xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 4e6df137e8e14049b5a701d249962c480449c141)
1 /*	$NetBSD: vfs_syscalls.c,v 1.404 2010/03/03 00:47:31 yamt 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 1.404 2010/03/03 00:47:31 yamt Exp $");
70 
71 #ifdef _KERNEL_OPT
72 #include "opt_fileassoc.h"
73 #include "veriexec.h"
74 #endif
75 
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/namei.h>
79 #include <sys/filedesc.h>
80 #include <sys/kernel.h>
81 #include <sys/file.h>
82 #include <sys/stat.h>
83 #include <sys/vnode.h>
84 #include <sys/mount.h>
85 #include <sys/proc.h>
86 #include <sys/uio.h>
87 #include <sys/kmem.h>
88 #include <sys/dirent.h>
89 #include <sys/sysctl.h>
90 #include <sys/syscallargs.h>
91 #include <sys/vfs_syscalls.h>
92 #include <sys/ktrace.h>
93 #ifdef FILEASSOC
94 #include <sys/fileassoc.h>
95 #endif /* FILEASSOC */
96 #include <sys/verified_exec.h>
97 #include <sys/kauth.h>
98 #include <sys/atomic.h>
99 #include <sys/module.h>
100 #include <sys/buf.h>
101 
102 #include <miscfs/genfs/genfs.h>
103 #include <miscfs/syncfs/syncfs.h>
104 #include <miscfs/specfs/specdev.h>
105 
106 #include <nfs/rpcv2.h>
107 #include <nfs/nfsproto.h>
108 #include <nfs/nfs.h>
109 #include <nfs/nfs_var.h>
110 
111 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
112 
113 static int change_flags(struct vnode *, u_long, struct lwp *);
114 static int change_mode(struct vnode *, int, struct lwp *l);
115 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
116 
117 void checkdirs(struct vnode *);
118 
119 /*
120  * Virtual File System System Calls
121  */
122 
123 /*
124  * Mount a file system.
125  */
126 
127 /*
128  * This table is used to maintain compatibility with 4.3BSD
129  * and NetBSD 0.9 mount syscalls - and possibly other systems.
130  * Note, the order is important!
131  *
132  * Do not modify this table. It should only contain filesystems
133  * supported by NetBSD 0.9 and 4.3BSD.
134  */
135 const char * const mountcompatnames[] = {
136 	NULL,		/* 0 = MOUNT_NONE */
137 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
138 	MOUNT_NFS,	/* 2 */
139 	MOUNT_MFS,	/* 3 */
140 	MOUNT_MSDOS,	/* 4 */
141 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
142 	MOUNT_FDESC,	/* 6 */
143 	MOUNT_KERNFS,	/* 7 */
144 	NULL,		/* 8 = MOUNT_DEVFS */
145 	MOUNT_AFS,	/* 9 */
146 };
147 const int nmountcompatnames = sizeof(mountcompatnames) /
148     sizeof(mountcompatnames[0]);
149 
150 static int
151 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
152     void *data, size_t *data_len)
153 {
154 	struct mount *mp;
155 	int error = 0, saved_flags;
156 
157 	mp = vp->v_mount;
158 	saved_flags = mp->mnt_flag;
159 
160 	/* We can operate only on VV_ROOT nodes. */
161 	if ((vp->v_vflag & VV_ROOT) == 0) {
162 		error = EINVAL;
163 		goto out;
164 	}
165 
166 	/*
167 	 * We only allow the filesystem to be reloaded if it
168 	 * is currently mounted read-only.  Additionally, we
169 	 * prevent read-write to read-only downgrades.
170 	 */
171 	if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
172 	    (mp->mnt_flag & MNT_RDONLY) == 0) {
173 		error = EOPNOTSUPP;	/* Needs translation */
174 		goto out;
175 	}
176 
177 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
178 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
179 	if (error)
180 		goto out;
181 
182 	if (vfs_busy(mp, NULL)) {
183 		error = EPERM;
184 		goto out;
185 	}
186 
187 	mutex_enter(&mp->mnt_updating);
188 
189 	mp->mnt_flag &= ~MNT_OP_FLAGS;
190 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
191 
192 	/*
193 	 * Set the mount level flags.
194 	 */
195 	if (flags & MNT_RDONLY)
196 		mp->mnt_flag |= MNT_RDONLY;
197 	else if (mp->mnt_flag & MNT_RDONLY)
198 		mp->mnt_iflag |= IMNT_WANTRDWR;
199 	mp->mnt_flag &=
200 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
201 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
202 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
203 	    MNT_LOG);
204 	mp->mnt_flag |= flags &
205 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
206 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
207 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
208 	    MNT_LOG | MNT_IGNORE);
209 
210 	error = VFS_MOUNT(mp, path, data, data_len);
211 
212 	if (error && data != NULL) {
213 		int error2;
214 
215 		/*
216 		 * Update failed; let's try and see if it was an
217 		 * export request.  For compat with 3.0 and earlier.
218 		 */
219 		error2 = vfs_hooks_reexport(mp, path, data);
220 
221 		/*
222 		 * Only update error code if the export request was
223 		 * understood but some problem occurred while
224 		 * processing it.
225 		 */
226 		if (error2 != EJUSTRETURN)
227 			error = error2;
228 	}
229 
230 	if (mp->mnt_iflag & IMNT_WANTRDWR)
231 		mp->mnt_flag &= ~MNT_RDONLY;
232 	if (error)
233 		mp->mnt_flag = saved_flags;
234 	mp->mnt_flag &= ~MNT_OP_FLAGS;
235 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
236 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
237 		if (mp->mnt_syncer == NULL)
238 			error = vfs_allocate_syncvnode(mp);
239 	} else {
240 		if (mp->mnt_syncer != NULL)
241 			vfs_deallocate_syncvnode(mp);
242 	}
243 	mutex_exit(&mp->mnt_updating);
244 	vfs_unbusy(mp, false, NULL);
245 
246  out:
247 	return (error);
248 }
249 
250 static int
251 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
252 {
253 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
254 	int error;
255 
256 	/* Copy file-system type from userspace.  */
257 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
258 	if (error) {
259 		/*
260 		 * Historically, filesystem types were identified by numbers.
261 		 * If we get an integer for the filesystem type instead of a
262 		 * string, we check to see if it matches one of the historic
263 		 * filesystem types.
264 		 */
265 		u_long fsindex = (u_long)fstype;
266 		if (fsindex >= nmountcompatnames ||
267 		    mountcompatnames[fsindex] == NULL)
268 			return ENODEV;
269 		strlcpy(fstypename, mountcompatnames[fsindex],
270 		    sizeof(fstypename));
271 	}
272 
273 	/* Accept `ufs' as an alias for `ffs', for compatibility. */
274 	if (strcmp(fstypename, "ufs") == 0)
275 		fstypename[0] = 'f';
276 
277 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
278 		return 0;
279 
280 	/* If we can autoload a vfs module, try again */
281 	mutex_enter(&module_lock);
282 	(void)module_autoload(fstypename, MODULE_CLASS_VFS);
283 	mutex_exit(&module_lock);
284 
285 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
286 		return 0;
287 
288 	return ENODEV;
289 }
290 
291 static int
292 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
293     const char *path, int flags, void *data, size_t *data_len, u_int recurse)
294 {
295 	struct mount *mp;
296 	struct vnode *vp = *vpp;
297 	struct vattr va;
298 	int error;
299 
300 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
301 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
302 	if (error) {
303 		vfs_delref(vfsops);
304 		return error;
305 	}
306 
307 	/* Can't make a non-dir a mount-point (from here anyway). */
308 	if (vp->v_type != VDIR) {
309 		vfs_delref(vfsops);
310 		return ENOTDIR;
311 	}
312 
313 	/*
314 	 * If the user is not root, ensure that they own the directory
315 	 * onto which we are attempting to mount.
316 	 */
317 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
318 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
319 	    (error = kauth_authorize_generic(l->l_cred,
320 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
321 		vfs_delref(vfsops);
322 		return error;
323 	}
324 
325 	if (flags & MNT_EXPORTED) {
326 		vfs_delref(vfsops);
327 		return EINVAL;
328 	}
329 
330 	if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) {
331 		vfs_delref(vfsops);
332 		return error;
333 	}
334 
335 	/*
336 	 * Check if a file-system is not already mounted on this vnode.
337 	 */
338 	if (vp->v_mountedhere != NULL) {
339 		vfs_delref(vfsops);
340 		return EBUSY;
341 	}
342 
343 	if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) {
344 		vfs_delref(vfsops);
345 		return ENOMEM;
346 	}
347 
348 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
349 
350 	/*
351 	 * The underlying file system may refuse the mount for
352 	 * various reasons.  Allow the user to force it to happen.
353 	 *
354 	 * Set the mount level flags.
355 	 */
356 	mp->mnt_flag = flags &
357 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
358 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
359 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
360 	    MNT_LOG | MNT_IGNORE | MNT_RDONLY);
361 
362 	mutex_enter(&mp->mnt_updating);
363 	error = VFS_MOUNT(mp, path, data, data_len);
364 	mp->mnt_flag &= ~MNT_OP_FLAGS;
365 
366 	/*
367 	 * Put the new filesystem on the mount list after root.
368 	 */
369 	cache_purge(vp);
370 	if (error != 0) {
371 		vp->v_mountedhere = NULL;
372 		mutex_exit(&mp->mnt_updating);
373 		vfs_unbusy(mp, false, NULL);
374 		vfs_destroy(mp);
375 		return error;
376 	}
377 
378 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
379 	mutex_enter(&mountlist_lock);
380 	vp->v_mountedhere = mp;
381 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
382 	mutex_exit(&mountlist_lock);
383     	vn_restorerecurse(vp, recurse);
384 	VOP_UNLOCK(vp, 0);
385 	checkdirs(vp);
386 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
387 		error = vfs_allocate_syncvnode(mp);
388 	/* Hold an additional reference to the mount across VFS_START(). */
389 	mutex_exit(&mp->mnt_updating);
390 	vfs_unbusy(mp, true, NULL);
391 	(void) VFS_STATVFS(mp, &mp->mnt_stat);
392 	error = VFS_START(mp, 0);
393 	if (error)
394 		vrele(vp);
395 	/* Drop reference held for VFS_START(). */
396 	vfs_destroy(mp);
397 	*vpp = NULL;
398 	return error;
399 }
400 
401 static int
402 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
403     void *data, size_t *data_len)
404 {
405 	struct mount *mp;
406 	int error;
407 
408 	/* If MNT_GETARGS is specified, it should be the only flag. */
409 	if (flags & ~MNT_GETARGS)
410 		return EINVAL;
411 
412 	mp = vp->v_mount;
413 
414 	/* XXX: probably some notion of "can see" here if we want isolation. */
415 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
416 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
417 	if (error)
418 		return error;
419 
420 	if ((vp->v_vflag & VV_ROOT) == 0)
421 		return EINVAL;
422 
423 	if (vfs_busy(mp, NULL))
424 		return EPERM;
425 
426 	mutex_enter(&mp->mnt_updating);
427 	mp->mnt_flag &= ~MNT_OP_FLAGS;
428 	mp->mnt_flag |= MNT_GETARGS;
429 	error = VFS_MOUNT(mp, path, data, data_len);
430 	mp->mnt_flag &= ~MNT_OP_FLAGS;
431 	mutex_exit(&mp->mnt_updating);
432 
433 	vfs_unbusy(mp, false, NULL);
434 	return (error);
435 }
436 
437 int
438 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
439 {
440 	/* {
441 		syscallarg(const char *) type;
442 		syscallarg(const char *) path;
443 		syscallarg(int) flags;
444 		syscallarg(void *) data;
445 		syscallarg(size_t) data_len;
446 	} */
447 
448 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
449 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
450 	    SCARG(uap, data_len), retval);
451 }
452 
453 int
454 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
455     const char *path, int flags, void *data, enum uio_seg data_seg,
456     size_t data_len, register_t *retval)
457 {
458 	struct vnode *vp;
459 	void *data_buf = data;
460 	u_int recurse;
461 	bool vfsopsrele = false;
462 	int error;
463 
464 	/* XXX: The calling convention of this routine is totally bizarre */
465 	if (vfsops)
466 		vfsopsrele = true;
467 
468 	/*
469 	 * Get vnode to be covered
470 	 */
471 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
472 	if (error != 0) {
473 		/* XXXgcc */
474 		vp = NULL;
475 		recurse = 0;
476 		goto done;
477 	}
478 
479 	/*
480 	 * A lookup in VFS_MOUNT might result in an attempt to
481 	 * lock this vnode again, so make the lock recursive.
482 	 */
483 	if (vfsops == NULL) {
484 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
485 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
486 			recurse = vn_setrecurse(vp);
487 			vfsops = vp->v_mount->mnt_op;
488 		} else {
489 			/* 'type' is userspace */
490 			error = mount_get_vfsops(type, &vfsops);
491 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
492 			recurse = vn_setrecurse(vp);
493 			if (error != 0)
494 				goto done;
495 			vfsopsrele = true;
496 		}
497 	} else {
498 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
499 		recurse = vn_setrecurse(vp);
500 	}
501 
502 	if (data != NULL && data_seg == UIO_USERSPACE) {
503 		if (data_len == 0) {
504 			/* No length supplied, use default for filesystem */
505 			data_len = vfsops->vfs_min_mount_data;
506 			if (data_len > VFS_MAX_MOUNT_DATA) {
507 				error = EINVAL;
508 				goto done;
509 			}
510 			/*
511 			 * Hopefully a longer buffer won't make copyin() fail.
512 			 * For compatibility with 3.0 and earlier.
513 			 */
514 			if (flags & MNT_UPDATE
515 			    && data_len < sizeof (struct mnt_export_args30))
516 				data_len = sizeof (struct mnt_export_args30);
517 		}
518 		data_buf = kmem_alloc(data_len, KM_SLEEP);
519 
520 		/* NFS needs the buffer even for mnt_getargs .... */
521 		error = copyin(data, data_buf, data_len);
522 		if (error != 0)
523 			goto done;
524 	}
525 
526 	if (flags & MNT_GETARGS) {
527 		if (data_len == 0) {
528 			error = EINVAL;
529 			goto done;
530 		}
531 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
532 		if (error != 0)
533 			goto done;
534 		if (data_seg == UIO_USERSPACE)
535 			error = copyout(data_buf, data, data_len);
536 		*retval = data_len;
537 	} else if (flags & MNT_UPDATE) {
538 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
539 	} else {
540 		/* Locking is handled internally in mount_domount(). */
541 		KASSERT(vfsopsrele == true);
542 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
543 		    &data_len, recurse);
544 		vfsopsrele = false;
545 	}
546 
547     done:
548 	if (vfsopsrele)
549 		vfs_delref(vfsops);
550     	if (vp != NULL) {
551 	    	vn_restorerecurse(vp, recurse);
552 	    	vput(vp);
553 	}
554 	if (data_buf != data)
555 		kmem_free(data_buf, data_len);
556 	return (error);
557 }
558 
559 /*
560  * Scan all active processes to see if any of them have a current
561  * or root directory onto which the new filesystem has just been
562  * mounted. If so, replace them with the new mount point.
563  */
564 void
565 checkdirs(struct vnode *olddp)
566 {
567 	struct cwdinfo *cwdi;
568 	struct vnode *newdp, *rele1, *rele2;
569 	struct proc *p;
570 	bool retry;
571 
572 	if (olddp->v_usecount == 1)
573 		return;
574 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
575 		panic("mount: lost mount");
576 
577 	do {
578 		retry = false;
579 		mutex_enter(proc_lock);
580 		PROCLIST_FOREACH(p, &allproc) {
581 			if ((cwdi = p->p_cwdi) == NULL)
582 				continue;
583 			/*
584 			 * Can't change to the old directory any more,
585 			 * so even if we see a stale value it's not a
586 			 * problem.
587 			 */
588 			if (cwdi->cwdi_cdir != olddp &&
589 			    cwdi->cwdi_rdir != olddp)
590 			    	continue;
591 			retry = true;
592 			rele1 = NULL;
593 			rele2 = NULL;
594 			atomic_inc_uint(&cwdi->cwdi_refcnt);
595 			mutex_exit(proc_lock);
596 			rw_enter(&cwdi->cwdi_lock, RW_WRITER);
597 			if (cwdi->cwdi_cdir == olddp) {
598 				rele1 = cwdi->cwdi_cdir;
599 				vref(newdp);
600 				cwdi->cwdi_cdir = newdp;
601 			}
602 			if (cwdi->cwdi_rdir == olddp) {
603 				rele2 = cwdi->cwdi_rdir;
604 				vref(newdp);
605 				cwdi->cwdi_rdir = newdp;
606 			}
607 			rw_exit(&cwdi->cwdi_lock);
608 			cwdfree(cwdi);
609 			if (rele1 != NULL)
610 				vrele(rele1);
611 			if (rele2 != NULL)
612 				vrele(rele2);
613 			mutex_enter(proc_lock);
614 			break;
615 		}
616 		mutex_exit(proc_lock);
617 	} while (retry);
618 
619 	if (rootvnode == olddp) {
620 		vrele(rootvnode);
621 		vref(newdp);
622 		rootvnode = newdp;
623 	}
624 	vput(newdp);
625 }
626 
627 /*
628  * Unmount a file system.
629  *
630  * Note: unmount takes a path to the vnode mounted on as argument,
631  * not special file (as before).
632  */
633 /* ARGSUSED */
634 int
635 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
636 {
637 	/* {
638 		syscallarg(const char *) path;
639 		syscallarg(int) flags;
640 	} */
641 	struct vnode *vp;
642 	struct mount *mp;
643 	int error;
644 	struct nameidata nd;
645 
646 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
647 	    SCARG(uap, path));
648 	if ((error = namei(&nd)) != 0)
649 		return (error);
650 	vp = nd.ni_vp;
651 	mp = vp->v_mount;
652 	atomic_inc_uint(&mp->mnt_refcnt);
653 	VOP_UNLOCK(vp, 0);
654 
655 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
656 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
657 	if (error) {
658 		vrele(vp);
659 		vfs_destroy(mp);
660 		return (error);
661 	}
662 
663 	/*
664 	 * Don't allow unmounting the root file system.
665 	 */
666 	if (mp->mnt_flag & MNT_ROOTFS) {
667 		vrele(vp);
668 		vfs_destroy(mp);
669 		return (EINVAL);
670 	}
671 
672 	/*
673 	 * Must be the root of the filesystem
674 	 */
675 	if ((vp->v_vflag & VV_ROOT) == 0) {
676 		vrele(vp);
677 		vfs_destroy(mp);
678 		return (EINVAL);
679 	}
680 
681 	vrele(vp);
682 	error = dounmount(mp, SCARG(uap, flags), l);
683 	vfs_destroy(mp);
684 	return error;
685 }
686 
687 /*
688  * Do the actual file system unmount.  File system is assumed to have
689  * been locked by the caller.
690  *
691  * => Caller hold reference to the mount, explicitly for dounmount().
692  */
693 int
694 dounmount(struct mount *mp, int flags, struct lwp *l)
695 {
696 	struct vnode *coveredvp;
697 	int error;
698 	int async;
699 	int used_syncer;
700 
701 #if NVERIEXEC > 0
702 	error = veriexec_unmountchk(mp);
703 	if (error)
704 		return (error);
705 #endif /* NVERIEXEC > 0 */
706 
707 	/*
708 	 * XXX Freeze syncer.  Must do this before locking the
709 	 * mount point.  See dounmount() for details.
710 	 */
711 	mutex_enter(&syncer_mutex);
712 	rw_enter(&mp->mnt_unmounting, RW_WRITER);
713 	if ((mp->mnt_iflag & IMNT_GONE) != 0) {
714 		rw_exit(&mp->mnt_unmounting);
715 		mutex_exit(&syncer_mutex);
716 		return ENOENT;
717 	}
718 
719 	used_syncer = (mp->mnt_syncer != NULL);
720 
721 	/*
722 	 * XXX Syncer must be frozen when we get here.  This should really
723 	 * be done on a per-mountpoint basis, but the syncer doesn't work
724 	 * like that.
725 	 *
726 	 * The caller of dounmount() must acquire syncer_mutex because
727 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
728 	 * order, and we must preserve that order to avoid deadlock.
729 	 *
730 	 * So, if the file system did not use the syncer, now is
731 	 * the time to release the syncer_mutex.
732 	 */
733 	if (used_syncer == 0)
734 		mutex_exit(&syncer_mutex);
735 
736 	mp->mnt_iflag |= IMNT_UNMOUNT;
737 	async = mp->mnt_flag & MNT_ASYNC;
738 	mp->mnt_flag &= ~MNT_ASYNC;
739 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
740 	if (mp->mnt_syncer != NULL)
741 		vfs_deallocate_syncvnode(mp);
742 	error = 0;
743 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
744 		error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
745 	}
746 	vfs_scrubvnlist(mp);
747 	if (error == 0 || (flags & MNT_FORCE))
748 		error = VFS_UNMOUNT(mp, flags);
749 	if (error) {
750 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
751 			(void) vfs_allocate_syncvnode(mp);
752 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
753 		mp->mnt_flag |= async;
754 		rw_exit(&mp->mnt_unmounting);
755 		if (used_syncer)
756 			mutex_exit(&syncer_mutex);
757 		return (error);
758 	}
759 	vfs_scrubvnlist(mp);
760 	mutex_enter(&mountlist_lock);
761 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
762 		coveredvp->v_mountedhere = NULL;
763 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
764 	mp->mnt_iflag |= IMNT_GONE;
765 	mutex_exit(&mountlist_lock);
766 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
767 		panic("unmount: dangling vnode");
768 	if (used_syncer)
769 		mutex_exit(&syncer_mutex);
770 	vfs_hooks_unmount(mp);
771 	rw_exit(&mp->mnt_unmounting);
772 	vfs_destroy(mp);	/* reference from mount() */
773 	if (coveredvp != NULLVP)
774 		vrele(coveredvp);
775 	return (0);
776 }
777 
778 /*
779  * Sync each mounted filesystem.
780  */
781 #ifdef DEBUG
782 int syncprt = 0;
783 struct ctldebug debug0 = { "syncprt", &syncprt };
784 #endif
785 
786 /* ARGSUSED */
787 int
788 sys_sync(struct lwp *l, const void *v, register_t *retval)
789 {
790 	struct mount *mp, *nmp;
791 	int asyncflag;
792 
793 	if (l == NULL)
794 		l = &lwp0;
795 
796 	mutex_enter(&mountlist_lock);
797 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
798 	     mp = nmp) {
799 		if (vfs_busy(mp, &nmp)) {
800 			continue;
801 		}
802 		mutex_enter(&mp->mnt_updating);
803 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
804 			asyncflag = mp->mnt_flag & MNT_ASYNC;
805 			mp->mnt_flag &= ~MNT_ASYNC;
806 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
807 			if (asyncflag)
808 				 mp->mnt_flag |= MNT_ASYNC;
809 		}
810 		mutex_exit(&mp->mnt_updating);
811 		vfs_unbusy(mp, false, &nmp);
812 	}
813 	mutex_exit(&mountlist_lock);
814 #ifdef DEBUG
815 	if (syncprt)
816 		vfs_bufstats();
817 #endif /* DEBUG */
818 	return (0);
819 }
820 
821 /*
822  * Change filesystem quotas.
823  */
824 /* ARGSUSED */
825 int
826 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
827 {
828 	/* {
829 		syscallarg(const char *) path;
830 		syscallarg(int) cmd;
831 		syscallarg(int) uid;
832 		syscallarg(void *) arg;
833 	} */
834 	struct mount *mp;
835 	int error;
836 	struct vnode *vp;
837 
838 	error = namei_simple_user(SCARG(uap, path),
839 				NSM_FOLLOW_TRYEMULROOT, &vp);
840 	if (error != 0)
841 		return (error);
842 	mp = vp->v_mount;
843 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
844 	    SCARG(uap, arg));
845 	vrele(vp);
846 	return (error);
847 }
848 
849 int
850 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
851     int root)
852 {
853 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
854 	int error = 0;
855 
856 	/*
857 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
858 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
859 	 * overrides MNT_NOWAIT.
860 	 */
861 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
862 	    (flags != MNT_WAIT && flags != 0)) {
863 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
864 		goto done;
865 	}
866 
867 	/* Get the filesystem stats now */
868 	memset(sp, 0, sizeof(*sp));
869 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
870 		return error;
871 	}
872 
873 	if (cwdi->cwdi_rdir == NULL)
874 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
875 done:
876 	if (cwdi->cwdi_rdir != NULL) {
877 		size_t len;
878 		char *bp;
879 		char c;
880 		char *path = PNBUF_GET();
881 
882 		bp = path + MAXPATHLEN;
883 		*--bp = '\0';
884 		rw_enter(&cwdi->cwdi_lock, RW_READER);
885 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
886 		    MAXPATHLEN / 2, 0, l);
887 		rw_exit(&cwdi->cwdi_lock);
888 		if (error) {
889 			PNBUF_PUT(path);
890 			return error;
891 		}
892 		len = strlen(bp);
893 		if (len != 1) {
894 			/*
895 			 * for mount points that are below our root, we can see
896 			 * them, so we fix up the pathname and return them. The
897 			 * rest we cannot see, so we don't allow viewing the
898 			 * data.
899 			 */
900 			if (strncmp(bp, sp->f_mntonname, len) == 0 &&
901 			    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
902 				(void)strlcpy(sp->f_mntonname,
903 				    c == '\0' ? "/" : &sp->f_mntonname[len],
904 				    sizeof(sp->f_mntonname));
905 			} else {
906 				if (root)
907 					(void)strlcpy(sp->f_mntonname, "/",
908 					    sizeof(sp->f_mntonname));
909 				else
910 					error = EPERM;
911 			}
912 		}
913 		PNBUF_PUT(path);
914 	}
915 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
916 	return error;
917 }
918 
919 /*
920  * Get filesystem statistics by path.
921  */
922 int
923 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
924 {
925 	struct mount *mp;
926 	int error;
927 	struct vnode *vp;
928 
929 	error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
930 	if (error != 0)
931 		return error;
932 	mp = vp->v_mount;
933 	error = dostatvfs(mp, sb, l, flags, 1);
934 	vrele(vp);
935 	return error;
936 }
937 
938 /* ARGSUSED */
939 int
940 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
941 {
942 	/* {
943 		syscallarg(const char *) path;
944 		syscallarg(struct statvfs *) buf;
945 		syscallarg(int) flags;
946 	} */
947 	struct statvfs *sb;
948 	int error;
949 
950 	sb = STATVFSBUF_GET();
951 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
952 	if (error == 0)
953 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
954 	STATVFSBUF_PUT(sb);
955 	return error;
956 }
957 
958 /*
959  * Get filesystem statistics by fd.
960  */
961 int
962 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
963 {
964 	file_t *fp;
965 	struct mount *mp;
966 	int error;
967 
968 	/* fd_getvnode() will use the descriptor for us */
969 	if ((error = fd_getvnode(fd, &fp)) != 0)
970 		return (error);
971 	mp = ((struct vnode *)fp->f_data)->v_mount;
972 	error = dostatvfs(mp, sb, curlwp, flags, 1);
973 	fd_putfile(fd);
974 	return error;
975 }
976 
977 /* ARGSUSED */
978 int
979 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
980 {
981 	/* {
982 		syscallarg(int) fd;
983 		syscallarg(struct statvfs *) buf;
984 		syscallarg(int) flags;
985 	} */
986 	struct statvfs *sb;
987 	int error;
988 
989 	sb = STATVFSBUF_GET();
990 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
991 	if (error == 0)
992 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
993 	STATVFSBUF_PUT(sb);
994 	return error;
995 }
996 
997 
998 /*
999  * Get statistics on all filesystems.
1000  */
1001 int
1002 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1003     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1004     register_t *retval)
1005 {
1006 	int root = 0;
1007 	struct proc *p = l->l_proc;
1008 	struct mount *mp, *nmp;
1009 	struct statvfs *sb;
1010 	size_t count, maxcount;
1011 	int error = 0;
1012 
1013 	sb = STATVFSBUF_GET();
1014 	maxcount = bufsize / entry_sz;
1015 	mutex_enter(&mountlist_lock);
1016 	count = 0;
1017 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1018 	     mp = nmp) {
1019 		if (vfs_busy(mp, &nmp)) {
1020 			continue;
1021 		}
1022 		if (sfsp && count < maxcount) {
1023 			error = dostatvfs(mp, sb, l, flags, 0);
1024 			if (error) {
1025 				vfs_unbusy(mp, false, &nmp);
1026 				error = 0;
1027 				continue;
1028 			}
1029 			error = copyfn(sb, sfsp, entry_sz);
1030 			if (error) {
1031 				vfs_unbusy(mp, false, NULL);
1032 				goto out;
1033 			}
1034 			sfsp = (char *)sfsp + entry_sz;
1035 			root |= strcmp(sb->f_mntonname, "/") == 0;
1036 		}
1037 		count++;
1038 		vfs_unbusy(mp, false, &nmp);
1039 	}
1040 	mutex_exit(&mountlist_lock);
1041 
1042 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
1043 		/*
1044 		 * fake a root entry
1045 		 */
1046 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1047 		    sb, l, flags, 1);
1048 		if (error != 0)
1049 			goto out;
1050 		if (sfsp) {
1051 			error = copyfn(sb, sfsp, entry_sz);
1052 			if (error != 0)
1053 				goto out;
1054 		}
1055 		count++;
1056 	}
1057 	if (sfsp && count > maxcount)
1058 		*retval = maxcount;
1059 	else
1060 		*retval = count;
1061 out:
1062 	STATVFSBUF_PUT(sb);
1063 	return error;
1064 }
1065 
1066 int
1067 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1068 {
1069 	/* {
1070 		syscallarg(struct statvfs *) buf;
1071 		syscallarg(size_t) bufsize;
1072 		syscallarg(int) flags;
1073 	} */
1074 
1075 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1076 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1077 }
1078 
1079 /*
1080  * Change current working directory to a given file descriptor.
1081  */
1082 /* ARGSUSED */
1083 int
1084 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1085 {
1086 	/* {
1087 		syscallarg(int) fd;
1088 	} */
1089 	struct proc *p = l->l_proc;
1090 	struct cwdinfo *cwdi;
1091 	struct vnode *vp, *tdp;
1092 	struct mount *mp;
1093 	file_t *fp;
1094 	int error, fd;
1095 
1096 	/* fd_getvnode() will use the descriptor for us */
1097 	fd = SCARG(uap, fd);
1098 	if ((error = fd_getvnode(fd, &fp)) != 0)
1099 		return (error);
1100 	vp = fp->f_data;
1101 
1102 	vref(vp);
1103 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1104 	if (vp->v_type != VDIR)
1105 		error = ENOTDIR;
1106 	else
1107 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1108 	if (error) {
1109 		vput(vp);
1110 		goto out;
1111 	}
1112 	while ((mp = vp->v_mountedhere) != NULL) {
1113 		error = vfs_busy(mp, NULL);
1114 		vput(vp);
1115 		if (error != 0)
1116 			goto out;
1117 		error = VFS_ROOT(mp, &tdp);
1118 		vfs_unbusy(mp, false, NULL);
1119 		if (error)
1120 			goto out;
1121 		vp = tdp;
1122 	}
1123 	VOP_UNLOCK(vp, 0);
1124 
1125 	/*
1126 	 * Disallow changing to a directory not under the process's
1127 	 * current root directory (if there is one).
1128 	 */
1129 	cwdi = p->p_cwdi;
1130 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1131 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1132 		vrele(vp);
1133 		error = EPERM;	/* operation not permitted */
1134 	} else {
1135 		vrele(cwdi->cwdi_cdir);
1136 		cwdi->cwdi_cdir = vp;
1137 	}
1138 	rw_exit(&cwdi->cwdi_lock);
1139 
1140  out:
1141 	fd_putfile(fd);
1142 	return (error);
1143 }
1144 
1145 /*
1146  * Change this process's notion of the root directory to a given file
1147  * descriptor.
1148  */
1149 int
1150 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1151 {
1152 	struct proc *p = l->l_proc;
1153 	struct vnode	*vp;
1154 	file_t	*fp;
1155 	int		 error, fd = SCARG(uap, fd);
1156 
1157 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1158  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1159 		return error;
1160 	/* fd_getvnode() will use the descriptor for us */
1161 	if ((error = fd_getvnode(fd, &fp)) != 0)
1162 		return error;
1163 	vp = fp->f_data;
1164 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1165 	if (vp->v_type != VDIR)
1166 		error = ENOTDIR;
1167 	else
1168 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1169 	VOP_UNLOCK(vp, 0);
1170 	if (error)
1171 		goto out;
1172 	vref(vp);
1173 
1174 	change_root(p->p_cwdi, vp, l);
1175 
1176  out:
1177 	fd_putfile(fd);
1178 	return (error);
1179 }
1180 
1181 /*
1182  * Change current working directory (``.'').
1183  */
1184 /* ARGSUSED */
1185 int
1186 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1187 {
1188 	/* {
1189 		syscallarg(const char *) path;
1190 	} */
1191 	struct proc *p = l->l_proc;
1192 	struct cwdinfo *cwdi;
1193 	int error;
1194 	struct vnode *vp;
1195 
1196 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1197 				  &vp, l)) != 0)
1198 		return (error);
1199 	cwdi = p->p_cwdi;
1200 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1201 	vrele(cwdi->cwdi_cdir);
1202 	cwdi->cwdi_cdir = vp;
1203 	rw_exit(&cwdi->cwdi_lock);
1204 	return (0);
1205 }
1206 
1207 /*
1208  * Change notion of root (``/'') directory.
1209  */
1210 /* ARGSUSED */
1211 int
1212 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1213 {
1214 	/* {
1215 		syscallarg(const char *) path;
1216 	} */
1217 	struct proc *p = l->l_proc;
1218 	int error;
1219 	struct vnode *vp;
1220 
1221 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1222 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1223 		return (error);
1224 	if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1225 				  &vp, l)) != 0)
1226 		return (error);
1227 
1228 	change_root(p->p_cwdi, vp, l);
1229 
1230 	return (0);
1231 }
1232 
1233 /*
1234  * Common routine for chroot and fchroot.
1235  * NB: callers need to properly authorize the change root operation.
1236  */
1237 void
1238 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1239 {
1240 
1241 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1242 	if (cwdi->cwdi_rdir != NULL)
1243 		vrele(cwdi->cwdi_rdir);
1244 	cwdi->cwdi_rdir = vp;
1245 
1246 	/*
1247 	 * Prevent escaping from chroot by putting the root under
1248 	 * the working directory.  Silently chdir to / if we aren't
1249 	 * already there.
1250 	 */
1251 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1252 		/*
1253 		 * XXX would be more failsafe to change directory to a
1254 		 * deadfs node here instead
1255 		 */
1256 		vrele(cwdi->cwdi_cdir);
1257 		vref(vp);
1258 		cwdi->cwdi_cdir = vp;
1259 	}
1260 	rw_exit(&cwdi->cwdi_lock);
1261 }
1262 
1263 /*
1264  * Common routine for chroot and chdir.
1265  */
1266 int
1267 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1268 {
1269 	struct nameidata nd;
1270 	int error;
1271 
1272 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, where,
1273 	    path);
1274 	if ((error = namei(&nd)) != 0)
1275 		return (error);
1276 	*vpp = nd.ni_vp;
1277 	if ((*vpp)->v_type != VDIR)
1278 		error = ENOTDIR;
1279 	else
1280 		error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1281 
1282 	if (error)
1283 		vput(*vpp);
1284 	else
1285 		VOP_UNLOCK(*vpp, 0);
1286 	return (error);
1287 }
1288 
1289 /*
1290  * Check permissions, allocate an open file structure,
1291  * and call the device open routine if any.
1292  */
1293 int
1294 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1295 {
1296 	/* {
1297 		syscallarg(const char *) path;
1298 		syscallarg(int) flags;
1299 		syscallarg(int) mode;
1300 	} */
1301 	struct proc *p = l->l_proc;
1302 	struct cwdinfo *cwdi = p->p_cwdi;
1303 	file_t *fp;
1304 	struct vnode *vp;
1305 	int flags, cmode;
1306 	int type, indx, error;
1307 	struct flock lf;
1308 	struct nameidata nd;
1309 
1310 	flags = FFLAGS(SCARG(uap, flags));
1311 	if ((flags & (FREAD | FWRITE)) == 0)
1312 		return (EINVAL);
1313 	if ((error = fd_allocfile(&fp, &indx)) != 0)
1314 		return (error);
1315 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1316 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1317 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1318 	    SCARG(uap, path));
1319 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1320 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1321 		fd_abort(p, fp, indx);
1322 		if ((error == EDUPFD || error == EMOVEFD) &&
1323 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1324 		    (error =
1325 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1326 			*retval = indx;
1327 			return (0);
1328 		}
1329 		if (error == ERESTART)
1330 			error = EINTR;
1331 		return (error);
1332 	}
1333 
1334 	l->l_dupfd = 0;
1335 	vp = nd.ni_vp;
1336 	fp->f_flag = flags & FMASK;
1337 	fp->f_type = DTYPE_VNODE;
1338 	fp->f_ops = &vnops;
1339 	fp->f_data = vp;
1340 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1341 		lf.l_whence = SEEK_SET;
1342 		lf.l_start = 0;
1343 		lf.l_len = 0;
1344 		if (flags & O_EXLOCK)
1345 			lf.l_type = F_WRLCK;
1346 		else
1347 			lf.l_type = F_RDLCK;
1348 		type = F_FLOCK;
1349 		if ((flags & FNONBLOCK) == 0)
1350 			type |= F_WAIT;
1351 		VOP_UNLOCK(vp, 0);
1352 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1353 		if (error) {
1354 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1355 			fd_abort(p, fp, indx);
1356 			return (error);
1357 		}
1358 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1359 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1360 	}
1361 	VOP_UNLOCK(vp, 0);
1362 	*retval = indx;
1363 	fd_affix(p, fp, indx);
1364 	return (0);
1365 }
1366 
1367 static void
1368 vfs__fhfree(fhandle_t *fhp)
1369 {
1370 	size_t fhsize;
1371 
1372 	if (fhp == NULL) {
1373 		return;
1374 	}
1375 	fhsize = FHANDLE_SIZE(fhp);
1376 	kmem_free(fhp, fhsize);
1377 }
1378 
1379 /*
1380  * vfs_composefh: compose a filehandle.
1381  */
1382 
1383 int
1384 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1385 {
1386 	struct mount *mp;
1387 	struct fid *fidp;
1388 	int error;
1389 	size_t needfhsize;
1390 	size_t fidsize;
1391 
1392 	mp = vp->v_mount;
1393 	fidp = NULL;
1394 	if (*fh_size < FHANDLE_SIZE_MIN) {
1395 		fidsize = 0;
1396 	} else {
1397 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1398 		if (fhp != NULL) {
1399 			memset(fhp, 0, *fh_size);
1400 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1401 			fidp = &fhp->fh_fid;
1402 		}
1403 	}
1404 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1405 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1406 	if (error == 0 && *fh_size < needfhsize) {
1407 		error = E2BIG;
1408 	}
1409 	*fh_size = needfhsize;
1410 	return error;
1411 }
1412 
1413 int
1414 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1415 {
1416 	struct mount *mp;
1417 	fhandle_t *fhp;
1418 	size_t fhsize;
1419 	size_t fidsize;
1420 	int error;
1421 
1422 	*fhpp = NULL;
1423 	mp = vp->v_mount;
1424 	fidsize = 0;
1425 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1426 	KASSERT(error != 0);
1427 	if (error != E2BIG) {
1428 		goto out;
1429 	}
1430 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1431 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1432 	if (fhp == NULL) {
1433 		error = ENOMEM;
1434 		goto out;
1435 	}
1436 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1437 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1438 	if (error == 0) {
1439 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1440 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1441 		*fhpp = fhp;
1442 	} else {
1443 		kmem_free(fhp, fhsize);
1444 	}
1445 out:
1446 	return error;
1447 }
1448 
1449 void
1450 vfs_composefh_free(fhandle_t *fhp)
1451 {
1452 
1453 	vfs__fhfree(fhp);
1454 }
1455 
1456 /*
1457  * vfs_fhtovp: lookup a vnode by a filehandle.
1458  */
1459 
1460 int
1461 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1462 {
1463 	struct mount *mp;
1464 	int error;
1465 
1466 	*vpp = NULL;
1467 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1468 	if (mp == NULL) {
1469 		error = ESTALE;
1470 		goto out;
1471 	}
1472 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1473 		error = EOPNOTSUPP;
1474 		goto out;
1475 	}
1476 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1477 out:
1478 	return error;
1479 }
1480 
1481 /*
1482  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1483  * the needed size.
1484  */
1485 
1486 int
1487 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1488 {
1489 	fhandle_t *fhp;
1490 	int error;
1491 
1492 	*fhpp = NULL;
1493 	if (fhsize > FHANDLE_SIZE_MAX) {
1494 		return EINVAL;
1495 	}
1496 	if (fhsize < FHANDLE_SIZE_MIN) {
1497 		return EINVAL;
1498 	}
1499 again:
1500 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1501 	if (fhp == NULL) {
1502 		return ENOMEM;
1503 	}
1504 	error = copyin(ufhp, fhp, fhsize);
1505 	if (error == 0) {
1506 		/* XXX this check shouldn't be here */
1507 		if (FHANDLE_SIZE(fhp) == fhsize) {
1508 			*fhpp = fhp;
1509 			return 0;
1510 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1511 			/*
1512 			 * a kludge for nfsv2 padded handles.
1513 			 */
1514 			size_t sz;
1515 
1516 			sz = FHANDLE_SIZE(fhp);
1517 			kmem_free(fhp, fhsize);
1518 			fhsize = sz;
1519 			goto again;
1520 		} else {
1521 			/*
1522 			 * userland told us wrong size.
1523 			 */
1524 		    	error = EINVAL;
1525 		}
1526 	}
1527 	kmem_free(fhp, fhsize);
1528 	return error;
1529 }
1530 
1531 void
1532 vfs_copyinfh_free(fhandle_t *fhp)
1533 {
1534 
1535 	vfs__fhfree(fhp);
1536 }
1537 
1538 /*
1539  * Get file handle system call
1540  */
1541 int
1542 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1543 {
1544 	/* {
1545 		syscallarg(char *) fname;
1546 		syscallarg(fhandle_t *) fhp;
1547 		syscallarg(size_t *) fh_size;
1548 	} */
1549 	struct vnode *vp;
1550 	fhandle_t *fh;
1551 	int error;
1552 	struct nameidata nd;
1553 	size_t sz;
1554 	size_t usz;
1555 
1556 	/*
1557 	 * Must be super user
1558 	 */
1559 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1560 	    0, NULL, NULL, NULL);
1561 	if (error)
1562 		return (error);
1563 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1564 	    SCARG(uap, fname));
1565 	error = namei(&nd);
1566 	if (error)
1567 		return (error);
1568 	vp = nd.ni_vp;
1569 	error = vfs_composefh_alloc(vp, &fh);
1570 	vput(vp);
1571 	if (error != 0) {
1572 		goto out;
1573 	}
1574 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1575 	if (error != 0) {
1576 		goto out;
1577 	}
1578 	sz = FHANDLE_SIZE(fh);
1579 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1580 	if (error != 0) {
1581 		goto out;
1582 	}
1583 	if (usz >= sz) {
1584 		error = copyout(fh, SCARG(uap, fhp), sz);
1585 	} else {
1586 		error = E2BIG;
1587 	}
1588 out:
1589 	vfs_composefh_free(fh);
1590 	return (error);
1591 }
1592 
1593 /*
1594  * Open a file given a file handle.
1595  *
1596  * Check permissions, allocate an open file structure,
1597  * and call the device open routine if any.
1598  */
1599 
1600 int
1601 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1602     register_t *retval)
1603 {
1604 	file_t *fp;
1605 	struct vnode *vp = NULL;
1606 	kauth_cred_t cred = l->l_cred;
1607 	file_t *nfp;
1608 	int type, indx, error=0;
1609 	struct flock lf;
1610 	struct vattr va;
1611 	fhandle_t *fh;
1612 	int flags;
1613 	proc_t *p;
1614 
1615 	p = curproc;
1616 
1617 	/*
1618 	 * Must be super user
1619 	 */
1620 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1621 	    0, NULL, NULL, NULL)))
1622 		return (error);
1623 
1624 	flags = FFLAGS(oflags);
1625 	if ((flags & (FREAD | FWRITE)) == 0)
1626 		return (EINVAL);
1627 	if ((flags & O_CREAT))
1628 		return (EINVAL);
1629 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1630 		return (error);
1631 	fp = nfp;
1632 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1633 	if (error != 0) {
1634 		goto bad;
1635 	}
1636 	error = vfs_fhtovp(fh, &vp);
1637 	if (error != 0) {
1638 		goto bad;
1639 	}
1640 
1641 	/* Now do an effective vn_open */
1642 
1643 	if (vp->v_type == VSOCK) {
1644 		error = EOPNOTSUPP;
1645 		goto bad;
1646 	}
1647 	error = vn_openchk(vp, cred, flags);
1648 	if (error != 0)
1649 		goto bad;
1650 	if (flags & O_TRUNC) {
1651 		VOP_UNLOCK(vp, 0);			/* XXX */
1652 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1653 		vattr_null(&va);
1654 		va.va_size = 0;
1655 		error = VOP_SETATTR(vp, &va, cred);
1656 		if (error)
1657 			goto bad;
1658 	}
1659 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1660 		goto bad;
1661 	if (flags & FWRITE) {
1662 		mutex_enter(&vp->v_interlock);
1663 		vp->v_writecount++;
1664 		mutex_exit(&vp->v_interlock);
1665 	}
1666 
1667 	/* done with modified vn_open, now finish what sys_open does. */
1668 
1669 	fp->f_flag = flags & FMASK;
1670 	fp->f_type = DTYPE_VNODE;
1671 	fp->f_ops = &vnops;
1672 	fp->f_data = vp;
1673 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1674 		lf.l_whence = SEEK_SET;
1675 		lf.l_start = 0;
1676 		lf.l_len = 0;
1677 		if (flags & O_EXLOCK)
1678 			lf.l_type = F_WRLCK;
1679 		else
1680 			lf.l_type = F_RDLCK;
1681 		type = F_FLOCK;
1682 		if ((flags & FNONBLOCK) == 0)
1683 			type |= F_WAIT;
1684 		VOP_UNLOCK(vp, 0);
1685 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1686 		if (error) {
1687 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1688 			fd_abort(p, fp, indx);
1689 			return (error);
1690 		}
1691 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1692 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1693 	}
1694 	VOP_UNLOCK(vp, 0);
1695 	*retval = indx;
1696 	fd_affix(p, fp, indx);
1697 	vfs_copyinfh_free(fh);
1698 	return (0);
1699 
1700 bad:
1701 	fd_abort(p, fp, indx);
1702 	if (vp != NULL)
1703 		vput(vp);
1704 	vfs_copyinfh_free(fh);
1705 	return (error);
1706 }
1707 
1708 int
1709 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1710 {
1711 	/* {
1712 		syscallarg(const void *) fhp;
1713 		syscallarg(size_t) fh_size;
1714 		syscallarg(int) flags;
1715 	} */
1716 
1717 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1718 	    SCARG(uap, flags), retval);
1719 }
1720 
1721 int
1722 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1723 {
1724 	int error;
1725 	fhandle_t *fh;
1726 	struct vnode *vp;
1727 
1728 	/*
1729 	 * Must be super user
1730 	 */
1731 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1732 	    0, NULL, NULL, NULL)))
1733 		return (error);
1734 
1735 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1736 	if (error != 0)
1737 		return error;
1738 
1739 	error = vfs_fhtovp(fh, &vp);
1740 	vfs_copyinfh_free(fh);
1741 	if (error != 0)
1742 		return error;
1743 
1744 	error = vn_stat(vp, sb);
1745 	vput(vp);
1746 	return error;
1747 }
1748 
1749 
1750 /* ARGSUSED */
1751 int
1752 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1753 {
1754 	/* {
1755 		syscallarg(const void *) fhp;
1756 		syscallarg(size_t) fh_size;
1757 		syscallarg(struct stat *) sb;
1758 	} */
1759 	struct stat sb;
1760 	int error;
1761 
1762 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1763 	if (error)
1764 		return error;
1765 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1766 }
1767 
1768 int
1769 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1770     int flags)
1771 {
1772 	fhandle_t *fh;
1773 	struct mount *mp;
1774 	struct vnode *vp;
1775 	int error;
1776 
1777 	/*
1778 	 * Must be super user
1779 	 */
1780 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1781 	    0, NULL, NULL, NULL)))
1782 		return error;
1783 
1784 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1785 	if (error != 0)
1786 		return error;
1787 
1788 	error = vfs_fhtovp(fh, &vp);
1789 	vfs_copyinfh_free(fh);
1790 	if (error != 0)
1791 		return error;
1792 
1793 	mp = vp->v_mount;
1794 	error = dostatvfs(mp, sb, l, flags, 1);
1795 	vput(vp);
1796 	return error;
1797 }
1798 
1799 /* ARGSUSED */
1800 int
1801 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1802 {
1803 	/* {
1804 		syscallarg(const void *) fhp;
1805 		syscallarg(size_t) fh_size;
1806 		syscallarg(struct statvfs *) buf;
1807 		syscallarg(int)	flags;
1808 	} */
1809 	struct statvfs *sb = STATVFSBUF_GET();
1810 	int error;
1811 
1812 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1813 	    SCARG(uap, flags));
1814 	if (error == 0)
1815 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1816 	STATVFSBUF_PUT(sb);
1817 	return error;
1818 }
1819 
1820 /*
1821  * Create a special file.
1822  */
1823 /* ARGSUSED */
1824 int
1825 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
1826     register_t *retval)
1827 {
1828 	/* {
1829 		syscallarg(const char *) path;
1830 		syscallarg(mode_t) mode;
1831 		syscallarg(dev_t) dev;
1832 	} */
1833 	return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
1834 	    SCARG(uap, dev), retval, UIO_USERSPACE);
1835 }
1836 
1837 int
1838 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
1839     register_t *retval, enum uio_seg seg)
1840 {
1841 	struct proc *p = l->l_proc;
1842 	struct vnode *vp;
1843 	struct vattr vattr;
1844 	int error, optype;
1845 	struct nameidata nd;
1846 	char *path;
1847 	const char *cpath;
1848 
1849 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1850 	    0, NULL, NULL, NULL)) != 0)
1851 		return (error);
1852 
1853 	optype = VOP_MKNOD_DESCOFFSET;
1854 
1855 	VERIEXEC_PATH_GET(pathname, seg, cpath, path);
1856 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
1857 
1858 	if ((error = namei(&nd)) != 0)
1859 		goto out;
1860 	vp = nd.ni_vp;
1861 	if (vp != NULL)
1862 		error = EEXIST;
1863 	else {
1864 		vattr_null(&vattr);
1865 		/* We will read cwdi->cwdi_cmask unlocked. */
1866 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1867 		vattr.va_rdev = dev;
1868 
1869 		switch (mode & S_IFMT) {
1870 		case S_IFMT:	/* used by badsect to flag bad sectors */
1871 			vattr.va_type = VBAD;
1872 			break;
1873 		case S_IFCHR:
1874 			vattr.va_type = VCHR;
1875 			break;
1876 		case S_IFBLK:
1877 			vattr.va_type = VBLK;
1878 			break;
1879 		case S_IFWHT:
1880 			optype = VOP_WHITEOUT_DESCOFFSET;
1881 			break;
1882 		case S_IFREG:
1883 #if NVERIEXEC > 0
1884 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1885 			    O_CREAT);
1886 #endif /* NVERIEXEC > 0 */
1887 			vattr.va_type = VREG;
1888 			vattr.va_rdev = VNOVAL;
1889 			optype = VOP_CREATE_DESCOFFSET;
1890 			break;
1891 		default:
1892 			error = EINVAL;
1893 			break;
1894 		}
1895 	}
1896 	if (!error) {
1897 		switch (optype) {
1898 		case VOP_WHITEOUT_DESCOFFSET:
1899 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1900 			if (error)
1901 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1902 			vput(nd.ni_dvp);
1903 			break;
1904 
1905 		case VOP_MKNOD_DESCOFFSET:
1906 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1907 						&nd.ni_cnd, &vattr);
1908 			if (error == 0)
1909 				vput(nd.ni_vp);
1910 			break;
1911 
1912 		case VOP_CREATE_DESCOFFSET:
1913 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1914 						&nd.ni_cnd, &vattr);
1915 			if (error == 0)
1916 				vput(nd.ni_vp);
1917 			break;
1918 		}
1919 	} else {
1920 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1921 		if (nd.ni_dvp == vp)
1922 			vrele(nd.ni_dvp);
1923 		else
1924 			vput(nd.ni_dvp);
1925 		if (vp)
1926 			vrele(vp);
1927 	}
1928 out:
1929 	VERIEXEC_PATH_PUT(path);
1930 	return (error);
1931 }
1932 
1933 /*
1934  * Create a named pipe.
1935  */
1936 /* ARGSUSED */
1937 int
1938 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1939 {
1940 	/* {
1941 		syscallarg(const char *) path;
1942 		syscallarg(int) mode;
1943 	} */
1944 	struct proc *p = l->l_proc;
1945 	struct vattr vattr;
1946 	int error;
1947 	struct nameidata nd;
1948 
1949 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1950 	    SCARG(uap, path));
1951 	if ((error = namei(&nd)) != 0)
1952 		return (error);
1953 	if (nd.ni_vp != NULL) {
1954 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1955 		if (nd.ni_dvp == nd.ni_vp)
1956 			vrele(nd.ni_dvp);
1957 		else
1958 			vput(nd.ni_dvp);
1959 		vrele(nd.ni_vp);
1960 		return (EEXIST);
1961 	}
1962 	vattr_null(&vattr);
1963 	vattr.va_type = VFIFO;
1964 	/* We will read cwdi->cwdi_cmask unlocked. */
1965 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1966 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1967 	if (error == 0)
1968 		vput(nd.ni_vp);
1969 	return (error);
1970 }
1971 
1972 /*
1973  * Make a hard file link.
1974  */
1975 /* ARGSUSED */
1976 int
1977 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
1978 {
1979 	/* {
1980 		syscallarg(const char *) path;
1981 		syscallarg(const char *) link;
1982 	} */
1983 	struct vnode *vp;
1984 	struct nameidata nd;
1985 	int error;
1986 
1987 	error = namei_simple_user(SCARG(uap, path),
1988 				NSM_FOLLOW_TRYEMULROOT, &vp);
1989 	if (error != 0)
1990 		return (error);
1991 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1992 	    SCARG(uap, link));
1993 	if ((error = namei(&nd)) != 0)
1994 		goto out;
1995 	if (nd.ni_vp) {
1996 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1997 		if (nd.ni_dvp == nd.ni_vp)
1998 			vrele(nd.ni_dvp);
1999 		else
2000 			vput(nd.ni_dvp);
2001 		vrele(nd.ni_vp);
2002 		error = EEXIST;
2003 		goto out;
2004 	}
2005 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2006 out:
2007 	vrele(vp);
2008 	return (error);
2009 }
2010 
2011 /*
2012  * Make a symbolic link.
2013  */
2014 /* ARGSUSED */
2015 int
2016 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2017 {
2018 	/* {
2019 		syscallarg(const char *) path;
2020 		syscallarg(const char *) link;
2021 	} */
2022 	struct proc *p = l->l_proc;
2023 	struct vattr vattr;
2024 	char *path;
2025 	int error;
2026 	struct nameidata nd;
2027 
2028 	path = PNBUF_GET();
2029 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
2030 	if (error)
2031 		goto out;
2032 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2033 	    SCARG(uap, link));
2034 	if ((error = namei(&nd)) != 0)
2035 		goto out;
2036 	if (nd.ni_vp) {
2037 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2038 		if (nd.ni_dvp == nd.ni_vp)
2039 			vrele(nd.ni_dvp);
2040 		else
2041 			vput(nd.ni_dvp);
2042 		vrele(nd.ni_vp);
2043 		error = EEXIST;
2044 		goto out;
2045 	}
2046 	vattr_null(&vattr);
2047 	vattr.va_type = VLNK;
2048 	/* We will read cwdi->cwdi_cmask unlocked. */
2049 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2050 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2051 	if (error == 0)
2052 		vput(nd.ni_vp);
2053 out:
2054 	PNBUF_PUT(path);
2055 	return (error);
2056 }
2057 
2058 /*
2059  * Delete a whiteout from the filesystem.
2060  */
2061 /* ARGSUSED */
2062 int
2063 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2064 {
2065 	/* {
2066 		syscallarg(const char *) path;
2067 	} */
2068 	int error;
2069 	struct nameidata nd;
2070 
2071 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
2072 	    UIO_USERSPACE, SCARG(uap, path));
2073 	error = namei(&nd);
2074 	if (error)
2075 		return (error);
2076 
2077 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2078 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2079 		if (nd.ni_dvp == nd.ni_vp)
2080 			vrele(nd.ni_dvp);
2081 		else
2082 			vput(nd.ni_dvp);
2083 		if (nd.ni_vp)
2084 			vrele(nd.ni_vp);
2085 		return (EEXIST);
2086 	}
2087 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2088 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2089 	vput(nd.ni_dvp);
2090 	return (error);
2091 }
2092 
2093 /*
2094  * Delete a name from the filesystem.
2095  */
2096 /* ARGSUSED */
2097 int
2098 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2099 {
2100 	/* {
2101 		syscallarg(const char *) path;
2102 	} */
2103 
2104 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2105 }
2106 
2107 int
2108 do_sys_unlink(const char *arg, enum uio_seg seg)
2109 {
2110 	struct vnode *vp;
2111 	int error;
2112 	struct nameidata nd;
2113 	char *path;
2114 	const char *cpath;
2115 
2116 	VERIEXEC_PATH_GET(arg, seg, cpath, path);
2117 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
2118 
2119 	if ((error = namei(&nd)) != 0)
2120 		goto out;
2121 	vp = nd.ni_vp;
2122 
2123 	/*
2124 	 * The root of a mounted filesystem cannot be deleted.
2125 	 */
2126 	if (vp->v_vflag & VV_ROOT) {
2127 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2128 		if (nd.ni_dvp == vp)
2129 			vrele(nd.ni_dvp);
2130 		else
2131 			vput(nd.ni_dvp);
2132 		vput(vp);
2133 		error = EBUSY;
2134 		goto out;
2135 	}
2136 
2137 #if NVERIEXEC > 0
2138 	/* Handle remove requests for veriexec entries. */
2139 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
2140 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2141 		if (nd.ni_dvp == vp)
2142 			vrele(nd.ni_dvp);
2143 		else
2144 			vput(nd.ni_dvp);
2145 		vput(vp);
2146 		goto out;
2147 	}
2148 #endif /* NVERIEXEC > 0 */
2149 
2150 #ifdef FILEASSOC
2151 	(void)fileassoc_file_delete(vp);
2152 #endif /* FILEASSOC */
2153 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2154 out:
2155 	VERIEXEC_PATH_PUT(path);
2156 	return (error);
2157 }
2158 
2159 /*
2160  * Reposition read/write file offset.
2161  */
2162 int
2163 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2164 {
2165 	/* {
2166 		syscallarg(int) fd;
2167 		syscallarg(int) pad;
2168 		syscallarg(off_t) offset;
2169 		syscallarg(int) whence;
2170 	} */
2171 	kauth_cred_t cred = l->l_cred;
2172 	file_t *fp;
2173 	struct vnode *vp;
2174 	struct vattr vattr;
2175 	off_t newoff;
2176 	int error, fd;
2177 
2178 	fd = SCARG(uap, fd);
2179 
2180 	if ((fp = fd_getfile(fd)) == NULL)
2181 		return (EBADF);
2182 
2183 	vp = fp->f_data;
2184 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2185 		error = ESPIPE;
2186 		goto out;
2187 	}
2188 
2189 	switch (SCARG(uap, whence)) {
2190 	case SEEK_CUR:
2191 		newoff = fp->f_offset + SCARG(uap, offset);
2192 		break;
2193 	case SEEK_END:
2194 		error = VOP_GETATTR(vp, &vattr, cred);
2195 		if (error) {
2196 			goto out;
2197 		}
2198 		newoff = SCARG(uap, offset) + vattr.va_size;
2199 		break;
2200 	case SEEK_SET:
2201 		newoff = SCARG(uap, offset);
2202 		break;
2203 	default:
2204 		error = EINVAL;
2205 		goto out;
2206 	}
2207 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2208 		*(off_t *)retval = fp->f_offset = newoff;
2209 	}
2210  out:
2211  	fd_putfile(fd);
2212 	return (error);
2213 }
2214 
2215 /*
2216  * Positional read system call.
2217  */
2218 int
2219 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2220 {
2221 	/* {
2222 		syscallarg(int) fd;
2223 		syscallarg(void *) buf;
2224 		syscallarg(size_t) nbyte;
2225 		syscallarg(off_t) offset;
2226 	} */
2227 	file_t *fp;
2228 	struct vnode *vp;
2229 	off_t offset;
2230 	int error, fd = SCARG(uap, fd);
2231 
2232 	if ((fp = fd_getfile(fd)) == NULL)
2233 		return (EBADF);
2234 
2235 	if ((fp->f_flag & FREAD) == 0) {
2236 		fd_putfile(fd);
2237 		return (EBADF);
2238 	}
2239 
2240 	vp = fp->f_data;
2241 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2242 		error = ESPIPE;
2243 		goto out;
2244 	}
2245 
2246 	offset = SCARG(uap, offset);
2247 
2248 	/*
2249 	 * XXX This works because no file systems actually
2250 	 * XXX take any action on the seek operation.
2251 	 */
2252 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2253 		goto out;
2254 
2255 	/* dofileread() will unuse the descriptor for us */
2256 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2257 	    &offset, 0, retval));
2258 
2259  out:
2260 	fd_putfile(fd);
2261 	return (error);
2262 }
2263 
2264 /*
2265  * Positional scatter read system call.
2266  */
2267 int
2268 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2269 {
2270 	/* {
2271 		syscallarg(int) fd;
2272 		syscallarg(const struct iovec *) iovp;
2273 		syscallarg(int) iovcnt;
2274 		syscallarg(off_t) offset;
2275 	} */
2276 	off_t offset = SCARG(uap, offset);
2277 
2278 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2279 	    SCARG(uap, iovcnt), &offset, 0, retval);
2280 }
2281 
2282 /*
2283  * Positional write system call.
2284  */
2285 int
2286 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2287 {
2288 	/* {
2289 		syscallarg(int) fd;
2290 		syscallarg(const void *) buf;
2291 		syscallarg(size_t) nbyte;
2292 		syscallarg(off_t) offset;
2293 	} */
2294 	file_t *fp;
2295 	struct vnode *vp;
2296 	off_t offset;
2297 	int error, fd = SCARG(uap, fd);
2298 
2299 	if ((fp = fd_getfile(fd)) == NULL)
2300 		return (EBADF);
2301 
2302 	if ((fp->f_flag & FWRITE) == 0) {
2303 		fd_putfile(fd);
2304 		return (EBADF);
2305 	}
2306 
2307 	vp = fp->f_data;
2308 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2309 		error = ESPIPE;
2310 		goto out;
2311 	}
2312 
2313 	offset = SCARG(uap, offset);
2314 
2315 	/*
2316 	 * XXX This works because no file systems actually
2317 	 * XXX take any action on the seek operation.
2318 	 */
2319 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2320 		goto out;
2321 
2322 	/* dofilewrite() will unuse the descriptor for us */
2323 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2324 	    &offset, 0, retval));
2325 
2326  out:
2327 	fd_putfile(fd);
2328 	return (error);
2329 }
2330 
2331 /*
2332  * Positional gather write system call.
2333  */
2334 int
2335 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2336 {
2337 	/* {
2338 		syscallarg(int) fd;
2339 		syscallarg(const struct iovec *) iovp;
2340 		syscallarg(int) iovcnt;
2341 		syscallarg(off_t) offset;
2342 	} */
2343 	off_t offset = SCARG(uap, offset);
2344 
2345 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2346 	    SCARG(uap, iovcnt), &offset, 0, retval);
2347 }
2348 
2349 /*
2350  * Check access permissions.
2351  */
2352 int
2353 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2354 {
2355 	/* {
2356 		syscallarg(const char *) path;
2357 		syscallarg(int) flags;
2358 	} */
2359 	kauth_cred_t cred;
2360 	struct vnode *vp;
2361 	int error, flags;
2362 	struct nameidata nd;
2363 
2364 	cred = kauth_cred_dup(l->l_cred);
2365 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2366 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2367 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2368 	    SCARG(uap, path));
2369 	/* Override default credentials */
2370 	nd.ni_cnd.cn_cred = cred;
2371 	if ((error = namei(&nd)) != 0)
2372 		goto out;
2373 	vp = nd.ni_vp;
2374 
2375 	/* Flags == 0 means only check for existence. */
2376 	if (SCARG(uap, flags)) {
2377 		flags = 0;
2378 		if (SCARG(uap, flags) & R_OK)
2379 			flags |= VREAD;
2380 		if (SCARG(uap, flags) & W_OK)
2381 			flags |= VWRITE;
2382 		if (SCARG(uap, flags) & X_OK)
2383 			flags |= VEXEC;
2384 
2385 		error = VOP_ACCESS(vp, flags, cred);
2386 		if (!error && (flags & VWRITE))
2387 			error = vn_writechk(vp);
2388 	}
2389 	vput(vp);
2390 out:
2391 	kauth_cred_free(cred);
2392 	return (error);
2393 }
2394 
2395 /*
2396  * Common code for all sys_stat functions, including compat versions.
2397  */
2398 int
2399 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
2400 {
2401 	int error;
2402 	struct nameidata nd;
2403 
2404 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
2405 	    UIO_USERSPACE, path);
2406 	error = namei(&nd);
2407 	if (error != 0)
2408 		return error;
2409 	error = vn_stat(nd.ni_vp, sb);
2410 	vput(nd.ni_vp);
2411 	return error;
2412 }
2413 
2414 /*
2415  * Get file status; this version follows links.
2416  */
2417 /* ARGSUSED */
2418 int
2419 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
2420 {
2421 	/* {
2422 		syscallarg(const char *) path;
2423 		syscallarg(struct stat *) ub;
2424 	} */
2425 	struct stat sb;
2426 	int error;
2427 
2428 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2429 	if (error)
2430 		return error;
2431 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2432 }
2433 
2434 /*
2435  * Get file status; this version does not follow links.
2436  */
2437 /* ARGSUSED */
2438 int
2439 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
2440 {
2441 	/* {
2442 		syscallarg(const char *) path;
2443 		syscallarg(struct stat *) ub;
2444 	} */
2445 	struct stat sb;
2446 	int error;
2447 
2448 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2449 	if (error)
2450 		return error;
2451 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2452 }
2453 
2454 /*
2455  * Get configurable pathname variables.
2456  */
2457 /* ARGSUSED */
2458 int
2459 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2460 {
2461 	/* {
2462 		syscallarg(const char *) path;
2463 		syscallarg(int) name;
2464 	} */
2465 	int error;
2466 	struct nameidata nd;
2467 
2468 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2469 	    SCARG(uap, path));
2470 	if ((error = namei(&nd)) != 0)
2471 		return (error);
2472 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2473 	vput(nd.ni_vp);
2474 	return (error);
2475 }
2476 
2477 /*
2478  * Return target name of a symbolic link.
2479  */
2480 /* ARGSUSED */
2481 int
2482 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2483 {
2484 	/* {
2485 		syscallarg(const char *) path;
2486 		syscallarg(char *) buf;
2487 		syscallarg(size_t) count;
2488 	} */
2489 	struct vnode *vp;
2490 	struct iovec aiov;
2491 	struct uio auio;
2492 	int error;
2493 	struct nameidata nd;
2494 
2495 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2496 	    SCARG(uap, path));
2497 	if ((error = namei(&nd)) != 0)
2498 		return (error);
2499 	vp = nd.ni_vp;
2500 	if (vp->v_type != VLNK)
2501 		error = EINVAL;
2502 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2503 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2504 		aiov.iov_base = SCARG(uap, buf);
2505 		aiov.iov_len = SCARG(uap, count);
2506 		auio.uio_iov = &aiov;
2507 		auio.uio_iovcnt = 1;
2508 		auio.uio_offset = 0;
2509 		auio.uio_rw = UIO_READ;
2510 		KASSERT(l == curlwp);
2511 		auio.uio_vmspace = l->l_proc->p_vmspace;
2512 		auio.uio_resid = SCARG(uap, count);
2513 		error = VOP_READLINK(vp, &auio, l->l_cred);
2514 	}
2515 	vput(vp);
2516 	*retval = SCARG(uap, count) - auio.uio_resid;
2517 	return (error);
2518 }
2519 
2520 /*
2521  * Change flags of a file given a path name.
2522  */
2523 /* ARGSUSED */
2524 int
2525 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2526 {
2527 	/* {
2528 		syscallarg(const char *) path;
2529 		syscallarg(u_long) flags;
2530 	} */
2531 	struct vnode *vp;
2532 	int error;
2533 
2534 	error = namei_simple_user(SCARG(uap, path),
2535 				NSM_FOLLOW_TRYEMULROOT, &vp);
2536 	if (error != 0)
2537 		return (error);
2538 	error = change_flags(vp, SCARG(uap, flags), l);
2539 	vput(vp);
2540 	return (error);
2541 }
2542 
2543 /*
2544  * Change flags of a file given a file descriptor.
2545  */
2546 /* ARGSUSED */
2547 int
2548 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
2549 {
2550 	/* {
2551 		syscallarg(int) fd;
2552 		syscallarg(u_long) flags;
2553 	} */
2554 	struct vnode *vp;
2555 	file_t *fp;
2556 	int error;
2557 
2558 	/* fd_getvnode() will use the descriptor for us */
2559 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2560 		return (error);
2561 	vp = fp->f_data;
2562 	error = change_flags(vp, SCARG(uap, flags), l);
2563 	VOP_UNLOCK(vp, 0);
2564 	fd_putfile(SCARG(uap, fd));
2565 	return (error);
2566 }
2567 
2568 /*
2569  * Change flags of a file given a path name; this version does
2570  * not follow links.
2571  */
2572 int
2573 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
2574 {
2575 	/* {
2576 		syscallarg(const char *) path;
2577 		syscallarg(u_long) flags;
2578 	} */
2579 	struct vnode *vp;
2580 	int error;
2581 
2582 	error = namei_simple_user(SCARG(uap, path),
2583 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2584 	if (error != 0)
2585 		return (error);
2586 	error = change_flags(vp, SCARG(uap, flags), l);
2587 	vput(vp);
2588 	return (error);
2589 }
2590 
2591 /*
2592  * Common routine to change flags of a file.
2593  */
2594 int
2595 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2596 {
2597 	struct vattr vattr;
2598 	int error;
2599 
2600 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2601 	/*
2602 	 * Non-superusers cannot change the flags on devices, even if they
2603 	 * own them.
2604 	 */
2605 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2606 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2607 			goto out;
2608 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2609 			error = EINVAL;
2610 			goto out;
2611 		}
2612 	}
2613 	vattr_null(&vattr);
2614 	vattr.va_flags = flags;
2615 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2616 out:
2617 	return (error);
2618 }
2619 
2620 /*
2621  * Change mode of a file given path name; this version follows links.
2622  */
2623 /* ARGSUSED */
2624 int
2625 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
2626 {
2627 	/* {
2628 		syscallarg(const char *) path;
2629 		syscallarg(int) mode;
2630 	} */
2631 	int error;
2632 	struct vnode *vp;
2633 
2634 	error = namei_simple_user(SCARG(uap, path),
2635 				NSM_FOLLOW_TRYEMULROOT, &vp);
2636 	if (error != 0)
2637 		return (error);
2638 
2639 	error = change_mode(vp, SCARG(uap, mode), l);
2640 
2641 	vrele(vp);
2642 	return (error);
2643 }
2644 
2645 /*
2646  * Change mode of a file given a file descriptor.
2647  */
2648 /* ARGSUSED */
2649 int
2650 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
2651 {
2652 	/* {
2653 		syscallarg(int) fd;
2654 		syscallarg(int) mode;
2655 	} */
2656 	file_t *fp;
2657 	int error;
2658 
2659 	/* fd_getvnode() will use the descriptor for us */
2660 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2661 		return (error);
2662 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
2663 	fd_putfile(SCARG(uap, fd));
2664 	return (error);
2665 }
2666 
2667 /*
2668  * Change mode of a file given path name; this version does not follow links.
2669  */
2670 /* ARGSUSED */
2671 int
2672 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
2673 {
2674 	/* {
2675 		syscallarg(const char *) path;
2676 		syscallarg(int) mode;
2677 	} */
2678 	int error;
2679 	struct vnode *vp;
2680 
2681 	error = namei_simple_user(SCARG(uap, path),
2682 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2683 	if (error != 0)
2684 		return (error);
2685 
2686 	error = change_mode(vp, SCARG(uap, mode), l);
2687 
2688 	vrele(vp);
2689 	return (error);
2690 }
2691 
2692 /*
2693  * Common routine to set mode given a vnode.
2694  */
2695 static int
2696 change_mode(struct vnode *vp, int mode, struct lwp *l)
2697 {
2698 	struct vattr vattr;
2699 	int error;
2700 
2701 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2702 	vattr_null(&vattr);
2703 	vattr.va_mode = mode & ALLPERMS;
2704 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2705 	VOP_UNLOCK(vp, 0);
2706 	return (error);
2707 }
2708 
2709 /*
2710  * Set ownership given a path name; this version follows links.
2711  */
2712 /* ARGSUSED */
2713 int
2714 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
2715 {
2716 	/* {
2717 		syscallarg(const char *) path;
2718 		syscallarg(uid_t) uid;
2719 		syscallarg(gid_t) gid;
2720 	} */
2721 	int error;
2722 	struct vnode *vp;
2723 
2724 	error = namei_simple_user(SCARG(uap, path),
2725 				NSM_FOLLOW_TRYEMULROOT, &vp);
2726 	if (error != 0)
2727 		return (error);
2728 
2729 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2730 
2731 	vrele(vp);
2732 	return (error);
2733 }
2734 
2735 /*
2736  * Set ownership given a path name; this version follows links.
2737  * Provides POSIX semantics.
2738  */
2739 /* ARGSUSED */
2740 int
2741 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
2742 {
2743 	/* {
2744 		syscallarg(const char *) path;
2745 		syscallarg(uid_t) uid;
2746 		syscallarg(gid_t) gid;
2747 	} */
2748 	int error;
2749 	struct vnode *vp;
2750 
2751 	error = namei_simple_user(SCARG(uap, path),
2752 				NSM_FOLLOW_TRYEMULROOT, &vp);
2753 	if (error != 0)
2754 		return (error);
2755 
2756 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2757 
2758 	vrele(vp);
2759 	return (error);
2760 }
2761 
2762 /*
2763  * Set ownership given a file descriptor.
2764  */
2765 /* ARGSUSED */
2766 int
2767 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
2768 {
2769 	/* {
2770 		syscallarg(int) fd;
2771 		syscallarg(uid_t) uid;
2772 		syscallarg(gid_t) gid;
2773 	} */
2774 	int error;
2775 	file_t *fp;
2776 
2777 	/* fd_getvnode() will use the descriptor for us */
2778 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2779 		return (error);
2780 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2781 	    l, 0);
2782 	fd_putfile(SCARG(uap, fd));
2783 	return (error);
2784 }
2785 
2786 /*
2787  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2788  */
2789 /* ARGSUSED */
2790 int
2791 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
2792 {
2793 	/* {
2794 		syscallarg(int) fd;
2795 		syscallarg(uid_t) uid;
2796 		syscallarg(gid_t) gid;
2797 	} */
2798 	int error;
2799 	file_t *fp;
2800 
2801 	/* fd_getvnode() will use the descriptor for us */
2802 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2803 		return (error);
2804 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2805 	    l, 1);
2806 	fd_putfile(SCARG(uap, fd));
2807 	return (error);
2808 }
2809 
2810 /*
2811  * Set ownership given a path name; this version does not follow links.
2812  */
2813 /* ARGSUSED */
2814 int
2815 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
2816 {
2817 	/* {
2818 		syscallarg(const char *) path;
2819 		syscallarg(uid_t) uid;
2820 		syscallarg(gid_t) gid;
2821 	} */
2822 	int error;
2823 	struct vnode *vp;
2824 
2825 	error = namei_simple_user(SCARG(uap, path),
2826 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2827 	if (error != 0)
2828 		return (error);
2829 
2830 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2831 
2832 	vrele(vp);
2833 	return (error);
2834 }
2835 
2836 /*
2837  * Set ownership given a path name; this version does not follow links.
2838  * Provides POSIX/XPG semantics.
2839  */
2840 /* ARGSUSED */
2841 int
2842 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
2843 {
2844 	/* {
2845 		syscallarg(const char *) path;
2846 		syscallarg(uid_t) uid;
2847 		syscallarg(gid_t) gid;
2848 	} */
2849 	int error;
2850 	struct vnode *vp;
2851 
2852 	error = namei_simple_user(SCARG(uap, path),
2853 				NSM_NOFOLLOW_TRYEMULROOT, &vp);
2854 	if (error != 0)
2855 		return (error);
2856 
2857 	error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2858 
2859 	vrele(vp);
2860 	return (error);
2861 }
2862 
2863 /*
2864  * Common routine to set ownership given a vnode.
2865  */
2866 static int
2867 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2868     int posix_semantics)
2869 {
2870 	struct vattr vattr;
2871 	mode_t newmode;
2872 	int error;
2873 
2874 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2875 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2876 		goto out;
2877 
2878 #define CHANGED(x) ((int)(x) != -1)
2879 	newmode = vattr.va_mode;
2880 	if (posix_semantics) {
2881 		/*
2882 		 * POSIX/XPG semantics: if the caller is not the super-user,
2883 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2884 		 * the XPG consider the behaviour for calls by the super-user
2885 		 * implementation-defined; we leave the set-user-id and set-
2886 		 * group-id settings intact in that case.
2887 		 */
2888 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2889 				      NULL) != 0)
2890 			newmode &= ~(S_ISUID | S_ISGID);
2891 	} else {
2892 		/*
2893 		 * NetBSD semantics: when changing owner and/or group,
2894 		 * clear the respective bit(s).
2895 		 */
2896 		if (CHANGED(uid))
2897 			newmode &= ~S_ISUID;
2898 		if (CHANGED(gid))
2899 			newmode &= ~S_ISGID;
2900 	}
2901 	/* Update va_mode iff altered. */
2902 	if (vattr.va_mode == newmode)
2903 		newmode = VNOVAL;
2904 
2905 	vattr_null(&vattr);
2906 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2907 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2908 	vattr.va_mode = newmode;
2909 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2910 #undef CHANGED
2911 
2912 out:
2913 	VOP_UNLOCK(vp, 0);
2914 	return (error);
2915 }
2916 
2917 /*
2918  * Set the access and modification times given a path name; this
2919  * version follows links.
2920  */
2921 /* ARGSUSED */
2922 int
2923 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
2924     register_t *retval)
2925 {
2926 	/* {
2927 		syscallarg(const char *) path;
2928 		syscallarg(const struct timeval *) tptr;
2929 	} */
2930 
2931 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
2932 	    SCARG(uap, tptr), UIO_USERSPACE);
2933 }
2934 
2935 /*
2936  * Set the access and modification times given a file descriptor.
2937  */
2938 /* ARGSUSED */
2939 int
2940 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
2941     register_t *retval)
2942 {
2943 	/* {
2944 		syscallarg(int) fd;
2945 		syscallarg(const struct timeval *) tptr;
2946 	} */
2947 	int error;
2948 	file_t *fp;
2949 
2950 	/* fd_getvnode() will use the descriptor for us */
2951 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2952 		return (error);
2953 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
2954 	    UIO_USERSPACE);
2955 	fd_putfile(SCARG(uap, fd));
2956 	return (error);
2957 }
2958 
2959 /*
2960  * Set the access and modification times given a path name; this
2961  * version does not follow links.
2962  */
2963 int
2964 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
2965     register_t *retval)
2966 {
2967 	/* {
2968 		syscallarg(const char *) path;
2969 		syscallarg(const struct timeval *) tptr;
2970 	} */
2971 
2972 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
2973 	    SCARG(uap, tptr), UIO_USERSPACE);
2974 }
2975 
2976 /*
2977  * Common routine to set access and modification times given a vnode.
2978  */
2979 int
2980 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
2981     const struct timeval *tptr, enum uio_seg seg)
2982 {
2983 	struct vattr vattr;
2984 	int error, dorele = 0;
2985 	namei_simple_flags_t sflags;
2986 
2987 	bool vanull, setbirthtime;
2988 	struct timespec ts[2];
2989 
2990 	/*
2991 	 * I have checked all callers and they pass either FOLLOW,
2992 	 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
2993 	 * is 0. More to the point, they don't pass anything else.
2994 	 * Let's keep it that way at least until the namei interfaces
2995 	 * are fully sanitized.
2996 	 */
2997 	KASSERT(flag == NOFOLLOW || flag == FOLLOW);
2998 	sflags = (flag == FOLLOW) ?
2999 		NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3000 
3001 	if (tptr == NULL) {
3002 		vanull = true;
3003 		nanotime(&ts[0]);
3004 		ts[1] = ts[0];
3005 	} else {
3006 		struct timeval tv[2];
3007 
3008 		vanull = false;
3009 		if (seg != UIO_SYSSPACE) {
3010 			error = copyin(tptr, tv, sizeof (tv));
3011 			if (error != 0)
3012 				return error;
3013 			tptr = tv;
3014 		}
3015 		TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3016 		TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3017 	}
3018 
3019 	if (vp == NULL) {
3020 		/* note: SEG describes TPTR, not PATH; PATH is always user */
3021 		error = namei_simple_user(path, sflags, &vp);
3022 		if (error != 0)
3023 			return error;
3024 		dorele = 1;
3025 	}
3026 
3027 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3028 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3029 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
3030 	vattr_null(&vattr);
3031 	vattr.va_atime = ts[0];
3032 	vattr.va_mtime = ts[1];
3033 	if (setbirthtime)
3034 		vattr.va_birthtime = ts[1];
3035 	if (vanull)
3036 		vattr.va_vaflags |= VA_UTIMES_NULL;
3037 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3038 	VOP_UNLOCK(vp, 0);
3039 
3040 	if (dorele != 0)
3041 		vrele(vp);
3042 
3043 	return error;
3044 }
3045 
3046 /*
3047  * Truncate a file given its path name.
3048  */
3049 /* ARGSUSED */
3050 int
3051 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3052 {
3053 	/* {
3054 		syscallarg(const char *) path;
3055 		syscallarg(int) pad;
3056 		syscallarg(off_t) length;
3057 	} */
3058 	struct vnode *vp;
3059 	struct vattr vattr;
3060 	int error;
3061 
3062 	error = namei_simple_user(SCARG(uap, path),
3063 				NSM_FOLLOW_TRYEMULROOT, &vp);
3064 	if (error != 0)
3065 		return (error);
3066 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3067 	if (vp->v_type == VDIR)
3068 		error = EISDIR;
3069 	else if ((error = vn_writechk(vp)) == 0 &&
3070 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3071 		vattr_null(&vattr);
3072 		vattr.va_size = SCARG(uap, length);
3073 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3074 	}
3075 	vput(vp);
3076 	return (error);
3077 }
3078 
3079 /*
3080  * Truncate a file given a file descriptor.
3081  */
3082 /* ARGSUSED */
3083 int
3084 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3085 {
3086 	/* {
3087 		syscallarg(int) fd;
3088 		syscallarg(int) pad;
3089 		syscallarg(off_t) length;
3090 	} */
3091 	struct vattr vattr;
3092 	struct vnode *vp;
3093 	file_t *fp;
3094 	int error;
3095 
3096 	/* fd_getvnode() will use the descriptor for us */
3097 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3098 		return (error);
3099 	if ((fp->f_flag & FWRITE) == 0) {
3100 		error = EINVAL;
3101 		goto out;
3102 	}
3103 	vp = fp->f_data;
3104 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3105 	if (vp->v_type == VDIR)
3106 		error = EISDIR;
3107 	else if ((error = vn_writechk(vp)) == 0) {
3108 		vattr_null(&vattr);
3109 		vattr.va_size = SCARG(uap, length);
3110 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3111 	}
3112 	VOP_UNLOCK(vp, 0);
3113  out:
3114 	fd_putfile(SCARG(uap, fd));
3115 	return (error);
3116 }
3117 
3118 /*
3119  * Sync an open file.
3120  */
3121 /* ARGSUSED */
3122 int
3123 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3124 {
3125 	/* {
3126 		syscallarg(int) fd;
3127 	} */
3128 	struct vnode *vp;
3129 	file_t *fp;
3130 	int error;
3131 
3132 	/* fd_getvnode() will use the descriptor for us */
3133 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3134 		return (error);
3135 	vp = fp->f_data;
3136 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3137 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3138 	VOP_UNLOCK(vp, 0);
3139 	fd_putfile(SCARG(uap, fd));
3140 	return (error);
3141 }
3142 
3143 /*
3144  * Sync a range of file data.  API modeled after that found in AIX.
3145  *
3146  * FDATASYNC indicates that we need only save enough metadata to be able
3147  * to re-read the written data.  Note we duplicate AIX's requirement that
3148  * the file be open for writing.
3149  */
3150 /* ARGSUSED */
3151 int
3152 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3153 {
3154 	/* {
3155 		syscallarg(int) fd;
3156 		syscallarg(int) flags;
3157 		syscallarg(off_t) start;
3158 		syscallarg(off_t) length;
3159 	} */
3160 	struct vnode *vp;
3161 	file_t *fp;
3162 	int flags, nflags;
3163 	off_t s, e, len;
3164 	int error;
3165 
3166 	/* fd_getvnode() will use the descriptor for us */
3167 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3168 		return (error);
3169 
3170 	if ((fp->f_flag & FWRITE) == 0) {
3171 		error = EBADF;
3172 		goto out;
3173 	}
3174 
3175 	flags = SCARG(uap, flags);
3176 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3177 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3178 		error = EINVAL;
3179 		goto out;
3180 	}
3181 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3182 	if (flags & FDATASYNC)
3183 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3184 	else
3185 		nflags = FSYNC_WAIT;
3186 	if (flags & FDISKSYNC)
3187 		nflags |= FSYNC_CACHE;
3188 
3189 	len = SCARG(uap, length);
3190 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3191 	if (len) {
3192 		s = SCARG(uap, start);
3193 		e = s + len;
3194 		if (e < s) {
3195 			error = EINVAL;
3196 			goto out;
3197 		}
3198 	} else {
3199 		e = 0;
3200 		s = 0;
3201 	}
3202 
3203 	vp = fp->f_data;
3204 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3205 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3206 	VOP_UNLOCK(vp, 0);
3207 out:
3208 	fd_putfile(SCARG(uap, fd));
3209 	return (error);
3210 }
3211 
3212 /*
3213  * Sync the data of an open file.
3214  */
3215 /* ARGSUSED */
3216 int
3217 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3218 {
3219 	/* {
3220 		syscallarg(int) fd;
3221 	} */
3222 	struct vnode *vp;
3223 	file_t *fp;
3224 	int error;
3225 
3226 	/* fd_getvnode() will use the descriptor for us */
3227 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3228 		return (error);
3229 	if ((fp->f_flag & FWRITE) == 0) {
3230 		fd_putfile(SCARG(uap, fd));
3231 		return (EBADF);
3232 	}
3233 	vp = fp->f_data;
3234 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3235 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3236 	VOP_UNLOCK(vp, 0);
3237 	fd_putfile(SCARG(uap, fd));
3238 	return (error);
3239 }
3240 
3241 /*
3242  * Rename files, (standard) BSD semantics frontend.
3243  */
3244 /* ARGSUSED */
3245 int
3246 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3247 {
3248 	/* {
3249 		syscallarg(const char *) from;
3250 		syscallarg(const char *) to;
3251 	} */
3252 
3253 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3254 }
3255 
3256 /*
3257  * Rename files, POSIX semantics frontend.
3258  */
3259 /* ARGSUSED */
3260 int
3261 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3262 {
3263 	/* {
3264 		syscallarg(const char *) from;
3265 		syscallarg(const char *) to;
3266 	} */
3267 
3268 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3269 }
3270 
3271 /*
3272  * Rename files.  Source and destination must either both be directories,
3273  * or both not be directories.  If target is a directory, it must be empty.
3274  * If `from' and `to' refer to the same object, the value of the `retain'
3275  * argument is used to determine whether `from' will be
3276  *
3277  * (retain == 0)	deleted unless `from' and `to' refer to the same
3278  *			object in the file system's name space (BSD).
3279  * (retain == 1)	always retained (POSIX).
3280  */
3281 int
3282 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3283 {
3284 	struct vnode *tvp, *fvp, *tdvp;
3285 	struct nameidata fromnd, tond;
3286 	struct mount *fs;
3287 	struct lwp *l = curlwp;
3288 	struct proc *p;
3289 	uint32_t saveflag;
3290 	int error;
3291 
3292 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT | INRENAME,
3293 	    seg, from);
3294 	if ((error = namei(&fromnd)) != 0)
3295 		return (error);
3296 	if (fromnd.ni_dvp != fromnd.ni_vp)
3297 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3298 	fvp = fromnd.ni_vp;
3299 
3300 	fs = fvp->v_mount;
3301 	error = VFS_RENAMELOCK_ENTER(fs);
3302 	if (error) {
3303 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3304 		vrele(fromnd.ni_dvp);
3305 		vrele(fvp);
3306 		goto out1;
3307 	}
3308 
3309 	/*
3310 	 * close, partially, yet another race - ideally we should only
3311 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3312 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3313 	 * that with namei as it stands.
3314 	 *
3315 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3316 	 * under us. The real fix is to get the locks in the right
3317 	 * order and do the lookups in the right places, but that's a
3318 	 * major rototill.
3319 	 *
3320 	 * Preserve the SAVESTART in cn_flags, because who knows what
3321 	 * might happen if we don't.
3322 	 *
3323 	 * Note: this logic (as well as this whole function) is cloned
3324 	 * in nfs_serv.c. Proceed accordingly.
3325 	 */
3326 	vrele(fvp);
3327 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3328 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3329 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3330 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3331 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3332 		error = EINVAL;
3333 		VFS_RENAMELOCK_EXIT(fs);
3334 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3335 		vrele(fromnd.ni_dvp);
3336 		goto out1;
3337 	}
3338 	saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
3339 	fromnd.ni_cnd.cn_flags &= ~SAVESTART;
3340 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3341 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
3342 	fromnd.ni_cnd.cn_flags |= saveflag;
3343 	if (error) {
3344 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3345 		VFS_RENAMELOCK_EXIT(fs);
3346 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3347 		vrele(fromnd.ni_dvp);
3348 		goto out1;
3349 	}
3350 	VOP_UNLOCK(fromnd.ni_vp, 0);
3351 	if (fromnd.ni_dvp != fromnd.ni_vp)
3352 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3353 	fvp = fromnd.ni_vp;
3354 
3355 	NDINIT(&tond, RENAME,
3356 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3357 	      | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0),
3358 	    seg, to);
3359 	if ((error = namei(&tond)) != 0) {
3360 		VFS_RENAMELOCK_EXIT(fs);
3361 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3362 		vrele(fromnd.ni_dvp);
3363 		vrele(fvp);
3364 		goto out1;
3365 	}
3366 	tdvp = tond.ni_dvp;
3367 	tvp = tond.ni_vp;
3368 
3369 	if (tvp != NULL) {
3370 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3371 			error = ENOTDIR;
3372 			goto out;
3373 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3374 			error = EISDIR;
3375 			goto out;
3376 		}
3377 	}
3378 
3379 	if (fvp == tdvp)
3380 		error = EINVAL;
3381 
3382 	/*
3383 	 * Source and destination refer to the same object.
3384 	 */
3385 	if (fvp == tvp) {
3386 		if (retain)
3387 			error = -1;
3388 		else if (fromnd.ni_dvp == tdvp &&
3389 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3390 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3391 		          tond.ni_cnd.cn_nameptr,
3392 		          fromnd.ni_cnd.cn_namelen))
3393 		error = -1;
3394 	}
3395 
3396 #if NVERIEXEC > 0
3397 	if (!error) {
3398 		char *f1, *f2;
3399 		size_t f1_len;
3400 		size_t f2_len;
3401 
3402 		f1_len = fromnd.ni_cnd.cn_namelen + 1;
3403 		f1 = kmem_alloc(f1_len, KM_SLEEP);
3404 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
3405 
3406 		f2_len = tond.ni_cnd.cn_namelen + 1;
3407 		f2 = kmem_alloc(f2_len, KM_SLEEP);
3408 		strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
3409 
3410 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
3411 
3412 		kmem_free(f1, f1_len);
3413 		kmem_free(f2, f2_len);
3414 	}
3415 #endif /* NVERIEXEC > 0 */
3416 
3417 out:
3418 	p = l->l_proc;
3419 	if (!error) {
3420 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3421 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3422 		VFS_RENAMELOCK_EXIT(fs);
3423 	} else {
3424 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3425 		if (tdvp == tvp)
3426 			vrele(tdvp);
3427 		else
3428 			vput(tdvp);
3429 		if (tvp)
3430 			vput(tvp);
3431 		VFS_RENAMELOCK_EXIT(fs);
3432 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3433 		vrele(fromnd.ni_dvp);
3434 		vrele(fvp);
3435 	}
3436 	vrele(tond.ni_startdir);
3437 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3438 out1:
3439 	if (fromnd.ni_startdir)
3440 		vrele(fromnd.ni_startdir);
3441 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3442 	return (error == -1 ? 0 : error);
3443 }
3444 
3445 /*
3446  * Make a directory file.
3447  */
3448 /* ARGSUSED */
3449 int
3450 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
3451 {
3452 	/* {
3453 		syscallarg(const char *) path;
3454 		syscallarg(int) mode;
3455 	} */
3456 
3457 	return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE);
3458 }
3459 
3460 int
3461 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
3462 {
3463 	struct proc *p = curlwp->l_proc;
3464 	struct vnode *vp;
3465 	struct vattr vattr;
3466 	int error;
3467 	struct nameidata nd;
3468 
3469 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT,
3470 	    seg, path);
3471 	if ((error = namei(&nd)) != 0)
3472 		return (error);
3473 	vp = nd.ni_vp;
3474 	if (vp != NULL) {
3475 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3476 		if (nd.ni_dvp == vp)
3477 			vrele(nd.ni_dvp);
3478 		else
3479 			vput(nd.ni_dvp);
3480 		vrele(vp);
3481 		return (EEXIST);
3482 	}
3483 	vattr_null(&vattr);
3484 	vattr.va_type = VDIR;
3485 	/* We will read cwdi->cwdi_cmask unlocked. */
3486 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3487 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3488 	if (!error)
3489 		vput(nd.ni_vp);
3490 	return (error);
3491 }
3492 
3493 /*
3494  * Remove a directory file.
3495  */
3496 /* ARGSUSED */
3497 int
3498 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
3499 {
3500 	/* {
3501 		syscallarg(const char *) path;
3502 	} */
3503 	struct vnode *vp;
3504 	int error;
3505 	struct nameidata nd;
3506 
3507 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
3508 	    SCARG(uap, path));
3509 	if ((error = namei(&nd)) != 0)
3510 		return (error);
3511 	vp = nd.ni_vp;
3512 	if (vp->v_type != VDIR) {
3513 		error = ENOTDIR;
3514 		goto out;
3515 	}
3516 	/*
3517 	 * No rmdir "." please.
3518 	 */
3519 	if (nd.ni_dvp == vp) {
3520 		error = EINVAL;
3521 		goto out;
3522 	}
3523 	/*
3524 	 * The root of a mounted filesystem cannot be deleted.
3525 	 */
3526 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
3527 		error = EBUSY;
3528 		goto out;
3529 	}
3530 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3531 	return (error);
3532 
3533 out:
3534 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3535 	if (nd.ni_dvp == vp)
3536 		vrele(nd.ni_dvp);
3537 	else
3538 		vput(nd.ni_dvp);
3539 	vput(vp);
3540 	return (error);
3541 }
3542 
3543 /*
3544  * Read a block of directory entries in a file system independent format.
3545  */
3546 int
3547 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
3548 {
3549 	/* {
3550 		syscallarg(int) fd;
3551 		syscallarg(char *) buf;
3552 		syscallarg(size_t) count;
3553 	} */
3554 	file_t *fp;
3555 	int error, done;
3556 
3557 	/* fd_getvnode() will use the descriptor for us */
3558 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3559 		return (error);
3560 	if ((fp->f_flag & FREAD) == 0) {
3561 		error = EBADF;
3562 		goto out;
3563 	}
3564 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3565 			SCARG(uap, count), &done, l, 0, 0);
3566 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3567 	*retval = done;
3568  out:
3569 	fd_putfile(SCARG(uap, fd));
3570 	return (error);
3571 }
3572 
3573 /*
3574  * Set the mode mask for creation of filesystem nodes.
3575  */
3576 int
3577 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
3578 {
3579 	/* {
3580 		syscallarg(mode_t) newmask;
3581 	} */
3582 	struct proc *p = l->l_proc;
3583 	struct cwdinfo *cwdi;
3584 
3585 	/*
3586 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3587 	 * important is that we serialize changes to the mask.  The
3588 	 * rw_exit() will issue a write memory barrier on our behalf,
3589 	 * and force the changes out to other CPUs (as it must use an
3590 	 * atomic operation, draining the local CPU's store buffers).
3591 	 */
3592 	cwdi = p->p_cwdi;
3593 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3594 	*retval = cwdi->cwdi_cmask;
3595 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3596 	rw_exit(&cwdi->cwdi_lock);
3597 
3598 	return (0);
3599 }
3600 
3601 int
3602 dorevoke(struct vnode *vp, kauth_cred_t cred)
3603 {
3604 	struct vattr vattr;
3605 	int error;
3606 
3607 	if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
3608 		return error;
3609 	if (kauth_cred_geteuid(cred) == vattr.va_uid ||
3610 	    (error = kauth_authorize_generic(cred,
3611 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
3612 		VOP_REVOKE(vp, REVOKEALL);
3613 	return (error);
3614 }
3615 
3616 /*
3617  * Void all references to file by ripping underlying filesystem
3618  * away from vnode.
3619  */
3620 /* ARGSUSED */
3621 int
3622 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
3623 {
3624 	/* {
3625 		syscallarg(const char *) path;
3626 	} */
3627 	struct vnode *vp;
3628 	int error;
3629 
3630 	error = namei_simple_user(SCARG(uap, path),
3631 				NSM_FOLLOW_TRYEMULROOT, &vp);
3632 	if (error != 0)
3633 		return (error);
3634 	error = dorevoke(vp, l->l_cred);
3635 	vrele(vp);
3636 	return (error);
3637 }
3638