xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 8ac07aec990b9d2e483062509d0a9fa5b4f57cf2)
1 /*	$NetBSD: vfs_syscalls.c,v 1.350 2008/04/25 13:40:55 joerg Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the NetBSD
18  *	Foundation, Inc. and its contributors.
19  * 4. Neither the name of The NetBSD Foundation nor the names of its
20  *    contributors may be used to endorse or promote products derived
21  *    from this software without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
27  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 /*
37  * Copyright (c) 1989, 1993
38  *	The Regents of the University of California.  All rights reserved.
39  * (c) UNIX System Laboratories, Inc.
40  * All or some portions of this file are derived from material licensed
41  * to the University of California by American Telephone and Telegraph
42  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
43  * the permission of UNIX System Laboratories, Inc.
44  *
45  * Redistribution and use in source and binary forms, with or without
46  * modification, are permitted provided that the following conditions
47  * are met:
48  * 1. Redistributions of source code must retain the above copyright
49  *    notice, this list of conditions and the following disclaimer.
50  * 2. Redistributions in binary form must reproduce the above copyright
51  *    notice, this list of conditions and the following disclaimer in the
52  *    documentation and/or other materials provided with the distribution.
53  * 3. Neither the name of the University nor the names of its contributors
54  *    may be used to endorse or promote products derived from this software
55  *    without specific prior written permission.
56  *
57  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67  * SUCH DAMAGE.
68  *
69  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
70  */
71 
72 #include <sys/cdefs.h>
73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.350 2008/04/25 13:40:55 joerg Exp $");
74 
75 #include "opt_compat_netbsd.h"
76 #include "opt_compat_43.h"
77 #include "opt_fileassoc.h"
78 #include "fss.h"
79 #include "veriexec.h"
80 
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/namei.h>
84 #include <sys/filedesc.h>
85 #include <sys/kernel.h>
86 #include <sys/file.h>
87 #include <sys/stat.h>
88 #include <sys/vnode.h>
89 #include <sys/mount.h>
90 #include <sys/proc.h>
91 #include <sys/uio.h>
92 #include <sys/malloc.h>
93 #include <sys/kmem.h>
94 #include <sys/dirent.h>
95 #include <sys/sysctl.h>
96 #include <sys/syscallargs.h>
97 #include <sys/vfs_syscalls.h>
98 #include <sys/ktrace.h>
99 #ifdef FILEASSOC
100 #include <sys/fileassoc.h>
101 #endif /* FILEASSOC */
102 #include <sys/verified_exec.h>
103 #include <sys/kauth.h>
104 #include <sys/atomic.h>
105 
106 #include <miscfs/genfs/genfs.h>
107 #include <miscfs/syncfs/syncfs.h>
108 #include <miscfs/specfs/specdev.h>
109 
110 #ifdef COMPAT_30
111 #include "opt_nfsserver.h"
112 #include <nfs/rpcv2.h>
113 #endif
114 #include <nfs/nfsproto.h>
115 #ifdef COMPAT_30
116 #include <nfs/nfs.h>
117 #include <nfs/nfs_var.h>
118 #endif
119 
120 #if NFSS > 0
121 #include <dev/fssvar.h>
122 #endif
123 
124 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
125 
126 static int change_dir(struct nameidata *, struct lwp *);
127 static int change_flags(struct vnode *, u_long, struct lwp *);
128 static int change_mode(struct vnode *, int, struct lwp *l);
129 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
130 
131 void checkdirs(struct vnode *);
132 
133 int dovfsusermount = 0;
134 
135 /*
136  * Virtual File System System Calls
137  */
138 
139 /*
140  * Mount a file system.
141  */
142 
143 #if defined(COMPAT_09) || defined(COMPAT_43)
144 /*
145  * This table is used to maintain compatibility with 4.3BSD
146  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
147  *
148  * Do not modify this table. It should only contain filesystems
149  * supported by NetBSD 0.9 and 4.3BSD.
150  */
151 const char * const mountcompatnames[] = {
152 	NULL,		/* 0 = MOUNT_NONE */
153 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
154 	MOUNT_NFS,	/* 2 */
155 	MOUNT_MFS,	/* 3 */
156 	MOUNT_MSDOS,	/* 4 */
157 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
158 	MOUNT_FDESC,	/* 6 */
159 	MOUNT_KERNFS,	/* 7 */
160 	NULL,		/* 8 = MOUNT_DEVFS */
161 	MOUNT_AFS,	/* 9 */
162 };
163 const int nmountcompatnames = sizeof(mountcompatnames) /
164     sizeof(mountcompatnames[0]);
165 #endif /* COMPAT_09 || COMPAT_43 */
166 
167 static int
168 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
169     void *data, size_t *data_len)
170 {
171 	struct mount *mp;
172 	int error = 0, saved_flags;
173 
174 	mp = vp->v_mount;
175 	saved_flags = mp->mnt_flag;
176 
177 	/* We can operate only on VV_ROOT nodes. */
178 	if ((vp->v_vflag & VV_ROOT) == 0) {
179 		error = EINVAL;
180 		goto out;
181 	}
182 
183 	/*
184 	 * We only allow the filesystem to be reloaded if it
185 	 * is currently mounted read-only.
186 	 */
187 	if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) {
188 		error = EOPNOTSUPP;	/* Needs translation */
189 		goto out;
190 	}
191 
192 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
193 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
194 	if (error)
195 		goto out;
196 
197 	if (vfs_trybusy(mp, RW_WRITER, 0)) {
198 		error = EPERM;
199 		goto out;
200 	}
201 
202 	mp->mnt_flag &= ~MNT_OP_FLAGS;
203 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
204 
205 	/*
206 	 * Set the mount level flags.
207 	 */
208 	if (flags & MNT_RDONLY)
209 		mp->mnt_flag |= MNT_RDONLY;
210 	else if (mp->mnt_flag & MNT_RDONLY)
211 		mp->mnt_iflag |= IMNT_WANTRDWR;
212 	mp->mnt_flag &=
213 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
214 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
215 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
216 	mp->mnt_flag |= flags &
217 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
218 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
219 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
220 	    MNT_IGNORE);
221 
222 	error = VFS_MOUNT(mp, path, data, data_len);
223 
224 #if defined(COMPAT_30) && defined(NFSSERVER)
225 	if (error && data != NULL) {
226 		int error2;
227 
228 		/* Update failed; let's try and see if it was an
229 		 * export request. */
230 		error2 = nfs_update_exports_30(mp, path, data, l);
231 
232 		/* Only update error code if the export request was
233 		 * understood but some problem occurred while
234 		 * processing it. */
235 		if (error2 != EJUSTRETURN)
236 			error = error2;
237 	}
238 #endif
239 	if (mp->mnt_iflag & IMNT_WANTRDWR)
240 		mp->mnt_flag &= ~MNT_RDONLY;
241 	if (error)
242 		mp->mnt_flag = saved_flags;
243 	mp->mnt_flag &= ~MNT_OP_FLAGS;
244 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
245 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
246 		if (mp->mnt_syncer == NULL)
247 			error = vfs_allocate_syncvnode(mp);
248 	} else {
249 		if (mp->mnt_syncer != NULL)
250 			vfs_deallocate_syncvnode(mp);
251 	}
252 	vfs_unbusy(mp, false);
253 
254  out:
255 	return (error);
256 }
257 
258 static int
259 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
260 {
261 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
262 	int error;
263 
264 	/* Copy file-system type from userspace.  */
265 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
266 	if (error) {
267 #if defined(COMPAT_09) || defined(COMPAT_43)
268 		/*
269 		 * Historically, filesystem types were identified by numbers.
270 		 * If we get an integer for the filesystem type instead of a
271 		 * string, we check to see if it matches one of the historic
272 		 * filesystem types.
273 		 */
274 		u_long fsindex = (u_long)fstype;
275 		if (fsindex >= nmountcompatnames ||
276 		    mountcompatnames[fsindex] == NULL)
277 			return ENODEV;
278 		strlcpy(fstypename, mountcompatnames[fsindex],
279 		    sizeof(fstypename));
280 #else
281 		return error;
282 #endif
283 	}
284 
285 #ifdef	COMPAT_10
286 	/* Accept `ufs' as an alias for `ffs'. */
287 	if (strcmp(fstypename, "ufs") == 0)
288 		fstypename[0] = 'f';
289 #endif
290 
291 	if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL)
292 		return ENODEV;
293 	return 0;
294 }
295 
296 static int
297 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
298     const char *path, int flags, void *data, size_t *data_len, u_int recurse)
299 {
300 	struct mount *mp = NULL;
301 	struct vnode *vp = *vpp;
302 	struct vattr va;
303 	int error;
304 
305 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
306 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
307 	if (error)
308 		return error;
309 
310 	/* Can't make a non-dir a mount-point (from here anyway). */
311 	if (vp->v_type != VDIR)
312 		return ENOTDIR;
313 
314 	/*
315 	 * If the user is not root, ensure that they own the directory
316 	 * onto which we are attempting to mount.
317 	 */
318 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
319 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
320 	    (error = kauth_authorize_generic(l->l_cred,
321 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
322 		return error;
323 	}
324 
325 	if (flags & MNT_EXPORTED)
326 		return EINVAL;
327 
328 	if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
329 		return error;
330 
331 	/*
332 	 * Check if a file-system is not already mounted on this vnode.
333 	 */
334 	if (vp->v_mountedhere != NULL)
335 		return EBUSY;
336 
337 	mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
338 	if (mp == NULL)
339 		return ENOMEM;
340 
341 	mp->mnt_op = vfsops;
342 	mp->mnt_refcnt = 1;
343 
344 	TAILQ_INIT(&mp->mnt_vnodelist);
345 	rw_init(&mp->mnt_lock);
346  	mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE);
347 	(void)vfs_busy(mp, RW_WRITER, 0);
348 
349 	mp->mnt_vnodecovered = vp;
350 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
351 	mount_initspecific(mp);
352 
353 	/*
354 	 * The underlying file system may refuse the mount for
355 	 * various reasons.  Allow the user to force it to happen.
356 	 *
357 	 * Set the mount level flags.
358 	 */
359 	mp->mnt_flag = flags &
360 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
361 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
362 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
363 	    MNT_IGNORE | MNT_RDONLY);
364 
365 	error = VFS_MOUNT(mp, path, data, data_len);
366 	mp->mnt_flag &= ~MNT_OP_FLAGS;
367 
368 	/*
369 	 * Put the new filesystem on the mount list after root.
370 	 */
371 	cache_purge(vp);
372 	if (error != 0) {
373 		vp->v_mountedhere = NULL;
374 		mp->mnt_op->vfs_refcount--;
375 		vfs_unbusy(mp, false);
376 		vfs_destroy(mp);
377 		return error;
378 	}
379 
380 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
381 	mutex_enter(&mountlist_lock);
382 	vp->v_mountedhere = mp;
383 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
384 	mutex_exit(&mountlist_lock);
385     	vn_restorerecurse(vp, recurse);
386 	VOP_UNLOCK(vp, 0);
387 	checkdirs(vp);
388 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
389 		error = vfs_allocate_syncvnode(mp);
390 	/* Hold an additional reference to the mount across VFS_START(). */
391 	vfs_unbusy(mp, true);
392 	(void) VFS_STATVFS(mp, &mp->mnt_stat);
393 	error = VFS_START(mp, 0);
394 	if (error) {
395 		vrele(vp);
396 		vfs_destroy(mp);
397 	}
398 	/* Drop reference held for VFS_START(). */
399 	vfs_destroy(mp);
400 	*vpp = NULL;
401 	return error;
402 }
403 
404 static int
405 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
406     void *data, size_t *data_len)
407 {
408 	struct mount *mp;
409 	int error;
410 
411 	/* If MNT_GETARGS is specified, it should be the only flag. */
412 	if (flags & ~MNT_GETARGS)
413 		return EINVAL;
414 
415 	mp = vp->v_mount;
416 
417 	/* XXX: probably some notion of "can see" here if we want isolation. */
418 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
419 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
420 	if (error)
421 		return error;
422 
423 	if ((vp->v_vflag & VV_ROOT) == 0)
424 		return EINVAL;
425 
426 	if (vfs_trybusy(mp, RW_WRITER, NULL))
427 		return EPERM;
428 
429 	mp->mnt_flag &= ~MNT_OP_FLAGS;
430 	mp->mnt_flag |= MNT_GETARGS;
431 	error = VFS_MOUNT(mp, path, data, data_len);
432 	mp->mnt_flag &= ~MNT_OP_FLAGS;
433 
434 	vfs_unbusy(mp, false);
435 	return (error);
436 }
437 
438 #ifdef COMPAT_40
439 /* ARGSUSED */
440 int
441 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval)
442 {
443 	/* {
444 		syscallarg(const char *) type;
445 		syscallarg(const char *) path;
446 		syscallarg(int) flags;
447 		syscallarg(void *) data;
448 	} */
449 	register_t dummy;
450 
451 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
452 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy);
453 }
454 #endif
455 
456 int
457 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
458 {
459 	/* {
460 		syscallarg(const char *) type;
461 		syscallarg(const char *) path;
462 		syscallarg(int) flags;
463 		syscallarg(void *) data;
464 		syscallarg(size_t) data_len;
465 	} */
466 
467 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
468 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
469 	    SCARG(uap, data_len), retval);
470 }
471 
472 int
473 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
474     const char *path, int flags, void *data, enum uio_seg data_seg,
475     size_t data_len, register_t *retval)
476 {
477 	struct vnode *vp;
478 	struct nameidata nd;
479 	void *data_buf = data;
480 	u_int recurse;
481 	int error;
482 
483 	/*
484 	 * Get vnode to be covered
485 	 */
486 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
487 	if ((error = namei(&nd)) != 0)
488 		return (error);
489 	vp = nd.ni_vp;
490 
491 	/*
492 	 * A lookup in VFS_MOUNT might result in an attempt to
493 	 * lock this vnode again, so make the lock recursive.
494 	 */
495 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
496 	recurse = vn_setrecurse(vp);
497 
498 	if (vfsops == NULL) {
499 		if (flags & (MNT_GETARGS | MNT_UPDATE))
500 			vfsops = vp->v_mount->mnt_op;
501 		else {
502 			/* 'type' is userspace */
503 			error = mount_get_vfsops(type, &vfsops);
504 			if (error != 0)
505 				goto done;
506 		}
507 	}
508 
509 	if (data != NULL && data_seg == UIO_USERSPACE) {
510 		if (data_len == 0) {
511 			/* No length supplied, use default for filesystem */
512 			data_len = vfsops->vfs_min_mount_data;
513 			if (data_len > VFS_MAX_MOUNT_DATA) {
514 				/* maybe a force loaded old LKM */
515 				error = EINVAL;
516 				goto done;
517 			}
518 #ifdef COMPAT_30
519 			/* Hopefully a longer buffer won't make copyin() fail */
520 			if (flags & MNT_UPDATE
521 			    && data_len < sizeof (struct mnt_export_args30))
522 				data_len = sizeof (struct mnt_export_args30);
523 #endif
524 		}
525 		data_buf = malloc(data_len, M_TEMP, M_WAITOK);
526 
527 		/* NFS needs the buffer even for mnt_getargs .... */
528 		error = copyin(data, data_buf, data_len);
529 		if (error != 0)
530 			goto done;
531 	}
532 
533 	if (flags & MNT_GETARGS) {
534 		if (data_len == 0) {
535 			error = EINVAL;
536 			goto done;
537 		}
538 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
539 		if (error != 0)
540 			goto done;
541 		if (data_seg == UIO_USERSPACE)
542 			error = copyout(data_buf, data, data_len);
543 		*retval = data_len;
544 	} else if (flags & MNT_UPDATE) {
545 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
546 	} else {
547 		/* Locking is handled internally in mount_domount(). */
548 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
549 		    &data_len, recurse);
550 	}
551 
552     done:
553     	if (vp != NULL) {
554 	    	vn_restorerecurse(vp, recurse);
555 	    	vput(vp);
556 	}
557 	if (data_buf != data)
558 		free(data_buf, M_TEMP);
559 	return (error);
560 }
561 
562 /*
563  * Scan all active processes to see if any of them have a current
564  * or root directory onto which the new filesystem has just been
565  * mounted. If so, replace them with the new mount point.
566  */
567 void
568 checkdirs(struct vnode *olddp)
569 {
570 	struct cwdinfo *cwdi;
571 	struct vnode *newdp;
572 	struct proc *p;
573 
574 	if (olddp->v_usecount == 1)
575 		return;
576 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
577 		panic("mount: lost mount");
578 	mutex_enter(proc_lock);
579 	/* XXXAD Should not be acquiring these locks with proc_lock held!! */
580 	PROCLIST_FOREACH(p, &allproc) {
581 		cwdi = p->p_cwdi;
582 		if (!cwdi)
583 			continue;
584 		rw_enter(&cwdi->cwdi_lock, RW_WRITER);
585 		if (cwdi->cwdi_cdir == olddp) {
586 			vrele(cwdi->cwdi_cdir);
587 			VREF(newdp);
588 			cwdi->cwdi_cdir = newdp;
589 		}
590 		if (cwdi->cwdi_rdir == olddp) {
591 			vrele(cwdi->cwdi_rdir);
592 			VREF(newdp);
593 			cwdi->cwdi_rdir = newdp;
594 		}
595 		rw_exit(&cwdi->cwdi_lock);
596 	}
597 	mutex_exit(proc_lock);
598 	if (rootvnode == olddp) {
599 		vrele(rootvnode);
600 		VREF(newdp);
601 		rootvnode = newdp;
602 	}
603 	vput(newdp);
604 }
605 
606 /*
607  * Unmount a file system.
608  *
609  * Note: unmount takes a path to the vnode mounted on as argument,
610  * not special file (as before).
611  */
612 /* ARGSUSED */
613 int
614 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
615 {
616 	/* {
617 		syscallarg(const char *) path;
618 		syscallarg(int) flags;
619 	} */
620 	struct vnode *vp;
621 	struct mount *mp;
622 	int error;
623 	struct nameidata nd;
624 
625 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
626 	    SCARG(uap, path));
627 	if ((error = namei(&nd)) != 0)
628 		return (error);
629 	vp = nd.ni_vp;
630 	mp = vp->v_mount;
631 	VOP_UNLOCK(vp, 0);
632 
633 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
634 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
635 	if (error) {
636 		vrele(vp);
637 		return (error);
638 	}
639 
640 	/*
641 	 * Don't allow unmounting the root file system.
642 	 */
643 	if (mp->mnt_flag & MNT_ROOTFS) {
644 		vrele(vp);
645 		return (EINVAL);
646 	}
647 
648 	/*
649 	 * Must be the root of the filesystem
650 	 */
651 	if ((vp->v_vflag & VV_ROOT) == 0) {
652 		vrele(vp);
653 		return (EINVAL);
654 	}
655 
656 	/*
657 	 * XXX Freeze syncer.  Must do this before locking the
658 	 * mount point.  See dounmount() for details.
659 	 */
660 	mutex_enter(&syncer_mutex);
661 	error = vfs_busy(mp, RW_WRITER, NULL);
662 	vrele(vp);
663 	if (error != 0) {
664 		mutex_exit(&syncer_mutex);
665 		return (error);
666 	}
667 
668 	return (dounmount(mp, SCARG(uap, flags), l));
669 }
670 
671 /*
672  * Do the actual file system unmount. File system is assumed to have been
673  * marked busy by the caller.
674  */
675 int
676 dounmount(struct mount *mp, int flags, struct lwp *l)
677 {
678 	struct vnode *coveredvp;
679 	int error;
680 	int async;
681 	int used_syncer;
682 
683 	KASSERT(rw_write_held(&mp->mnt_lock));
684 
685 #if NVERIEXEC > 0
686 	error = veriexec_unmountchk(mp);
687 	if (error)
688 		return (error);
689 #endif /* NVERIEXEC > 0 */
690 
691 	used_syncer = (mp->mnt_syncer != NULL);
692 
693 	/*
694 	 * XXX Syncer must be frozen when we get here.  This should really
695 	 * be done on a per-mountpoint basis, but especially the softdep
696 	 * code possibly called from the syncer doesn't exactly work on a
697 	 * per-mountpoint basis, so the softdep code would become a maze
698 	 * of vfs_busy() calls.
699 	 *
700 	 * The caller of dounmount() must acquire syncer_mutex because
701 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
702 	 * order, and we must preserve that order to avoid deadlock.
703 	 *
704 	 * So, if the file system did not use the syncer, now is
705 	 * the time to release the syncer_mutex.
706 	 */
707 	if (used_syncer == 0)
708 		mutex_exit(&syncer_mutex);
709 
710 	mp->mnt_iflag |= IMNT_UNMOUNT;
711 	async = mp->mnt_flag & MNT_ASYNC;
712 	mp->mnt_flag &= ~MNT_ASYNC;
713 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
714 	if (mp->mnt_syncer != NULL)
715 		vfs_deallocate_syncvnode(mp);
716 	error = 0;
717 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
718 #if NFSS > 0
719 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
720 #endif
721 		if (error == 0)
722 			error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
723 	}
724 	vfs_scrubvnlist(mp);
725 	if (error == 0 || (flags & MNT_FORCE))
726 		error = VFS_UNMOUNT(mp, flags);
727 	if (error) {
728 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
729 			(void) vfs_allocate_syncvnode(mp);
730 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
731 		mp->mnt_flag |= async;
732 		if (used_syncer)
733 			mutex_exit(&syncer_mutex);
734 		vfs_unbusy(mp, false);
735 		return (error);
736 	}
737 	vfs_scrubvnlist(mp);
738 	mutex_enter(&mountlist_lock);
739 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
740 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
741 		coveredvp->v_mountedhere = NULL;
742 	mutex_exit(&mountlist_lock);
743 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
744 		panic("unmount: dangling vnode");
745 	mp->mnt_iflag |= IMNT_GONE;
746 	if (used_syncer)
747 		mutex_exit(&syncer_mutex);
748 	vfs_hooks_unmount(mp);
749 	vfs_unbusy(mp, false);
750 	vfs_destroy(mp);
751 	if (coveredvp != NULLVP)
752 		vrele(coveredvp);
753 	return (0);
754 }
755 
756 /*
757  * Sync each mounted filesystem.
758  */
759 #ifdef DEBUG
760 int syncprt = 0;
761 struct ctldebug debug0 = { "syncprt", &syncprt };
762 #endif
763 
764 /* ARGSUSED */
765 int
766 sys_sync(struct lwp *l, const void *v, register_t *retval)
767 {
768 	struct mount *mp, *nmp;
769 	int asyncflag;
770 
771 	if (l == NULL)
772 		l = &lwp0;
773 
774 	mutex_enter(&mountlist_lock);
775 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
776 		if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) {
777 			nmp = mp->mnt_list.cqe_prev;
778 			continue;
779 		}
780 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
781 			asyncflag = mp->mnt_flag & MNT_ASYNC;
782 			mp->mnt_flag &= ~MNT_ASYNC;
783 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
784 			if (asyncflag)
785 				 mp->mnt_flag |= MNT_ASYNC;
786 		}
787 		mutex_enter(&mountlist_lock);
788 		nmp = mp->mnt_list.cqe_prev;
789 		vfs_unbusy(mp, false);
790 
791 	}
792 	mutex_exit(&mountlist_lock);
793 #ifdef DEBUG
794 	if (syncprt)
795 		vfs_bufstats();
796 #endif /* DEBUG */
797 	return (0);
798 }
799 
800 /*
801  * Change filesystem quotas.
802  */
803 /* ARGSUSED */
804 int
805 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
806 {
807 	/* {
808 		syscallarg(const char *) path;
809 		syscallarg(int) cmd;
810 		syscallarg(int) uid;
811 		syscallarg(void *) arg;
812 	} */
813 	struct mount *mp;
814 	int error;
815 	struct nameidata nd;
816 
817 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
818 	    SCARG(uap, path));
819 	if ((error = namei(&nd)) != 0)
820 		return (error);
821 	mp = nd.ni_vp->v_mount;
822 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
823 	    SCARG(uap, arg));
824 	vrele(nd.ni_vp);
825 	return (error);
826 }
827 
828 int
829 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
830     int root)
831 {
832 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
833 	int error = 0;
834 
835 	/*
836 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
837 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
838 	 * overrides MNT_NOWAIT.
839 	 */
840 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
841 	    (flags != MNT_WAIT && flags != 0)) {
842 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
843 		goto done;
844 	}
845 
846 	/* Get the filesystem stats now */
847 	memset(sp, 0, sizeof(*sp));
848 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
849 		return error;
850 	}
851 
852 	if (cwdi->cwdi_rdir == NULL)
853 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
854 done:
855 	if (cwdi->cwdi_rdir != NULL) {
856 		size_t len;
857 		char *bp;
858 		char *path = PNBUF_GET();
859 
860 		bp = path + MAXPATHLEN;
861 		*--bp = '\0';
862 		rw_enter(&cwdi->cwdi_lock, RW_READER);
863 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
864 		    MAXPATHLEN / 2, 0, l);
865 		rw_exit(&cwdi->cwdi_lock);
866 		if (error) {
867 			PNBUF_PUT(path);
868 			return error;
869 		}
870 		len = strlen(bp);
871 		/*
872 		 * for mount points that are below our root, we can see
873 		 * them, so we fix up the pathname and return them. The
874 		 * rest we cannot see, so we don't allow viewing the
875 		 * data.
876 		 */
877 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
878 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
879 			    sizeof(sp->f_mntonname));
880 			if (sp->f_mntonname[0] == '\0')
881 				(void)strlcpy(sp->f_mntonname, "/",
882 				    sizeof(sp->f_mntonname));
883 		} else {
884 			if (root)
885 				(void)strlcpy(sp->f_mntonname, "/",
886 				    sizeof(sp->f_mntonname));
887 			else
888 				error = EPERM;
889 		}
890 		PNBUF_PUT(path);
891 	}
892 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
893 	return error;
894 }
895 
896 /*
897  * Get filesystem statistics by path.
898  */
899 int
900 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
901 {
902 	struct mount *mp;
903 	int error;
904 	struct nameidata nd;
905 
906 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
907 	if ((error = namei(&nd)) != 0)
908 		return error;
909 	mp = nd.ni_vp->v_mount;
910 	error = dostatvfs(mp, sb, l, flags, 1);
911 	vrele(nd.ni_vp);
912 	return error;
913 }
914 
915 /* ARGSUSED */
916 int
917 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
918 {
919 	/* {
920 		syscallarg(const char *) path;
921 		syscallarg(struct statvfs *) buf;
922 		syscallarg(int) flags;
923 	} */
924 	struct statvfs *sb;
925 	int error;
926 
927 	sb = STATVFSBUF_GET();
928 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
929 	if (error == 0)
930 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
931 	STATVFSBUF_PUT(sb);
932 	return error;
933 }
934 
935 /*
936  * Get filesystem statistics by fd.
937  */
938 int
939 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
940 {
941 	file_t *fp;
942 	struct mount *mp;
943 	int error;
944 
945 	/* fd_getvnode() will use the descriptor for us */
946 	if ((error = fd_getvnode(fd, &fp)) != 0)
947 		return (error);
948 	mp = ((struct vnode *)fp->f_data)->v_mount;
949 	error = dostatvfs(mp, sb, curlwp, flags, 1);
950 	fd_putfile(fd);
951 	return error;
952 }
953 
954 /* ARGSUSED */
955 int
956 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
957 {
958 	/* {
959 		syscallarg(int) fd;
960 		syscallarg(struct statvfs *) buf;
961 		syscallarg(int) flags;
962 	} */
963 	struct statvfs *sb;
964 	int error;
965 
966 	sb = STATVFSBUF_GET();
967 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
968 	if (error == 0)
969 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
970 	STATVFSBUF_PUT(sb);
971 	return error;
972 }
973 
974 
975 /*
976  * Get statistics on all filesystems.
977  */
978 int
979 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
980     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
981     register_t *retval)
982 {
983 	int root = 0;
984 	struct proc *p = l->l_proc;
985 	struct mount *mp, *nmp;
986 	struct statvfs *sb;
987 	size_t count, maxcount;
988 	int error = 0;
989 
990 	sb = STATVFSBUF_GET();
991 	maxcount = bufsize / entry_sz;
992 	mutex_enter(&mountlist_lock);
993 	count = 0;
994 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
995 	     mp = nmp) {
996 		if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) {
997 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
998 			continue;
999 		}
1000 		if (sfsp && count < maxcount) {
1001 			error = dostatvfs(mp, sb, l, flags, 0);
1002 			if (error) {
1003 				mutex_enter(&mountlist_lock);
1004 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
1005 				vfs_unbusy(mp, false);
1006 				continue;
1007 			}
1008 			error = copyfn(sb, sfsp, entry_sz);
1009 			if (error) {
1010 				vfs_unbusy(mp, false);
1011 				goto out;
1012 			}
1013 			sfsp = (char *)sfsp + entry_sz;
1014 			root |= strcmp(sb->f_mntonname, "/") == 0;
1015 		}
1016 		count++;
1017 		mutex_enter(&mountlist_lock);
1018 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
1019 		vfs_unbusy(mp, false);
1020 	}
1021 
1022 	mutex_exit(&mountlist_lock);
1023 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
1024 		/*
1025 		 * fake a root entry
1026 		 */
1027 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1028 		    sb, l, flags, 1);
1029 		if (error != 0)
1030 			goto out;
1031 		if (sfsp)
1032 			error = copyfn(sb, sfsp, entry_sz);
1033 		count++;
1034 	}
1035 	if (sfsp && count > maxcount)
1036 		*retval = maxcount;
1037 	else
1038 		*retval = count;
1039 out:
1040 	STATVFSBUF_PUT(sb);
1041 	return error;
1042 }
1043 
1044 int
1045 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1046 {
1047 	/* {
1048 		syscallarg(struct statvfs *) buf;
1049 		syscallarg(size_t) bufsize;
1050 		syscallarg(int) flags;
1051 	} */
1052 
1053 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1054 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1055 }
1056 
1057 /*
1058  * Change current working directory to a given file descriptor.
1059  */
1060 /* ARGSUSED */
1061 int
1062 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1063 {
1064 	/* {
1065 		syscallarg(int) fd;
1066 	} */
1067 	struct proc *p = l->l_proc;
1068 	struct cwdinfo *cwdi;
1069 	struct vnode *vp, *tdp;
1070 	struct mount *mp;
1071 	file_t *fp;
1072 	int error, fd;
1073 
1074 	/* fd_getvnode() will use the descriptor for us */
1075 	fd = SCARG(uap, fd);
1076 	if ((error = fd_getvnode(fd, &fp)) != 0)
1077 		return (error);
1078 	vp = fp->f_data;
1079 
1080 	VREF(vp);
1081 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1082 	if (vp->v_type != VDIR)
1083 		error = ENOTDIR;
1084 	else
1085 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1086 	if (error) {
1087 		vput(vp);
1088 		goto out;
1089 	}
1090 	while ((mp = vp->v_mountedhere) != NULL) {
1091 		if (vfs_busy(mp, RW_READER, NULL))
1092 			continue;
1093 		vput(vp);
1094 		error = VFS_ROOT(mp, &tdp);
1095 		vfs_unbusy(mp, false);
1096 		if (error)
1097 			goto out;
1098 		vp = tdp;
1099 	}
1100 	VOP_UNLOCK(vp, 0);
1101 
1102 	/*
1103 	 * Disallow changing to a directory not under the process's
1104 	 * current root directory (if there is one).
1105 	 */
1106 	cwdi = p->p_cwdi;
1107 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1108 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1109 		vrele(vp);
1110 		error = EPERM;	/* operation not permitted */
1111 	} else {
1112 		vrele(cwdi->cwdi_cdir);
1113 		cwdi->cwdi_cdir = vp;
1114 	}
1115 	rw_exit(&cwdi->cwdi_lock);
1116 
1117  out:
1118 	fd_putfile(fd);
1119 	return (error);
1120 }
1121 
1122 /*
1123  * Change this process's notion of the root directory to a given file
1124  * descriptor.
1125  */
1126 int
1127 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1128 {
1129 	struct proc *p = l->l_proc;
1130 	struct cwdinfo *cwdi;
1131 	struct vnode	*vp;
1132 	file_t	*fp;
1133 	int		 error, fd = SCARG(uap, fd);
1134 
1135 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1136  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1137 		return error;
1138 	/* fd_getvnode() will use the descriptor for us */
1139 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1140 		return error;
1141 	vp = fp->f_data;
1142 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1143 	if (vp->v_type != VDIR)
1144 		error = ENOTDIR;
1145 	else
1146 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1147 	VOP_UNLOCK(vp, 0);
1148 	if (error)
1149 		goto out;
1150 	VREF(vp);
1151 
1152 	/*
1153 	 * Prevent escaping from chroot by putting the root under
1154 	 * the working directory.  Silently chdir to / if we aren't
1155 	 * already there.
1156 	 */
1157 	cwdi = p->p_cwdi;
1158 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1159 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1160 		/*
1161 		 * XXX would be more failsafe to change directory to a
1162 		 * deadfs node here instead
1163 		 */
1164 		vrele(cwdi->cwdi_cdir);
1165 		VREF(vp);
1166 		cwdi->cwdi_cdir = vp;
1167 	}
1168 
1169 	if (cwdi->cwdi_rdir != NULL)
1170 		vrele(cwdi->cwdi_rdir);
1171 	cwdi->cwdi_rdir = vp;
1172 	rw_exit(&cwdi->cwdi_lock);
1173 
1174  out:
1175 	fd_putfile(fd);
1176 	return (error);
1177 }
1178 
1179 /*
1180  * Change current working directory (``.'').
1181  */
1182 /* ARGSUSED */
1183 int
1184 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1185 {
1186 	/* {
1187 		syscallarg(const char *) path;
1188 	} */
1189 	struct proc *p = l->l_proc;
1190 	struct cwdinfo *cwdi;
1191 	int error;
1192 	struct nameidata nd;
1193 
1194 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1195 	    SCARG(uap, path));
1196 	if ((error = change_dir(&nd, l)) != 0)
1197 		return (error);
1198 	cwdi = p->p_cwdi;
1199 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1200 	vrele(cwdi->cwdi_cdir);
1201 	cwdi->cwdi_cdir = nd.ni_vp;
1202 	rw_exit(&cwdi->cwdi_lock);
1203 	return (0);
1204 }
1205 
1206 /*
1207  * Change notion of root (``/'') directory.
1208  */
1209 /* ARGSUSED */
1210 int
1211 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1212 {
1213 	/* {
1214 		syscallarg(const char *) path;
1215 	} */
1216 	struct proc *p = l->l_proc;
1217 	struct cwdinfo *cwdi;
1218 	struct vnode *vp;
1219 	int error;
1220 	struct nameidata nd;
1221 
1222 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1223 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1224 		return (error);
1225 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1226 	    SCARG(uap, path));
1227 	if ((error = change_dir(&nd, l)) != 0)
1228 		return (error);
1229 
1230 	cwdi = p->p_cwdi;
1231 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1232 	if (cwdi->cwdi_rdir != NULL)
1233 		vrele(cwdi->cwdi_rdir);
1234 	vp = nd.ni_vp;
1235 	cwdi->cwdi_rdir = vp;
1236 
1237 	/*
1238 	 * Prevent escaping from chroot by putting the root under
1239 	 * the working directory.  Silently chdir to / if we aren't
1240 	 * already there.
1241 	 */
1242 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1243 		/*
1244 		 * XXX would be more failsafe to change directory to a
1245 		 * deadfs node here instead
1246 		 */
1247 		vrele(cwdi->cwdi_cdir);
1248 		VREF(vp);
1249 		cwdi->cwdi_cdir = vp;
1250 	}
1251 	rw_exit(&cwdi->cwdi_lock);
1252 
1253 	return (0);
1254 }
1255 
1256 /*
1257  * Common routine for chroot and chdir.
1258  */
1259 static int
1260 change_dir(struct nameidata *ndp, struct lwp *l)
1261 {
1262 	struct vnode *vp;
1263 	int error;
1264 
1265 	if ((error = namei(ndp)) != 0)
1266 		return (error);
1267 	vp = ndp->ni_vp;
1268 	if (vp->v_type != VDIR)
1269 		error = ENOTDIR;
1270 	else
1271 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1272 
1273 	if (error)
1274 		vput(vp);
1275 	else
1276 		VOP_UNLOCK(vp, 0);
1277 	return (error);
1278 }
1279 
1280 /*
1281  * Check permissions, allocate an open file structure,
1282  * and call the device open routine if any.
1283  */
1284 int
1285 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1286 {
1287 	/* {
1288 		syscallarg(const char *) path;
1289 		syscallarg(int) flags;
1290 		syscallarg(int) mode;
1291 	} */
1292 	struct proc *p = l->l_proc;
1293 	struct cwdinfo *cwdi = p->p_cwdi;
1294 	file_t *fp;
1295 	struct vnode *vp;
1296 	int flags, cmode;
1297 	int type, indx, error;
1298 	struct flock lf;
1299 	struct nameidata nd;
1300 
1301 	flags = FFLAGS(SCARG(uap, flags));
1302 	if ((flags & (FREAD | FWRITE)) == 0)
1303 		return (EINVAL);
1304 	if ((error = fd_allocfile(&fp, &indx)) != 0)
1305 		return (error);
1306 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1307 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1308 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1309 	    SCARG(uap, path));
1310 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1311 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1312 		fd_abort(p, fp, indx);
1313 		if ((error == EDUPFD || error == EMOVEFD) &&
1314 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1315 		    (error =
1316 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1317 			*retval = indx;
1318 			return (0);
1319 		}
1320 		if (error == ERESTART)
1321 			error = EINTR;
1322 		return (error);
1323 	}
1324 
1325 	l->l_dupfd = 0;
1326 	vp = nd.ni_vp;
1327 	fp->f_flag = flags & FMASK;
1328 	fp->f_type = DTYPE_VNODE;
1329 	fp->f_ops = &vnops;
1330 	fp->f_data = vp;
1331 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1332 		lf.l_whence = SEEK_SET;
1333 		lf.l_start = 0;
1334 		lf.l_len = 0;
1335 		if (flags & O_EXLOCK)
1336 			lf.l_type = F_WRLCK;
1337 		else
1338 			lf.l_type = F_RDLCK;
1339 		type = F_FLOCK;
1340 		if ((flags & FNONBLOCK) == 0)
1341 			type |= F_WAIT;
1342 		VOP_UNLOCK(vp, 0);
1343 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1344 		if (error) {
1345 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1346 			fd_abort(p, fp, indx);
1347 			return (error);
1348 		}
1349 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1350 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1351 	}
1352 	VOP_UNLOCK(vp, 0);
1353 	*retval = indx;
1354 	fd_affix(p, fp, indx);
1355 	return (0);
1356 }
1357 
1358 static void
1359 vfs__fhfree(fhandle_t *fhp)
1360 {
1361 	size_t fhsize;
1362 
1363 	if (fhp == NULL) {
1364 		return;
1365 	}
1366 	fhsize = FHANDLE_SIZE(fhp);
1367 	kmem_free(fhp, fhsize);
1368 }
1369 
1370 /*
1371  * vfs_composefh: compose a filehandle.
1372  */
1373 
1374 int
1375 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1376 {
1377 	struct mount *mp;
1378 	struct fid *fidp;
1379 	int error;
1380 	size_t needfhsize;
1381 	size_t fidsize;
1382 
1383 	mp = vp->v_mount;
1384 	fidp = NULL;
1385 	if (*fh_size < FHANDLE_SIZE_MIN) {
1386 		fidsize = 0;
1387 	} else {
1388 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1389 		if (fhp != NULL) {
1390 			memset(fhp, 0, *fh_size);
1391 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1392 			fidp = &fhp->fh_fid;
1393 		}
1394 	}
1395 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1396 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1397 	if (error == 0 && *fh_size < needfhsize) {
1398 		error = E2BIG;
1399 	}
1400 	*fh_size = needfhsize;
1401 	return error;
1402 }
1403 
1404 int
1405 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1406 {
1407 	struct mount *mp;
1408 	fhandle_t *fhp;
1409 	size_t fhsize;
1410 	size_t fidsize;
1411 	int error;
1412 
1413 	*fhpp = NULL;
1414 	mp = vp->v_mount;
1415 	fidsize = 0;
1416 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1417 	KASSERT(error != 0);
1418 	if (error != E2BIG) {
1419 		goto out;
1420 	}
1421 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1422 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1423 	if (fhp == NULL) {
1424 		error = ENOMEM;
1425 		goto out;
1426 	}
1427 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1428 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1429 	if (error == 0) {
1430 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1431 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1432 		*fhpp = fhp;
1433 	} else {
1434 		kmem_free(fhp, fhsize);
1435 	}
1436 out:
1437 	return error;
1438 }
1439 
1440 void
1441 vfs_composefh_free(fhandle_t *fhp)
1442 {
1443 
1444 	vfs__fhfree(fhp);
1445 }
1446 
1447 /*
1448  * vfs_fhtovp: lookup a vnode by a filehandle.
1449  */
1450 
1451 int
1452 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1453 {
1454 	struct mount *mp;
1455 	int error;
1456 
1457 	*vpp = NULL;
1458 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1459 	if (mp == NULL) {
1460 		error = ESTALE;
1461 		goto out;
1462 	}
1463 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1464 		error = EOPNOTSUPP;
1465 		goto out;
1466 	}
1467 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1468 out:
1469 	return error;
1470 }
1471 
1472 /*
1473  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1474  * the needed size.
1475  */
1476 
1477 int
1478 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1479 {
1480 	fhandle_t *fhp;
1481 	int error;
1482 
1483 	*fhpp = NULL;
1484 	if (fhsize > FHANDLE_SIZE_MAX) {
1485 		return EINVAL;
1486 	}
1487 	if (fhsize < FHANDLE_SIZE_MIN) {
1488 		return EINVAL;
1489 	}
1490 again:
1491 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1492 	if (fhp == NULL) {
1493 		return ENOMEM;
1494 	}
1495 	error = copyin(ufhp, fhp, fhsize);
1496 	if (error == 0) {
1497 		/* XXX this check shouldn't be here */
1498 		if (FHANDLE_SIZE(fhp) == fhsize) {
1499 			*fhpp = fhp;
1500 			return 0;
1501 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1502 			/*
1503 			 * a kludge for nfsv2 padded handles.
1504 			 */
1505 			size_t sz;
1506 
1507 			sz = FHANDLE_SIZE(fhp);
1508 			kmem_free(fhp, fhsize);
1509 			fhsize = sz;
1510 			goto again;
1511 		} else {
1512 			/*
1513 			 * userland told us wrong size.
1514 			 */
1515 		    	error = EINVAL;
1516 		}
1517 	}
1518 	kmem_free(fhp, fhsize);
1519 	return error;
1520 }
1521 
1522 void
1523 vfs_copyinfh_free(fhandle_t *fhp)
1524 {
1525 
1526 	vfs__fhfree(fhp);
1527 }
1528 
1529 /*
1530  * Get file handle system call
1531  */
1532 int
1533 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1534 {
1535 	/* {
1536 		syscallarg(char *) fname;
1537 		syscallarg(fhandle_t *) fhp;
1538 		syscallarg(size_t *) fh_size;
1539 	} */
1540 	struct vnode *vp;
1541 	fhandle_t *fh;
1542 	int error;
1543 	struct nameidata nd;
1544 	size_t sz;
1545 	size_t usz;
1546 
1547 	/*
1548 	 * Must be super user
1549 	 */
1550 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1551 	    0, NULL, NULL, NULL);
1552 	if (error)
1553 		return (error);
1554 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1555 	    SCARG(uap, fname));
1556 	error = namei(&nd);
1557 	if (error)
1558 		return (error);
1559 	vp = nd.ni_vp;
1560 	error = vfs_composefh_alloc(vp, &fh);
1561 	vput(vp);
1562 	if (error != 0) {
1563 		goto out;
1564 	}
1565 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1566 	if (error != 0) {
1567 		goto out;
1568 	}
1569 	sz = FHANDLE_SIZE(fh);
1570 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1571 	if (error != 0) {
1572 		goto out;
1573 	}
1574 	if (usz >= sz) {
1575 		error = copyout(fh, SCARG(uap, fhp), sz);
1576 	} else {
1577 		error = E2BIG;
1578 	}
1579 out:
1580 	vfs_composefh_free(fh);
1581 	return (error);
1582 }
1583 
1584 /*
1585  * Open a file given a file handle.
1586  *
1587  * Check permissions, allocate an open file structure,
1588  * and call the device open routine if any.
1589  */
1590 
1591 int
1592 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1593     register_t *retval)
1594 {
1595 	file_t *fp;
1596 	struct vnode *vp = NULL;
1597 	kauth_cred_t cred = l->l_cred;
1598 	file_t *nfp;
1599 	int type, indx, error=0;
1600 	struct flock lf;
1601 	struct vattr va;
1602 	fhandle_t *fh;
1603 	int flags;
1604 	proc_t *p;
1605 
1606 	p = curproc;
1607 
1608 	/*
1609 	 * Must be super user
1610 	 */
1611 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1612 	    0, NULL, NULL, NULL)))
1613 		return (error);
1614 
1615 	flags = FFLAGS(oflags);
1616 	if ((flags & (FREAD | FWRITE)) == 0)
1617 		return (EINVAL);
1618 	if ((flags & O_CREAT))
1619 		return (EINVAL);
1620 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1621 		return (error);
1622 	fp = nfp;
1623 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1624 	if (error != 0) {
1625 		goto bad;
1626 	}
1627 	error = vfs_fhtovp(fh, &vp);
1628 	if (error != 0) {
1629 		goto bad;
1630 	}
1631 
1632 	/* Now do an effective vn_open */
1633 
1634 	if (vp->v_type == VSOCK) {
1635 		error = EOPNOTSUPP;
1636 		goto bad;
1637 	}
1638 	error = vn_openchk(vp, cred, flags);
1639 	if (error != 0)
1640 		goto bad;
1641 	if (flags & O_TRUNC) {
1642 		VOP_UNLOCK(vp, 0);			/* XXX */
1643 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1644 		VATTR_NULL(&va);
1645 		va.va_size = 0;
1646 		error = VOP_SETATTR(vp, &va, cred);
1647 		if (error)
1648 			goto bad;
1649 	}
1650 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1651 		goto bad;
1652 	if (flags & FWRITE) {
1653 		mutex_enter(&vp->v_interlock);
1654 		vp->v_writecount++;
1655 		mutex_exit(&vp->v_interlock);
1656 	}
1657 
1658 	/* done with modified vn_open, now finish what sys_open does. */
1659 
1660 	fp->f_flag = flags & FMASK;
1661 	fp->f_type = DTYPE_VNODE;
1662 	fp->f_ops = &vnops;
1663 	fp->f_data = vp;
1664 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1665 		lf.l_whence = SEEK_SET;
1666 		lf.l_start = 0;
1667 		lf.l_len = 0;
1668 		if (flags & O_EXLOCK)
1669 			lf.l_type = F_WRLCK;
1670 		else
1671 			lf.l_type = F_RDLCK;
1672 		type = F_FLOCK;
1673 		if ((flags & FNONBLOCK) == 0)
1674 			type |= F_WAIT;
1675 		VOP_UNLOCK(vp, 0);
1676 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1677 		if (error) {
1678 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1679 			fd_abort(p, fp, indx);
1680 			return (error);
1681 		}
1682 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1683 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1684 	}
1685 	VOP_UNLOCK(vp, 0);
1686 	*retval = indx;
1687 	fd_affix(p, fp, indx);
1688 	vfs_copyinfh_free(fh);
1689 	return (0);
1690 
1691 bad:
1692 	fd_abort(p, fp, indx);
1693 	if (vp != NULL)
1694 		vput(vp);
1695 	vfs_copyinfh_free(fh);
1696 	return (error);
1697 }
1698 
1699 int
1700 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1701 {
1702 	/* {
1703 		syscallarg(const void *) fhp;
1704 		syscallarg(size_t) fh_size;
1705 		syscallarg(int) flags;
1706 	} */
1707 
1708 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1709 	    SCARG(uap, flags), retval);
1710 }
1711 
1712 int
1713 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1714 {
1715 	int error;
1716 	fhandle_t *fh;
1717 	struct vnode *vp;
1718 
1719 	/*
1720 	 * Must be super user
1721 	 */
1722 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1723 	    0, NULL, NULL, NULL)))
1724 		return (error);
1725 
1726 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1727 	if (error != 0)
1728 		return error;
1729 
1730 	error = vfs_fhtovp(fh, &vp);
1731 	vfs_copyinfh_free(fh);
1732 	if (error != 0)
1733 		return error;
1734 
1735 	error = vn_stat(vp, sb);
1736 	vput(vp);
1737 	return error;
1738 }
1739 
1740 
1741 /* ARGSUSED */
1742 int
1743 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval)
1744 {
1745 	/* {
1746 		syscallarg(const void *) fhp;
1747 		syscallarg(size_t) fh_size;
1748 		syscallarg(struct stat *) sb;
1749 	} */
1750 	struct stat sb;
1751 	int error;
1752 
1753 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1754 	if (error)
1755 		return error;
1756 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1757 }
1758 
1759 int
1760 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1761     int flags)
1762 {
1763 	fhandle_t *fh;
1764 	struct mount *mp;
1765 	struct vnode *vp;
1766 	int error;
1767 
1768 	/*
1769 	 * Must be super user
1770 	 */
1771 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1772 	    0, NULL, NULL, NULL)))
1773 		return error;
1774 
1775 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1776 	if (error != 0)
1777 		return error;
1778 
1779 	error = vfs_fhtovp(fh, &vp);
1780 	vfs_copyinfh_free(fh);
1781 	if (error != 0)
1782 		return error;
1783 
1784 	mp = vp->v_mount;
1785 	error = dostatvfs(mp, sb, l, flags, 1);
1786 	vput(vp);
1787 	return error;
1788 }
1789 
1790 /* ARGSUSED */
1791 int
1792 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1793 {
1794 	/* {
1795 		syscallarg(const void *) fhp;
1796 		syscallarg(size_t) fh_size;
1797 		syscallarg(struct statvfs *) buf;
1798 		syscallarg(int)	flags;
1799 	} */
1800 	struct statvfs *sb = STATVFSBUF_GET();
1801 	int error;
1802 
1803 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1804 	    SCARG(uap, flags));
1805 	if (error == 0)
1806 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1807 	STATVFSBUF_PUT(sb);
1808 	return error;
1809 }
1810 
1811 /*
1812  * Create a special file.
1813  */
1814 /* ARGSUSED */
1815 int
1816 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval)
1817 {
1818 	/* {
1819 		syscallarg(const char *) path;
1820 		syscallarg(int) mode;
1821 		syscallarg(int) dev;
1822 	} */
1823 	struct proc *p = l->l_proc;
1824 	struct vnode *vp;
1825 	struct vattr vattr;
1826 	int error, optype;
1827 	struct nameidata nd;
1828 	char *path;
1829 	const char *cpath;
1830 	enum uio_seg seg = UIO_USERSPACE;
1831 
1832 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1833 	    0, NULL, NULL, NULL)) != 0)
1834 		return (error);
1835 
1836 	optype = VOP_MKNOD_DESCOFFSET;
1837 
1838 	VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path);
1839 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
1840 
1841 	if ((error = namei(&nd)) != 0)
1842 		goto out;
1843 	vp = nd.ni_vp;
1844 	if (vp != NULL)
1845 		error = EEXIST;
1846 	else {
1847 		VATTR_NULL(&vattr);
1848 		/* We will read cwdi->cwdi_cmask unlocked. */
1849 		vattr.va_mode =
1850 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1851 		vattr.va_rdev = SCARG(uap, dev);
1852 
1853 		switch (SCARG(uap, mode) & S_IFMT) {
1854 		case S_IFMT:	/* used by badsect to flag bad sectors */
1855 			vattr.va_type = VBAD;
1856 			break;
1857 		case S_IFCHR:
1858 			vattr.va_type = VCHR;
1859 			break;
1860 		case S_IFBLK:
1861 			vattr.va_type = VBLK;
1862 			break;
1863 		case S_IFWHT:
1864 			optype = VOP_WHITEOUT_DESCOFFSET;
1865 			break;
1866 		case S_IFREG:
1867 #if NVERIEXEC > 0
1868 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1869 			    O_CREAT);
1870 #endif /* NVERIEXEC > 0 */
1871 			vattr.va_type = VREG;
1872 			vattr.va_rdev = VNOVAL;
1873 			optype = VOP_CREATE_DESCOFFSET;
1874 			break;
1875 		default:
1876 			error = EINVAL;
1877 			break;
1878 		}
1879 	}
1880 	if (!error) {
1881 		switch (optype) {
1882 		case VOP_WHITEOUT_DESCOFFSET:
1883 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1884 			if (error)
1885 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1886 			vput(nd.ni_dvp);
1887 			break;
1888 
1889 		case VOP_MKNOD_DESCOFFSET:
1890 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1891 						&nd.ni_cnd, &vattr);
1892 			if (error == 0)
1893 				vput(nd.ni_vp);
1894 			break;
1895 
1896 		case VOP_CREATE_DESCOFFSET:
1897 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1898 						&nd.ni_cnd, &vattr);
1899 			if (error == 0)
1900 				vput(nd.ni_vp);
1901 			break;
1902 		}
1903 	} else {
1904 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1905 		if (nd.ni_dvp == vp)
1906 			vrele(nd.ni_dvp);
1907 		else
1908 			vput(nd.ni_dvp);
1909 		if (vp)
1910 			vrele(vp);
1911 	}
1912 out:
1913 	VERIEXEC_PATH_PUT(path);
1914 	return (error);
1915 }
1916 
1917 /*
1918  * Create a named pipe.
1919  */
1920 /* ARGSUSED */
1921 int
1922 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1923 {
1924 	/* {
1925 		syscallarg(const char *) path;
1926 		syscallarg(int) mode;
1927 	} */
1928 	struct proc *p = l->l_proc;
1929 	struct vattr vattr;
1930 	int error;
1931 	struct nameidata nd;
1932 
1933 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1934 	    SCARG(uap, path));
1935 	if ((error = namei(&nd)) != 0)
1936 		return (error);
1937 	if (nd.ni_vp != NULL) {
1938 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1939 		if (nd.ni_dvp == nd.ni_vp)
1940 			vrele(nd.ni_dvp);
1941 		else
1942 			vput(nd.ni_dvp);
1943 		vrele(nd.ni_vp);
1944 		return (EEXIST);
1945 	}
1946 	VATTR_NULL(&vattr);
1947 	vattr.va_type = VFIFO;
1948 	/* We will read cwdi->cwdi_cmask unlocked. */
1949 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1950 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1951 	if (error == 0)
1952 		vput(nd.ni_vp);
1953 	return (error);
1954 }
1955 
1956 /*
1957  * Make a hard file link.
1958  */
1959 /* ARGSUSED */
1960 int
1961 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
1962 {
1963 	/* {
1964 		syscallarg(const char *) path;
1965 		syscallarg(const char *) link;
1966 	} */
1967 	struct vnode *vp;
1968 	struct nameidata nd;
1969 	int error;
1970 
1971 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1972 	    SCARG(uap, path));
1973 	if ((error = namei(&nd)) != 0)
1974 		return (error);
1975 	vp = nd.ni_vp;
1976 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1977 	    SCARG(uap, link));
1978 	if ((error = namei(&nd)) != 0)
1979 		goto out;
1980 	if (nd.ni_vp) {
1981 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1982 		if (nd.ni_dvp == nd.ni_vp)
1983 			vrele(nd.ni_dvp);
1984 		else
1985 			vput(nd.ni_dvp);
1986 		vrele(nd.ni_vp);
1987 		error = EEXIST;
1988 		goto out;
1989 	}
1990 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1991 out:
1992 	vrele(vp);
1993 	return (error);
1994 }
1995 
1996 /*
1997  * Make a symbolic link.
1998  */
1999 /* ARGSUSED */
2000 int
2001 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2002 {
2003 	/* {
2004 		syscallarg(const char *) path;
2005 		syscallarg(const char *) link;
2006 	} */
2007 	struct proc *p = l->l_proc;
2008 	struct vattr vattr;
2009 	char *path;
2010 	int error;
2011 	struct nameidata nd;
2012 
2013 	path = PNBUF_GET();
2014 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
2015 	if (error)
2016 		goto out;
2017 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2018 	    SCARG(uap, link));
2019 	if ((error = namei(&nd)) != 0)
2020 		goto out;
2021 	if (nd.ni_vp) {
2022 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2023 		if (nd.ni_dvp == nd.ni_vp)
2024 			vrele(nd.ni_dvp);
2025 		else
2026 			vput(nd.ni_dvp);
2027 		vrele(nd.ni_vp);
2028 		error = EEXIST;
2029 		goto out;
2030 	}
2031 	VATTR_NULL(&vattr);
2032 	vattr.va_type = VLNK;
2033 	/* We will read cwdi->cwdi_cmask unlocked. */
2034 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2035 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2036 	if (error == 0)
2037 		vput(nd.ni_vp);
2038 out:
2039 	PNBUF_PUT(path);
2040 	return (error);
2041 }
2042 
2043 /*
2044  * Delete a whiteout from the filesystem.
2045  */
2046 /* ARGSUSED */
2047 int
2048 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2049 {
2050 	/* {
2051 		syscallarg(const char *) path;
2052 	} */
2053 	int error;
2054 	struct nameidata nd;
2055 
2056 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
2057 	    UIO_USERSPACE, SCARG(uap, path));
2058 	error = namei(&nd);
2059 	if (error)
2060 		return (error);
2061 
2062 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2063 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2064 		if (nd.ni_dvp == nd.ni_vp)
2065 			vrele(nd.ni_dvp);
2066 		else
2067 			vput(nd.ni_dvp);
2068 		if (nd.ni_vp)
2069 			vrele(nd.ni_vp);
2070 		return (EEXIST);
2071 	}
2072 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2073 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2074 	vput(nd.ni_dvp);
2075 	return (error);
2076 }
2077 
2078 /*
2079  * Delete a name from the filesystem.
2080  */
2081 /* ARGSUSED */
2082 int
2083 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2084 {
2085 	/* {
2086 		syscallarg(const char *) path;
2087 	} */
2088 
2089 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2090 }
2091 
2092 int
2093 do_sys_unlink(const char *arg, enum uio_seg seg)
2094 {
2095 	struct vnode *vp;
2096 	int error;
2097 	struct nameidata nd;
2098 	kauth_cred_t cred;
2099 	char *path;
2100 	const char *cpath;
2101 
2102 	VERIEXEC_PATH_GET(arg, seg, cpath, path);
2103 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
2104 
2105 	if ((error = namei(&nd)) != 0)
2106 		goto out;
2107 	vp = nd.ni_vp;
2108 
2109 	/*
2110 	 * The root of a mounted filesystem cannot be deleted.
2111 	 */
2112 	if (vp->v_vflag & VV_ROOT) {
2113 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2114 		if (nd.ni_dvp == vp)
2115 			vrele(nd.ni_dvp);
2116 		else
2117 			vput(nd.ni_dvp);
2118 		vput(vp);
2119 		error = EBUSY;
2120 		goto out;
2121 	}
2122 
2123 #if NVERIEXEC > 0
2124 	/* Handle remove requests for veriexec entries. */
2125 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
2126 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2127 		if (nd.ni_dvp == vp)
2128 			vrele(nd.ni_dvp);
2129 		else
2130 			vput(nd.ni_dvp);
2131 		vput(vp);
2132 		goto out;
2133 	}
2134 #endif /* NVERIEXEC > 0 */
2135 
2136 	cred = kauth_cred_get();
2137 #ifdef FILEASSOC
2138 	(void)fileassoc_file_delete(vp);
2139 #endif /* FILEASSOC */
2140 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2141 out:
2142 	VERIEXEC_PATH_PUT(path);
2143 	return (error);
2144 }
2145 
2146 /*
2147  * Reposition read/write file offset.
2148  */
2149 int
2150 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2151 {
2152 	/* {
2153 		syscallarg(int) fd;
2154 		syscallarg(int) pad;
2155 		syscallarg(off_t) offset;
2156 		syscallarg(int) whence;
2157 	} */
2158 	kauth_cred_t cred = l->l_cred;
2159 	file_t *fp;
2160 	struct vnode *vp;
2161 	struct vattr vattr;
2162 	off_t newoff;
2163 	int error, fd;
2164 
2165 	fd = SCARG(uap, fd);
2166 
2167 	if ((fp = fd_getfile(fd)) == NULL)
2168 		return (EBADF);
2169 
2170 	vp = fp->f_data;
2171 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2172 		error = ESPIPE;
2173 		goto out;
2174 	}
2175 
2176 	switch (SCARG(uap, whence)) {
2177 	case SEEK_CUR:
2178 		newoff = fp->f_offset + SCARG(uap, offset);
2179 		break;
2180 	case SEEK_END:
2181 		error = VOP_GETATTR(vp, &vattr, cred);
2182 		if (error) {
2183 			goto out;
2184 		}
2185 		newoff = SCARG(uap, offset) + vattr.va_size;
2186 		break;
2187 	case SEEK_SET:
2188 		newoff = SCARG(uap, offset);
2189 		break;
2190 	default:
2191 		error = EINVAL;
2192 		goto out;
2193 	}
2194 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2195 		*(off_t *)retval = fp->f_offset = newoff;
2196 	}
2197  out:
2198  	fd_putfile(fd);
2199 	return (error);
2200 }
2201 
2202 /*
2203  * Positional read system call.
2204  */
2205 int
2206 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2207 {
2208 	/* {
2209 		syscallarg(int) fd;
2210 		syscallarg(void *) buf;
2211 		syscallarg(size_t) nbyte;
2212 		syscallarg(off_t) offset;
2213 	} */
2214 	file_t *fp;
2215 	struct vnode *vp;
2216 	off_t offset;
2217 	int error, fd = SCARG(uap, fd);
2218 
2219 	if ((fp = fd_getfile(fd)) == NULL)
2220 		return (EBADF);
2221 
2222 	if ((fp->f_flag & FREAD) == 0) {
2223 		fd_putfile(fd);
2224 		return (EBADF);
2225 	}
2226 
2227 	vp = fp->f_data;
2228 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2229 		error = ESPIPE;
2230 		goto out;
2231 	}
2232 
2233 	offset = SCARG(uap, offset);
2234 
2235 	/*
2236 	 * XXX This works because no file systems actually
2237 	 * XXX take any action on the seek operation.
2238 	 */
2239 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2240 		goto out;
2241 
2242 	/* dofileread() will unuse the descriptor for us */
2243 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2244 	    &offset, 0, retval));
2245 
2246  out:
2247 	fd_putfile(fd);
2248 	return (error);
2249 }
2250 
2251 /*
2252  * Positional scatter read system call.
2253  */
2254 int
2255 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2256 {
2257 	/* {
2258 		syscallarg(int) fd;
2259 		syscallarg(const struct iovec *) iovp;
2260 		syscallarg(int) iovcnt;
2261 		syscallarg(off_t) offset;
2262 	} */
2263 	off_t offset = SCARG(uap, offset);
2264 
2265 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2266 	    SCARG(uap, iovcnt), &offset, 0, retval);
2267 }
2268 
2269 /*
2270  * Positional write system call.
2271  */
2272 int
2273 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2274 {
2275 	/* {
2276 		syscallarg(int) fd;
2277 		syscallarg(const void *) buf;
2278 		syscallarg(size_t) nbyte;
2279 		syscallarg(off_t) offset;
2280 	} */
2281 	file_t *fp;
2282 	struct vnode *vp;
2283 	off_t offset;
2284 	int error, fd = SCARG(uap, fd);
2285 
2286 	if ((fp = fd_getfile(fd)) == NULL)
2287 		return (EBADF);
2288 
2289 	if ((fp->f_flag & FWRITE) == 0) {
2290 		fd_putfile(fd);
2291 		return (EBADF);
2292 	}
2293 
2294 	vp = fp->f_data;
2295 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2296 		error = ESPIPE;
2297 		goto out;
2298 	}
2299 
2300 	offset = SCARG(uap, offset);
2301 
2302 	/*
2303 	 * XXX This works because no file systems actually
2304 	 * XXX take any action on the seek operation.
2305 	 */
2306 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2307 		goto out;
2308 
2309 	/* dofilewrite() will unuse the descriptor for us */
2310 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2311 	    &offset, 0, retval));
2312 
2313  out:
2314 	fd_putfile(fd);
2315 	return (error);
2316 }
2317 
2318 /*
2319  * Positional gather write system call.
2320  */
2321 int
2322 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2323 {
2324 	/* {
2325 		syscallarg(int) fd;
2326 		syscallarg(const struct iovec *) iovp;
2327 		syscallarg(int) iovcnt;
2328 		syscallarg(off_t) offset;
2329 	} */
2330 	off_t offset = SCARG(uap, offset);
2331 
2332 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2333 	    SCARG(uap, iovcnt), &offset, 0, retval);
2334 }
2335 
2336 /*
2337  * Check access permissions.
2338  */
2339 int
2340 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2341 {
2342 	/* {
2343 		syscallarg(const char *) path;
2344 		syscallarg(int) flags;
2345 	} */
2346 	kauth_cred_t cred;
2347 	struct vnode *vp;
2348 	int error, flags;
2349 	struct nameidata nd;
2350 
2351 	cred = kauth_cred_dup(l->l_cred);
2352 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2353 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2354 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2355 	    SCARG(uap, path));
2356 	/* Override default credentials */
2357 	nd.ni_cnd.cn_cred = cred;
2358 	if ((error = namei(&nd)) != 0)
2359 		goto out;
2360 	vp = nd.ni_vp;
2361 
2362 	/* Flags == 0 means only check for existence. */
2363 	if (SCARG(uap, flags)) {
2364 		flags = 0;
2365 		if (SCARG(uap, flags) & R_OK)
2366 			flags |= VREAD;
2367 		if (SCARG(uap, flags) & W_OK)
2368 			flags |= VWRITE;
2369 		if (SCARG(uap, flags) & X_OK)
2370 			flags |= VEXEC;
2371 
2372 		error = VOP_ACCESS(vp, flags, cred);
2373 		if (!error && (flags & VWRITE))
2374 			error = vn_writechk(vp);
2375 	}
2376 	vput(vp);
2377 out:
2378 	kauth_cred_free(cred);
2379 	return (error);
2380 }
2381 
2382 /*
2383  * Common code for all sys_stat functions, including compat versions.
2384  */
2385 int
2386 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
2387 {
2388 	int error;
2389 	struct nameidata nd;
2390 
2391 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
2392 	    UIO_USERSPACE, path);
2393 	error = namei(&nd);
2394 	if (error != 0)
2395 		return error;
2396 	error = vn_stat(nd.ni_vp, sb);
2397 	vput(nd.ni_vp);
2398 	return error;
2399 }
2400 
2401 /*
2402  * Get file status; this version follows links.
2403  */
2404 /* ARGSUSED */
2405 int
2406 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval)
2407 {
2408 	/* {
2409 		syscallarg(const char *) path;
2410 		syscallarg(struct stat *) ub;
2411 	} */
2412 	struct stat sb;
2413 	int error;
2414 
2415 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2416 	if (error)
2417 		return error;
2418 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2419 }
2420 
2421 /*
2422  * Get file status; this version does not follow links.
2423  */
2424 /* ARGSUSED */
2425 int
2426 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval)
2427 {
2428 	/* {
2429 		syscallarg(const char *) path;
2430 		syscallarg(struct stat *) ub;
2431 	} */
2432 	struct stat sb;
2433 	int error;
2434 
2435 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2436 	if (error)
2437 		return error;
2438 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2439 }
2440 
2441 /*
2442  * Get configurable pathname variables.
2443  */
2444 /* ARGSUSED */
2445 int
2446 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2447 {
2448 	/* {
2449 		syscallarg(const char *) path;
2450 		syscallarg(int) name;
2451 	} */
2452 	int error;
2453 	struct nameidata nd;
2454 
2455 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2456 	    SCARG(uap, path));
2457 	if ((error = namei(&nd)) != 0)
2458 		return (error);
2459 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2460 	vput(nd.ni_vp);
2461 	return (error);
2462 }
2463 
2464 /*
2465  * Return target name of a symbolic link.
2466  */
2467 /* ARGSUSED */
2468 int
2469 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2470 {
2471 	/* {
2472 		syscallarg(const char *) path;
2473 		syscallarg(char *) buf;
2474 		syscallarg(size_t) count;
2475 	} */
2476 	struct vnode *vp;
2477 	struct iovec aiov;
2478 	struct uio auio;
2479 	int error;
2480 	struct nameidata nd;
2481 
2482 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2483 	    SCARG(uap, path));
2484 	if ((error = namei(&nd)) != 0)
2485 		return (error);
2486 	vp = nd.ni_vp;
2487 	if (vp->v_type != VLNK)
2488 		error = EINVAL;
2489 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2490 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2491 		aiov.iov_base = SCARG(uap, buf);
2492 		aiov.iov_len = SCARG(uap, count);
2493 		auio.uio_iov = &aiov;
2494 		auio.uio_iovcnt = 1;
2495 		auio.uio_offset = 0;
2496 		auio.uio_rw = UIO_READ;
2497 		KASSERT(l == curlwp);
2498 		auio.uio_vmspace = l->l_proc->p_vmspace;
2499 		auio.uio_resid = SCARG(uap, count);
2500 		error = VOP_READLINK(vp, &auio, l->l_cred);
2501 	}
2502 	vput(vp);
2503 	*retval = SCARG(uap, count) - auio.uio_resid;
2504 	return (error);
2505 }
2506 
2507 /*
2508  * Change flags of a file given a path name.
2509  */
2510 /* ARGSUSED */
2511 int
2512 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2513 {
2514 	/* {
2515 		syscallarg(const char *) path;
2516 		syscallarg(u_long) flags;
2517 	} */
2518 	struct vnode *vp;
2519 	int error;
2520 	struct nameidata nd;
2521 
2522 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2523 	    SCARG(uap, path));
2524 	if ((error = namei(&nd)) != 0)
2525 		return (error);
2526 	vp = nd.ni_vp;
2527 	error = change_flags(vp, SCARG(uap, flags), l);
2528 	vput(vp);
2529 	return (error);
2530 }
2531 
2532 /*
2533  * Change flags of a file given a file descriptor.
2534  */
2535 /* ARGSUSED */
2536 int
2537 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
2538 {
2539 	/* {
2540 		syscallarg(int) fd;
2541 		syscallarg(u_long) flags;
2542 	} */
2543 	struct vnode *vp;
2544 	file_t *fp;
2545 	int error;
2546 
2547 	/* fd_getvnode() will use the descriptor for us */
2548 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2549 		return (error);
2550 	vp = fp->f_data;
2551 	error = change_flags(vp, SCARG(uap, flags), l);
2552 	VOP_UNLOCK(vp, 0);
2553 	fd_putfile(SCARG(uap, fd));
2554 	return (error);
2555 }
2556 
2557 /*
2558  * Change flags of a file given a path name; this version does
2559  * not follow links.
2560  */
2561 int
2562 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
2563 {
2564 	/* {
2565 		syscallarg(const char *) path;
2566 		syscallarg(u_long) flags;
2567 	} */
2568 	struct vnode *vp;
2569 	int error;
2570 	struct nameidata nd;
2571 
2572 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2573 	    SCARG(uap, path));
2574 	if ((error = namei(&nd)) != 0)
2575 		return (error);
2576 	vp = nd.ni_vp;
2577 	error = change_flags(vp, SCARG(uap, flags), l);
2578 	vput(vp);
2579 	return (error);
2580 }
2581 
2582 /*
2583  * Common routine to change flags of a file.
2584  */
2585 int
2586 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2587 {
2588 	struct vattr vattr;
2589 	int error;
2590 
2591 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2592 	/*
2593 	 * Non-superusers cannot change the flags on devices, even if they
2594 	 * own them.
2595 	 */
2596 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2597 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2598 			goto out;
2599 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2600 			error = EINVAL;
2601 			goto out;
2602 		}
2603 	}
2604 	VATTR_NULL(&vattr);
2605 	vattr.va_flags = flags;
2606 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2607 out:
2608 	return (error);
2609 }
2610 
2611 /*
2612  * Change mode of a file given path name; this version follows links.
2613  */
2614 /* ARGSUSED */
2615 int
2616 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
2617 {
2618 	/* {
2619 		syscallarg(const char *) path;
2620 		syscallarg(int) mode;
2621 	} */
2622 	int error;
2623 	struct nameidata nd;
2624 
2625 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2626 	    SCARG(uap, path));
2627 	if ((error = namei(&nd)) != 0)
2628 		return (error);
2629 
2630 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2631 
2632 	vrele(nd.ni_vp);
2633 	return (error);
2634 }
2635 
2636 /*
2637  * Change mode of a file given a file descriptor.
2638  */
2639 /* ARGSUSED */
2640 int
2641 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
2642 {
2643 	/* {
2644 		syscallarg(int) fd;
2645 		syscallarg(int) mode;
2646 	} */
2647 	file_t *fp;
2648 	int error;
2649 
2650 	/* fd_getvnode() will use the descriptor for us */
2651 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2652 		return (error);
2653 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
2654 	fd_putfile(SCARG(uap, fd));
2655 	return (error);
2656 }
2657 
2658 /*
2659  * Change mode of a file given path name; this version does not follow links.
2660  */
2661 /* ARGSUSED */
2662 int
2663 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
2664 {
2665 	/* {
2666 		syscallarg(const char *) path;
2667 		syscallarg(int) mode;
2668 	} */
2669 	int error;
2670 	struct nameidata nd;
2671 
2672 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2673 	    SCARG(uap, path));
2674 	if ((error = namei(&nd)) != 0)
2675 		return (error);
2676 
2677 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2678 
2679 	vrele(nd.ni_vp);
2680 	return (error);
2681 }
2682 
2683 /*
2684  * Common routine to set mode given a vnode.
2685  */
2686 static int
2687 change_mode(struct vnode *vp, int mode, struct lwp *l)
2688 {
2689 	struct vattr vattr;
2690 	int error;
2691 
2692 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2693 	VATTR_NULL(&vattr);
2694 	vattr.va_mode = mode & ALLPERMS;
2695 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2696 	VOP_UNLOCK(vp, 0);
2697 	return (error);
2698 }
2699 
2700 /*
2701  * Set ownership given a path name; this version follows links.
2702  */
2703 /* ARGSUSED */
2704 int
2705 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
2706 {
2707 	/* {
2708 		syscallarg(const char *) path;
2709 		syscallarg(uid_t) uid;
2710 		syscallarg(gid_t) gid;
2711 	} */
2712 	int error;
2713 	struct nameidata nd;
2714 
2715 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2716 	    SCARG(uap, path));
2717 	if ((error = namei(&nd)) != 0)
2718 		return (error);
2719 
2720 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2721 
2722 	vrele(nd.ni_vp);
2723 	return (error);
2724 }
2725 
2726 /*
2727  * Set ownership given a path name; this version follows links.
2728  * Provides POSIX semantics.
2729  */
2730 /* ARGSUSED */
2731 int
2732 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
2733 {
2734 	/* {
2735 		syscallarg(const char *) path;
2736 		syscallarg(uid_t) uid;
2737 		syscallarg(gid_t) gid;
2738 	} */
2739 	int error;
2740 	struct nameidata nd;
2741 
2742 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2743 	    SCARG(uap, path));
2744 	if ((error = namei(&nd)) != 0)
2745 		return (error);
2746 
2747 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2748 
2749 	vrele(nd.ni_vp);
2750 	return (error);
2751 }
2752 
2753 /*
2754  * Set ownership given a file descriptor.
2755  */
2756 /* ARGSUSED */
2757 int
2758 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
2759 {
2760 	/* {
2761 		syscallarg(int) fd;
2762 		syscallarg(uid_t) uid;
2763 		syscallarg(gid_t) gid;
2764 	} */
2765 	int error;
2766 	file_t *fp;
2767 
2768 	/* fd_getvnode() will use the descriptor for us */
2769 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2770 		return (error);
2771 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2772 	    l, 0);
2773 	fd_putfile(SCARG(uap, fd));
2774 	return (error);
2775 }
2776 
2777 /*
2778  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2779  */
2780 /* ARGSUSED */
2781 int
2782 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
2783 {
2784 	/* {
2785 		syscallarg(int) fd;
2786 		syscallarg(uid_t) uid;
2787 		syscallarg(gid_t) gid;
2788 	} */
2789 	int error;
2790 	file_t *fp;
2791 
2792 	/* fd_getvnode() will use the descriptor for us */
2793 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2794 		return (error);
2795 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2796 	    l, 1);
2797 	fd_putfile(SCARG(uap, fd));
2798 	return (error);
2799 }
2800 
2801 /*
2802  * Set ownership given a path name; this version does not follow links.
2803  */
2804 /* ARGSUSED */
2805 int
2806 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
2807 {
2808 	/* {
2809 		syscallarg(const char *) path;
2810 		syscallarg(uid_t) uid;
2811 		syscallarg(gid_t) gid;
2812 	} */
2813 	int error;
2814 	struct nameidata nd;
2815 
2816 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2817 	    SCARG(uap, path));
2818 	if ((error = namei(&nd)) != 0)
2819 		return (error);
2820 
2821 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2822 
2823 	vrele(nd.ni_vp);
2824 	return (error);
2825 }
2826 
2827 /*
2828  * Set ownership given a path name; this version does not follow links.
2829  * Provides POSIX/XPG semantics.
2830  */
2831 /* ARGSUSED */
2832 int
2833 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
2834 {
2835 	/* {
2836 		syscallarg(const char *) path;
2837 		syscallarg(uid_t) uid;
2838 		syscallarg(gid_t) gid;
2839 	} */
2840 	int error;
2841 	struct nameidata nd;
2842 
2843 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2844 	    SCARG(uap, path));
2845 	if ((error = namei(&nd)) != 0)
2846 		return (error);
2847 
2848 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2849 
2850 	vrele(nd.ni_vp);
2851 	return (error);
2852 }
2853 
2854 /*
2855  * Common routine to set ownership given a vnode.
2856  */
2857 static int
2858 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2859     int posix_semantics)
2860 {
2861 	struct vattr vattr;
2862 	mode_t newmode;
2863 	int error;
2864 
2865 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2866 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2867 		goto out;
2868 
2869 #define CHANGED(x) ((int)(x) != -1)
2870 	newmode = vattr.va_mode;
2871 	if (posix_semantics) {
2872 		/*
2873 		 * POSIX/XPG semantics: if the caller is not the super-user,
2874 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2875 		 * the XPG consider the behaviour for calls by the super-user
2876 		 * implementation-defined; we leave the set-user-id and set-
2877 		 * group-id settings intact in that case.
2878 		 */
2879 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2880 				      NULL) != 0)
2881 			newmode &= ~(S_ISUID | S_ISGID);
2882 	} else {
2883 		/*
2884 		 * NetBSD semantics: when changing owner and/or group,
2885 		 * clear the respective bit(s).
2886 		 */
2887 		if (CHANGED(uid))
2888 			newmode &= ~S_ISUID;
2889 		if (CHANGED(gid))
2890 			newmode &= ~S_ISGID;
2891 	}
2892 	/* Update va_mode iff altered. */
2893 	if (vattr.va_mode == newmode)
2894 		newmode = VNOVAL;
2895 
2896 	VATTR_NULL(&vattr);
2897 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2898 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2899 	vattr.va_mode = newmode;
2900 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2901 #undef CHANGED
2902 
2903 out:
2904 	VOP_UNLOCK(vp, 0);
2905 	return (error);
2906 }
2907 
2908 /*
2909  * Set the access and modification times given a path name; this
2910  * version follows links.
2911  */
2912 /* ARGSUSED */
2913 int
2914 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval)
2915 {
2916 	/* {
2917 		syscallarg(const char *) path;
2918 		syscallarg(const struct timeval *) tptr;
2919 	} */
2920 
2921 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
2922 	    SCARG(uap, tptr), UIO_USERSPACE);
2923 }
2924 
2925 /*
2926  * Set the access and modification times given a file descriptor.
2927  */
2928 /* ARGSUSED */
2929 int
2930 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval)
2931 {
2932 	/* {
2933 		syscallarg(int) fd;
2934 		syscallarg(const struct timeval *) tptr;
2935 	} */
2936 	int error;
2937 	file_t *fp;
2938 
2939 	/* fd_getvnode() will use the descriptor for us */
2940 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2941 		return (error);
2942 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
2943 	    UIO_USERSPACE);
2944 	fd_putfile(SCARG(uap, fd));
2945 	return (error);
2946 }
2947 
2948 /*
2949  * Set the access and modification times given a path name; this
2950  * version does not follow links.
2951  */
2952 int
2953 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval)
2954 {
2955 	/* {
2956 		syscallarg(const char *) path;
2957 		syscallarg(const struct timeval *) tptr;
2958 	} */
2959 
2960 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
2961 	    SCARG(uap, tptr), UIO_USERSPACE);
2962 }
2963 
2964 /*
2965  * Common routine to set access and modification times given a vnode.
2966  */
2967 int
2968 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
2969     const struct timeval *tptr, enum uio_seg seg)
2970 {
2971 	struct vattr vattr;
2972 	struct nameidata nd;
2973 	int error;
2974 
2975 	VATTR_NULL(&vattr);
2976 	if (tptr == NULL) {
2977 		nanotime(&vattr.va_atime);
2978 		vattr.va_mtime = vattr.va_atime;
2979 		vattr.va_vaflags |= VA_UTIMES_NULL;
2980 	} else {
2981 		struct timeval tv[2];
2982 
2983 		if (seg != UIO_SYSSPACE) {
2984 			error = copyin(tptr, &tv, sizeof (tv));
2985 			if (error != 0)
2986 				return error;
2987 			tptr = tv;
2988 		}
2989 		TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime);
2990 		TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime);
2991 	}
2992 
2993 	if (vp == NULL) {
2994 		NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path);
2995 		if ((error = namei(&nd)) != 0)
2996 			return (error);
2997 		vp = nd.ni_vp;
2998 	} else
2999 		nd.ni_vp = NULL;
3000 
3001 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3002 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3003 	VOP_UNLOCK(vp, 0);
3004 
3005 	if (nd.ni_vp != NULL)
3006 		vrele(nd.ni_vp);
3007 
3008 	return (error);
3009 }
3010 
3011 /*
3012  * Truncate a file given its path name.
3013  */
3014 /* ARGSUSED */
3015 int
3016 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3017 {
3018 	/* {
3019 		syscallarg(const char *) path;
3020 		syscallarg(int) pad;
3021 		syscallarg(off_t) length;
3022 	} */
3023 	struct vnode *vp;
3024 	struct vattr vattr;
3025 	int error;
3026 	struct nameidata nd;
3027 
3028 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
3029 	    SCARG(uap, path));
3030 	if ((error = namei(&nd)) != 0)
3031 		return (error);
3032 	vp = nd.ni_vp;
3033 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3034 	if (vp->v_type == VDIR)
3035 		error = EISDIR;
3036 	else if ((error = vn_writechk(vp)) == 0 &&
3037 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3038 		VATTR_NULL(&vattr);
3039 		vattr.va_size = SCARG(uap, length);
3040 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3041 	}
3042 	vput(vp);
3043 	return (error);
3044 }
3045 
3046 /*
3047  * Truncate a file given a file descriptor.
3048  */
3049 /* ARGSUSED */
3050 int
3051 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3052 {
3053 	/* {
3054 		syscallarg(int) fd;
3055 		syscallarg(int) pad;
3056 		syscallarg(off_t) length;
3057 	} */
3058 	struct vattr vattr;
3059 	struct vnode *vp;
3060 	file_t *fp;
3061 	int error;
3062 
3063 	/* fd_getvnode() will use the descriptor for us */
3064 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3065 		return (error);
3066 	if ((fp->f_flag & FWRITE) == 0) {
3067 		error = EINVAL;
3068 		goto out;
3069 	}
3070 	vp = fp->f_data;
3071 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3072 	if (vp->v_type == VDIR)
3073 		error = EISDIR;
3074 	else if ((error = vn_writechk(vp)) == 0) {
3075 		VATTR_NULL(&vattr);
3076 		vattr.va_size = SCARG(uap, length);
3077 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3078 	}
3079 	VOP_UNLOCK(vp, 0);
3080  out:
3081 	fd_putfile(SCARG(uap, fd));
3082 	return (error);
3083 }
3084 
3085 /*
3086  * Sync an open file.
3087  */
3088 /* ARGSUSED */
3089 int
3090 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3091 {
3092 	/* {
3093 		syscallarg(int) fd;
3094 	} */
3095 	struct vnode *vp;
3096 	file_t *fp;
3097 	int error;
3098 
3099 	/* fd_getvnode() will use the descriptor for us */
3100 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3101 		return (error);
3102 	vp = fp->f_data;
3103 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3104 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3105 	if (error == 0 && bioopsp != NULL &&
3106 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3107 		(*bioopsp->io_fsync)(vp, 0);
3108 	VOP_UNLOCK(vp, 0);
3109 	fd_putfile(SCARG(uap, fd));
3110 	return (error);
3111 }
3112 
3113 /*
3114  * Sync a range of file data.  API modeled after that found in AIX.
3115  *
3116  * FDATASYNC indicates that we need only save enough metadata to be able
3117  * to re-read the written data.  Note we duplicate AIX's requirement that
3118  * the file be open for writing.
3119  */
3120 /* ARGSUSED */
3121 int
3122 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3123 {
3124 	/* {
3125 		syscallarg(int) fd;
3126 		syscallarg(int) flags;
3127 		syscallarg(off_t) start;
3128 		syscallarg(off_t) length;
3129 	} */
3130 	struct vnode *vp;
3131 	file_t *fp;
3132 	int flags, nflags;
3133 	off_t s, e, len;
3134 	int error;
3135 
3136 	/* fd_getvnode() will use the descriptor for us */
3137 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3138 		return (error);
3139 
3140 	if ((fp->f_flag & FWRITE) == 0) {
3141 		error = EBADF;
3142 		goto out;
3143 	}
3144 
3145 	flags = SCARG(uap, flags);
3146 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3147 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3148 		error = EINVAL;
3149 		goto out;
3150 	}
3151 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3152 	if (flags & FDATASYNC)
3153 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3154 	else
3155 		nflags = FSYNC_WAIT;
3156 	if (flags & FDISKSYNC)
3157 		nflags |= FSYNC_CACHE;
3158 
3159 	len = SCARG(uap, length);
3160 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3161 	if (len) {
3162 		s = SCARG(uap, start);
3163 		e = s + len;
3164 		if (e < s) {
3165 			error = EINVAL;
3166 			goto out;
3167 		}
3168 	} else {
3169 		e = 0;
3170 		s = 0;
3171 	}
3172 
3173 	vp = fp->f_data;
3174 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3175 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3176 
3177 	if (error == 0 && bioopsp != NULL &&
3178 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3179 		(*bioopsp->io_fsync)(vp, nflags);
3180 
3181 	VOP_UNLOCK(vp, 0);
3182 out:
3183 	fd_putfile(SCARG(uap, fd));
3184 	return (error);
3185 }
3186 
3187 /*
3188  * Sync the data of an open file.
3189  */
3190 /* ARGSUSED */
3191 int
3192 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3193 {
3194 	/* {
3195 		syscallarg(int) fd;
3196 	} */
3197 	struct vnode *vp;
3198 	file_t *fp;
3199 	int error;
3200 
3201 	/* fd_getvnode() will use the descriptor for us */
3202 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3203 		return (error);
3204 	if ((fp->f_flag & FWRITE) == 0) {
3205 		fd_putfile(SCARG(uap, fd));
3206 		return (EBADF);
3207 	}
3208 	vp = fp->f_data;
3209 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3210 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3211 	VOP_UNLOCK(vp, 0);
3212 	fd_putfile(SCARG(uap, fd));
3213 	return (error);
3214 }
3215 
3216 /*
3217  * Rename files, (standard) BSD semantics frontend.
3218  */
3219 /* ARGSUSED */
3220 int
3221 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3222 {
3223 	/* {
3224 		syscallarg(const char *) from;
3225 		syscallarg(const char *) to;
3226 	} */
3227 
3228 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3229 }
3230 
3231 /*
3232  * Rename files, POSIX semantics frontend.
3233  */
3234 /* ARGSUSED */
3235 int
3236 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3237 {
3238 	/* {
3239 		syscallarg(const char *) from;
3240 		syscallarg(const char *) to;
3241 	} */
3242 
3243 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3244 }
3245 
3246 /*
3247  * Rename files.  Source and destination must either both be directories,
3248  * or both not be directories.  If target is a directory, it must be empty.
3249  * If `from' and `to' refer to the same object, the value of the `retain'
3250  * argument is used to determine whether `from' will be
3251  *
3252  * (retain == 0)	deleted unless `from' and `to' refer to the same
3253  *			object in the file system's name space (BSD).
3254  * (retain == 1)	always retained (POSIX).
3255  */
3256 int
3257 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3258 {
3259 	struct vnode *tvp, *fvp, *tdvp;
3260 	struct nameidata fromnd, tond;
3261 	struct mount *fs;
3262 	struct lwp *l = curlwp;
3263 	struct proc *p;
3264 	uint32_t saveflag;
3265 	int error;
3266 
3267 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT,
3268 	    seg, from);
3269 	if ((error = namei(&fromnd)) != 0)
3270 		return (error);
3271 	if (fromnd.ni_dvp != fromnd.ni_vp)
3272 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3273 	fvp = fromnd.ni_vp;
3274 
3275 	fs = fvp->v_mount;
3276 	error = VFS_RENAMELOCK_ENTER(fs);
3277 	if (error) {
3278 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3279 		vrele(fromnd.ni_dvp);
3280 		vrele(fvp);
3281 		goto out1;
3282 	}
3283 
3284 	/*
3285 	 * close, partially, yet another race - ideally we should only
3286 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3287 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3288 	 * that with namei as it stands.
3289 	 *
3290 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3291 	 * under us. The real fix is to get the locks in the right
3292 	 * order and do the lookups in the right places, but that's a
3293 	 * major rototill.
3294 	 *
3295 	 * Preserve the SAVESTART in cn_flags, because who knows what
3296 	 * might happen if we don't.
3297 	 *
3298 	 * Note: this logic (as well as this whole function) is cloned
3299 	 * in nfs_serv.c. Proceed accordingly.
3300 	 */
3301 	vrele(fvp);
3302 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3303 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3304 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3305 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3306 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3307 		error = EINVAL;
3308 		VFS_RENAMELOCK_EXIT(fs);
3309 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3310 		vrele(fromnd.ni_dvp);
3311 		goto out1;
3312 	}
3313 	saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
3314 	fromnd.ni_cnd.cn_flags &= ~SAVESTART;
3315 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3316 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
3317 	fromnd.ni_cnd.cn_flags |= saveflag;
3318 	if (error) {
3319 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3320 		VFS_RENAMELOCK_EXIT(fs);
3321 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3322 		vrele(fromnd.ni_dvp);
3323 		goto out1;
3324 	}
3325 	VOP_UNLOCK(fromnd.ni_vp, 0);
3326 	if (fromnd.ni_dvp != fromnd.ni_vp)
3327 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3328 	fvp = fromnd.ni_vp;
3329 
3330 	NDINIT(&tond, RENAME,
3331 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3332 	      | (fvp->v_type == VDIR ? CREATEDIR : 0),
3333 	    seg, to);
3334 	if ((error = namei(&tond)) != 0) {
3335 		VFS_RENAMELOCK_EXIT(fs);
3336 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3337 		vrele(fromnd.ni_dvp);
3338 		vrele(fvp);
3339 		goto out1;
3340 	}
3341 	tdvp = tond.ni_dvp;
3342 	tvp = tond.ni_vp;
3343 
3344 	if (tvp != NULL) {
3345 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3346 			error = ENOTDIR;
3347 			goto out;
3348 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3349 			error = EISDIR;
3350 			goto out;
3351 		}
3352 	}
3353 
3354 	if (fvp == tdvp)
3355 		error = EINVAL;
3356 
3357 	/*
3358 	 * Source and destination refer to the same object.
3359 	 */
3360 	if (fvp == tvp) {
3361 		if (retain)
3362 			error = -1;
3363 		else if (fromnd.ni_dvp == tdvp &&
3364 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3365 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3366 		          tond.ni_cnd.cn_nameptr,
3367 		          fromnd.ni_cnd.cn_namelen))
3368 		error = -1;
3369 	}
3370 
3371 #if NVERIEXEC > 0
3372 	if (!error) {
3373 		char *f1, *f2;
3374 
3375 		f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
3376 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen);
3377 
3378 		f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
3379 		strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen);
3380 
3381 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
3382 
3383 		free(f1, M_TEMP);
3384 		free(f2, M_TEMP);
3385 	}
3386 #endif /* NVERIEXEC > 0 */
3387 
3388 out:
3389 	p = l->l_proc;
3390 	if (!error) {
3391 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3392 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3393 		VFS_RENAMELOCK_EXIT(fs);
3394 	} else {
3395 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3396 		if (tdvp == tvp)
3397 			vrele(tdvp);
3398 		else
3399 			vput(tdvp);
3400 		if (tvp)
3401 			vput(tvp);
3402 		VFS_RENAMELOCK_EXIT(fs);
3403 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3404 		vrele(fromnd.ni_dvp);
3405 		vrele(fvp);
3406 	}
3407 	vrele(tond.ni_startdir);
3408 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3409 out1:
3410 	if (fromnd.ni_startdir)
3411 		vrele(fromnd.ni_startdir);
3412 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3413 	return (error == -1 ? 0 : error);
3414 }
3415 
3416 /*
3417  * Make a directory file.
3418  */
3419 /* ARGSUSED */
3420 int
3421 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
3422 {
3423 	/* {
3424 		syscallarg(const char *) path;
3425 		syscallarg(int) mode;
3426 	} */
3427 	struct proc *p = l->l_proc;
3428 	struct vnode *vp;
3429 	struct vattr vattr;
3430 	int error;
3431 	struct nameidata nd;
3432 
3433 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE,
3434 	    SCARG(uap, path));
3435 	if ((error = namei(&nd)) != 0)
3436 		return (error);
3437 	vp = nd.ni_vp;
3438 	if (vp != NULL) {
3439 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3440 		if (nd.ni_dvp == vp)
3441 			vrele(nd.ni_dvp);
3442 		else
3443 			vput(nd.ni_dvp);
3444 		vrele(vp);
3445 		return (EEXIST);
3446 	}
3447 	VATTR_NULL(&vattr);
3448 	vattr.va_type = VDIR;
3449 	/* We will read cwdi->cwdi_cmask unlocked. */
3450 	vattr.va_mode =
3451 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3452 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3453 	if (!error)
3454 		vput(nd.ni_vp);
3455 	return (error);
3456 }
3457 
3458 /*
3459  * Remove a directory file.
3460  */
3461 /* ARGSUSED */
3462 int
3463 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
3464 {
3465 	/* {
3466 		syscallarg(const char *) path;
3467 	} */
3468 	struct vnode *vp;
3469 	int error;
3470 	struct nameidata nd;
3471 
3472 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
3473 	    SCARG(uap, path));
3474 	if ((error = namei(&nd)) != 0)
3475 		return (error);
3476 	vp = nd.ni_vp;
3477 	if (vp->v_type != VDIR) {
3478 		error = ENOTDIR;
3479 		goto out;
3480 	}
3481 	/*
3482 	 * No rmdir "." please.
3483 	 */
3484 	if (nd.ni_dvp == vp) {
3485 		error = EINVAL;
3486 		goto out;
3487 	}
3488 	/*
3489 	 * The root of a mounted filesystem cannot be deleted.
3490 	 */
3491 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
3492 		error = EBUSY;
3493 		goto out;
3494 	}
3495 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3496 	return (error);
3497 
3498 out:
3499 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3500 	if (nd.ni_dvp == vp)
3501 		vrele(nd.ni_dvp);
3502 	else
3503 		vput(nd.ni_dvp);
3504 	vput(vp);
3505 	return (error);
3506 }
3507 
3508 /*
3509  * Read a block of directory entries in a file system independent format.
3510  */
3511 int
3512 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
3513 {
3514 	/* {
3515 		syscallarg(int) fd;
3516 		syscallarg(char *) buf;
3517 		syscallarg(size_t) count;
3518 	} */
3519 	file_t *fp;
3520 	int error, done;
3521 
3522 	/* fd_getvnode() will use the descriptor for us */
3523 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3524 		return (error);
3525 	if ((fp->f_flag & FREAD) == 0) {
3526 		error = EBADF;
3527 		goto out;
3528 	}
3529 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3530 			SCARG(uap, count), &done, l, 0, 0);
3531 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3532 	*retval = done;
3533  out:
3534 	fd_putfile(SCARG(uap, fd));
3535 	return (error);
3536 }
3537 
3538 /*
3539  * Set the mode mask for creation of filesystem nodes.
3540  */
3541 int
3542 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
3543 {
3544 	/* {
3545 		syscallarg(mode_t) newmask;
3546 	} */
3547 	struct proc *p = l->l_proc;
3548 	struct cwdinfo *cwdi;
3549 
3550 	/*
3551 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3552 	 * important is that we serialize changes to the mask.  The
3553 	 * rw_exit() will issue a write memory barrier on our behalf,
3554 	 * and force the changes out to other CPUs (as it must use an
3555 	 * atomic operation, draining the local CPU's store buffers).
3556 	 */
3557 	cwdi = p->p_cwdi;
3558 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3559 	*retval = cwdi->cwdi_cmask;
3560 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3561 	rw_exit(&cwdi->cwdi_lock);
3562 
3563 	return (0);
3564 }
3565 
3566 int
3567 dorevoke(struct vnode *vp, kauth_cred_t cred)
3568 {
3569 	struct vattr vattr;
3570 	int error;
3571 
3572 	if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
3573 		return error;
3574 	if (kauth_cred_geteuid(cred) != vattr.va_uid &&
3575 	    (error = kauth_authorize_generic(cred,
3576 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
3577 		VOP_REVOKE(vp, REVOKEALL);
3578 	return (error);
3579 }
3580 
3581 /*
3582  * Void all references to file by ripping underlying filesystem
3583  * away from vnode.
3584  */
3585 /* ARGSUSED */
3586 int
3587 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
3588 {
3589 	/* {
3590 		syscallarg(const char *) path;
3591 	} */
3592 	struct vnode *vp;
3593 	int error;
3594 	struct nameidata nd;
3595 
3596 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
3597 	    SCARG(uap, path));
3598 	if ((error = namei(&nd)) != 0)
3599 		return (error);
3600 	vp = nd.ni_vp;
3601 	error = dorevoke(vp, l->l_cred);
3602 	vrele(vp);
3603 	return (error);
3604 }
3605 
3606 /*
3607  * Convert a user file descriptor to a kernel file entry.
3608  */
3609 int
3610 getvnode(int fd, file_t **fpp)
3611 {
3612 	struct vnode *vp;
3613 	file_t *fp;
3614 
3615 	if ((fp = fd_getfile(fd)) == NULL)
3616 		return (EBADF);
3617 
3618 	if (fp->f_type != DTYPE_VNODE) {
3619 		fd_putfile(fd);
3620 		return (EINVAL);
3621 	}
3622 
3623 	vp = fp->f_data;
3624 	if (vp->v_type == VBAD) {
3625 		fd_putfile(fd);
3626 		return (EBADF);
3627 	}
3628 
3629 	*fpp = fp;
3630 	return (0);
3631 }
3632