xref: /freebsd/sys/kern/vfs_extattr.c (revision 77a0943ded95b9e6438f7db70c4a28e4d93946d4)
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  * $FreeBSD$
40  */
41 
42 /* For 4.3 integer FS ID compatibility */
43 #include "opt_compat.h"
44 #include "opt_ffs.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/bio.h>
49 #include <sys/buf.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/linker.h>
61 #include <sys/stat.h>
62 #include <sys/unistd.h>
63 #include <sys/vnode.h>
64 #include <sys/proc.h>
65 #include <sys/dirent.h>
66 #include <sys/extattr.h>
67 
68 #include <machine/limits.h>
69 #include <miscfs/union/union.h>
70 #include <sys/sysctl.h>
71 #include <vm/vm.h>
72 #include <vm/vm_object.h>
73 #include <vm/vm_zone.h>
74 #include <vm/vm_page.h>
75 
76 static int change_dir __P((struct nameidata *ndp, struct proc *p));
77 static void checkdirs __P((struct vnode *olddp));
78 static int chroot_refuse_vdir_fds __P((struct filedesc *fdp));
79 static int getutimes __P((const struct timeval *, struct timespec *));
80 static int setfown __P((struct proc *, struct vnode *, uid_t, gid_t));
81 static int setfmode __P((struct proc *, struct vnode *, int));
82 static int setfflags __P((struct proc *, struct vnode *, int));
83 static int setutimes __P((struct proc *, struct vnode *,
84     const struct timespec *, int));
85 static int	usermount = 0;	/* if 1, non-root can mount fs. */
86 
87 int (*union_dircheckp) __P((struct proc *, struct vnode **, struct file *));
88 
89 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
90 
91 /*
92  * Virtual File System System Calls
93  */
94 
95 /*
96  * Mount a file system.
97  */
98 #ifndef _SYS_SYSPROTO_H_
99 struct mount_args {
100 	char	*type;
101 	char	*path;
102 	int	flags;
103 	caddr_t	data;
104 };
105 #endif
106 /* ARGSUSED */
107 int
108 mount(p, uap)
109 	struct proc *p;
110 	register struct mount_args /* {
111 		syscallarg(char *) type;
112 		syscallarg(char *) path;
113 		syscallarg(int) flags;
114 		syscallarg(caddr_t) data;
115 	} */ *uap;
116 {
117 	struct vnode *vp;
118 	struct mount *mp;
119 	struct vfsconf *vfsp;
120 	int error, flag = 0, flag2 = 0;
121 	struct vattr va;
122 	struct nameidata nd;
123 	char fstypename[MFSNAMELEN];
124 
125 	if (usermount == 0 && (error = suser(p)))
126 		return (error);
127 	/*
128 	 * Do not allow NFS export by non-root users.
129 	 */
130 	if (SCARG(uap, flags) & MNT_EXPORTED) {
131 		error = suser(p);
132 		if (error)
133 			return (error);
134 	}
135 	/*
136 	 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
137 	 */
138 	if (suser_xxx(p->p_ucred, 0, 0))
139 		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
140 	/*
141 	 * Get vnode to be covered
142 	 */
143 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
144 	    SCARG(uap, path), p);
145 	if ((error = namei(&nd)) != 0)
146 		return (error);
147 	NDFREE(&nd, NDF_ONLY_PNBUF);
148 	vp = nd.ni_vp;
149 	if (SCARG(uap, flags) & MNT_UPDATE) {
150 		if ((vp->v_flag & VROOT) == 0) {
151 			vput(vp);
152 			return (EINVAL);
153 		}
154 		mp = vp->v_mount;
155 		flag = mp->mnt_flag;
156 		flag2 = mp->mnt_kern_flag;
157 		/*
158 		 * We only allow the filesystem to be reloaded if it
159 		 * is currently mounted read-only.
160 		 */
161 		if ((SCARG(uap, flags) & MNT_RELOAD) &&
162 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
163 			vput(vp);
164 			return (EOPNOTSUPP);	/* Needs translation */
165 		}
166 		/*
167 		 * Only root, or the user that did the original mount is
168 		 * permitted to update it.
169 		 */
170 		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
171 		    (error = suser(p))) {
172 			vput(vp);
173 			return (error);
174 		}
175 		if (vfs_busy(mp, LK_NOWAIT, 0, p)) {
176 			vput(vp);
177 			return (EBUSY);
178 		}
179 		mtx_enter(&vp->v_interlock, MTX_DEF);
180 		if ((vp->v_flag & VMOUNT) != 0 ||
181 		    vp->v_mountedhere != NULL) {
182 			mtx_exit(&vp->v_interlock, MTX_DEF);
183 			vfs_unbusy(mp, p);
184 			vput(vp);
185 			return (EBUSY);
186 		}
187 		vp->v_flag |= VMOUNT;
188 		mtx_exit(&vp->v_interlock, MTX_DEF);
189 		mp->mnt_flag |= SCARG(uap, flags) &
190 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT);
191 		VOP_UNLOCK(vp, 0, p);
192 		goto update;
193 	}
194 	/*
195 	 * If the user is not root, ensure that they own the directory
196 	 * onto which we are attempting to mount.
197 	 */
198 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) ||
199 	    (va.va_uid != p->p_ucred->cr_uid &&
200 	     (error = suser(p)))) {
201 		vput(vp);
202 		return (error);
203 	}
204 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) {
205 		vput(vp);
206 		return (error);
207 	}
208 	if (vp->v_type != VDIR) {
209 		vput(vp);
210 		return (ENOTDIR);
211 	}
212 	if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) {
213 		vput(vp);
214 		return (error);
215 	}
216 	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
217 		if (!strcmp(vfsp->vfc_name, fstypename))
218 			break;
219 	if (vfsp == NULL) {
220 		linker_file_t lf;
221 
222 		/* Only load modules for root (very important!) */
223 		if ((error = suser(p)) != 0) {
224 			vput(vp);
225 			return error;
226 		}
227 		error = linker_load_file(fstypename, &lf);
228 		if (error || lf == NULL) {
229 			vput(vp);
230 			if (lf == NULL)
231 				error = ENODEV;
232 			return error;
233 		}
234 		lf->userrefs++;
235 		/* lookup again, see if the VFS was loaded */
236 		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
237 			if (!strcmp(vfsp->vfc_name, fstypename))
238 				break;
239 		if (vfsp == NULL) {
240 			lf->userrefs--;
241 			linker_file_unload(lf);
242 			vput(vp);
243 			return (ENODEV);
244 		}
245 	}
246 	mtx_enter(&vp->v_interlock, MTX_DEF);
247 	if ((vp->v_flag & VMOUNT) != 0 ||
248 	    vp->v_mountedhere != NULL) {
249 		mtx_exit(&vp->v_interlock, MTX_DEF);
250 		vput(vp);
251 		return (EBUSY);
252 	}
253 	vp->v_flag |= VMOUNT;
254 	mtx_exit(&vp->v_interlock, MTX_DEF);
255 
256 	/*
257 	 * Allocate and initialize the filesystem.
258 	 */
259 	mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK);
260 	bzero((char *)mp, (u_long)sizeof(struct mount));
261 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE);
262 	(void)vfs_busy(mp, LK_NOWAIT, 0, p);
263 	mp->mnt_op = vfsp->vfc_vfsops;
264 	mp->mnt_vfc = vfsp;
265 	vfsp->vfc_refcount++;
266 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
267 	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
268 	strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
269 	mp->mnt_vnodecovered = vp;
270 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
271 	mp->mnt_iosize_max = DFLTPHYS;
272 	VOP_UNLOCK(vp, 0, p);
273 update:
274 	/*
275 	 * Set the mount level flags.
276 	 */
277 	if (SCARG(uap, flags) & MNT_RDONLY)
278 		mp->mnt_flag |= MNT_RDONLY;
279 	else if (mp->mnt_flag & MNT_RDONLY)
280 		mp->mnt_kern_flag |= MNTK_WANTRDWR;
281 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
282 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
283 	    MNT_NOSYMFOLLOW | MNT_IGNORE |
284 	    MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
285 	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
286 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
287 	    MNT_NOSYMFOLLOW | MNT_IGNORE |
288 	    MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
289 	/*
290 	 * Mount the filesystem.
291 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
292 	 * get.  No freeing of cn_pnbuf.
293 	 */
294 	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
295 	if (mp->mnt_flag & MNT_UPDATE) {
296 		if (mp->mnt_kern_flag & MNTK_WANTRDWR)
297 			mp->mnt_flag &= ~MNT_RDONLY;
298 		mp->mnt_flag &=~
299 		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT);
300 		mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
301 		if (error) {
302 			mp->mnt_flag = flag;
303 			mp->mnt_kern_flag = flag2;
304 		}
305 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
306 			if (mp->mnt_syncer == NULL)
307 				error = vfs_allocate_syncvnode(mp);
308 		} else {
309 			if (mp->mnt_syncer != NULL)
310 				vrele(mp->mnt_syncer);
311 			mp->mnt_syncer = NULL;
312 		}
313 		vfs_unbusy(mp, p);
314 		mtx_enter(&vp->v_interlock, MTX_DEF);
315 		vp->v_flag &= ~VMOUNT;
316 		mtx_exit(&vp->v_interlock, MTX_DEF);
317 		vrele(vp);
318 		return (error);
319 	}
320 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
321 	/*
322 	 * Put the new filesystem on the mount list after root.
323 	 */
324 	cache_purge(vp);
325 	if (!error) {
326 		mtx_enter(&vp->v_interlock, MTX_DEF);
327 		vp->v_flag &= ~VMOUNT;
328 		vp->v_mountedhere = mp;
329 		mtx_exit(&vp->v_interlock, MTX_DEF);
330 		mtx_enter(&mountlist_mtx, MTX_DEF);
331 		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
332 		mtx_exit(&mountlist_mtx, MTX_DEF);
333 		checkdirs(vp);
334 		VOP_UNLOCK(vp, 0, p);
335 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
336 			error = vfs_allocate_syncvnode(mp);
337 		vfs_unbusy(mp, p);
338 		if ((error = VFS_START(mp, 0, p)) != 0)
339 			vrele(vp);
340 	} else {
341 		mtx_enter(&vp->v_interlock, MTX_DEF);
342 		vp->v_flag &= ~VMOUNT;
343 		mtx_exit(&vp->v_interlock, MTX_DEF);
344 		mp->mnt_vfc->vfc_refcount--;
345 		vfs_unbusy(mp, p);
346 		free((caddr_t)mp, M_MOUNT);
347 		vput(vp);
348 	}
349 	return (error);
350 }
351 
352 /*
353  * Scan all active processes to see if any of them have a current
354  * or root directory onto which the new filesystem has just been
355  * mounted. If so, replace them with the new mount point.
356  */
357 static void
358 checkdirs(olddp)
359 	struct vnode *olddp;
360 {
361 	struct filedesc *fdp;
362 	struct vnode *newdp;
363 	struct proc *p;
364 
365 	if (olddp->v_usecount == 1)
366 		return;
367 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
368 		panic("mount: lost mount");
369 	lockmgr(&allproc_lock, LK_SHARED, NULL, CURPROC);
370 	LIST_FOREACH(p, &allproc, p_list) {
371 		fdp = p->p_fd;
372 		if (fdp->fd_cdir == olddp) {
373 			vrele(fdp->fd_cdir);
374 			VREF(newdp);
375 			fdp->fd_cdir = newdp;
376 		}
377 		if (fdp->fd_rdir == olddp) {
378 			vrele(fdp->fd_rdir);
379 			VREF(newdp);
380 			fdp->fd_rdir = newdp;
381 		}
382 	}
383 	lockmgr(&allproc_lock, LK_RELEASE, NULL, CURPROC);
384 	if (rootvnode == olddp) {
385 		vrele(rootvnode);
386 		VREF(newdp);
387 		rootvnode = newdp;
388 	}
389 	vput(newdp);
390 }
391 
392 /*
393  * Unmount a file system.
394  *
395  * Note: unmount takes a path to the vnode mounted on as argument,
396  * not special file (as before).
397  */
398 #ifndef _SYS_SYSPROTO_H_
399 struct unmount_args {
400 	char	*path;
401 	int	flags;
402 };
403 #endif
404 /* ARGSUSED */
405 int
406 unmount(p, uap)
407 	struct proc *p;
408 	register struct unmount_args /* {
409 		syscallarg(char *) path;
410 		syscallarg(int) flags;
411 	} */ *uap;
412 {
413 	register struct vnode *vp;
414 	struct mount *mp;
415 	int error;
416 	struct nameidata nd;
417 
418 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
419 	    SCARG(uap, path), p);
420 	if ((error = namei(&nd)) != 0)
421 		return (error);
422 	vp = nd.ni_vp;
423 	NDFREE(&nd, NDF_ONLY_PNBUF);
424 	mp = vp->v_mount;
425 
426 	/*
427 	 * Only root, or the user that did the original mount is
428 	 * permitted to unmount this filesystem.
429 	 */
430 	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
431 	    (error = suser(p))) {
432 		vput(vp);
433 		return (error);
434 	}
435 
436 	/*
437 	 * Don't allow unmounting the root file system.
438 	 */
439 	if (mp->mnt_flag & MNT_ROOTFS) {
440 		vput(vp);
441 		return (EINVAL);
442 	}
443 
444 	/*
445 	 * Must be the root of the filesystem
446 	 */
447 	if ((vp->v_flag & VROOT) == 0) {
448 		vput(vp);
449 		return (EINVAL);
450 	}
451 	vput(vp);
452 	return (dounmount(mp, SCARG(uap, flags), p));
453 }
454 
455 /*
456  * Do the actual file system unmount.
457  */
458 int
459 dounmount(mp, flags, p)
460 	struct mount *mp;
461 	int flags;
462 	struct proc *p;
463 {
464 	struct vnode *coveredvp;
465 	int error;
466 	int async_flag;
467 
468 	mtx_enter(&mountlist_mtx, MTX_DEF);
469 	mp->mnt_kern_flag |= MNTK_UNMOUNT;
470 	lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_mtx, p);
471 	vn_start_write(NULL, &mp, V_WAIT);
472 
473 	if (mp->mnt_flag & MNT_EXPUBLIC)
474 		vfs_setpublicfs(NULL, NULL, NULL);
475 
476 	vfs_msync(mp, MNT_WAIT);
477 	async_flag = mp->mnt_flag & MNT_ASYNC;
478 	mp->mnt_flag &=~ MNT_ASYNC;
479 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
480 	if (mp->mnt_syncer != NULL)
481 		vrele(mp->mnt_syncer);
482 	if (((mp->mnt_flag & MNT_RDONLY) ||
483 	     (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
484 	    (flags & MNT_FORCE)) {
485 		error = VFS_UNMOUNT(mp, flags, p);
486 	}
487 	vn_finished_write(mp);
488 	mtx_enter(&mountlist_mtx, MTX_DEF);
489 	if (error) {
490 		if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
491 			(void) vfs_allocate_syncvnode(mp);
492 		mp->mnt_kern_flag &= ~MNTK_UNMOUNT;
493 		mp->mnt_flag |= async_flag;
494 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
495 		    &mountlist_mtx, p);
496 		if (mp->mnt_kern_flag & MNTK_MWAIT)
497 			wakeup((caddr_t)mp);
498 		return (error);
499 	}
500 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
501 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
502 		coveredvp->v_mountedhere = (struct mount *)0;
503 		vrele(coveredvp);
504 	}
505 	mp->mnt_vfc->vfc_refcount--;
506 	if (!LIST_EMPTY(&mp->mnt_vnodelist))
507 		panic("unmount: dangling vnode");
508 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_mtx, p);
509 	lockdestroy(&mp->mnt_lock);
510 	if (mp->mnt_kern_flag & MNTK_MWAIT)
511 		wakeup((caddr_t)mp);
512 	free((caddr_t)mp, M_MOUNT);
513 	return (0);
514 }
515 
516 /*
517  * Sync each mounted filesystem.
518  */
519 #ifndef _SYS_SYSPROTO_H_
520 struct sync_args {
521         int     dummy;
522 };
523 #endif
524 
525 #ifdef DEBUG
526 static int syncprt = 0;
527 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
528 #endif
529 
530 /* ARGSUSED */
531 int
532 sync(p, uap)
533 	struct proc *p;
534 	struct sync_args *uap;
535 {
536 	struct mount *mp, *nmp;
537 	int asyncflag;
538 
539 	mtx_enter(&mountlist_mtx, MTX_DEF);
540 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
541 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, p)) {
542 			nmp = TAILQ_NEXT(mp, mnt_list);
543 			continue;
544 		}
545 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
546 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
547 			asyncflag = mp->mnt_flag & MNT_ASYNC;
548 			mp->mnt_flag &= ~MNT_ASYNC;
549 			vfs_msync(mp, MNT_NOWAIT);
550 			VFS_SYNC(mp, MNT_NOWAIT,
551 			    ((p != NULL) ? p->p_ucred : NOCRED), p);
552 			mp->mnt_flag |= asyncflag;
553 			vn_finished_write(mp);
554 		}
555 		mtx_enter(&mountlist_mtx, MTX_DEF);
556 		nmp = TAILQ_NEXT(mp, mnt_list);
557 		vfs_unbusy(mp, p);
558 	}
559 	mtx_exit(&mountlist_mtx, MTX_DEF);
560 #if 0
561 /*
562  * XXX don't call vfs_bufstats() yet because that routine
563  * was not imported in the Lite2 merge.
564  */
565 #ifdef DIAGNOSTIC
566 	if (syncprt)
567 		vfs_bufstats();
568 #endif /* DIAGNOSTIC */
569 #endif
570 	return (0);
571 }
572 
573 /* XXX PRISON: could be per prison flag */
574 static int prison_quotas;
575 #if 0
576 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
577 #endif
578 
579 /*
580  * Change filesystem quotas.
581  */
582 #ifndef _SYS_SYSPROTO_H_
583 struct quotactl_args {
584 	char *path;
585 	int cmd;
586 	int uid;
587 	caddr_t arg;
588 };
589 #endif
590 /* ARGSUSED */
591 int
592 quotactl(p, uap)
593 	struct proc *p;
594 	register struct quotactl_args /* {
595 		syscallarg(char *) path;
596 		syscallarg(int) cmd;
597 		syscallarg(int) uid;
598 		syscallarg(caddr_t) arg;
599 	} */ *uap;
600 {
601 	struct mount *mp;
602 	int error;
603 	struct nameidata nd;
604 
605 	if (p->p_prison && !prison_quotas)
606 		return (EPERM);
607 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
608 	if ((error = namei(&nd)) != 0)
609 		return (error);
610 	NDFREE(&nd, NDF_ONLY_PNBUF);
611 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH);
612 	vrele(nd.ni_vp);
613 	if (error)
614 		return (error);
615 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
616 	    SCARG(uap, arg), p);
617 	vn_finished_write(mp);
618 	return (error);
619 }
620 
621 /*
622  * Get filesystem statistics.
623  */
624 #ifndef _SYS_SYSPROTO_H_
625 struct statfs_args {
626 	char *path;
627 	struct statfs *buf;
628 };
629 #endif
630 /* ARGSUSED */
631 int
632 statfs(p, uap)
633 	struct proc *p;
634 	register struct statfs_args /* {
635 		syscallarg(char *) path;
636 		syscallarg(struct statfs *) buf;
637 	} */ *uap;
638 {
639 	register struct mount *mp;
640 	register struct statfs *sp;
641 	int error;
642 	struct nameidata nd;
643 	struct statfs sb;
644 
645 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
646 	if ((error = namei(&nd)) != 0)
647 		return (error);
648 	mp = nd.ni_vp->v_mount;
649 	sp = &mp->mnt_stat;
650 	NDFREE(&nd, NDF_ONLY_PNBUF);
651 	vrele(nd.ni_vp);
652 	error = VFS_STATFS(mp, sp, p);
653 	if (error)
654 		return (error);
655 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
656 	if (suser_xxx(p->p_ucred, 0, 0)) {
657 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
658 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
659 		sp = &sb;
660 	}
661 	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
662 }
663 
664 /*
665  * Get filesystem statistics.
666  */
667 #ifndef _SYS_SYSPROTO_H_
668 struct fstatfs_args {
669 	int fd;
670 	struct statfs *buf;
671 };
672 #endif
673 /* ARGSUSED */
674 int
675 fstatfs(p, uap)
676 	struct proc *p;
677 	register struct fstatfs_args /* {
678 		syscallarg(int) fd;
679 		syscallarg(struct statfs *) buf;
680 	} */ *uap;
681 {
682 	struct file *fp;
683 	struct mount *mp;
684 	register struct statfs *sp;
685 	int error;
686 	struct statfs sb;
687 
688 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
689 		return (error);
690 	mp = ((struct vnode *)fp->f_data)->v_mount;
691 	sp = &mp->mnt_stat;
692 	error = VFS_STATFS(mp, sp, p);
693 	if (error)
694 		return (error);
695 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
696 	if (suser_xxx(p->p_ucred, 0, 0)) {
697 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
698 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
699 		sp = &sb;
700 	}
701 	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
702 }
703 
704 /*
705  * Get statistics on all filesystems.
706  */
707 #ifndef _SYS_SYSPROTO_H_
708 struct getfsstat_args {
709 	struct statfs *buf;
710 	long bufsize;
711 	int flags;
712 };
713 #endif
714 int
715 getfsstat(p, uap)
716 	struct proc *p;
717 	register struct getfsstat_args /* {
718 		syscallarg(struct statfs *) buf;
719 		syscallarg(long) bufsize;
720 		syscallarg(int) flags;
721 	} */ *uap;
722 {
723 	register struct mount *mp, *nmp;
724 	register struct statfs *sp;
725 	caddr_t sfsp;
726 	long count, maxcount, error;
727 
728 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
729 	sfsp = (caddr_t)SCARG(uap, buf);
730 	count = 0;
731 	mtx_enter(&mountlist_mtx, MTX_DEF);
732 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
733 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, p)) {
734 			nmp = TAILQ_NEXT(mp, mnt_list);
735 			continue;
736 		}
737 		if (sfsp && count < maxcount) {
738 			sp = &mp->mnt_stat;
739 			/*
740 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
741 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
742 			 * overrides MNT_WAIT.
743 			 */
744 			if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
745 			    (SCARG(uap, flags) & MNT_WAIT)) &&
746 			    (error = VFS_STATFS(mp, sp, p))) {
747 				mtx_enter(&mountlist_mtx, MTX_DEF);
748 				nmp = TAILQ_NEXT(mp, mnt_list);
749 				vfs_unbusy(mp, p);
750 				continue;
751 			}
752 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
753 			error = copyout((caddr_t)sp, sfsp, sizeof(*sp));
754 			if (error) {
755 				vfs_unbusy(mp, p);
756 				return (error);
757 			}
758 			sfsp += sizeof(*sp);
759 		}
760 		count++;
761 		mtx_enter(&mountlist_mtx, MTX_DEF);
762 		nmp = TAILQ_NEXT(mp, mnt_list);
763 		vfs_unbusy(mp, p);
764 	}
765 	mtx_exit(&mountlist_mtx, MTX_DEF);
766 	if (sfsp && count > maxcount)
767 		p->p_retval[0] = maxcount;
768 	else
769 		p->p_retval[0] = count;
770 	return (0);
771 }
772 
773 /*
774  * Change current working directory to a given file descriptor.
775  */
776 #ifndef _SYS_SYSPROTO_H_
777 struct fchdir_args {
778 	int	fd;
779 };
780 #endif
781 /* ARGSUSED */
782 int
783 fchdir(p, uap)
784 	struct proc *p;
785 	struct fchdir_args /* {
786 		syscallarg(int) fd;
787 	} */ *uap;
788 {
789 	register struct filedesc *fdp = p->p_fd;
790 	struct vnode *vp, *tdp;
791 	struct mount *mp;
792 	struct file *fp;
793 	int error;
794 
795 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
796 		return (error);
797 	vp = (struct vnode *)fp->f_data;
798 	VREF(vp);
799 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
800 	if (vp->v_type != VDIR)
801 		error = ENOTDIR;
802 	else
803 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
804 	while (!error && (mp = vp->v_mountedhere) != NULL) {
805 		if (vfs_busy(mp, 0, 0, p))
806 			continue;
807 		error = VFS_ROOT(mp, &tdp);
808 		vfs_unbusy(mp, p);
809 		if (error)
810 			break;
811 		vput(vp);
812 		vp = tdp;
813 	}
814 	if (error) {
815 		vput(vp);
816 		return (error);
817 	}
818 	VOP_UNLOCK(vp, 0, p);
819 	vrele(fdp->fd_cdir);
820 	fdp->fd_cdir = vp;
821 	return (0);
822 }
823 
824 /*
825  * Change current working directory (``.'').
826  */
827 #ifndef _SYS_SYSPROTO_H_
828 struct chdir_args {
829 	char	*path;
830 };
831 #endif
832 /* ARGSUSED */
833 int
834 chdir(p, uap)
835 	struct proc *p;
836 	struct chdir_args /* {
837 		syscallarg(char *) path;
838 	} */ *uap;
839 {
840 	register struct filedesc *fdp = p->p_fd;
841 	int error;
842 	struct nameidata nd;
843 
844 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
845 	    SCARG(uap, path), p);
846 	if ((error = change_dir(&nd, p)) != 0)
847 		return (error);
848 	NDFREE(&nd, NDF_ONLY_PNBUF);
849 	vrele(fdp->fd_cdir);
850 	fdp->fd_cdir = nd.ni_vp;
851 	return (0);
852 }
853 
854 /*
855  * Helper function for raised chroot(2) security function:  Refuse if
856  * any filedescriptors are open directories.
857  */
858 static int
859 chroot_refuse_vdir_fds(fdp)
860 	struct filedesc *fdp;
861 {
862 	struct vnode *vp;
863 	struct file *fp;
864 	int error;
865 	int fd;
866 
867 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
868 		error = getvnode(fdp, fd, &fp);
869 		if (error)
870 			continue;
871 		vp = (struct vnode *)fp->f_data;
872 		if (vp->v_type != VDIR)
873 			continue;
874 		return(EPERM);
875 	}
876 	return (0);
877 }
878 
879 /*
880  * This sysctl determines if we will allow a process to chroot(2) if it
881  * has a directory open:
882  *	0: disallowed for all processes.
883  *	1: allowed for processes that were not already chroot(2)'ed.
884  *	2: allowed for all processes.
885  */
886 
887 static int chroot_allow_open_directories = 1;
888 
889 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
890      &chroot_allow_open_directories, 0, "");
891 
892 /*
893  * Change notion of root (``/'') directory.
894  */
895 #ifndef _SYS_SYSPROTO_H_
896 struct chroot_args {
897 	char	*path;
898 };
899 #endif
900 /* ARGSUSED */
901 int
902 chroot(p, uap)
903 	struct proc *p;
904 	struct chroot_args /* {
905 		syscallarg(char *) path;
906 	} */ *uap;
907 {
908 	register struct filedesc *fdp = p->p_fd;
909 	int error;
910 	struct nameidata nd;
911 
912 	error = suser_xxx(0, p, PRISON_ROOT);
913 	if (error)
914 		return (error);
915 	if (chroot_allow_open_directories == 0 ||
916 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode))
917 		error = chroot_refuse_vdir_fds(fdp);
918 	if (error)
919 		return (error);
920 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
921 	    SCARG(uap, path), p);
922 	if ((error = change_dir(&nd, p)) != 0)
923 		return (error);
924 	NDFREE(&nd, NDF_ONLY_PNBUF);
925 	vrele(fdp->fd_rdir);
926 	fdp->fd_rdir = nd.ni_vp;
927 	if (!fdp->fd_jdir) {
928 		fdp->fd_jdir = nd.ni_vp;
929                 VREF(fdp->fd_jdir);
930 	}
931 	return (0);
932 }
933 
934 /*
935  * Common routine for chroot and chdir.
936  */
937 static int
938 change_dir(ndp, p)
939 	register struct nameidata *ndp;
940 	struct proc *p;
941 {
942 	struct vnode *vp;
943 	int error;
944 
945 	error = namei(ndp);
946 	if (error)
947 		return (error);
948 	vp = ndp->ni_vp;
949 	if (vp->v_type != VDIR)
950 		error = ENOTDIR;
951 	else
952 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
953 	if (error)
954 		vput(vp);
955 	else
956 		VOP_UNLOCK(vp, 0, p);
957 	return (error);
958 }
959 
960 /*
961  * Check permissions, allocate an open file structure,
962  * and call the device open routine if any.
963  */
964 #ifndef _SYS_SYSPROTO_H_
965 struct open_args {
966 	char	*path;
967 	int	flags;
968 	int	mode;
969 };
970 #endif
971 int
972 open(p, uap)
973 	struct proc *p;
974 	register struct open_args /* {
975 		syscallarg(char *) path;
976 		syscallarg(int) flags;
977 		syscallarg(int) mode;
978 	} */ *uap;
979 {
980 	struct filedesc *fdp = p->p_fd;
981 	struct file *fp;
982 	struct vnode *vp;
983 	struct vattr vat;
984 	struct mount *mp;
985 	int cmode, flags, oflags;
986 	struct file *nfp;
987 	int type, indx, error;
988 	struct flock lf;
989 	struct nameidata nd;
990 
991 	oflags = SCARG(uap, flags);
992 	if ((oflags & O_ACCMODE) == O_ACCMODE)
993 		return (EINVAL);
994 	flags = FFLAGS(oflags);
995 	error = falloc(p, &nfp, &indx);
996 	if (error)
997 		return (error);
998 	fp = nfp;
999 	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1000 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1001 	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
1002 	/*
1003 	 * Bump the ref count to prevent another process from closing
1004 	 * the descriptor while we are blocked in vn_open()
1005 	 */
1006 	fhold(fp);
1007 	error = vn_open(&nd, &flags, cmode);
1008 	if (error) {
1009 		/*
1010 		 * release our own reference
1011 		 */
1012 		fdrop(fp, p);
1013 
1014 		/*
1015 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1016 		 * responsible for dropping the old contents of ofiles[indx]
1017 		 * if it succeeds.
1018 		 */
1019 		if ((error == ENODEV || error == ENXIO) &&
1020 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
1021 		    (error =
1022 			dupfdopen(p, fdp, indx, p->p_dupfd, flags, error)) == 0) {
1023 			p->p_retval[0] = indx;
1024 			return (0);
1025 		}
1026 		/*
1027 		 * Clean up the descriptor, but only if another thread hadn't
1028 		 * replaced or closed it.
1029 		 */
1030 		if (fdp->fd_ofiles[indx] == fp) {
1031 			fdp->fd_ofiles[indx] = NULL;
1032 			fdrop(fp, p);
1033 		}
1034 
1035 		if (error == ERESTART)
1036 			error = EINTR;
1037 		return (error);
1038 	}
1039 	p->p_dupfd = 0;
1040 	NDFREE(&nd, NDF_ONLY_PNBUF);
1041 	vp = nd.ni_vp;
1042 
1043 	/*
1044 	 * There should be 2 references on the file, one from the descriptor
1045 	 * table, and one for us.
1046 	 *
1047 	 * Handle the case where someone closed the file (via its file
1048 	 * descriptor) while we were blocked.  The end result should look
1049 	 * like opening the file succeeded but it was immediately closed.
1050 	 */
1051 	if (fp->f_count == 1) {
1052 		KASSERT(fdp->fd_ofiles[indx] != fp,
1053 		    ("Open file descriptor lost all refs"));
1054 		VOP_UNLOCK(vp, 0, p);
1055 		vn_close(vp, flags & FMASK, fp->f_cred, p);
1056 		fdrop(fp, p);
1057 		p->p_retval[0] = indx;
1058 		return 0;
1059 	}
1060 
1061 	fp->f_data = (caddr_t)vp;
1062 	fp->f_flag = flags & FMASK;
1063 	fp->f_ops = &vnops;
1064 	fp->f_type = (vp->v_type == VFIFO ? DTYPE_FIFO : DTYPE_VNODE);
1065 	VOP_UNLOCK(vp, 0, p);
1066 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1067 		lf.l_whence = SEEK_SET;
1068 		lf.l_start = 0;
1069 		lf.l_len = 0;
1070 		if (flags & O_EXLOCK)
1071 			lf.l_type = F_WRLCK;
1072 		else
1073 			lf.l_type = F_RDLCK;
1074 		type = F_FLOCK;
1075 		if ((flags & FNONBLOCK) == 0)
1076 			type |= F_WAIT;
1077 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0)
1078 			goto bad;
1079 		fp->f_flag |= FHASLOCK;
1080 	}
1081 	if (flags & O_TRUNC) {
1082 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1083 			goto bad;
1084 		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1085 		VATTR_NULL(&vat);
1086 		vat.va_size = 0;
1087 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1088 		error = VOP_SETATTR(vp, &vat, p->p_ucred, p);
1089 		VOP_UNLOCK(vp, 0, p);
1090 		vn_finished_write(mp);
1091 		if (error)
1092 			goto bad;
1093 	}
1094 	/* assert that vn_open created a backing object if one is needed */
1095 	KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1096 		("open: vmio vnode has no backing object after vn_open"));
1097 	/*
1098 	 * Release our private reference, leaving the one associated with
1099 	 * the descriptor table intact.
1100 	 */
1101 	fdrop(fp, p);
1102 	p->p_retval[0] = indx;
1103 	return (0);
1104 bad:
1105 	if (fdp->fd_ofiles[indx] == fp) {
1106 		fdp->fd_ofiles[indx] = NULL;
1107 		fdrop(fp, p);
1108 	}
1109 	fdrop(fp, p);
1110 	return (error);
1111 }
1112 
1113 #ifdef COMPAT_43
1114 /*
1115  * Create a file.
1116  */
1117 #ifndef _SYS_SYSPROTO_H_
1118 struct ocreat_args {
1119 	char	*path;
1120 	int	mode;
1121 };
1122 #endif
1123 int
1124 ocreat(p, uap)
1125 	struct proc *p;
1126 	register struct ocreat_args /* {
1127 		syscallarg(char *) path;
1128 		syscallarg(int) mode;
1129 	} */ *uap;
1130 {
1131 	struct open_args /* {
1132 		syscallarg(char *) path;
1133 		syscallarg(int) flags;
1134 		syscallarg(int) mode;
1135 	} */ nuap;
1136 
1137 	SCARG(&nuap, path) = SCARG(uap, path);
1138 	SCARG(&nuap, mode) = SCARG(uap, mode);
1139 	SCARG(&nuap, flags) = O_WRONLY | O_CREAT | O_TRUNC;
1140 	return (open(p, &nuap));
1141 }
1142 #endif /* COMPAT_43 */
1143 
1144 /*
1145  * Create a special file.
1146  */
1147 #ifndef _SYS_SYSPROTO_H_
1148 struct mknod_args {
1149 	char	*path;
1150 	int	mode;
1151 	int	dev;
1152 };
1153 #endif
1154 /* ARGSUSED */
1155 int
1156 mknod(p, uap)
1157 	struct proc *p;
1158 	register struct mknod_args /* {
1159 		syscallarg(char *) path;
1160 		syscallarg(int) mode;
1161 		syscallarg(int) dev;
1162 	} */ *uap;
1163 {
1164 	struct vnode *vp;
1165 	struct mount *mp;
1166 	struct vattr vattr;
1167 	int error;
1168 	int whiteout = 0;
1169 	struct nameidata nd;
1170 
1171 	switch (SCARG(uap, mode) & S_IFMT) {
1172 	case S_IFCHR:
1173 	case S_IFBLK:
1174 		error = suser(p);
1175 		break;
1176 	default:
1177 		error = suser_xxx(0, p, PRISON_ROOT);
1178 		break;
1179 	}
1180 	if (error)
1181 		return (error);
1182 restart:
1183 	bwillwrite();
1184 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1185 	if ((error = namei(&nd)) != 0)
1186 		return (error);
1187 	vp = nd.ni_vp;
1188 	if (vp != NULL) {
1189 		vrele(vp);
1190 		error = EEXIST;
1191 	} else {
1192 		VATTR_NULL(&vattr);
1193 		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1194 		vattr.va_rdev = SCARG(uap, dev);
1195 		whiteout = 0;
1196 
1197 		switch (SCARG(uap, mode) & S_IFMT) {
1198 		case S_IFMT:	/* used by badsect to flag bad sectors */
1199 			vattr.va_type = VBAD;
1200 			break;
1201 		case S_IFCHR:
1202 			vattr.va_type = VCHR;
1203 			break;
1204 		case S_IFBLK:
1205 			vattr.va_type = VBLK;
1206 			break;
1207 		case S_IFWHT:
1208 			whiteout = 1;
1209 			break;
1210 		default:
1211 			error = EINVAL;
1212 			break;
1213 		}
1214 	}
1215 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1216 		NDFREE(&nd, NDF_ONLY_PNBUF);
1217 		vput(nd.ni_dvp);
1218 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1219 			return (error);
1220 		goto restart;
1221 	}
1222 	if (!error) {
1223 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1224 		if (whiteout)
1225 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1226 		else {
1227 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1228 						&nd.ni_cnd, &vattr);
1229 			if (error == 0)
1230 				vput(nd.ni_vp);
1231 		}
1232 	}
1233 	NDFREE(&nd, NDF_ONLY_PNBUF);
1234 	vput(nd.ni_dvp);
1235 	vn_finished_write(mp);
1236 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mknod");
1237 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "mknod");
1238 	return (error);
1239 }
1240 
1241 /*
1242  * Create a named pipe.
1243  */
1244 #ifndef _SYS_SYSPROTO_H_
1245 struct mkfifo_args {
1246 	char	*path;
1247 	int	mode;
1248 };
1249 #endif
1250 /* ARGSUSED */
1251 int
1252 mkfifo(p, uap)
1253 	struct proc *p;
1254 	register struct mkfifo_args /* {
1255 		syscallarg(char *) path;
1256 		syscallarg(int) mode;
1257 	} */ *uap;
1258 {
1259 	struct mount *mp;
1260 	struct vattr vattr;
1261 	int error;
1262 	struct nameidata nd;
1263 
1264 restart:
1265 	bwillwrite();
1266 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1267 	if ((error = namei(&nd)) != 0)
1268 		return (error);
1269 	if (nd.ni_vp != NULL) {
1270 		NDFREE(&nd, NDF_ONLY_PNBUF);
1271 		vrele(nd.ni_vp);
1272 		vput(nd.ni_dvp);
1273 		return (EEXIST);
1274 	}
1275 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1276 		NDFREE(&nd, NDF_ONLY_PNBUF);
1277 		vput(nd.ni_dvp);
1278 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1279 			return (error);
1280 		goto restart;
1281 	}
1282 	VATTR_NULL(&vattr);
1283 	vattr.va_type = VFIFO;
1284 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1285 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1286 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1287 	if (error == 0)
1288 		vput(nd.ni_vp);
1289 	NDFREE(&nd, NDF_ONLY_PNBUF);
1290 	vput(nd.ni_dvp);
1291 	vn_finished_write(mp);
1292 	return (error);
1293 }
1294 
1295 /*
1296  * Make a hard file link.
1297  */
1298 #ifndef _SYS_SYSPROTO_H_
1299 struct link_args {
1300 	char	*path;
1301 	char	*link;
1302 };
1303 #endif
1304 /* ARGSUSED */
1305 int
1306 link(p, uap)
1307 	struct proc *p;
1308 	register struct link_args /* {
1309 		syscallarg(char *) path;
1310 		syscallarg(char *) link;
1311 	} */ *uap;
1312 {
1313 	struct vnode *vp;
1314 	struct mount *mp;
1315 	struct nameidata nd;
1316 	int error;
1317 
1318 	bwillwrite();
1319 	NDINIT(&nd, LOOKUP, FOLLOW|NOOBJ, UIO_USERSPACE, SCARG(uap, path), p);
1320 	if ((error = namei(&nd)) != 0)
1321 		return (error);
1322 	NDFREE(&nd, NDF_ONLY_PNBUF);
1323 	vp = nd.ni_vp;
1324 	if (vp->v_type == VDIR) {
1325 		vrele(vp);
1326 		return (EPERM);		/* POSIX */
1327 	}
1328 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1329 		vrele(vp);
1330 		return (error);
1331 	}
1332 	NDINIT(&nd, CREATE, LOCKPARENT|NOOBJ, UIO_USERSPACE, SCARG(uap, link), p);
1333 	if ((error = namei(&nd)) == 0) {
1334 		if (nd.ni_vp != NULL) {
1335 			vrele(nd.ni_vp);
1336 			error = EEXIST;
1337 		} else {
1338 			VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1339 			VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1340 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1341 		}
1342 		NDFREE(&nd, NDF_ONLY_PNBUF);
1343 		vput(nd.ni_dvp);
1344 	}
1345 	vrele(vp);
1346 	vn_finished_write(mp);
1347 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "link");
1348 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "link");
1349 	return (error);
1350 }
1351 
1352 /*
1353  * Make a symbolic link.
1354  */
1355 #ifndef _SYS_SYSPROTO_H_
1356 struct symlink_args {
1357 	char	*path;
1358 	char	*link;
1359 };
1360 #endif
1361 /* ARGSUSED */
1362 int
1363 symlink(p, uap)
1364 	struct proc *p;
1365 	register struct symlink_args /* {
1366 		syscallarg(char *) path;
1367 		syscallarg(char *) link;
1368 	} */ *uap;
1369 {
1370 	struct mount *mp;
1371 	struct vattr vattr;
1372 	char *path;
1373 	int error;
1374 	struct nameidata nd;
1375 
1376 	path = zalloc(namei_zone);
1377 	if ((error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL)) != 0)
1378 		goto out;
1379 restart:
1380 	bwillwrite();
1381 	NDINIT(&nd, CREATE, LOCKPARENT|NOOBJ, UIO_USERSPACE, SCARG(uap, link), p);
1382 	if ((error = namei(&nd)) != 0)
1383 		goto out;
1384 	if (nd.ni_vp) {
1385 		NDFREE(&nd, NDF_ONLY_PNBUF);
1386 		vrele(nd.ni_vp);
1387 		vput(nd.ni_dvp);
1388 		error = EEXIST;
1389 		goto out;
1390 	}
1391 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1392 		NDFREE(&nd, NDF_ONLY_PNBUF);
1393 		vput(nd.ni_dvp);
1394 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1395 			return (error);
1396 		goto restart;
1397 	}
1398 	VATTR_NULL(&vattr);
1399 	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1400 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1401 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1402 	NDFREE(&nd, NDF_ONLY_PNBUF);
1403 	if (error == 0)
1404 		vput(nd.ni_vp);
1405 	vput(nd.ni_dvp);
1406 	vn_finished_write(mp);
1407 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "symlink");
1408 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "symlink");
1409 out:
1410 	zfree(namei_zone, path);
1411 	return (error);
1412 }
1413 
1414 /*
1415  * Delete a whiteout from the filesystem.
1416  */
1417 /* ARGSUSED */
1418 int
1419 undelete(p, uap)
1420 	struct proc *p;
1421 	register struct undelete_args /* {
1422 		syscallarg(char *) path;
1423 	} */ *uap;
1424 {
1425 	int error;
1426 	struct mount *mp;
1427 	struct nameidata nd;
1428 
1429 restart:
1430 	bwillwrite();
1431 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1432 	    SCARG(uap, path), p);
1433 	error = namei(&nd);
1434 	if (error)
1435 		return (error);
1436 
1437 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1438 		NDFREE(&nd, NDF_ONLY_PNBUF);
1439 		if (nd.ni_vp)
1440 			vrele(nd.ni_vp);
1441 		vput(nd.ni_dvp);
1442 		return (EEXIST);
1443 	}
1444 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1445 		NDFREE(&nd, NDF_ONLY_PNBUF);
1446 		vput(nd.ni_dvp);
1447 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1448 			return (error);
1449 		goto restart;
1450 	}
1451 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1452 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1453 	NDFREE(&nd, NDF_ONLY_PNBUF);
1454 	vput(nd.ni_dvp);
1455 	vn_finished_write(mp);
1456 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "undelete");
1457 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "undelete");
1458 	return (error);
1459 }
1460 
1461 /*
1462  * Delete a name from the filesystem.
1463  */
1464 #ifndef _SYS_SYSPROTO_H_
1465 struct unlink_args {
1466 	char	*path;
1467 };
1468 #endif
1469 /* ARGSUSED */
1470 int
1471 unlink(p, uap)
1472 	struct proc *p;
1473 	struct unlink_args /* {
1474 		syscallarg(char *) path;
1475 	} */ *uap;
1476 {
1477 	struct mount *mp;
1478 	struct vnode *vp;
1479 	int error;
1480 	struct nameidata nd;
1481 
1482 restart:
1483 	bwillwrite();
1484 	NDINIT(&nd, DELETE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1485 	if ((error = namei(&nd)) != 0)
1486 		return (error);
1487 	vp = nd.ni_vp;
1488 	if (vp->v_type == VDIR)
1489 		error = EPERM;		/* POSIX */
1490 	else {
1491 		/*
1492 		 * The root of a mounted filesystem cannot be deleted.
1493 		 *
1494 		 * XXX: can this only be a VDIR case?
1495 		 */
1496 		if (vp->v_flag & VROOT)
1497 			error = EBUSY;
1498 	}
1499 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1500 		NDFREE(&nd, NDF_ONLY_PNBUF);
1501 		vrele(vp);
1502 		vput(nd.ni_dvp);
1503 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1504 			return (error);
1505 		goto restart;
1506 	}
1507 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1508 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1509 	if (!error) {
1510 		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1511 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1512 	}
1513 	NDFREE(&nd, NDF_ONLY_PNBUF);
1514 	vput(nd.ni_dvp);
1515 	vput(vp);
1516 	vn_finished_write(mp);
1517 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "unlink");
1518 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "unlink");
1519 	return (error);
1520 }
1521 
1522 /*
1523  * Reposition read/write file offset.
1524  */
1525 #ifndef _SYS_SYSPROTO_H_
1526 struct lseek_args {
1527 	int	fd;
1528 	int	pad;
1529 	off_t	offset;
1530 	int	whence;
1531 };
1532 #endif
1533 int
1534 lseek(p, uap)
1535 	struct proc *p;
1536 	register struct lseek_args /* {
1537 		syscallarg(int) fd;
1538 		syscallarg(int) pad;
1539 		syscallarg(off_t) offset;
1540 		syscallarg(int) whence;
1541 	} */ *uap;
1542 {
1543 	struct ucred *cred = p->p_ucred;
1544 	register struct filedesc *fdp = p->p_fd;
1545 	register struct file *fp;
1546 	struct vattr vattr;
1547 	int error;
1548 
1549 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1550 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
1551 		return (EBADF);
1552 	if (fp->f_type != DTYPE_VNODE)
1553 		return (ESPIPE);
1554 	switch (SCARG(uap, whence)) {
1555 	case L_INCR:
1556 		fp->f_offset += SCARG(uap, offset);
1557 		break;
1558 	case L_XTND:
1559 		error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, cred, p);
1560 		if (error)
1561 			return (error);
1562 		fp->f_offset = SCARG(uap, offset) + vattr.va_size;
1563 		break;
1564 	case L_SET:
1565 		fp->f_offset = SCARG(uap, offset);
1566 		break;
1567 	default:
1568 		return (EINVAL);
1569 	}
1570 	*(off_t *)(p->p_retval) = fp->f_offset;
1571 	return (0);
1572 }
1573 
1574 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
1575 /*
1576  * Reposition read/write file offset.
1577  */
1578 #ifndef _SYS_SYSPROTO_H_
1579 struct olseek_args {
1580 	int	fd;
1581 	long	offset;
1582 	int	whence;
1583 };
1584 #endif
1585 int
1586 olseek(p, uap)
1587 	struct proc *p;
1588 	register struct olseek_args /* {
1589 		syscallarg(int) fd;
1590 		syscallarg(long) offset;
1591 		syscallarg(int) whence;
1592 	} */ *uap;
1593 {
1594 	struct lseek_args /* {
1595 		syscallarg(int) fd;
1596 		syscallarg(int) pad;
1597 		syscallarg(off_t) offset;
1598 		syscallarg(int) whence;
1599 	} */ nuap;
1600 	int error;
1601 
1602 	SCARG(&nuap, fd) = SCARG(uap, fd);
1603 	SCARG(&nuap, offset) = SCARG(uap, offset);
1604 	SCARG(&nuap, whence) = SCARG(uap, whence);
1605 	error = lseek(p, &nuap);
1606 	return (error);
1607 }
1608 #endif /* COMPAT_43 */
1609 
1610 /*
1611  * Check access permissions.
1612  */
1613 #ifndef _SYS_SYSPROTO_H_
1614 struct access_args {
1615 	char	*path;
1616 	int	flags;
1617 };
1618 #endif
1619 int
1620 access(p, uap)
1621 	struct proc *p;
1622 	register struct access_args /* {
1623 		syscallarg(char *) path;
1624 		syscallarg(int) flags;
1625 	} */ *uap;
1626 {
1627 	struct ucred *cred, *tmpcred;
1628 	register struct vnode *vp;
1629 	int error, flags;
1630 	struct nameidata nd;
1631 
1632 	cred = p->p_ucred;
1633 	/*
1634 	 * Create and modify a temporary credential instead of one that
1635 	 * is potentially shared.  This could also mess up socket
1636 	 * buffer accounting which can run in an interrupt context.
1637 	 *
1638 	 * XXX - Depending on how "threads" are finally implemented, it
1639 	 * may be better to explicitly pass the credential to namei()
1640 	 * rather than to modify the potentially shared process structure.
1641 	 */
1642 	tmpcred = crdup(cred);
1643 	tmpcred->cr_uid = p->p_cred->p_ruid;
1644 	tmpcred->cr_groups[0] = p->p_cred->p_rgid;
1645 	p->p_ucred = tmpcred;
1646 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1647 	    SCARG(uap, path), p);
1648 	if ((error = namei(&nd)) != 0)
1649 		goto out1;
1650 	vp = nd.ni_vp;
1651 
1652 	/* Flags == 0 means only check for existence. */
1653 	if (SCARG(uap, flags)) {
1654 		flags = 0;
1655 		if (SCARG(uap, flags) & R_OK)
1656 			flags |= VREAD;
1657 		if (SCARG(uap, flags) & W_OK)
1658 			flags |= VWRITE;
1659 		if (SCARG(uap, flags) & X_OK)
1660 			flags |= VEXEC;
1661 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1662 			error = VOP_ACCESS(vp, flags, cred, p);
1663 	}
1664 	NDFREE(&nd, NDF_ONLY_PNBUF);
1665 	vput(vp);
1666 out1:
1667 	p->p_ucred = cred;
1668 	crfree(tmpcred);
1669 	return (error);
1670 }
1671 
1672 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
1673 /*
1674  * Get file status; this version follows links.
1675  */
1676 #ifndef _SYS_SYSPROTO_H_
1677 struct ostat_args {
1678 	char	*path;
1679 	struct ostat *ub;
1680 };
1681 #endif
1682 /* ARGSUSED */
1683 int
1684 ostat(p, uap)
1685 	struct proc *p;
1686 	register struct ostat_args /* {
1687 		syscallarg(char *) path;
1688 		syscallarg(struct ostat *) ub;
1689 	} */ *uap;
1690 {
1691 	struct stat sb;
1692 	struct ostat osb;
1693 	int error;
1694 	struct nameidata nd;
1695 
1696 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1697 	    SCARG(uap, path), p);
1698 	if ((error = namei(&nd)) != 0)
1699 		return (error);
1700 	NDFREE(&nd, NDF_ONLY_PNBUF);
1701 	error = vn_stat(nd.ni_vp, &sb, p);
1702 	vput(nd.ni_vp);
1703 	if (error)
1704 		return (error);
1705 	cvtstat(&sb, &osb);
1706 	error = copyout((caddr_t)&osb, (caddr_t)SCARG(uap, ub), sizeof (osb));
1707 	return (error);
1708 }
1709 
1710 /*
1711  * Get file status; this version does not follow links.
1712  */
1713 #ifndef _SYS_SYSPROTO_H_
1714 struct olstat_args {
1715 	char	*path;
1716 	struct ostat *ub;
1717 };
1718 #endif
1719 /* ARGSUSED */
1720 int
1721 olstat(p, uap)
1722 	struct proc *p;
1723 	register struct olstat_args /* {
1724 		syscallarg(char *) path;
1725 		syscallarg(struct ostat *) ub;
1726 	} */ *uap;
1727 {
1728 	struct vnode *vp;
1729 	struct stat sb;
1730 	struct ostat osb;
1731 	int error;
1732 	struct nameidata nd;
1733 
1734 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1735 	    SCARG(uap, path), p);
1736 	if ((error = namei(&nd)) != 0)
1737 		return (error);
1738 	vp = nd.ni_vp;
1739 	error = vn_stat(vp, &sb, p);
1740 	NDFREE(&nd, NDF_ONLY_PNBUF);
1741 	vput(vp);
1742 	if (error)
1743 		return (error);
1744 	cvtstat(&sb, &osb);
1745 	error = copyout((caddr_t)&osb, (caddr_t)SCARG(uap, ub), sizeof (osb));
1746 	return (error);
1747 }
1748 
1749 /*
1750  * Convert from an old to a new stat structure.
1751  */
1752 void
1753 cvtstat(st, ost)
1754 	struct stat *st;
1755 	struct ostat *ost;
1756 {
1757 
1758 	ost->st_dev = st->st_dev;
1759 	ost->st_ino = st->st_ino;
1760 	ost->st_mode = st->st_mode;
1761 	ost->st_nlink = st->st_nlink;
1762 	ost->st_uid = st->st_uid;
1763 	ost->st_gid = st->st_gid;
1764 	ost->st_rdev = st->st_rdev;
1765 	if (st->st_size < (quad_t)1 << 32)
1766 		ost->st_size = st->st_size;
1767 	else
1768 		ost->st_size = -2;
1769 	ost->st_atime = st->st_atime;
1770 	ost->st_mtime = st->st_mtime;
1771 	ost->st_ctime = st->st_ctime;
1772 	ost->st_blksize = st->st_blksize;
1773 	ost->st_blocks = st->st_blocks;
1774 	ost->st_flags = st->st_flags;
1775 	ost->st_gen = st->st_gen;
1776 }
1777 #endif /* COMPAT_43 || COMPAT_SUNOS */
1778 
1779 /*
1780  * Get file status; this version follows links.
1781  */
1782 #ifndef _SYS_SYSPROTO_H_
1783 struct stat_args {
1784 	char	*path;
1785 	struct stat *ub;
1786 };
1787 #endif
1788 /* ARGSUSED */
1789 int
1790 stat(p, uap)
1791 	struct proc *p;
1792 	register struct stat_args /* {
1793 		syscallarg(char *) path;
1794 		syscallarg(struct stat *) ub;
1795 	} */ *uap;
1796 {
1797 	struct stat sb;
1798 	int error;
1799 	struct nameidata nd;
1800 
1801 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1802 	    SCARG(uap, path), p);
1803 	if ((error = namei(&nd)) != 0)
1804 		return (error);
1805 	error = vn_stat(nd.ni_vp, &sb, p);
1806 	NDFREE(&nd, NDF_ONLY_PNBUF);
1807 	vput(nd.ni_vp);
1808 	if (error)
1809 		return (error);
1810 	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1811 	return (error);
1812 }
1813 
1814 /*
1815  * Get file status; this version does not follow links.
1816  */
1817 #ifndef _SYS_SYSPROTO_H_
1818 struct lstat_args {
1819 	char	*path;
1820 	struct stat *ub;
1821 };
1822 #endif
1823 /* ARGSUSED */
1824 int
1825 lstat(p, uap)
1826 	struct proc *p;
1827 	register struct lstat_args /* {
1828 		syscallarg(char *) path;
1829 		syscallarg(struct stat *) ub;
1830 	} */ *uap;
1831 {
1832 	int error;
1833 	struct vnode *vp;
1834 	struct stat sb;
1835 	struct nameidata nd;
1836 
1837 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1838 	    SCARG(uap, path), p);
1839 	if ((error = namei(&nd)) != 0)
1840 		return (error);
1841 	vp = nd.ni_vp;
1842 	error = vn_stat(vp, &sb, p);
1843 	NDFREE(&nd, NDF_ONLY_PNBUF);
1844 	vput(vp);
1845 	if (error)
1846 		return (error);
1847 	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1848 	return (error);
1849 }
1850 
1851 /*
1852  * Implementation of the NetBSD stat() function.
1853  * XXX This should probably be collapsed with the FreeBSD version,
1854  * as the differences are only due to vn_stat() clearing spares at
1855  * the end of the structures.  vn_stat could be split to avoid this,
1856  * and thus collapse the following to close to zero code.
1857  */
1858 void
1859 cvtnstat(sb, nsb)
1860 	struct stat *sb;
1861 	struct nstat *nsb;
1862 {
1863 	nsb->st_dev = sb->st_dev;
1864 	nsb->st_ino = sb->st_ino;
1865 	nsb->st_mode = sb->st_mode;
1866 	nsb->st_nlink = sb->st_nlink;
1867 	nsb->st_uid = sb->st_uid;
1868 	nsb->st_gid = sb->st_gid;
1869 	nsb->st_rdev = sb->st_rdev;
1870 	nsb->st_atimespec = sb->st_atimespec;
1871 	nsb->st_mtimespec = sb->st_mtimespec;
1872 	nsb->st_ctimespec = sb->st_ctimespec;
1873 	nsb->st_size = sb->st_size;
1874 	nsb->st_blocks = sb->st_blocks;
1875 	nsb->st_blksize = sb->st_blksize;
1876 	nsb->st_flags = sb->st_flags;
1877 	nsb->st_gen = sb->st_gen;
1878 	nsb->st_qspare[0] = sb->st_qspare[0];
1879 	nsb->st_qspare[1] = sb->st_qspare[1];
1880 }
1881 
1882 #ifndef _SYS_SYSPROTO_H_
1883 struct nstat_args {
1884 	char	*path;
1885 	struct nstat *ub;
1886 };
1887 #endif
1888 /* ARGSUSED */
1889 int
1890 nstat(p, uap)
1891 	struct proc *p;
1892 	register struct nstat_args /* {
1893 		syscallarg(char *) path;
1894 		syscallarg(struct nstat *) ub;
1895 	} */ *uap;
1896 {
1897 	struct stat sb;
1898 	struct nstat nsb;
1899 	int error;
1900 	struct nameidata nd;
1901 
1902 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1903 	    SCARG(uap, path), p);
1904 	if ((error = namei(&nd)) != 0)
1905 		return (error);
1906 	NDFREE(&nd, NDF_ONLY_PNBUF);
1907 	error = vn_stat(nd.ni_vp, &sb, p);
1908 	vput(nd.ni_vp);
1909 	if (error)
1910 		return (error);
1911 	cvtnstat(&sb, &nsb);
1912 	error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb));
1913 	return (error);
1914 }
1915 
1916 /*
1917  * NetBSD lstat.  Get file status; this version does not follow links.
1918  */
1919 #ifndef _SYS_SYSPROTO_H_
1920 struct lstat_args {
1921 	char	*path;
1922 	struct stat *ub;
1923 };
1924 #endif
1925 /* ARGSUSED */
1926 int
1927 nlstat(p, uap)
1928 	struct proc *p;
1929 	register struct nlstat_args /* {
1930 		syscallarg(char *) path;
1931 		syscallarg(struct nstat *) ub;
1932 	} */ *uap;
1933 {
1934 	int error;
1935 	struct vnode *vp;
1936 	struct stat sb;
1937 	struct nstat nsb;
1938 	struct nameidata nd;
1939 
1940 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1941 	    SCARG(uap, path), p);
1942 	if ((error = namei(&nd)) != 0)
1943 		return (error);
1944 	vp = nd.ni_vp;
1945 	NDFREE(&nd, NDF_ONLY_PNBUF);
1946 	error = vn_stat(vp, &sb, p);
1947 	vput(vp);
1948 	if (error)
1949 		return (error);
1950 	cvtnstat(&sb, &nsb);
1951 	error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb));
1952 	return (error);
1953 }
1954 
1955 /*
1956  * Get configurable pathname variables.
1957  */
1958 #ifndef _SYS_SYSPROTO_H_
1959 struct pathconf_args {
1960 	char	*path;
1961 	int	name;
1962 };
1963 #endif
1964 /* ARGSUSED */
1965 int
1966 pathconf(p, uap)
1967 	struct proc *p;
1968 	register struct pathconf_args /* {
1969 		syscallarg(char *) path;
1970 		syscallarg(int) name;
1971 	} */ *uap;
1972 {
1973 	int error;
1974 	struct nameidata nd;
1975 
1976 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1977 	    SCARG(uap, path), p);
1978 	if ((error = namei(&nd)) != 0)
1979 		return (error);
1980 	NDFREE(&nd, NDF_ONLY_PNBUF);
1981 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), p->p_retval);
1982 	vput(nd.ni_vp);
1983 	return (error);
1984 }
1985 
1986 /*
1987  * Return target name of a symbolic link.
1988  */
1989 #ifndef _SYS_SYSPROTO_H_
1990 struct readlink_args {
1991 	char	*path;
1992 	char	*buf;
1993 	int	count;
1994 };
1995 #endif
1996 /* ARGSUSED */
1997 int
1998 readlink(p, uap)
1999 	struct proc *p;
2000 	register struct readlink_args /* {
2001 		syscallarg(char *) path;
2002 		syscallarg(char *) buf;
2003 		syscallarg(int) count;
2004 	} */ *uap;
2005 {
2006 	register struct vnode *vp;
2007 	struct iovec aiov;
2008 	struct uio auio;
2009 	int error;
2010 	struct nameidata nd;
2011 
2012 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2013 	    SCARG(uap, path), p);
2014 	if ((error = namei(&nd)) != 0)
2015 		return (error);
2016 	NDFREE(&nd, NDF_ONLY_PNBUF);
2017 	vp = nd.ni_vp;
2018 	if (vp->v_type != VLNK)
2019 		error = EINVAL;
2020 	else {
2021 		aiov.iov_base = SCARG(uap, buf);
2022 		aiov.iov_len = SCARG(uap, count);
2023 		auio.uio_iov = &aiov;
2024 		auio.uio_iovcnt = 1;
2025 		auio.uio_offset = 0;
2026 		auio.uio_rw = UIO_READ;
2027 		auio.uio_segflg = UIO_USERSPACE;
2028 		auio.uio_procp = p;
2029 		auio.uio_resid = SCARG(uap, count);
2030 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2031 	}
2032 	vput(vp);
2033 	p->p_retval[0] = SCARG(uap, count) - auio.uio_resid;
2034 	return (error);
2035 }
2036 
2037 /*
2038  * Common implementation code for chflags() and fchflags().
2039  */
2040 static int
2041 setfflags(p, vp, flags)
2042 	struct proc *p;
2043 	struct vnode *vp;
2044 	int flags;
2045 {
2046 	int error;
2047 	struct mount *mp;
2048 	struct vattr vattr;
2049 
2050 	/*
2051 	 * Prevent non-root users from setting flags on devices.  When
2052 	 * a device is reused, users can retain ownership of the device
2053 	 * if they are allowed to set flags and programs assume that
2054 	 * chown can't fail when done as root.
2055 	 */
2056 	if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2057 	    ((error = suser_xxx(p->p_ucred, p, PRISON_ROOT)) != 0))
2058 		return (error);
2059 
2060 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2061 		return (error);
2062 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2063 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2064 	VATTR_NULL(&vattr);
2065 	vattr.va_flags = flags;
2066 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2067 	VOP_UNLOCK(vp, 0, p);
2068 	vn_finished_write(mp);
2069 	return (error);
2070 }
2071 
2072 /*
2073  * Change flags of a file given a path name.
2074  */
2075 #ifndef _SYS_SYSPROTO_H_
2076 struct chflags_args {
2077 	char	*path;
2078 	int	flags;
2079 };
2080 #endif
2081 /* ARGSUSED */
2082 int
2083 chflags(p, uap)
2084 	struct proc *p;
2085 	register struct chflags_args /* {
2086 		syscallarg(char *) path;
2087 		syscallarg(int) flags;
2088 	} */ *uap;
2089 {
2090 	int error;
2091 	struct nameidata nd;
2092 
2093 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2094 	if ((error = namei(&nd)) != 0)
2095 		return (error);
2096 	NDFREE(&nd, NDF_ONLY_PNBUF);
2097 	error = setfflags(p, nd.ni_vp, SCARG(uap, flags));
2098 	vrele(nd.ni_vp);
2099 	return error;
2100 }
2101 
2102 /*
2103  * Change flags of a file given a file descriptor.
2104  */
2105 #ifndef _SYS_SYSPROTO_H_
2106 struct fchflags_args {
2107 	int	fd;
2108 	int	flags;
2109 };
2110 #endif
2111 /* ARGSUSED */
2112 int
2113 fchflags(p, uap)
2114 	struct proc *p;
2115 	register struct fchflags_args /* {
2116 		syscallarg(int) fd;
2117 		syscallarg(int) flags;
2118 	} */ *uap;
2119 {
2120 	struct file *fp;
2121 	int error;
2122 
2123 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2124 		return (error);
2125 	return setfflags(p, (struct vnode *) fp->f_data, SCARG(uap, flags));
2126 }
2127 
2128 /*
2129  * Common implementation code for chmod(), lchmod() and fchmod().
2130  */
2131 static int
2132 setfmode(p, vp, mode)
2133 	struct proc *p;
2134 	struct vnode *vp;
2135 	int mode;
2136 {
2137 	int error;
2138 	struct mount *mp;
2139 	struct vattr vattr;
2140 
2141 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2142 		return (error);
2143 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2144 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2145 	VATTR_NULL(&vattr);
2146 	vattr.va_mode = mode & ALLPERMS;
2147 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2148 	VOP_UNLOCK(vp, 0, p);
2149 	vn_finished_write(mp);
2150 	return error;
2151 }
2152 
2153 /*
2154  * Change mode of a file given path name.
2155  */
2156 #ifndef _SYS_SYSPROTO_H_
2157 struct chmod_args {
2158 	char	*path;
2159 	int	mode;
2160 };
2161 #endif
2162 /* ARGSUSED */
2163 int
2164 chmod(p, uap)
2165 	struct proc *p;
2166 	register struct chmod_args /* {
2167 		syscallarg(char *) path;
2168 		syscallarg(int) mode;
2169 	} */ *uap;
2170 {
2171 	int error;
2172 	struct nameidata nd;
2173 
2174 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2175 	if ((error = namei(&nd)) != 0)
2176 		return (error);
2177 	NDFREE(&nd, NDF_ONLY_PNBUF);
2178 	error = setfmode(p, nd.ni_vp, SCARG(uap, mode));
2179 	vrele(nd.ni_vp);
2180 	return error;
2181 }
2182 
2183 /*
2184  * Change mode of a file given path name (don't follow links.)
2185  */
2186 #ifndef _SYS_SYSPROTO_H_
2187 struct lchmod_args {
2188 	char	*path;
2189 	int	mode;
2190 };
2191 #endif
2192 /* ARGSUSED */
2193 int
2194 lchmod(p, uap)
2195 	struct proc *p;
2196 	register struct lchmod_args /* {
2197 		syscallarg(char *) path;
2198 		syscallarg(int) mode;
2199 	} */ *uap;
2200 {
2201 	int error;
2202 	struct nameidata nd;
2203 
2204 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2205 	if ((error = namei(&nd)) != 0)
2206 		return (error);
2207 	NDFREE(&nd, NDF_ONLY_PNBUF);
2208 	error = setfmode(p, nd.ni_vp, SCARG(uap, mode));
2209 	vrele(nd.ni_vp);
2210 	return error;
2211 }
2212 
2213 /*
2214  * Change mode of a file given a file descriptor.
2215  */
2216 #ifndef _SYS_SYSPROTO_H_
2217 struct fchmod_args {
2218 	int	fd;
2219 	int	mode;
2220 };
2221 #endif
2222 /* ARGSUSED */
2223 int
2224 fchmod(p, uap)
2225 	struct proc *p;
2226 	register struct fchmod_args /* {
2227 		syscallarg(int) fd;
2228 		syscallarg(int) mode;
2229 	} */ *uap;
2230 {
2231 	struct file *fp;
2232 	int error;
2233 
2234 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2235 		return (error);
2236 	return setfmode(p, (struct vnode *)fp->f_data, SCARG(uap, mode));
2237 }
2238 
2239 /*
2240  * Common implementation for chown(), lchown(), and fchown()
2241  */
2242 static int
2243 setfown(p, vp, uid, gid)
2244 	struct proc *p;
2245 	struct vnode *vp;
2246 	uid_t uid;
2247 	gid_t gid;
2248 {
2249 	int error;
2250 	struct mount *mp;
2251 	struct vattr vattr;
2252 
2253 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2254 		return (error);
2255 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2256 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2257 	VATTR_NULL(&vattr);
2258 	vattr.va_uid = uid;
2259 	vattr.va_gid = gid;
2260 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2261 	VOP_UNLOCK(vp, 0, p);
2262 	vn_finished_write(mp);
2263 	return error;
2264 }
2265 
2266 /*
2267  * Set ownership given a path name.
2268  */
2269 #ifndef _SYS_SYSPROTO_H_
2270 struct chown_args {
2271 	char	*path;
2272 	int	uid;
2273 	int	gid;
2274 };
2275 #endif
2276 /* ARGSUSED */
2277 int
2278 chown(p, uap)
2279 	struct proc *p;
2280 	register struct chown_args /* {
2281 		syscallarg(char *) path;
2282 		syscallarg(int) uid;
2283 		syscallarg(int) gid;
2284 	} */ *uap;
2285 {
2286 	int error;
2287 	struct nameidata nd;
2288 
2289 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2290 	if ((error = namei(&nd)) != 0)
2291 		return (error);
2292 	NDFREE(&nd, NDF_ONLY_PNBUF);
2293 	error = setfown(p, nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid));
2294 	vrele(nd.ni_vp);
2295 	return (error);
2296 }
2297 
2298 /*
2299  * Set ownership given a path name, do not cross symlinks.
2300  */
2301 #ifndef _SYS_SYSPROTO_H_
2302 struct lchown_args {
2303 	char	*path;
2304 	int	uid;
2305 	int	gid;
2306 };
2307 #endif
2308 /* ARGSUSED */
2309 int
2310 lchown(p, uap)
2311 	struct proc *p;
2312 	register struct lchown_args /* {
2313 		syscallarg(char *) path;
2314 		syscallarg(int) uid;
2315 		syscallarg(int) gid;
2316 	} */ *uap;
2317 {
2318 	int error;
2319 	struct nameidata nd;
2320 
2321 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2322 	if ((error = namei(&nd)) != 0)
2323 		return (error);
2324 	NDFREE(&nd, NDF_ONLY_PNBUF);
2325 	error = setfown(p, nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid));
2326 	vrele(nd.ni_vp);
2327 	return (error);
2328 }
2329 
2330 /*
2331  * Set ownership given a file descriptor.
2332  */
2333 #ifndef _SYS_SYSPROTO_H_
2334 struct fchown_args {
2335 	int	fd;
2336 	int	uid;
2337 	int	gid;
2338 };
2339 #endif
2340 /* ARGSUSED */
2341 int
2342 fchown(p, uap)
2343 	struct proc *p;
2344 	register struct fchown_args /* {
2345 		syscallarg(int) fd;
2346 		syscallarg(int) uid;
2347 		syscallarg(int) gid;
2348 	} */ *uap;
2349 {
2350 	struct file *fp;
2351 	int error;
2352 
2353 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2354 		return (error);
2355 	return setfown(p, (struct vnode *)fp->f_data,
2356 		SCARG(uap, uid), SCARG(uap, gid));
2357 }
2358 
2359 /*
2360  * Common implementation code for utimes(), lutimes(), and futimes().
2361  */
2362 static int
2363 getutimes(usrtvp, tsp)
2364 	const struct timeval *usrtvp;
2365 	struct timespec *tsp;
2366 {
2367 	struct timeval tv[2];
2368 	int error;
2369 
2370 	if (usrtvp == NULL) {
2371 		microtime(&tv[0]);
2372 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2373 		tsp[1] = tsp[0];
2374 	} else {
2375 		if ((error = copyin(usrtvp, tv, sizeof (tv))) != 0)
2376 			return (error);
2377 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2378 		TIMEVAL_TO_TIMESPEC(&tv[1], &tsp[1]);
2379 	}
2380 	return 0;
2381 }
2382 
2383 /*
2384  * Common implementation code for utimes(), lutimes(), and futimes().
2385  */
2386 static int
2387 setutimes(p, vp, ts, nullflag)
2388 	struct proc *p;
2389 	struct vnode *vp;
2390 	const struct timespec *ts;
2391 	int nullflag;
2392 {
2393 	int error;
2394 	struct mount *mp;
2395 	struct vattr vattr;
2396 
2397 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2398 		return (error);
2399 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2400 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2401 	VATTR_NULL(&vattr);
2402 	vattr.va_atime = ts[0];
2403 	vattr.va_mtime = ts[1];
2404 	if (nullflag)
2405 		vattr.va_vaflags |= VA_UTIMES_NULL;
2406 	error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2407 	VOP_UNLOCK(vp, 0, p);
2408 	vn_finished_write(mp);
2409 	return error;
2410 }
2411 
2412 /*
2413  * Set the access and modification times of a file.
2414  */
2415 #ifndef _SYS_SYSPROTO_H_
2416 struct utimes_args {
2417 	char	*path;
2418 	struct	timeval *tptr;
2419 };
2420 #endif
2421 /* ARGSUSED */
2422 int
2423 utimes(p, uap)
2424 	struct proc *p;
2425 	register struct utimes_args /* {
2426 		syscallarg(char *) path;
2427 		syscallarg(struct timeval *) tptr;
2428 	} */ *uap;
2429 {
2430 	struct timespec ts[2];
2431 	struct timeval *usrtvp;
2432 	int error;
2433 	struct nameidata nd;
2434 
2435 	usrtvp = SCARG(uap, tptr);
2436 	if ((error = getutimes(usrtvp, ts)) != 0)
2437 		return (error);
2438 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2439 	if ((error = namei(&nd)) != 0)
2440 		return (error);
2441 	NDFREE(&nd, NDF_ONLY_PNBUF);
2442 	error = setutimes(p, nd.ni_vp, ts, usrtvp == NULL);
2443 	vrele(nd.ni_vp);
2444 	return (error);
2445 }
2446 
2447 /*
2448  * Set the access and modification times of a file.
2449  */
2450 #ifndef _SYS_SYSPROTO_H_
2451 struct lutimes_args {
2452 	char	*path;
2453 	struct	timeval *tptr;
2454 };
2455 #endif
2456 /* ARGSUSED */
2457 int
2458 lutimes(p, uap)
2459 	struct proc *p;
2460 	register struct lutimes_args /* {
2461 		syscallarg(char *) path;
2462 		syscallarg(struct timeval *) tptr;
2463 	} */ *uap;
2464 {
2465 	struct timespec ts[2];
2466 	struct timeval *usrtvp;
2467 	int error;
2468 	struct nameidata nd;
2469 
2470 	usrtvp = SCARG(uap, tptr);
2471 	if ((error = getutimes(usrtvp, ts)) != 0)
2472 		return (error);
2473 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2474 	if ((error = namei(&nd)) != 0)
2475 		return (error);
2476 	NDFREE(&nd, NDF_ONLY_PNBUF);
2477 	error = setutimes(p, nd.ni_vp, ts, usrtvp == NULL);
2478 	vrele(nd.ni_vp);
2479 	return (error);
2480 }
2481 
2482 /*
2483  * Set the access and modification times of a file.
2484  */
2485 #ifndef _SYS_SYSPROTO_H_
2486 struct futimes_args {
2487 	int	fd;
2488 	struct	timeval *tptr;
2489 };
2490 #endif
2491 /* ARGSUSED */
2492 int
2493 futimes(p, uap)
2494 	struct proc *p;
2495 	register struct futimes_args /* {
2496 		syscallarg(int ) fd;
2497 		syscallarg(struct timeval *) tptr;
2498 	} */ *uap;
2499 {
2500 	struct timespec ts[2];
2501 	struct file *fp;
2502 	struct timeval *usrtvp;
2503 	int error;
2504 
2505 	usrtvp = SCARG(uap, tptr);
2506 	if ((error = getutimes(usrtvp, ts)) != 0)
2507 		return (error);
2508 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2509 		return (error);
2510 	return setutimes(p, (struct vnode *)fp->f_data, ts, usrtvp == NULL);
2511 }
2512 
2513 /*
2514  * Truncate a file given its path name.
2515  */
2516 #ifndef _SYS_SYSPROTO_H_
2517 struct truncate_args {
2518 	char	*path;
2519 	int	pad;
2520 	off_t	length;
2521 };
2522 #endif
2523 /* ARGSUSED */
2524 int
2525 truncate(p, uap)
2526 	struct proc *p;
2527 	register struct truncate_args /* {
2528 		syscallarg(char *) path;
2529 		syscallarg(int) pad;
2530 		syscallarg(off_t) length;
2531 	} */ *uap;
2532 {
2533 	struct mount *mp;
2534 	struct vnode *vp;
2535 	struct vattr vattr;
2536 	int error;
2537 	struct nameidata nd;
2538 
2539 	if (uap->length < 0)
2540 		return(EINVAL);
2541 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2542 	if ((error = namei(&nd)) != 0)
2543 		return (error);
2544 	vp = nd.ni_vp;
2545 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
2546 		vrele(vp);
2547 		return (error);
2548 	}
2549 	NDFREE(&nd, NDF_ONLY_PNBUF);
2550 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2551 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2552 	if (vp->v_type == VDIR)
2553 		error = EISDIR;
2554 	else if ((error = vn_writechk(vp)) == 0 &&
2555 	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2556 		VATTR_NULL(&vattr);
2557 		vattr.va_size = SCARG(uap, length);
2558 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2559 	}
2560 	vput(vp);
2561 	vn_finished_write(mp);
2562 	return (error);
2563 }
2564 
2565 /*
2566  * Truncate a file given a file descriptor.
2567  */
2568 #ifndef _SYS_SYSPROTO_H_
2569 struct ftruncate_args {
2570 	int	fd;
2571 	int	pad;
2572 	off_t	length;
2573 };
2574 #endif
2575 /* ARGSUSED */
2576 int
2577 ftruncate(p, uap)
2578 	struct proc *p;
2579 	register struct ftruncate_args /* {
2580 		syscallarg(int) fd;
2581 		syscallarg(int) pad;
2582 		syscallarg(off_t) length;
2583 	} */ *uap;
2584 {
2585 	struct mount *mp;
2586 	struct vattr vattr;
2587 	struct vnode *vp;
2588 	struct file *fp;
2589 	int error;
2590 
2591 	if (uap->length < 0)
2592 		return(EINVAL);
2593 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2594 		return (error);
2595 	if ((fp->f_flag & FWRITE) == 0)
2596 		return (EINVAL);
2597 	vp = (struct vnode *)fp->f_data;
2598 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2599 		return (error);
2600 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2601 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2602 	if (vp->v_type == VDIR)
2603 		error = EISDIR;
2604 	else if ((error = vn_writechk(vp)) == 0) {
2605 		VATTR_NULL(&vattr);
2606 		vattr.va_size = SCARG(uap, length);
2607 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2608 	}
2609 	VOP_UNLOCK(vp, 0, p);
2610 	vn_finished_write(mp);
2611 	return (error);
2612 }
2613 
2614 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
2615 /*
2616  * Truncate a file given its path name.
2617  */
2618 #ifndef _SYS_SYSPROTO_H_
2619 struct otruncate_args {
2620 	char	*path;
2621 	long	length;
2622 };
2623 #endif
2624 /* ARGSUSED */
2625 int
2626 otruncate(p, uap)
2627 	struct proc *p;
2628 	register struct otruncate_args /* {
2629 		syscallarg(char *) path;
2630 		syscallarg(long) length;
2631 	} */ *uap;
2632 {
2633 	struct truncate_args /* {
2634 		syscallarg(char *) path;
2635 		syscallarg(int) pad;
2636 		syscallarg(off_t) length;
2637 	} */ nuap;
2638 
2639 	SCARG(&nuap, path) = SCARG(uap, path);
2640 	SCARG(&nuap, length) = SCARG(uap, length);
2641 	return (truncate(p, &nuap));
2642 }
2643 
2644 /*
2645  * Truncate a file given a file descriptor.
2646  */
2647 #ifndef _SYS_SYSPROTO_H_
2648 struct oftruncate_args {
2649 	int	fd;
2650 	long	length;
2651 };
2652 #endif
2653 /* ARGSUSED */
2654 int
2655 oftruncate(p, uap)
2656 	struct proc *p;
2657 	register struct oftruncate_args /* {
2658 		syscallarg(int) fd;
2659 		syscallarg(long) length;
2660 	} */ *uap;
2661 {
2662 	struct ftruncate_args /* {
2663 		syscallarg(int) fd;
2664 		syscallarg(int) pad;
2665 		syscallarg(off_t) length;
2666 	} */ nuap;
2667 
2668 	SCARG(&nuap, fd) = SCARG(uap, fd);
2669 	SCARG(&nuap, length) = SCARG(uap, length);
2670 	return (ftruncate(p, &nuap));
2671 }
2672 #endif /* COMPAT_43 || COMPAT_SUNOS */
2673 
2674 /*
2675  * Sync an open file.
2676  */
2677 #ifndef _SYS_SYSPROTO_H_
2678 struct fsync_args {
2679 	int	fd;
2680 };
2681 #endif
2682 /* ARGSUSED */
2683 int
2684 fsync(p, uap)
2685 	struct proc *p;
2686 	struct fsync_args /* {
2687 		syscallarg(int) fd;
2688 	} */ *uap;
2689 {
2690 	struct vnode *vp;
2691 	struct mount *mp;
2692 	struct file *fp;
2693 	vm_object_t obj;
2694 	int error;
2695 
2696 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2697 		return (error);
2698 	vp = (struct vnode *)fp->f_data;
2699 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2700 		return (error);
2701 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2702 	if (VOP_GETVOBJECT(vp, &obj) == 0)
2703 		vm_object_page_clean(obj, 0, 0, 0);
2704 	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
2705 #ifdef SOFTUPDATES
2706 	if (error == 0 && vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2707 	    error = softdep_fsync(vp);
2708 #endif
2709 
2710 	VOP_UNLOCK(vp, 0, p);
2711 	vn_finished_write(mp);
2712 	return (error);
2713 }
2714 
2715 /*
2716  * Rename files.  Source and destination must either both be directories,
2717  * or both not be directories.  If target is a directory, it must be empty.
2718  */
2719 #ifndef _SYS_SYSPROTO_H_
2720 struct rename_args {
2721 	char	*from;
2722 	char	*to;
2723 };
2724 #endif
2725 /* ARGSUSED */
2726 int
2727 rename(p, uap)
2728 	struct proc *p;
2729 	register struct rename_args /* {
2730 		syscallarg(char *) from;
2731 		syscallarg(char *) to;
2732 	} */ *uap;
2733 {
2734 	struct mount *mp;
2735 	struct vnode *tvp, *fvp, *tdvp;
2736 	struct nameidata fromnd, tond;
2737 	int error;
2738 
2739 	bwillwrite();
2740 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2741 	    SCARG(uap, from), p);
2742 	if ((error = namei(&fromnd)) != 0)
2743 		return (error);
2744 	fvp = fromnd.ni_vp;
2745 	if ((error = vn_start_write(fvp, &mp, V_WAIT | PCATCH)) != 0) {
2746 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2747 		vrele(fromnd.ni_dvp);
2748 		vrele(fvp);
2749 		goto out1;
2750 	}
2751 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | NOOBJ,
2752 	    UIO_USERSPACE, SCARG(uap, to), p);
2753 	if (fromnd.ni_vp->v_type == VDIR)
2754 		tond.ni_cnd.cn_flags |= WILLBEDIR;
2755 	if ((error = namei(&tond)) != 0) {
2756 		/* Translate error code for rename("dir1", "dir2/."). */
2757 		if (error == EISDIR && fvp->v_type == VDIR)
2758 			error = EINVAL;
2759 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2760 		vrele(fromnd.ni_dvp);
2761 		vrele(fvp);
2762 		goto out1;
2763 	}
2764 	tdvp = tond.ni_dvp;
2765 	tvp = tond.ni_vp;
2766 	if (tvp != NULL) {
2767 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2768 			error = ENOTDIR;
2769 			goto out;
2770 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2771 			error = EISDIR;
2772 			goto out;
2773 		}
2774 	}
2775 	if (fvp == tdvp)
2776 		error = EINVAL;
2777 	/*
2778 	 * If source is the same as the destination (that is the
2779 	 * same inode number with the same name in the same directory),
2780 	 * then there is nothing to do.
2781 	 */
2782 	if (fvp == tvp && fromnd.ni_dvp == tdvp &&
2783 	    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2784 	    !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
2785 	      fromnd.ni_cnd.cn_namelen))
2786 		error = -1;
2787 out:
2788 	if (!error) {
2789 		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2790 		if (fromnd.ni_dvp != tdvp) {
2791 			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2792 		}
2793 		if (tvp) {
2794 			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2795 		}
2796 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2797 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2798 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2799 		NDFREE(&tond, NDF_ONLY_PNBUF);
2800 	} else {
2801 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
2802 		NDFREE(&tond, NDF_ONLY_PNBUF);
2803 		if (tdvp == tvp)
2804 			vrele(tdvp);
2805 		else
2806 			vput(tdvp);
2807 		if (tvp)
2808 			vput(tvp);
2809 		vrele(fromnd.ni_dvp);
2810 		vrele(fvp);
2811 	}
2812 	vrele(tond.ni_startdir);
2813 	vn_finished_write(mp);
2814 	ASSERT_VOP_UNLOCKED(fromnd.ni_dvp, "rename");
2815 	ASSERT_VOP_UNLOCKED(fromnd.ni_vp, "rename");
2816 	ASSERT_VOP_UNLOCKED(tond.ni_dvp, "rename");
2817 	ASSERT_VOP_UNLOCKED(tond.ni_vp, "rename");
2818 out1:
2819 	if (fromnd.ni_startdir)
2820 		vrele(fromnd.ni_startdir);
2821 	if (error == -1)
2822 		return (0);
2823 	return (error);
2824 }
2825 
2826 /*
2827  * Make a directory file.
2828  */
2829 #ifndef _SYS_SYSPROTO_H_
2830 struct mkdir_args {
2831 	char	*path;
2832 	int	mode;
2833 };
2834 #endif
2835 /* ARGSUSED */
2836 int
2837 mkdir(p, uap)
2838 	struct proc *p;
2839 	register struct mkdir_args /* {
2840 		syscallarg(char *) path;
2841 		syscallarg(int) mode;
2842 	} */ *uap;
2843 {
2844 	struct mount *mp;
2845 	struct vnode *vp;
2846 	struct vattr vattr;
2847 	int error;
2848 	struct nameidata nd;
2849 
2850 restart:
2851 	bwillwrite();
2852 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2853 	nd.ni_cnd.cn_flags |= WILLBEDIR;
2854 	if ((error = namei(&nd)) != 0)
2855 		return (error);
2856 	vp = nd.ni_vp;
2857 	if (vp != NULL) {
2858 		NDFREE(&nd, NDF_ONLY_PNBUF);
2859 		vrele(vp);
2860 		vput(nd.ni_dvp);
2861 		return (EEXIST);
2862 	}
2863 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
2864 		NDFREE(&nd, NDF_ONLY_PNBUF);
2865 		vput(nd.ni_dvp);
2866 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
2867 			return (error);
2868 		goto restart;
2869 	}
2870 	VATTR_NULL(&vattr);
2871 	vattr.va_type = VDIR;
2872 	vattr.va_mode = (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2873 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2874 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2875 	NDFREE(&nd, NDF_ONLY_PNBUF);
2876 	vput(nd.ni_dvp);
2877 	if (!error)
2878 		vput(nd.ni_vp);
2879 	vn_finished_write(mp);
2880 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mkdir");
2881 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "mkdir");
2882 	return (error);
2883 }
2884 
2885 /*
2886  * Remove a directory file.
2887  */
2888 #ifndef _SYS_SYSPROTO_H_
2889 struct rmdir_args {
2890 	char	*path;
2891 };
2892 #endif
2893 /* ARGSUSED */
2894 int
2895 rmdir(p, uap)
2896 	struct proc *p;
2897 	struct rmdir_args /* {
2898 		syscallarg(char *) path;
2899 	} */ *uap;
2900 {
2901 	struct mount *mp;
2902 	struct vnode *vp;
2903 	int error;
2904 	struct nameidata nd;
2905 
2906 restart:
2907 	bwillwrite();
2908 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2909 	    SCARG(uap, path), p);
2910 	if ((error = namei(&nd)) != 0)
2911 		return (error);
2912 	vp = nd.ni_vp;
2913 	if (vp->v_type != VDIR) {
2914 		error = ENOTDIR;
2915 		goto out;
2916 	}
2917 	/*
2918 	 * No rmdir "." please.
2919 	 */
2920 	if (nd.ni_dvp == vp) {
2921 		error = EINVAL;
2922 		goto out;
2923 	}
2924 	/*
2925 	 * The root of a mounted filesystem cannot be deleted.
2926 	 */
2927 	if (vp->v_flag & VROOT) {
2928 		error = EBUSY;
2929 		goto out;
2930 	}
2931 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
2932 		NDFREE(&nd, NDF_ONLY_PNBUF);
2933 		if (nd.ni_dvp == vp)
2934 			vrele(nd.ni_dvp);
2935 		else
2936 			vput(nd.ni_dvp);
2937 		vput(vp);
2938 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
2939 			return (error);
2940 		goto restart;
2941 	}
2942 	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2943 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2944 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2945 	vn_finished_write(mp);
2946 out:
2947 	NDFREE(&nd, NDF_ONLY_PNBUF);
2948 	if (nd.ni_dvp == vp)
2949 		vrele(nd.ni_dvp);
2950 	else
2951 		vput(nd.ni_dvp);
2952 	vput(vp);
2953 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "rmdir");
2954 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "rmdir");
2955 	return (error);
2956 }
2957 
2958 #ifdef COMPAT_43
2959 /*
2960  * Read a block of directory entries in a file system independent format.
2961  */
2962 #ifndef _SYS_SYSPROTO_H_
2963 struct ogetdirentries_args {
2964 	int	fd;
2965 	char	*buf;
2966 	u_int	count;
2967 	long	*basep;
2968 };
2969 #endif
2970 int
2971 ogetdirentries(p, uap)
2972 	struct proc *p;
2973 	register struct ogetdirentries_args /* {
2974 		syscallarg(int) fd;
2975 		syscallarg(char *) buf;
2976 		syscallarg(u_int) count;
2977 		syscallarg(long *) basep;
2978 	} */ *uap;
2979 {
2980 	struct vnode *vp;
2981 	struct file *fp;
2982 	struct uio auio, kuio;
2983 	struct iovec aiov, kiov;
2984 	struct dirent *dp, *edp;
2985 	caddr_t dirbuf;
2986 	int error, eofflag, readcnt;
2987 	long loff;
2988 
2989 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2990 		return (error);
2991 	if ((fp->f_flag & FREAD) == 0)
2992 		return (EBADF);
2993 	vp = (struct vnode *)fp->f_data;
2994 unionread:
2995 	if (vp->v_type != VDIR)
2996 		return (EINVAL);
2997 	aiov.iov_base = SCARG(uap, buf);
2998 	aiov.iov_len = SCARG(uap, count);
2999 	auio.uio_iov = &aiov;
3000 	auio.uio_iovcnt = 1;
3001 	auio.uio_rw = UIO_READ;
3002 	auio.uio_segflg = UIO_USERSPACE;
3003 	auio.uio_procp = p;
3004 	auio.uio_resid = SCARG(uap, count);
3005 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3006 	loff = auio.uio_offset = fp->f_offset;
3007 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3008 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3009 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3010 			    NULL, NULL);
3011 			fp->f_offset = auio.uio_offset;
3012 		} else
3013 #	endif
3014 	{
3015 		kuio = auio;
3016 		kuio.uio_iov = &kiov;
3017 		kuio.uio_segflg = UIO_SYSSPACE;
3018 		kiov.iov_len = SCARG(uap, count);
3019 		MALLOC(dirbuf, caddr_t, SCARG(uap, count), M_TEMP, M_WAITOK);
3020 		kiov.iov_base = dirbuf;
3021 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3022 			    NULL, NULL);
3023 		fp->f_offset = kuio.uio_offset;
3024 		if (error == 0) {
3025 			readcnt = SCARG(uap, count) - kuio.uio_resid;
3026 			edp = (struct dirent *)&dirbuf[readcnt];
3027 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3028 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3029 					/*
3030 					 * The expected low byte of
3031 					 * dp->d_namlen is our dp->d_type.
3032 					 * The high MBZ byte of dp->d_namlen
3033 					 * is our dp->d_namlen.
3034 					 */
3035 					dp->d_type = dp->d_namlen;
3036 					dp->d_namlen = 0;
3037 #				else
3038 					/*
3039 					 * The dp->d_type is the high byte
3040 					 * of the expected dp->d_namlen,
3041 					 * so must be zero'ed.
3042 					 */
3043 					dp->d_type = 0;
3044 #				endif
3045 				if (dp->d_reclen > 0) {
3046 					dp = (struct dirent *)
3047 					    ((char *)dp + dp->d_reclen);
3048 				} else {
3049 					error = EIO;
3050 					break;
3051 				}
3052 			}
3053 			if (dp >= edp)
3054 				error = uiomove(dirbuf, readcnt, &auio);
3055 		}
3056 		FREE(dirbuf, M_TEMP);
3057 	}
3058 	VOP_UNLOCK(vp, 0, p);
3059 	if (error)
3060 		return (error);
3061 	if (SCARG(uap, count) == auio.uio_resid) {
3062 		if (union_dircheckp) {
3063 			error = union_dircheckp(p, &vp, fp);
3064 			if (error == -1)
3065 				goto unionread;
3066 			if (error)
3067 				return (error);
3068 		}
3069 		if ((vp->v_flag & VROOT) &&
3070 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
3071 			struct vnode *tvp = vp;
3072 			vp = vp->v_mount->mnt_vnodecovered;
3073 			VREF(vp);
3074 			fp->f_data = (caddr_t) vp;
3075 			fp->f_offset = 0;
3076 			vrele(tvp);
3077 			goto unionread;
3078 		}
3079 	}
3080 	error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
3081 	    sizeof(long));
3082 	p->p_retval[0] = SCARG(uap, count) - auio.uio_resid;
3083 	return (error);
3084 }
3085 #endif /* COMPAT_43 */
3086 
3087 /*
3088  * Read a block of directory entries in a file system independent format.
3089  */
3090 #ifndef _SYS_SYSPROTO_H_
3091 struct getdirentries_args {
3092 	int	fd;
3093 	char	*buf;
3094 	u_int	count;
3095 	long	*basep;
3096 };
3097 #endif
3098 int
3099 getdirentries(p, uap)
3100 	struct proc *p;
3101 	register struct getdirentries_args /* {
3102 		syscallarg(int) fd;
3103 		syscallarg(char *) buf;
3104 		syscallarg(u_int) count;
3105 		syscallarg(long *) basep;
3106 	} */ *uap;
3107 {
3108 	struct vnode *vp;
3109 	struct file *fp;
3110 	struct uio auio;
3111 	struct iovec aiov;
3112 	long loff;
3113 	int error, eofflag;
3114 
3115 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3116 		return (error);
3117 	if ((fp->f_flag & FREAD) == 0)
3118 		return (EBADF);
3119 	vp = (struct vnode *)fp->f_data;
3120 unionread:
3121 	if (vp->v_type != VDIR)
3122 		return (EINVAL);
3123 	aiov.iov_base = SCARG(uap, buf);
3124 	aiov.iov_len = SCARG(uap, count);
3125 	auio.uio_iov = &aiov;
3126 	auio.uio_iovcnt = 1;
3127 	auio.uio_rw = UIO_READ;
3128 	auio.uio_segflg = UIO_USERSPACE;
3129 	auio.uio_procp = p;
3130 	auio.uio_resid = SCARG(uap, count);
3131 	/* vn_lock(vp, LK_SHARED | LK_RETRY, p); */
3132 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3133 	loff = auio.uio_offset = fp->f_offset;
3134 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
3135 	fp->f_offset = auio.uio_offset;
3136 	VOP_UNLOCK(vp, 0, p);
3137 	if (error)
3138 		return (error);
3139 	if (SCARG(uap, count) == auio.uio_resid) {
3140 		if (union_dircheckp) {
3141 			error = union_dircheckp(p, &vp, fp);
3142 			if (error == -1)
3143 				goto unionread;
3144 			if (error)
3145 				return (error);
3146 		}
3147 		if ((vp->v_flag & VROOT) &&
3148 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
3149 			struct vnode *tvp = vp;
3150 			vp = vp->v_mount->mnt_vnodecovered;
3151 			VREF(vp);
3152 			fp->f_data = (caddr_t) vp;
3153 			fp->f_offset = 0;
3154 			vrele(tvp);
3155 			goto unionread;
3156 		}
3157 	}
3158 	if (SCARG(uap, basep) != NULL) {
3159 		error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
3160 		    sizeof(long));
3161 	}
3162 	p->p_retval[0] = SCARG(uap, count) - auio.uio_resid;
3163 	return (error);
3164 }
3165 #ifndef _SYS_SYSPROTO_H_
3166 struct getdents_args {
3167 	int fd;
3168 	char *buf;
3169 	size_t count;
3170 };
3171 #endif
3172 int
3173 getdents(p, uap)
3174 	struct proc *p;
3175 	register struct getdents_args /* {
3176 		syscallarg(int) fd;
3177 		syscallarg(char *) buf;
3178 		syscallarg(u_int) count;
3179 	} */ *uap;
3180 {
3181 	struct getdirentries_args ap;
3182 	ap.fd = uap->fd;
3183 	ap.buf = uap->buf;
3184 	ap.count = uap->count;
3185 	ap.basep = NULL;
3186 	return getdirentries(p, &ap);
3187 }
3188 
3189 /*
3190  * Set the mode mask for creation of filesystem nodes.
3191  *
3192  * MP SAFE
3193  */
3194 #ifndef _SYS_SYSPROTO_H_
3195 struct umask_args {
3196 	int	newmask;
3197 };
3198 #endif
3199 int
3200 umask(p, uap)
3201 	struct proc *p;
3202 	struct umask_args /* {
3203 		syscallarg(int) newmask;
3204 	} */ *uap;
3205 {
3206 	register struct filedesc *fdp;
3207 
3208 	fdp = p->p_fd;
3209 	p->p_retval[0] = fdp->fd_cmask;
3210 	fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
3211 	return (0);
3212 }
3213 
3214 /*
3215  * Void all references to file by ripping underlying filesystem
3216  * away from vnode.
3217  */
3218 #ifndef _SYS_SYSPROTO_H_
3219 struct revoke_args {
3220 	char	*path;
3221 };
3222 #endif
3223 /* ARGSUSED */
3224 int
3225 revoke(p, uap)
3226 	struct proc *p;
3227 	register struct revoke_args /* {
3228 		syscallarg(char *) path;
3229 	} */ *uap;
3230 {
3231 	struct mount *mp;
3232 	struct vnode *vp;
3233 	struct vattr vattr;
3234 	int error;
3235 	struct nameidata nd;
3236 
3237 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3238 	if ((error = namei(&nd)) != 0)
3239 		return (error);
3240 	vp = nd.ni_vp;
3241 	NDFREE(&nd, NDF_ONLY_PNBUF);
3242 	if (vp->v_type != VCHR) {
3243 		error = EINVAL;
3244 		goto out;
3245 	}
3246 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3247 		goto out;
3248 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3249 	    (error = suser_xxx(0, p, PRISON_ROOT)))
3250 		goto out;
3251 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3252 		goto out;
3253 	if (vcount(vp) > 1)
3254 		VOP_REVOKE(vp, REVOKEALL);
3255 	vn_finished_write(mp);
3256 out:
3257 	vrele(vp);
3258 	return (error);
3259 }
3260 
3261 /*
3262  * Convert a user file descriptor to a kernel file entry.
3263  */
3264 int
3265 getvnode(fdp, fd, fpp)
3266 	struct filedesc *fdp;
3267 	int fd;
3268 	struct file **fpp;
3269 {
3270 	struct file *fp;
3271 
3272 	if ((u_int)fd >= fdp->fd_nfiles ||
3273 	    (fp = fdp->fd_ofiles[fd]) == NULL)
3274 		return (EBADF);
3275 	if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO)
3276 		return (EINVAL);
3277 	*fpp = fp;
3278 	return (0);
3279 }
3280 /*
3281  * Get (NFS) file handle
3282  */
3283 #ifndef _SYS_SYSPROTO_H_
3284 struct getfh_args {
3285 	char	*fname;
3286 	fhandle_t *fhp;
3287 };
3288 #endif
3289 int
3290 getfh(p, uap)
3291 	struct proc *p;
3292 	register struct getfh_args *uap;
3293 {
3294 	struct nameidata nd;
3295 	fhandle_t fh;
3296 	register struct vnode *vp;
3297 	int error;
3298 
3299 	/*
3300 	 * Must be super user
3301 	 */
3302 	error = suser(p);
3303 	if (error)
3304 		return (error);
3305 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, p);
3306 	error = namei(&nd);
3307 	if (error)
3308 		return (error);
3309 	NDFREE(&nd, NDF_ONLY_PNBUF);
3310 	vp = nd.ni_vp;
3311 	bzero(&fh, sizeof(fh));
3312 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3313 	error = VFS_VPTOFH(vp, &fh.fh_fid);
3314 	vput(vp);
3315 	if (error)
3316 		return (error);
3317 	error = copyout(&fh, uap->fhp, sizeof (fh));
3318 	return (error);
3319 }
3320 
3321 /*
3322  * syscall for the rpc.lockd to use to translate a NFS file handle into
3323  * an open descriptor.
3324  *
3325  * warning: do not remove the suser() call or this becomes one giant
3326  * security hole.
3327  */
3328 #ifndef _SYS_SYSPROTO_H_
3329 struct fhopen_args {
3330 	const struct fhandle *u_fhp;
3331 	int flags;
3332 };
3333 #endif
3334 int
3335 fhopen(p, uap)
3336 	struct proc *p;
3337 	struct fhopen_args /* {
3338 		syscallarg(const struct fhandle *) u_fhp;
3339 		syscallarg(int) flags;
3340 	} */ *uap;
3341 {
3342 	struct mount *mp;
3343 	struct vnode *vp;
3344 	struct fhandle fhp;
3345 	struct vattr vat;
3346 	struct vattr *vap = &vat;
3347 	struct flock lf;
3348 	struct file *fp;
3349 	register struct filedesc *fdp = p->p_fd;
3350 	int fmode, mode, error, type;
3351 	struct file *nfp;
3352 	int indx;
3353 
3354 	/*
3355 	 * Must be super user
3356 	 */
3357 	error = suser(p);
3358 	if (error)
3359 		return (error);
3360 
3361 	fmode = FFLAGS(SCARG(uap, flags));
3362 	/* why not allow a non-read/write open for our lockd? */
3363 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3364 		return (EINVAL);
3365 	error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp));
3366 	if (error)
3367 		return(error);
3368 	/* find the mount point */
3369 	mp = vfs_getvfs(&fhp.fh_fsid);
3370 	if (mp == NULL)
3371 		return (ESTALE);
3372 	/* now give me my vnode, it gets returned to me locked */
3373 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3374 	if (error)
3375 		return (error);
3376  	/*
3377 	 * from now on we have to make sure not
3378 	 * to forget about the vnode
3379 	 * any error that causes an abort must vput(vp)
3380 	 * just set error = err and 'goto bad;'.
3381 	 */
3382 
3383 	/*
3384 	 * from vn_open
3385 	 */
3386 	if (vp->v_type == VLNK) {
3387 		error = EMLINK;
3388 		goto bad;
3389 	}
3390 	if (vp->v_type == VSOCK) {
3391 		error = EOPNOTSUPP;
3392 		goto bad;
3393 	}
3394 	mode = 0;
3395 	if (fmode & (FWRITE | O_TRUNC)) {
3396 		if (vp->v_type == VDIR) {
3397 			error = EISDIR;
3398 			goto bad;
3399 		}
3400 		error = vn_writechk(vp);
3401 		if (error)
3402 			goto bad;
3403 		mode |= VWRITE;
3404 	}
3405 	if (fmode & FREAD)
3406 		mode |= VREAD;
3407 	if (mode) {
3408 		error = VOP_ACCESS(vp, mode, p->p_ucred, p);
3409 		if (error)
3410 			goto bad;
3411 	}
3412 	if (fmode & O_TRUNC) {
3413 		VOP_UNLOCK(vp, 0, p);				/* XXX */
3414 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
3415 			vrele(vp);
3416 			return (error);
3417 		}
3418 		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3419 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);	/* XXX */
3420 		VATTR_NULL(vap);
3421 		vap->va_size = 0;
3422 		error = VOP_SETATTR(vp, vap, p->p_ucred, p);
3423 		vn_finished_write(mp);
3424 		if (error)
3425 			goto bad;
3426 	}
3427 	error = VOP_OPEN(vp, fmode, p->p_ucred, p);
3428 	if (error)
3429 		goto bad;
3430 	/*
3431 	 * Make sure that a VM object is created for VMIO support.
3432 	 */
3433 	if (vn_canvmio(vp) == TRUE) {
3434 		if ((error = vfs_object_create(vp, p, p->p_ucred)) != 0)
3435 			goto bad;
3436 	}
3437 	if (fmode & FWRITE)
3438 		vp->v_writecount++;
3439 
3440 	/*
3441 	 * end of vn_open code
3442 	 */
3443 
3444 	if ((error = falloc(p, &nfp, &indx)) != 0)
3445 		goto bad;
3446 	fp = nfp;
3447 
3448 	/*
3449 	 * Hold an extra reference to avoid having fp ripped out
3450 	 * from under us while we block in the lock op
3451 	 */
3452 	fhold(fp);
3453 	nfp->f_data = (caddr_t)vp;
3454 	nfp->f_flag = fmode & FMASK;
3455 	nfp->f_ops = &vnops;
3456 	nfp->f_type = DTYPE_VNODE;
3457 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
3458 		lf.l_whence = SEEK_SET;
3459 		lf.l_start = 0;
3460 		lf.l_len = 0;
3461 		if (fmode & O_EXLOCK)
3462 			lf.l_type = F_WRLCK;
3463 		else
3464 			lf.l_type = F_RDLCK;
3465 		type = F_FLOCK;
3466 		if ((fmode & FNONBLOCK) == 0)
3467 			type |= F_WAIT;
3468 		VOP_UNLOCK(vp, 0, p);
3469 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3470 			/*
3471 			 * The lock request failed.  Normally close the
3472 			 * descriptor but handle the case where someone might
3473 			 * have dup()d or close()d it when we weren't looking.
3474 			 */
3475 			if (fdp->fd_ofiles[indx] == fp) {
3476 				fdp->fd_ofiles[indx] = NULL;
3477 				fdrop(fp, p);
3478 			}
3479 			/*
3480 			 * release our private reference
3481 			 */
3482 			fdrop(fp, p);
3483 			return(error);
3484 		}
3485 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3486 		fp->f_flag |= FHASLOCK;
3487 	}
3488 	if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
3489 		vfs_object_create(vp, p, p->p_ucred);
3490 
3491 	VOP_UNLOCK(vp, 0, p);
3492 	fdrop(fp, p);
3493 	p->p_retval[0] = indx;
3494 	return (0);
3495 
3496 bad:
3497 	vput(vp);
3498 	return (error);
3499 }
3500 
3501 /*
3502  * Stat an (NFS) file handle.
3503  */
3504 #ifndef _SYS_SYSPROTO_H_
3505 struct fhstat_args {
3506 	struct fhandle *u_fhp;
3507 	struct stat *sb;
3508 };
3509 #endif
3510 int
3511 fhstat(p, uap)
3512 	struct proc *p;
3513 	register struct fhstat_args /* {
3514 		syscallarg(struct fhandle *) u_fhp;
3515 		syscallarg(struct stat *) sb;
3516 	} */ *uap;
3517 {
3518 	struct stat sb;
3519 	fhandle_t fh;
3520 	struct mount *mp;
3521 	struct vnode *vp;
3522 	int error;
3523 
3524 	/*
3525 	 * Must be super user
3526 	 */
3527 	error = suser(p);
3528 	if (error)
3529 		return (error);
3530 
3531 	error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t));
3532 	if (error)
3533 		return (error);
3534 
3535 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3536 		return (ESTALE);
3537 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3538 		return (error);
3539 	error = vn_stat(vp, &sb, p);
3540 	vput(vp);
3541 	if (error)
3542 		return (error);
3543 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
3544 	return (error);
3545 }
3546 
3547 /*
3548  * Implement fstatfs() for (NFS) file handles.
3549  */
3550 #ifndef _SYS_SYSPROTO_H_
3551 struct fhstatfs_args {
3552 	struct fhandle *u_fhp;
3553 	struct statfs *buf;
3554 };
3555 #endif
3556 int
3557 fhstatfs(p, uap)
3558 	struct proc *p;
3559 	struct fhstatfs_args /* {
3560 		syscallarg(struct fhandle) *u_fhp;
3561 		syscallarg(struct statfs) *buf;
3562 	} */ *uap;
3563 {
3564 	struct statfs *sp;
3565 	struct mount *mp;
3566 	struct vnode *vp;
3567 	struct statfs sb;
3568 	fhandle_t fh;
3569 	int error;
3570 
3571 	/*
3572 	 * Must be super user
3573 	 */
3574 	if ((error = suser(p)))
3575 		return (error);
3576 
3577 	if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0)
3578 		return (error);
3579 
3580 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3581 		return (ESTALE);
3582 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3583 		return (error);
3584 	mp = vp->v_mount;
3585 	sp = &mp->mnt_stat;
3586 	vput(vp);
3587 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
3588 		return (error);
3589 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3590 	if (suser_xxx(p->p_ucred, 0, 0)) {
3591 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
3592 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3593 		sp = &sb;
3594 	}
3595 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
3596 }
3597 
3598 /*
3599  * Syscall to push extended attribute configuration information into the
3600  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
3601  * a command (int cmd), and attribute name and misc data.  For now, the
3602  * attribute name is left in userspace for consumption by the VFS_op.
3603  * It will probably be changed to be copied into sysspace by the
3604  * syscall in the future, once issues with various consumers of the
3605  * attribute code have raised their hands.
3606  *
3607  * Currently this is used only by UFS Extended Attributes.
3608  */
3609 int
3610 extattrctl(p, uap)
3611 	struct proc *p;
3612 	struct extattrctl_args *uap;
3613 {
3614 	struct nameidata nd;
3615 	struct mount *mp;
3616 	int error;
3617 
3618 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3619 	if ((error = namei(&nd)) != 0)
3620 		return (error);
3621 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH);
3622 	NDFREE(&nd, 0);
3623 	if (error)
3624 		return (error);
3625 	error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), SCARG(uap, attrname),
3626 	    SCARG(uap, arg), p);
3627 	vn_finished_write(mp);
3628 	return (error);
3629 }
3630 
3631 /*
3632  * Syscall to set a named extended attribute on a file or directory.
3633  * Accepts attribute name, and a uio structure pointing to the data to set.
3634  * The uio is consumed in the style of writev().  The real work happens
3635  * in VOP_SETEXTATTR().
3636  */
3637 int
3638 extattr_set_file(p, uap)
3639 	struct proc *p;
3640 	struct extattr_set_file_args *uap;
3641 {
3642 	struct nameidata nd;
3643 	struct mount *mp;
3644 	struct uio auio;
3645 	struct iovec *iov, *needfree = NULL, aiov[UIO_SMALLIOV];
3646 	char attrname[EXTATTR_MAXNAMELEN];
3647 	u_int iovlen, cnt;
3648 	int error, i;
3649 
3650 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3651 	if (error)
3652 		return (error);
3653 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3654 	    SCARG(uap, path), p);
3655 	if ((error = namei(&nd)) != 0)
3656 		return(error);
3657 	if ((error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH)) != 0) {
3658 		NDFREE(&nd, 0);
3659 		return (error);
3660 	}
3661 	iovlen = uap->iovcnt * sizeof(struct iovec);
3662 	if (uap->iovcnt > UIO_SMALLIOV) {
3663 		if (uap->iovcnt > UIO_MAXIOV) {
3664 			error = EINVAL;
3665 			goto done;
3666 		}
3667 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3668 		needfree = iov;
3669 	} else
3670 		iov = aiov;
3671 	auio.uio_iov = iov;
3672 	auio.uio_iovcnt = uap->iovcnt;
3673 	auio.uio_rw = UIO_WRITE;
3674 	auio.uio_segflg = UIO_USERSPACE;
3675 	auio.uio_procp = p;
3676 	auio.uio_offset = 0;
3677 	if ((error = copyin((caddr_t)uap->iovp, (caddr_t)iov, iovlen)))
3678 		goto done;
3679 	auio.uio_resid = 0;
3680 	for (i = 0; i < uap->iovcnt; i++) {
3681 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
3682 			error = EINVAL;
3683 			goto done;
3684 		}
3685 		auio.uio_resid += iov->iov_len;
3686 		iov++;
3687 	}
3688 	cnt = auio.uio_resid;
3689 	error = VOP_SETEXTATTR(nd.ni_vp, attrname, &auio, p->p_cred->pc_ucred,
3690 	    p);
3691 	cnt -= auio.uio_resid;
3692 	p->p_retval[0] = cnt;
3693 done:
3694 	if (needfree)
3695 		FREE(needfree, M_IOV);
3696 	NDFREE(&nd, 0);
3697 	vn_finished_write(mp);
3698 	return (error);
3699 }
3700 
3701 /*
3702  * Syscall to get a named extended attribute on a file or directory.
3703  * Accepts attribute name, and a uio structure pointing to a buffer for the
3704  * data.  The uio is consumed in the style of readv().  The real work
3705  * happens in VOP_GETEXTATTR();
3706  */
3707 int
3708 extattr_get_file(p, uap)
3709 	struct proc *p;
3710 	struct extattr_get_file_args *uap;
3711 {
3712 	struct nameidata nd;
3713 	struct uio auio;
3714 	struct iovec *iov, *needfree, aiov[UIO_SMALLIOV];
3715 	char attrname[EXTATTR_MAXNAMELEN];
3716 	u_int iovlen, cnt;
3717 	int error, i;
3718 
3719 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3720 	if (error)
3721 		return (error);
3722 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3723 	    SCARG(uap, path), p);
3724 	if ((error = namei(&nd)) != 0)
3725 		return (error);
3726 	iovlen = uap->iovcnt * sizeof (struct iovec);
3727 	if (uap->iovcnt > UIO_SMALLIOV) {
3728 		if (uap->iovcnt > UIO_MAXIOV) {
3729 			NDFREE(&nd, 0);
3730 			return (EINVAL);
3731 		}
3732 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3733 		needfree = iov;
3734 	} else {
3735 		iov = aiov;
3736 		needfree = NULL;
3737 	}
3738 	auio.uio_iov = iov;
3739 	auio.uio_iovcnt = uap->iovcnt;
3740 	auio.uio_rw = UIO_READ;
3741 	auio.uio_segflg = UIO_USERSPACE;
3742 	auio.uio_procp = p;
3743 	auio.uio_offset = 0;
3744 	if ((error = copyin((caddr_t)uap->iovp, (caddr_t)iov, iovlen)))
3745 		goto done;
3746 	auio.uio_resid = 0;
3747 	for (i = 0; i < uap->iovcnt; i++) {
3748 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
3749 			error = EINVAL;
3750 			goto done;
3751 		}
3752 		auio.uio_resid += iov->iov_len;
3753 		iov++;
3754 	}
3755 	cnt = auio.uio_resid;
3756 	error = VOP_GETEXTATTR(nd.ni_vp, attrname, &auio, p->p_cred->pc_ucred,
3757 	    p);
3758 	cnt -= auio.uio_resid;
3759 	p->p_retval[0] = cnt;
3760 done:
3761 	if (needfree)
3762 		FREE(needfree, M_IOV);
3763 	NDFREE(&nd, 0);
3764 	return(error);
3765 }
3766 
3767 /*
3768  * Syscall to delete a named extended attribute from a file or directory.
3769  * Accepts attribute name.  The real work happens in VOP_SETEXTATTR().
3770  */
3771 int
3772 extattr_delete_file(p, uap)
3773 	struct proc *p;
3774 	struct extattr_delete_file_args *uap;
3775 {
3776 	struct mount *mp;
3777 	struct nameidata nd;
3778 	char attrname[EXTATTR_MAXNAMELEN];
3779 	int	error;
3780 
3781 	error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3782 	if (error)
3783 		return(error);
3784 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3785 	    SCARG(uap, path), p);
3786 	if ((error = namei(&nd)) != 0)
3787 		return(error);
3788 	if ((error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH)) != 0) {
3789 		NDFREE(&nd, 0);
3790 		return (error);
3791 	}
3792 	error = VOP_SETEXTATTR(nd.ni_vp, attrname, NULL, p->p_cred->pc_ucred,
3793 	    p);
3794 	NDFREE(&nd, 0);
3795 	vn_finished_write(mp);
3796 	return(error);
3797 }
3798