xref: /dflybsd-src/sys/kern/vfs_syscalls.c (revision ac2e3f5effc58aa364c7e5c199f35ebbae7cda81)
1 /*
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
39  * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
40  * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.15 2003/08/07 21:17:23 dillon Exp $
41  */
42 
43 /* For 4.3 integer FS ID compatibility */
44 #include "opt_compat.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/buf.h>
49 #include <sys/sysent.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/sysproto.h>
53 #include <sys/filedesc.h>
54 #include <sys/kernel.h>
55 #include <sys/fcntl.h>
56 #include <sys/file.h>
57 #include <sys/linker.h>
58 #include <sys/stat.h>
59 #include <sys/unistd.h>
60 #include <sys/vnode.h>
61 #include <sys/proc.h>
62 #include <sys/namei.h>
63 #include <sys/dirent.h>
64 #include <sys/extattr.h>
65 
66 #include <machine/limits.h>
67 #include <vfs/union/union.h>
68 #include <sys/sysctl.h>
69 #include <vm/vm.h>
70 #include <vm/vm_object.h>
71 #include <vm/vm_zone.h>
72 #include <vm/vm_page.h>
73 
74 #include <sys/file2.h>
75 
76 static int change_dir __P((struct nameidata *ndp, struct thread *td));
77 static void checkdirs __P((struct vnode *olddp));
78 static int chroot_refuse_vdir_fds __P((struct filedesc *fdp));
79 static int getutimes __P((const struct timeval *, struct timespec *));
80 static int setfown __P((struct vnode *, uid_t, gid_t));
81 static int setfmode __P((struct vnode *, int));
82 static int setfflags __P((struct vnode *, int));
83 static int setutimes __P((struct vnode *, const struct timespec *, int));
84 static int	usermount = 0;	/* if 1, non-root can mount fs. */
85 
86 int (*union_dircheckp) __P((struct thread *, struct vnode **, struct file *));
87 
88 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
89 
90 /*
91  * Virtual File System System Calls
92  */
93 
94 /*
95  * Mount a file system.
96  */
97 /*
98  * mount_args(char *type, char *path, int flags, caddr_t data)
99  */
100 /* ARGSUSED */
101 int
102 mount(struct mount_args *uap)
103 {
104 	struct thread *td = curthread;
105 	struct proc *p = td->td_proc;
106 	struct vnode *vp;
107 	struct mount *mp;
108 	struct vfsconf *vfsp;
109 	int error, flag = 0, flag2 = 0;
110 	struct vattr va;
111 #ifdef COMPAT_43
112 	u_long fstypenum;
113 #endif
114 	struct nameidata nd;
115 	char fstypename[MFSNAMELEN];
116 
117 	if (usermount == 0 && (error = suser(td)))
118 		return (error);
119 	/*
120 	 * Do not allow NFS export by non-root users.
121 	 */
122 	if (SCARG(uap, flags) & MNT_EXPORTED) {
123 		error = suser(td);
124 		if (error)
125 			return (error);
126 	}
127 	/*
128 	 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
129 	 */
130 	if (suser(td))
131 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
132 	/*
133 	 * Get vnode to be covered
134 	 */
135 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
136 	    SCARG(uap, path), td);
137 	if ((error = namei(&nd)) != 0)
138 		return (error);
139 	NDFREE(&nd, NDF_ONLY_PNBUF);
140 	vp = nd.ni_vp;
141 	if (SCARG(uap, flags) & MNT_UPDATE) {
142 		if ((vp->v_flag & VROOT) == 0) {
143 			vput(vp);
144 			return (EINVAL);
145 		}
146 		mp = vp->v_mount;
147 		flag = mp->mnt_flag;
148 		flag2 = mp->mnt_kern_flag;
149 		/*
150 		 * We only allow the filesystem to be reloaded if it
151 		 * is currently mounted read-only.
152 		 */
153 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
154 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
155 			vput(vp);
156 			return (EOPNOTSUPP);	/* Needs translation */
157 		}
158 		/*
159 		 * Only root, or the user that did the original mount is
160 		 * permitted to update it.
161 		 */
162 		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
163 		    (error = suser(td))) {
164 			vput(vp);
165 			return (error);
166 		}
167 		if (vfs_busy(mp, LK_NOWAIT, 0, td)) {
168 			vput(vp);
169 			return (EBUSY);
170 		}
171 		lwkt_gettoken(&vp->v_interlock);
172 		if ((vp->v_flag & VMOUNT) != 0 ||
173 		    vp->v_mountedhere != NULL) {
174 			lwkt_reltoken(&vp->v_interlock);
175 			vfs_unbusy(mp, td);
176 			vput(vp);
177 			return (EBUSY);
178 		}
179 		vp->v_flag |= VMOUNT;
180 		lwkt_reltoken(&vp->v_interlock);
181 		mp->mnt_flag |=
182 		    SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
183 		VOP_UNLOCK(vp, 0, td);
184 		goto update;
185 	}
186 	/*
187 	 * If the user is not root, ensure that they own the directory
188 	 * onto which we are attempting to mount.
189 	 */
190 	if ((error = VOP_GETATTR(vp, &va, td)) ||
191 	    (va.va_uid != p->p_ucred->cr_uid &&
192 	     (error = suser(td)))) {
193 		vput(vp);
194 		return (error);
195 	}
196 	if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) {
197 		vput(vp);
198 		return (error);
199 	}
200 	if (vp->v_type != VDIR) {
201 		vput(vp);
202 		return (ENOTDIR);
203 	}
204 #ifdef COMPAT_43
205 	/*
206 	 * Historically filesystem types were identified by number. If we
207 	 * get an integer for the filesystem type instead of a string, we
208 	 * check to see if it matches one of the historic filesystem types.
209 	 */
210 	fstypenum = (uintptr_t)SCARG(uap, type);
211 	if (fstypenum < maxvfsconf) {
212 		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
213 			if (vfsp->vfc_typenum == fstypenum)
214 				break;
215 		if (vfsp == NULL) {
216 			vput(vp);
217 			return (ENODEV);
218 		}
219 		strncpy(fstypename, vfsp->vfc_name, MFSNAMELEN);
220 	} else
221 #endif /* COMPAT_43 */
222 	if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) {
223 		vput(vp);
224 		return (error);
225 	}
226 	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
227 		if (!strcmp(vfsp->vfc_name, fstypename))
228 			break;
229 	if (vfsp == NULL) {
230 		linker_file_t lf;
231 
232 		/* Only load modules for root (very important!) */
233 		if ((error = suser(td)) != 0) {
234 			vput(vp);
235 			return error;
236 		}
237 		error = linker_load_file(fstypename, &lf);
238 		if (error || lf == NULL) {
239 			vput(vp);
240 			if (lf == NULL)
241 				error = ENODEV;
242 			return error;
243 		}
244 		lf->userrefs++;
245 		/* lookup again, see if the VFS was loaded */
246 		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
247 			if (!strcmp(vfsp->vfc_name, fstypename))
248 				break;
249 		if (vfsp == NULL) {
250 			lf->userrefs--;
251 			linker_file_unload(lf);
252 			vput(vp);
253 			return (ENODEV);
254 		}
255 	}
256 	lwkt_gettoken(&vp->v_interlock);
257 	if ((vp->v_flag & VMOUNT) != 0 ||
258 	    vp->v_mountedhere != NULL) {
259 		lwkt_reltoken(&vp->v_interlock);
260 		vput(vp);
261 		return (EBUSY);
262 	}
263 	vp->v_flag |= VMOUNT;
264 	lwkt_reltoken(&vp->v_interlock);
265 
266 	/*
267 	 * Allocate and initialize the filesystem.
268 	 */
269 	mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK);
270 	bzero((char *)mp, (u_long)sizeof(struct mount));
271 	TAILQ_INIT(&mp->mnt_nvnodelist);
272 	TAILQ_INIT(&mp->mnt_reservedvnlist);
273 	mp->mnt_nvnodelistsize = 0;
274 	lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE);
275 	(void)vfs_busy(mp, LK_NOWAIT, 0, td);
276 	mp->mnt_op = vfsp->vfc_vfsops;
277 	mp->mnt_vfc = vfsp;
278 	vfsp->vfc_refcount++;
279 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
280 	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
281 	strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
282 	mp->mnt_vnodecovered = vp;
283 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
284 	mp->mnt_iosize_max = DFLTPHYS;
285 	VOP_UNLOCK(vp, 0, td);
286 update:
287 	/*
288 	 * Set the mount level flags.
289 	 */
290 	if (SCARG(uap, flags) & MNT_RDONLY)
291 		mp->mnt_flag |= MNT_RDONLY;
292 	else if (mp->mnt_flag & MNT_RDONLY)
293 		mp->mnt_kern_flag |= MNTK_WANTRDWR;
294 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
295 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
296 	    MNT_NOSYMFOLLOW | MNT_IGNORE |
297 	    MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
298 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
299 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
300 	    MNT_NOSYMFOLLOW | MNT_IGNORE |
301 	    MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
302 	/*
303 	 * Mount the filesystem.
304 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
305 	 * get.  No freeing of cn_pnbuf.
306 	 */
307 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, td);
308 	if (mp->mnt_flag & MNT_UPDATE) {
309 		if (mp->mnt_kern_flag & MNTK_WANTRDWR)
310 			mp->mnt_flag &= ~MNT_RDONLY;
311 		mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
312 		mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
313 		if (error) {
314 			mp->mnt_flag = flag;
315 			mp->mnt_kern_flag = flag2;
316 		}
317 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
318 			if (mp->mnt_syncer == NULL)
319 				error = vfs_allocate_syncvnode(mp);
320 		} else {
321 			if (mp->mnt_syncer != NULL)
322 				vrele(mp->mnt_syncer);
323 			mp->mnt_syncer = NULL;
324 		}
325 		vfs_unbusy(mp, td);
326 		lwkt_gettoken(&vp->v_interlock);
327 		vp->v_flag &= ~VMOUNT;
328 		lwkt_reltoken(&vp->v_interlock);
329 		vrele(vp);
330 		return (error);
331 	}
332 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
333 	/*
334 	 * Put the new filesystem on the mount list after root.
335 	 */
336 	cache_purge(vp);
337 	if (!error) {
338 		lwkt_gettoken(&vp->v_interlock);
339 		vp->v_flag &= ~VMOUNT;
340 		vp->v_mountedhere = mp;
341 		lwkt_reltoken(&vp->v_interlock);
342 		lwkt_gettoken(&mountlist_token);
343 		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
344 		lwkt_reltoken(&mountlist_token);
345 		checkdirs(vp);
346 		VOP_UNLOCK(vp, 0, td);
347 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
348 			error = vfs_allocate_syncvnode(mp);
349 		vfs_unbusy(mp, td);
350 		if ((error = VFS_START(mp, 0, td)) != 0)
351 			vrele(vp);
352 	} else {
353 		lwkt_gettoken(&vp->v_interlock);
354 		vp->v_flag &= ~VMOUNT;
355 		lwkt_reltoken(&vp->v_interlock);
356 		mp->mnt_vfc->vfc_refcount--;
357 		vfs_unbusy(mp, td);
358 		free((caddr_t)mp, M_MOUNT);
359 		vput(vp);
360 	}
361 	return (error);
362 }
363 
364 /*
365  * Scan all active processes to see if any of them have a current
366  * or root directory onto which the new filesystem has just been
367  * mounted. If so, replace them with the new mount point.
368  */
369 static void
370 checkdirs(struct vnode *olddp)
371 {
372 	struct filedesc *fdp;
373 	struct vnode *newdp;
374 	struct proc *p;
375 
376 	if (olddp->v_usecount == 1)
377 		return;
378 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
379 		panic("mount: lost mount");
380 	FOREACH_PROC_IN_SYSTEM(p) {
381 		fdp = p->p_fd;
382 		if (fdp->fd_cdir == olddp) {
383 			vrele(fdp->fd_cdir);
384 			VREF(newdp);
385 			fdp->fd_cdir = newdp;
386 		}
387 		if (fdp->fd_rdir == olddp) {
388 			vrele(fdp->fd_rdir);
389 			VREF(newdp);
390 			fdp->fd_rdir = newdp;
391 		}
392 	}
393 	if (rootvnode == olddp) {
394 		vrele(rootvnode);
395 		VREF(newdp);
396 		rootvnode = newdp;
397 	}
398 	vput(newdp);
399 }
400 
401 /*
402  * Unmount a file system.
403  *
404  * Note: unmount takes a path to the vnode mounted on as argument,
405  * not special file (as before).
406  */
407 /*
408  * umount_args(char *path, int flags)
409  */
410 /* ARGSUSED */
411 int
412 unmount(struct unmount_args *uap)
413 {
414 	struct thread *td = curthread;
415 	struct proc *p = td->td_proc;
416 	struct vnode *vp;
417 	struct mount *mp;
418 	int error;
419 	struct nameidata nd;
420 
421 	KKASSERT(p);
422 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
423 	    SCARG(uap, path), td);
424 	if ((error = namei(&nd)) != 0)
425 		return (error);
426 	vp = nd.ni_vp;
427 	NDFREE(&nd, NDF_ONLY_PNBUF);
428 	mp = vp->v_mount;
429 
430 	/*
431 	 * Only root, or the user that did the original mount is
432 	 * permitted to unmount this filesystem.
433 	 */
434 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
435 	    (error = suser(td))) {
436 		vput(vp);
437 		return (error);
438 	}
439 
440 	/*
441 	 * Don't allow unmounting the root file system.
442 	 */
443 	if (mp->mnt_flag & MNT_ROOTFS) {
444 		vput(vp);
445 		return (EINVAL);
446 	}
447 
448 	/*
449 	 * Must be the root of the filesystem
450 	 */
451 	if ((vp->v_flag & VROOT) == 0) {
452 		vput(vp);
453 		return (EINVAL);
454 	}
455 	vput(vp);
456 	return (dounmount(mp, SCARG(uap, flags), td));
457 }
458 
459 /*
460  * Do the actual file system unmount.
461  */
462 int
463 dounmount(struct mount *mp, int flags, struct thread *td)
464 {
465 	struct vnode *coveredvp;
466 	int error;
467 	int async_flag;
468 
469 	lwkt_gettoken(&mountlist_token);
470 	if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
471 		lwkt_reltoken(&mountlist_token);
472 		return (EBUSY);
473 	}
474 	mp->mnt_kern_flag |= MNTK_UNMOUNT;
475 	/* Allow filesystems to detect that a forced unmount is in progress. */
476 	if (flags & MNT_FORCE)
477 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
478 	error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
479 	    ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_token, td);
480 	if (error) {
481 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
482 		if (mp->mnt_kern_flag & MNTK_MWAIT)
483 			wakeup((caddr_t)mp);
484 		return (error);
485 	}
486 
487 	if (mp->mnt_flag & MNT_EXPUBLIC)
488 		vfs_setpublicfs(NULL, NULL, NULL);
489 
490 	vfs_msync(mp, MNT_WAIT);
491 	async_flag = mp->mnt_flag & MNT_ASYNC;
492 	mp->mnt_flag &=~ MNT_ASYNC;
493 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
494 	if (mp->mnt_syncer != NULL)
495 		vrele(mp->mnt_syncer);
496 	if (((mp->mnt_flag & MNT_RDONLY) ||
497 	     (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
498 	    (flags & MNT_FORCE))
499 		error = VFS_UNMOUNT(mp, flags, td);
500 	lwkt_gettoken(&mountlist_token);
501 	if (error) {
502 		if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
503 			(void) vfs_allocate_syncvnode(mp);
504 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
505 		mp->mnt_flag |= async_flag;
506 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
507 		    &mountlist_token, td);
508 		if (mp->mnt_kern_flag & MNTK_MWAIT)
509 			wakeup((caddr_t)mp);
510 		return (error);
511 	}
512 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
513 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
514 		coveredvp->v_mountedhere = (struct mount *)0;
515 		vrele(coveredvp);
516 	}
517 	mp->mnt_vfc->vfc_refcount--;
518 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
519 		panic("unmount: dangling vnode");
520 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_token, td);
521 	if (mp->mnt_kern_flag & MNTK_MWAIT)
522 		wakeup((caddr_t)mp);
523 	free((caddr_t)mp, M_MOUNT);
524 	return (0);
525 }
526 
527 /*
528  * Sync each mounted filesystem.
529  */
530 
531 #ifdef DEBUG
532 static int syncprt = 0;
533 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
534 #endif
535 
536 /* ARGSUSED */
537 int
538 sync(struct sync_args *uap)
539 {
540 	struct thread *td = curthread;
541 	struct mount *mp, *nmp;
542 	int asyncflag;
543 
544 	lwkt_gettoken(&mountlist_token);
545 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
546 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_token, td)) {
547 			nmp = TAILQ_NEXT(mp, mnt_list);
548 			continue;
549 		}
550 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
551 			asyncflag = mp->mnt_flag & MNT_ASYNC;
552 			mp->mnt_flag &= ~MNT_ASYNC;
553 			vfs_msync(mp, MNT_NOWAIT);
554 			VFS_SYNC(mp, MNT_NOWAIT, td);
555 			mp->mnt_flag |= asyncflag;
556 		}
557 		lwkt_gettoken(&mountlist_token);
558 		nmp = TAILQ_NEXT(mp, mnt_list);
559 		vfs_unbusy(mp, td);
560 	}
561 	lwkt_reltoken(&mountlist_token);
562 #if 0
563 /*
564  * XXX don't call vfs_bufstats() yet because that routine
565  * was not imported in the Lite2 merge.
566  */
567 #ifdef DIAGNOSTIC
568 	if (syncprt)
569 		vfs_bufstats();
570 #endif /* DIAGNOSTIC */
571 #endif
572 	return (0);
573 }
574 
575 /* XXX PRISON: could be per prison flag */
576 static int prison_quotas;
577 #if 0
578 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
579 #endif
580 
581 /*
582  *  quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
583  *
584  * Change filesystem quotas.
585  */
586 /* ARGSUSED */
587 int
588 quotactl(struct quotactl_args *uap)
589 {
590 	struct thread *td = curthread;
591 	struct proc *p = td->td_proc;
592 	struct mount *mp;
593 	int error;
594 	struct nameidata nd;
595 
596 	KKASSERT(p);
597 	if (p->p_ucred->cr_prison && !prison_quotas)
598 		return (EPERM);
599 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
600 	if ((error = namei(&nd)) != 0)
601 		return (error);
602 	mp = nd.ni_vp->v_mount;
603 	NDFREE(&nd, NDF_ONLY_PNBUF);
604 	vrele(nd.ni_vp);
605 	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
606 	    SCARG(uap, arg), td));
607 }
608 
609 /*
610  * statfs_args(char *path, struct statfs *buf)
611  *
612  * Get filesystem statistics.
613  */
614 /* ARGSUSED */
615 int
616 statfs(struct statfs_args *uap)
617 {
618 	struct thread *td = curthread;
619 	struct mount *mp;
620 	struct statfs *sp;
621 	int error;
622 	struct nameidata nd;
623 	struct statfs sb;
624 
625 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
626 	if ((error = namei(&nd)) != 0)
627 		return (error);
628 	mp = nd.ni_vp->v_mount;
629 	sp = &mp->mnt_stat;
630 	NDFREE(&nd, NDF_ONLY_PNBUF);
631 	vrele(nd.ni_vp);
632 	error = VFS_STATFS(mp, sp, td);
633 	if (error)
634 		return (error);
635 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
636 	if (suser(td)) {
637 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
638 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
639 		sp = &sb;
640 	}
641 	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
642 }
643 
644 /*
645  * fstatfs_args(int fd, struct statfs *buf)
646  *
647  * Get filesystem statistics.
648  */
649 /* ARGSUSED */
650 int
651 fstatfs(struct fstatfs_args *uap)
652 {
653 	struct thread *td = curthread;
654 	struct proc *p = td->td_proc;
655 	struct file *fp;
656 	struct mount *mp;
657 	struct statfs *sp;
658 	int error;
659 	struct statfs sb;
660 
661 	KKASSERT(p);
662 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
663 		return (error);
664 	mp = ((struct vnode *)fp->f_data)->v_mount;
665 	if (mp == NULL)
666 		return (EBADF);
667 	sp = &mp->mnt_stat;
668 	error = VFS_STATFS(mp, sp, td);
669 	if (error)
670 		return (error);
671 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
672 	if (suser(td)) {
673 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
674 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
675 		sp = &sb;
676 	}
677 	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
678 }
679 
680 /*
681  * getfsstat_args(struct statfs *buf, long bufsize, int flags)
682  *
683  * Get statistics on all filesystems.
684  */
685 /* ARGSUSED */
686 int
687 getfsstat(struct getfsstat_args *uap)
688 {
689 	struct thread *td = curthread;
690 	struct mount *mp, *nmp;
691 	struct statfs *sp;
692 	caddr_t sfsp;
693 	long count, maxcount, error;
694 
695 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
696 	sfsp = (caddr_t)SCARG(uap, buf);
697 	count = 0;
698 	lwkt_gettoken(&mountlist_token);
699 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
700 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_token, td)) {
701 			nmp = TAILQ_NEXT(mp, mnt_list);
702 			continue;
703 		}
704 		if (sfsp && count < maxcount) {
705 			sp = &mp->mnt_stat;
706 			/*
707 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
708 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
709 			 * overrides MNT_WAIT.
710 			 */
711 			if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
712 			    (SCARG(uap, flags) & MNT_WAIT)) &&
713 			    (error = VFS_STATFS(mp, sp, td))) {
714 				lwkt_gettoken(&mountlist_token);
715 				nmp = TAILQ_NEXT(mp, mnt_list);
716 				vfs_unbusy(mp, td);
717 				continue;
718 			}
719 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
720 			error = copyout((caddr_t)sp, sfsp, sizeof(*sp));
721 			if (error) {
722 				vfs_unbusy(mp, td);
723 				return (error);
724 			}
725 			sfsp += sizeof(*sp);
726 		}
727 		count++;
728 		lwkt_gettoken(&mountlist_token);
729 		nmp = TAILQ_NEXT(mp, mnt_list);
730 		vfs_unbusy(mp, td);
731 	}
732 	lwkt_reltoken(&mountlist_token);
733 	if (sfsp && count > maxcount)
734 		uap->sysmsg_result = maxcount;
735 	else
736 		uap->sysmsg_result = count;
737 	return (0);
738 }
739 
740 /*
741  * fchdir_args(int fd)
742  *
743  * Change current working directory to a given file descriptor.
744  */
745 /* ARGSUSED */
746 int
747 fchdir(struct fchdir_args *uap)
748 {
749 	struct thread *td = curthread;
750 	struct proc *p = td->td_proc;
751 	struct filedesc *fdp = p->p_fd;
752 	struct vnode *vp, *tdp;
753 	struct mount *mp;
754 	struct file *fp;
755 	int error;
756 
757 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
758 		return (error);
759 	vp = (struct vnode *)fp->f_data;
760 	VREF(vp);
761 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
762 	if (vp->v_type != VDIR)
763 		error = ENOTDIR;
764 	else
765 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
766 	while (!error && (mp = vp->v_mountedhere) != NULL) {
767 		if (vfs_busy(mp, 0, 0, td))
768 			continue;
769 		error = VFS_ROOT(mp, &tdp);
770 		vfs_unbusy(mp, td);
771 		if (error)
772 			break;
773 		vput(vp);
774 		vp = tdp;
775 	}
776 	if (error) {
777 		vput(vp);
778 		return (error);
779 	}
780 	VOP_UNLOCK(vp, 0, td);
781 	vrele(fdp->fd_cdir);
782 	fdp->fd_cdir = vp;
783 	return (0);
784 }
785 
786 /*
787  * chdir_args(char *path)
788  *
789  * Change current working directory (``.'').
790  */
791 /* ARGSUSED */
792 int
793 chdir(struct chdir_args *uap)
794 {
795 	struct thread *td = curthread;
796 	struct proc *p = td->td_proc;
797 	struct filedesc *fdp = p->p_fd;
798 	int error;
799 	struct nameidata nd;
800 
801 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
802 	    SCARG(uap, path), td);
803 	if ((error = change_dir(&nd, td)) != 0)
804 		return (error);
805 	NDFREE(&nd, NDF_ONLY_PNBUF);
806 	vrele(fdp->fd_cdir);
807 	fdp->fd_cdir = nd.ni_vp;
808 	return (0);
809 }
810 
811 /*
812  * Helper function for raised chroot(2) security function:  Refuse if
813  * any filedescriptors are open directories.
814  */
815 static int
816 chroot_refuse_vdir_fds(fdp)
817 	struct filedesc *fdp;
818 {
819 	struct vnode *vp;
820 	struct file *fp;
821 	int error;
822 	int fd;
823 
824 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
825 		error = getvnode(fdp, fd, &fp);
826 		if (error)
827 			continue;
828 		vp = (struct vnode *)fp->f_data;
829 		if (vp->v_type != VDIR)
830 			continue;
831 		return(EPERM);
832 	}
833 	return (0);
834 }
835 
836 /*
837  * This sysctl determines if we will allow a process to chroot(2) if it
838  * has a directory open:
839  *	0: disallowed for all processes.
840  *	1: allowed for processes that were not already chroot(2)'ed.
841  *	2: allowed for all processes.
842  */
843 
844 static int chroot_allow_open_directories = 1;
845 
846 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
847      &chroot_allow_open_directories, 0, "");
848 
849 /*
850  * chroot_args(char *path)
851  *
852  * Change notion of root (``/'') directory.
853  */
854 /* ARGSUSED */
855 int
856 chroot(struct chroot_args *uap)
857 {
858 	struct thread *td = curthread;
859 	struct proc *p = td->td_proc;
860 	struct filedesc *fdp = p->p_fd;
861 	int error;
862 	struct nameidata nd;
863 
864 	KKASSERT(p);
865 	error = suser_cred(p->p_ucred, PRISON_ROOT);
866 	if (error)
867 		return (error);
868 	if (chroot_allow_open_directories == 0 ||
869 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode))
870 		error = chroot_refuse_vdir_fds(fdp);
871 	if (error)
872 		return (error);
873 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
874 	    SCARG(uap, path), td);
875 	if ((error = change_dir(&nd, td)) != 0)
876 		return (error);
877 	NDFREE(&nd, NDF_ONLY_PNBUF);
878 	vrele(fdp->fd_rdir);
879 	fdp->fd_rdir = nd.ni_vp;
880 	if (!fdp->fd_jdir) {
881 		fdp->fd_jdir = nd.ni_vp;
882                 VREF(fdp->fd_jdir);
883 	}
884 	return (0);
885 }
886 
887 /*
888  * Common routine for chroot and chdir.
889  */
890 static int
891 change_dir(struct nameidata *ndp, struct thread *td)
892 {
893 	struct vnode *vp;
894 	int error;
895 
896 	error = namei(ndp);
897 	if (error)
898 		return (error);
899 	vp = ndp->ni_vp;
900 	if (vp->v_type != VDIR)
901 		error = ENOTDIR;
902 	else
903 		error = VOP_ACCESS(vp, VEXEC, ndp->ni_cnd.cn_cred, td);
904 	if (error)
905 		vput(vp);
906 	else
907 		VOP_UNLOCK(vp, 0, td);
908 	return (error);
909 }
910 
911 /*
912  * open_args(char *path, int flags, int mode)
913  *
914  * Check permissions, allocate an open file structure,
915  * and call the device open routine if any.
916  */
917 int
918 open(struct open_args *uap)
919 {
920 	struct thread *td = curthread;
921 	struct proc *p = td->td_proc;
922 	struct filedesc *fdp = p->p_fd;
923 	struct file *fp;
924 	struct vnode *vp;
925 	int cmode, flags, oflags;
926 	struct file *nfp;
927 	int type, indx, error;
928 	struct flock lf;
929 	struct nameidata nd;
930 
931 	oflags = SCARG(uap, flags);
932 	if ((oflags & O_ACCMODE) == O_ACCMODE)
933 		return (EINVAL);
934 	flags = FFLAGS(oflags);
935 	error = falloc(p, &nfp, &indx);
936 	if (error)
937 		return (error);
938 	fp = nfp;
939 	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
940 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
941 	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
942 	/*
943 	 * Bump the ref count to prevent another process from closing
944 	 * the descriptor while we are blocked in vn_open()
945 	 */
946 	fhold(fp);
947 	error = vn_open(&nd, flags, cmode);
948 	if (error) {
949 		/*
950 		 * release our own reference
951 		 */
952 		fdrop(fp, td);
953 
954 		/*
955 		 * handle special fdopen() case.  bleh.  dupfdopen() is
956 		 * responsible for dropping the old contents of ofiles[indx]
957 		 * if it succeeds.
958 		 */
959 		if ((error == ENODEV || error == ENXIO) &&
960 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
961 		    (error =
962 			dupfdopen(fdp, indx, p->p_dupfd, flags, error)) == 0) {
963 			uap->sysmsg_result = indx;
964 			return (0);
965 		}
966 		/*
967 		 * Clean up the descriptor, but only if another thread hadn't
968 		 * replaced or closed it.
969 		 */
970 		if (fdp->fd_ofiles[indx] == fp) {
971 			fdp->fd_ofiles[indx] = NULL;
972 			fdrop(fp, td);
973 		}
974 
975 		if (error == ERESTART)
976 			error = EINTR;
977 		return (error);
978 	}
979 	p->p_dupfd = 0;
980 	NDFREE(&nd, NDF_ONLY_PNBUF);
981 	vp = nd.ni_vp;
982 
983 	/*
984 	 * There should be 2 references on the file, one from the descriptor
985 	 * table, and one for us.
986 	 *
987 	 * Handle the case where someone closed the file (via its file
988 	 * descriptor) while we were blocked.  The end result should look
989 	 * like opening the file succeeded but it was immediately closed.
990 	 */
991 	if (fp->f_count == 1) {
992 		KASSERT(fdp->fd_ofiles[indx] != fp,
993 		    ("Open file descriptor lost all refs"));
994 		VOP_UNLOCK(vp, 0, td);
995 		vn_close(vp, flags & FMASK, td);
996 		fdrop(fp, td);
997 		uap->sysmsg_result = indx;
998 		return 0;
999 	}
1000 
1001 	fp->f_data = (caddr_t)vp;
1002 	fp->f_flag = flags & FMASK;
1003 	fp->f_ops = &vnops;
1004 	fp->f_type = (vp->v_type == VFIFO ? DTYPE_FIFO : DTYPE_VNODE);
1005 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1006 		lf.l_whence = SEEK_SET;
1007 		lf.l_start = 0;
1008 		lf.l_len = 0;
1009 		if (flags & O_EXLOCK)
1010 			lf.l_type = F_WRLCK;
1011 		else
1012 			lf.l_type = F_RDLCK;
1013 		type = F_FLOCK;
1014 		if ((flags & FNONBLOCK) == 0)
1015 			type |= F_WAIT;
1016 		VOP_UNLOCK(vp, 0, td);
1017 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1018 			/*
1019 			 * lock request failed.  Normally close the descriptor
1020 			 * but handle the case where someone might have dup()d
1021 			 * it when we weren't looking.  One reference is
1022 			 * owned by the descriptor array, the other by us.
1023 			 */
1024 			if (fdp->fd_ofiles[indx] == fp) {
1025 				fdp->fd_ofiles[indx] = NULL;
1026 				fdrop(fp, td);
1027 			}
1028 			fdrop(fp, td);
1029 			return (error);
1030 		}
1031 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1032 		fp->f_flag |= FHASLOCK;
1033 	}
1034 	/* assert that vn_open created a backing object if one is needed */
1035 	KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1036 		("open: vmio vnode has no backing object after vn_open"));
1037 	VOP_UNLOCK(vp, 0, td);
1038 
1039 	/*
1040 	 * release our private reference, leaving the one associated with the
1041 	 * descriptor table intact.
1042 	 */
1043 	fdrop(fp, td);
1044 	uap->sysmsg_result = indx;
1045 	return (0);
1046 }
1047 
1048 #ifdef COMPAT_43
1049 /*
1050  * ocreat(char *path, int mode)
1051  *
1052  * Create a file.
1053  */
1054 int
1055 ocreat(struct ocreat_args *uap)
1056 {
1057 	struct open_args /* {
1058 		syscallarg(char *) path;
1059 		syscallarg(int) flags;
1060 		syscallarg(int) mode;
1061 	} */ nuap;
1062 
1063 	SCARG(&nuap, path) = SCARG(uap, path);
1064 	SCARG(&nuap, mode) = SCARG(uap, mode);
1065 	SCARG(&nuap, flags) = O_WRONLY | O_CREAT | O_TRUNC;
1066 	return (open(&nuap));
1067 }
1068 #endif /* COMPAT_43 */
1069 
1070 /*
1071  * mknod_args(char *path, int mode, int dev)
1072  *
1073  * Create a special file.
1074  */
1075 /* ARGSUSED */
1076 int
1077 mknod(struct mknod_args *uap)
1078 {
1079 	struct thread *td = curthread;
1080 	struct proc *p = td->td_proc;
1081 	struct vnode *vp;
1082 	struct vattr vattr;
1083 	int error;
1084 	int whiteout = 0;
1085 	struct nameidata nd;
1086 
1087 	KKASSERT(p);
1088 
1089 	switch (SCARG(uap, mode) & S_IFMT) {
1090 	case S_IFCHR:
1091 	case S_IFBLK:
1092 		error = suser(td);
1093 		break;
1094 	default:
1095 		error = suser_cred(p->p_ucred, PRISON_ROOT);
1096 		break;
1097 	}
1098 	if (error)
1099 		return (error);
1100 	bwillwrite();
1101 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
1102 	if ((error = namei(&nd)) != 0)
1103 		return (error);
1104 	vp = nd.ni_vp;
1105 	if (vp != NULL)
1106 		error = EEXIST;
1107 	else {
1108 		VATTR_NULL(&vattr);
1109 		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1110 		vattr.va_rdev = SCARG(uap, dev);
1111 		whiteout = 0;
1112 
1113 		switch (SCARG(uap, mode) & S_IFMT) {
1114 		case S_IFMT:	/* used by badsect to flag bad sectors */
1115 			vattr.va_type = VBAD;
1116 			break;
1117 		case S_IFCHR:
1118 			vattr.va_type = VCHR;
1119 			break;
1120 		case S_IFBLK:
1121 			vattr.va_type = VBLK;
1122 			break;
1123 		case S_IFWHT:
1124 			whiteout = 1;
1125 			break;
1126 		default:
1127 			error = EINVAL;
1128 			break;
1129 		}
1130 	}
1131 	if (!error) {
1132 		VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
1133 		if (whiteout)
1134 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1135 		else {
1136 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1137 						&nd.ni_cnd, &vattr);
1138 			if (error == 0)
1139 				vput(nd.ni_vp);
1140 		}
1141 		NDFREE(&nd, NDF_ONLY_PNBUF);
1142 		vput(nd.ni_dvp);
1143 	} else {
1144 		NDFREE(&nd, NDF_ONLY_PNBUF);
1145 		if (nd.ni_dvp == vp)
1146 			vrele(nd.ni_dvp);
1147 		else
1148 			vput(nd.ni_dvp);
1149 		if (vp)
1150 			vrele(vp);
1151 	}
1152 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mknod");
1153 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "mknod");
1154 	return (error);
1155 }
1156 
1157 /*
1158  * mkfifo_args(char *path, int mode)
1159  *
1160  * Create a named pipe.
1161  */
1162 /* ARGSUSED */
1163 int
1164 mkfifo(struct mkfifo_args *uap)
1165 {
1166 	struct thread *td = curthread;
1167 	struct proc *p = td->td_proc;
1168 	struct vattr vattr;
1169 	int error;
1170 	struct nameidata nd;
1171 
1172 	bwillwrite();
1173 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
1174 	if ((error = namei(&nd)) != 0)
1175 		return (error);
1176 	if (nd.ni_vp != NULL) {
1177 		NDFREE(&nd, NDF_ONLY_PNBUF);
1178 		if (nd.ni_dvp == nd.ni_vp)
1179 			vrele(nd.ni_dvp);
1180 		else
1181 			vput(nd.ni_dvp);
1182 		vrele(nd.ni_vp);
1183 		return (EEXIST);
1184 	}
1185 	VATTR_NULL(&vattr);
1186 	vattr.va_type = VFIFO;
1187 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1188 	VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
1189 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1190 	if (error == 0)
1191 		vput(nd.ni_vp);
1192 	NDFREE(&nd, NDF_ONLY_PNBUF);
1193 	vput(nd.ni_dvp);
1194 	return (error);
1195 }
1196 
1197 /*
1198  * link_args(char *path, char *link)
1199  *
1200  * Make a hard file link.
1201  */
1202 /* ARGSUSED */
1203 int
1204 link(struct link_args *uap)
1205 {
1206 	struct thread *td = curthread;
1207 	struct proc *p = td->td_proc;
1208 	struct vnode *vp;
1209 	struct nameidata nd;
1210 	int error;
1211 
1212 	bwillwrite();
1213 	NDINIT(&nd, LOOKUP, FOLLOW|NOOBJ, UIO_USERSPACE, SCARG(uap, path), td);
1214 	if ((error = namei(&nd)) != 0)
1215 		return (error);
1216 	NDFREE(&nd, NDF_ONLY_PNBUF);
1217 	vp = nd.ni_vp;
1218 	if (vp->v_type == VDIR)
1219 		error = EPERM;		/* POSIX */
1220 	else {
1221 		NDINIT(&nd, CREATE, LOCKPARENT|NOOBJ, UIO_USERSPACE, SCARG(uap, link), td);
1222 		error = namei(&nd);
1223 		if (!error) {
1224 			if (nd.ni_vp != NULL) {
1225 				if (nd.ni_vp)
1226 					vrele(nd.ni_vp);
1227 				error = EEXIST;
1228 			} else {
1229 				VOP_LEASE(nd.ni_dvp, td, p->p_ucred,
1230 				    LEASE_WRITE);
1231 				VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
1232 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1233 			}
1234 			NDFREE(&nd, NDF_ONLY_PNBUF);
1235 			if (nd.ni_dvp == nd.ni_vp)
1236 				vrele(nd.ni_dvp);
1237 			else
1238 				vput(nd.ni_dvp);
1239 		}
1240 	}
1241 	vrele(vp);
1242 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "link");
1243 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "link");
1244 	return (error);
1245 }
1246 
1247 /*
1248  * symlink(char *path, char *link)
1249  *
1250  * Make a symbolic link.
1251  */
1252 /* ARGSUSED */
1253 int
1254 symlink(struct symlink_args *uap)
1255 {
1256 	struct thread *td = curthread;
1257 	struct proc *p = td->td_proc;
1258 	struct vattr vattr;
1259 	char *path;
1260 	int error;
1261 	struct nameidata nd;
1262 
1263 	path = zalloc(namei_zone);
1264 	if ((error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL)) != 0)
1265 		goto out;
1266 	bwillwrite();
1267 	NDINIT(&nd, CREATE, LOCKPARENT|NOOBJ, UIO_USERSPACE, SCARG(uap, link), td);
1268 	if ((error = namei(&nd)) != 0)
1269 		goto out;
1270 	if (nd.ni_vp) {
1271 		NDFREE(&nd, NDF_ONLY_PNBUF);
1272 		if (nd.ni_dvp == nd.ni_vp)
1273 			vrele(nd.ni_dvp);
1274 		else
1275 			vput(nd.ni_dvp);
1276 		vrele(nd.ni_vp);
1277 		error = EEXIST;
1278 		goto out;
1279 	}
1280 	VATTR_NULL(&vattr);
1281 	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1282 	VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
1283 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1284 	NDFREE(&nd, NDF_ONLY_PNBUF);
1285 	if (error == 0)
1286 		vput(nd.ni_vp);
1287 	vput(nd.ni_dvp);
1288 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "symlink");
1289 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "symlink");
1290 out:
1291 	zfree(namei_zone, path);
1292 	return (error);
1293 }
1294 
1295 /*
1296  * undelete_args(char *path)
1297  *
1298  * Delete a whiteout from the filesystem.
1299  */
1300 /* ARGSUSED */
1301 int
1302 undelete(struct undelete_args *uap)
1303 {
1304 	struct thread *td = curthread;
1305 	struct proc *p = td->td_proc;
1306 	int error;
1307 	struct nameidata nd;
1308 
1309 	bwillwrite();
1310 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1311 	    SCARG(uap, path), td);
1312 	error = namei(&nd);
1313 	if (error)
1314 		return (error);
1315 
1316 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1317 		NDFREE(&nd, NDF_ONLY_PNBUF);
1318 		if (nd.ni_dvp == nd.ni_vp)
1319 			vrele(nd.ni_dvp);
1320 		else
1321 			vput(nd.ni_dvp);
1322 		if (nd.ni_vp)
1323 			vrele(nd.ni_vp);
1324 		return (EEXIST);
1325 	}
1326 
1327 	VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
1328 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1329 	NDFREE(&nd, NDF_ONLY_PNBUF);
1330 	vput(nd.ni_dvp);
1331 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "undelete");
1332 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "undelete");
1333 	return (error);
1334 }
1335 
1336 /*
1337  * unlink_args(char *path)
1338  *
1339  * Delete a name from the filesystem.
1340  */
1341 int
1342 unlink(struct unlink_args *uap)
1343 {
1344 	struct thread *td = curthread;
1345 	struct proc *p = td->td_proc;
1346 	struct vnode *vp;
1347 	int error;
1348 	struct nameidata nd;
1349 
1350 	bwillwrite();
1351 	NDINIT(&nd, DELETE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
1352 	if ((error = namei(&nd)) != 0)
1353 		return (error);
1354 	vp = nd.ni_vp;
1355 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
1356 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1357 
1358 	if (vp->v_type == VDIR)
1359 		error = EPERM;		/* POSIX */
1360 	else {
1361 		/*
1362 		 * The root of a mounted filesystem cannot be deleted.
1363 		 *
1364 		 * XXX: can this only be a VDIR case?
1365 		 */
1366 		if (vp->v_flag & VROOT)
1367 			error = EBUSY;
1368 	}
1369 
1370 	if (!error) {
1371 		VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
1372 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1373 	}
1374 	NDFREE(&nd, NDF_ONLY_PNBUF);
1375 	if (nd.ni_dvp == vp)
1376 		vrele(nd.ni_dvp);
1377 	else
1378 		vput(nd.ni_dvp);
1379 	if (vp != NULLVP)
1380 		vput(vp);
1381 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "unlink");
1382 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "unlink");
1383 	return (error);
1384 }
1385 
1386 /*
1387  * lseek_args(int fd, int pad, off_t offset, int whence)
1388  *
1389  * Reposition read/write file offset.
1390  */
1391 int
1392 lseek(struct lseek_args *uap)
1393 {
1394 	struct thread *td = curthread;
1395 	struct proc *p = td->td_proc;
1396 	struct filedesc *fdp = p->p_fd;
1397 	struct file *fp;
1398 	struct vattr vattr;
1399 	int error;
1400 
1401 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1402 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
1403 		return (EBADF);
1404 	if (fp->f_type != DTYPE_VNODE)
1405 		return (ESPIPE);
1406 	switch (SCARG(uap, whence)) {
1407 	case L_INCR:
1408 		fp->f_offset += SCARG(uap, offset);
1409 		break;
1410 	case L_XTND:
1411 		error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td);
1412 		if (error)
1413 			return (error);
1414 		fp->f_offset = SCARG(uap, offset) + vattr.va_size;
1415 		break;
1416 	case L_SET:
1417 		fp->f_offset = SCARG(uap, offset);
1418 		break;
1419 	default:
1420 		return (EINVAL);
1421 	}
1422 	uap->sysmsg_offset = fp->f_offset;
1423 	return (0);
1424 }
1425 
1426 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
1427 /*
1428  * Reposition read/write file offset.
1429  *
1430  * olseek_args(int fd, long offset, int whence)
1431  */
1432 int
1433 olseek(struct olseek_args *uap)
1434 {
1435 	struct lseek_args /* {
1436 		syscallarg(int) fd;
1437 		syscallarg(int) pad;
1438 		syscallarg(off_t) offset;
1439 		syscallarg(int) whence;
1440 	} */ nuap;
1441 	int error;
1442 
1443 	SCARG(&nuap, fd) = SCARG(uap, fd);
1444 	SCARG(&nuap, offset) = SCARG(uap, offset);
1445 	SCARG(&nuap, whence) = SCARG(uap, whence);
1446 	error = lseek(&nuap);
1447 	return (error);
1448 }
1449 #endif /* COMPAT_43 */
1450 
1451 /*
1452  * access_args(char *path, int flags)
1453  *
1454  * Check access permissions.
1455  */
1456 int
1457 access(struct access_args *uap)
1458 {
1459 	struct thread *td = curthread;
1460 	struct proc *p = td->td_proc;
1461 	struct ucred *cred, *tmpcred;
1462 	struct vnode *vp;
1463 	int error, flags;
1464 	struct nameidata nd;
1465 
1466 	cred = p->p_ucred;
1467 	/*
1468 	 * Create and modify a temporary credential instead of one that
1469 	 * is potentially shared.  This could also mess up socket
1470 	 * buffer accounting which can run in an interrupt context.
1471 	 */
1472 	tmpcred = crdup(cred);
1473 	tmpcred->cr_uid = p->p_ucred->cr_ruid;
1474 	tmpcred->cr_groups[0] = p->p_ucred->cr_rgid;
1475 	p->p_ucred = tmpcred;
1476 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1477 	    SCARG(uap, path), td);
1478 	if ((error = namei(&nd)) != 0)
1479 		goto out1;
1480 	vp = nd.ni_vp;
1481 
1482 	/* Flags == 0 means only check for existence. */
1483 	if (SCARG(uap, flags)) {
1484 		flags = 0;
1485 		if (SCARG(uap, flags) & R_OK)
1486 			flags |= VREAD;
1487 		if (SCARG(uap, flags) & W_OK)
1488 			flags |= VWRITE;
1489 		if (SCARG(uap, flags) & X_OK)
1490 			flags |= VEXEC;
1491 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1492 			error = VOP_ACCESS(vp, flags, tmpcred, td);
1493 	}
1494 	NDFREE(&nd, NDF_ONLY_PNBUF);
1495 	vput(vp);
1496 out1:
1497 	p->p_ucred = cred;
1498 	crfree(tmpcred);
1499 	return (error);
1500 }
1501 
1502 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
1503 /*
1504  * ostat_args(char *path, struct ostat *ub)
1505  *
1506  * Get file status; this version follows links.
1507  */
1508 /* ARGSUSED */
1509 int
1510 ostat(struct ostat_args *uap)
1511 {
1512 	struct thread *td = curthread;
1513 	struct stat sb;
1514 	struct ostat osb;
1515 	int error;
1516 	struct nameidata nd;
1517 
1518 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1519 	    SCARG(uap, path), td);
1520 	if ((error = namei(&nd)) != 0)
1521 		return (error);
1522 	NDFREE(&nd, NDF_ONLY_PNBUF);
1523 	error = vn_stat(nd.ni_vp, &sb, td);
1524 	vput(nd.ni_vp);
1525 	if (error)
1526 		return (error);
1527 	cvtstat(&sb, &osb);
1528 	error = copyout((caddr_t)&osb, (caddr_t)SCARG(uap, ub), sizeof (osb));
1529 	return (error);
1530 }
1531 
1532 /*
1533  * olstat_args(char *path, struct ostat *ub)
1534  *
1535  * Get file status; this version does not follow links.
1536  */
1537 /* ARGSUSED */
1538 int
1539 olstat(struct olstat_args *uap)
1540 {
1541 	struct thread *td = curthread;
1542 	struct vnode *vp;
1543 	struct stat sb;
1544 	struct ostat osb;
1545 	int error;
1546 	struct nameidata nd;
1547 
1548 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1549 	    SCARG(uap, path), td);
1550 	if ((error = namei(&nd)) != 0)
1551 		return (error);
1552 	vp = nd.ni_vp;
1553 	error = vn_stat(vp, &sb, td);
1554 	NDFREE(&nd, NDF_ONLY_PNBUF);
1555 	vput(vp);
1556 	if (error)
1557 		return (error);
1558 	cvtstat(&sb, &osb);
1559 	error = copyout((caddr_t)&osb, (caddr_t)SCARG(uap, ub), sizeof (osb));
1560 	return (error);
1561 }
1562 
1563 /*
1564  * Convert from an old to a new stat structure.
1565  */
1566 void
1567 cvtstat(st, ost)
1568 	struct stat *st;
1569 	struct ostat *ost;
1570 {
1571 	ost->st_dev = st->st_dev;
1572 	ost->st_ino = st->st_ino;
1573 	ost->st_mode = st->st_mode;
1574 	ost->st_nlink = st->st_nlink;
1575 	ost->st_uid = st->st_uid;
1576 	ost->st_gid = st->st_gid;
1577 	ost->st_rdev = st->st_rdev;
1578 	if (st->st_size < (quad_t)1 << 32)
1579 		ost->st_size = st->st_size;
1580 	else
1581 		ost->st_size = -2;
1582 	ost->st_atime = st->st_atime;
1583 	ost->st_mtime = st->st_mtime;
1584 	ost->st_ctime = st->st_ctime;
1585 	ost->st_blksize = st->st_blksize;
1586 	ost->st_blocks = st->st_blocks;
1587 	ost->st_flags = st->st_flags;
1588 	ost->st_gen = st->st_gen;
1589 }
1590 #endif /* COMPAT_43 || COMPAT_SUNOS */
1591 
1592 /*
1593  * stat_args(char *path, struct stat *ub)
1594  *
1595  * Get file status; this version follows links.
1596  */
1597 /* ARGSUSED */
1598 int
1599 stat(struct stat_args *uap)
1600 {
1601 	struct thread *td = curthread;
1602 	struct stat sb;
1603 	int error;
1604 	struct nameidata nd;
1605 
1606 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1607 	    SCARG(uap, path), td);
1608 	if ((error = namei(&nd)) != 0)
1609 		return (error);
1610 	error = vn_stat(nd.ni_vp, &sb, td);
1611 	NDFREE(&nd, NDF_ONLY_PNBUF);
1612 	vput(nd.ni_vp);
1613 	if (error)
1614 		return (error);
1615 	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1616 	return (error);
1617 }
1618 
1619 /*
1620  * lstat_args(char *path, struct stat *ub)
1621  *
1622  * Get file status; this version does not follow links.
1623  */
1624 /* ARGSUSED */
1625 int
1626 lstat(struct lstat_args *uap)
1627 {
1628 	struct thread *td = curthread;
1629 	int error;
1630 	struct vnode *vp;
1631 	struct stat sb;
1632 	struct nameidata nd;
1633 
1634 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1635 	    SCARG(uap, path), td);
1636 	if ((error = namei(&nd)) != 0)
1637 		return (error);
1638 	vp = nd.ni_vp;
1639 	error = vn_stat(vp, &sb, td);
1640 	NDFREE(&nd, NDF_ONLY_PNBUF);
1641 	vput(vp);
1642 	if (error)
1643 		return (error);
1644 	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1645 	return (error);
1646 }
1647 
1648 void
1649 cvtnstat(sb, nsb)
1650 	struct stat *sb;
1651 	struct nstat *nsb;
1652 {
1653 	nsb->st_dev = sb->st_dev;
1654 	nsb->st_ino = sb->st_ino;
1655 	nsb->st_mode = sb->st_mode;
1656 	nsb->st_nlink = sb->st_nlink;
1657 	nsb->st_uid = sb->st_uid;
1658 	nsb->st_gid = sb->st_gid;
1659 	nsb->st_rdev = sb->st_rdev;
1660 	nsb->st_atimespec = sb->st_atimespec;
1661 	nsb->st_mtimespec = sb->st_mtimespec;
1662 	nsb->st_ctimespec = sb->st_ctimespec;
1663 	nsb->st_size = sb->st_size;
1664 	nsb->st_blocks = sb->st_blocks;
1665 	nsb->st_blksize = sb->st_blksize;
1666 	nsb->st_flags = sb->st_flags;
1667 	nsb->st_gen = sb->st_gen;
1668 	nsb->st_qspare[0] = sb->st_qspare[0];
1669 	nsb->st_qspare[1] = sb->st_qspare[1];
1670 }
1671 
1672 /*
1673  * nstat_args(char *path, struct nstat *ub)
1674  */
1675 /* ARGSUSED */
1676 int
1677 nstat(struct nstat_args *uap)
1678 {
1679 	struct thread *td = curthread;
1680 	struct stat sb;
1681 	struct nstat nsb;
1682 	int error;
1683 	struct nameidata nd;
1684 
1685 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1686 	    SCARG(uap, path), td);
1687 	if ((error = namei(&nd)) != 0)
1688 		return (error);
1689 	NDFREE(&nd, NDF_ONLY_PNBUF);
1690 	error = vn_stat(nd.ni_vp, &sb, td);
1691 	vput(nd.ni_vp);
1692 	if (error)
1693 		return (error);
1694 	cvtnstat(&sb, &nsb);
1695 	error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb));
1696 	return (error);
1697 }
1698 
1699 /*
1700  * lstat_args(char *path, struct stat *ub)
1701  *
1702  * Get file status; this version does not follow links.
1703  */
1704 /* ARGSUSED */
1705 int
1706 nlstat(struct nlstat_args *uap)
1707 {
1708 	struct thread *td = curthread;
1709 	int error;
1710 	struct vnode *vp;
1711 	struct stat sb;
1712 	struct nstat nsb;
1713 	struct nameidata nd;
1714 
1715 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1716 	    SCARG(uap, path), td);
1717 	if ((error = namei(&nd)) != 0)
1718 		return (error);
1719 	vp = nd.ni_vp;
1720 	NDFREE(&nd, NDF_ONLY_PNBUF);
1721 	error = vn_stat(vp, &sb, td);
1722 	vput(vp);
1723 	if (error)
1724 		return (error);
1725 	cvtnstat(&sb, &nsb);
1726 	error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb));
1727 	return (error);
1728 }
1729 
1730 /*
1731  * pathconf_Args(char *path, int name)
1732  *
1733  * Get configurable pathname variables.
1734  */
1735 /* ARGSUSED */
1736 int
1737 pathconf(struct pathconf_args *uap)
1738 {
1739 	struct thread *td = curthread;
1740 	int error;
1741 	struct nameidata nd;
1742 
1743 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1744 	    SCARG(uap, path), td);
1745 	if ((error = namei(&nd)) != 0)
1746 		return (error);
1747 	NDFREE(&nd, NDF_ONLY_PNBUF);
1748 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), uap->sysmsg_fds);
1749 	vput(nd.ni_vp);
1750 	return (error);
1751 }
1752 
1753 /*
1754  * readlink_args(char *path, char *buf, int count)
1755  *
1756  * Return target name of a symbolic link.
1757  */
1758 /* ARGSUSED */
1759 int
1760 readlink(struct readlink_args *uap)
1761 {
1762 	struct thread *td = curthread;
1763 	struct proc *p = td->td_proc;
1764 	struct vnode *vp;
1765 	struct iovec aiov;
1766 	struct uio auio;
1767 	int error;
1768 	struct nameidata nd;
1769 
1770 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1771 	    SCARG(uap, path), td);
1772 	if ((error = namei(&nd)) != 0)
1773 		return (error);
1774 	NDFREE(&nd, NDF_ONLY_PNBUF);
1775 	vp = nd.ni_vp;
1776 	if (vp->v_type != VLNK)
1777 		error = EINVAL;
1778 	else {
1779 		aiov.iov_base = SCARG(uap, buf);
1780 		aiov.iov_len = SCARG(uap, count);
1781 		auio.uio_iov = &aiov;
1782 		auio.uio_iovcnt = 1;
1783 		auio.uio_offset = 0;
1784 		auio.uio_rw = UIO_READ;
1785 		auio.uio_segflg = UIO_USERSPACE;
1786 		auio.uio_td = td;
1787 		auio.uio_resid = SCARG(uap, count);
1788 		error = VOP_READLINK(vp, &auio, p->p_ucred);
1789 	}
1790 	vput(vp);
1791 	uap->sysmsg_result = SCARG(uap, count) - auio.uio_resid;
1792 	return (error);
1793 }
1794 
1795 static int
1796 setfflags(struct vnode *vp, int flags)
1797 {
1798 	struct thread *td = curthread;
1799 	struct proc *p = td->td_proc;
1800 	int error;
1801 	struct vattr vattr;
1802 
1803 	/*
1804 	 * Prevent non-root users from setting flags on devices.  When
1805 	 * a device is reused, users can retain ownership of the device
1806 	 * if they are allowed to set flags and programs assume that
1807 	 * chown can't fail when done as root.
1808 	 */
1809 	if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
1810 	    ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
1811 		return (error);
1812 
1813 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
1814 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1815 	VATTR_NULL(&vattr);
1816 	vattr.va_flags = flags;
1817 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
1818 	VOP_UNLOCK(vp, 0, td);
1819 	return (error);
1820 }
1821 
1822 /*
1823  * chflags(char *path, int flags)
1824  *
1825  * Change flags of a file given a path name.
1826  */
1827 /* ARGSUSED */
1828 int
1829 chflags(struct chflags_args *uap)
1830 {
1831 	struct thread *td = curthread;
1832 	int error;
1833 	struct nameidata nd;
1834 
1835 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
1836 	if ((error = namei(&nd)) != 0)
1837 		return (error);
1838 	NDFREE(&nd, NDF_ONLY_PNBUF);
1839 	error = setfflags(nd.ni_vp, SCARG(uap, flags));
1840 	vrele(nd.ni_vp);
1841 	return error;
1842 }
1843 
1844 /*
1845  * fchflags_args(int fd, int flags)
1846  *
1847  * Change flags of a file given a file descriptor.
1848  */
1849 /* ARGSUSED */
1850 int
1851 fchflags(struct fchflags_args *uap)
1852 {
1853 	struct thread *td = curthread;
1854 	struct proc *p = td->td_proc;
1855 	struct file *fp;
1856 	int error;
1857 
1858 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1859 		return (error);
1860 	return setfflags((struct vnode *) fp->f_data, SCARG(uap, flags));
1861 }
1862 
1863 static int
1864 setfmode(struct vnode *vp, int mode)
1865 {
1866 	struct thread *td = curthread;
1867 	struct proc *p = td->td_proc;
1868 	int error;
1869 	struct vattr vattr;
1870 
1871 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
1872 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1873 	VATTR_NULL(&vattr);
1874 	vattr.va_mode = mode & ALLPERMS;
1875 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
1876 	VOP_UNLOCK(vp, 0, td);
1877 	return error;
1878 }
1879 
1880 /*
1881  * chmod_args(char *path, int mode)
1882  *
1883  * Change mode of a file given path name.
1884  */
1885 /* ARGSUSED */
1886 int
1887 chmod(struct chmod_args *uap)
1888 {
1889 	struct thread *td = curthread;
1890 	int error;
1891 	struct nameidata nd;
1892 
1893 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
1894 	if ((error = namei(&nd)) != 0)
1895 		return (error);
1896 	NDFREE(&nd, NDF_ONLY_PNBUF);
1897 	error = setfmode(nd.ni_vp, SCARG(uap, mode));
1898 	vrele(nd.ni_vp);
1899 	return error;
1900 }
1901 
1902 /*
1903  * lchmod_args(char *path, int mode)
1904  *
1905  * Change mode of a file given path name (don't follow links.)
1906  */
1907 /* ARGSUSED */
1908 int
1909 lchmod(struct lchmod_args *uap)
1910 {
1911 	struct thread *td = curthread;
1912 	int error;
1913 	struct nameidata nd;
1914 
1915 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
1916 	if ((error = namei(&nd)) != 0)
1917 		return (error);
1918 	NDFREE(&nd, NDF_ONLY_PNBUF);
1919 	error = setfmode(nd.ni_vp, SCARG(uap, mode));
1920 	vrele(nd.ni_vp);
1921 	return error;
1922 }
1923 
1924 /*
1925  * fchmod_args(int fd, int mode)
1926  *
1927  * Change mode of a file given a file descriptor.
1928  */
1929 /* ARGSUSED */
1930 int
1931 fchmod(struct fchmod_args *uap)
1932 {
1933 	struct thread *td = curthread;
1934 	struct proc *p = td->td_proc;
1935 	struct file *fp;
1936 	int error;
1937 
1938 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1939 		return (error);
1940 	return setfmode((struct vnode *)fp->f_data, SCARG(uap, mode));
1941 }
1942 
1943 static int
1944 setfown(struct vnode *vp, uid_t uid, gid_t gid)
1945 {
1946 	struct thread *td = curthread;
1947 	struct proc *p = td->td_proc;
1948 	int error;
1949 	struct vattr vattr;
1950 
1951 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
1952 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1953 	VATTR_NULL(&vattr);
1954 	vattr.va_uid = uid;
1955 	vattr.va_gid = gid;
1956 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
1957 	VOP_UNLOCK(vp, 0, td);
1958 	return error;
1959 }
1960 
1961 /*
1962  * chown(char *path, int uid, int gid)
1963  *
1964  * Set ownership given a path name.
1965  */
1966 /* ARGSUSED */
1967 int
1968 chown(struct chown_args *uap)
1969 {
1970 	struct thread *td = curthread;
1971 	int error;
1972 	struct nameidata nd;
1973 
1974 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
1975 	if ((error = namei(&nd)) != 0)
1976 		return (error);
1977 	NDFREE(&nd, NDF_ONLY_PNBUF);
1978 	error = setfown(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid));
1979 	vrele(nd.ni_vp);
1980 	return (error);
1981 }
1982 
1983 /*
1984  * lchown_args(char *path, int uid, int gid)
1985  *
1986  * Set ownership given a path name, do not cross symlinks.
1987  */
1988 /* ARGSUSED */
1989 int
1990 lchown(struct lchown_args *uap)
1991 {
1992 	struct thread *td = curthread;
1993 	int error;
1994 	struct nameidata nd;
1995 
1996 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
1997 	if ((error = namei(&nd)) != 0)
1998 		return (error);
1999 	NDFREE(&nd, NDF_ONLY_PNBUF);
2000 	error = setfown(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid));
2001 	vrele(nd.ni_vp);
2002 	return (error);
2003 }
2004 
2005 /*
2006  * fchown_args(int fd, int uid, int gid)
2007  *
2008  * Set ownership given a file descriptor.
2009  */
2010 /* ARGSUSED */
2011 int
2012 fchown(struct fchown_args *uap)
2013 {
2014 	struct thread *td = curthread;
2015 	struct proc *p = td->td_proc;
2016 	struct file *fp;
2017 	int error;
2018 
2019 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2020 		return (error);
2021 	return setfown((struct vnode *)fp->f_data,
2022 		SCARG(uap, uid), SCARG(uap, gid));
2023 }
2024 
2025 static int
2026 getutimes(const struct timeval *usrtvp, struct timespec *tsp)
2027 {
2028 	struct timeval tv[2];
2029 	int error;
2030 
2031 	if (usrtvp == NULL) {
2032 		microtime(&tv[0]);
2033 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2034 		tsp[1] = tsp[0];
2035 	} else {
2036 		if ((error = copyin(usrtvp, tv, sizeof (tv))) != 0)
2037 			return (error);
2038 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2039 		TIMEVAL_TO_TIMESPEC(&tv[1], &tsp[1]);
2040 	}
2041 	return 0;
2042 }
2043 
2044 static int
2045 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
2046 {
2047 	struct thread *td = curthread;
2048 	struct proc *p = td->td_proc;
2049 	int error;
2050 	struct vattr vattr;
2051 
2052 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2053 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2054 	VATTR_NULL(&vattr);
2055 	vattr.va_atime = ts[0];
2056 	vattr.va_mtime = ts[1];
2057 	if (nullflag)
2058 		vattr.va_vaflags |= VA_UTIMES_NULL;
2059 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2060 	VOP_UNLOCK(vp, 0, td);
2061 	return error;
2062 }
2063 
2064 /*
2065  * utimes_args(char *path, struct timeval *tptr)
2066  *
2067  * Set the access and modification times of a file.
2068  */
2069 /* ARGSUSED */
2070 int
2071 utimes(struct utimes_args *uap)
2072 {
2073 	struct thread *td = curthread;
2074 	struct timespec ts[2];
2075 	struct timeval *usrtvp;
2076 	int error;
2077 	struct nameidata nd;
2078 
2079 	usrtvp = SCARG(uap, tptr);
2080 	if ((error = getutimes(usrtvp, ts)) != 0)
2081 		return (error);
2082 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2083 	if ((error = namei(&nd)) != 0)
2084 		return (error);
2085 	NDFREE(&nd, NDF_ONLY_PNBUF);
2086 	error = setutimes(nd.ni_vp, ts, usrtvp == NULL);
2087 	vrele(nd.ni_vp);
2088 	return (error);
2089 }
2090 
2091 /*
2092  * lutimes_args(char *path, struct timeval *tptr)
2093  *
2094  * Set the access and modification times of a file.
2095  */
2096 /* ARGSUSED */
2097 int
2098 lutimes(struct lutimes_args *uap)
2099 {
2100 	struct thread *td = curthread;
2101 	struct timespec ts[2];
2102 	struct timeval *usrtvp;
2103 	int error;
2104 	struct nameidata nd;
2105 
2106 	usrtvp = SCARG(uap, tptr);
2107 	if ((error = getutimes(usrtvp, ts)) != 0)
2108 		return (error);
2109 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2110 	if ((error = namei(&nd)) != 0)
2111 		return (error);
2112 	NDFREE(&nd, NDF_ONLY_PNBUF);
2113 	error = setutimes(nd.ni_vp, ts, usrtvp == NULL);
2114 	vrele(nd.ni_vp);
2115 	return (error);
2116 }
2117 
2118 /*
2119  * futimes_args(int fd, struct timeval *tptr)
2120  *
2121  * Set the access and modification times of a file.
2122  */
2123 /* ARGSUSED */
2124 int
2125 futimes(struct futimes_args *uap)
2126 {
2127 	struct thread *td = curthread;
2128 	struct proc *p = td->td_proc;
2129 	struct timespec ts[2];
2130 	struct file *fp;
2131 	struct timeval *usrtvp;
2132 	int error;
2133 
2134 	usrtvp = SCARG(uap, tptr);
2135 	if ((error = getutimes(usrtvp, ts)) != 0)
2136 		return (error);
2137 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2138 		return (error);
2139 	return setutimes((struct vnode *)fp->f_data, ts, usrtvp == NULL);
2140 }
2141 
2142 /*
2143  * truncate(char *path, int pad, off_t length)
2144  *
2145  * Truncate a file given its path name.
2146  */
2147 /* ARGSUSED */
2148 int
2149 truncate(struct truncate_args *uap)
2150 {
2151 	struct thread *td = curthread;
2152 	struct proc *p = td->td_proc;
2153 	struct vnode *vp;
2154 	struct vattr vattr;
2155 	int error;
2156 	struct nameidata nd;
2157 
2158 	if (uap->length < 0)
2159 		return(EINVAL);
2160 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2161 	if ((error = namei(&nd)) != 0)
2162 		return (error);
2163 	vp = nd.ni_vp;
2164 	NDFREE(&nd, NDF_ONLY_PNBUF);
2165 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2166 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2167 	if (vp->v_type == VDIR)
2168 		error = EISDIR;
2169 	else if ((error = vn_writechk(vp)) == 0 &&
2170 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, td)) == 0) {
2171 		VATTR_NULL(&vattr);
2172 		vattr.va_size = SCARG(uap, length);
2173 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2174 	}
2175 	vput(vp);
2176 	return (error);
2177 }
2178 
2179 /*
2180  * ftruncate_args(int fd, int pad, off_t length)
2181  *
2182  * Truncate a file given a file descriptor.
2183  */
2184 /* ARGSUSED */
2185 int
2186 ftruncate(struct ftruncate_args *uap)
2187 {
2188 	struct thread *td = curthread;
2189 	struct proc *p = td->td_proc;
2190 	struct vattr vattr;
2191 	struct vnode *vp;
2192 	struct file *fp;
2193 	int error;
2194 
2195 	if (uap->length < 0)
2196 		return(EINVAL);
2197 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2198 		return (error);
2199 	if ((fp->f_flag & FWRITE) == 0)
2200 		return (EINVAL);
2201 	vp = (struct vnode *)fp->f_data;
2202 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2203 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2204 	if (vp->v_type == VDIR)
2205 		error = EISDIR;
2206 	else if ((error = vn_writechk(vp)) == 0) {
2207 		VATTR_NULL(&vattr);
2208 		vattr.va_size = SCARG(uap, length);
2209 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
2210 	}
2211 	VOP_UNLOCK(vp, 0, td);
2212 	return (error);
2213 }
2214 
2215 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
2216 /*
2217  * otruncate_args(char *path, long length)
2218  *
2219  * Truncate a file given its path name.
2220  */
2221 /* ARGSUSED */
2222 int
2223 otruncate(struct otruncate_args *uap)
2224 {
2225 	struct truncate_args /* {
2226 		syscallarg(char *) path;
2227 		syscallarg(int) pad;
2228 		syscallarg(off_t) length;
2229 	} */ nuap;
2230 
2231 	SCARG(&nuap, path) = SCARG(uap, path);
2232 	SCARG(&nuap, length) = SCARG(uap, length);
2233 	return (truncate(&nuap));
2234 }
2235 
2236 /*
2237  * oftruncate_args(int fd, long length)
2238  *
2239  * Truncate a file given a file descriptor.
2240  */
2241 /* ARGSUSED */
2242 int
2243 oftruncate(struct oftruncate_args *uap)
2244 {
2245 	struct ftruncate_args /* {
2246 		syscallarg(int) fd;
2247 		syscallarg(int) pad;
2248 		syscallarg(off_t) length;
2249 	} */ nuap;
2250 
2251 	SCARG(&nuap, fd) = SCARG(uap, fd);
2252 	SCARG(&nuap, length) = SCARG(uap, length);
2253 	return (ftruncate(&nuap));
2254 }
2255 #endif /* COMPAT_43 || COMPAT_SUNOS */
2256 
2257 /*
2258  * fsync(int fd)
2259  *
2260  * Sync an open file.
2261  */
2262 /* ARGSUSED */
2263 int
2264 fsync(struct fsync_args *uap)
2265 {
2266 	struct thread *td = curthread;
2267 	struct proc *p = td->td_proc;
2268 	struct vnode *vp;
2269 	struct file *fp;
2270 	vm_object_t obj;
2271 	int error;
2272 
2273 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2274 		return (error);
2275 	vp = (struct vnode *)fp->f_data;
2276 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2277 	if (VOP_GETVOBJECT(vp, &obj) == 0)
2278 		vm_object_page_clean(obj, 0, 0, 0);
2279 	if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 &&
2280 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
2281 	    bioops.io_fsync)
2282 		error = (*bioops.io_fsync)(vp);
2283 	VOP_UNLOCK(vp, 0, td);
2284 	return (error);
2285 }
2286 
2287 /*
2288  * rename_args(char *from, char *to)
2289  *
2290  * Rename files.  Source and destination must either both be directories,
2291  * or both not be directories.  If target is a directory, it must be empty.
2292  */
2293 /* ARGSUSED */
2294 int
2295 rename(struct rename_args *uap)
2296 {
2297 	struct thread *td = curthread;
2298 	struct proc *p = td->td_proc;
2299 	struct vnode *tvp, *fvp, *tdvp;
2300 	struct nameidata fromnd, tond;
2301 	int error;
2302 
2303 	bwillwrite();
2304 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2305 	    SCARG(uap, from), td);
2306 	if ((error = namei(&fromnd)) != 0)
2307 		return (error);
2308 	fvp = fromnd.ni_vp;
2309 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | NOOBJ,
2310 	    UIO_USERSPACE, SCARG(uap, to), td);
2311 	if (fromnd.ni_vp->v_type == VDIR)
2312 		tond.ni_cnd.cn_flags |= WILLBEDIR;
2313 	if ((error = namei(&tond)) != 0) {
2314 		/* Translate error code for rename("dir1", "dir2/."). */
2315 		if (error == EISDIR && fvp->v_type == VDIR)
2316 			error = EINVAL;
2317 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2318 		vrele(fromnd.ni_dvp);
2319 		vrele(fvp);
2320 		goto out1;
2321 	}
2322 	tdvp = tond.ni_dvp;
2323 	tvp = tond.ni_vp;
2324 	if (tvp != NULL) {
2325 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2326 			error = ENOTDIR;
2327 			goto out;
2328 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2329 			error = EISDIR;
2330 			goto out;
2331 		}
2332 	}
2333 	if (fvp == tdvp)
2334 		error = EINVAL;
2335 	/*
2336 	 * If the source is the same as the destination (that is, if they
2337 	 * are links to the same vnode), then there is nothing to do.
2338 	 */
2339 	if (fvp == tvp)
2340 		error = -1;
2341 out:
2342 	if (!error) {
2343 		VOP_LEASE(tdvp, td, p->p_ucred, LEASE_WRITE);
2344 		if (fromnd.ni_dvp != tdvp) {
2345 			VOP_LEASE(fromnd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
2346 		}
2347 		if (tvp) {
2348 			VOP_LEASE(tvp, td, p->p_ucred, LEASE_WRITE);
2349 		}
2350 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2351 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2352 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2353 		NDFREE(&tond, NDF_ONLY_PNBUF);
2354 	} else {
2355 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2356 		NDFREE(&tond, NDF_ONLY_PNBUF);
2357 		if (tdvp == tvp)
2358 			vrele(tdvp);
2359 		else
2360 			vput(tdvp);
2361 		if (tvp)
2362 			vput(tvp);
2363 		vrele(fromnd.ni_dvp);
2364 		vrele(fvp);
2365 	}
2366 	vrele(tond.ni_startdir);
2367 	ASSERT_VOP_UNLOCKED(fromnd.ni_dvp, "rename");
2368 	ASSERT_VOP_UNLOCKED(fromnd.ni_vp, "rename");
2369 	ASSERT_VOP_UNLOCKED(tond.ni_dvp, "rename");
2370 	ASSERT_VOP_UNLOCKED(tond.ni_vp, "rename");
2371 out1:
2372 	if (fromnd.ni_startdir)
2373 		vrele(fromnd.ni_startdir);
2374 	if (error == -1)
2375 		return (0);
2376 	return (error);
2377 }
2378 
2379 /*
2380  * mkdir_args(char *path, int mode)
2381  *
2382  * Make a directory file.
2383  */
2384 /* ARGSUSED */
2385 int
2386 mkdir(struct mkdir_args *uap)
2387 {
2388 	struct thread *td = curthread;
2389 	struct proc *p = td->td_proc;
2390 	struct vnode *vp;
2391 	struct vattr vattr;
2392 	int error;
2393 	struct nameidata nd;
2394 
2395 	bwillwrite();
2396 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
2397 	nd.ni_cnd.cn_flags |= WILLBEDIR;
2398 	if ((error = namei(&nd)) != 0)
2399 		return (error);
2400 	vp = nd.ni_vp;
2401 	if (vp != NULL) {
2402 		NDFREE(&nd, NDF_ONLY_PNBUF);
2403 		if (nd.ni_dvp == vp)
2404 			vrele(nd.ni_dvp);
2405 		else
2406 			vput(nd.ni_dvp);
2407 		vrele(vp);
2408 		return (EEXIST);
2409 	}
2410 	VATTR_NULL(&vattr);
2411 	vattr.va_type = VDIR;
2412 	vattr.va_mode = (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2413 	VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
2414 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2415 	NDFREE(&nd, NDF_ONLY_PNBUF);
2416 	vput(nd.ni_dvp);
2417 	if (!error)
2418 		vput(nd.ni_vp);
2419 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mkdir");
2420 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "mkdir");
2421 	return (error);
2422 }
2423 
2424 /*
2425  * rmdir_args(char *path)
2426  *
2427  * Remove a directory file.
2428  */
2429 /* ARGSUSED */
2430 int
2431 rmdir(struct rmdir_args *uap)
2432 {
2433 	struct thread *td = curthread;
2434 	struct proc *p = td->td_proc;
2435 	struct vnode *vp;
2436 	int error;
2437 	struct nameidata nd;
2438 
2439 	bwillwrite();
2440 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2441 	    SCARG(uap, path), td);
2442 	if ((error = namei(&nd)) != 0)
2443 		return (error);
2444 	vp = nd.ni_vp;
2445 	if (vp->v_type != VDIR) {
2446 		error = ENOTDIR;
2447 		goto out;
2448 	}
2449 	/*
2450 	 * No rmdir "." please.
2451 	 */
2452 	if (nd.ni_dvp == vp) {
2453 		error = EINVAL;
2454 		goto out;
2455 	}
2456 	/*
2457 	 * The root of a mounted filesystem cannot be deleted.
2458 	 */
2459 	if (vp->v_flag & VROOT)
2460 		error = EBUSY;
2461 	else {
2462 		VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE);
2463 		VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2464 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2465 	}
2466 out:
2467 	NDFREE(&nd, NDF_ONLY_PNBUF);
2468 	if (nd.ni_dvp == vp)
2469 		vrele(nd.ni_dvp);
2470 	else
2471 		vput(nd.ni_dvp);
2472 	if (vp != NULLVP)
2473 		vput(vp);
2474 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "rmdir");
2475 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "rmdir");
2476 	return (error);
2477 }
2478 
2479 #ifdef COMPAT_43
2480 /*
2481  * ogetdirentries_args(int fd, char *buf, u_int count, long *basep)
2482  *
2483  * Read a block of directory entries in a file system independent format.
2484  */
2485 int
2486 ogetdirentries(struct ogetdirentries_args *uap)
2487 {
2488 	struct thread *td = curthread;
2489 	struct proc *p = td->td_proc;
2490 	struct vnode *vp;
2491 	struct file *fp;
2492 	struct uio auio, kuio;
2493 	struct iovec aiov, kiov;
2494 	struct dirent *dp, *edp;
2495 	caddr_t dirbuf;
2496 	int error, eofflag, readcnt;
2497 	long loff;
2498 
2499 	/* XXX arbitrary sanity limit on `count'. */
2500 	if (SCARG(uap, count) > 64 * 1024)
2501 		return (EINVAL);
2502 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2503 		return (error);
2504 	if ((fp->f_flag & FREAD) == 0)
2505 		return (EBADF);
2506 	vp = (struct vnode *)fp->f_data;
2507 unionread:
2508 	if (vp->v_type != VDIR)
2509 		return (EINVAL);
2510 	aiov.iov_base = SCARG(uap, buf);
2511 	aiov.iov_len = SCARG(uap, count);
2512 	auio.uio_iov = &aiov;
2513 	auio.uio_iovcnt = 1;
2514 	auio.uio_rw = UIO_READ;
2515 	auio.uio_segflg = UIO_USERSPACE;
2516 	auio.uio_td = td;
2517 	auio.uio_resid = SCARG(uap, count);
2518 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2519 	loff = auio.uio_offset = fp->f_offset;
2520 #	if (BYTE_ORDER != LITTLE_ENDIAN)
2521 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
2522 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
2523 			    NULL, NULL);
2524 			fp->f_offset = auio.uio_offset;
2525 		} else
2526 #	endif
2527 	{
2528 		kuio = auio;
2529 		kuio.uio_iov = &kiov;
2530 		kuio.uio_segflg = UIO_SYSSPACE;
2531 		kiov.iov_len = SCARG(uap, count);
2532 		MALLOC(dirbuf, caddr_t, SCARG(uap, count), M_TEMP, M_WAITOK);
2533 		kiov.iov_base = dirbuf;
2534 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
2535 			    NULL, NULL);
2536 		fp->f_offset = kuio.uio_offset;
2537 		if (error == 0) {
2538 			readcnt = SCARG(uap, count) - kuio.uio_resid;
2539 			edp = (struct dirent *)&dirbuf[readcnt];
2540 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
2541 #				if (BYTE_ORDER == LITTLE_ENDIAN)
2542 					/*
2543 					 * The expected low byte of
2544 					 * dp->d_namlen is our dp->d_type.
2545 					 * The high MBZ byte of dp->d_namlen
2546 					 * is our dp->d_namlen.
2547 					 */
2548 					dp->d_type = dp->d_namlen;
2549 					dp->d_namlen = 0;
2550 #				else
2551 					/*
2552 					 * The dp->d_type is the high byte
2553 					 * of the expected dp->d_namlen,
2554 					 * so must be zero'ed.
2555 					 */
2556 					dp->d_type = 0;
2557 #				endif
2558 				if (dp->d_reclen > 0) {
2559 					dp = (struct dirent *)
2560 					    ((char *)dp + dp->d_reclen);
2561 				} else {
2562 					error = EIO;
2563 					break;
2564 				}
2565 			}
2566 			if (dp >= edp)
2567 				error = uiomove(dirbuf, readcnt, &auio);
2568 		}
2569 		FREE(dirbuf, M_TEMP);
2570 	}
2571 	VOP_UNLOCK(vp, 0, td);
2572 	if (error)
2573 		return (error);
2574 	if (SCARG(uap, count) == auio.uio_resid) {
2575 		if (union_dircheckp) {
2576 			error = union_dircheckp(td, &vp, fp);
2577 			if (error == -1)
2578 				goto unionread;
2579 			if (error)
2580 				return (error);
2581 		}
2582 		if ((vp->v_flag & VROOT) &&
2583 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
2584 			struct vnode *tvp = vp;
2585 			vp = vp->v_mount->mnt_vnodecovered;
2586 			VREF(vp);
2587 			fp->f_data = (caddr_t) vp;
2588 			fp->f_offset = 0;
2589 			vrele(tvp);
2590 			goto unionread;
2591 		}
2592 	}
2593 	error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
2594 	    sizeof(long));
2595 	uap->sysmsg_result = SCARG(uap, count) - auio.uio_resid;
2596 	return (error);
2597 }
2598 #endif /* COMPAT_43 */
2599 
2600 /*
2601  * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2602  *
2603  * Read a block of directory entries in a file system independent format.
2604  */
2605 int
2606 getdirentries(struct getdirentries_args *uap)
2607 {
2608 	struct thread *td = curthread;
2609 	struct proc *p = td->td_proc;
2610 	struct vnode *vp;
2611 	struct file *fp;
2612 	struct uio auio;
2613 	struct iovec aiov;
2614 	long loff;
2615 	int error, eofflag;
2616 
2617 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2618 		return (error);
2619 	if ((fp->f_flag & FREAD) == 0)
2620 		return (EBADF);
2621 	vp = (struct vnode *)fp->f_data;
2622 unionread:
2623 	if (vp->v_type != VDIR)
2624 		return (EINVAL);
2625 	aiov.iov_base = SCARG(uap, buf);
2626 	aiov.iov_len = SCARG(uap, count);
2627 	auio.uio_iov = &aiov;
2628 	auio.uio_iovcnt = 1;
2629 	auio.uio_rw = UIO_READ;
2630 	auio.uio_segflg = UIO_USERSPACE;
2631 	auio.uio_td = td;
2632 	auio.uio_resid = SCARG(uap, count);
2633 	/* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
2634 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2635 	loff = auio.uio_offset = fp->f_offset;
2636 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2637 	fp->f_offset = auio.uio_offset;
2638 	VOP_UNLOCK(vp, 0, td);
2639 	if (error)
2640 		return (error);
2641 	if (SCARG(uap, count) == auio.uio_resid) {
2642 		if (union_dircheckp) {
2643 			error = union_dircheckp(td, &vp, fp);
2644 			if (error == -1)
2645 				goto unionread;
2646 			if (error)
2647 				return (error);
2648 		}
2649 		if ((vp->v_flag & VROOT) &&
2650 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
2651 			struct vnode *tvp = vp;
2652 			vp = vp->v_mount->mnt_vnodecovered;
2653 			VREF(vp);
2654 			fp->f_data = (caddr_t) vp;
2655 			fp->f_offset = 0;
2656 			vrele(tvp);
2657 			goto unionread;
2658 		}
2659 	}
2660 	if (SCARG(uap, basep) != NULL) {
2661 		error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
2662 		    sizeof(long));
2663 	}
2664 	uap->sysmsg_result = SCARG(uap, count) - auio.uio_resid;
2665 	return (error);
2666 }
2667 
2668 /*
2669  * getdents_args(int fd, char *buf, size_t count)
2670  */
2671 int
2672 getdents(struct getdents_args *uap)
2673 {
2674 	struct getdirentries_args ap;
2675 
2676 	ap.fd = uap->fd;
2677 	ap.buf = uap->buf;
2678 	ap.count = uap->count;
2679 	ap.basep = NULL;
2680 	return getdirentries(&ap);
2681 }
2682 
2683 /*
2684  * umask(int newmask)
2685  *
2686  * Set the mode mask for creation of filesystem nodes.
2687  *
2688  * MP SAFE
2689  */
2690 int
2691 umask(struct umask_args *uap)
2692 {
2693 	struct thread *td = curthread;
2694 	struct proc *p = td->td_proc;
2695 	struct filedesc *fdp;
2696 
2697 	fdp = p->p_fd;
2698 	uap->sysmsg_result = fdp->fd_cmask;
2699 	fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
2700 	return (0);
2701 }
2702 
2703 /*
2704  * revoke(char *path)
2705  *
2706  * Void all references to file by ripping underlying filesystem
2707  * away from vnode.
2708  */
2709 /* ARGSUSED */
2710 int
2711 revoke(struct revoke_args *uap)
2712 {
2713 	struct thread *td = curthread;
2714 	struct proc *p = td->td_proc;
2715 	struct vnode *vp;
2716 	struct vattr vattr;
2717 	int error;
2718 	struct nameidata nd;
2719 
2720 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2721 	if ((error = namei(&nd)) != 0)
2722 		return (error);
2723 	vp = nd.ni_vp;
2724 	NDFREE(&nd, NDF_ONLY_PNBUF);
2725 	if (vp->v_type != VCHR && vp->v_type != VBLK) {
2726 		error = EINVAL;
2727 		goto out;
2728 	}
2729 	if ((error = VOP_GETATTR(vp, &vattr, td)) != 0)
2730 		goto out;
2731 	if (p->p_ucred->cr_uid != vattr.va_uid &&
2732 	    (error = suser_cred(p->p_ucred, PRISON_ROOT)))
2733 		goto out;
2734 	if (vcount(vp) > 1)
2735 		VOP_REVOKE(vp, REVOKEALL);
2736 out:
2737 	vrele(vp);
2738 	return (error);
2739 }
2740 
2741 /*
2742  * Convert a user file descriptor to a kernel file entry.
2743  */
2744 int
2745 getvnode(struct filedesc *fdp, int fd, struct file **fpp)
2746 {
2747 	struct file *fp;
2748 
2749 	if ((u_int)fd >= fdp->fd_nfiles ||
2750 	    (fp = fdp->fd_ofiles[fd]) == NULL)
2751 		return (EBADF);
2752 	if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO)
2753 		return (EINVAL);
2754 	*fpp = fp;
2755 	return (0);
2756 }
2757 /*
2758  * getfh_args(char *fname, fhandle_t *fhp)
2759  *
2760  * Get (NFS) file handle
2761  */
2762 int
2763 getfh(struct getfh_args *uap)
2764 {
2765 	struct thread *td = curthread;
2766 	struct nameidata nd;
2767 	fhandle_t fh;
2768 	struct vnode *vp;
2769 	int error;
2770 
2771 	/*
2772 	 * Must be super user
2773 	 */
2774 	error = suser(td);
2775 	if (error)
2776 		return (error);
2777 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, td);
2778 	error = namei(&nd);
2779 	if (error)
2780 		return (error);
2781 	NDFREE(&nd, NDF_ONLY_PNBUF);
2782 	vp = nd.ni_vp;
2783 	bzero(&fh, sizeof(fh));
2784 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2785 	error = VFS_VPTOFH(vp, &fh.fh_fid);
2786 	vput(vp);
2787 	if (error)
2788 		return (error);
2789 	error = copyout(&fh, uap->fhp, sizeof (fh));
2790 	return (error);
2791 }
2792 
2793 /*
2794  * fhopen_args(const struct fhandle *u_fhp, int flags)
2795  *
2796  * syscall for the rpc.lockd to use to translate a NFS file handle into
2797  * an open descriptor.
2798  *
2799  * warning: do not remove the suser() call or this becomes one giant
2800  * security hole.
2801  */
2802 int
2803 fhopen(struct fhopen_args *uap)
2804 {
2805 	struct thread *td = curthread;
2806 	struct proc *p = td->td_proc;
2807 	struct mount *mp;
2808 	struct vnode *vp;
2809 	struct fhandle fhp;
2810 	struct vattr vat;
2811 	struct vattr *vap = &vat;
2812 	struct flock lf;
2813 	struct file *fp;
2814 	struct filedesc *fdp = p->p_fd;
2815 	int fmode, mode, error, type;
2816 	struct file *nfp;
2817 	int indx;
2818 
2819 	/*
2820 	 * Must be super user
2821 	 */
2822 	error = suser(td);
2823 	if (error)
2824 		return (error);
2825 
2826 	fmode = FFLAGS(SCARG(uap, flags));
2827 	/* why not allow a non-read/write open for our lockd? */
2828 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
2829 		return (EINVAL);
2830 	error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp));
2831 	if (error)
2832 		return(error);
2833 	/* find the mount point */
2834 	mp = vfs_getvfs(&fhp.fh_fsid);
2835 	if (mp == NULL)
2836 		return (ESTALE);
2837 	/* now give me my vnode, it gets returned to me locked */
2838 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
2839 	if (error)
2840 		return (error);
2841  	/*
2842 	 * from now on we have to make sure not
2843 	 * to forget about the vnode
2844 	 * any error that causes an abort must vput(vp)
2845 	 * just set error = err and 'goto bad;'.
2846 	 */
2847 
2848 	/*
2849 	 * from vn_open
2850 	 */
2851 	if (vp->v_type == VLNK) {
2852 		error = EMLINK;
2853 		goto bad;
2854 	}
2855 	if (vp->v_type == VSOCK) {
2856 		error = EOPNOTSUPP;
2857 		goto bad;
2858 	}
2859 	mode = 0;
2860 	if (fmode & (FWRITE | O_TRUNC)) {
2861 		if (vp->v_type == VDIR) {
2862 			error = EISDIR;
2863 			goto bad;
2864 		}
2865 		error = vn_writechk(vp);
2866 		if (error)
2867 			goto bad;
2868 		mode |= VWRITE;
2869 	}
2870 	if (fmode & FREAD)
2871 		mode |= VREAD;
2872 	if (mode) {
2873 		error = VOP_ACCESS(vp, mode, p->p_ucred, td);
2874 		if (error)
2875 			goto bad;
2876 	}
2877 	if (fmode & O_TRUNC) {
2878 		VOP_UNLOCK(vp, 0, td);				/* XXX */
2879 		VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2880 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);	/* XXX */
2881 		VATTR_NULL(vap);
2882 		vap->va_size = 0;
2883 		error = VOP_SETATTR(vp, vap, p->p_ucred, td);
2884 		if (error)
2885 			goto bad;
2886 	}
2887 	error = VOP_OPEN(vp, fmode, p->p_ucred, td);
2888 	if (error)
2889 		goto bad;
2890 	/*
2891 	 * Make sure that a VM object is created for VMIO support.
2892 	 */
2893 	if (vn_canvmio(vp) == TRUE) {
2894 		if ((error = vfs_object_create(vp, td)) != 0)
2895 			goto bad;
2896 	}
2897 	if (fmode & FWRITE)
2898 		vp->v_writecount++;
2899 
2900 	/*
2901 	 * end of vn_open code
2902 	 */
2903 
2904 	if ((error = falloc(p, &nfp, &indx)) != 0) {
2905 		if (fmode & FWRITE)
2906 			vp->v_writecount--;
2907 		goto bad;
2908 	}
2909 	fp = nfp;
2910 
2911 	/*
2912 	 * hold an extra reference to avoid having fp ripped out
2913 	 * from under us while we block in the lock op.
2914 	 */
2915 	fhold(fp);
2916 	nfp->f_data = (caddr_t)vp;
2917 	nfp->f_flag = fmode & FMASK;
2918 	nfp->f_ops = &vnops;
2919 	nfp->f_type = DTYPE_VNODE;
2920 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
2921 		lf.l_whence = SEEK_SET;
2922 		lf.l_start = 0;
2923 		lf.l_len = 0;
2924 		if (fmode & O_EXLOCK)
2925 			lf.l_type = F_WRLCK;
2926 		else
2927 			lf.l_type = F_RDLCK;
2928 		type = F_FLOCK;
2929 		if ((fmode & FNONBLOCK) == 0)
2930 			type |= F_WAIT;
2931 		VOP_UNLOCK(vp, 0, td);
2932 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
2933 			/*
2934 			 * lock request failed.  Normally close the descriptor
2935 			 * but handle the case where someone might have dup()d
2936 			 * or close()d it when we weren't looking.
2937 			 */
2938 			if (fdp->fd_ofiles[indx] == fp) {
2939 				fdp->fd_ofiles[indx] = NULL;
2940 				fdrop(fp, td);
2941 			}
2942 
2943 			/*
2944 			 * release our private reference.
2945 			 */
2946 			fdrop(fp, td);
2947 			return (error);
2948 		}
2949 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2950 		fp->f_flag |= FHASLOCK;
2951 	}
2952 	if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
2953 		vfs_object_create(vp, td);
2954 
2955 	VOP_UNLOCK(vp, 0, td);
2956 	fdrop(fp, td);
2957 	uap->sysmsg_result = indx;
2958 	return (0);
2959 
2960 bad:
2961 	vput(vp);
2962 	return (error);
2963 }
2964 
2965 /*
2966  * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
2967  */
2968 int
2969 fhstat(struct fhstat_args *uap)
2970 {
2971 	struct thread *td = curthread;
2972 	struct stat sb;
2973 	fhandle_t fh;
2974 	struct mount *mp;
2975 	struct vnode *vp;
2976 	int error;
2977 
2978 	/*
2979 	 * Must be super user
2980 	 */
2981 	error = suser(td);
2982 	if (error)
2983 		return (error);
2984 
2985 	error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t));
2986 	if (error)
2987 		return (error);
2988 
2989 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
2990 		return (ESTALE);
2991 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
2992 		return (error);
2993 	error = vn_stat(vp, &sb, td);
2994 	vput(vp);
2995 	if (error)
2996 		return (error);
2997 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
2998 	return (error);
2999 }
3000 
3001 /*
3002  * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3003  */
3004 int
3005 fhstatfs(struct fhstatfs_args *uap)
3006 {
3007 	struct thread *td = curthread;
3008 	struct statfs *sp;
3009 	struct mount *mp;
3010 	struct vnode *vp;
3011 	struct statfs sb;
3012 	fhandle_t fh;
3013 	int error;
3014 
3015 	/*
3016 	 * Must be super user
3017 	 */
3018 	if ((error = suser(td)))
3019 		return (error);
3020 
3021 	if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0)
3022 		return (error);
3023 
3024 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3025 		return (ESTALE);
3026 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3027 		return (error);
3028 	mp = vp->v_mount;
3029 	sp = &mp->mnt_stat;
3030 	vput(vp);
3031 	if ((error = VFS_STATFS(mp, sp, td)) != 0)
3032 		return (error);
3033 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3034 	if (suser(td)) {
3035 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
3036 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3037 		sp = &sb;
3038 	}
3039 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
3040 }
3041 
3042 /*
3043  * Syscall to push extended attribute configuration information into the
3044  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
3045  * a command (int cmd), and attribute name and misc data.  For now, the
3046  * attribute name is left in userspace for consumption by the VFS_op.
3047  * It will probably be changed to be copied into sysspace by the
3048  * syscall in the future, once issues with various consumers of the
3049  * attribute code have raised their hands.
3050  *
3051  * Currently this is used only by UFS Extended Attributes.
3052  */
3053 int
3054 extattrctl(struct extattrctl_args *uap)
3055 {
3056 	struct thread *td = curthread;
3057 	struct nameidata nd;
3058 	struct mount *mp;
3059 	int error;
3060 
3061 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
3062 	if ((error = namei(&nd)) != 0)
3063 		return (error);
3064 	mp = nd.ni_vp->v_mount;
3065 	NDFREE(&nd, 0);
3066 	return (VFS_EXTATTRCTL(mp, SCARG(uap, cmd), SCARG(uap, attrname),
3067 	    SCARG(uap, arg), td));
3068 }
3069 
3070 /*
3071  * Syscall to set a named extended attribute on a file or directory.
3072  * Accepts attribute name, and a uio structure pointing to the data to set.
3073  * The uio is consumed in the style of writev().  The real work happens
3074  * in VOP_SETEXTATTR().
3075  */
3076 int
3077 extattr_set_file(struct extattr_set_file_args *uap)
3078 {
3079 	struct thread *td = curthread;
3080 	struct proc *p = td->td_proc;
3081 	struct nameidata nd;
3082 	struct uio auio;
3083 	struct iovec *iov, *needfree = NULL, aiov[UIO_SMALLIOV];
3084 	char attrname[EXTATTR_MAXNAMELEN];
3085 	u_int iovlen, cnt;
3086 	int error, i;
3087 
3088 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3089 	if (error)
3090 		return (error);
3091 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3092 	    SCARG(uap, path), td);
3093 	if ((error = namei(&nd)) != 0)
3094 		return(error);
3095 	iovlen = uap->iovcnt * sizeof(struct iovec);
3096 	if (uap->iovcnt > UIO_SMALLIOV) {
3097 		if (uap->iovcnt > UIO_MAXIOV) {
3098 			error = EINVAL;
3099 			goto done;
3100 		}
3101 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3102 		needfree = iov;
3103 	} else
3104 		iov = aiov;
3105 	auio.uio_iov = iov;
3106 	auio.uio_iovcnt = uap->iovcnt;
3107 	auio.uio_rw = UIO_WRITE;
3108 	auio.uio_segflg = UIO_USERSPACE;
3109 	auio.uio_td = td;
3110 	auio.uio_offset = 0;
3111 	if ((error = copyin((caddr_t)uap->iovp, (caddr_t)iov, iovlen)))
3112 		goto done;
3113 	auio.uio_resid = 0;
3114 	for (i = 0; i < uap->iovcnt; i++) {
3115 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
3116 			error = EINVAL;
3117 			goto done;
3118 		}
3119 		auio.uio_resid += iov->iov_len;
3120 		iov++;
3121 	}
3122 	cnt = auio.uio_resid;
3123 	error = VOP_SETEXTATTR(nd.ni_vp, attrname, &auio, p->p_ucred, td);
3124 	cnt -= auio.uio_resid;
3125 	uap->sysmsg_result = cnt;
3126 done:
3127 	if (needfree)
3128 		FREE(needfree, M_IOV);
3129 	NDFREE(&nd, 0);
3130 	return (error);
3131 }
3132 
3133 /*
3134  * Syscall to get a named extended attribute on a file or directory.
3135  * Accepts attribute name, and a uio structure pointing to a buffer for the
3136  * data.  The uio is consumed in the style of readv().  The real work
3137  * happens in VOP_GETEXTATTR();
3138  */
3139 int
3140 extattr_get_file(struct extattr_get_file_args *uap)
3141 {
3142 	struct thread *td = curthread;
3143 	struct proc *p = td->td_proc;
3144 	struct nameidata nd;
3145 	struct uio auio;
3146 	struct iovec *iov, *needfree, aiov[UIO_SMALLIOV];
3147 	char attrname[EXTATTR_MAXNAMELEN];
3148 	u_int iovlen, cnt;
3149 	int error, i;
3150 
3151 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3152 	if (error)
3153 		return (error);
3154 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3155 	    SCARG(uap, path), td);
3156 	if ((error = namei(&nd)) != 0)
3157 		return (error);
3158 	iovlen = uap->iovcnt * sizeof (struct iovec);
3159 	if (uap->iovcnt > UIO_SMALLIOV) {
3160 		if (uap->iovcnt > UIO_MAXIOV) {
3161 			NDFREE(&nd, 0);
3162 			return (EINVAL);
3163 		}
3164 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3165 		needfree = iov;
3166 	} else {
3167 		iov = aiov;
3168 		needfree = NULL;
3169 	}
3170 	auio.uio_iov = iov;
3171 	auio.uio_iovcnt = uap->iovcnt;
3172 	auio.uio_rw = UIO_READ;
3173 	auio.uio_segflg = UIO_USERSPACE;
3174 	auio.uio_td = td;
3175 	auio.uio_offset = 0;
3176 	if ((error = copyin((caddr_t)uap->iovp, (caddr_t)iov, iovlen)))
3177 		goto done;
3178 	auio.uio_resid = 0;
3179 	for (i = 0; i < uap->iovcnt; i++) {
3180 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
3181 			error = EINVAL;
3182 			goto done;
3183 		}
3184 		auio.uio_resid += iov->iov_len;
3185 		iov++;
3186 	}
3187 	cnt = auio.uio_resid;
3188 	error = VOP_GETEXTATTR(nd.ni_vp, attrname, &auio, p->p_ucred, td);
3189 	cnt -= auio.uio_resid;
3190 	uap->sysmsg_result = cnt;
3191 done:
3192 	if (needfree)
3193 		FREE(needfree, M_IOV);
3194 	NDFREE(&nd, 0);
3195 	return(error);
3196 }
3197 
3198 /*
3199  * Syscall to delete a named extended attribute from a file or directory.
3200  * Accepts attribute name.  The real work happens in VOP_SETEXTATTR().
3201  */
3202 int
3203 extattr_delete_file(struct extattr_delete_file_args *uap)
3204 {
3205 	struct thread *td = curthread;
3206 	struct proc *p = td->td_proc;
3207 	struct nameidata nd;
3208 	char attrname[EXTATTR_MAXNAMELEN];
3209 	int	error;
3210 
3211 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3212 	if (error)
3213 		return(error);
3214 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3215 	    SCARG(uap, path), td);
3216 	if ((error = namei(&nd)) != 0)
3217 		return(error);
3218 	error = VOP_SETEXTATTR(nd.ni_vp, attrname, NULL, p->p_ucred, td);
3219 	NDFREE(&nd, 0);
3220 	return(error);
3221 }
3222