xref: /dflybsd-src/sys/kern/vfs_syscalls.c (revision 17ea22213f86a5c5966c1e6bf8e95f022ebb92b9)
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 1.55 2005/01/31 17:20:48 joerg Exp $
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/buf.h>
46 #include <sys/conf.h>
47 #include <sys/sysent.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/sysproto.h>
51 #include <sys/filedesc.h>
52 #include <sys/kernel.h>
53 #include <sys/fcntl.h>
54 #include <sys/file.h>
55 #include <sys/linker.h>
56 #include <sys/stat.h>
57 #include <sys/unistd.h>
58 #include <sys/vnode.h>
59 #include <sys/proc.h>
60 #include <sys/namei.h>
61 #include <sys/nlookup.h>
62 #include <sys/dirent.h>
63 #include <sys/extattr.h>
64 #include <sys/kern_syscall.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 checkvp_chdir (struct vnode *vn, struct thread *td);
77 static void checkdirs (struct vnode *olddp, struct namecache *ncp);
78 static int chroot_refuse_vdir_fds (struct filedesc *fdp);
79 static int getutimes (const struct timeval *, struct timespec *);
80 static int setfown (struct vnode *, uid_t, gid_t);
81 static int setfmode (struct vnode *, int);
82 static int setfflags (struct vnode *, int);
83 static int setutimes (struct vnode *, const struct timespec *, int);
84 static int	usermount = 0;	/* if 1, non-root can mount fs. */
85 
86 int (*union_dircheckp) (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 namecache *ncp;
108 	struct mount *mp;
109 	struct vfsconf *vfsp;
110 	int error, flag = 0, flag2 = 0;
111 	struct vattr va;
112 	struct nlookupdata nd;
113 	char fstypename[MFSNAMELEN];
114 	lwkt_tokref ilock;
115 	struct nlcomponent nlc;
116 
117 	KKASSERT(p);
118 	if (p->p_ucred->cr_prison != NULL)
119 		return (EPERM);
120 	if (usermount == 0 && (error = suser(td)))
121 		return (error);
122 	/*
123 	 * Do not allow NFS export by non-root users.
124 	 */
125 	if (SCARG(uap, flags) & MNT_EXPORTED) {
126 		error = suser(td);
127 		if (error)
128 			return (error);
129 	}
130 	/*
131 	 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
132 	 */
133 	if (suser(td))
134 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
135 
136 	/*
137 	 * Lookup the requested path and extract the ncp and vnode.
138 	 */
139 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
140 	if (error == 0) {
141 		if ((error = nlookup(&nd)) == 0) {
142 			if (nd.nl_ncp->nc_vp == NULL)
143 				error = ENOENT;
144 		}
145 	}
146 	if (error) {
147 		nlookup_done(&nd);
148 		return (error);
149 	}
150 
151 	/*
152 	 * Extract the locked+refd ncp and cleanup the nd structure
153 	 */
154 	ncp = nd.nl_ncp;
155 	nd.nl_ncp = NULL;
156 	nlookup_done(&nd);
157 
158 	/*
159 	 * now we have the locked ref'd ncp and unreferenced vnode.
160 	 */
161 	vp = ncp->nc_vp;
162 	if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) {
163 		cache_put(ncp);
164 		return (error);
165 	}
166 	cache_unlock(ncp);
167 
168 	/*
169 	 * Now we have an unlocked ref'd ncp and a locked ref'd vp
170 	 */
171 	if (SCARG(uap, flags) & MNT_UPDATE) {
172 		if ((vp->v_flag & VROOT) == 0) {
173 			cache_drop(ncp);
174 			vput(vp);
175 			return (EINVAL);
176 		}
177 		mp = vp->v_mount;
178 		flag = mp->mnt_flag;
179 		flag2 = mp->mnt_kern_flag;
180 		/*
181 		 * We only allow the filesystem to be reloaded if it
182 		 * is currently mounted read-only.
183 		 */
184 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
185 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
186 			cache_drop(ncp);
187 			vput(vp);
188 			return (EOPNOTSUPP);	/* Needs translation */
189 		}
190 		/*
191 		 * Only root, or the user that did the original mount is
192 		 * permitted to update it.
193 		 */
194 		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
195 		    (error = suser(td))) {
196 			cache_drop(ncp);
197 			vput(vp);
198 			return (error);
199 		}
200 		if (vfs_busy(mp, LK_NOWAIT, NULL, td)) {
201 			cache_drop(ncp);
202 			vput(vp);
203 			return (EBUSY);
204 		}
205 		if ((vp->v_flag & VMOUNT) != 0 ||
206 		    vp->v_mountedhere != NULL) {
207 			cache_drop(ncp);
208 			vfs_unbusy(mp, td);
209 			vput(vp);
210 			return (EBUSY);
211 		}
212 		vp->v_flag |= VMOUNT;
213 		mp->mnt_flag |=
214 		    SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
215 		VOP_UNLOCK(vp, 0, td);
216 		goto update;
217 	}
218 	/*
219 	 * If the user is not root, ensure that they own the directory
220 	 * onto which we are attempting to mount.
221 	 */
222 	if ((error = VOP_GETATTR(vp, &va, td)) ||
223 	    (va.va_uid != p->p_ucred->cr_uid &&
224 	     (error = suser(td)))) {
225 		cache_drop(ncp);
226 		vput(vp);
227 		return (error);
228 	}
229 	if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) {
230 		cache_drop(ncp);
231 		vput(vp);
232 		return (error);
233 	}
234 	if (vp->v_type != VDIR) {
235 		cache_drop(ncp);
236 		vput(vp);
237 		return (ENOTDIR);
238 	}
239 	if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) {
240 		cache_drop(ncp);
241 		vput(vp);
242 		return (error);
243 	}
244 	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
245 		if (!strcmp(vfsp->vfc_name, fstypename))
246 			break;
247 	}
248 	if (vfsp == NULL) {
249 		linker_file_t lf;
250 
251 		/* Only load modules for root (very important!) */
252 		if ((error = suser(td)) != 0) {
253 			cache_drop(ncp);
254 			vput(vp);
255 			return error;
256 		}
257 		error = linker_load_file(fstypename, &lf);
258 		if (error || lf == NULL) {
259 			cache_drop(ncp);
260 			vput(vp);
261 			if (lf == NULL)
262 				error = ENODEV;
263 			return error;
264 		}
265 		lf->userrefs++;
266 		/* lookup again, see if the VFS was loaded */
267 		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
268 			if (!strcmp(vfsp->vfc_name, fstypename))
269 				break;
270 		}
271 		if (vfsp == NULL) {
272 			lf->userrefs--;
273 			linker_file_unload(lf);
274 			cache_drop(ncp);
275 			vput(vp);
276 			return (ENODEV);
277 		}
278 	}
279 	if ((vp->v_flag & VMOUNT) != 0 ||
280 	    vp->v_mountedhere != NULL) {
281 		cache_drop(ncp);
282 		vput(vp);
283 		return (EBUSY);
284 	}
285 	vp->v_flag |= VMOUNT;
286 
287 	/*
288 	 * Allocate and initialize the filesystem.
289 	 */
290 	mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
291 	TAILQ_INIT(&mp->mnt_nvnodelist);
292 	TAILQ_INIT(&mp->mnt_reservedvnlist);
293 	TAILQ_INIT(&mp->mnt_jlist);
294 	mp->mnt_nvnodelistsize = 0;
295 	lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE);
296 	vfs_busy(mp, LK_NOWAIT, NULL, td);
297 	mp->mnt_op = vfsp->vfc_vfsops;
298 	mp->mnt_vfc = vfsp;
299 	vfsp->vfc_refcount++;
300 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
301 	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
302 	strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
303 	mp->mnt_vnodecovered = vp;
304 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
305 	mp->mnt_iosize_max = DFLTPHYS;
306 	VOP_UNLOCK(vp, 0, td);
307 update:
308 	/*
309 	 * Set the mount level flags.
310 	 */
311 	if (SCARG(uap, flags) & MNT_RDONLY)
312 		mp->mnt_flag |= MNT_RDONLY;
313 	else if (mp->mnt_flag & MNT_RDONLY)
314 		mp->mnt_kern_flag |= MNTK_WANTRDWR;
315 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
316 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
317 	    MNT_NOSYMFOLLOW | MNT_IGNORE |
318 	    MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
319 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
320 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
321 	    MNT_NOSYMFOLLOW | MNT_IGNORE |
322 	    MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
323 	/*
324 	 * Mount the filesystem.
325 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
326 	 * get.
327 	 */
328 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), td);
329 	if (mp->mnt_flag & MNT_UPDATE) {
330 		if (mp->mnt_kern_flag & MNTK_WANTRDWR)
331 			mp->mnt_flag &= ~MNT_RDONLY;
332 		mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
333 		mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
334 		if (error) {
335 			mp->mnt_flag = flag;
336 			mp->mnt_kern_flag = flag2;
337 		}
338 		vfs_unbusy(mp, td);
339 		vp->v_flag &= ~VMOUNT;
340 		vrele(vp);
341 		cache_drop(ncp);
342 		return (error);
343 	}
344 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
345 	/*
346 	 * Put the new filesystem on the mount list after root.  The mount
347 	 * point gets its own mnt_ncp which is a special ncp linking the
348 	 * vnode-under to the root of the new mount.  The lookup code
349 	 * detects the mount point going forward and detects the special
350 	 * mnt_ncp via NCP_MOUNTPT going backwards.
351 	 *
352 	 * It is not necessary to invalidate or purge the vnode underneath
353 	 * because elements under the mount will be given their own glue
354 	 * namecache record.
355 	 */
356 	if (!error) {
357 		nlc.nlc_nameptr = "";
358 		nlc.nlc_namelen = 0;
359 		mp->mnt_ncp = cache_nlookup(ncp, &nlc);
360 		cache_setunresolved(mp->mnt_ncp);
361 		mp->mnt_ncp->nc_flag |= NCF_MOUNTPT;
362 		mp->mnt_ncp->nc_mount = mp;
363 		cache_drop(ncp);
364 		/* XXX get the root of the fs and cache_setvp(mnt_ncp...) */
365 		vp->v_flag &= ~VMOUNT;
366 		vp->v_mountedhere = mp;
367 		lwkt_gettoken(&ilock, &mountlist_token);
368 		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
369 		lwkt_reltoken(&ilock);
370 		checkdirs(vp, mp->mnt_ncp);
371 		cache_unlock(mp->mnt_ncp);	/* leave ref intact */
372 		VOP_UNLOCK(vp, 0, td);
373 		error = vfs_allocate_syncvnode(mp);
374 		vfs_unbusy(mp, td);
375 		if ((error = VFS_START(mp, 0, td)) != 0)
376 			vrele(vp);
377 	} else {
378 		vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
379 		vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
380 		vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
381 		vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
382 		vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
383 		vp->v_flag &= ~VMOUNT;
384 		mp->mnt_vfc->vfc_refcount--;
385 		vfs_unbusy(mp, td);
386 		free(mp, M_MOUNT);
387 		cache_drop(ncp);
388 		vput(vp);
389 	}
390 	return (error);
391 }
392 
393 /*
394  * Scan all active processes to see if any of them have a current
395  * or root directory onto which the new filesystem has just been
396  * mounted. If so, replace them with the new mount point.
397  *
398  * The passed ncp is ref'd and locked (from the mount code) and
399  * must be associated with the vnode representing the root of the
400  * mount point.
401  */
402 static void
403 checkdirs(struct vnode *olddp, struct namecache *ncp)
404 {
405 	struct filedesc *fdp;
406 	struct vnode *newdp;
407 	struct mount *mp;
408 	struct proc *p;
409 
410 	if (olddp->v_usecount == 1)
411 		return;
412 	mp = olddp->v_mountedhere;
413 	if (VFS_ROOT(mp, &newdp))
414 		panic("mount: lost mount");
415 	cache_setvp(ncp, newdp);
416 
417 	if (rootvnode == olddp) {
418 		vref(newdp);
419 		vfs_cache_setroot(newdp, cache_hold(ncp));
420 	}
421 
422 	FOREACH_PROC_IN_SYSTEM(p) {
423 		fdp = p->p_fd;
424 		if (fdp->fd_cdir == olddp) {
425 			vrele(fdp->fd_cdir);
426 			vref(newdp);
427 			fdp->fd_cdir = newdp;
428 			cache_drop(fdp->fd_ncdir);
429 			fdp->fd_ncdir = cache_hold(ncp);
430 		}
431 		if (fdp->fd_rdir == olddp) {
432 			vrele(fdp->fd_rdir);
433 			vref(newdp);
434 			fdp->fd_rdir = newdp;
435 			cache_drop(fdp->fd_nrdir);
436 			fdp->fd_nrdir = cache_hold(ncp);
437 		}
438 	}
439 	vput(newdp);
440 }
441 
442 /*
443  * Unmount a file system.
444  *
445  * Note: unmount takes a path to the vnode mounted on as argument,
446  * not special file (as before).
447  */
448 /*
449  * umount_args(char *path, int flags)
450  */
451 /* ARGSUSED */
452 int
453 unmount(struct unmount_args *uap)
454 {
455 	struct thread *td = curthread;
456 	struct proc *p = td->td_proc;
457 	struct vnode *vp;
458 	struct mount *mp;
459 	int error;
460 	struct nlookupdata nd;
461 
462 	KKASSERT(p);
463 	if (p->p_ucred->cr_prison != NULL)
464 		return (EPERM);
465 	if (usermount == 0 && (error = suser(td)))
466 		return (error);
467 
468 	vp = NULL;
469 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
470 	if (error == 0)
471 		error = nlookup(&nd);
472 	if (error == 0)
473 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
474 	nlookup_done(&nd);
475 	if (error)
476 		return (error);
477 
478 	mp = vp->v_mount;
479 
480 	/*
481 	 * Only root, or the user that did the original mount is
482 	 * permitted to unmount this filesystem.
483 	 */
484 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
485 	    (error = suser(td))) {
486 		vput(vp);
487 		return (error);
488 	}
489 
490 	/*
491 	 * Don't allow unmounting the root file system.
492 	 */
493 	if (mp->mnt_flag & MNT_ROOTFS) {
494 		vput(vp);
495 		return (EINVAL);
496 	}
497 
498 	/*
499 	 * Must be the root of the filesystem
500 	 */
501 	if ((vp->v_flag & VROOT) == 0) {
502 		vput(vp);
503 		return (EINVAL);
504 	}
505 	vput(vp);
506 	return (dounmount(mp, SCARG(uap, flags), td));
507 }
508 
509 /*
510  * Do the actual file system unmount.
511  */
512 int
513 dounmount(struct mount *mp, int flags, struct thread *td)
514 {
515 	struct vnode *coveredvp;
516 	int error;
517 	int async_flag;
518 	lwkt_tokref ilock;
519 
520 	lwkt_gettoken(&ilock, &mountlist_token);
521 	if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
522 		lwkt_reltoken(&ilock);
523 		return (EBUSY);
524 	}
525 	mp->mnt_kern_flag |= MNTK_UNMOUNT;
526 	/* Allow filesystems to detect that a forced unmount is in progress. */
527 	if (flags & MNT_FORCE)
528 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
529 	error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
530 	    ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &ilock, td);
531 	if (error) {
532 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
533 		if (mp->mnt_kern_flag & MNTK_MWAIT)
534 			wakeup(mp);
535 		return (error);
536 	}
537 
538 	if (mp->mnt_flag & MNT_EXPUBLIC)
539 		vfs_setpublicfs(NULL, NULL, NULL);
540 
541 	vfs_msync(mp, MNT_WAIT);
542 	async_flag = mp->mnt_flag & MNT_ASYNC;
543 	mp->mnt_flag &=~ MNT_ASYNC;
544 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
545 	if (mp->mnt_syncer != NULL)
546 		vrele(mp->mnt_syncer);
547 	if (((mp->mnt_flag & MNT_RDONLY) ||
548 	     (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
549 	    (flags & MNT_FORCE))
550 		error = VFS_UNMOUNT(mp, flags, td);
551 	lwkt_gettokref(&ilock);
552 	if (error) {
553 		if (mp->mnt_syncer == NULL)
554 			vfs_allocate_syncvnode(mp);
555 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
556 		mp->mnt_flag |= async_flag;
557 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
558 		    &ilock, td);
559 		if (mp->mnt_kern_flag & MNTK_MWAIT)
560 			wakeup(mp);
561 		return (error);
562 	}
563 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
564 
565 	/*
566 	 * Remove any installed vnode ops here so the individual VFSs don't
567 	 * have to.
568 	 */
569 	vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
570 	vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
571 	vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
572 	vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
573 	vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
574 
575 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
576 		coveredvp->v_mountedhere = NULL;
577 		vrele(coveredvp);
578 		cache_drop(mp->mnt_ncp);
579 		mp->mnt_ncp = NULL;
580 	}
581 	mp->mnt_vfc->vfc_refcount--;
582 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
583 		panic("unmount: dangling vnode");
584 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &ilock, td);
585 	if (mp->mnt_kern_flag & MNTK_MWAIT)
586 		wakeup(mp);
587 	free(mp, M_MOUNT);
588 	return (0);
589 }
590 
591 /*
592  * Sync each mounted filesystem.
593  */
594 
595 #ifdef DEBUG
596 static int syncprt = 0;
597 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
598 #endif /* DEBUG */
599 
600 /* ARGSUSED */
601 int
602 sync(struct sync_args *uap)
603 {
604 	struct thread *td = curthread;
605 	struct mount *mp, *nmp;
606 	lwkt_tokref ilock;
607 	int asyncflag;
608 
609 	lwkt_gettoken(&ilock, &mountlist_token);
610 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
611 		if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
612 			nmp = TAILQ_NEXT(mp, mnt_list);
613 			continue;
614 		}
615 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
616 			asyncflag = mp->mnt_flag & MNT_ASYNC;
617 			mp->mnt_flag &= ~MNT_ASYNC;
618 			vfs_msync(mp, MNT_NOWAIT);
619 			VFS_SYNC(mp, MNT_NOWAIT, td);
620 			mp->mnt_flag |= asyncflag;
621 		}
622 		lwkt_gettokref(&ilock);
623 		nmp = TAILQ_NEXT(mp, mnt_list);
624 		vfs_unbusy(mp, td);
625 	}
626 	lwkt_reltoken(&ilock);
627 /*
628  * print out buffer pool stat information on each sync() call.
629  */
630 #ifdef DEBUG
631 	if (syncprt)
632 		vfs_bufstats();
633 #endif /* DEBUG */
634 	return (0);
635 }
636 
637 /* XXX PRISON: could be per prison flag */
638 static int prison_quotas;
639 #if 0
640 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
641 #endif
642 
643 /*
644  *  quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
645  *
646  * Change filesystem quotas.
647  */
648 /* ARGSUSED */
649 int
650 quotactl(struct quotactl_args *uap)
651 {
652 	struct nlookupdata nd;
653 	struct thread *td;
654 	struct proc *p;
655 	struct mount *mp;
656 	int error;
657 
658 	td = curthread;
659 	p = td->td_proc;
660 	if (p->p_ucred->cr_prison && !prison_quotas)
661 		return (EPERM);
662 
663 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
664 	if (error == 0)
665 		error = nlookup(&nd);
666 	if (error == 0) {
667 		mp = nd.nl_ncp->nc_mount;
668 		error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
669 				    SCARG(uap, arg), nd.nl_td);
670 	}
671 	nlookup_done(&nd);
672 	return (error);
673 }
674 
675 /*
676  * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
677  *		void *buf, int buflen)
678  *
679  * This function operates on a mount point and executes the specified
680  * operation using the specified control data, and possibly returns data.
681  *
682  * The actual number of bytes stored in the result buffer is returned, 0
683  * if none, otherwise an error is returned.
684  */
685 /* ARGSUSED */
686 int
687 mountctl(struct mountctl_args *uap)
688 {
689 	struct thread *td = curthread;
690 	struct proc *p = td->td_proc;
691 	struct filedesc *fdp = p->p_fd;
692 	struct file *fp;
693 	void *ctl = NULL;
694 	void *buf = NULL;
695 	char *path = NULL;
696 	int error;
697 
698 	/*
699 	 * Sanity and permissions checks.  We must be root.
700 	 */
701 	KKASSERT(p);
702 	if (p->p_ucred->cr_prison != NULL)
703 		return (EPERM);
704 	if ((error = suser(td)) != 0)
705 		return (error);
706 
707 	/*
708 	 * Argument length checks
709 	 */
710 	if (uap->ctllen < 0 || uap->ctllen > 1024)
711 		return (EINVAL);
712 	if (uap->buflen < 0 || uap->buflen > 16 * 1024)
713 		return (EINVAL);
714 	if (uap->path == NULL)
715 		return (EINVAL);
716 
717 	/*
718 	 * Allocate the necessary buffers and copyin data
719 	 */
720 	path = zalloc(namei_zone);
721 	error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
722 	if (error)
723 		goto done;
724 
725 	if (uap->ctllen) {
726 		ctl = malloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
727 		error = copyin(uap->ctl, ctl, uap->ctllen);
728 		if (error)
729 			goto done;
730 	}
731 	if (uap->buflen)
732 		buf = malloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
733 
734 	/*
735 	 * Validate the descriptor
736 	 */
737 	if (uap->fd == -1) {
738 		fp = NULL;
739 	} else if ((u_int)uap->fd >= fdp->fd_nfiles ||
740 	    (fp = fdp->fd_ofiles[uap->fd]) == NULL) {
741 		error = EBADF;
742 		goto done;
743 	}
744 	if (fp)
745 		fhold(fp);
746 
747 	/*
748 	 * Execute the internal kernel function and clean up.
749 	 */
750 	error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
751 	if (fp)
752 		fdrop(fp, td);
753 	if (error == 0 && uap->sysmsg_result > 0)
754 		error = copyout(buf, uap->buf, uap->sysmsg_result);
755 done:
756 	if (path)
757 		zfree(namei_zone, path);
758 	if (ctl)
759 		free(ctl, M_TEMP);
760 	if (buf)
761 		free(buf, M_TEMP);
762 	return (error);
763 }
764 
765 /*
766  * Execute a mount control operation by resolving the path to a mount point
767  * and calling vop_mountctl().
768  */
769 int
770 kern_mountctl(const char *path, int op, struct file *fp,
771 		const void *ctl, int ctllen,
772 		void *buf, int buflen, int *res)
773 {
774 	struct vnode *vp;
775 	struct mount *mp;
776 	struct nlookupdata nd;
777 	int error;
778 
779 	*res = 0;
780 	vp = NULL;
781 	error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
782 	if (error == 0)
783 		error = nlookup(&nd);
784 	if (error == 0)
785 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
786 	nlookup_done(&nd);
787 	if (error)
788 		return (error);
789 
790 	mp = vp->v_mount;
791 
792 	/*
793 	 * Must be the root of the filesystem
794 	 */
795 	if ((vp->v_flag & VROOT) == 0) {
796 		vput(vp);
797 		return (EINVAL);
798 	}
799 	error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
800 				buf, buflen, res);
801 	vput(vp);
802 	return (error);
803 }
804 
805 int
806 kern_statfs(struct nlookupdata *nd, struct statfs *buf)
807 {
808 	struct thread *td = curthread;
809 	struct mount *mp;
810 	struct statfs *sp;
811 	int error;
812 
813 	if ((error = nlookup(nd)) != 0)
814 		return (error);
815 	mp = nd->nl_ncp->nc_mount;
816 	sp = &mp->mnt_stat;
817 	if ((error = VFS_STATFS(mp, sp, td)) != 0)
818 		return (error);
819 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
820 	bcopy(sp, buf, sizeof(*buf));
821 	/* Only root should have access to the fsid's. */
822 	if (suser(td))
823 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
824 	return (0);
825 }
826 
827 /*
828  * statfs_args(char *path, struct statfs *buf)
829  *
830  * Get filesystem statistics.
831  */
832 int
833 statfs(struct statfs_args *uap)
834 {
835 	struct nlookupdata nd;
836 	struct statfs buf;
837 	int error;
838 
839 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
840 	if (error == 0)
841 		error = kern_statfs(&nd, &buf);
842 	nlookup_done(&nd);
843 	if (error == 0)
844 		error = copyout(&buf, uap->buf, sizeof(*uap->buf));
845 	return (error);
846 }
847 
848 int
849 kern_fstatfs(int fd, struct statfs *buf)
850 {
851 	struct thread *td = curthread;
852 	struct proc *p = td->td_proc;
853 	struct file *fp;
854 	struct mount *mp;
855 	struct statfs *sp;
856 	int error;
857 
858 	KKASSERT(p);
859 	error = getvnode(p->p_fd, fd, &fp);
860 	if (error)
861 		return (error);
862 	mp = ((struct vnode *)fp->f_data)->v_mount;
863 	if (mp == NULL)
864 		return (EBADF);
865 	sp = &mp->mnt_stat;
866 	error = VFS_STATFS(mp, sp, td);
867 	if (error)
868 		return (error);
869 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
870 	bcopy(sp, buf, sizeof(*buf));
871 	/* Only root should have access to the fsid's. */
872 	if (suser(td))
873 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
874 	return (0);
875 }
876 
877 /*
878  * fstatfs_args(int fd, struct statfs *buf)
879  *
880  * Get filesystem statistics.
881  */
882 int
883 fstatfs(struct fstatfs_args *uap)
884 {
885 	struct statfs buf;
886 	int error;
887 
888 	error = kern_fstatfs(uap->fd, &buf);
889 
890 	if (error == 0)
891 		error = copyout(&buf, uap->buf, sizeof(*uap->buf));
892 	return (error);
893 }
894 
895 /*
896  * getfsstat_args(struct statfs *buf, long bufsize, int flags)
897  *
898  * Get statistics on all filesystems.
899  */
900 /* ARGSUSED */
901 int
902 getfsstat(struct getfsstat_args *uap)
903 {
904 	struct thread *td = curthread;
905 	struct mount *mp, *nmp;
906 	struct statfs *sp;
907 	caddr_t sfsp;
908 	lwkt_tokref ilock;
909 	long count, maxcount, error;
910 
911 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
912 	sfsp = (caddr_t)SCARG(uap, buf);
913 	count = 0;
914 	lwkt_gettoken(&ilock, &mountlist_token);
915 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
916 		if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
917 			nmp = TAILQ_NEXT(mp, mnt_list);
918 			continue;
919 		}
920 		if (sfsp && count < maxcount) {
921 			sp = &mp->mnt_stat;
922 			/*
923 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
924 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
925 			 * overrides MNT_WAIT.
926 			 */
927 			if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
928 			    (SCARG(uap, flags) & MNT_WAIT)) &&
929 			    (error = VFS_STATFS(mp, sp, td))) {
930 				lwkt_gettokref(&ilock);
931 				nmp = TAILQ_NEXT(mp, mnt_list);
932 				vfs_unbusy(mp, td);
933 				continue;
934 			}
935 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
936 			error = copyout(sp, sfsp, sizeof(*sp));
937 			if (error) {
938 				vfs_unbusy(mp, td);
939 				return (error);
940 			}
941 			sfsp += sizeof(*sp);
942 		}
943 		count++;
944 		lwkt_gettokref(&ilock);
945 		nmp = TAILQ_NEXT(mp, mnt_list);
946 		vfs_unbusy(mp, td);
947 	}
948 	lwkt_reltoken(&ilock);
949 	if (sfsp && count > maxcount)
950 		uap->sysmsg_result = maxcount;
951 	else
952 		uap->sysmsg_result = count;
953 	return (0);
954 }
955 
956 /*
957  * fchdir_args(int fd)
958  *
959  * Change current working directory to a given file descriptor.
960  */
961 /* ARGSUSED */
962 int
963 fchdir(struct fchdir_args *uap)
964 {
965 	struct thread *td = curthread;
966 	struct proc *p = td->td_proc;
967 	struct filedesc *fdp = p->p_fd;
968 	struct vnode *vp, *ovp;
969 	struct mount *mp;
970 	struct file *fp;
971 	struct namecache *ncp, *oncp;
972 	struct namecache *nct;
973 	int error;
974 
975 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
976 		return (error);
977 	vp = (struct vnode *)fp->f_data;
978 	vref(vp);
979 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
980 	if (vp->v_type != VDIR || fp->f_ncp == NULL)
981 		error = ENOTDIR;
982 	else
983 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
984 	if (error) {
985 		vput(vp);
986 		return (error);
987 	}
988 	ncp = cache_hold(fp->f_ncp);
989 	while (!error && (mp = vp->v_mountedhere) != NULL) {
990 		error = nlookup_mp(mp, &nct);
991 		if (error == 0) {
992 			cache_unlock(nct);	/* leave ref intact */
993 			vput(vp);
994 			vp = nct->nc_vp;
995 			error = vget(vp, LK_SHARED, td);
996 			KKASSERT(error == 0);
997 			cache_drop(ncp);
998 			ncp = nct;
999 		}
1000 	}
1001 	if (error == 0) {
1002 		ovp = fdp->fd_cdir;
1003 		oncp = fdp->fd_ncdir;
1004 		VOP_UNLOCK(vp, 0, td);	/* leave ref intact */
1005 		fdp->fd_cdir = vp;
1006 		fdp->fd_ncdir = ncp;
1007 		cache_drop(oncp);
1008 		vrele(ovp);
1009 	} else {
1010 		cache_drop(ncp);
1011 		vput(vp);
1012 	}
1013 	return (error);
1014 }
1015 
1016 int
1017 kern_chdir(struct nlookupdata *nd)
1018 {
1019 	struct thread *td = curthread;
1020 	struct proc *p = td->td_proc;
1021 	struct filedesc *fdp = p->p_fd;
1022 	struct vnode *vp, *ovp;
1023 	struct namecache *oncp;
1024 	int error;
1025 
1026 	if ((error = nlookup(nd)) != 0)
1027 		return (error);
1028 	if ((vp = nd->nl_ncp->nc_vp) == NULL)
1029 		return (ENOENT);
1030 	if ((error = vget(vp, LK_SHARED, td)) != 0)
1031 		return (error);
1032 
1033 	error = checkvp_chdir(vp, td);
1034 	VOP_UNLOCK(vp, 0, td);
1035 	if (error == 0) {
1036 		ovp = fdp->fd_cdir;
1037 		oncp = fdp->fd_ncdir;
1038 		cache_unlock(nd->nl_ncp);	/* leave reference intact */
1039 		fdp->fd_ncdir = nd->nl_ncp;
1040 		fdp->fd_cdir = vp;
1041 		cache_drop(oncp);
1042 		vrele(ovp);
1043 		nd->nl_ncp = NULL;
1044 	} else {
1045 		vrele(vp);
1046 	}
1047 	return (error);
1048 }
1049 
1050 /*
1051  * chdir_args(char *path)
1052  *
1053  * Change current working directory (``.'').
1054  */
1055 int
1056 chdir(struct chdir_args *uap)
1057 {
1058 	struct nlookupdata nd;
1059 	int error;
1060 
1061 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1062 	if (error == 0)
1063 		error = kern_chdir(&nd);
1064 	nlookup_done(&nd);
1065 	return (error);
1066 }
1067 
1068 /*
1069  * Helper function for raised chroot(2) security function:  Refuse if
1070  * any filedescriptors are open directories.
1071  */
1072 static int
1073 chroot_refuse_vdir_fds(fdp)
1074 	struct filedesc *fdp;
1075 {
1076 	struct vnode *vp;
1077 	struct file *fp;
1078 	int error;
1079 	int fd;
1080 
1081 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1082 		error = getvnode(fdp, fd, &fp);
1083 		if (error)
1084 			continue;
1085 		vp = (struct vnode *)fp->f_data;
1086 		if (vp->v_type != VDIR)
1087 			continue;
1088 		return(EPERM);
1089 	}
1090 	return (0);
1091 }
1092 
1093 /*
1094  * This sysctl determines if we will allow a process to chroot(2) if it
1095  * has a directory open:
1096  *	0: disallowed for all processes.
1097  *	1: allowed for processes that were not already chroot(2)'ed.
1098  *	2: allowed for all processes.
1099  */
1100 
1101 static int chroot_allow_open_directories = 1;
1102 
1103 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1104      &chroot_allow_open_directories, 0, "");
1105 
1106 /*
1107  * chroot to the specified namecache entry.  We obtain the vp from the
1108  * namecache data.  The passed ncp must be locked and referenced and will
1109  * remain locked and referenced on return.
1110  */
1111 int
1112 kern_chroot(struct namecache *ncp)
1113 {
1114 	struct thread *td = curthread;
1115 	struct proc *p = td->td_proc;
1116 	struct filedesc *fdp = p->p_fd;
1117 	struct vnode *vp;
1118 	int error;
1119 
1120 	/*
1121 	 * Only root can chroot
1122 	 */
1123 	if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1124 		return (error);
1125 
1126 	/*
1127 	 * Disallow open directory descriptors (fchdir() breakouts).
1128 	 */
1129 	if (chroot_allow_open_directories == 0 ||
1130 	   (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1131 		if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1132 			return (error);
1133 	}
1134 	if ((vp = ncp->nc_vp) == NULL)
1135 		return (ENOENT);
1136 
1137 	if ((error = vget(vp, LK_SHARED, td)) != 0)
1138 		return (error);
1139 
1140 	/*
1141 	 * Check the validity of vp as a directory to change to and
1142 	 * associate it with rdir/jdir.
1143 	 */
1144 	error = checkvp_chdir(vp, td);
1145 	VOP_UNLOCK(vp, 0, td);	/* leave reference intact */
1146 	if (error == 0) {
1147 		vrele(fdp->fd_rdir);
1148 		fdp->fd_rdir = vp;	/* reference inherited by fd_rdir */
1149 		cache_drop(fdp->fd_nrdir);
1150 		fdp->fd_nrdir = cache_hold(ncp);
1151 		if (fdp->fd_jdir == NULL) {
1152 			fdp->fd_jdir = vp;
1153 			vref(fdp->fd_jdir);
1154 			fdp->fd_njdir = cache_hold(ncp);
1155 		}
1156 	} else {
1157 		vrele(vp);
1158 	}
1159 	return (error);
1160 }
1161 
1162 /*
1163  * chroot_args(char *path)
1164  *
1165  * Change notion of root (``/'') directory.
1166  */
1167 /* ARGSUSED */
1168 int
1169 chroot(struct chroot_args *uap)
1170 {
1171 	struct thread *td = curthread;
1172 	struct nlookupdata nd;
1173 	int error;
1174 
1175 	KKASSERT(td->td_proc);
1176 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1177 	if (error) {
1178 		nlookup_done(&nd);
1179 		return(error);
1180 	}
1181 	error = nlookup(&nd);
1182 	if (error == 0)
1183 		error = kern_chroot(nd.nl_ncp);
1184 	nlookup_done(&nd);
1185 	return(error);
1186 }
1187 
1188 /*
1189  * Common routine for chroot and chdir.  Given a locked, referenced vnode,
1190  * determine whether it is legal to chdir to the vnode.  The vnode's state
1191  * is not changed by this call.
1192  */
1193 int
1194 checkvp_chdir(struct vnode *vp, struct thread *td)
1195 {
1196 	int error;
1197 
1198 	if (vp->v_type != VDIR)
1199 		error = ENOTDIR;
1200 	else
1201 		error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td);
1202 	return (error);
1203 }
1204 
1205 int
1206 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
1207 {
1208 	struct thread *td = curthread;
1209 	struct proc *p = td->td_proc;
1210 	struct filedesc *fdp = p->p_fd;
1211 	int cmode, flags;
1212 	struct file *nfp;
1213 	struct file *fp;
1214 	struct vnode *vp;
1215 	int type, indx, error;
1216 	struct flock lf;
1217 
1218 	if ((oflags & O_ACCMODE) == O_ACCMODE)
1219 		return (EINVAL);
1220 	flags = FFLAGS(oflags);
1221 	error = falloc(p, &nfp, NULL);
1222 	if (error)
1223 		return (error);
1224 	fp = nfp;
1225 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1226 
1227 	/*
1228 	 * XXX p_dupfd is a real mess.  It allows a device to return a
1229 	 * file descriptor to be duplicated rather then doing the open
1230 	 * itself.
1231 	 */
1232 	p->p_dupfd = -1;
1233 
1234 	/*
1235 	 * Call vn_open() to do the lookup and assign the vnode to the
1236 	 * file pointer.  vn_open() does not change the ref count on fp
1237 	 * and the vnode, on success, will be inherited by the file pointer
1238 	 * and unlocked.
1239 	 */
1240 	nd->nl_flags |= NLC_LOCKVP;
1241 	error = vn_open(nd, fp, flags, cmode);
1242 	nlookup_done(nd);
1243 	if (error) {
1244 		/*
1245 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1246 		 * responsible for dropping the old contents of ofiles[indx]
1247 		 * if it succeeds.
1248 		 *
1249 		 * Note that if fsetfd() succeeds it will add a ref to fp
1250 		 * which represents the fd_ofiles[] assignment.  We must still
1251 		 * drop our reference.
1252 		 */
1253 		if ((error == ENODEV || error == ENXIO) && p->p_dupfd >= 0) {
1254 			if (fsetfd(p, fp, &indx) == 0) {
1255 				error = dupfdopen(fdp, indx, p->p_dupfd, flags, error);
1256 				if (error == 0) {
1257 					*res = indx;
1258 					fdrop(fp, td);	/* our ref */
1259 					return (0);
1260 				}
1261 				if (fdp->fd_ofiles[indx] == fp) {
1262 					fdp->fd_ofiles[indx] = NULL;
1263 					fdrop(fp, td);	/* fd_ofiles[] ref */
1264 				}
1265 			}
1266 		}
1267 		fdrop(fp, td);	/* our ref */
1268 		if (error == ERESTART)
1269 			error = EINTR;
1270 		return (error);
1271 	}
1272 
1273 	/*
1274 	 * ref the vnode for ourselves so it can't be ripped out from under
1275 	 * is.  XXX need an ND flag to request that the vnode be returned
1276 	 * anyway.
1277 	 */
1278 	vp = (struct vnode *)fp->f_data;
1279 	vref(vp);
1280 	if ((error = fsetfd(p, fp, &indx)) != 0) {
1281 		fdrop(fp, td);
1282 		vrele(vp);
1283 		return (error);
1284 	}
1285 
1286 	/*
1287 	 * If no error occurs the vp will have been assigned to the file
1288 	 * pointer.
1289 	 */
1290 	p->p_dupfd = 0;
1291 
1292 	/*
1293 	 * There should be 2 references on the file, one from the descriptor
1294 	 * table, and one for us.
1295 	 *
1296 	 * Handle the case where someone closed the file (via its file
1297 	 * descriptor) while we were blocked.  The end result should look
1298 	 * like opening the file succeeded but it was immediately closed.
1299 	 */
1300 	if (fp->f_count == 1) {
1301 		KASSERT(fdp->fd_ofiles[indx] != fp,
1302 		    ("Open file descriptor lost all refs"));
1303 		vrele(vp);
1304 		fo_close(fp, td);
1305 		fdrop(fp, td);
1306 		*res = indx;
1307 		return 0;
1308 	}
1309 
1310 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1311 		lf.l_whence = SEEK_SET;
1312 		lf.l_start = 0;
1313 		lf.l_len = 0;
1314 		if (flags & O_EXLOCK)
1315 			lf.l_type = F_WRLCK;
1316 		else
1317 			lf.l_type = F_RDLCK;
1318 		type = F_FLOCK;
1319 		if ((flags & FNONBLOCK) == 0)
1320 			type |= F_WAIT;
1321 
1322 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1323 			/*
1324 			 * lock request failed.  Normally close the descriptor
1325 			 * but handle the case where someone might have dup()d
1326 			 * it when we weren't looking.  One reference is
1327 			 * owned by the descriptor array, the other by us.
1328 			 */
1329 			vrele(vp);
1330 			if (fdp->fd_ofiles[indx] == fp) {
1331 				fdp->fd_ofiles[indx] = NULL;
1332 				fdrop(fp, td);
1333 			}
1334 			fdrop(fp, td);
1335 			return (error);
1336 		}
1337 		fp->f_flag |= FHASLOCK;
1338 	}
1339 	/* assert that vn_open created a backing object if one is needed */
1340 	KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1341 		("open: vmio vnode has no backing object after vn_open"));
1342 
1343 	vrele(vp);
1344 
1345 	/*
1346 	 * release our private reference, leaving the one associated with the
1347 	 * descriptor table intact.
1348 	 */
1349 	fdrop(fp, td);
1350 	*res = indx;
1351 	return (0);
1352 }
1353 
1354 /*
1355  * open_args(char *path, int flags, int mode)
1356  *
1357  * Check permissions, allocate an open file structure,
1358  * and call the device open routine if any.
1359  */
1360 int
1361 open(struct open_args *uap)
1362 {
1363 	struct nlookupdata nd;
1364 	int error;
1365 
1366 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1367 	if (error == 0) {
1368 		error = kern_open(&nd, uap->flags,
1369 				    uap->mode, &uap->sysmsg_result);
1370 	}
1371 	nlookup_done(&nd);
1372 	return (error);
1373 }
1374 
1375 int
1376 kern_mknod(struct nlookupdata *nd, int mode, int dev)
1377 {
1378 	struct namecache *ncp;
1379 	struct thread *td = curthread;
1380 	struct proc *p = td->td_proc;
1381 	struct vnode *vp;
1382 	struct vattr vattr;
1383 	int error;
1384 	int whiteout = 0;
1385 
1386 	KKASSERT(p);
1387 
1388 	switch (mode & S_IFMT) {
1389 	case S_IFCHR:
1390 	case S_IFBLK:
1391 		error = suser(td);
1392 		break;
1393 	default:
1394 		error = suser_cred(p->p_ucred, PRISON_ROOT);
1395 		break;
1396 	}
1397 	if (error)
1398 		return (error);
1399 
1400 	bwillwrite();
1401 	nd->nl_flags |= NLC_CREATE;
1402 	if ((error = nlookup(nd)) != 0)
1403 		return (error);
1404 	ncp = nd->nl_ncp;
1405 	if (ncp->nc_vp)
1406 		return (EEXIST);
1407 
1408 	VATTR_NULL(&vattr);
1409 	vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1410 	vattr.va_rdev = dev;
1411 	whiteout = 0;
1412 
1413 	switch (mode & S_IFMT) {
1414 	case S_IFMT:	/* used by badsect to flag bad sectors */
1415 		vattr.va_type = VBAD;
1416 		break;
1417 	case S_IFCHR:
1418 		vattr.va_type = VCHR;
1419 		break;
1420 	case S_IFBLK:
1421 		vattr.va_type = VBLK;
1422 		break;
1423 	case S_IFWHT:
1424 		whiteout = 1;
1425 		break;
1426 	default:
1427 		error = EINVAL;
1428 		break;
1429 	}
1430 	if (error == 0) {
1431 		if (whiteout) {
1432 			error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE);
1433 		} else {
1434 			vp = NULL;
1435 			error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
1436 			if (error == 0)
1437 				vput(vp);
1438 		}
1439 	}
1440 	return (error);
1441 }
1442 
1443 /*
1444  * mknod_args(char *path, int mode, int dev)
1445  *
1446  * Create a special file.
1447  */
1448 int
1449 mknod(struct mknod_args *uap)
1450 {
1451 	struct nlookupdata nd;
1452 	int error;
1453 
1454 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1455 	if (error == 0)
1456 		error = kern_mknod(&nd, uap->mode, uap->dev);
1457 	nlookup_done(&nd);
1458 	return (error);
1459 }
1460 
1461 int
1462 kern_mkfifo(struct nlookupdata *nd, int mode)
1463 {
1464 	struct namecache *ncp;
1465 	struct thread *td = curthread;
1466 	struct proc *p = td->td_proc;
1467 	struct vattr vattr;
1468 	struct vnode *vp;
1469 	int error;
1470 
1471 	bwillwrite();
1472 
1473 	nd->nl_flags |= NLC_CREATE;
1474 	if ((error = nlookup(nd)) != 0)
1475 		return (error);
1476 	ncp = nd->nl_ncp;
1477 	if (ncp->nc_vp)
1478 		return (EEXIST);
1479 
1480 	VATTR_NULL(&vattr);
1481 	vattr.va_type = VFIFO;
1482 	vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1483 	vp = NULL;
1484 	error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
1485 	if (error == 0)
1486 		vput(vp);
1487 	return (error);
1488 }
1489 
1490 /*
1491  * mkfifo_args(char *path, int mode)
1492  *
1493  * Create a named pipe.
1494  */
1495 int
1496 mkfifo(struct mkfifo_args *uap)
1497 {
1498 	struct nlookupdata nd;
1499 	int error;
1500 
1501 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1502 	if (error == 0)
1503 		error = kern_mkfifo(&nd, uap->mode);
1504 	nlookup_done(&nd);
1505 	return (error);
1506 }
1507 
1508 int
1509 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
1510 {
1511 	struct thread *td = curthread;
1512 	struct vnode *vp;
1513 	int error;
1514 
1515 	/*
1516 	 * Lookup the source and obtained a locked vnode.
1517 	 *
1518 	 * XXX relookup on vget failure / race ?
1519 	 */
1520 	bwillwrite();
1521 	if ((error = nlookup(nd)) != 0)
1522 		return (error);
1523 	vp = nd->nl_ncp->nc_vp;
1524 	KKASSERT(vp != NULL);
1525 	if (vp->v_type == VDIR)
1526 		return (EPERM);		/* POSIX */
1527 	if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0)
1528 		return (error);
1529 
1530 	/*
1531 	 * Unlock the source so we can lookup the target without deadlocking
1532 	 * (XXX vp is locked already, possible other deadlock?).  The target
1533 	 * must not exist.
1534 	 */
1535 	KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1536 	nd->nl_flags &= ~NLC_NCPISLOCKED;
1537 	cache_unlock(nd->nl_ncp);
1538 
1539 	linknd->nl_flags |= NLC_CREATE;
1540 	if ((error = nlookup(linknd)) != 0) {
1541 		vput(vp);
1542 		return (error);
1543 	}
1544 	if (linknd->nl_ncp->nc_vp) {
1545 		vput(vp);
1546 		return (EEXIST);
1547 	}
1548 
1549 	/*
1550 	 * Finally run the new API VOP.
1551 	 */
1552 	error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred);
1553 	vput(vp);
1554 	return (error);
1555 }
1556 
1557 /*
1558  * link_args(char *path, char *link)
1559  *
1560  * Make a hard file link.
1561  */
1562 int
1563 link(struct link_args *uap)
1564 {
1565 	struct nlookupdata nd, linknd;
1566 	int error;
1567 
1568 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1569 	if (error == 0) {
1570 		error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1571 		if (error == 0)
1572 			error = kern_link(&nd, &linknd);
1573 		nlookup_done(&linknd);
1574 	}
1575 	nlookup_done(&nd);
1576 	return (error);
1577 }
1578 
1579 int
1580 kern_symlink(struct nlookupdata *nd, char *path, int mode)
1581 {
1582 	struct namecache *ncp;
1583 	struct vattr vattr;
1584 	struct vnode *vp;
1585 	int error;
1586 
1587 	bwillwrite();
1588 	nd->nl_flags |= NLC_CREATE;
1589 	if ((error = nlookup(nd)) != 0)
1590 		return (error);
1591 	ncp = nd->nl_ncp;
1592 	if (ncp->nc_vp)
1593 		return (EEXIST);
1594 
1595 	VATTR_NULL(&vattr);
1596 	vattr.va_mode = mode;
1597 	error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path);
1598 	if (error == 0)
1599 		vput(vp);
1600 	return (error);
1601 }
1602 
1603 /*
1604  * symlink(char *path, char *link)
1605  *
1606  * Make a symbolic link.
1607  */
1608 int
1609 symlink(struct symlink_args *uap)
1610 {
1611 	struct thread *td = curthread;
1612 	struct nlookupdata nd;
1613 	char *path;
1614 	int error;
1615 	int mode;
1616 
1617 	path = zalloc(namei_zone);
1618 	error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
1619 	if (error == 0) {
1620 		error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1621 		if (error == 0) {
1622 			mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1623 			error = kern_symlink(&nd, path, mode);
1624 		}
1625 		nlookup_done(&nd);
1626 	}
1627 	zfree(namei_zone, path);
1628 	return (error);
1629 }
1630 
1631 /*
1632  * undelete_args(char *path)
1633  *
1634  * Delete a whiteout from the filesystem.
1635  */
1636 /* ARGSUSED */
1637 int
1638 undelete(struct undelete_args *uap)
1639 {
1640 	struct nlookupdata nd;
1641 	int error;
1642 
1643 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0);
1644 	bwillwrite();
1645 	nd.nl_flags |= NLC_DELETE;
1646 	if (error == 0)
1647 		error = nlookup(&nd);
1648 	if (error == 0)
1649 		error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE);
1650 	nlookup_done(&nd);
1651 	return (error);
1652 }
1653 
1654 int
1655 kern_unlink(struct nlookupdata *nd)
1656 {
1657 	struct namecache *ncp;
1658 	int error;
1659 
1660 	bwillwrite();
1661 	nd->nl_flags |= NLC_DELETE;
1662 	if ((error = nlookup(nd)) != 0)
1663 		return (error);
1664 	ncp = nd->nl_ncp;
1665 	error = VOP_NREMOVE(ncp, nd->nl_cred);
1666 	return (error);
1667 }
1668 
1669 /*
1670  * unlink_args(char *path)
1671  *
1672  * Delete a name from the filesystem.
1673  */
1674 int
1675 unlink(struct unlink_args *uap)
1676 {
1677 	struct nlookupdata nd;
1678 	int error;
1679 
1680 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1681 	if (error == 0)
1682 		error = kern_unlink(&nd);
1683 	nlookup_done(&nd);
1684 	return (error);
1685 }
1686 
1687 int
1688 kern_lseek(int fd, off_t offset, int whence, off_t *res)
1689 {
1690 	struct thread *td = curthread;
1691 	struct proc *p = td->td_proc;
1692 	struct filedesc *fdp = p->p_fd;
1693 	struct file *fp;
1694 	struct vattr vattr;
1695 	int error;
1696 
1697 	if ((u_int)fd >= fdp->fd_nfiles ||
1698 	    (fp = fdp->fd_ofiles[fd]) == NULL)
1699 		return (EBADF);
1700 	if (fp->f_type != DTYPE_VNODE)
1701 		return (ESPIPE);
1702 	switch (whence) {
1703 	case L_INCR:
1704 		fp->f_offset += offset;
1705 		break;
1706 	case L_XTND:
1707 		error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td);
1708 		if (error)
1709 			return (error);
1710 		fp->f_offset = offset + vattr.va_size;
1711 		break;
1712 	case L_SET:
1713 		fp->f_offset = offset;
1714 		break;
1715 	default:
1716 		return (EINVAL);
1717 	}
1718 	*res = fp->f_offset;
1719 	return (0);
1720 }
1721 
1722 /*
1723  * lseek_args(int fd, int pad, off_t offset, int whence)
1724  *
1725  * Reposition read/write file offset.
1726  */
1727 int
1728 lseek(struct lseek_args *uap)
1729 {
1730 	int error;
1731 
1732 	error = kern_lseek(uap->fd, uap->offset, uap->whence,
1733 	    &uap->sysmsg_offset);
1734 
1735 	return (error);
1736 }
1737 
1738 int
1739 kern_access(struct nlookupdata *nd, int aflags)
1740 {
1741 	struct thread *td = curthread;
1742 	struct vnode *vp;
1743 	int error, flags;
1744 
1745 	if ((error = nlookup(nd)) != 0)
1746 		return (error);
1747 retry:
1748 	error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
1749 	if (error)
1750 		return (error);
1751 
1752 	/* Flags == 0 means only check for existence. */
1753 	if (aflags) {
1754 		flags = 0;
1755 		if (aflags & R_OK)
1756 			flags |= VREAD;
1757 		if (aflags & W_OK)
1758 			flags |= VWRITE;
1759 		if (aflags & X_OK)
1760 			flags |= VEXEC;
1761 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1762 			error = VOP_ACCESS(vp, flags, nd->nl_cred, td);
1763 
1764 		/*
1765 		 * If the file handle is stale we have to re-resolve the
1766 		 * entry.  This is a hack at the moment.
1767 		 */
1768 		if (error == ESTALE) {
1769 			cache_setunresolved(nd->nl_ncp);
1770 			error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1771 			if (error == 0) {
1772 				vput(vp);
1773 				vp = NULL;
1774 				goto retry;
1775 			}
1776 		}
1777 	}
1778 	vput(vp);
1779 	return (error);
1780 }
1781 
1782 /*
1783  * access_args(char *path, int flags)
1784  *
1785  * Check access permissions.
1786  */
1787 int
1788 access(struct access_args *uap)
1789 {
1790 	struct nlookupdata nd;
1791 	int error;
1792 
1793 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1794 	if (error == 0)
1795 		error = kern_access(&nd, uap->flags);
1796 	nlookup_done(&nd);
1797 	return (error);
1798 }
1799 
1800 int
1801 kern_stat(struct nlookupdata *nd, struct stat *st)
1802 {
1803 	int error;
1804 	struct vnode *vp;
1805 	thread_t td;
1806 
1807 	if ((error = nlookup(nd)) != 0)
1808 		return (error);
1809 again:
1810 	if ((vp = nd->nl_ncp->nc_vp) == NULL)
1811 		return (ENOENT);
1812 
1813 	td = curthread;
1814 	if ((error = vget(vp, LK_SHARED, td)) != 0)
1815 		return (error);
1816 	error = vn_stat(vp, st, td);
1817 
1818 	/*
1819 	 * If the file handle is stale we have to re-resolve the entry.  This
1820 	 * is a hack at the moment.
1821 	 */
1822 	if (error == ESTALE) {
1823 		cache_setunresolved(nd->nl_ncp);
1824 		error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1825 		if (error == 0) {
1826 			vput(vp);
1827 			goto again;
1828 		}
1829 	}
1830 	vput(vp);
1831 	return (error);
1832 }
1833 
1834 /*
1835  * stat_args(char *path, struct stat *ub)
1836  *
1837  * Get file status; this version follows links.
1838  */
1839 int
1840 stat(struct stat_args *uap)
1841 {
1842 	struct nlookupdata nd;
1843 	struct stat st;
1844 	int error;
1845 
1846 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1847 	if (error == 0) {
1848 		error = kern_stat(&nd, &st);
1849 		if (error == 0)
1850 			error = copyout(&st, uap->ub, sizeof(*uap->ub));
1851 	}
1852 	nlookup_done(&nd);
1853 	return (error);
1854 }
1855 
1856 /*
1857  * lstat_args(char *path, struct stat *ub)
1858  *
1859  * Get file status; this version does not follow links.
1860  */
1861 int
1862 lstat(struct lstat_args *uap)
1863 {
1864 	struct nlookupdata nd;
1865 	struct stat st;
1866 	int error;
1867 
1868 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1869 	if (error == 0) {
1870 		error = kern_stat(&nd, &st);
1871 		if (error == 0)
1872 			error = copyout(&st, uap->ub, sizeof(*uap->ub));
1873 	}
1874 	nlookup_done(&nd);
1875 	return (error);
1876 }
1877 
1878 void
1879 cvtnstat(sb, nsb)
1880 	struct stat *sb;
1881 	struct nstat *nsb;
1882 {
1883 	nsb->st_dev = sb->st_dev;
1884 	nsb->st_ino = sb->st_ino;
1885 	nsb->st_mode = sb->st_mode;
1886 	nsb->st_nlink = sb->st_nlink;
1887 	nsb->st_uid = sb->st_uid;
1888 	nsb->st_gid = sb->st_gid;
1889 	nsb->st_rdev = sb->st_rdev;
1890 	nsb->st_atimespec = sb->st_atimespec;
1891 	nsb->st_mtimespec = sb->st_mtimespec;
1892 	nsb->st_ctimespec = sb->st_ctimespec;
1893 	nsb->st_size = sb->st_size;
1894 	nsb->st_blocks = sb->st_blocks;
1895 	nsb->st_blksize = sb->st_blksize;
1896 	nsb->st_flags = sb->st_flags;
1897 	nsb->st_gen = sb->st_gen;
1898 	nsb->st_qspare[0] = sb->st_qspare[0];
1899 	nsb->st_qspare[1] = sb->st_qspare[1];
1900 }
1901 
1902 /*
1903  * nstat_args(char *path, struct nstat *ub)
1904  */
1905 /* ARGSUSED */
1906 int
1907 nstat(struct nstat_args *uap)
1908 {
1909 	struct thread *td = curthread;
1910 	struct vnode *vp;
1911 	struct stat sb;
1912 	struct nstat nsb;
1913 	struct nlookupdata nd;
1914 	int error;
1915 
1916 	vp = NULL;
1917 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1918 	if (error == 0)
1919 		error = nlookup(&nd);
1920 	if (error == 0)
1921 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1922 	nlookup_done(&nd);
1923 	if (error == 0) {
1924 		error = vn_stat(vp, &sb, td);
1925 		vput(vp);
1926 		if (error == 0) {
1927 			cvtnstat(&sb, &nsb);
1928 			error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb));
1929 		}
1930 	}
1931 	return (error);
1932 }
1933 
1934 /*
1935  * lstat_args(char *path, struct stat *ub)
1936  *
1937  * Get file status; this version does not follow links.
1938  */
1939 /* ARGSUSED */
1940 int
1941 nlstat(struct nlstat_args *uap)
1942 {
1943 	struct thread *td = curthread;
1944 	struct vnode *vp;
1945 	struct stat sb;
1946 	struct nstat nsb;
1947 	struct nlookupdata nd;
1948 	int error;
1949 
1950 	vp = NULL;
1951 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0);
1952 	if (error == 0)
1953 		error = nlookup(&nd);
1954 	if (error == 0)
1955 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1956 	nlookup_done(&nd);
1957 	if (error == 0) {
1958 		error = vn_stat(vp, &sb, td);
1959 		vput(vp);
1960 		if (error == 0) {
1961 			cvtnstat(&sb, &nsb);
1962 			error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb));
1963 		}
1964 	}
1965 	return (error);
1966 }
1967 
1968 /*
1969  * pathconf_Args(char *path, int name)
1970  *
1971  * Get configurable pathname variables.
1972  */
1973 /* ARGSUSED */
1974 int
1975 pathconf(struct pathconf_args *uap)
1976 {
1977 	struct nlookupdata nd;
1978 	struct vnode *vp;
1979 	int error;
1980 
1981 	vp = NULL;
1982 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1983 	if (error == 0)
1984 		error = nlookup(&nd);
1985 	if (error == 0)
1986 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1987 	nlookup_done(&nd);
1988 	if (error == 0) {
1989 		error = VOP_PATHCONF(vp, SCARG(uap, name), uap->sysmsg_fds);
1990 		vput(vp);
1991 	}
1992 	return (error);
1993 }
1994 
1995 /*
1996  * XXX: daver
1997  * kern_readlink isn't properly split yet.  There is a copyin burried
1998  * in VOP_READLINK().
1999  */
2000 int
2001 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
2002 {
2003 	struct thread *td = curthread;
2004 	struct proc *p = td->td_proc;
2005 	struct vnode *vp;
2006 	struct iovec aiov;
2007 	struct uio auio;
2008 	int error;
2009 
2010 	if ((error = nlookup(nd)) != 0)
2011 		return (error);
2012 	error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
2013 	if (error)
2014 		return (error);
2015 	if (vp->v_type != VLNK) {
2016 		error = EINVAL;
2017 	} else {
2018 		aiov.iov_base = buf;
2019 		aiov.iov_len = count;
2020 		auio.uio_iov = &aiov;
2021 		auio.uio_iovcnt = 1;
2022 		auio.uio_offset = 0;
2023 		auio.uio_rw = UIO_READ;
2024 		auio.uio_segflg = UIO_USERSPACE;
2025 		auio.uio_td = td;
2026 		auio.uio_resid = count;
2027 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2028 	}
2029 	vput(vp);
2030 	*res = count - auio.uio_resid;
2031 	return (error);
2032 }
2033 
2034 /*
2035  * readlink_args(char *path, char *buf, int count)
2036  *
2037  * Return target name of a symbolic link.
2038  */
2039 int
2040 readlink(struct readlink_args *uap)
2041 {
2042 	struct nlookupdata nd;
2043 	int error;
2044 
2045 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2046 	if (error == 0) {
2047 		error = kern_readlink(&nd, uap->buf, uap->count,
2048 					&uap->sysmsg_result);
2049 	}
2050 	nlookup_done(&nd);
2051 	return (error);
2052 }
2053 
2054 static int
2055 setfflags(struct vnode *vp, int flags)
2056 {
2057 	struct thread *td = curthread;
2058 	struct proc *p = td->td_proc;
2059 	int error;
2060 	struct vattr vattr;
2061 
2062 	/*
2063 	 * Prevent non-root users from setting flags on devices.  When
2064 	 * a device is reused, users can retain ownership of the device
2065 	 * if they are allowed to set flags and programs assume that
2066 	 * chown can't fail when done as root.
2067 	 */
2068 	if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2069 	    ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
2070 		return (error);
2071 
2072 	/*
2073 	 * note: vget is required for any operation that might mod the vnode
2074 	 * so VINACTIVE is properly cleared.
2075 	 */
2076 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2077 	if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2078 		VATTR_NULL(&vattr);
2079 		vattr.va_flags = flags;
2080 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2081 		vput(vp);
2082 	}
2083 	return (error);
2084 }
2085 
2086 /*
2087  * chflags(char *path, int flags)
2088  *
2089  * Change flags of a file given a path name.
2090  */
2091 /* ARGSUSED */
2092 int
2093 chflags(struct chflags_args *uap)
2094 {
2095 	struct nlookupdata nd;
2096 	struct vnode *vp;
2097 	int error;
2098 
2099 	vp = NULL;
2100 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
2101 	/* XXX Add NLC flag indicating modifying operation? */
2102 	if (error == 0)
2103 		error = nlookup(&nd);
2104 	if (error == 0)
2105 		error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2106 	nlookup_done(&nd);
2107 	if (error == 0) {
2108 		error = setfflags(vp, SCARG(uap, flags));
2109 		vrele(vp);
2110 	}
2111 	return (error);
2112 }
2113 
2114 /*
2115  * fchflags_args(int fd, int flags)
2116  *
2117  * Change flags of a file given a file descriptor.
2118  */
2119 /* ARGSUSED */
2120 int
2121 fchflags(struct fchflags_args *uap)
2122 {
2123 	struct thread *td = curthread;
2124 	struct proc *p = td->td_proc;
2125 	struct file *fp;
2126 	int error;
2127 
2128 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2129 		return (error);
2130 	return setfflags((struct vnode *) fp->f_data, SCARG(uap, flags));
2131 }
2132 
2133 static int
2134 setfmode(struct vnode *vp, int mode)
2135 {
2136 	struct thread *td = curthread;
2137 	struct proc *p = td->td_proc;
2138 	int error;
2139 	struct vattr vattr;
2140 
2141 	/*
2142 	 * note: vget is required for any operation that might mod the vnode
2143 	 * so VINACTIVE is properly cleared.
2144 	 */
2145 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2146 	if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2147 		VATTR_NULL(&vattr);
2148 		vattr.va_mode = mode & ALLPERMS;
2149 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2150 		vput(vp);
2151 	}
2152 	return error;
2153 }
2154 
2155 int
2156 kern_chmod(struct nlookupdata *nd, int mode)
2157 {
2158 	struct vnode *vp;
2159 	int error;
2160 
2161 	/* XXX Add NLC flag indicating modifying operation? */
2162 	if ((error = nlookup(nd)) != 0)
2163 		return (error);
2164 	if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2165 		return (error);
2166 	error = setfmode(vp, mode);
2167 	vrele(vp);
2168 	return (error);
2169 }
2170 
2171 /*
2172  * chmod_args(char *path, int mode)
2173  *
2174  * Change mode of a file given path name.
2175  */
2176 /* ARGSUSED */
2177 int
2178 chmod(struct chmod_args *uap)
2179 {
2180 	struct nlookupdata nd;
2181 	int error;
2182 
2183 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2184 	if (error == 0)
2185 		error = kern_chmod(&nd, uap->mode);
2186 	nlookup_done(&nd);
2187 	return (error);
2188 }
2189 
2190 /*
2191  * lchmod_args(char *path, int mode)
2192  *
2193  * Change mode of a file given path name (don't follow links.)
2194  */
2195 /* ARGSUSED */
2196 int
2197 lchmod(struct lchmod_args *uap)
2198 {
2199 	struct nlookupdata nd;
2200 	int error;
2201 
2202 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2203 	if (error == 0)
2204 		error = kern_chmod(&nd, uap->mode);
2205 	nlookup_done(&nd);
2206 	return (error);
2207 }
2208 
2209 /*
2210  * fchmod_args(int fd, int mode)
2211  *
2212  * Change mode of a file given a file descriptor.
2213  */
2214 /* ARGSUSED */
2215 int
2216 fchmod(struct fchmod_args *uap)
2217 {
2218 	struct thread *td = curthread;
2219 	struct proc *p = td->td_proc;
2220 	struct file *fp;
2221 	int error;
2222 
2223 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2224 		return (error);
2225 	return setfmode((struct vnode *)fp->f_data, SCARG(uap, mode));
2226 }
2227 
2228 static int
2229 setfown(struct vnode *vp, uid_t uid, gid_t gid)
2230 {
2231 	struct thread *td = curthread;
2232 	struct proc *p = td->td_proc;
2233 	int error;
2234 	struct vattr vattr;
2235 
2236 	/*
2237 	 * note: vget is required for any operation that might mod the vnode
2238 	 * so VINACTIVE is properly cleared.
2239 	 */
2240 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2241 	if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2242 		VATTR_NULL(&vattr);
2243 		vattr.va_uid = uid;
2244 		vattr.va_gid = gid;
2245 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2246 		vput(vp);
2247 	}
2248 	return error;
2249 }
2250 
2251 int
2252 kern_chown(struct nlookupdata *nd, int uid, int gid)
2253 {
2254 	struct vnode *vp;
2255 	int error;
2256 
2257 	/* XXX Add NLC flag indicating modifying operation? */
2258 	if ((error = nlookup(nd)) != 0)
2259 		return (error);
2260 	if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2261 		return (error);
2262 	error = setfown(vp, uid, gid);
2263 	vrele(vp);
2264 	return (error);
2265 }
2266 
2267 /*
2268  * chown(char *path, int uid, int gid)
2269  *
2270  * Set ownership given a path name.
2271  */
2272 int
2273 chown(struct chown_args *uap)
2274 {
2275 	struct nlookupdata nd;
2276 	int error;
2277 
2278 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2279 	if (error == 0)
2280 		error = kern_chown(&nd, uap->uid, uap->gid);
2281 	nlookup_done(&nd);
2282 	return (error);
2283 }
2284 
2285 /*
2286  * lchown_args(char *path, int uid, int gid)
2287  *
2288  * Set ownership given a path name, do not cross symlinks.
2289  */
2290 int
2291 lchown(struct lchown_args *uap)
2292 {
2293 	struct nlookupdata nd;
2294 	int error;
2295 
2296 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2297 	if (error == 0)
2298 		error = kern_chown(&nd, uap->uid, uap->gid);
2299 	nlookup_done(&nd);
2300 	return (error);
2301 }
2302 
2303 /*
2304  * fchown_args(int fd, int uid, int gid)
2305  *
2306  * Set ownership given a file descriptor.
2307  */
2308 /* ARGSUSED */
2309 int
2310 fchown(struct fchown_args *uap)
2311 {
2312 	struct thread *td = curthread;
2313 	struct proc *p = td->td_proc;
2314 	struct file *fp;
2315 	int error;
2316 
2317 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2318 		return (error);
2319 	return setfown((struct vnode *)fp->f_data,
2320 		SCARG(uap, uid), SCARG(uap, gid));
2321 }
2322 
2323 static int
2324 getutimes(const struct timeval *tvp, struct timespec *tsp)
2325 {
2326 	struct timeval tv[2];
2327 
2328 	if (tvp == NULL) {
2329 		microtime(&tv[0]);
2330 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2331 		tsp[1] = tsp[0];
2332 	} else {
2333 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2334 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2335 	}
2336 	return 0;
2337 }
2338 
2339 static int
2340 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
2341 {
2342 	struct thread *td = curthread;
2343 	struct proc *p = td->td_proc;
2344 	int error;
2345 	struct vattr vattr;
2346 
2347 	/*
2348 	 * note: vget is required for any operation that might mod the vnode
2349 	 * so VINACTIVE is properly cleared.
2350 	 */
2351 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2352 	if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2353 		VATTR_NULL(&vattr);
2354 		vattr.va_atime = ts[0];
2355 		vattr.va_mtime = ts[1];
2356 		if (nullflag)
2357 			vattr.va_vaflags |= VA_UTIMES_NULL;
2358 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2359 		vput(vp);
2360 	}
2361 	return error;
2362 }
2363 
2364 int
2365 kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
2366 {
2367 	struct timespec ts[2];
2368 	struct vnode *vp;
2369 	int error;
2370 
2371 	if ((error = getutimes(tptr, ts)) != 0)
2372 		return (error);
2373 	/* XXX Add NLC flag indicating modifying operation? */
2374 	if ((error = nlookup(nd)) != 0)
2375 		return (error);
2376 	if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2377 		return (error);
2378 	error = setutimes(vp, ts, tptr == NULL);
2379 	vrele(vp);
2380 	return (error);
2381 }
2382 
2383 /*
2384  * utimes_args(char *path, struct timeval *tptr)
2385  *
2386  * Set the access and modification times of a file.
2387  */
2388 int
2389 utimes(struct utimes_args *uap)
2390 {
2391 	struct timeval tv[2];
2392 	struct nlookupdata nd;
2393 	int error;
2394 
2395 	if (uap->tptr) {
2396  		error = copyin(uap->tptr, tv, sizeof(tv));
2397 		if (error)
2398 			return (error);
2399 	}
2400 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2401 	if (error == 0)
2402 		error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2403 	nlookup_done(&nd);
2404 	return (error);
2405 }
2406 
2407 /*
2408  * lutimes_args(char *path, struct timeval *tptr)
2409  *
2410  * Set the access and modification times of a file.
2411  */
2412 int
2413 lutimes(struct lutimes_args *uap)
2414 {
2415 	struct timeval tv[2];
2416 	struct nlookupdata nd;
2417 	int error;
2418 
2419 	if (uap->tptr) {
2420 		error = copyin(uap->tptr, tv, sizeof(tv));
2421 		if (error)
2422 			return (error);
2423 	}
2424 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2425 	if (error == 0)
2426 		error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2427 	nlookup_done(&nd);
2428 	return (error);
2429 }
2430 
2431 int
2432 kern_futimes(int fd, struct timeval *tptr)
2433 {
2434 	struct thread *td = curthread;
2435 	struct proc *p = td->td_proc;
2436 	struct timespec ts[2];
2437 	struct file *fp;
2438 	int error;
2439 
2440 	error = getutimes(tptr, ts);
2441 	if (error)
2442 		return (error);
2443 	error = getvnode(p->p_fd, fd, &fp);
2444 	if (error)
2445 		return (error);
2446 	error =  setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2447 	return (error);
2448 }
2449 
2450 /*
2451  * futimes_args(int fd, struct timeval *tptr)
2452  *
2453  * Set the access and modification times of a file.
2454  */
2455 int
2456 futimes(struct futimes_args *uap)
2457 {
2458 	struct timeval tv[2];
2459 	int error;
2460 
2461 	if (uap->tptr) {
2462 		error = copyin(uap->tptr, tv, sizeof(tv));
2463 		if (error)
2464 			return (error);
2465 	}
2466 
2467 	error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
2468 
2469 	return (error);
2470 }
2471 
2472 int
2473 kern_truncate(struct nlookupdata *nd, off_t length)
2474 {
2475 	struct vnode *vp;
2476 	struct vattr vattr;
2477 	int error;
2478 
2479 	if (length < 0)
2480 		return(EINVAL);
2481 	/* XXX Add NLC flag indicating modifying operation? */
2482 	if ((error = nlookup(nd)) != 0)
2483 		return (error);
2484 	if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2485 		return (error);
2486 	VOP_LEASE(vp, nd->nl_td, nd->nl_cred, LEASE_WRITE);
2487 	if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, nd->nl_td)) != 0) {
2488 		vrele(vp);
2489 		return (error);
2490 	}
2491 	if (vp->v_type == VDIR) {
2492 		error = EISDIR;
2493 	} else if ((error = vn_writechk(vp)) == 0 &&
2494 	    (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred, nd->nl_td)) == 0) {
2495 		VATTR_NULL(&vattr);
2496 		vattr.va_size = length;
2497 		error = VOP_SETATTR(vp, &vattr, nd->nl_cred, nd->nl_td);
2498 	}
2499 	vput(vp);
2500 	return (error);
2501 }
2502 
2503 /*
2504  * truncate(char *path, int pad, off_t length)
2505  *
2506  * Truncate a file given its path name.
2507  */
2508 int
2509 truncate(struct truncate_args *uap)
2510 {
2511 	struct nlookupdata nd;
2512 	int error;
2513 
2514 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2515 	if (error == 0)
2516 		error = kern_truncate(&nd, uap->length);
2517 	nlookup_done(&nd);
2518 	return error;
2519 }
2520 
2521 int
2522 kern_ftruncate(int fd, off_t length)
2523 {
2524 	struct thread *td = curthread;
2525 	struct proc *p = td->td_proc;
2526 	struct vattr vattr;
2527 	struct vnode *vp;
2528 	struct file *fp;
2529 	int error;
2530 
2531 	if (length < 0)
2532 		return(EINVAL);
2533 	if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
2534 		return (error);
2535 	if ((fp->f_flag & FWRITE) == 0)
2536 		return (EINVAL);
2537 	vp = (struct vnode *)fp->f_data;
2538 	VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2539 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2540 	if (vp->v_type == VDIR)
2541 		error = EISDIR;
2542 	else if ((error = vn_writechk(vp)) == 0) {
2543 		VATTR_NULL(&vattr);
2544 		vattr.va_size = length;
2545 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
2546 	}
2547 	VOP_UNLOCK(vp, 0, td);
2548 	return (error);
2549 }
2550 
2551 /*
2552  * ftruncate_args(int fd, int pad, off_t length)
2553  *
2554  * Truncate a file given a file descriptor.
2555  */
2556 int
2557 ftruncate(struct ftruncate_args *uap)
2558 {
2559 	int error;
2560 
2561 	error = kern_ftruncate(uap->fd, uap->length);
2562 
2563 	return (error);
2564 }
2565 
2566 /*
2567  * fsync(int fd)
2568  *
2569  * Sync an open file.
2570  */
2571 /* ARGSUSED */
2572 int
2573 fsync(struct fsync_args *uap)
2574 {
2575 	struct thread *td = curthread;
2576 	struct proc *p = td->td_proc;
2577 	struct vnode *vp;
2578 	struct file *fp;
2579 	vm_object_t obj;
2580 	int error;
2581 
2582 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2583 		return (error);
2584 	vp = (struct vnode *)fp->f_data;
2585 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2586 	if (VOP_GETVOBJECT(vp, &obj) == 0)
2587 		vm_object_page_clean(obj, 0, 0, 0);
2588 	if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 &&
2589 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
2590 	    bioops.io_fsync)
2591 		error = (*bioops.io_fsync)(vp);
2592 	VOP_UNLOCK(vp, 0, td);
2593 	return (error);
2594 }
2595 
2596 int
2597 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
2598 {
2599 	struct namecache *fncpd;
2600 	struct namecache *tncpd;
2601 	struct namecache *ncp;
2602 	struct mount *mp;
2603 	int error;
2604 
2605 	bwillwrite();
2606 	if ((error = nlookup(fromnd)) != 0)
2607 		return (error);
2608 	if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL)
2609 		return (ENOENT);
2610 	cache_hold(fncpd);
2611 
2612 	/*
2613 	 * unlock the source ncp so we can lookup the target ncp without
2614 	 * deadlocking.  The target may or may not exist so we do not check
2615 	 * for a target vp like kern_mkdir() and other creation functions do.
2616 	 *
2617 	 * The source and target directories are ref'd and rechecked after
2618 	 * everything is relocked to determine if the source or target file
2619 	 * has been renamed.
2620 	 */
2621 	KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2622 	fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2623 	cache_unlock(fromnd->nl_ncp);
2624 
2625 	tond->nl_flags |= NLC_CREATE;
2626 	if ((error = nlookup(tond)) != 0) {
2627 		cache_drop(fncpd);
2628 		return (error);
2629 	}
2630 	if ((tncpd = tond->nl_ncp->nc_parent) == NULL) {
2631 		cache_drop(fncpd);
2632 		return (ENOENT);
2633 	}
2634 	cache_hold(tncpd);
2635 
2636 	/*
2637 	 * If the source and target are the same there is nothing to do
2638 	 */
2639 	if (fromnd->nl_ncp == tond->nl_ncp) {
2640 		cache_drop(fncpd);
2641 		cache_drop(tncpd);
2642 		return (0);
2643 	}
2644 
2645 	/*
2646 	 * relock the source ncp
2647 	 */
2648 	if (cache_lock_nonblock(fromnd->nl_ncp) == 0) {
2649 		cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2650 	} else if (fromnd->nl_ncp > tond->nl_ncp) {
2651 		cache_lock(fromnd->nl_ncp);
2652 		cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2653 	} else {
2654 		cache_unlock(tond->nl_ncp);
2655 		cache_lock(fromnd->nl_ncp);
2656 		cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2657 		cache_lock(tond->nl_ncp);
2658 		cache_resolve(tond->nl_ncp, tond->nl_cred);
2659 	}
2660 	fromnd->nl_flags |= NLC_NCPISLOCKED;
2661 
2662 	/*
2663 	 * make sure the parent directories linkages are the same
2664 	 */
2665 	if (fncpd != fromnd->nl_ncp->nc_parent ||
2666 	    tncpd != tond->nl_ncp->nc_parent) {
2667 		cache_drop(fncpd);
2668 		cache_drop(tncpd);
2669 		return (ENOENT);
2670 	}
2671 
2672 	/*
2673 	 * Both the source and target must be within the same filesystem and
2674 	 * in the same filesystem as their parent directories within the
2675 	 * namecache topology.
2676 	 */
2677 	mp = fncpd->nc_mount;
2678 	if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount ||
2679 	    mp != tond->nl_ncp->nc_mount) {
2680 		cache_drop(fncpd);
2681 		cache_drop(tncpd);
2682 		return (EXDEV);
2683 	}
2684 
2685 	/*
2686 	 * If the target exists and either the source or target is a directory,
2687 	 * then both must be directories.
2688 	 */
2689 	if (tond->nl_ncp->nc_vp) {
2690 		if (fromnd->nl_ncp->nc_vp->v_type == VDIR) {
2691 			if (tond->nl_ncp->nc_vp->v_type != VDIR)
2692 				error = ENOTDIR;
2693 		} else if (tond->nl_ncp->nc_vp->v_type == VDIR) {
2694 			error = EISDIR;
2695 		}
2696 	}
2697 
2698 	/*
2699 	 * You cannot rename a source into itself or a subdirectory of itself.
2700 	 * We check this by travsersing the target directory upwards looking
2701 	 * for a match against the source.
2702 	 */
2703 	if (error == 0) {
2704 		for (ncp = tncpd; ncp; ncp = ncp->nc_parent) {
2705 			if (fromnd->nl_ncp == ncp) {
2706 				error = EINVAL;
2707 				break;
2708 			}
2709 		}
2710 	}
2711 
2712 	cache_drop(fncpd);
2713 	cache_drop(tncpd);
2714 	if (error)
2715 		return (error);
2716 	error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp, tond->nl_cred);
2717 	return (error);
2718 }
2719 
2720 /*
2721  * rename_args(char *from, char *to)
2722  *
2723  * Rename files.  Source and destination must either both be directories,
2724  * or both not be directories.  If target is a directory, it must be empty.
2725  */
2726 int
2727 rename(struct rename_args *uap)
2728 {
2729 	struct nlookupdata fromnd, tond;
2730 	int error;
2731 
2732 	error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
2733 	if (error == 0) {
2734 		error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
2735 		if (error == 0)
2736 			error = kern_rename(&fromnd, &tond);
2737 		nlookup_done(&tond);
2738 	}
2739 	nlookup_done(&fromnd);
2740 	return (error);
2741 }
2742 
2743 int
2744 kern_mkdir(struct nlookupdata *nd, int mode)
2745 {
2746 	struct thread *td = curthread;
2747 	struct proc *p = td->td_proc;
2748 	struct namecache *ncp;
2749 	struct vnode *vp;
2750 	struct vattr vattr;
2751 	int error;
2752 
2753 	bwillwrite();
2754 	nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
2755 	if ((error = nlookup(nd)) != 0)
2756 		return (error);
2757 
2758 	ncp = nd->nl_ncp;
2759 	if (ncp->nc_vp)
2760 		return (EEXIST);
2761 
2762 	VATTR_NULL(&vattr);
2763 	vattr.va_type = VDIR;
2764 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2765 
2766 	vp = NULL;
2767 	error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr);
2768 	if (error == 0)
2769 		vput(vp);
2770 	return (error);
2771 }
2772 
2773 /*
2774  * mkdir_args(char *path, int mode)
2775  *
2776  * Make a directory file.
2777  */
2778 /* ARGSUSED */
2779 int
2780 mkdir(struct mkdir_args *uap)
2781 {
2782 	struct nlookupdata nd;
2783 	int error;
2784 
2785 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2786 	if (error == 0)
2787 		error = kern_mkdir(&nd, uap->mode);
2788 	nlookup_done(&nd);
2789 	return (error);
2790 }
2791 
2792 int
2793 kern_rmdir(struct nlookupdata *nd)
2794 {
2795 	struct namecache *ncp;
2796 	int error;
2797 
2798 	bwillwrite();
2799 	nd->nl_flags |= NLC_DELETE;
2800 	if ((error = nlookup(nd)) != 0)
2801 		return (error);
2802 
2803 	ncp = nd->nl_ncp;
2804 	error = VOP_NRMDIR(ncp, nd->nl_cred);
2805 	return (error);
2806 }
2807 
2808 /*
2809  * rmdir_args(char *path)
2810  *
2811  * Remove a directory file.
2812  */
2813 /* ARGSUSED */
2814 int
2815 rmdir(struct rmdir_args *uap)
2816 {
2817 	struct nlookupdata nd;
2818 	int error;
2819 
2820 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2821 	if (error == 0)
2822 		error = kern_rmdir(&nd);
2823 	nlookup_done(&nd);
2824 	return (error);
2825 }
2826 
2827 int
2828 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res)
2829 {
2830 	struct thread *td = curthread;
2831 	struct proc *p = td->td_proc;
2832 	struct vnode *vp;
2833 	struct file *fp;
2834 	struct uio auio;
2835 	struct iovec aiov;
2836 	long loff;
2837 	int error, eofflag;
2838 
2839 	if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
2840 		return (error);
2841 	if ((fp->f_flag & FREAD) == 0)
2842 		return (EBADF);
2843 	vp = (struct vnode *)fp->f_data;
2844 unionread:
2845 	if (vp->v_type != VDIR)
2846 		return (EINVAL);
2847 	aiov.iov_base = buf;
2848 	aiov.iov_len = count;
2849 	auio.uio_iov = &aiov;
2850 	auio.uio_iovcnt = 1;
2851 	auio.uio_rw = UIO_READ;
2852 	auio.uio_segflg = UIO_USERSPACE;
2853 	auio.uio_td = td;
2854 	auio.uio_resid = count;
2855 	/* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
2856 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2857 	loff = auio.uio_offset = fp->f_offset;
2858 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2859 	fp->f_offset = auio.uio_offset;
2860 	VOP_UNLOCK(vp, 0, td);
2861 	if (error)
2862 		return (error);
2863 	if (count == auio.uio_resid) {
2864 		if (union_dircheckp) {
2865 			error = union_dircheckp(td, &vp, fp);
2866 			if (error == -1)
2867 				goto unionread;
2868 			if (error)
2869 				return (error);
2870 		}
2871 		if ((vp->v_flag & VROOT) &&
2872 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
2873 			struct vnode *tvp = vp;
2874 			vp = vp->v_mount->mnt_vnodecovered;
2875 			vref(vp);
2876 			fp->f_data = (caddr_t)vp;
2877 			fp->f_offset = 0;
2878 			vrele(tvp);
2879 			goto unionread;
2880 		}
2881 	}
2882 	if (basep) {
2883 		*basep = loff;
2884 	}
2885 	*res = count - auio.uio_resid;
2886 	return (error);
2887 }
2888 
2889 /*
2890  * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2891  *
2892  * Read a block of directory entries in a file system independent format.
2893  */
2894 int
2895 getdirentries(struct getdirentries_args *uap)
2896 {
2897 	long base;
2898 	int error;
2899 
2900 	error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
2901 	    &uap->sysmsg_result);
2902 
2903 	if (error == 0)
2904 		error = copyout(&base, uap->basep, sizeof(*uap->basep));
2905 	return (error);
2906 }
2907 
2908 /*
2909  * getdents_args(int fd, char *buf, size_t count)
2910  */
2911 int
2912 getdents(struct getdents_args *uap)
2913 {
2914 	int error;
2915 
2916 	error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
2917 	    &uap->sysmsg_result);
2918 
2919 	return (error);
2920 }
2921 
2922 /*
2923  * umask(int newmask)
2924  *
2925  * Set the mode mask for creation of filesystem nodes.
2926  *
2927  * MP SAFE
2928  */
2929 int
2930 umask(struct umask_args *uap)
2931 {
2932 	struct thread *td = curthread;
2933 	struct proc *p = td->td_proc;
2934 	struct filedesc *fdp;
2935 
2936 	fdp = p->p_fd;
2937 	uap->sysmsg_result = fdp->fd_cmask;
2938 	fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
2939 	return (0);
2940 }
2941 
2942 /*
2943  * revoke(char *path)
2944  *
2945  * Void all references to file by ripping underlying filesystem
2946  * away from vnode.
2947  */
2948 /* ARGSUSED */
2949 int
2950 revoke(struct revoke_args *uap)
2951 {
2952 	struct thread *td = curthread;
2953 	struct nlookupdata nd;
2954 	struct vattr vattr;
2955 	struct vnode *vp;
2956 	struct ucred *cred;
2957 	int error;
2958 
2959 	vp = NULL;
2960 	error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
2961 	if (error == 0)
2962 		error = nlookup(&nd);
2963 	if (error == 0)
2964 		error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2965 	cred = crhold(nd.nl_cred);
2966 	nlookup_done(&nd);
2967 	if (error == 0) {
2968 		if (vp->v_type != VCHR && vp->v_type != VBLK)
2969 			error = EINVAL;
2970 		if (error == 0)
2971 			error = VOP_GETATTR(vp, &vattr, td);
2972 		if (error == 0 && cred->cr_uid != vattr.va_uid)
2973 			error = suser_cred(cred, PRISON_ROOT);
2974 		if (error == 0 && count_udev(vp->v_udev) > 0) {
2975 			if ((error = vx_lock(vp)) == 0) {
2976 				VOP_REVOKE(vp, REVOKEALL);
2977 				vx_unlock(vp);
2978 			}
2979 		}
2980 		vrele(vp);
2981 	}
2982 	crfree(cred);
2983 	return (error);
2984 }
2985 
2986 /*
2987  * Convert a user file descriptor to a kernel file entry.
2988  */
2989 int
2990 getvnode(struct filedesc *fdp, int fd, struct file **fpp)
2991 {
2992 	struct file *fp;
2993 
2994 	if ((u_int)fd >= fdp->fd_nfiles ||
2995 	    (fp = fdp->fd_ofiles[fd]) == NULL)
2996 		return (EBADF);
2997 	if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO)
2998 		return (EINVAL);
2999 	*fpp = fp;
3000 	return (0);
3001 }
3002 /*
3003  * getfh_args(char *fname, fhandle_t *fhp)
3004  *
3005  * Get (NFS) file handle
3006  */
3007 int
3008 getfh(struct getfh_args *uap)
3009 {
3010 	struct thread *td = curthread;
3011 	struct nlookupdata nd;
3012 	fhandle_t fh;
3013 	struct vnode *vp;
3014 	int error;
3015 
3016 	/*
3017 	 * Must be super user
3018 	 */
3019 	if ((error = suser(td)) != 0)
3020 		return (error);
3021 
3022 	vp = NULL;
3023 	error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3024 	if (error == 0)
3025 		error = nlookup(&nd);
3026 	if (error == 0)
3027 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3028 	nlookup_done(&nd);
3029 	if (error == 0) {
3030 		bzero(&fh, sizeof(fh));
3031 		fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3032 		error = VFS_VPTOFH(vp, &fh.fh_fid);
3033 		vput(vp);
3034 		if (error == 0)
3035 			error = copyout(&fh, uap->fhp, sizeof(fh));
3036 	}
3037 	return (error);
3038 }
3039 
3040 /*
3041  * fhopen_args(const struct fhandle *u_fhp, int flags)
3042  *
3043  * syscall for the rpc.lockd to use to translate a NFS file handle into
3044  * an open descriptor.
3045  *
3046  * warning: do not remove the suser() call or this becomes one giant
3047  * security hole.
3048  */
3049 int
3050 fhopen(struct fhopen_args *uap)
3051 {
3052 	struct thread *td = curthread;
3053 	struct proc *p = td->td_proc;
3054 	struct mount *mp;
3055 	struct vnode *vp;
3056 	struct fhandle fhp;
3057 	struct vattr vat;
3058 	struct vattr *vap = &vat;
3059 	struct flock lf;
3060 	struct filedesc *fdp = p->p_fd;
3061 	int fmode, mode, error, type;
3062 	struct file *nfp;
3063 	struct file *fp;
3064 	int indx;
3065 
3066 	/*
3067 	 * Must be super user
3068 	 */
3069 	error = suser(td);
3070 	if (error)
3071 		return (error);
3072 
3073 	fmode = FFLAGS(SCARG(uap, flags));
3074 	/* why not allow a non-read/write open for our lockd? */
3075 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3076 		return (EINVAL);
3077 	error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp));
3078 	if (error)
3079 		return(error);
3080 	/* find the mount point */
3081 	mp = vfs_getvfs(&fhp.fh_fsid);
3082 	if (mp == NULL)
3083 		return (ESTALE);
3084 	/* now give me my vnode, it gets returned to me locked */
3085 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3086 	if (error)
3087 		return (error);
3088  	/*
3089 	 * from now on we have to make sure not
3090 	 * to forget about the vnode
3091 	 * any error that causes an abort must vput(vp)
3092 	 * just set error = err and 'goto bad;'.
3093 	 */
3094 
3095 	/*
3096 	 * from vn_open
3097 	 */
3098 	if (vp->v_type == VLNK) {
3099 		error = EMLINK;
3100 		goto bad;
3101 	}
3102 	if (vp->v_type == VSOCK) {
3103 		error = EOPNOTSUPP;
3104 		goto bad;
3105 	}
3106 	mode = 0;
3107 	if (fmode & (FWRITE | O_TRUNC)) {
3108 		if (vp->v_type == VDIR) {
3109 			error = EISDIR;
3110 			goto bad;
3111 		}
3112 		error = vn_writechk(vp);
3113 		if (error)
3114 			goto bad;
3115 		mode |= VWRITE;
3116 	}
3117 	if (fmode & FREAD)
3118 		mode |= VREAD;
3119 	if (mode) {
3120 		error = VOP_ACCESS(vp, mode, p->p_ucred, td);
3121 		if (error)
3122 			goto bad;
3123 	}
3124 	if (fmode & O_TRUNC) {
3125 		VOP_UNLOCK(vp, 0, td);			/* XXX */
3126 		VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
3127 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);	/* XXX */
3128 		VATTR_NULL(vap);
3129 		vap->va_size = 0;
3130 		error = VOP_SETATTR(vp, vap, p->p_ucred, td);
3131 		if (error)
3132 			goto bad;
3133 	}
3134 
3135 	/*
3136 	 * VOP_OPEN needs the file pointer so it can potentially override
3137 	 * it.
3138 	 *
3139 	 * WARNING! no f_ncp will be associated when fhopen()ing a directory.
3140 	 * XXX
3141 	 */
3142 	if ((error = falloc(p, &nfp, NULL)) != 0)
3143 		goto bad;
3144 	fp = nfp;
3145 
3146 	fp->f_data = (caddr_t)vp;
3147 	fp->f_flag = fmode & FMASK;
3148 	fp->f_ops = &vnode_fileops;
3149 	fp->f_type = DTYPE_VNODE;
3150 
3151 	error = VOP_OPEN(vp, fmode, p->p_ucred, fp, td);
3152 	if (error) {
3153 		/*
3154 		 * setting f_ops this way prevents VOP_CLOSE from being
3155 		 * called or fdrop() releasing the vp from v_data.   Since
3156 		 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3157 		 */
3158 		fp->f_ops = &badfileops;
3159 		fp->f_data = NULL;
3160 		fdrop(fp, td);
3161 		goto bad;
3162 	}
3163 	if (fmode & FWRITE)
3164 		vp->v_writecount++;
3165 
3166 	/*
3167 	 * The fp now owns a reference on the vnode.  We still have our own
3168 	 * ref+lock.
3169 	 */
3170 	vref(vp);
3171 
3172 	/*
3173 	 * Make sure that a VM object is created for VMIO support.  If this
3174 	 * fails just fdrop() normally to clean up.
3175 	 */
3176 	if (vn_canvmio(vp) == TRUE) {
3177 		if ((error = vfs_object_create(vp, td)) != 0) {
3178 			fdrop(fp, td);
3179 			goto bad;
3180 		}
3181 	}
3182 
3183 	/*
3184 	 * The open was successful, associate it with a file descriptor.
3185 	 */
3186 	if ((error = fsetfd(p, fp, &indx)) != 0) {
3187 		if (fmode & FWRITE)
3188 			vp->v_writecount--;
3189 		fdrop(fp, td);
3190 		goto bad;
3191 	}
3192 
3193 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
3194 		lf.l_whence = SEEK_SET;
3195 		lf.l_start = 0;
3196 		lf.l_len = 0;
3197 		if (fmode & O_EXLOCK)
3198 			lf.l_type = F_WRLCK;
3199 		else
3200 			lf.l_type = F_RDLCK;
3201 		type = F_FLOCK;
3202 		if ((fmode & FNONBLOCK) == 0)
3203 			type |= F_WAIT;
3204 		VOP_UNLOCK(vp, 0, td);
3205 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3206 			/*
3207 			 * lock request failed.  Normally close the descriptor
3208 			 * but handle the case where someone might have dup()d
3209 			 * or close()d it when we weren't looking.
3210 			 */
3211 			if (fdp->fd_ofiles[indx] == fp) {
3212 				fdp->fd_ofiles[indx] = NULL;
3213 				fdrop(fp, td);
3214 			}
3215 
3216 			/*
3217 			 * release our private reference.
3218 			 */
3219 			fdrop(fp, td);
3220 			vrele(vp);
3221 			return (error);
3222 		}
3223 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
3224 		fp->f_flag |= FHASLOCK;
3225 	}
3226 	if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
3227 		vfs_object_create(vp, td);
3228 
3229 	vput(vp);
3230 	fdrop(fp, td);
3231 	uap->sysmsg_result = indx;
3232 	return (0);
3233 
3234 bad:
3235 	vput(vp);
3236 	return (error);
3237 }
3238 
3239 /*
3240  * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3241  */
3242 int
3243 fhstat(struct fhstat_args *uap)
3244 {
3245 	struct thread *td = curthread;
3246 	struct stat sb;
3247 	fhandle_t fh;
3248 	struct mount *mp;
3249 	struct vnode *vp;
3250 	int error;
3251 
3252 	/*
3253 	 * Must be super user
3254 	 */
3255 	error = suser(td);
3256 	if (error)
3257 		return (error);
3258 
3259 	error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t));
3260 	if (error)
3261 		return (error);
3262 
3263 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3264 		return (ESTALE);
3265 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3266 		return (error);
3267 	error = vn_stat(vp, &sb, td);
3268 	vput(vp);
3269 	if (error)
3270 		return (error);
3271 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
3272 	return (error);
3273 }
3274 
3275 /*
3276  * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3277  */
3278 int
3279 fhstatfs(struct fhstatfs_args *uap)
3280 {
3281 	struct thread *td = curthread;
3282 	struct statfs *sp;
3283 	struct mount *mp;
3284 	struct vnode *vp;
3285 	struct statfs sb;
3286 	fhandle_t fh;
3287 	int error;
3288 
3289 	/*
3290 	 * Must be super user
3291 	 */
3292 	if ((error = suser(td)))
3293 		return (error);
3294 
3295 	if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0)
3296 		return (error);
3297 
3298 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3299 		return (ESTALE);
3300 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3301 		return (error);
3302 	mp = vp->v_mount;
3303 	sp = &mp->mnt_stat;
3304 	vput(vp);
3305 	if ((error = VFS_STATFS(mp, sp, td)) != 0)
3306 		return (error);
3307 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3308 	if (suser(td)) {
3309 		bcopy(sp, &sb, sizeof(sb));
3310 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3311 		sp = &sb;
3312 	}
3313 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
3314 }
3315 
3316 /*
3317  * Syscall to push extended attribute configuration information into the
3318  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
3319  * a command (int cmd), and attribute name and misc data.  For now, the
3320  * attribute name is left in userspace for consumption by the VFS_op.
3321  * It will probably be changed to be copied into sysspace by the
3322  * syscall in the future, once issues with various consumers of the
3323  * attribute code have raised their hands.
3324  *
3325  * Currently this is used only by UFS Extended Attributes.
3326  */
3327 int
3328 extattrctl(struct extattrctl_args *uap)
3329 {
3330 	struct nlookupdata nd;
3331 	struct mount *mp;
3332 	struct vnode *vp;
3333 	int error;
3334 
3335 	vp = NULL;
3336 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3337 	if (error == 0)
3338 		error = nlookup(&nd);
3339 	if (error == 0) {
3340 		mp = nd.nl_ncp->nc_mount;
3341 		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd),
3342 				SCARG(uap, attrname), SCARG(uap, arg),
3343 				nd.nl_td);
3344 	}
3345 	nlookup_done(&nd);
3346 	return (error);
3347 }
3348 
3349 /*
3350  * Syscall to set a named extended attribute on a file or directory.
3351  * Accepts attribute name, and a uio structure pointing to the data to set.
3352  * The uio is consumed in the style of writev().  The real work happens
3353  * in VOP_SETEXTATTR().
3354  */
3355 int
3356 extattr_set_file(struct extattr_set_file_args *uap)
3357 {
3358 	char attrname[EXTATTR_MAXNAMELEN];
3359 	struct iovec aiov[UIO_SMALLIOV];
3360 	struct iovec *needfree;
3361 	struct nlookupdata nd;
3362 	struct iovec *iov;
3363 	struct vnode *vp;
3364 	struct uio auio;
3365 	u_int iovlen;
3366 	u_int cnt;
3367 	int error;
3368 	int i;
3369 
3370 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3371 	if (error)
3372 		return (error);
3373 
3374 	vp = NULL;
3375 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3376 	if (error == 0)
3377 		error = nlookup(&nd);
3378 	if (error == 0)
3379 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3380 	if (error) {
3381 		nlookup_done(&nd);
3382 		return (error);
3383 	}
3384 
3385 	needfree = NULL;
3386 	iovlen = uap->iovcnt * sizeof(struct iovec);
3387 	if (uap->iovcnt > UIO_SMALLIOV) {
3388 		if (uap->iovcnt > UIO_MAXIOV) {
3389 			error = EINVAL;
3390 			goto done;
3391 		}
3392 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3393 		needfree = iov;
3394 	} else {
3395 		iov = aiov;
3396 	}
3397 	auio.uio_iov = iov;
3398 	auio.uio_iovcnt = uap->iovcnt;
3399 	auio.uio_rw = UIO_WRITE;
3400 	auio.uio_segflg = UIO_USERSPACE;
3401 	auio.uio_td = nd.nl_td;
3402 	auio.uio_offset = 0;
3403 	if ((error = copyin(uap->iovp, iov, iovlen)))
3404 		goto done;
3405 	auio.uio_resid = 0;
3406 	for (i = 0; i < uap->iovcnt; i++) {
3407 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
3408 			error = EINVAL;
3409 			goto done;
3410 		}
3411 		auio.uio_resid += iov->iov_len;
3412 		iov++;
3413 	}
3414 	cnt = auio.uio_resid;
3415 	error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td);
3416 	cnt -= auio.uio_resid;
3417 	uap->sysmsg_result = cnt;
3418 done:
3419 	vput(vp);
3420 	nlookup_done(&nd);
3421 	if (needfree)
3422 		FREE(needfree, M_IOV);
3423 	return (error);
3424 }
3425 
3426 /*
3427  * Syscall to get a named extended attribute on a file or directory.
3428  * Accepts attribute name, and a uio structure pointing to a buffer for the
3429  * data.  The uio is consumed in the style of readv().  The real work
3430  * happens in VOP_GETEXTATTR();
3431  */
3432 int
3433 extattr_get_file(struct extattr_get_file_args *uap)
3434 {
3435 	char attrname[EXTATTR_MAXNAMELEN];
3436 	struct iovec aiov[UIO_SMALLIOV];
3437 	struct iovec *needfree;
3438 	struct nlookupdata nd;
3439 	struct iovec *iov;
3440 	struct vnode *vp;
3441 	struct uio auio;
3442 	u_int iovlen;
3443 	u_int cnt;
3444 	int error;
3445 	int i;
3446 
3447 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3448 	if (error)
3449 		return (error);
3450 
3451 	vp = NULL;
3452 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3453 	if (error == 0)
3454 		error = nlookup(&nd);
3455 	if (error == 0)
3456 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3457 	if (error) {
3458 		nlookup_done(&nd);
3459 		return (error);
3460 	}
3461 
3462 	iovlen = uap->iovcnt * sizeof (struct iovec);
3463 	needfree = NULL;
3464 	if (uap->iovcnt > UIO_SMALLIOV) {
3465 		if (uap->iovcnt > UIO_MAXIOV) {
3466 			error = EINVAL;
3467 			goto done;
3468 		}
3469 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3470 		needfree = iov;
3471 	} else {
3472 		iov = aiov;
3473 	}
3474 	auio.uio_iov = iov;
3475 	auio.uio_iovcnt = uap->iovcnt;
3476 	auio.uio_rw = UIO_READ;
3477 	auio.uio_segflg = UIO_USERSPACE;
3478 	auio.uio_td = nd.nl_td;
3479 	auio.uio_offset = 0;
3480 	if ((error = copyin(uap->iovp, iov, iovlen)))
3481 		goto done;
3482 	auio.uio_resid = 0;
3483 	for (i = 0; i < uap->iovcnt; i++) {
3484 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
3485 			error = EINVAL;
3486 			goto done;
3487 		}
3488 		auio.uio_resid += iov->iov_len;
3489 		iov++;
3490 	}
3491 	cnt = auio.uio_resid;
3492 	error = VOP_GETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td);
3493 	cnt -= auio.uio_resid;
3494 	uap->sysmsg_result = cnt;
3495 done:
3496 	vput(vp);
3497 	nlookup_done(&nd);
3498 	if (needfree)
3499 		FREE(needfree, M_IOV);
3500 	return(error);
3501 }
3502 
3503 /*
3504  * Syscall to delete a named extended attribute from a file or directory.
3505  * Accepts attribute name.  The real work happens in VOP_SETEXTATTR().
3506  */
3507 int
3508 extattr_delete_file(struct extattr_delete_file_args *uap)
3509 {
3510 	char attrname[EXTATTR_MAXNAMELEN];
3511 	struct nlookupdata nd;
3512 	struct vnode *vp;
3513 	int error;
3514 
3515 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3516 	if (error)
3517 		return(error);
3518 
3519 	vp = NULL;
3520 	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3521 	if (error == 0)
3522 		error = nlookup(&nd);
3523 	if (error == 0)
3524 		error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3525 	if (error) {
3526 		nlookup_done(&nd);
3527 		return (error);
3528 	}
3529 
3530 	error = VOP_SETEXTATTR(vp, attrname, NULL, nd.nl_cred, nd.nl_td);
3531 	vput(vp);
3532 	nlookup_done(&nd);
3533 	return(error);
3534 }
3535 
3536 /*
3537  * print out statistics from the current status of the buffer pool
3538  * this can be toggeled by the system control option debug.syncprt
3539  */
3540 #ifdef DEBUG
3541 void
3542 vfs_bufstats(void)
3543 {
3544         int s, i, j, count;
3545         struct buf *bp;
3546         struct bqueues *dp;
3547         int counts[(MAXBSIZE / PAGE_SIZE) + 1];
3548         static char *bname[3] = { "LOCKED", "LRU", "AGE" };
3549 
3550         for (dp = bufqueues, i = 0; dp < &bufqueues[3]; dp++, i++) {
3551                 count = 0;
3552                 for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++)
3553                         counts[j] = 0;
3554                 s = splbio();
3555                 TAILQ_FOREACH(bp, dp, b_freelist) {
3556                         counts[bp->b_bufsize/PAGE_SIZE]++;
3557                         count++;
3558                 }
3559                 splx(s);
3560                 printf("%s: total-%d", bname[i], count);
3561                 for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++)
3562                         if (counts[j] != 0)
3563                                 printf(", %d-%d", j * PAGE_SIZE, counts[j]);
3564                 printf("\n");
3565         }
3566 }
3567 #endif
3568