval)
2991 {
2992 	/* {
2993 		syscallarg(const char *) path;
2994 		syscallarg(const struct timeval *) tptr;
2995 	} */
2996 
2997 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
2998 	    SCARG(uap, tptr), UIO_USERSPACE);
2999 }
3000 
3001 /*
3002  * Common routine to set access and modification times given a vnode.
3003  */
3004 int
3005 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3006     const struct timeval *tptr, enum uio_seg seg)
3007 {
3008 	struct vattr vattr;
3009 	int error, dorele = 0;
3010 	namei_simple_flags_t sflags;
3011 
3012 	bool vanull, setbirthtime;
3013 	struct timespec ts[2];
3014 
3015 	/*
3016 	 * I have checked all callers and they pass either FOLLOW,
3017 	 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
3018 	 * is 0. More to the point, they don't pass anything else.
3019 	 * Let's keep it that way at least until the namei interfaces
3020 	 * are fully sanitized.
3021 	 */
3022 	KASSERT(flag == NOFOLLOW || flag == FOLLOW);
3023 	sflags = (flag == FOLLOW) ?
3024 		NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3025 
3026 	if (tptr == NULL) {
3027 		vanull = true;
3028 		nanotime(&ts[0]);
3029 		ts[1] = ts[0];
3030 	} else {
3031 		struct timeval tv[2];
3032 
3033 		vanull = false;
3034 		if (seg != UIO_SYSSPACE) {
3035 			error = copyin(tptr, tv, sizeof (tv));
3036 			if (error != 0)
3037 				return error;
3038 			tptr = tv;
3039 		}
3040 		TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3041 		TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3042 	}
3043 
3044 	if (vp == NULL) {
3045 		/* note: SEG describes TPTR, not PATH; PATH is always user */
3046 		error = namei_simple_user(path, sflags, &vp);
3047 		if (error != 0)
3048 			return error;
3049 		dorele = 1;
3050 	}
3051 
3052 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3053 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3054 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
3055 	vattr_null(&vattr);
3056 	vattr.va_atime = ts[0];
3057 	vattr.va_mtime = ts[1];
3058 	if (setbirthtime)
3059 		vattr.va_birthtime = ts[1];
3060 	if (vanull)
3061 		vattr.va_vaflags |= VA_UTIMES_NULL;
3062 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3063 	VOP_UNLOCK(vp);
3064 
3065 	if (dorele != 0)
3066 		vrele(vp);
3067 
3068 	return error;
3069 }
3070 
3071 /*
3072  * Truncate a file given its path name.
3073  */
3074 /* ARGSUSED */
3075 int
3076 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3077 {
3078 	/* {
3079 		syscallarg(const char *) path;
3080 		syscallarg(int) pad;
3081 		syscallarg(off_t) length;
3082 	} */
3083 	struct vnode *vp;
3084 	struct vattr vattr;
3085 	int error;
3086 
3087 	error = namei_simple_user(SCARG(uap, path),
3088 				NSM_FOLLOW_TRYEMULROOT, &vp);
3089 	if (error != 0)
3090 		return (error);
3091 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3092 	if (vp->v_type == VDIR)
3093 		error = EISDIR;
3094 	else if ((error = vn_writechk(vp)) == 0 &&
3095 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3096 		vattr_null(&vattr);
3097 		vattr.va_size = SCARG(uap, length);
3098 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3099 	}
3100 	vput(vp);
3101 	return (error);
3102 }
3103 
3104 /*
3105  * Truncate a file given a file descriptor.
3106  */
3107 /* ARGSUSED */
3108 int
3109 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3110 {
3111 	/* {
3112 		syscallarg(int) fd;
3113 		syscallarg(int) pad;
3114 		syscallarg(off_t) length;
3115 	} */
3116 	struct vattr vattr;
3117 	struct vnode *vp;
3118 	file_t *fp;
3119 	int error;
3120 
3121 	/* fd_getvnode() will use the descriptor for us */
3122 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3123 		return (error);
3124 	if ((fp->f_flag & FWRITE) == 0) {
3125 		error = EINVAL;
3126 		goto out;
3127 	}
3128 	vp = fp->f_data;
3129 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3130 	if (vp->v_type == VDIR)
3131 		error = EISDIR;
3132 	else if ((error = vn_writechk(vp)) == 0) {
3133 		vattr_null(&vattr);
3134 		vattr.va_size = SCARG(uap, length);
3135 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3136 	}
3137 	VOP_UNLOCK(vp);
3138  out:
3139 	fd_putfile(SCARG(uap, fd));
3140 	return (error);
3141 }
3142 
3143 /*
3144  * Sync an open file.
3145  */
3146 /* ARGSUSED */
3147 int
3148 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3149 {
3150 	/* {
3151 		syscallarg(int) fd;
3152 	} */
3153 	struct vnode *vp;
3154 	file_t *fp;
3155 	int error;
3156 
3157 	/* fd_getvnode() will use the descriptor for us */
3158 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3159 		return (error);
3160 	vp = fp->f_data;
3161 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3162 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3163 	VOP_UNLOCK(vp);
3164 	fd_putfile(SCARG(uap, fd));
3165 	return (error);
3166 }
3167 
3168 /*
3169  * Sync a range of file data.  API modeled after that found in AIX.
3170  *
3171  * FDATASYNC indicates that we need only save enough metadata to be able
3172  * to re-read the written data.  Note we duplicate AIX's requirement that
3173  * the file be open for writing.
3174  */
3175 /* ARGSUSED */
3176 int
3177 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3178 {
3179 	/* {
3180 		syscallarg(int) fd;
3181 		syscallarg(int) flags;
3182 		syscallarg(off_t) start;
3183 		syscallarg(off_t) length;
3184 	} */
3185 	struct vnode *vp;
3186 	file_t *fp;
3187 	int flags, nflags;
3188 	off_t s, e, len;
3189 	int error;
3190 
3191 	/* fd_getvnode() will use the descriptor for us */
3192 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3193 		return (error);
3194 
3195 	if ((fp->f_flag & FWRITE) == 0) {
3196 		error = EBADF;
3197 		goto out;
3198 	}
3199 
3200 	flags = SCARG(uap, flags);
3201 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3202 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3203 		error = EINVAL;
3204 		goto out;
3205 	}
3206 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3207 	if (flags & FDATASYNC)
3208 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3209 	else
3210 		nflags = FSYNC_WAIT;
3211 	if (flags & FDISKSYNC)
3212 		nflags |= FSYNC_CACHE;
3213 
3214 	len = SCARG(uap, length);
3215 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3216 	if (len) {
3217 		s = SCARG(uap, start);
3218 		e = s + len;
3219 		if (e < s) {
3220 			error = EINVAL;
3221 			goto out;
3222 		}
3223 	} else {
3224 		e = 0;
3225 		s = 0;
3226 	}
3227 
3228 	vp = fp->f_data;
3229 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3230 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3231 	VOP_UNLOCK(vp);
3232 out:
3233 	fd_putfile(SCARG(uap, fd));
3234 	return (error);
3235 }
3236 
3237 /*
3238  * Sync the data of an open file.
3239  */
3240 /* ARGSUSED */
3241 int
3242 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3243 {
3244 	/* {
3245 		syscallarg(int) fd;
3246 	} */
3247 	struct vnode *vp;
3248 	file_t *fp;
3249 	int error;
3250 
3251 	/* fd_getvnode() will use the descriptor for us */
3252 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3253 		return (error);
3254 	if ((fp->f_flag & FWRITE) == 0) {
3255 		fd_putfile(SCARG(uap, fd));
3256 		return (EBADF);
3257 	}
3258 	vp = fp->f_data;
3259 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3260 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3261 	VOP_UNLOCK(vp);
3262 	fd_putfile(SCARG(uap, fd));
3263 	return (error);
3264 }
3265 
3266 /*
3267  * Rename files, (standard) BSD semantics frontend.
3268  */
3269 /* ARGSUSED */
3270 int
3271 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3272 {
3273 	/* {
3274 		syscallarg(const char *) from;
3275 		syscallarg(const char *) to;
3276 	} */
3277 
3278 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3279 }
3280 
3281 /*
3282  * Rename files, POSIX semantics frontend.
3283  */
3284 /* ARGSUSED */
3285 int
3286 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3287 {
3288 	/* {
3289 		syscallarg(const char *) from;
3290 		syscallarg(const char *) to;
3291 	} */
3292 
3293 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3294 }
3295 
3296 /*
3297  * Rename files.  Source and destination must either both be directories,
3298  * or both not be directories.  If target is a directory, it must be empty.
3299  * If `from' and `to' refer to the same object, the value of the `retain'
3300  * argument is used to determine whether `from' will be
3301  *
3302  * (retain == 0)	deleted unless `from' and `to' refer to the same
3303  *			object in the file system's name space (BSD).
3304  * (retain == 1)	always retained (POSIX).
3305  */
3306 int
3307 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3308 {
3309 	struct vnode *tvp, *fvp, *tdvp;
3310 	struct nameidata fromnd, tond;
3311 	struct mount *fs;
3312 	struct lwp *l = curlwp;
3313 	struct proc *p;
3314 	uint32_t saveflag;
3315 	int error;
3316 
3317 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT | INRENAME,
3318 	    seg, from);
3319 	if ((error = namei(&fromnd)) != 0)
3320 		return (error);
3321 	if (fromnd.ni_dvp != fromnd.ni_vp)
3322 		VOP_UNLOCK(fromnd.ni_dvp);
3323 	fvp = fromnd.ni_vp;
3324 
3325 	fs = fvp->v_mount;
3326 	error = VFS_RENAMELOCK_ENTER(fs);
3327 	if (error) {
3328 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3329 		vrele(fromnd.ni_dvp);
3330 		vrele(fvp);
3331 		goto out1;
3332 	}
3333 
3334 	/*
3335 	 * close, partially, yet another race - ideally we should only
3336 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3337 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3338 	 * that with namei as it stands.
3339 	 *
3340 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3341 	 * under us. The real fix is to get the locks in the right
3342 	 * order and do the lookups in the right places, but that's a
3343 	 * major rototill.
3344 	 *
3345 	 * Preserve the SAVESTART in cn_flags, because who knows what
3346 	 * might happen if we don't.
3347 	 *
3348 	 * Note: this logic (as well as this whole function) is cloned
3349 	 * in nfs_serv.c. Proceed accordingly.
3350 	 */
3351 	vrele(fvp);
3352 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3353 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3354 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3355 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3356 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3357 		error = EINVAL;
3358 		VFS_RENAMELOCK_EXIT(fs);
3359 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3360 		vrele(fromnd.ni_dvp);
3361 		goto out1;
3362 	}
3363 	saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
3364 	fromnd.ni_cnd.cn_flags &= ~SAVESTART;
3365 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3366 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
3367 	fromnd.ni_cnd.cn_flags |= saveflag;
3368 	if (error) {
3369 		VOP_UNLOCK(fromnd.ni_dvp);
3370 		VFS_RENAMELOCK_EXIT(fs);
3371 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3372 		vrele(fromnd.ni_dvp);
3373 		goto out1;
3374 	}
3375 	VOP_UNLOCK(fromnd.ni_vp);
3376 	if (fromnd.ni_dvp != fromnd.ni_vp)
3377 		VOP_UNLOCK(fromnd.ni_dvp);
3378 	fvp = fromnd.ni_vp;
3379 
3380 	NDINIT(&tond, RENAME,
3381 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3382 	      | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0),
3383 	    seg, to);
3384 	if ((error = namei(&tond)) != 0) {
3385 		VFS_RENAMELOCK_EXIT(fs);
3386 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3387 		vrele(fromnd.ni_dvp);
3388 		vrele(fvp);
3389 		goto out1;
3390 	}
3391 	tdvp = tond.ni_dvp;
3392 	tvp = tond.ni_vp;
3393 
3394 	if (tvp != NULL) {
3395 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3396 			error = ENOTDIR;
3397 			goto out;
3398 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3399 			error = EISDIR;
3400 			goto out;
3401 		}
3402 	}
3403 
3404 	if (fvp == tdvp)
3405 		error = EINVAL;
3406 
3407 	/*
3408 	 * Source and destination refer to the same object.
3409 	 */
3410 	if (fvp == tvp) {
3411 		if (retain)
3412 			error = -1;
3413 		else if (fromnd.ni_dvp == tdvp &&
3414 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3415 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3416 		          tond.ni_cnd.cn_nameptr,
3417 		          fromnd.ni_cnd.cn_namelen))
3418 		error = -1;
3419 	}
3420 
3421 #if NVERIEXEC > 0
3422 	if (!error) {
3423 		char *f1, *f2;
3424 		size_t f1_len;
3425 		size_t f2_len;
3426 
3427 		f1_len = fromnd.ni_cnd.cn_namelen + 1;
3428 		f1 = kmem_alloc(f1_len, KM_SLEEP);
3429 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
3430 
3431 		f2_len = tond.ni_cnd.cn_namelen + 1;
3432 		f2 = kmem_alloc(f2_len, KM_SLEEP);
3433 		strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
3434 
3435 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
3436 
3437 		kmem_free(f1, f1_len);
3438 		kmem_free(f2, f2_len);
3439 	}
3440 #endif /* NVERIEXEC > 0 */
3441 
3442 out:
3443 	p = l->l_proc;
3444 	if (!error) {
3445 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3446 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3447 		VFS_RENAMELOCK_EXIT(fs);
3448 	} else {
3449 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3450 		if (tdvp == tvp)
3451 			vrele(tdvp);
3452 		else
3453 			vput(tdvp);
3454 		if (tvp)
3455 			vput(tvp);
3456 		VFS_RENAMELOCK_EXIT(fs);
3457 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3458 		vrele(fromnd.ni_dvp);
3459 		vrele(fvp);
3460 	}
3461 	vrele(tond.ni_startdir);
3462 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3463 out1:
3464 	if (fromnd.ni_startdir)
3465 		vrele(fromnd.ni_startdir);
3466 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3467 	return (error == -1 ? 0 : error);
3468 }
3469 
3470 /*
3471  * Make a directory file.
3472  */
3473 /* ARGSUSED */
3474 int
3475 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
3476 {
3477 	/* {
3478 		syscallarg(const char *) path;
3479 		syscallarg(int) mode;
3480 	} */
3481 
3482 	return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE);
3483 }
3484 
3485 int
3486 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
3487 {
3488 	struct proc *p = curlwp->l_proc;
3489 	struct vnode *vp;
3490 	struct vattr vattr;
3491 	int error;
3492 	struct nameidata nd;
3493 
3494 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT,
3495 	    seg, path);
3496 	if ((error = namei(&nd)) != 0)
3497 		return (error);
3498 	vp = nd.ni_vp;
3499 	if (vp != NULL) {
3500 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3501 		if (nd.ni_dvp == vp)
3502 			vrele(nd.ni_dvp);
3503 		else
3504 			vput(nd.ni_dvp);
3505 		vrele(vp);
3506 		return (EEXIST);
3507 	}
3508 	vattr_null(&vattr);
3509 	vattr.va_type = VDIR;
3510 	/* We will read cwdi->cwdi_cmask unlocked. */
3511 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3512 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3513 	if (!error)
3514 		vput(nd.ni_vp);
3515 	return (error);
3516 }
3517 
3518 /*
3519  * Remove a directory file.
3520  */
3521 /* ARGSUSED */
3522 int
3523 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
3524 {
3525 	/* {
3526 		syscallarg(const char *) path;
3527 	} */
3528 	struct vnode *vp;
3529 	int error;
3530 	struct nameidata nd;
3531 
3532 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
3533 	    SCARG(uap, path));
3534 	if ((error = namei(&nd)) != 0)
3535 		return (error);
3536 	vp = nd.ni_vp;
3537 	if (vp->v_type != VDIR) {
3538 		error = ENOTDIR;
3539 		goto out;
3540 	}
3541 	/*
3542 	 * No rmdir "." please.
3543 	 */
3544 	if (nd.ni_dvp == vp) {
3545 		error = EINVAL;
3546 		goto out;
3547 	}
3548 	/*
3549 	 * The root of a mounted filesystem cannot be deleted.
3550 	 */
3551 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
3552 		error = EBUSY;
3553 		goto out;
3554 	}
3555 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3556 	return (error);
3557 
3558 out:
3559 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3560 	if (nd.ni_dvp == vp)
3561 		vrele(nd.ni_dvp);
3562 	else
3563 		vput(nd.ni_dvp);
3564 	vput(vp);
3565 	return (error);
3566 }
3567 
3568 /*
3569  * Read a block of directory entries in a file system independent format.
3570  */
3571 int
3572 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
3573 {
3574 	/* {
3575 		syscallarg(int) fd;
3576 		syscallarg(char *) buf;
3577 		syscallarg(size_t) count;
3578 	} */
3579 	file_t *fp;
3580 	int error, done;
3581 
3582 	/* fd_getvnode() will use the descriptor for us */
3583 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3584 		return (error);
3585 	if ((fp->f_flag & FREAD) == 0) {
3586 		error = EBADF;
3587 		goto out;
3588 	}
3589 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3590 			SCARG(uap, count), &done, l, 0, 0);
3591 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3592 	*retval = done;
3593  out:
3594 	fd_putfile(SCARG(uap, fd));
3595 	return (error);
3596 }
3597 
3598 /*
3599  * Set the mode mask for creation of filesystem nodes.
3600  */
3601 int
3602 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
3603 {
3604 	/* {
3605 		syscallarg(mode_t) newmask;
3606 	} */
3607 	struct proc *p = l->l_proc;
3608 	struct cwdinfo *cwdi;
3609 
3610 	/*
3611 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3612 	 * important is that we serialize changes to the mask.  The
3613 	 * rw_exit() will issue a write memory barrier on our behalf,
3614 	 * and force the changes out to other CPUs (as it must use an
3615 	 * atomic operation, draining the local CPU's store buffers).
3616 	 */
3617 	cwdi = p->p_cwdi;
3618 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3619 	*retval = cwdi->cwdi_cmask;
3620 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3621 	rw_exit(&cwdi->cwdi_lock);
3622 
3623 	return (0);
3624 }
3625 
3626 int
3627 dorevoke(struct vnode *vp, kauth_cred_t cred)
3628 {
3629 	struct vattr vattr;
3630 	int error;
3631 
3632 	if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
3633 		return error;
3634 	if (kauth_cred_geteuid(cred) == vattr.va_uid ||
3635 	    (error = kauth_authorize_generic(cred,
3636 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
3637 		VOP_REVOKE(vp, REVOKEALL);
3638 	return (error);
3639 }
3640 
3641 /*
3642  * Void all references to file by ripping underlying filesystem
3643  * away from vnode.
3644  */
3645 /* ARGSUSED */
3646 int
3647 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
3648 {
3649 	/* {
3650 		syscallarg(const char *) path;
3651 	} */
3652 	struct vnode *vp;
3653 	int error;
3654 
3655 	error = namei_simple_user(SCARG(uap, path),
3656 				NSM_FOLLOW_TRYEMULROOT, &vp);
3657 	if (error != 0)
3658 		return (error);
3659 	error = dorevoke(vp, l->l_cred);
3660 	vrele(vp);
3661 	return (error);
3662 }
3663