xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 96230fab84e26a6435963032070e916a951a8b2e)
1 /*	$NetBSD: vfs_syscalls.c,v 1.375 2008/09/25 14:37:30 wiz 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  *
16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26  * POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 /*
30  * Copyright (c) 1989, 1993
31  *	The Regents of the University of California.  All rights reserved.
32  * (c) UNIX System Laboratories, Inc.
33  * All or some portions of this file are derived from material licensed
34  * to the University of California by American Telephone and Telegraph
35  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36  * the permission of UNIX System Laboratories, Inc.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
63  */
64 
65 #include <sys/cdefs.h>
66 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.375 2008/09/25 14:37:30 wiz Exp $");
67 
68 #include "opt_compat_netbsd.h"
69 #include "opt_compat_43.h"
70 #include "opt_fileassoc.h"
71 #include "veriexec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/namei.h>
76 #include <sys/filedesc.h>
77 #include <sys/kernel.h>
78 #include <sys/file.h>
79 #include <sys/stat.h>
80 #include <sys/vnode.h>
81 #include <sys/mount.h>
82 #include <sys/proc.h>
83 #include <sys/uio.h>
84 #include <sys/malloc.h>
85 #include <sys/kmem.h>
86 #include <sys/dirent.h>
87 #include <sys/sysctl.h>
88 #include <sys/syscallargs.h>
89 #include <sys/vfs_syscalls.h>
90 #include <sys/ktrace.h>
91 #ifdef FILEASSOC
92 #include <sys/fileassoc.h>
93 #endif /* FILEASSOC */
94 #include <sys/verified_exec.h>
95 #include <sys/kauth.h>
96 #include <sys/atomic.h>
97 #include <sys/module.h>
98 
99 #include <miscfs/genfs/genfs.h>
100 #include <miscfs/syncfs/syncfs.h>
101 #include <miscfs/specfs/specdev.h>
102 
103 #ifdef COMPAT_30
104 #include "opt_nfsserver.h"
105 #include <nfs/rpcv2.h>
106 #endif
107 #include <nfs/nfsproto.h>
108 #ifdef COMPAT_30
109 #include <nfs/nfs.h>
110 #include <nfs/nfs_var.h>
111 #endif
112 
113 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
114 
115 static int change_dir(struct nameidata *, struct lwp *);
116 static int change_flags(struct vnode *, u_long, struct lwp *);
117 static int change_mode(struct vnode *, int, struct lwp *l);
118 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
119 
120 void checkdirs(struct vnode *);
121 
122 int dovfsusermount = 0;
123 
124 /*
125  * Virtual File System System Calls
126  */
127 
128 /*
129  * Mount a file system.
130  */
131 
132 #if defined(COMPAT_09) || defined(COMPAT_43)
133 /*
134  * This table is used to maintain compatibility with 4.3BSD
135  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
136  *
137  * Do not modify this table. It should only contain filesystems
138  * supported by NetBSD 0.9 and 4.3BSD.
139  */
140 const char * const mountcompatnames[] = {
141 	NULL,		/* 0 = MOUNT_NONE */
142 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
143 	MOUNT_NFS,	/* 2 */
144 	MOUNT_MFS,	/* 3 */
145 	MOUNT_MSDOS,	/* 4 */
146 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
147 	MOUNT_FDESC,	/* 6 */
148 	MOUNT_KERNFS,	/* 7 */
149 	NULL,		/* 8 = MOUNT_DEVFS */
150 	MOUNT_AFS,	/* 9 */
151 };
152 const int nmountcompatnames = sizeof(mountcompatnames) /
153     sizeof(mountcompatnames[0]);
154 #endif /* COMPAT_09 || COMPAT_43 */
155 
156 static int
157 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
158     void *data, size_t *data_len)
159 {
160 	struct mount *mp;
161 	int error = 0, saved_flags;
162 
163 	mp = vp->v_mount;
164 	saved_flags = mp->mnt_flag;
165 
166 	/* We can operate only on VV_ROOT nodes. */
167 	if ((vp->v_vflag & VV_ROOT) == 0) {
168 		error = EINVAL;
169 		goto out;
170 	}
171 
172 	/*
173 	 * We only allow the filesystem to be reloaded if it
174 	 * is currently mounted read-only.  Additionally, we
175 	 * prevent read-write to read-only downgrades.
176 	 */
177 	if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
178 	    (mp->mnt_flag & MNT_RDONLY) == 0) {
179 		error = EOPNOTSUPP;	/* Needs translation */
180 		goto out;
181 	}
182 
183 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
184 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
185 	if (error)
186 		goto out;
187 
188 	if (vfs_busy(mp, NULL)) {
189 		error = EPERM;
190 		goto out;
191 	}
192 
193 	mutex_enter(&mp->mnt_updating);
194 
195 	mp->mnt_flag &= ~MNT_OP_FLAGS;
196 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
197 
198 	/*
199 	 * Set the mount level flags.
200 	 */
201 	if (flags & MNT_RDONLY)
202 		mp->mnt_flag |= MNT_RDONLY;
203 	else if (mp->mnt_flag & MNT_RDONLY)
204 		mp->mnt_iflag |= IMNT_WANTRDWR;
205 	mp->mnt_flag &=
206 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
207 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
208 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
209 	    MNT_LOG);
210 	mp->mnt_flag |= flags &
211 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
212 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
213 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
214 	    MNT_LOG | MNT_IGNORE);
215 
216 	error = VFS_MOUNT(mp, path, data, data_len);
217 
218 #if defined(COMPAT_30) && defined(NFSSERVER)
219 	if (error && data != NULL) {
220 		int error2;
221 
222 		/* Update failed; let's try and see if it was an
223 		 * export request. */
224 		error2 = nfs_update_exports_30(mp, path, data, l);
225 
226 		/* Only update error code if the export request was
227 		 * understood but some problem occurred while
228 		 * processing it. */
229 		if (error2 != EJUSTRETURN)
230 			error = error2;
231 	}
232 #endif
233 	if (mp->mnt_iflag & IMNT_WANTRDWR)
234 		mp->mnt_flag &= ~MNT_RDONLY;
235 	if (error)
236 		mp->mnt_flag = saved_flags;
237 	mp->mnt_flag &= ~MNT_OP_FLAGS;
238 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
239 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
240 		if (mp->mnt_syncer == NULL)
241 			error = vfs_allocate_syncvnode(mp);
242 	} else {
243 		if (mp->mnt_syncer != NULL)
244 			vfs_deallocate_syncvnode(mp);
245 	}
246 	mutex_exit(&mp->mnt_updating);
247 	vfs_unbusy(mp, false, NULL);
248 
249  out:
250 	return (error);
251 }
252 
253 static int
254 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
255 {
256 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
257 	int error;
258 
259 	/* Copy file-system type from userspace.  */
260 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
261 	if (error) {
262 #if defined(COMPAT_09) || defined(COMPAT_43)
263 		/*
264 		 * Historically, filesystem types were identified by numbers.
265 		 * If we get an integer for the filesystem type instead of a
266 		 * string, we check to see if it matches one of the historic
267 		 * filesystem types.
268 		 */
269 		u_long fsindex = (u_long)fstype;
270 		if (fsindex >= nmountcompatnames ||
271 		    mountcompatnames[fsindex] == NULL)
272 			return ENODEV;
273 		strlcpy(fstypename, mountcompatnames[fsindex],
274 		    sizeof(fstypename));
275 #else
276 		return error;
277 #endif
278 	}
279 
280 #ifdef	COMPAT_10
281 	/* Accept `ufs' as an alias for `ffs'. */
282 	if (strcmp(fstypename, "ufs") == 0)
283 		fstypename[0] = 'f';
284 #endif
285 
286 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
287 		return 0;
288 
289 	/* If we can autoload a vfs module, try again */
290 	(void)module_load(fstype, 0, NULL, MODULE_CLASS_VFS, true);
291 
292 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
293 		return 0;
294 
295 	return ENODEV;
296 }
297 
298 static int
299 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
300     const char *path, int flags, void *data, size_t *data_len, u_int recurse)
301 {
302 	struct mount *mp;
303 	struct vnode *vp = *vpp;
304 	struct vattr va;
305 	int error;
306 
307 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
308 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
309 	if (error)
310 		return error;
311 
312 	/* Can't make a non-dir a mount-point (from here anyway). */
313 	if (vp->v_type != VDIR)
314 		return ENOTDIR;
315 
316 	/*
317 	 * If the user is not root, ensure that they own the directory
318 	 * onto which we are attempting to mount.
319 	 */
320 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
321 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
322 	    (error = kauth_authorize_generic(l->l_cred,
323 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
324 		return error;
325 	}
326 
327 	if (flags & MNT_EXPORTED)
328 		return EINVAL;
329 
330 	if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
331 		return error;
332 
333 	/*
334 	 * Check if a file-system is not already mounted on this vnode.
335 	 */
336 	if (vp->v_mountedhere != NULL)
337 		return EBUSY;
338 
339 	mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
340 	if (mp == NULL)
341 		return ENOMEM;
342 
343 	mp->mnt_op = vfsops;
344 	mp->mnt_refcnt = 1;
345 
346 	TAILQ_INIT(&mp->mnt_vnodelist);
347 	rw_init(&mp->mnt_unmounting);
348  	mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE);
349 	mutex_init(&mp->mnt_updating, MUTEX_DEFAULT, IPL_NONE);
350 	error = vfs_busy(mp, NULL);
351 	KASSERT(error == 0);
352 	mutex_enter(&mp->mnt_updating);
353 
354 	mp->mnt_vnodecovered = vp;
355 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
356 	mount_initspecific(mp);
357 
358 	/*
359 	 * The underlying file system may refuse the mount for
360 	 * various reasons.  Allow the user to force it to happen.
361 	 *
362 	 * Set the mount level flags.
363 	 */
364 	mp->mnt_flag = flags &
365 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
366 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
367 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
368 	    MNT_LOG | MNT_IGNORE | MNT_RDONLY);
369 
370 	error = VFS_MOUNT(mp, path, data, data_len);
371 	mp->mnt_flag &= ~MNT_OP_FLAGS;
372 
373 	/*
374 	 * Put the new filesystem on the mount list after root.
375 	 */
376 	cache_purge(vp);
377 	if (error != 0) {
378 		vp->v_mountedhere = NULL;
379 		mutex_exit(&mp->mnt_updating);
380 		vfs_unbusy(mp, false, NULL);
381 		vfs_destroy(mp);
382 		return error;
383 	}
384 
385 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
386 	mutex_enter(&mountlist_lock);
387 	vp->v_mountedhere = mp;
388 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
389 	mutex_exit(&mountlist_lock);
390     	vn_restorerecurse(vp, recurse);
391 	VOP_UNLOCK(vp, 0);
392 	checkdirs(vp);
393 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
394 		error = vfs_allocate_syncvnode(mp);
395 	/* Hold an additional reference to the mount across VFS_START(). */
396 	mutex_exit(&mp->mnt_updating);
397 	vfs_unbusy(mp, true, NULL);
398 	(void) VFS_STATVFS(mp, &mp->mnt_stat);
399 	error = VFS_START(mp, 0);
400 	if (error)
401 		vrele(vp);
402 	/* Drop reference held for VFS_START(). */
403 	vfs_destroy(mp);
404 	*vpp = NULL;
405 	return error;
406 }
407 
408 static int
409 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
410     void *data, size_t *data_len)
411 {
412 	struct mount *mp;
413 	int error;
414 
415 	/* If MNT_GETARGS is specified, it should be the only flag. */
416 	if (flags & ~MNT_GETARGS)
417 		return EINVAL;
418 
419 	mp = vp->v_mount;
420 
421 	/* XXX: probably some notion of "can see" here if we want isolation. */
422 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
423 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
424 	if (error)
425 		return error;
426 
427 	if ((vp->v_vflag & VV_ROOT) == 0)
428 		return EINVAL;
429 
430 	if (vfs_busy(mp, NULL))
431 		return EPERM;
432 
433 	mutex_enter(&mp->mnt_updating);
434 	mp->mnt_flag &= ~MNT_OP_FLAGS;
435 	mp->mnt_flag |= MNT_GETARGS;
436 	error = VFS_MOUNT(mp, path, data, data_len);
437 	mp->mnt_flag &= ~MNT_OP_FLAGS;
438 	mutex_exit(&mp->mnt_updating);
439 
440 	vfs_unbusy(mp, false, NULL);
441 	return (error);
442 }
443 
444 #ifdef COMPAT_40
445 /* ARGSUSED */
446 int
447 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval)
448 {
449 	/* {
450 		syscallarg(const char *) type;
451 		syscallarg(const char *) path;
452 		syscallarg(int) flags;
453 		syscallarg(void *) data;
454 	} */
455 	register_t dummy;
456 
457 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
458 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy);
459 }
460 #endif
461 
462 int
463 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
464 {
465 	/* {
466 		syscallarg(const char *) type;
467 		syscallarg(const char *) path;
468 		syscallarg(int) flags;
469 		syscallarg(void *) data;
470 		syscallarg(size_t) data_len;
471 	} */
472 
473 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
474 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
475 	    SCARG(uap, data_len), retval);
476 }
477 
478 int
479 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
480     const char *path, int flags, void *data, enum uio_seg data_seg,
481     size_t data_len, register_t *retval)
482 {
483 	struct vnode *vp;
484 	struct nameidata nd;
485 	void *data_buf = data;
486 	u_int recurse;
487 	int error;
488 
489 	/*
490 	 * Get vnode to be covered
491 	 */
492 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
493 	if ((error = namei(&nd)) != 0)
494 		return (error);
495 	vp = nd.ni_vp;
496 
497 	/*
498 	 * A lookup in VFS_MOUNT might result in an attempt to
499 	 * lock this vnode again, so make the lock recursive.
500 	 */
501 	if (vfsops == NULL) {
502 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
503 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
504 			recurse = vn_setrecurse(vp);
505 			vfsops = vp->v_mount->mnt_op;
506 		} else {
507 			/* 'type' is userspace */
508 			error = mount_get_vfsops(type, &vfsops);
509 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
510 			recurse = vn_setrecurse(vp);
511 			if (error != 0)
512 				goto done;
513 		}
514 	} else {
515 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
516 		recurse = vn_setrecurse(vp);
517 	}
518 
519 	if (data != NULL && data_seg == UIO_USERSPACE) {
520 		if (data_len == 0) {
521 			/* No length supplied, use default for filesystem */
522 			data_len = vfsops->vfs_min_mount_data;
523 			if (data_len > VFS_MAX_MOUNT_DATA) {
524 				/* maybe a force loaded old LKM */
525 				error = EINVAL;
526 				goto done;
527 			}
528 #ifdef COMPAT_30
529 			/* Hopefully a longer buffer won't make copyin() fail */
530 			if (flags & MNT_UPDATE
531 			    && data_len < sizeof (struct mnt_export_args30))
532 				data_len = sizeof (struct mnt_export_args30);
533 #endif
534 		}
535 		data_buf = malloc(data_len, M_TEMP, M_WAITOK);
536 
537 		/* NFS needs the buffer even for mnt_getargs .... */
538 		error = copyin(data, data_buf, data_len);
539 		if (error != 0)
540 			goto done;
541 	}
542 
543 	if (flags & MNT_GETARGS) {
544 		if (data_len == 0) {
545 			error = EINVAL;
546 			goto done;
547 		}
548 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
549 		if (error != 0)
550 			goto done;
551 		if (data_seg == UIO_USERSPACE)
552 			error = copyout(data_buf, data, data_len);
553 		*retval = data_len;
554 	} else if (flags & MNT_UPDATE) {
555 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
556 	} else {
557 		/* Locking is handled internally in mount_domount(). */
558 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
559 		    &data_len, recurse);
560 	}
561 
562     done:
563     	if (vp != NULL) {
564 	    	vn_restorerecurse(vp, recurse);
565 	    	vput(vp);
566 	}
567 	if (data_buf != data)
568 		free(data_buf, M_TEMP);
569 	return (error);
570 }
571 
572 /*
573  * Scan all active processes to see if any of them have a current
574  * or root directory onto which the new filesystem has just been
575  * mounted. If so, replace them with the new mount point.
576  */
577 void
578 checkdirs(struct vnode *olddp)
579 {
580 	struct cwdinfo *cwdi;
581 	struct vnode *newdp, *rele1, *rele2;
582 	struct proc *p;
583 	bool retry;
584 
585 	if (olddp->v_usecount == 1)
586 		return;
587 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
588 		panic("mount: lost mount");
589 
590 	do {
591 		retry = false;
592 		mutex_enter(proc_lock);
593 		PROCLIST_FOREACH(p, &allproc) {
594 			if ((p->p_flag & PK_MARKER) != 0)
595 				continue;
596 			if ((cwdi = p->p_cwdi) == NULL)
597 				continue;
598 			/*
599 			 * Can't change to the old directory any more,
600 			 * so even if we see a stale value it's not a
601 			 * problem.
602 			 */
603 			if (cwdi->cwdi_cdir != olddp &&
604 			    cwdi->cwdi_rdir != olddp)
605 			    	continue;
606 			retry = true;
607 			rele1 = NULL;
608 			rele2 = NULL;
609 			atomic_inc_uint(&cwdi->cwdi_refcnt);
610 			mutex_exit(proc_lock);
611 			rw_enter(&cwdi->cwdi_lock, RW_WRITER);
612 			if (cwdi->cwdi_cdir == olddp) {
613 				rele1 = cwdi->cwdi_cdir;
614 				VREF(newdp);
615 				cwdi->cwdi_cdir = newdp;
616 			}
617 			if (cwdi->cwdi_rdir == olddp) {
618 				rele2 = cwdi->cwdi_rdir;
619 				VREF(newdp);
620 				cwdi->cwdi_rdir = newdp;
621 			}
622 			rw_exit(&cwdi->cwdi_lock);
623 			cwdfree(cwdi);
624 			if (rele1 != NULL)
625 				vrele(rele1);
626 			if (rele2 != NULL)
627 				vrele(rele2);
628 			mutex_enter(proc_lock);
629 			break;
630 		}
631 		mutex_exit(proc_lock);
632 	} while (retry);
633 
634 	if (rootvnode == olddp) {
635 		vrele(rootvnode);
636 		VREF(newdp);
637 		rootvnode = newdp;
638 	}
639 	vput(newdp);
640 }
641 
642 /*
643  * Unmount a file system.
644  *
645  * Note: unmount takes a path to the vnode mounted on as argument,
646  * not special file (as before).
647  */
648 /* ARGSUSED */
649 int
650 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
651 {
652 	/* {
653 		syscallarg(const char *) path;
654 		syscallarg(int) flags;
655 	} */
656 	struct vnode *vp;
657 	struct mount *mp;
658 	int error;
659 	struct nameidata nd;
660 
661 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
662 	    SCARG(uap, path));
663 	if ((error = namei(&nd)) != 0)
664 		return (error);
665 	vp = nd.ni_vp;
666 	mp = vp->v_mount;
667 	atomic_inc_uint(&mp->mnt_refcnt);
668 	VOP_UNLOCK(vp, 0);
669 
670 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
671 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
672 	if (error) {
673 		vrele(vp);
674 		vfs_destroy(mp);
675 		return (error);
676 	}
677 
678 	/*
679 	 * Don't allow unmounting the root file system.
680 	 */
681 	if (mp->mnt_flag & MNT_ROOTFS) {
682 		vrele(vp);
683 		vfs_destroy(mp);
684 		return (EINVAL);
685 	}
686 
687 	/*
688 	 * Must be the root of the filesystem
689 	 */
690 	if ((vp->v_vflag & VV_ROOT) == 0) {
691 		vrele(vp);
692 		vfs_destroy(mp);
693 		return (EINVAL);
694 	}
695 
696 	vrele(vp);
697 	error = dounmount(mp, SCARG(uap, flags), l);
698 	vfs_destroy(mp);
699 	return error;
700 }
701 
702 /*
703  * Do the actual file system unmount.  File system is assumed to have
704  * been locked by the caller.
705  *
706  * => Caller hold reference to the mount, explicitly for dounmount().
707  */
708 int
709 dounmount(struct mount *mp, int flags, struct lwp *l)
710 {
711 	struct vnode *coveredvp;
712 	int error;
713 	int async;
714 	int used_syncer;
715 
716 #if NVERIEXEC > 0
717 	error = veriexec_unmountchk(mp);
718 	if (error)
719 		return (error);
720 #endif /* NVERIEXEC > 0 */
721 
722 	/*
723 	 * XXX Freeze syncer.  Must do this before locking the
724 	 * mount point.  See dounmount() for details.
725 	 */
726 	mutex_enter(&syncer_mutex);
727 	rw_enter(&mp->mnt_unmounting, RW_WRITER);
728 	if ((mp->mnt_iflag & IMNT_GONE) != 0) {
729 		rw_exit(&mp->mnt_unmounting);
730 		mutex_exit(&syncer_mutex);
731 		return ENOENT;
732 	}
733 
734 	used_syncer = (mp->mnt_syncer != NULL);
735 
736 	/*
737 	 * XXX Syncer must be frozen when we get here.  This should really
738 	 * be done on a per-mountpoint basis, but especially the softdep
739 	 * code possibly called from the syncer doesn't exactly work on a
740 	 * per-mountpoint basis, so the softdep code would become a maze
741 	 * of vfs_busy() calls.
742 	 *
743 	 * The caller of dounmount() must acquire syncer_mutex because
744 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
745 	 * order, and we must preserve that order to avoid deadlock.
746 	 *
747 	 * So, if the file system did not use the syncer, now is
748 	 * the time to release the syncer_mutex.
749 	 */
750 	if (used_syncer == 0)
751 		mutex_exit(&syncer_mutex);
752 
753 	mp->mnt_iflag |= IMNT_UNMOUNT;
754 	async = mp->mnt_flag & MNT_ASYNC;
755 	mp->mnt_flag &= ~MNT_ASYNC;
756 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
757 	if (mp->mnt_syncer != NULL)
758 		vfs_deallocate_syncvnode(mp);
759 	error = 0;
760 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
761 		error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
762 	}
763 	vfs_scrubvnlist(mp);
764 	if (error == 0 || (flags & MNT_FORCE))
765 		error = VFS_UNMOUNT(mp, flags);
766 	if (error) {
767 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
768 			(void) vfs_allocate_syncvnode(mp);
769 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
770 		mp->mnt_flag |= async;
771 		rw_exit(&mp->mnt_unmounting);
772 		if (used_syncer)
773 			mutex_exit(&syncer_mutex);
774 		return (error);
775 	}
776 	vfs_scrubvnlist(mp);
777 	mutex_enter(&mountlist_lock);
778 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
779 		coveredvp->v_mountedhere = NULL;
780 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
781 	mp->mnt_iflag |= IMNT_GONE;
782 	mutex_exit(&mountlist_lock);
783 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
784 		panic("unmount: dangling vnode");
785 	if (used_syncer)
786 		mutex_exit(&syncer_mutex);
787 	vfs_hooks_unmount(mp);
788 	rw_exit(&mp->mnt_unmounting);
789 	vfs_destroy(mp);	/* reference from mount() */
790 	if (coveredvp != NULLVP)
791 		vrele(coveredvp);
792 	return (0);
793 }
794 
795 /*
796  * Sync each mounted filesystem.
797  */
798 #ifdef DEBUG
799 int syncprt = 0;
800 struct ctldebug debug0 = { "syncprt", &syncprt };
801 #endif
802 
803 /* ARGSUSED */
804 int
805 sys_sync(struct lwp *l, const void *v, register_t *retval)
806 {
807 	struct mount *mp, *nmp;
808 	int asyncflag;
809 
810 	if (l == NULL)
811 		l = &lwp0;
812 
813 	mutex_enter(&mountlist_lock);
814 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
815 	     mp = nmp) {
816 		if (vfs_busy(mp, &nmp)) {
817 			continue;
818 		}
819 		mutex_enter(&mp->mnt_updating);
820 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
821 			asyncflag = mp->mnt_flag & MNT_ASYNC;
822 			mp->mnt_flag &= ~MNT_ASYNC;
823 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
824 			if (asyncflag)
825 				 mp->mnt_flag |= MNT_ASYNC;
826 		}
827 		mutex_exit(&mp->mnt_updating);
828 		vfs_unbusy(mp, false, &nmp);
829 	}
830 	mutex_exit(&mountlist_lock);
831 #ifdef DEBUG
832 	if (syncprt)
833 		vfs_bufstats();
834 #endif /* DEBUG */
835 	return (0);
836 }
837 
838 /*
839  * Change filesystem quotas.
840  */
841 /* ARGSUSED */
842 int
843 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
844 {
845 	/* {
846 		syscallarg(const char *) path;
847 		syscallarg(int) cmd;
848 		syscallarg(int) uid;
849 		syscallarg(void *) arg;
850 	} */
851 	struct mount *mp;
852 	int error;
853 	struct nameidata nd;
854 
855 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
856 	    SCARG(uap, path));
857 	if ((error = namei(&nd)) != 0)
858 		return (error);
859 	mp = nd.ni_vp->v_mount;
860 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
861 	    SCARG(uap, arg));
862 	vrele(nd.ni_vp);
863 	return (error);
864 }
865 
866 int
867 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
868     int root)
869 {
870 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
871 	int error = 0;
872 
873 	/*
874 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
875 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
876 	 * overrides MNT_NOWAIT.
877 	 */
878 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
879 	    (flags != MNT_WAIT && flags != 0)) {
880 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
881 		goto done;
882 	}
883 
884 	/* Get the filesystem stats now */
885 	memset(sp, 0, sizeof(*sp));
886 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
887 		return error;
888 	}
889 
890 	if (cwdi->cwdi_rdir == NULL)
891 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
892 done:
893 	if (cwdi->cwdi_rdir != NULL) {
894 		size_t len;
895 		char *bp;
896 		char c;
897 		char *path = PNBUF_GET();
898 
899 		bp = path + MAXPATHLEN;
900 		*--bp = '\0';
901 		rw_enter(&cwdi->cwdi_lock, RW_READER);
902 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
903 		    MAXPATHLEN / 2, 0, l);
904 		rw_exit(&cwdi->cwdi_lock);
905 		if (error) {
906 			PNBUF_PUT(path);
907 			return error;
908 		}
909 		len = strlen(bp);
910 		/*
911 		 * for mount points that are below our root, we can see
912 		 * them, so we fix up the pathname and return them. The
913 		 * rest we cannot see, so we don't allow viewing the
914 		 * data.
915 		 */
916 		if (strncmp(bp, sp->f_mntonname, len) == 0 &&
917 		    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
918 			(void)strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
919 			    sizeof(sp->f_mntonname));
920 			if (sp->f_mntonname[0] == '\0')
921 				(void)strlcpy(sp->f_mntonname, "/",
922 				    sizeof(sp->f_mntonname));
923 		} else {
924 			if (root)
925 				(void)strlcpy(sp->f_mntonname, "/",
926 				    sizeof(sp->f_mntonname));
927 			else
928 				error = EPERM;
929 		}
930 		PNBUF_PUT(path);
931 	}
932 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
933 	return error;
934 }
935 
936 /*
937  * Get filesystem statistics by path.
938  */
939 int
940 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
941 {
942 	struct mount *mp;
943 	int error;
944 	struct nameidata nd;
945 
946 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
947 	if ((error = namei(&nd)) != 0)
948 		return error;
949 	mp = nd.ni_vp->v_mount;
950 	error = dostatvfs(mp, sb, l, flags, 1);
951 	vrele(nd.ni_vp);
952 	return error;
953 }
954 
955 /* ARGSUSED */
956 int
957 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
958 {
959 	/* {
960 		syscallarg(const char *) path;
961 		syscallarg(struct statvfs *) buf;
962 		syscallarg(int) flags;
963 	} */
964 	struct statvfs *sb;
965 	int error;
966 
967 	sb = STATVFSBUF_GET();
968 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
969 	if (error == 0)
970 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
971 	STATVFSBUF_PUT(sb);
972 	return error;
973 }
974 
975 /*
976  * Get filesystem statistics by fd.
977  */
978 int
979 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
980 {
981 	file_t *fp;
982 	struct mount *mp;
983 	int error;
984 
985 	/* fd_getvnode() will use the descriptor for us */
986 	if ((error = fd_getvnode(fd, &fp)) != 0)
987 		return (error);
988 	mp = ((struct vnode *)fp->f_data)->v_mount;
989 	error = dostatvfs(mp, sb, curlwp, flags, 1);
990 	fd_putfile(fd);
991 	return error;
992 }
993 
994 /* ARGSUSED */
995 int
996 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
997 {
998 	/* {
999 		syscallarg(int) fd;
1000 		syscallarg(struct statvfs *) buf;
1001 		syscallarg(int) flags;
1002 	} */
1003 	struct statvfs *sb;
1004 	int error;
1005 
1006 	sb = STATVFSBUF_GET();
1007 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1008 	if (error == 0)
1009 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1010 	STATVFSBUF_PUT(sb);
1011 	return error;
1012 }
1013 
1014 
1015 /*
1016  * Get statistics on all filesystems.
1017  */
1018 int
1019 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1020     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1021     register_t *retval)
1022 {
1023 	int root = 0;
1024 	struct proc *p = l->l_proc;
1025 	struct mount *mp, *nmp;
1026 	struct statvfs *sb;
1027 	size_t count, maxcount;
1028 	int error = 0;
1029 
1030 	sb = STATVFSBUF_GET();
1031 	maxcount = bufsize / entry_sz;
1032 	mutex_enter(&mountlist_lock);
1033 	count = 0;
1034 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1035 	     mp = nmp) {
1036 		if (vfs_busy(mp, &nmp)) {
1037 			continue;
1038 		}
1039 		if (sfsp && count < maxcount) {
1040 			error = dostatvfs(mp, sb, l, flags, 0);
1041 			if (error) {
1042 				vfs_unbusy(mp, false, &nmp);
1043 				error = 0;
1044 				continue;
1045 			}
1046 			error = copyfn(sb, sfsp, entry_sz);
1047 			if (error) {
1048 				vfs_unbusy(mp, false, NULL);
1049 				goto out;
1050 			}
1051 			sfsp = (char *)sfsp + entry_sz;
1052 			root |= strcmp(sb->f_mntonname, "/") == 0;
1053 		}
1054 		count++;
1055 		vfs_unbusy(mp, false, &nmp);
1056 	}
1057 	mutex_exit(&mountlist_lock);
1058 
1059 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
1060 		/*
1061 		 * fake a root entry
1062 		 */
1063 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1064 		    sb, l, flags, 1);
1065 		if (error != 0)
1066 			goto out;
1067 		if (sfsp) {
1068 			error = copyfn(sb, sfsp, entry_sz);
1069 			if (error != 0)
1070 				goto out;
1071 		}
1072 		count++;
1073 	}
1074 	if (sfsp && count > maxcount)
1075 		*retval = maxcount;
1076 	else
1077 		*retval = count;
1078 out:
1079 	STATVFSBUF_PUT(sb);
1080 	return error;
1081 }
1082 
1083 int
1084 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1085 {
1086 	/* {
1087 		syscallarg(struct statvfs *) buf;
1088 		syscallarg(size_t) bufsize;
1089 		syscallarg(int) flags;
1090 	} */
1091 
1092 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1093 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1094 }
1095 
1096 /*
1097  * Change current working directory to a given file descriptor.
1098  */
1099 /* ARGSUSED */
1100 int
1101 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1102 {
1103 	/* {
1104 		syscallarg(int) fd;
1105 	} */
1106 	struct proc *p = l->l_proc;
1107 	struct cwdinfo *cwdi;
1108 	struct vnode *vp, *tdp;
1109 	struct mount *mp;
1110 	file_t *fp;
1111 	int error, fd;
1112 
1113 	/* fd_getvnode() will use the descriptor for us */
1114 	fd = SCARG(uap, fd);
1115 	if ((error = fd_getvnode(fd, &fp)) != 0)
1116 		return (error);
1117 	vp = fp->f_data;
1118 
1119 	VREF(vp);
1120 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1121 	if (vp->v_type != VDIR)
1122 		error = ENOTDIR;
1123 	else
1124 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1125 	if (error) {
1126 		vput(vp);
1127 		goto out;
1128 	}
1129 	while ((mp = vp->v_mountedhere) != NULL) {
1130 		error = vfs_busy(mp, NULL);
1131 		vput(vp);
1132 		if (error != 0)
1133 			goto out;
1134 		error = VFS_ROOT(mp, &tdp);
1135 		vfs_unbusy(mp, false, NULL);
1136 		if (error)
1137 			goto out;
1138 		vp = tdp;
1139 	}
1140 	VOP_UNLOCK(vp, 0);
1141 
1142 	/*
1143 	 * Disallow changing to a directory not under the process's
1144 	 * current root directory (if there is one).
1145 	 */
1146 	cwdi = p->p_cwdi;
1147 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1148 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1149 		vrele(vp);
1150 		error = EPERM;	/* operation not permitted */
1151 	} else {
1152 		vrele(cwdi->cwdi_cdir);
1153 		cwdi->cwdi_cdir = vp;
1154 	}
1155 	rw_exit(&cwdi->cwdi_lock);
1156 
1157  out:
1158 	fd_putfile(fd);
1159 	return (error);
1160 }
1161 
1162 /*
1163  * Change this process's notion of the root directory to a given file
1164  * descriptor.
1165  */
1166 int
1167 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1168 {
1169 	struct proc *p = l->l_proc;
1170 	struct cwdinfo *cwdi;
1171 	struct vnode	*vp;
1172 	file_t	*fp;
1173 	int		 error, fd = SCARG(uap, fd);
1174 
1175 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1176  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1177 		return error;
1178 	/* fd_getvnode() will use the descriptor for us */
1179 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
1180 		return error;
1181 	vp = fp->f_data;
1182 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1183 	if (vp->v_type != VDIR)
1184 		error = ENOTDIR;
1185 	else
1186 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1187 	VOP_UNLOCK(vp, 0);
1188 	if (error)
1189 		goto out;
1190 	VREF(vp);
1191 
1192 	/*
1193 	 * Prevent escaping from chroot by putting the root under
1194 	 * the working directory.  Silently chdir to / if we aren't
1195 	 * already there.
1196 	 */
1197 	cwdi = p->p_cwdi;
1198 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1199 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1200 		/*
1201 		 * XXX would be more failsafe to change directory to a
1202 		 * deadfs node here instead
1203 		 */
1204 		vrele(cwdi->cwdi_cdir);
1205 		VREF(vp);
1206 		cwdi->cwdi_cdir = vp;
1207 	}
1208 
1209 	if (cwdi->cwdi_rdir != NULL)
1210 		vrele(cwdi->cwdi_rdir);
1211 	cwdi->cwdi_rdir = vp;
1212 	rw_exit(&cwdi->cwdi_lock);
1213 
1214  out:
1215 	fd_putfile(fd);
1216 	return (error);
1217 }
1218 
1219 /*
1220  * Change current working directory (``.'').
1221  */
1222 /* ARGSUSED */
1223 int
1224 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1225 {
1226 	/* {
1227 		syscallarg(const char *) path;
1228 	} */
1229 	struct proc *p = l->l_proc;
1230 	struct cwdinfo *cwdi;
1231 	int error;
1232 	struct nameidata nd;
1233 
1234 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1235 	    SCARG(uap, path));
1236 	if ((error = change_dir(&nd, l)) != 0)
1237 		return (error);
1238 	cwdi = p->p_cwdi;
1239 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1240 	vrele(cwdi->cwdi_cdir);
1241 	cwdi->cwdi_cdir = nd.ni_vp;
1242 	rw_exit(&cwdi->cwdi_lock);
1243 	return (0);
1244 }
1245 
1246 /*
1247  * Change notion of root (``/'') directory.
1248  */
1249 /* ARGSUSED */
1250 int
1251 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1252 {
1253 	/* {
1254 		syscallarg(const char *) path;
1255 	} */
1256 	struct proc *p = l->l_proc;
1257 	struct cwdinfo *cwdi;
1258 	struct vnode *vp;
1259 	int error;
1260 	struct nameidata nd;
1261 
1262 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1263 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1264 		return (error);
1265 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1266 	    SCARG(uap, path));
1267 	if ((error = change_dir(&nd, l)) != 0)
1268 		return (error);
1269 
1270 	cwdi = p->p_cwdi;
1271 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1272 	if (cwdi->cwdi_rdir != NULL)
1273 		vrele(cwdi->cwdi_rdir);
1274 	vp = nd.ni_vp;
1275 	cwdi->cwdi_rdir = vp;
1276 
1277 	/*
1278 	 * Prevent escaping from chroot by putting the root under
1279 	 * the working directory.  Silently chdir to / if we aren't
1280 	 * already there.
1281 	 */
1282 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1283 		/*
1284 		 * XXX would be more failsafe to change directory to a
1285 		 * deadfs node here instead
1286 		 */
1287 		vrele(cwdi->cwdi_cdir);
1288 		VREF(vp);
1289 		cwdi->cwdi_cdir = vp;
1290 	}
1291 	rw_exit(&cwdi->cwdi_lock);
1292 
1293 	return (0);
1294 }
1295 
1296 /*
1297  * Common routine for chroot and chdir.
1298  */
1299 static int
1300 change_dir(struct nameidata *ndp, struct lwp *l)
1301 {
1302 	struct vnode *vp;
1303 	int error;
1304 
1305 	if ((error = namei(ndp)) != 0)
1306 		return (error);
1307 	vp = ndp->ni_vp;
1308 	if (vp->v_type != VDIR)
1309 		error = ENOTDIR;
1310 	else
1311 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1312 
1313 	if (error)
1314 		vput(vp);
1315 	else
1316 		VOP_UNLOCK(vp, 0);
1317 	return (error);
1318 }
1319 
1320 /*
1321  * Check permissions, allocate an open file structure,
1322  * and call the device open routine if any.
1323  */
1324 int
1325 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1326 {
1327 	/* {
1328 		syscallarg(const char *) path;
1329 		syscallarg(int) flags;
1330 		syscallarg(int) mode;
1331 	} */
1332 	struct proc *p = l->l_proc;
1333 	struct cwdinfo *cwdi = p->p_cwdi;
1334 	file_t *fp;
1335 	struct vnode *vp;
1336 	int flags, cmode;
1337 	int type, indx, error;
1338 	struct flock lf;
1339 	struct nameidata nd;
1340 
1341 	flags = FFLAGS(SCARG(uap, flags));
1342 	if ((flags & (FREAD | FWRITE)) == 0)
1343 		return (EINVAL);
1344 	if ((error = fd_allocfile(&fp, &indx)) != 0)
1345 		return (error);
1346 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1347 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1348 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1349 	    SCARG(uap, path));
1350 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1351 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1352 		fd_abort(p, fp, indx);
1353 		if ((error == EDUPFD || error == EMOVEFD) &&
1354 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1355 		    (error =
1356 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1357 			*retval = indx;
1358 			return (0);
1359 		}
1360 		if (error == ERESTART)
1361 			error = EINTR;
1362 		return (error);
1363 	}
1364 
1365 	l->l_dupfd = 0;
1366 	vp = nd.ni_vp;
1367 	fp->f_flag = flags & FMASK;
1368 	fp->f_type = DTYPE_VNODE;
1369 	fp->f_ops = &vnops;
1370 	fp->f_data = vp;
1371 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1372 		lf.l_whence = SEEK_SET;
1373 		lf.l_start = 0;
1374 		lf.l_len = 0;
1375 		if (flags & O_EXLOCK)
1376 			lf.l_type = F_WRLCK;
1377 		else
1378 			lf.l_type = F_RDLCK;
1379 		type = F_FLOCK;
1380 		if ((flags & FNONBLOCK) == 0)
1381 			type |= F_WAIT;
1382 		VOP_UNLOCK(vp, 0);
1383 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1384 		if (error) {
1385 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1386 			fd_abort(p, fp, indx);
1387 			return (error);
1388 		}
1389 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1390 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1391 	}
1392 	VOP_UNLOCK(vp, 0);
1393 	*retval = indx;
1394 	fd_affix(p, fp, indx);
1395 	return (0);
1396 }
1397 
1398 static void
1399 vfs__fhfree(fhandle_t *fhp)
1400 {
1401 	size_t fhsize;
1402 
1403 	if (fhp == NULL) {
1404 		return;
1405 	}
1406 	fhsize = FHANDLE_SIZE(fhp);
1407 	kmem_free(fhp, fhsize);
1408 }
1409 
1410 /*
1411  * vfs_composefh: compose a filehandle.
1412  */
1413 
1414 int
1415 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1416 {
1417 	struct mount *mp;
1418 	struct fid *fidp;
1419 	int error;
1420 	size_t needfhsize;
1421 	size_t fidsize;
1422 
1423 	mp = vp->v_mount;
1424 	fidp = NULL;
1425 	if (*fh_size < FHANDLE_SIZE_MIN) {
1426 		fidsize = 0;
1427 	} else {
1428 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1429 		if (fhp != NULL) {
1430 			memset(fhp, 0, *fh_size);
1431 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1432 			fidp = &fhp->fh_fid;
1433 		}
1434 	}
1435 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1436 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1437 	if (error == 0 && *fh_size < needfhsize) {
1438 		error = E2BIG;
1439 	}
1440 	*fh_size = needfhsize;
1441 	return error;
1442 }
1443 
1444 int
1445 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1446 {
1447 	struct mount *mp;
1448 	fhandle_t *fhp;
1449 	size_t fhsize;
1450 	size_t fidsize;
1451 	int error;
1452 
1453 	*fhpp = NULL;
1454 	mp = vp->v_mount;
1455 	fidsize = 0;
1456 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1457 	KASSERT(error != 0);
1458 	if (error != E2BIG) {
1459 		goto out;
1460 	}
1461 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1462 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1463 	if (fhp == NULL) {
1464 		error = ENOMEM;
1465 		goto out;
1466 	}
1467 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1468 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1469 	if (error == 0) {
1470 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1471 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1472 		*fhpp = fhp;
1473 	} else {
1474 		kmem_free(fhp, fhsize);
1475 	}
1476 out:
1477 	return error;
1478 }
1479 
1480 void
1481 vfs_composefh_free(fhandle_t *fhp)
1482 {
1483 
1484 	vfs__fhfree(fhp);
1485 }
1486 
1487 /*
1488  * vfs_fhtovp: lookup a vnode by a filehandle.
1489  */
1490 
1491 int
1492 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1493 {
1494 	struct mount *mp;
1495 	int error;
1496 
1497 	*vpp = NULL;
1498 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1499 	if (mp == NULL) {
1500 		error = ESTALE;
1501 		goto out;
1502 	}
1503 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1504 		error = EOPNOTSUPP;
1505 		goto out;
1506 	}
1507 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1508 out:
1509 	return error;
1510 }
1511 
1512 /*
1513  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1514  * the needed size.
1515  */
1516 
1517 int
1518 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1519 {
1520 	fhandle_t *fhp;
1521 	int error;
1522 
1523 	*fhpp = NULL;
1524 	if (fhsize > FHANDLE_SIZE_MAX) {
1525 		return EINVAL;
1526 	}
1527 	if (fhsize < FHANDLE_SIZE_MIN) {
1528 		return EINVAL;
1529 	}
1530 again:
1531 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1532 	if (fhp == NULL) {
1533 		return ENOMEM;
1534 	}
1535 	error = copyin(ufhp, fhp, fhsize);
1536 	if (error == 0) {
1537 		/* XXX this check shouldn't be here */
1538 		if (FHANDLE_SIZE(fhp) == fhsize) {
1539 			*fhpp = fhp;
1540 			return 0;
1541 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1542 			/*
1543 			 * a kludge for nfsv2 padded handles.
1544 			 */
1545 			size_t sz;
1546 
1547 			sz = FHANDLE_SIZE(fhp);
1548 			kmem_free(fhp, fhsize);
1549 			fhsize = sz;
1550 			goto again;
1551 		} else {
1552 			/*
1553 			 * userland told us wrong size.
1554 			 */
1555 		    	error = EINVAL;
1556 		}
1557 	}
1558 	kmem_free(fhp, fhsize);
1559 	return error;
1560 }
1561 
1562 void
1563 vfs_copyinfh_free(fhandle_t *fhp)
1564 {
1565 
1566 	vfs__fhfree(fhp);
1567 }
1568 
1569 /*
1570  * Get file handle system call
1571  */
1572 int
1573 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1574 {
1575 	/* {
1576 		syscallarg(char *) fname;
1577 		syscallarg(fhandle_t *) fhp;
1578 		syscallarg(size_t *) fh_size;
1579 	} */
1580 	struct vnode *vp;
1581 	fhandle_t *fh;
1582 	int error;
1583 	struct nameidata nd;
1584 	size_t sz;
1585 	size_t usz;
1586 
1587 	/*
1588 	 * Must be super user
1589 	 */
1590 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1591 	    0, NULL, NULL, NULL);
1592 	if (error)
1593 		return (error);
1594 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1595 	    SCARG(uap, fname));
1596 	error = namei(&nd);
1597 	if (error)
1598 		return (error);
1599 	vp = nd.ni_vp;
1600 	error = vfs_composefh_alloc(vp, &fh);
1601 	vput(vp);
1602 	if (error != 0) {
1603 		goto out;
1604 	}
1605 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1606 	if (error != 0) {
1607 		goto out;
1608 	}
1609 	sz = FHANDLE_SIZE(fh);
1610 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1611 	if (error != 0) {
1612 		goto out;
1613 	}
1614 	if (usz >= sz) {
1615 		error = copyout(fh, SCARG(uap, fhp), sz);
1616 	} else {
1617 		error = E2BIG;
1618 	}
1619 out:
1620 	vfs_composefh_free(fh);
1621 	return (error);
1622 }
1623 
1624 /*
1625  * Open a file given a file handle.
1626  *
1627  * Check permissions, allocate an open file structure,
1628  * and call the device open routine if any.
1629  */
1630 
1631 int
1632 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1633     register_t *retval)
1634 {
1635 	file_t *fp;
1636 	struct vnode *vp = NULL;
1637 	kauth_cred_t cred = l->l_cred;
1638 	file_t *nfp;
1639 	int type, indx, error=0;
1640 	struct flock lf;
1641 	struct vattr va;
1642 	fhandle_t *fh;
1643 	int flags;
1644 	proc_t *p;
1645 
1646 	p = curproc;
1647 
1648 	/*
1649 	 * Must be super user
1650 	 */
1651 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1652 	    0, NULL, NULL, NULL)))
1653 		return (error);
1654 
1655 	flags = FFLAGS(oflags);
1656 	if ((flags & (FREAD | FWRITE)) == 0)
1657 		return (EINVAL);
1658 	if ((flags & O_CREAT))
1659 		return (EINVAL);
1660 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
1661 		return (error);
1662 	fp = nfp;
1663 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1664 	if (error != 0) {
1665 		goto bad;
1666 	}
1667 	error = vfs_fhtovp(fh, &vp);
1668 	if (error != 0) {
1669 		goto bad;
1670 	}
1671 
1672 	/* Now do an effective vn_open */
1673 
1674 	if (vp->v_type == VSOCK) {
1675 		error = EOPNOTSUPP;
1676 		goto bad;
1677 	}
1678 	error = vn_openchk(vp, cred, flags);
1679 	if (error != 0)
1680 		goto bad;
1681 	if (flags & O_TRUNC) {
1682 		VOP_UNLOCK(vp, 0);			/* XXX */
1683 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1684 		VATTR_NULL(&va);
1685 		va.va_size = 0;
1686 		error = VOP_SETATTR(vp, &va, cred);
1687 		if (error)
1688 			goto bad;
1689 	}
1690 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1691 		goto bad;
1692 	if (flags & FWRITE) {
1693 		mutex_enter(&vp->v_interlock);
1694 		vp->v_writecount++;
1695 		mutex_exit(&vp->v_interlock);
1696 	}
1697 
1698 	/* done with modified vn_open, now finish what sys_open does. */
1699 
1700 	fp->f_flag = flags & FMASK;
1701 	fp->f_type = DTYPE_VNODE;
1702 	fp->f_ops = &vnops;
1703 	fp->f_data = vp;
1704 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1705 		lf.l_whence = SEEK_SET;
1706 		lf.l_start = 0;
1707 		lf.l_len = 0;
1708 		if (flags & O_EXLOCK)
1709 			lf.l_type = F_WRLCK;
1710 		else
1711 			lf.l_type = F_RDLCK;
1712 		type = F_FLOCK;
1713 		if ((flags & FNONBLOCK) == 0)
1714 			type |= F_WAIT;
1715 		VOP_UNLOCK(vp, 0);
1716 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1717 		if (error) {
1718 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
1719 			fd_abort(p, fp, indx);
1720 			return (error);
1721 		}
1722 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1723 		atomic_or_uint(&fp->f_flag, FHASLOCK);
1724 	}
1725 	VOP_UNLOCK(vp, 0);
1726 	*retval = indx;
1727 	fd_affix(p, fp, indx);
1728 	vfs_copyinfh_free(fh);
1729 	return (0);
1730 
1731 bad:
1732 	fd_abort(p, fp, indx);
1733 	if (vp != NULL)
1734 		vput(vp);
1735 	vfs_copyinfh_free(fh);
1736 	return (error);
1737 }
1738 
1739 int
1740 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1741 {
1742 	/* {
1743 		syscallarg(const void *) fhp;
1744 		syscallarg(size_t) fh_size;
1745 		syscallarg(int) flags;
1746 	} */
1747 
1748 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1749 	    SCARG(uap, flags), retval);
1750 }
1751 
1752 int
1753 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1754 {
1755 	int error;
1756 	fhandle_t *fh;
1757 	struct vnode *vp;
1758 
1759 	/*
1760 	 * Must be super user
1761 	 */
1762 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1763 	    0, NULL, NULL, NULL)))
1764 		return (error);
1765 
1766 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1767 	if (error != 0)
1768 		return error;
1769 
1770 	error = vfs_fhtovp(fh, &vp);
1771 	vfs_copyinfh_free(fh);
1772 	if (error != 0)
1773 		return error;
1774 
1775 	error = vn_stat(vp, sb);
1776 	vput(vp);
1777 	return error;
1778 }
1779 
1780 
1781 /* ARGSUSED */
1782 int
1783 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval)
1784 {
1785 	/* {
1786 		syscallarg(const void *) fhp;
1787 		syscallarg(size_t) fh_size;
1788 		syscallarg(struct stat *) sb;
1789 	} */
1790 	struct stat sb;
1791 	int error;
1792 
1793 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1794 	if (error)
1795 		return error;
1796 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1797 }
1798 
1799 int
1800 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1801     int flags)
1802 {
1803 	fhandle_t *fh;
1804 	struct mount *mp;
1805 	struct vnode *vp;
1806 	int error;
1807 
1808 	/*
1809 	 * Must be super user
1810 	 */
1811 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1812 	    0, NULL, NULL, NULL)))
1813 		return error;
1814 
1815 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1816 	if (error != 0)
1817 		return error;
1818 
1819 	error = vfs_fhtovp(fh, &vp);
1820 	vfs_copyinfh_free(fh);
1821 	if (error != 0)
1822 		return error;
1823 
1824 	mp = vp->v_mount;
1825 	error = dostatvfs(mp, sb, l, flags, 1);
1826 	vput(vp);
1827 	return error;
1828 }
1829 
1830 /* ARGSUSED */
1831 int
1832 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
1833 {
1834 	/* {
1835 		syscallarg(const void *) fhp;
1836 		syscallarg(size_t) fh_size;
1837 		syscallarg(struct statvfs *) buf;
1838 		syscallarg(int)	flags;
1839 	} */
1840 	struct statvfs *sb = STATVFSBUF_GET();
1841 	int error;
1842 
1843 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1844 	    SCARG(uap, flags));
1845 	if (error == 0)
1846 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1847 	STATVFSBUF_PUT(sb);
1848 	return error;
1849 }
1850 
1851 /*
1852  * Create a special file.
1853  */
1854 /* ARGSUSED */
1855 int
1856 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval)
1857 {
1858 	/* {
1859 		syscallarg(const char *) path;
1860 		syscallarg(int) mode;
1861 		syscallarg(int) dev;
1862 	} */
1863 	struct proc *p = l->l_proc;
1864 	struct vnode *vp;
1865 	struct vattr vattr;
1866 	int error, optype;
1867 	struct nameidata nd;
1868 	char *path;
1869 	const char *cpath;
1870 	enum uio_seg seg = UIO_USERSPACE;
1871 
1872 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1873 	    0, NULL, NULL, NULL)) != 0)
1874 		return (error);
1875 
1876 	optype = VOP_MKNOD_DESCOFFSET;
1877 
1878 	VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path);
1879 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
1880 
1881 	if ((error = namei(&nd)) != 0)
1882 		goto out;
1883 	vp = nd.ni_vp;
1884 	if (vp != NULL)
1885 		error = EEXIST;
1886 	else {
1887 		VATTR_NULL(&vattr);
1888 		/* We will read cwdi->cwdi_cmask unlocked. */
1889 		vattr.va_mode =
1890 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1891 		vattr.va_rdev = SCARG(uap, dev);
1892 
1893 		switch (SCARG(uap, mode) & S_IFMT) {
1894 		case S_IFMT:	/* used by badsect to flag bad sectors */
1895 			vattr.va_type = VBAD;
1896 			break;
1897 		case S_IFCHR:
1898 			vattr.va_type = VCHR;
1899 			break;
1900 		case S_IFBLK:
1901 			vattr.va_type = VBLK;
1902 			break;
1903 		case S_IFWHT:
1904 			optype = VOP_WHITEOUT_DESCOFFSET;
1905 			break;
1906 		case S_IFREG:
1907 #if NVERIEXEC > 0
1908 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1909 			    O_CREAT);
1910 #endif /* NVERIEXEC > 0 */
1911 			vattr.va_type = VREG;
1912 			vattr.va_rdev = VNOVAL;
1913 			optype = VOP_CREATE_DESCOFFSET;
1914 			break;
1915 		default:
1916 			error = EINVAL;
1917 			break;
1918 		}
1919 	}
1920 	if (!error) {
1921 		switch (optype) {
1922 		case VOP_WHITEOUT_DESCOFFSET:
1923 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1924 			if (error)
1925 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1926 			vput(nd.ni_dvp);
1927 			break;
1928 
1929 		case VOP_MKNOD_DESCOFFSET:
1930 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1931 						&nd.ni_cnd, &vattr);
1932 			if (error == 0)
1933 				vput(nd.ni_vp);
1934 			break;
1935 
1936 		case VOP_CREATE_DESCOFFSET:
1937 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1938 						&nd.ni_cnd, &vattr);
1939 			if (error == 0)
1940 				vput(nd.ni_vp);
1941 			break;
1942 		}
1943 	} else {
1944 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1945 		if (nd.ni_dvp == vp)
1946 			vrele(nd.ni_dvp);
1947 		else
1948 			vput(nd.ni_dvp);
1949 		if (vp)
1950 			vrele(vp);
1951 	}
1952 out:
1953 	VERIEXEC_PATH_PUT(path);
1954 	return (error);
1955 }
1956 
1957 /*
1958  * Create a named pipe.
1959  */
1960 /* ARGSUSED */
1961 int
1962 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
1963 {
1964 	/* {
1965 		syscallarg(const char *) path;
1966 		syscallarg(int) mode;
1967 	} */
1968 	struct proc *p = l->l_proc;
1969 	struct vattr vattr;
1970 	int error;
1971 	struct nameidata nd;
1972 
1973 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1974 	    SCARG(uap, path));
1975 	if ((error = namei(&nd)) != 0)
1976 		return (error);
1977 	if (nd.ni_vp != NULL) {
1978 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1979 		if (nd.ni_dvp == nd.ni_vp)
1980 			vrele(nd.ni_dvp);
1981 		else
1982 			vput(nd.ni_dvp);
1983 		vrele(nd.ni_vp);
1984 		return (EEXIST);
1985 	}
1986 	VATTR_NULL(&vattr);
1987 	vattr.va_type = VFIFO;
1988 	/* We will read cwdi->cwdi_cmask unlocked. */
1989 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1990 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1991 	if (error == 0)
1992 		vput(nd.ni_vp);
1993 	return (error);
1994 }
1995 
1996 /*
1997  * Make a hard file link.
1998  */
1999 /* ARGSUSED */
2000 int
2001 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
2002 {
2003 	/* {
2004 		syscallarg(const char *) path;
2005 		syscallarg(const char *) link;
2006 	} */
2007 	struct vnode *vp;
2008 	struct nameidata nd;
2009 	int error;
2010 
2011 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2012 	    SCARG(uap, path));
2013 	if ((error = namei(&nd)) != 0)
2014 		return (error);
2015 	vp = nd.ni_vp;
2016 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2017 	    SCARG(uap, link));
2018 	if ((error = namei(&nd)) != 0)
2019 		goto out;
2020 	if (nd.ni_vp) {
2021 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2022 		if (nd.ni_dvp == nd.ni_vp)
2023 			vrele(nd.ni_dvp);
2024 		else
2025 			vput(nd.ni_dvp);
2026 		vrele(nd.ni_vp);
2027 		error = EEXIST;
2028 		goto out;
2029 	}
2030 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2031 out:
2032 	vrele(vp);
2033 	return (error);
2034 }
2035 
2036 /*
2037  * Make a symbolic link.
2038  */
2039 /* ARGSUSED */
2040 int
2041 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2042 {
2043 	/* {
2044 		syscallarg(const char *) path;
2045 		syscallarg(const char *) link;
2046 	} */
2047 	struct proc *p = l->l_proc;
2048 	struct vattr vattr;
2049 	char *path;
2050 	int error;
2051 	struct nameidata nd;
2052 
2053 	path = PNBUF_GET();
2054 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
2055 	if (error)
2056 		goto out;
2057 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2058 	    SCARG(uap, link));
2059 	if ((error = namei(&nd)) != 0)
2060 		goto out;
2061 	if (nd.ni_vp) {
2062 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2063 		if (nd.ni_dvp == nd.ni_vp)
2064 			vrele(nd.ni_dvp);
2065 		else
2066 			vput(nd.ni_dvp);
2067 		vrele(nd.ni_vp);
2068 		error = EEXIST;
2069 		goto out;
2070 	}
2071 	VATTR_NULL(&vattr);
2072 	vattr.va_type = VLNK;
2073 	/* We will read cwdi->cwdi_cmask unlocked. */
2074 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2075 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2076 	if (error == 0)
2077 		vput(nd.ni_vp);
2078 out:
2079 	PNBUF_PUT(path);
2080 	return (error);
2081 }
2082 
2083 /*
2084  * Delete a whiteout from the filesystem.
2085  */
2086 /* ARGSUSED */
2087 int
2088 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2089 {
2090 	/* {
2091 		syscallarg(const char *) path;
2092 	} */
2093 	int error;
2094 	struct nameidata nd;
2095 
2096 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
2097 	    UIO_USERSPACE, SCARG(uap, path));
2098 	error = namei(&nd);
2099 	if (error)
2100 		return (error);
2101 
2102 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2103 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2104 		if (nd.ni_dvp == nd.ni_vp)
2105 			vrele(nd.ni_dvp);
2106 		else
2107 			vput(nd.ni_dvp);
2108 		if (nd.ni_vp)
2109 			vrele(nd.ni_vp);
2110 		return (EEXIST);
2111 	}
2112 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2113 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2114 	vput(nd.ni_dvp);
2115 	return (error);
2116 }
2117 
2118 /*
2119  * Delete a name from the filesystem.
2120  */
2121 /* ARGSUSED */
2122 int
2123 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2124 {
2125 	/* {
2126 		syscallarg(const char *) path;
2127 	} */
2128 
2129 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2130 }
2131 
2132 int
2133 do_sys_unlink(const char *arg, enum uio_seg seg)
2134 {
2135 	struct vnode *vp;
2136 	int error;
2137 	struct nameidata nd;
2138 	kauth_cred_t cred;
2139 	char *path;
2140 	const char *cpath;
2141 
2142 	VERIEXEC_PATH_GET(arg, seg, cpath, path);
2143 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
2144 
2145 	if ((error = namei(&nd)) != 0)
2146 		goto out;
2147 	vp = nd.ni_vp;
2148 
2149 	/*
2150 	 * The root of a mounted filesystem cannot be deleted.
2151 	 */
2152 	if (vp->v_vflag & VV_ROOT) {
2153 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2154 		if (nd.ni_dvp == vp)
2155 			vrele(nd.ni_dvp);
2156 		else
2157 			vput(nd.ni_dvp);
2158 		vput(vp);
2159 		error = EBUSY;
2160 		goto out;
2161 	}
2162 
2163 #if NVERIEXEC > 0
2164 	/* Handle remove requests for veriexec entries. */
2165 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
2166 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2167 		if (nd.ni_dvp == vp)
2168 			vrele(nd.ni_dvp);
2169 		else
2170 			vput(nd.ni_dvp);
2171 		vput(vp);
2172 		goto out;
2173 	}
2174 #endif /* NVERIEXEC > 0 */
2175 
2176 	cred = kauth_cred_get();
2177 #ifdef FILEASSOC
2178 	(void)fileassoc_file_delete(vp);
2179 #endif /* FILEASSOC */
2180 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2181 out:
2182 	VERIEXEC_PATH_PUT(path);
2183 	return (error);
2184 }
2185 
2186 /*
2187  * Reposition read/write file offset.
2188  */
2189 int
2190 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2191 {
2192 	/* {
2193 		syscallarg(int) fd;
2194 		syscallarg(int) pad;
2195 		syscallarg(off_t) offset;
2196 		syscallarg(int) whence;
2197 	} */
2198 	kauth_cred_t cred = l->l_cred;
2199 	file_t *fp;
2200 	struct vnode *vp;
2201 	struct vattr vattr;
2202 	off_t newoff;
2203 	int error, fd;
2204 
2205 	fd = SCARG(uap, fd);
2206 
2207 	if ((fp = fd_getfile(fd)) == NULL)
2208 		return (EBADF);
2209 
2210 	vp = fp->f_data;
2211 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2212 		error = ESPIPE;
2213 		goto out;
2214 	}
2215 
2216 	switch (SCARG(uap, whence)) {
2217 	case SEEK_CUR:
2218 		newoff = fp->f_offset + SCARG(uap, offset);
2219 		break;
2220 	case SEEK_END:
2221 		error = VOP_GETATTR(vp, &vattr, cred);
2222 		if (error) {
2223 			goto out;
2224 		}
2225 		newoff = SCARG(uap, offset) + vattr.va_size;
2226 		break;
2227 	case SEEK_SET:
2228 		newoff = SCARG(uap, offset);
2229 		break;
2230 	default:
2231 		error = EINVAL;
2232 		goto out;
2233 	}
2234 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2235 		*(off_t *)retval = fp->f_offset = newoff;
2236 	}
2237  out:
2238  	fd_putfile(fd);
2239 	return (error);
2240 }
2241 
2242 /*
2243  * Positional read system call.
2244  */
2245 int
2246 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2247 {
2248 	/* {
2249 		syscallarg(int) fd;
2250 		syscallarg(void *) buf;
2251 		syscallarg(size_t) nbyte;
2252 		syscallarg(off_t) offset;
2253 	} */
2254 	file_t *fp;
2255 	struct vnode *vp;
2256 	off_t offset;
2257 	int error, fd = SCARG(uap, fd);
2258 
2259 	if ((fp = fd_getfile(fd)) == NULL)
2260 		return (EBADF);
2261 
2262 	if ((fp->f_flag & FREAD) == 0) {
2263 		fd_putfile(fd);
2264 		return (EBADF);
2265 	}
2266 
2267 	vp = fp->f_data;
2268 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2269 		error = ESPIPE;
2270 		goto out;
2271 	}
2272 
2273 	offset = SCARG(uap, offset);
2274 
2275 	/*
2276 	 * XXX This works because no file systems actually
2277 	 * XXX take any action on the seek operation.
2278 	 */
2279 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2280 		goto out;
2281 
2282 	/* dofileread() will unuse the descriptor for us */
2283 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2284 	    &offset, 0, retval));
2285 
2286  out:
2287 	fd_putfile(fd);
2288 	return (error);
2289 }
2290 
2291 /*
2292  * Positional scatter read system call.
2293  */
2294 int
2295 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2296 {
2297 	/* {
2298 		syscallarg(int) fd;
2299 		syscallarg(const struct iovec *) iovp;
2300 		syscallarg(int) iovcnt;
2301 		syscallarg(off_t) offset;
2302 	} */
2303 	off_t offset = SCARG(uap, offset);
2304 
2305 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2306 	    SCARG(uap, iovcnt), &offset, 0, retval);
2307 }
2308 
2309 /*
2310  * Positional write system call.
2311  */
2312 int
2313 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2314 {
2315 	/* {
2316 		syscallarg(int) fd;
2317 		syscallarg(const void *) buf;
2318 		syscallarg(size_t) nbyte;
2319 		syscallarg(off_t) offset;
2320 	} */
2321 	file_t *fp;
2322 	struct vnode *vp;
2323 	off_t offset;
2324 	int error, fd = SCARG(uap, fd);
2325 
2326 	if ((fp = fd_getfile(fd)) == NULL)
2327 		return (EBADF);
2328 
2329 	if ((fp->f_flag & FWRITE) == 0) {
2330 		fd_putfile(fd);
2331 		return (EBADF);
2332 	}
2333 
2334 	vp = fp->f_data;
2335 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2336 		error = ESPIPE;
2337 		goto out;
2338 	}
2339 
2340 	offset = SCARG(uap, offset);
2341 
2342 	/*
2343 	 * XXX This works because no file systems actually
2344 	 * XXX take any action on the seek operation.
2345 	 */
2346 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2347 		goto out;
2348 
2349 	/* dofilewrite() will unuse the descriptor for us */
2350 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2351 	    &offset, 0, retval));
2352 
2353  out:
2354 	fd_putfile(fd);
2355 	return (error);
2356 }
2357 
2358 /*
2359  * Positional gather write system call.
2360  */
2361 int
2362 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2363 {
2364 	/* {
2365 		syscallarg(int) fd;
2366 		syscallarg(const struct iovec *) iovp;
2367 		syscallarg(int) iovcnt;
2368 		syscallarg(off_t) offset;
2369 	} */
2370 	off_t offset = SCARG(uap, offset);
2371 
2372 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2373 	    SCARG(uap, iovcnt), &offset, 0, retval);
2374 }
2375 
2376 /*
2377  * Check access permissions.
2378  */
2379 int
2380 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2381 {
2382 	/* {
2383 		syscallarg(const char *) path;
2384 		syscallarg(int) flags;
2385 	} */
2386 	kauth_cred_t cred;
2387 	struct vnode *vp;
2388 	int error, flags;
2389 	struct nameidata nd;
2390 
2391 	cred = kauth_cred_dup(l->l_cred);
2392 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2393 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2394 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2395 	    SCARG(uap, path));
2396 	/* Override default credentials */
2397 	nd.ni_cnd.cn_cred = cred;
2398 	if ((error = namei(&nd)) != 0)
2399 		goto out;
2400 	vp = nd.ni_vp;
2401 
2402 	/* Flags == 0 means only check for existence. */
2403 	if (SCARG(uap, flags)) {
2404 		flags = 0;
2405 		if (SCARG(uap, flags) & R_OK)
2406 			flags |= VREAD;
2407 		if (SCARG(uap, flags) & W_OK)
2408 			flags |= VWRITE;
2409 		if (SCARG(uap, flags) & X_OK)
2410 			flags |= VEXEC;
2411 
2412 		error = VOP_ACCESS(vp, flags, cred);
2413 		if (!error && (flags & VWRITE))
2414 			error = vn_writechk(vp);
2415 	}
2416 	vput(vp);
2417 out:
2418 	kauth_cred_free(cred);
2419 	return (error);
2420 }
2421 
2422 /*
2423  * Common code for all sys_stat functions, including compat versions.
2424  */
2425 int
2426 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
2427 {
2428 	int error;
2429 	struct nameidata nd;
2430 
2431 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
2432 	    UIO_USERSPACE, path);
2433 	error = namei(&nd);
2434 	if (error != 0)
2435 		return error;
2436 	error = vn_stat(nd.ni_vp, sb);
2437 	vput(nd.ni_vp);
2438 	return error;
2439 }
2440 
2441 /*
2442  * Get file status; this version follows links.
2443  */
2444 /* ARGSUSED */
2445 int
2446 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval)
2447 {
2448 	/* {
2449 		syscallarg(const char *) path;
2450 		syscallarg(struct stat *) ub;
2451 	} */
2452 	struct stat sb;
2453 	int error;
2454 
2455 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2456 	if (error)
2457 		return error;
2458 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2459 }
2460 
2461 /*
2462  * Get file status; this version does not follow links.
2463  */
2464 /* ARGSUSED */
2465 int
2466 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval)
2467 {
2468 	/* {
2469 		syscallarg(const char *) path;
2470 		syscallarg(struct stat *) ub;
2471 	} */
2472 	struct stat sb;
2473 	int error;
2474 
2475 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2476 	if (error)
2477 		return error;
2478 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2479 }
2480 
2481 /*
2482  * Get configurable pathname variables.
2483  */
2484 /* ARGSUSED */
2485 int
2486 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2487 {
2488 	/* {
2489 		syscallarg(const char *) path;
2490 		syscallarg(int) name;
2491 	} */
2492 	int error;
2493 	struct nameidata nd;
2494 
2495 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2496 	    SCARG(uap, path));
2497 	if ((error = namei(&nd)) != 0)
2498 		return (error);
2499 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2500 	vput(nd.ni_vp);
2501 	return (error);
2502 }
2503 
2504 /*
2505  * Return target name of a symbolic link.
2506  */
2507 /* ARGSUSED */
2508 int
2509 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2510 {
2511 	/* {
2512 		syscallarg(const char *) path;
2513 		syscallarg(char *) buf;
2514 		syscallarg(size_t) count;
2515 	} */
2516 	struct vnode *vp;
2517 	struct iovec aiov;
2518 	struct uio auio;
2519 	int error;
2520 	struct nameidata nd;
2521 
2522 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2523 	    SCARG(uap, path));
2524 	if ((error = namei(&nd)) != 0)
2525 		return (error);
2526 	vp = nd.ni_vp;
2527 	if (vp->v_type != VLNK)
2528 		error = EINVAL;
2529 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2530 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2531 		aiov.iov_base = SCARG(uap, buf);
2532 		aiov.iov_len = SCARG(uap, count);
2533 		auio.uio_iov = &aiov;
2534 		auio.uio_iovcnt = 1;
2535 		auio.uio_offset = 0;
2536 		auio.uio_rw = UIO_READ;
2537 		KASSERT(l == curlwp);
2538 		auio.uio_vmspace = l->l_proc->p_vmspace;
2539 		auio.uio_resid = SCARG(uap, count);
2540 		error = VOP_READLINK(vp, &auio, l->l_cred);
2541 	}
2542 	vput(vp);
2543 	*retval = SCARG(uap, count) - auio.uio_resid;
2544 	return (error);
2545 }
2546 
2547 /*
2548  * Change flags of a file given a path name.
2549  */
2550 /* ARGSUSED */
2551 int
2552 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2553 {
2554 	/* {
2555 		syscallarg(const char *) path;
2556 		syscallarg(u_long) flags;
2557 	} */
2558 	struct vnode *vp;
2559 	int error;
2560 	struct nameidata nd;
2561 
2562 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2563 	    SCARG(uap, path));
2564 	if ((error = namei(&nd)) != 0)
2565 		return (error);
2566 	vp = nd.ni_vp;
2567 	error = change_flags(vp, SCARG(uap, flags), l);
2568 	vput(vp);
2569 	return (error);
2570 }
2571 
2572 /*
2573  * Change flags of a file given a file descriptor.
2574  */
2575 /* ARGSUSED */
2576 int
2577 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
2578 {
2579 	/* {
2580 		syscallarg(int) fd;
2581 		syscallarg(u_long) flags;
2582 	} */
2583 	struct vnode *vp;
2584 	file_t *fp;
2585 	int error;
2586 
2587 	/* fd_getvnode() will use the descriptor for us */
2588 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2589 		return (error);
2590 	vp = fp->f_data;
2591 	error = change_flags(vp, SCARG(uap, flags), l);
2592 	VOP_UNLOCK(vp, 0);
2593 	fd_putfile(SCARG(uap, fd));
2594 	return (error);
2595 }
2596 
2597 /*
2598  * Change flags of a file given a path name; this version does
2599  * not follow links.
2600  */
2601 int
2602 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
2603 {
2604 	/* {
2605 		syscallarg(const char *) path;
2606 		syscallarg(u_long) flags;
2607 	} */
2608 	struct vnode *vp;
2609 	int error;
2610 	struct nameidata nd;
2611 
2612 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2613 	    SCARG(uap, path));
2614 	if ((error = namei(&nd)) != 0)
2615 		return (error);
2616 	vp = nd.ni_vp;
2617 	error = change_flags(vp, SCARG(uap, flags), l);
2618 	vput(vp);
2619 	return (error);
2620 }
2621 
2622 /*
2623  * Common routine to change flags of a file.
2624  */
2625 int
2626 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2627 {
2628 	struct vattr vattr;
2629 	int error;
2630 
2631 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2632 	/*
2633 	 * Non-superusers cannot change the flags on devices, even if they
2634 	 * own them.
2635 	 */
2636 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2637 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2638 			goto out;
2639 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2640 			error = EINVAL;
2641 			goto out;
2642 		}
2643 	}
2644 	VATTR_NULL(&vattr);
2645 	vattr.va_flags = flags;
2646 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2647 out:
2648 	return (error);
2649 }
2650 
2651 /*
2652  * Change mode of a file given path name; this version follows links.
2653  */
2654 /* ARGSUSED */
2655 int
2656 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
2657 {
2658 	/* {
2659 		syscallarg(const char *) path;
2660 		syscallarg(int) mode;
2661 	} */
2662 	int error;
2663 	struct nameidata nd;
2664 
2665 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2666 	    SCARG(uap, path));
2667 	if ((error = namei(&nd)) != 0)
2668 		return (error);
2669 
2670 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2671 
2672 	vrele(nd.ni_vp);
2673 	return (error);
2674 }
2675 
2676 /*
2677  * Change mode of a file given a file descriptor.
2678  */
2679 /* ARGSUSED */
2680 int
2681 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
2682 {
2683 	/* {
2684 		syscallarg(int) fd;
2685 		syscallarg(int) mode;
2686 	} */
2687 	file_t *fp;
2688 	int error;
2689 
2690 	/* fd_getvnode() will use the descriptor for us */
2691 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2692 		return (error);
2693 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
2694 	fd_putfile(SCARG(uap, fd));
2695 	return (error);
2696 }
2697 
2698 /*
2699  * Change mode of a file given path name; this version does not follow links.
2700  */
2701 /* ARGSUSED */
2702 int
2703 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
2704 {
2705 	/* {
2706 		syscallarg(const char *) path;
2707 		syscallarg(int) mode;
2708 	} */
2709 	int error;
2710 	struct nameidata nd;
2711 
2712 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2713 	    SCARG(uap, path));
2714 	if ((error = namei(&nd)) != 0)
2715 		return (error);
2716 
2717 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2718 
2719 	vrele(nd.ni_vp);
2720 	return (error);
2721 }
2722 
2723 /*
2724  * Common routine to set mode given a vnode.
2725  */
2726 static int
2727 change_mode(struct vnode *vp, int mode, struct lwp *l)
2728 {
2729 	struct vattr vattr;
2730 	int error;
2731 
2732 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2733 	VATTR_NULL(&vattr);
2734 	vattr.va_mode = mode & ALLPERMS;
2735 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2736 	VOP_UNLOCK(vp, 0);
2737 	return (error);
2738 }
2739 
2740 /*
2741  * Set ownership given a path name; this version follows links.
2742  */
2743 /* ARGSUSED */
2744 int
2745 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
2746 {
2747 	/* {
2748 		syscallarg(const char *) path;
2749 		syscallarg(uid_t) uid;
2750 		syscallarg(gid_t) gid;
2751 	} */
2752 	int error;
2753 	struct nameidata nd;
2754 
2755 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2756 	    SCARG(uap, path));
2757 	if ((error = namei(&nd)) != 0)
2758 		return (error);
2759 
2760 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2761 
2762 	vrele(nd.ni_vp);
2763 	return (error);
2764 }
2765 
2766 /*
2767  * Set ownership given a path name; this version follows links.
2768  * Provides POSIX semantics.
2769  */
2770 /* ARGSUSED */
2771 int
2772 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
2773 {
2774 	/* {
2775 		syscallarg(const char *) path;
2776 		syscallarg(uid_t) uid;
2777 		syscallarg(gid_t) gid;
2778 	} */
2779 	int error;
2780 	struct nameidata nd;
2781 
2782 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2783 	    SCARG(uap, path));
2784 	if ((error = namei(&nd)) != 0)
2785 		return (error);
2786 
2787 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2788 
2789 	vrele(nd.ni_vp);
2790 	return (error);
2791 }
2792 
2793 /*
2794  * Set ownership given a file descriptor.
2795  */
2796 /* ARGSUSED */
2797 int
2798 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
2799 {
2800 	/* {
2801 		syscallarg(int) fd;
2802 		syscallarg(uid_t) uid;
2803 		syscallarg(gid_t) gid;
2804 	} */
2805 	int error;
2806 	file_t *fp;
2807 
2808 	/* fd_getvnode() will use the descriptor for us */
2809 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2810 		return (error);
2811 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2812 	    l, 0);
2813 	fd_putfile(SCARG(uap, fd));
2814 	return (error);
2815 }
2816 
2817 /*
2818  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2819  */
2820 /* ARGSUSED */
2821 int
2822 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
2823 {
2824 	/* {
2825 		syscallarg(int) fd;
2826 		syscallarg(uid_t) uid;
2827 		syscallarg(gid_t) gid;
2828 	} */
2829 	int error;
2830 	file_t *fp;
2831 
2832 	/* fd_getvnode() will use the descriptor for us */
2833 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2834 		return (error);
2835 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
2836 	    l, 1);
2837 	fd_putfile(SCARG(uap, fd));
2838 	return (error);
2839 }
2840 
2841 /*
2842  * Set ownership given a path name; this version does not follow links.
2843  */
2844 /* ARGSUSED */
2845 int
2846 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
2847 {
2848 	/* {
2849 		syscallarg(const char *) path;
2850 		syscallarg(uid_t) uid;
2851 		syscallarg(gid_t) gid;
2852 	} */
2853 	int error;
2854 	struct nameidata nd;
2855 
2856 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2857 	    SCARG(uap, path));
2858 	if ((error = namei(&nd)) != 0)
2859 		return (error);
2860 
2861 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2862 
2863 	vrele(nd.ni_vp);
2864 	return (error);
2865 }
2866 
2867 /*
2868  * Set ownership given a path name; this version does not follow links.
2869  * Provides POSIX/XPG semantics.
2870  */
2871 /* ARGSUSED */
2872 int
2873 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
2874 {
2875 	/* {
2876 		syscallarg(const char *) path;
2877 		syscallarg(uid_t) uid;
2878 		syscallarg(gid_t) gid;
2879 	} */
2880 	int error;
2881 	struct nameidata nd;
2882 
2883 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2884 	    SCARG(uap, path));
2885 	if ((error = namei(&nd)) != 0)
2886 		return (error);
2887 
2888 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2889 
2890 	vrele(nd.ni_vp);
2891 	return (error);
2892 }
2893 
2894 /*
2895  * Common routine to set ownership given a vnode.
2896  */
2897 static int
2898 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2899     int posix_semantics)
2900 {
2901 	struct vattr vattr;
2902 	mode_t newmode;
2903 	int error;
2904 
2905 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2906 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
2907 		goto out;
2908 
2909 #define CHANGED(x) ((int)(x) != -1)
2910 	newmode = vattr.va_mode;
2911 	if (posix_semantics) {
2912 		/*
2913 		 * POSIX/XPG semantics: if the caller is not the super-user,
2914 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2915 		 * the XPG consider the behaviour for calls by the super-user
2916 		 * implementation-defined; we leave the set-user-id and set-
2917 		 * group-id settings intact in that case.
2918 		 */
2919 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2920 				      NULL) != 0)
2921 			newmode &= ~(S_ISUID | S_ISGID);
2922 	} else {
2923 		/*
2924 		 * NetBSD semantics: when changing owner and/or group,
2925 		 * clear the respective bit(s).
2926 		 */
2927 		if (CHANGED(uid))
2928 			newmode &= ~S_ISUID;
2929 		if (CHANGED(gid))
2930 			newmode &= ~S_ISGID;
2931 	}
2932 	/* Update va_mode iff altered. */
2933 	if (vattr.va_mode == newmode)
2934 		newmode = VNOVAL;
2935 
2936 	VATTR_NULL(&vattr);
2937 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2938 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2939 	vattr.va_mode = newmode;
2940 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
2941 #undef CHANGED
2942 
2943 out:
2944 	VOP_UNLOCK(vp, 0);
2945 	return (error);
2946 }
2947 
2948 /*
2949  * Set the access and modification times given a path name; this
2950  * version follows links.
2951  */
2952 /* ARGSUSED */
2953 int
2954 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval)
2955 {
2956 	/* {
2957 		syscallarg(const char *) path;
2958 		syscallarg(const struct timeval *) tptr;
2959 	} */
2960 
2961 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
2962 	    SCARG(uap, tptr), UIO_USERSPACE);
2963 }
2964 
2965 /*
2966  * Set the access and modification times given a file descriptor.
2967  */
2968 /* ARGSUSED */
2969 int
2970 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval)
2971 {
2972 	/* {
2973 		syscallarg(int) fd;
2974 		syscallarg(const struct timeval *) tptr;
2975 	} */
2976 	int error;
2977 	file_t *fp;
2978 
2979 	/* fd_getvnode() will use the descriptor for us */
2980 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
2981 		return (error);
2982 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
2983 	    UIO_USERSPACE);
2984 	fd_putfile(SCARG(uap, fd));
2985 	return (error);
2986 }
2987 
2988 /*
2989  * Set the access and modification times given a path name; this
2990  * version does not follow links.
2991  */
2992 int
2993 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval)
2994 {
2995 	/* {
2996 		syscallarg(const char *) path;
2997 		syscallarg(const struct timeval *) tptr;
2998 	} */
2999 
3000 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
3001 	    SCARG(uap, tptr), UIO_USERSPACE);
3002 }
3003 
3004 /*
3005  * Common routine to set access and modification times given a vnode.
3006  */
3007 int
3008 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3009     const struct timeval *tptr, enum uio_seg seg)
3010 {
3011 	struct vattr vattr;
3012 	struct nameidata nd;
3013 	int error;
3014 	bool vanull, setbirthtime;
3015 	struct timespec ts[2];
3016 
3017 	if (tptr == NULL) {
3018 		vanull = true;
3019 		nanotime(&ts[0]);
3020 		ts[1] = ts[0];
3021 	} else {
3022 		struct timeval tv[2];
3023 
3024 		vanull = false;
3025 		if (seg != UIO_SYSSPACE) {
3026 			error = copyin(tptr, &tv, sizeof (tv));
3027 			if (error != 0)
3028 				return error;
3029 			tptr = tv;
3030 		}
3031 		TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3032 		TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3033 	}
3034 
3035 	if (vp == NULL) {
3036 		NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path);
3037 		if ((error = namei(&nd)) != 0)
3038 			return error;
3039 		vp = nd.ni_vp;
3040 	} else
3041 		nd.ni_vp = NULL;
3042 
3043 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3044 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3045 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
3046 	VATTR_NULL(&vattr);
3047 	vattr.va_atime = ts[0];
3048 	vattr.va_mtime = ts[1];
3049 	if (setbirthtime)
3050 		vattr.va_birthtime = ts[1];
3051 	if (vanull)
3052 		vattr.va_flags |= VA_UTIMES_NULL;
3053 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
3054 	VOP_UNLOCK(vp, 0);
3055 
3056 	if (nd.ni_vp != NULL)
3057 		vrele(nd.ni_vp);
3058 
3059 	return error;
3060 }
3061 
3062 /*
3063  * Truncate a file given its path name.
3064  */
3065 /* ARGSUSED */
3066 int
3067 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3068 {
3069 	/* {
3070 		syscallarg(const char *) path;
3071 		syscallarg(int) pad;
3072 		syscallarg(off_t) length;
3073 	} */
3074 	struct vnode *vp;
3075 	struct vattr vattr;
3076 	int error;
3077 	struct nameidata nd;
3078 
3079 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
3080 	    SCARG(uap, path));
3081 	if ((error = namei(&nd)) != 0)
3082 		return (error);
3083 	vp = nd.ni_vp;
3084 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3085 	if (vp->v_type == VDIR)
3086 		error = EISDIR;
3087 	else if ((error = vn_writechk(vp)) == 0 &&
3088 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3089 		VATTR_NULL(&vattr);
3090 		vattr.va_size = SCARG(uap, length);
3091 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
3092 	}
3093 	vput(vp);
3094 	return (error);
3095 }
3096 
3097 /*
3098  * Truncate a file given a file descriptor.
3099  */
3100 /* ARGSUSED */
3101 int
3102 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3103 {
3104 	/* {
3105 		syscallarg(int) fd;
3106 		syscallarg(int) pad;
3107 		syscallarg(off_t) length;
3108 	} */
3109 	struct vattr vattr;
3110 	struct vnode *vp;
3111 	file_t *fp;
3112 	int error;
3113 
3114 	/* fd_getvnode() will use the descriptor for us */
3115 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3116 		return (error);
3117 	if ((fp->f_flag & FWRITE) == 0) {
3118 		error = EINVAL;
3119 		goto out;
3120 	}
3121 	vp = fp->f_data;
3122 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3123 	if (vp->v_type == VDIR)
3124 		error = EISDIR;
3125 	else if ((error = vn_writechk(vp)) == 0) {
3126 		VATTR_NULL(&vattr);
3127 		vattr.va_size = SCARG(uap, length);
3128 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3129 	}
3130 	VOP_UNLOCK(vp, 0);
3131  out:
3132 	fd_putfile(SCARG(uap, fd));
3133 	return (error);
3134 }
3135 
3136 /*
3137  * Sync an open file.
3138  */
3139 /* ARGSUSED */
3140 int
3141 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3142 {
3143 	/* {
3144 		syscallarg(int) fd;
3145 	} */
3146 	struct vnode *vp;
3147 	file_t *fp;
3148 	int error;
3149 
3150 	/* fd_getvnode() will use the descriptor for us */
3151 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3152 		return (error);
3153 	vp = fp->f_data;
3154 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3155 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3156 	if (error == 0 && bioopsp != NULL &&
3157 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3158 		(*bioopsp->io_fsync)(vp, 0);
3159 	VOP_UNLOCK(vp, 0);
3160 	fd_putfile(SCARG(uap, fd));
3161 	return (error);
3162 }
3163 
3164 /*
3165  * Sync a range of file data.  API modeled after that found in AIX.
3166  *
3167  * FDATASYNC indicates that we need only save enough metadata to be able
3168  * to re-read the written data.  Note we duplicate AIX's requirement that
3169  * the file be open for writing.
3170  */
3171 /* ARGSUSED */
3172 int
3173 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3174 {
3175 	/* {
3176 		syscallarg(int) fd;
3177 		syscallarg(int) flags;
3178 		syscallarg(off_t) start;
3179 		syscallarg(off_t) length;
3180 	} */
3181 	struct vnode *vp;
3182 	file_t *fp;
3183 	int flags, nflags;
3184 	off_t s, e, len;
3185 	int error;
3186 
3187 	/* fd_getvnode() will use the descriptor for us */
3188 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3189 		return (error);
3190 
3191 	if ((fp->f_flag & FWRITE) == 0) {
3192 		error = EBADF;
3193 		goto out;
3194 	}
3195 
3196 	flags = SCARG(uap, flags);
3197 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3198 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3199 		error = EINVAL;
3200 		goto out;
3201 	}
3202 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3203 	if (flags & FDATASYNC)
3204 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3205 	else
3206 		nflags = FSYNC_WAIT;
3207 	if (flags & FDISKSYNC)
3208 		nflags |= FSYNC_CACHE;
3209 
3210 	len = SCARG(uap, length);
3211 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3212 	if (len) {
3213 		s = SCARG(uap, start);
3214 		e = s + len;
3215 		if (e < s) {
3216 			error = EINVAL;
3217 			goto out;
3218 		}
3219 	} else {
3220 		e = 0;
3221 		s = 0;
3222 	}
3223 
3224 	vp = fp->f_data;
3225 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3226 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3227 
3228 	if (error == 0 && bioopsp != NULL &&
3229 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3230 		(*bioopsp->io_fsync)(vp, nflags);
3231 
3232 	VOP_UNLOCK(vp, 0);
3233 out:
3234 	fd_putfile(SCARG(uap, fd));
3235 	return (error);
3236 }
3237 
3238 /*
3239  * Sync the data of an open file.
3240  */
3241 /* ARGSUSED */
3242 int
3243 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3244 {
3245 	/* {
3246 		syscallarg(int) fd;
3247 	} */
3248 	struct vnode *vp;
3249 	file_t *fp;
3250 	int error;
3251 
3252 	/* fd_getvnode() will use the descriptor for us */
3253 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3254 		return (error);
3255 	if ((fp->f_flag & FWRITE) == 0) {
3256 		fd_putfile(SCARG(uap, fd));
3257 		return (EBADF);
3258 	}
3259 	vp = fp->f_data;
3260 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3261 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3262 	VOP_UNLOCK(vp, 0);
3263 	fd_putfile(SCARG(uap, fd));
3264 	return (error);
3265 }
3266 
3267 /*
3268  * Rename files, (standard) BSD semantics frontend.
3269  */
3270 /* ARGSUSED */
3271 int
3272 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3273 {
3274 	/* {
3275 		syscallarg(const char *) from;
3276 		syscallarg(const char *) to;
3277 	} */
3278 
3279 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3280 }
3281 
3282 /*
3283  * Rename files, POSIX semantics frontend.
3284  */
3285 /* ARGSUSED */
3286 int
3287 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3288 {
3289 	/* {
3290 		syscallarg(const char *) from;
3291 		syscallarg(const char *) to;
3292 	} */
3293 
3294 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3295 }
3296 
3297 /*
3298  * Rename files.  Source and destination must either both be directories,
3299  * or both not be directories.  If target is a directory, it must be empty.
3300  * If `from' and `to' refer to the same object, the value of the `retain'
3301  * argument is used to determine whether `from' will be
3302  *
3303  * (retain == 0)	deleted unless `from' and `to' refer to the same
3304  *			object in the file system's name space (BSD).
3305  * (retain == 1)	always retained (POSIX).
3306  */
3307 int
3308 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3309 {
3310 	struct vnode *tvp, *fvp, *tdvp;
3311 	struct nameidata fromnd, tond;
3312 	struct mount *fs;
3313 	struct lwp *l = curlwp;
3314 	struct proc *p;
3315 	uint32_t saveflag;
3316 	int error;
3317 
3318 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT,
3319 	    seg, from);
3320 	if ((error = namei(&fromnd)) != 0)
3321 		return (error);
3322 	if (fromnd.ni_dvp != fromnd.ni_vp)
3323 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3324 	fvp = fromnd.ni_vp;
3325 
3326 	fs = fvp->v_mount;
3327 	error = VFS_RENAMELOCK_ENTER(fs);
3328 	if (error) {
3329 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3330 		vrele(fromnd.ni_dvp);
3331 		vrele(fvp);
3332 		goto out1;
3333 	}
3334 
3335 	/*
3336 	 * close, partially, yet another race - ideally we should only
3337 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
3338 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
3339 	 * that with namei as it stands.
3340 	 *
3341 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
3342 	 * under us. The real fix is to get the locks in the right
3343 	 * order and do the lookups in the right places, but that's a
3344 	 * major rototill.
3345 	 *
3346 	 * Preserve the SAVESTART in cn_flags, because who knows what
3347 	 * might happen if we don't.
3348 	 *
3349 	 * Note: this logic (as well as this whole function) is cloned
3350 	 * in nfs_serv.c. Proceed accordingly.
3351 	 */
3352 	vrele(fvp);
3353 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
3354 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3355 	    (fromnd.ni_cnd.cn_namelen == 2 &&
3356 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3357 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3358 		error = EINVAL;
3359 		VFS_RENAMELOCK_EXIT(fs);
3360 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3361 		vrele(fromnd.ni_dvp);
3362 		goto out1;
3363 	}
3364 	saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
3365 	fromnd.ni_cnd.cn_flags &= ~SAVESTART;
3366 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3367 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
3368 	fromnd.ni_cnd.cn_flags |= saveflag;
3369 	if (error) {
3370 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3371 		VFS_RENAMELOCK_EXIT(fs);
3372 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3373 		vrele(fromnd.ni_dvp);
3374 		goto out1;
3375 	}
3376 	VOP_UNLOCK(fromnd.ni_vp, 0);
3377 	if (fromnd.ni_dvp != fromnd.ni_vp)
3378 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3379 	fvp = fromnd.ni_vp;
3380 
3381 	NDINIT(&tond, RENAME,
3382 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3383 	      | (fvp->v_type == VDIR ? CREATEDIR : 0),
3384 	    seg, to);
3385 	if ((error = namei(&tond)) != 0) {
3386 		VFS_RENAMELOCK_EXIT(fs);
3387 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3388 		vrele(fromnd.ni_dvp);
3389 		vrele(fvp);
3390 		goto out1;
3391 	}
3392 	tdvp = tond.ni_dvp;
3393 	tvp = tond.ni_vp;
3394 
3395 	if (tvp != NULL) {
3396 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3397 			error = ENOTDIR;
3398 			goto out;
3399 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3400 			error = EISDIR;
3401 			goto out;
3402 		}
3403 	}
3404 
3405 	if (fvp == tdvp)
3406 		error = EINVAL;
3407 
3408 	/*
3409 	 * Source and destination refer to the same object.
3410 	 */
3411 	if (fvp == tvp) {
3412 		if (retain)
3413 			error = -1;
3414 		else if (fromnd.ni_dvp == tdvp &&
3415 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3416 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3417 		          tond.ni_cnd.cn_nameptr,
3418 		          fromnd.ni_cnd.cn_namelen))
3419 		error = -1;
3420 	}
3421 
3422 #if NVERIEXEC > 0
3423 	if (!error) {
3424 		char *f1, *f2;
3425 
3426 		f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
3427 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen);
3428 
3429 		f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
3430 		strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen);
3431 
3432 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
3433 
3434 		free(f1, M_TEMP);
3435 		free(f2, M_TEMP);
3436 	}
3437 #endif /* NVERIEXEC > 0 */
3438 
3439 out:
3440 	p = l->l_proc;
3441 	if (!error) {
3442 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3443 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3444 		VFS_RENAMELOCK_EXIT(fs);
3445 	} else {
3446 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3447 		if (tdvp == tvp)
3448 			vrele(tdvp);
3449 		else
3450 			vput(tdvp);
3451 		if (tvp)
3452 			vput(tvp);
3453 		VFS_RENAMELOCK_EXIT(fs);
3454 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3455 		vrele(fromnd.ni_dvp);
3456 		vrele(fvp);
3457 	}
3458 	vrele(tond.ni_startdir);
3459 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3460 out1:
3461 	if (fromnd.ni_startdir)
3462 		vrele(fromnd.ni_startdir);
3463 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3464 	return (error == -1 ? 0 : error);
3465 }
3466 
3467 /*
3468  * Make a directory file.
3469  */
3470 /* ARGSUSED */
3471 int
3472 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
3473 {
3474 	/* {
3475 		syscallarg(const char *) path;
3476 		syscallarg(int) mode;
3477 	} */
3478 	struct proc *p = l->l_proc;
3479 	struct vnode *vp;
3480 	struct vattr vattr;
3481 	int error;
3482 	struct nameidata nd;
3483 
3484 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE,
3485 	    SCARG(uap, path));
3486 	if ((error = namei(&nd)) != 0)
3487 		return (error);
3488 	vp = nd.ni_vp;
3489 	if (vp != NULL) {
3490 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3491 		if (nd.ni_dvp == vp)
3492 			vrele(nd.ni_dvp);
3493 		else
3494 			vput(nd.ni_dvp);
3495 		vrele(vp);
3496 		return (EEXIST);
3497 	}
3498 	VATTR_NULL(&vattr);
3499 	vattr.va_type = VDIR;
3500 	/* We will read cwdi->cwdi_cmask unlocked. */
3501 	vattr.va_mode =
3502 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3503 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3504 	if (!error)
3505 		vput(nd.ni_vp);
3506 	return (error);
3507 }
3508 
3509 /*
3510  * Remove a directory file.
3511  */
3512 /* ARGSUSED */
3513 int
3514 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
3515 {
3516 	/* {
3517 		syscallarg(const char *) path;
3518 	} */
3519 	struct vnode *vp;
3520 	int error;
3521 	struct nameidata nd;
3522 
3523 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
3524 	    SCARG(uap, path));
3525 	if ((error = namei(&nd)) != 0)
3526 		return (error);
3527 	vp = nd.ni_vp;
3528 	if (vp->v_type != VDIR) {
3529 		error = ENOTDIR;
3530 		goto out;
3531 	}
3532 	/*
3533 	 * No rmdir "." please.
3534 	 */
3535 	if (nd.ni_dvp == vp) {
3536 		error = EINVAL;
3537 		goto out;
3538 	}
3539 	/*
3540 	 * The root of a mounted filesystem cannot be deleted.
3541 	 */
3542 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
3543 		error = EBUSY;
3544 		goto out;
3545 	}
3546 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3547 	return (error);
3548 
3549 out:
3550 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3551 	if (nd.ni_dvp == vp)
3552 		vrele(nd.ni_dvp);
3553 	else
3554 		vput(nd.ni_dvp);
3555 	vput(vp);
3556 	return (error);
3557 }
3558 
3559 /*
3560  * Read a block of directory entries in a file system independent format.
3561  */
3562 int
3563 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
3564 {
3565 	/* {
3566 		syscallarg(int) fd;
3567 		syscallarg(char *) buf;
3568 		syscallarg(size_t) count;
3569 	} */
3570 	file_t *fp;
3571 	int error, done;
3572 
3573 	/* fd_getvnode() will use the descriptor for us */
3574 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3575 		return (error);
3576 	if ((fp->f_flag & FREAD) == 0) {
3577 		error = EBADF;
3578 		goto out;
3579 	}
3580 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3581 			SCARG(uap, count), &done, l, 0, 0);
3582 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3583 	*retval = done;
3584  out:
3585 	fd_putfile(SCARG(uap, fd));
3586 	return (error);
3587 }
3588 
3589 /*
3590  * Set the mode mask for creation of filesystem nodes.
3591  */
3592 int
3593 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
3594 {
3595 	/* {
3596 		syscallarg(mode_t) newmask;
3597 	} */
3598 	struct proc *p = l->l_proc;
3599 	struct cwdinfo *cwdi;
3600 
3601 	/*
3602 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3603 	 * important is that we serialize changes to the mask.  The
3604 	 * rw_exit() will issue a write memory barrier on our behalf,
3605 	 * and force the changes out to other CPUs (as it must use an
3606 	 * atomic operation, draining the local CPU's store buffers).
3607 	 */
3608 	cwdi = p->p_cwdi;
3609 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3610 	*retval = cwdi->cwdi_cmask;
3611 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3612 	rw_exit(&cwdi->cwdi_lock);
3613 
3614 	return (0);
3615 }
3616 
3617 int
3618 dorevoke(struct vnode *vp, kauth_cred_t cred)
3619 {
3620 	struct vattr vattr;
3621 	int error;
3622 
3623 	if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
3624 		return error;
3625 	if (kauth_cred_geteuid(cred) != vattr.va_uid &&
3626 	    (error = kauth_authorize_generic(cred,
3627 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
3628 		VOP_REVOKE(vp, REVOKEALL);
3629 	return (error);
3630 }
3631 
3632 /*
3633  * Void all references to file by ripping underlying filesystem
3634  * away from vnode.
3635  */
3636 /* ARGSUSED */
3637 int
3638 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
3639 {
3640 	/* {
3641 		syscallarg(const char *) path;
3642 	} */
3643 	struct vnode *vp;
3644 	int error;
3645 	struct nameidata nd;
3646 
3647 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
3648 	    SCARG(uap, path));
3649 	if ((error = namei(&nd)) != 0)
3650 		return (error);
3651 	vp = nd.ni_vp;
3652 	error = dorevoke(vp, l->l_cred);
3653 	vrele(vp);
3654 	return (error);
3655 }
3656