xref: /freebsd/sys/kern/vfs_extattr.c (revision 70fe064ad7cab6c0444b91622f60ec6a462f308a)
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/sx.h>
63 #include <sys/unistd.h>
64 #include <sys/vnode.h>
65 #include <sys/proc.h>
66 #include <sys/dirent.h>
67 #include <sys/extattr.h>
68 #include <sys/jail.h>
69 #include <sys/sysctl.h>
70 
71 #include <machine/limits.h>
72 
73 #include <vm/vm.h>
74 #include <vm/vm_object.h>
75 #include <vm/vm_zone.h>
76 #include <vm/vm_page.h>
77 
78 static int change_dir __P((struct nameidata *ndp, struct thread *td));
79 static void checkdirs __P((struct vnode *olddp, struct vnode *newdp));
80 static int chroot_refuse_vdir_fds __P((struct filedesc *fdp));
81 static int getutimes __P((const struct timeval *, struct timespec *));
82 static int setfown __P((struct thread *td, struct vnode *, uid_t, gid_t));
83 static int setfmode __P((struct thread *td, struct vnode *, int));
84 static int setfflags __P((struct thread *td, struct vnode *, int));
85 static int setutimes __P((struct thread *td, struct vnode *,
86     const struct timespec *, int));
87 static int vn_access __P((struct vnode *vp, int user_flags, struct ucred *cred,
88     struct thread *td));
89 
90 static int	usermount = 0;	/* if 1, non-root can mount fs. */
91 
92 int (*union_dircheckp) __P((struct thread *td, struct vnode **, struct file *));
93 
94 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
95 
96 /*
97  * Virtual File System System Calls
98  */
99 
100 #ifndef _SYS_SYSPROTO_H_
101 struct nmount_args {
102 	struct iovec    *iovp;
103 	unsigned int    iovcnt;
104 	int             flags;
105 	};
106 #endif
107 /* ARGSUSED */
108 int
109 nmount(td, uap)
110 	struct thread *td;
111 	struct nmount_args /* {
112 		syscallarg(struct iovec *) iovp;
113 		syscallarg(unsigned int) iovcnt;
114 		syscallarg(int) flags;
115 	} */ *uap;
116 {
117 
118 	return(EOPNOTSUPP);
119 }
120 
121 /*
122  * Mount a file system.
123  */
124 #ifndef _SYS_SYSPROTO_H_
125 struct mount_args {
126 	char	*type;
127 	char	*path;
128 	int	flags;
129 	caddr_t	data;
130 };
131 #endif
132 /* ARGSUSED */
133 int
134 mount(td, uap)
135 	struct thread *td;
136 	struct mount_args /* {
137 		syscallarg(char *) type;
138 		syscallarg(char *) path;
139 		syscallarg(int) flags;
140 		syscallarg(caddr_t) data;
141 	} */ *uap;
142 {
143 	char *fstype;
144 	char *fspath;
145 	int error;
146 
147 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK | M_ZERO);
148 	fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK | M_ZERO);
149 
150 	/*
151 	 * vfs_mount() actually takes a kernel string for `type' and
152 	 * `path' now, so extract them.
153 	 */
154 	error = copyinstr(SCARG(uap, type), fstype, MFSNAMELEN, NULL);
155 	if (error)
156 		goto finish;
157 	error = copyinstr(SCARG(uap, path), fspath, MNAMELEN, NULL);
158 	if (error)
159 		goto finish;
160 	error = vfs_mount(td, fstype, fspath, SCARG(uap, flags),
161 	    SCARG(uap, data));
162 finish:
163 	free(fstype, M_TEMP);
164 	free(fspath, M_TEMP);
165 	return (error);
166 }
167 
168 /*
169  * vfs_mount(): actually attempt a filesystem mount.
170  *
171  * This routine is designed to be a "generic" entry point for routines
172  * that wish to mount a filesystem. All parameters except `fsdata' are
173  * pointers into kernel space. `fsdata' is currently still a pointer
174  * into userspace.
175  */
176 int
177 vfs_mount(td, fstype, fspath, fsflags, fsdata)
178 	struct thread *td;
179 	const char *fstype;
180 	char *fspath;
181 	int fsflags;
182 	void *fsdata;
183 {
184 	struct vnode *vp;
185 	struct mount *mp;
186 	struct vfsconf *vfsp;
187 	int error, flag = 0, flag2 = 0;
188 	struct vattr va;
189 	struct nameidata nd;
190 
191 	/*
192 	 * Be ultra-paranoid about making sure the type and fspath
193 	 * variables will fit in our mp buffers, including the
194 	 * terminating NUL.
195 	 */
196 	if ((strlen(fstype) >= MFSNAMELEN - 1) ||
197 	    (strlen(fspath) >= MNAMELEN - 1))
198 		return (ENAMETOOLONG);
199 
200 	if (usermount == 0) {
201 		error = suser_td(td);
202 		if (error)
203 			return (error);
204 	}
205 	/*
206 	 * Do not allow NFS export by non-root users.
207 	 */
208 	if (fsflags & MNT_EXPORTED) {
209 		error = suser_td(td);
210 		if (error)
211 			return (error);
212 	}
213 	/*
214 	 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
215 	 */
216 	if (suser_xxx(td->td_proc->p_ucred, 0, 0))
217 		fsflags |= MNT_NOSUID | MNT_NODEV;
218 	/*
219 	 * Get vnode to be covered
220 	 */
221 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td);
222 	if ((error = namei(&nd)) != 0)
223 		return (error);
224 	NDFREE(&nd, NDF_ONLY_PNBUF);
225 	vp = nd.ni_vp;
226 	if (fsflags & MNT_UPDATE) {
227 		if ((vp->v_flag & VROOT) == 0) {
228 			vput(vp);
229 			return (EINVAL);
230 		}
231 		mp = vp->v_mount;
232 		flag = mp->mnt_flag;
233 		flag2 = mp->mnt_kern_flag;
234 		/*
235 		 * We only allow the filesystem to be reloaded if it
236 		 * is currently mounted read-only.
237 		 */
238 		if ((fsflags & MNT_RELOAD) &&
239 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
240 			vput(vp);
241 			return (EOPNOTSUPP);	/* Needs translation */
242 		}
243 		/*
244 		 * Only root, or the user that did the original mount is
245 		 * permitted to update it.
246 		 */
247 		if (mp->mnt_stat.f_owner != td->td_proc->p_ucred->cr_uid) {
248 			error = suser_td(td);
249 			if (error) {
250 				vput(vp);
251 				return (error);
252 			}
253 		}
254 		if (vfs_busy(mp, LK_NOWAIT, 0, td)) {
255 			vput(vp);
256 			return (EBUSY);
257 		}
258 		mtx_lock(&vp->v_interlock);
259 		if ((vp->v_flag & VMOUNT) != 0 ||
260 		    vp->v_mountedhere != NULL) {
261 			mtx_unlock(&vp->v_interlock);
262 			vfs_unbusy(mp, td);
263 			vput(vp);
264 			return (EBUSY);
265 		}
266 		vp->v_flag |= VMOUNT;
267 		mtx_unlock(&vp->v_interlock);
268 		mp->mnt_flag |= fsflags &
269 		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT);
270 		VOP_UNLOCK(vp, 0, td);
271 		goto update;
272 	}
273 	/*
274 	 * If the user is not root, ensure that they own the directory
275 	 * onto which we are attempting to mount.
276 	 */
277 	error = VOP_GETATTR(vp, &va, td->td_proc->p_ucred, td);
278 	if (error) {
279 		vput(vp);
280 		return (error);
281 	}
282 	if (va.va_uid != td->td_proc->p_ucred->cr_uid) {
283 		error = suser_td(td);
284 		if (error) {
285 			vput(vp);
286 			return (error);
287 		}
288 	}
289 	if ((error = vinvalbuf(vp, V_SAVE, td->td_proc->p_ucred, td, 0, 0))
290 	    != 0) {
291 		vput(vp);
292 		return (error);
293 	}
294 	if (vp->v_type != VDIR) {
295 		vput(vp);
296 		return (ENOTDIR);
297 	}
298 	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
299 		if (!strcmp(vfsp->vfc_name, fstype))
300 			break;
301 	if (vfsp == NULL) {
302 		linker_file_t lf;
303 
304 		/* Only load modules for root (very important!) */
305 		error = suser_td(td);
306 		if (error) {
307 			vput(vp);
308 			return error;
309 		}
310 		error = linker_load_file(fstype, &lf);
311 		if (error || lf == NULL) {
312 			vput(vp);
313 			if (lf == NULL)
314 				error = ENODEV;
315 			return error;
316 		}
317 		lf->userrefs++;
318 		/* lookup again, see if the VFS was loaded */
319 		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
320 			if (!strcmp(vfsp->vfc_name, fstype))
321 				break;
322 		if (vfsp == NULL) {
323 			lf->userrefs--;
324 			linker_file_unload(lf);
325 			vput(vp);
326 			return (ENODEV);
327 		}
328 	}
329 	mtx_lock(&vp->v_interlock);
330 	if ((vp->v_flag & VMOUNT) != 0 ||
331 	    vp->v_mountedhere != NULL) {
332 		mtx_unlock(&vp->v_interlock);
333 		vput(vp);
334 		return (EBUSY);
335 	}
336 	vp->v_flag |= VMOUNT;
337 	mtx_unlock(&vp->v_interlock);
338 
339 	/*
340 	 * Allocate and initialize the filesystem.
341 	 */
342 	mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO);
343 	TAILQ_INIT(&mp->mnt_nvnodelist);
344 	TAILQ_INIT(&mp->mnt_reservedvnlist);
345 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE);
346 	(void)vfs_busy(mp, LK_NOWAIT, 0, td);
347 	mp->mnt_op = vfsp->vfc_vfsops;
348 	mp->mnt_vfc = vfsp;
349 	vfsp->vfc_refcount++;
350 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
351 	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
352 	strncpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN);
353 	mp->mnt_stat.f_fstypename[MFSNAMELEN - 1] = '\0';
354 	mp->mnt_vnodecovered = vp;
355 	mp->mnt_stat.f_owner = td->td_proc->p_ucred->cr_uid;
356 	strncpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
357 	mp->mnt_stat.f_mntonname[MNAMELEN - 1] = '\0';
358 	mp->mnt_iosize_max = DFLTPHYS;
359 	VOP_UNLOCK(vp, 0, td);
360 update:
361 	/*
362 	 * Set the mount level flags.
363 	 */
364 	if (fsflags & MNT_RDONLY)
365 		mp->mnt_flag |= MNT_RDONLY;
366 	else if (mp->mnt_flag & MNT_RDONLY)
367 		mp->mnt_kern_flag |= MNTK_WANTRDWR;
368 	mp->mnt_flag &=~ MNT_UPDATEMASK;
369 	mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE);
370 	/*
371 	 * Mount the filesystem.
372 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
373 	 * get.  No freeing of cn_pnbuf.
374 	 */
375 	error = VFS_MOUNT(mp, fspath, fsdata, &nd, td);
376 	if (mp->mnt_flag & MNT_UPDATE) {
377 		if (mp->mnt_kern_flag & MNTK_WANTRDWR)
378 			mp->mnt_flag &= ~MNT_RDONLY;
379 		mp->mnt_flag &=~
380 		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT);
381 		mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
382 		if (error) {
383 			mp->mnt_flag = flag;
384 			mp->mnt_kern_flag = flag2;
385 		}
386 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
387 			if (mp->mnt_syncer == NULL)
388 				error = vfs_allocate_syncvnode(mp);
389 		} else {
390 			if (mp->mnt_syncer != NULL)
391 				vrele(mp->mnt_syncer);
392 			mp->mnt_syncer = NULL;
393 		}
394 		vfs_unbusy(mp, td);
395 		mtx_lock(&vp->v_interlock);
396 		vp->v_flag &= ~VMOUNT;
397 		mtx_unlock(&vp->v_interlock);
398 		vrele(vp);
399 		return (error);
400 	}
401 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
402 	/*
403 	 * Put the new filesystem on the mount list after root.
404 	 */
405 	cache_purge(vp);
406 	if (!error) {
407 		struct vnode *newdp;
408 
409 		mtx_lock(&vp->v_interlock);
410 		vp->v_flag &= ~VMOUNT;
411 		vp->v_mountedhere = mp;
412 		mtx_unlock(&vp->v_interlock);
413 		mtx_lock(&mountlist_mtx);
414 		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
415 		mtx_unlock(&mountlist_mtx);
416 		if (VFS_ROOT(mp, &newdp))
417 			panic("mount: lost mount");
418 		checkdirs(vp, newdp);
419 		vput(newdp);
420 		VOP_UNLOCK(vp, 0, td);
421 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
422 			error = vfs_allocate_syncvnode(mp);
423 		vfs_unbusy(mp, td);
424 		if ((error = VFS_START(mp, 0, td)) != 0)
425 			vrele(vp);
426 	} else {
427 		mtx_lock(&vp->v_interlock);
428 		vp->v_flag &= ~VMOUNT;
429 		mtx_unlock(&vp->v_interlock);
430 		mp->mnt_vfc->vfc_refcount--;
431 		vfs_unbusy(mp, td);
432 		free((caddr_t)mp, M_MOUNT);
433 		vput(vp);
434 	}
435 	return (error);
436 }
437 
438 /*
439  * Scan all active processes to see if any of them have a current
440  * or root directory of `olddp'. If so, replace them with the new
441  * mount point.
442  */
443 static void
444 checkdirs(olddp, newdp)
445 	struct vnode *olddp, *newdp;
446 {
447 	struct filedesc *fdp;
448 	struct proc *p;
449 
450 	if (olddp->v_usecount == 1)
451 		return;
452 	sx_slock(&allproc_lock);
453 	LIST_FOREACH(p, &allproc, p_list) {
454 		fdp = p->p_fd;
455 		if (fdp == NULL)
456 			continue;
457 		FILEDESC_LOCK(fdp);
458 		if (fdp->fd_cdir == olddp) {
459 			VREF(newdp);
460 			fdp->fd_cdir = newdp;
461 			FILEDESC_UNLOCK(fdp);
462 			vrele(olddp);
463 			FILEDESC_LOCK(fdp);
464 		}
465 		if (fdp->fd_rdir == olddp) {
466 			VREF(newdp);
467 			fdp->fd_rdir = newdp;
468 			FILEDESC_UNLOCK(fdp);
469 			vrele(olddp);
470 		} else
471 			FILEDESC_UNLOCK(fdp);
472 	}
473 	sx_sunlock(&allproc_lock);
474 	if (rootvnode == olddp) {
475 		vrele(rootvnode);
476 		VREF(newdp);
477 		rootvnode = newdp;
478 	}
479 }
480 
481 /*
482  * Unmount a file system.
483  *
484  * Note: unmount takes a path to the vnode mounted on as argument,
485  * not special file (as before).
486  */
487 #ifndef _SYS_SYSPROTO_H_
488 struct unmount_args {
489 	char	*path;
490 	int	flags;
491 };
492 #endif
493 /* ARGSUSED */
494 int
495 unmount(td, uap)
496 	struct thread *td;
497 	register struct unmount_args /* {
498 		syscallarg(char *) path;
499 		syscallarg(int) flags;
500 	} */ *uap;
501 {
502 	register struct vnode *vp;
503 	struct mount *mp;
504 	int error;
505 	struct nameidata nd;
506 
507 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
508 	    SCARG(uap, path), td);
509 	if ((error = namei(&nd)) != 0)
510 		return (error);
511 	vp = nd.ni_vp;
512 	NDFREE(&nd, NDF_ONLY_PNBUF);
513 	mp = vp->v_mount;
514 
515 	/*
516 	 * Only root, or the user that did the original mount is
517 	 * permitted to unmount this filesystem.
518 	 */
519 	if (mp->mnt_stat.f_owner != td->td_proc->p_ucred->cr_uid) {
520 		error = suser_td(td);
521 		if (error) {
522 			vput(vp);
523 			return (error);
524 		}
525 	}
526 
527 	/*
528 	 * Don't allow unmounting the root file system.
529 	 */
530 	if (mp->mnt_flag & MNT_ROOTFS) {
531 		vput(vp);
532 		return (EINVAL);
533 	}
534 
535 	/*
536 	 * Must be the root of the filesystem
537 	 */
538 	if ((vp->v_flag & VROOT) == 0) {
539 		vput(vp);
540 		return (EINVAL);
541 	}
542 	vput(vp);
543 	return (dounmount(mp, SCARG(uap, flags), td));
544 }
545 
546 /*
547  * Do the actual file system unmount.
548  */
549 int
550 dounmount(mp, flags, td)
551 	struct mount *mp;
552 	int flags;
553 	struct thread *td;
554 {
555 	struct vnode *coveredvp, *fsrootvp;
556 	int error;
557 	int async_flag;
558 
559 	mtx_lock(&mountlist_mtx);
560 	mp->mnt_kern_flag |= MNTK_UNMOUNT;
561 	error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
562 	    ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td);
563 	if (error) {
564 		mp->mnt_kern_flag &= ~MNTK_UNMOUNT;
565 		if (mp->mnt_kern_flag & MNTK_MWAIT)
566 			wakeup((caddr_t)mp);
567 		return (error);
568 	}
569 	vn_start_write(NULL, &mp, V_WAIT);
570 
571 	if (mp->mnt_flag & MNT_EXPUBLIC)
572 		vfs_setpublicfs(NULL, NULL, NULL);
573 
574 	vfs_msync(mp, MNT_WAIT);
575 	async_flag = mp->mnt_flag & MNT_ASYNC;
576 	mp->mnt_flag &=~ MNT_ASYNC;
577 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
578 	if (mp->mnt_syncer != NULL)
579 		vrele(mp->mnt_syncer);
580 	/* Move process cdir/rdir refs on fs root to underlying vnode. */
581 	if (VFS_ROOT(mp, &fsrootvp) == 0) {
582 		if (mp->mnt_vnodecovered != NULL)
583 			checkdirs(fsrootvp, mp->mnt_vnodecovered);
584 		if (fsrootvp == rootvnode) {
585 			vrele(rootvnode);
586 			rootvnode = NULL;
587 		}
588 		vput(fsrootvp);
589 	}
590 	if (((mp->mnt_flag & MNT_RDONLY) ||
591 	     (error = VFS_SYNC(mp, MNT_WAIT, td->td_proc->p_ucred, td)) == 0) ||
592 	    (flags & MNT_FORCE)) {
593 		error = VFS_UNMOUNT(mp, flags, td);
594 	}
595 	vn_finished_write(mp);
596 	if (error) {
597 		/* Undo cdir/rdir and rootvnode changes made above. */
598 		if (VFS_ROOT(mp, &fsrootvp) == 0) {
599 			if (mp->mnt_vnodecovered != NULL)
600 				checkdirs(mp->mnt_vnodecovered, fsrootvp);
601 			if (rootvnode == NULL) {
602 				rootvnode = fsrootvp;
603 				vref(rootvnode);
604 			}
605 			vput(fsrootvp);
606 		}
607 		if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
608 			(void) vfs_allocate_syncvnode(mp);
609 		mtx_lock(&mountlist_mtx);
610 		mp->mnt_kern_flag &= ~MNTK_UNMOUNT;
611 		mp->mnt_flag |= async_flag;
612 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK,
613 		    &mountlist_mtx, td);
614 		if (mp->mnt_kern_flag & MNTK_MWAIT)
615 			wakeup((caddr_t)mp);
616 		return (error);
617 	}
618 	mtx_lock(&mountlist_mtx);
619 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
620 	if ((coveredvp = mp->mnt_vnodecovered) != NULL)
621 		coveredvp->v_mountedhere = NULL;
622 	mp->mnt_vfc->vfc_refcount--;
623 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
624 		panic("unmount: dangling vnode");
625 	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_mtx, td);
626 	lockdestroy(&mp->mnt_lock);
627 	if (coveredvp != NULL)
628 		vrele(coveredvp);
629 	if (mp->mnt_kern_flag & MNTK_MWAIT)
630 		wakeup((caddr_t)mp);
631 	free((caddr_t)mp, M_MOUNT);
632 	return (0);
633 }
634 
635 /*
636  * Sync each mounted filesystem.
637  */
638 #ifndef _SYS_SYSPROTO_H_
639 struct sync_args {
640         int     dummy;
641 };
642 #endif
643 
644 #ifdef DEBUG
645 static int syncprt = 0;
646 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
647 #endif
648 
649 /* ARGSUSED */
650 int
651 sync(td, uap)
652 	struct thread *td;
653 	struct sync_args *uap;
654 {
655 	struct mount *mp, *nmp;
656 	int asyncflag;
657 
658 	mtx_lock(&mountlist_mtx);
659 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
660 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
661 			nmp = TAILQ_NEXT(mp, mnt_list);
662 			continue;
663 		}
664 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
665 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
666 			asyncflag = mp->mnt_flag & MNT_ASYNC;
667 			mp->mnt_flag &= ~MNT_ASYNC;
668 			vfs_msync(mp, MNT_NOWAIT);
669 			VFS_SYNC(mp, MNT_NOWAIT,
670 			    ((td != NULL) ? td->td_proc->p_ucred : NOCRED), td);
671 			mp->mnt_flag |= asyncflag;
672 			vn_finished_write(mp);
673 		}
674 		mtx_lock(&mountlist_mtx);
675 		nmp = TAILQ_NEXT(mp, mnt_list);
676 		vfs_unbusy(mp, td);
677 	}
678 	mtx_unlock(&mountlist_mtx);
679 #if 0
680 /*
681  * XXX don't call vfs_bufstats() yet because that routine
682  * was not imported in the Lite2 merge.
683  */
684 #ifdef DIAGNOSTIC
685 	if (syncprt)
686 		vfs_bufstats();
687 #endif /* DIAGNOSTIC */
688 #endif
689 	return (0);
690 }
691 
692 /* XXX PRISON: could be per prison flag */
693 static int prison_quotas;
694 #if 0
695 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
696 #endif
697 
698 /*
699  * Change filesystem quotas.
700  */
701 #ifndef _SYS_SYSPROTO_H_
702 struct quotactl_args {
703 	char *path;
704 	int cmd;
705 	int uid;
706 	caddr_t arg;
707 };
708 #endif
709 /* ARGSUSED */
710 int
711 quotactl(td, uap)
712 	struct thread *td;
713 	register struct quotactl_args /* {
714 		syscallarg(char *) path;
715 		syscallarg(int) cmd;
716 		syscallarg(int) uid;
717 		syscallarg(caddr_t) arg;
718 	} */ *uap;
719 {
720 	struct mount *mp;
721 	int error;
722 	struct nameidata nd;
723 
724 	if (jailed(td->td_proc->p_ucred) && !prison_quotas)
725 		return (EPERM);
726 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
727 	if ((error = namei(&nd)) != 0)
728 		return (error);
729 	NDFREE(&nd, NDF_ONLY_PNBUF);
730 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH);
731 	vrele(nd.ni_vp);
732 	if (error)
733 		return (error);
734 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
735 	    SCARG(uap, arg), td);
736 	vn_finished_write(mp);
737 	return (error);
738 }
739 
740 /*
741  * Get filesystem statistics.
742  */
743 #ifndef _SYS_SYSPROTO_H_
744 struct statfs_args {
745 	char *path;
746 	struct statfs *buf;
747 };
748 #endif
749 /* ARGSUSED */
750 int
751 statfs(td, uap)
752 	struct thread *td;
753 	register struct statfs_args /* {
754 		syscallarg(char *) path;
755 		syscallarg(struct statfs *) buf;
756 	} */ *uap;
757 {
758 	register struct mount *mp;
759 	register struct statfs *sp;
760 	int error;
761 	struct nameidata nd;
762 	struct statfs sb;
763 
764 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
765 	if ((error = namei(&nd)) != 0)
766 		return (error);
767 	mp = nd.ni_vp->v_mount;
768 	sp = &mp->mnt_stat;
769 	NDFREE(&nd, NDF_ONLY_PNBUF);
770 	vrele(nd.ni_vp);
771 	error = VFS_STATFS(mp, sp, td);
772 	if (error)
773 		return (error);
774 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
775 	if (suser_xxx(td->td_proc->p_ucred, 0, 0)) {
776 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
777 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
778 		sp = &sb;
779 	}
780 	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
781 }
782 
783 /*
784  * Get filesystem statistics.
785  */
786 #ifndef _SYS_SYSPROTO_H_
787 struct fstatfs_args {
788 	int fd;
789 	struct statfs *buf;
790 };
791 #endif
792 /* ARGSUSED */
793 int
794 fstatfs(td, uap)
795 	struct thread *td;
796 	register struct fstatfs_args /* {
797 		syscallarg(int) fd;
798 		syscallarg(struct statfs *) buf;
799 	} */ *uap;
800 {
801 	struct file *fp;
802 	struct mount *mp;
803 	register struct statfs *sp;
804 	int error;
805 	struct statfs sb;
806 
807 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
808 		return (error);
809 	mp = ((struct vnode *)fp->f_data)->v_mount;
810 	fdrop(fp, td);
811 	if (mp == NULL)
812 		return (EBADF);
813 	sp = &mp->mnt_stat;
814 	error = VFS_STATFS(mp, sp, td);
815 	if (error)
816 		return (error);
817 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
818 	if (suser_xxx(td->td_proc->p_ucred, 0, 0)) {
819 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
820 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
821 		sp = &sb;
822 	}
823 	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
824 }
825 
826 /*
827  * Get statistics on all filesystems.
828  */
829 #ifndef _SYS_SYSPROTO_H_
830 struct getfsstat_args {
831 	struct statfs *buf;
832 	long bufsize;
833 	int flags;
834 };
835 #endif
836 int
837 getfsstat(td, uap)
838 	struct thread *td;
839 	register struct getfsstat_args /* {
840 		syscallarg(struct statfs *) buf;
841 		syscallarg(long) bufsize;
842 		syscallarg(int) flags;
843 	} */ *uap;
844 {
845 	register struct mount *mp, *nmp;
846 	register struct statfs *sp;
847 	caddr_t sfsp;
848 	long count, maxcount, error;
849 
850 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
851 	sfsp = (caddr_t)SCARG(uap, buf);
852 	count = 0;
853 	mtx_lock(&mountlist_mtx);
854 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
855 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
856 			nmp = TAILQ_NEXT(mp, mnt_list);
857 			continue;
858 		}
859 		if (sfsp && count < maxcount) {
860 			sp = &mp->mnt_stat;
861 			/*
862 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
863 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
864 			 * overrides MNT_WAIT.
865 			 */
866 			if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
867 			    (SCARG(uap, flags) & MNT_WAIT)) &&
868 			    (error = VFS_STATFS(mp, sp, td))) {
869 				mtx_lock(&mountlist_mtx);
870 				nmp = TAILQ_NEXT(mp, mnt_list);
871 				vfs_unbusy(mp, td);
872 				continue;
873 			}
874 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
875 			error = copyout((caddr_t)sp, sfsp, sizeof(*sp));
876 			if (error) {
877 				vfs_unbusy(mp, td);
878 				return (error);
879 			}
880 			sfsp += sizeof(*sp);
881 		}
882 		count++;
883 		mtx_lock(&mountlist_mtx);
884 		nmp = TAILQ_NEXT(mp, mnt_list);
885 		vfs_unbusy(mp, td);
886 	}
887 	mtx_unlock(&mountlist_mtx);
888 	if (sfsp && count > maxcount)
889 		td->td_retval[0] = maxcount;
890 	else
891 		td->td_retval[0] = count;
892 	return (0);
893 }
894 
895 /*
896  * Change current working directory to a given file descriptor.
897  */
898 #ifndef _SYS_SYSPROTO_H_
899 struct fchdir_args {
900 	int	fd;
901 };
902 #endif
903 /* ARGSUSED */
904 int
905 fchdir(td, uap)
906 	struct thread *td;
907 	struct fchdir_args /* {
908 		syscallarg(int) fd;
909 	} */ *uap;
910 {
911 	register struct filedesc *fdp = td->td_proc->p_fd;
912 	struct vnode *vp, *tdp, *vpold;
913 	struct mount *mp;
914 	struct file *fp;
915 	int error;
916 
917 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
918 		return (error);
919 	vp = (struct vnode *)fp->f_data;
920 	VREF(vp);
921 	fdrop(fp, td);
922 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
923 	if (vp->v_type != VDIR)
924 		error = ENOTDIR;
925 	else
926 		error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td);
927 	while (!error && (mp = vp->v_mountedhere) != NULL) {
928 		if (vfs_busy(mp, 0, 0, td))
929 			continue;
930 		error = VFS_ROOT(mp, &tdp);
931 		vfs_unbusy(mp, td);
932 		if (error)
933 			break;
934 		vput(vp);
935 		vp = tdp;
936 	}
937 	if (error) {
938 		vput(vp);
939 		return (error);
940 	}
941 	VOP_UNLOCK(vp, 0, td);
942 	FILEDESC_LOCK(fdp);
943 	vpold = fdp->fd_cdir;
944 	fdp->fd_cdir = vp;
945 	FILEDESC_UNLOCK(fdp);
946 	vrele(vpold);
947 	return (0);
948 }
949 
950 /*
951  * Change current working directory (``.'').
952  */
953 #ifndef _SYS_SYSPROTO_H_
954 struct chdir_args {
955 	char	*path;
956 };
957 #endif
958 /* ARGSUSED */
959 int
960 chdir(td, uap)
961 	struct thread *td;
962 	struct chdir_args /* {
963 		syscallarg(char *) path;
964 	} */ *uap;
965 {
966 	register struct filedesc *fdp = td->td_proc->p_fd;
967 	int error;
968 	struct nameidata nd;
969 	struct vnode *vp;
970 
971 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
972 	    SCARG(uap, path), td);
973 	if ((error = change_dir(&nd, td)) != 0)
974 		return (error);
975 	NDFREE(&nd, NDF_ONLY_PNBUF);
976 	FILEDESC_LOCK(fdp);
977 	vp = fdp->fd_cdir;
978 	fdp->fd_cdir = nd.ni_vp;
979 	FILEDESC_UNLOCK(fdp);
980 	vrele(vp);
981 	return (0);
982 }
983 
984 /*
985  * Helper function for raised chroot(2) security function:  Refuse if
986  * any filedescriptors are open directories.
987  */
988 static int
989 chroot_refuse_vdir_fds(fdp)
990 	struct filedesc *fdp;
991 {
992 	struct vnode *vp;
993 	struct file *fp;
994 	struct thread *td = curthread;
995 	int error;
996 	int fd;
997 
998 	FILEDESC_LOCK(fdp);
999 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1000 		error = getvnode(fdp, fd, &fp);
1001 		if (error)
1002 			continue;
1003 		vp = (struct vnode *)fp->f_data;
1004 		fdrop(fp, td);
1005 		if (vp->v_type != VDIR)
1006 			continue;
1007 		FILEDESC_UNLOCK(fdp);
1008 		return(EPERM);
1009 	}
1010 	FILEDESC_UNLOCK(fdp);
1011 	return (0);
1012 }
1013 
1014 /*
1015  * This sysctl determines if we will allow a process to chroot(2) if it
1016  * has a directory open:
1017  *	0: disallowed for all processes.
1018  *	1: allowed for processes that were not already chroot(2)'ed.
1019  *	2: allowed for all processes.
1020  */
1021 
1022 static int chroot_allow_open_directories = 1;
1023 
1024 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1025      &chroot_allow_open_directories, 0, "");
1026 
1027 /*
1028  * Change notion of root (``/'') directory.
1029  */
1030 #ifndef _SYS_SYSPROTO_H_
1031 struct chroot_args {
1032 	char	*path;
1033 };
1034 #endif
1035 /* ARGSUSED */
1036 int
1037 chroot(td, uap)
1038 	struct thread *td;
1039 	struct chroot_args /* {
1040 		syscallarg(char *) path;
1041 	} */ *uap;
1042 {
1043 	register struct filedesc *fdp = td->td_proc->p_fd;
1044 	int error;
1045 	struct nameidata nd;
1046 	struct vnode *vp;
1047 
1048 	error = suser_xxx(0, td->td_proc, PRISON_ROOT);
1049 	if (error)
1050 		return (error);
1051 	FILEDESC_LOCK(fdp);
1052 	if (chroot_allow_open_directories == 0 ||
1053 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1054 		FILEDESC_UNLOCK(fdp);
1055 		error = chroot_refuse_vdir_fds(fdp);
1056 	} else
1057 		FILEDESC_UNLOCK(fdp);
1058 	if (error)
1059 		return (error);
1060 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1061 	    SCARG(uap, path), td);
1062 	if ((error = change_dir(&nd, td)) != 0)
1063 		return (error);
1064 	NDFREE(&nd, NDF_ONLY_PNBUF);
1065 	FILEDESC_LOCK(fdp);
1066 	vp = fdp->fd_rdir;
1067 	fdp->fd_rdir = nd.ni_vp;
1068 	if (!fdp->fd_jdir) {
1069 		fdp->fd_jdir = nd.ni_vp;
1070                 VREF(fdp->fd_jdir);
1071 	}
1072 	FILEDESC_UNLOCK(fdp);
1073 	vrele(vp);
1074 	return (0);
1075 }
1076 
1077 /*
1078  * Common routine for chroot and chdir.
1079  */
1080 static int
1081 change_dir(ndp, td)
1082 	register struct nameidata *ndp;
1083 	struct thread *td;
1084 {
1085 	struct vnode *vp;
1086 	int error;
1087 
1088 	error = namei(ndp);
1089 	if (error)
1090 		return (error);
1091 	vp = ndp->ni_vp;
1092 	if (vp->v_type != VDIR)
1093 		error = ENOTDIR;
1094 	else
1095 		error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td);
1096 	if (error)
1097 		vput(vp);
1098 	else
1099 		VOP_UNLOCK(vp, 0, td);
1100 	return (error);
1101 }
1102 
1103 /*
1104  * Check permissions, allocate an open file structure,
1105  * and call the device open routine if any.
1106  */
1107 #ifndef _SYS_SYSPROTO_H_
1108 struct open_args {
1109 	char	*path;
1110 	int	flags;
1111 	int	mode;
1112 };
1113 #endif
1114 int
1115 open(td, uap)
1116 	struct thread *td;
1117 	register struct open_args /* {
1118 		syscallarg(char *) path;
1119 		syscallarg(int) flags;
1120 		syscallarg(int) mode;
1121 	} */ *uap;
1122 {
1123 	struct proc *p = td->td_proc;
1124 	struct filedesc *fdp = p->p_fd;
1125 	struct file *fp;
1126 	struct vnode *vp;
1127 	struct vattr vat;
1128 	struct mount *mp;
1129 	int cmode, flags, oflags;
1130 	struct file *nfp;
1131 	int type, indx, error;
1132 	struct flock lf;
1133 	struct nameidata nd;
1134 
1135 	oflags = SCARG(uap, flags);
1136 	if ((oflags & O_ACCMODE) == O_ACCMODE)
1137 		return (EINVAL);
1138 	flags = FFLAGS(oflags);
1139 	error = falloc(td, &nfp, &indx);
1140 	if (error)
1141 		return (error);
1142 	fp = nfp;
1143 	FILEDESC_LOCK(fdp);
1144 	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1145 	FILEDESC_UNLOCK(fdp);
1146 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
1147 	td->td_dupfd = -indx - 1;		/* XXX check for fdopen */
1148 	/*
1149 	 * Bump the ref count to prevent another process from closing
1150 	 * the descriptor while we are blocked in vn_open()
1151 	 */
1152 	fhold(fp);
1153 	error = vn_open(&nd, &flags, cmode);
1154 	if (error) {
1155 		/*
1156 		 * release our own reference
1157 		 */
1158 		fdrop(fp, td);
1159 
1160 		/*
1161 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1162 		 * responsible for dropping the old contents of ofiles[indx]
1163 		 * if it succeeds.
1164 		 */
1165 		if ((error == ENODEV || error == ENXIO) &&
1166 		    td->td_dupfd >= 0 &&		/* XXX from fdopen */
1167 		    (error =
1168 			dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1169 			td->td_retval[0] = indx;
1170 			return (0);
1171 		}
1172 		/*
1173 		 * Clean up the descriptor, but only if another thread hadn't
1174 		 * replaced or closed it.
1175 		 */
1176 		FILEDESC_LOCK(fdp);
1177 		if (fdp->fd_ofiles[indx] == fp) {
1178 			fdp->fd_ofiles[indx] = NULL;
1179 			FILEDESC_UNLOCK(fdp);
1180 			fdrop(fp, td);
1181 		} else
1182 			FILEDESC_UNLOCK(fdp);
1183 
1184 		if (error == ERESTART)
1185 			error = EINTR;
1186 		return (error);
1187 	}
1188 	td->td_dupfd = 0;
1189 	NDFREE(&nd, NDF_ONLY_PNBUF);
1190 	vp = nd.ni_vp;
1191 
1192 	/*
1193 	 * There should be 2 references on the file, one from the descriptor
1194 	 * table, and one for us.
1195 	 *
1196 	 * Handle the case where someone closed the file (via its file
1197 	 * descriptor) while we were blocked.  The end result should look
1198 	 * like opening the file succeeded but it was immediately closed.
1199 	 */
1200 	FILEDESC_LOCK(fdp);
1201 	FILE_LOCK(fp);
1202 	if (fp->f_count == 1) {
1203 		KASSERT(fdp->fd_ofiles[indx] != fp,
1204 		    ("Open file descriptor lost all refs"));
1205 		FILEDESC_UNLOCK(fdp);
1206 		FILE_UNLOCK(fp);
1207 		VOP_UNLOCK(vp, 0, td);
1208 		vn_close(vp, flags & FMASK, fp->f_cred, td);
1209 		fdrop(fp, td);
1210 		td->td_retval[0] = indx;
1211 		return 0;
1212 	}
1213 
1214 	fp->f_data = (caddr_t)vp;
1215 	fp->f_flag = flags & FMASK;
1216 	fp->f_ops = &vnops;
1217 	fp->f_type = (vp->v_type == VFIFO ? DTYPE_FIFO : DTYPE_VNODE);
1218 	FILEDESC_UNLOCK(fdp);
1219 	FILE_UNLOCK(fp);
1220 	VOP_UNLOCK(vp, 0, td);
1221 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1222 		lf.l_whence = SEEK_SET;
1223 		lf.l_start = 0;
1224 		lf.l_len = 0;
1225 		if (flags & O_EXLOCK)
1226 			lf.l_type = F_WRLCK;
1227 		else
1228 			lf.l_type = F_RDLCK;
1229 		type = F_FLOCK;
1230 		if ((flags & FNONBLOCK) == 0)
1231 			type |= F_WAIT;
1232 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0)
1233 			goto bad;
1234 		fp->f_flag |= FHASLOCK;
1235 	}
1236 	if (flags & O_TRUNC) {
1237 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1238 			goto bad;
1239 		VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
1240 		VATTR_NULL(&vat);
1241 		vat.va_size = 0;
1242 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1243 		error = VOP_SETATTR(vp, &vat, p->p_ucred, td);
1244 		VOP_UNLOCK(vp, 0, td);
1245 		vn_finished_write(mp);
1246 		if (error)
1247 			goto bad;
1248 	}
1249 	/* assert that vn_open created a backing object if one is needed */
1250 	KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1251 		("open: vmio vnode has no backing object after vn_open"));
1252 	/*
1253 	 * Release our private reference, leaving the one associated with
1254 	 * the descriptor table intact.
1255 	 */
1256 	fdrop(fp, td);
1257 	td->td_retval[0] = indx;
1258 	return (0);
1259 bad:
1260 	FILEDESC_LOCK(fdp);
1261 	if (fdp->fd_ofiles[indx] == fp) {
1262 		fdp->fd_ofiles[indx] = NULL;
1263 		FILEDESC_UNLOCK(fdp);
1264 		fdrop(fp, td);
1265 	} else
1266 		FILEDESC_UNLOCK(fdp);
1267 	return (error);
1268 }
1269 
1270 #ifdef COMPAT_43
1271 /*
1272  * Create a file.
1273  */
1274 #ifndef _SYS_SYSPROTO_H_
1275 struct ocreat_args {
1276 	char	*path;
1277 	int	mode;
1278 };
1279 #endif
1280 int
1281 ocreat(td, uap)
1282 	struct thread *td;
1283 	register struct ocreat_args /* {
1284 		syscallarg(char *) path;
1285 		syscallarg(int) mode;
1286 	} */ *uap;
1287 {
1288 	struct open_args /* {
1289 		syscallarg(char *) path;
1290 		syscallarg(int) flags;
1291 		syscallarg(int) mode;
1292 	} */ nuap;
1293 
1294 	SCARG(&nuap, path) = SCARG(uap, path);
1295 	SCARG(&nuap, mode) = SCARG(uap, mode);
1296 	SCARG(&nuap, flags) = O_WRONLY | O_CREAT | O_TRUNC;
1297 	return (open(td, &nuap));
1298 }
1299 #endif /* COMPAT_43 */
1300 
1301 /*
1302  * Create a special file.
1303  */
1304 #ifndef _SYS_SYSPROTO_H_
1305 struct mknod_args {
1306 	char	*path;
1307 	int	mode;
1308 	int	dev;
1309 };
1310 #endif
1311 /* ARGSUSED */
1312 int
1313 mknod(td, uap)
1314 	struct thread *td;
1315 	register struct mknod_args /* {
1316 		syscallarg(char *) path;
1317 		syscallarg(int) mode;
1318 		syscallarg(int) dev;
1319 	} */ *uap;
1320 {
1321 	struct vnode *vp;
1322 	struct mount *mp;
1323 	struct vattr vattr;
1324 	int error;
1325 	int whiteout = 0;
1326 	struct nameidata nd;
1327 
1328 	switch (SCARG(uap, mode) & S_IFMT) {
1329 	case S_IFCHR:
1330 	case S_IFBLK:
1331 		error = suser_td(td);
1332 		break;
1333 	default:
1334 		error = suser_xxx(0, td->td_proc, PRISON_ROOT);
1335 		break;
1336 	}
1337 	if (error)
1338 		return (error);
1339 restart:
1340 	bwillwrite();
1341 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
1342 	if ((error = namei(&nd)) != 0)
1343 		return (error);
1344 	vp = nd.ni_vp;
1345 	if (vp != NULL) {
1346 		vrele(vp);
1347 		error = EEXIST;
1348 	} else {
1349 		VATTR_NULL(&vattr);
1350 		FILEDESC_LOCK(td->td_proc->p_fd);
1351 		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ td->td_proc->p_fd->fd_cmask;
1352 		FILEDESC_UNLOCK(td->td_proc->p_fd);
1353 		vattr.va_rdev = SCARG(uap, dev);
1354 		whiteout = 0;
1355 
1356 		switch (SCARG(uap, mode) & S_IFMT) {
1357 		case S_IFMT:	/* used by badsect to flag bad sectors */
1358 			vattr.va_type = VBAD;
1359 			break;
1360 		case S_IFCHR:
1361 			vattr.va_type = VCHR;
1362 			break;
1363 		case S_IFBLK:
1364 			vattr.va_type = VBLK;
1365 			break;
1366 		case S_IFWHT:
1367 			whiteout = 1;
1368 			break;
1369 		default:
1370 			error = EINVAL;
1371 			break;
1372 		}
1373 	}
1374 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1375 		NDFREE(&nd, NDF_ONLY_PNBUF);
1376 		vput(nd.ni_dvp);
1377 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1378 			return (error);
1379 		goto restart;
1380 	}
1381 	if (!error) {
1382 		VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
1383 		if (whiteout)
1384 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1385 		else {
1386 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1387 						&nd.ni_cnd, &vattr);
1388 			if (error == 0)
1389 				vput(nd.ni_vp);
1390 		}
1391 	}
1392 	NDFREE(&nd, NDF_ONLY_PNBUF);
1393 	vput(nd.ni_dvp);
1394 	vn_finished_write(mp);
1395 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mknod");
1396 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "mknod");
1397 	return (error);
1398 }
1399 
1400 /*
1401  * Create a named pipe.
1402  */
1403 #ifndef _SYS_SYSPROTO_H_
1404 struct mkfifo_args {
1405 	char	*path;
1406 	int	mode;
1407 };
1408 #endif
1409 /* ARGSUSED */
1410 int
1411 mkfifo(td, uap)
1412 	struct thread *td;
1413 	register struct mkfifo_args /* {
1414 		syscallarg(char *) path;
1415 		syscallarg(int) mode;
1416 	} */ *uap;
1417 {
1418 	struct mount *mp;
1419 	struct vattr vattr;
1420 	int error;
1421 	struct nameidata nd;
1422 
1423 restart:
1424 	bwillwrite();
1425 	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
1426 	if ((error = namei(&nd)) != 0)
1427 		return (error);
1428 	if (nd.ni_vp != NULL) {
1429 		NDFREE(&nd, NDF_ONLY_PNBUF);
1430 		vrele(nd.ni_vp);
1431 		vput(nd.ni_dvp);
1432 		return (EEXIST);
1433 	}
1434 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1435 		NDFREE(&nd, NDF_ONLY_PNBUF);
1436 		vput(nd.ni_dvp);
1437 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1438 			return (error);
1439 		goto restart;
1440 	}
1441 	VATTR_NULL(&vattr);
1442 	vattr.va_type = VFIFO;
1443 	FILEDESC_LOCK(td->td_proc->p_fd);
1444 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ td->td_proc->p_fd->fd_cmask;
1445 	FILEDESC_UNLOCK(td->td_proc->p_fd);
1446 	VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
1447 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1448 	if (error == 0)
1449 		vput(nd.ni_vp);
1450 	NDFREE(&nd, NDF_ONLY_PNBUF);
1451 	vput(nd.ni_dvp);
1452 	vn_finished_write(mp);
1453 	return (error);
1454 }
1455 
1456 /*
1457  * Make a hard file link.
1458  */
1459 #ifndef _SYS_SYSPROTO_H_
1460 struct link_args {
1461 	char	*path;
1462 	char	*link;
1463 };
1464 #endif
1465 /* ARGSUSED */
1466 int
1467 link(td, uap)
1468 	struct thread *td;
1469 	register struct link_args /* {
1470 		syscallarg(char *) path;
1471 		syscallarg(char *) link;
1472 	} */ *uap;
1473 {
1474 	struct vnode *vp;
1475 	struct mount *mp;
1476 	struct nameidata nd;
1477 	int error;
1478 
1479 	bwillwrite();
1480 	NDINIT(&nd, LOOKUP, FOLLOW|NOOBJ, UIO_USERSPACE, SCARG(uap, path), td);
1481 	if ((error = namei(&nd)) != 0)
1482 		return (error);
1483 	NDFREE(&nd, NDF_ONLY_PNBUF);
1484 	vp = nd.ni_vp;
1485 	if (vp->v_type == VDIR) {
1486 		vrele(vp);
1487 		return (EPERM);		/* POSIX */
1488 	}
1489 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1490 		vrele(vp);
1491 		return (error);
1492 	}
1493 	NDINIT(&nd, CREATE, LOCKPARENT|NOOBJ, UIO_USERSPACE, SCARG(uap, link), td);
1494 	if ((error = namei(&nd)) == 0) {
1495 		if (nd.ni_vp != NULL) {
1496 			vrele(nd.ni_vp);
1497 			error = EEXIST;
1498 		} else {
1499 			VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
1500 			VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
1501 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1502 		}
1503 		NDFREE(&nd, NDF_ONLY_PNBUF);
1504 		vput(nd.ni_dvp);
1505 	}
1506 	vrele(vp);
1507 	vn_finished_write(mp);
1508 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "link");
1509 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "link");
1510 	return (error);
1511 }
1512 
1513 /*
1514  * Make a symbolic link.
1515  */
1516 #ifndef _SYS_SYSPROTO_H_
1517 struct symlink_args {
1518 	char	*path;
1519 	char	*link;
1520 };
1521 #endif
1522 /* ARGSUSED */
1523 int
1524 symlink(td, uap)
1525 	struct thread *td;
1526 	register struct symlink_args /* {
1527 		syscallarg(char *) path;
1528 		syscallarg(char *) link;
1529 	} */ *uap;
1530 {
1531 	struct mount *mp;
1532 	struct vattr vattr;
1533 	char *path;
1534 	int error;
1535 	struct nameidata nd;
1536 
1537 	path = zalloc(namei_zone);
1538 	if ((error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL)) != 0)
1539 		goto out;
1540 restart:
1541 	bwillwrite();
1542 	NDINIT(&nd, CREATE, LOCKPARENT|NOOBJ, UIO_USERSPACE, SCARG(uap, link), td);
1543 	if ((error = namei(&nd)) != 0)
1544 		goto out;
1545 	if (nd.ni_vp) {
1546 		NDFREE(&nd, NDF_ONLY_PNBUF);
1547 		vrele(nd.ni_vp);
1548 		vput(nd.ni_dvp);
1549 		error = EEXIST;
1550 		goto out;
1551 	}
1552 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1553 		NDFREE(&nd, NDF_ONLY_PNBUF);
1554 		vput(nd.ni_dvp);
1555 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1556 			return (error);
1557 		goto restart;
1558 	}
1559 	VATTR_NULL(&vattr);
1560 	FILEDESC_LOCK(td->td_proc->p_fd);
1561 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1562 	FILEDESC_UNLOCK(td->td_proc->p_fd);
1563 	VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
1564 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1565 	NDFREE(&nd, NDF_ONLY_PNBUF);
1566 	if (error == 0)
1567 		vput(nd.ni_vp);
1568 	vput(nd.ni_dvp);
1569 	vn_finished_write(mp);
1570 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "symlink");
1571 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "symlink");
1572 out:
1573 	zfree(namei_zone, path);
1574 	return (error);
1575 }
1576 
1577 /*
1578  * Delete a whiteout from the filesystem.
1579  */
1580 /* ARGSUSED */
1581 int
1582 undelete(td, uap)
1583 	struct thread *td;
1584 	register struct undelete_args /* {
1585 		syscallarg(char *) path;
1586 	} */ *uap;
1587 {
1588 	int error;
1589 	struct mount *mp;
1590 	struct nameidata nd;
1591 
1592 restart:
1593 	bwillwrite();
1594 	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1595 	    SCARG(uap, path), td);
1596 	error = namei(&nd);
1597 	if (error)
1598 		return (error);
1599 
1600 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1601 		NDFREE(&nd, NDF_ONLY_PNBUF);
1602 		if (nd.ni_vp)
1603 			vrele(nd.ni_vp);
1604 		vput(nd.ni_dvp);
1605 		return (EEXIST);
1606 	}
1607 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1608 		NDFREE(&nd, NDF_ONLY_PNBUF);
1609 		vput(nd.ni_dvp);
1610 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1611 			return (error);
1612 		goto restart;
1613 	}
1614 	VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
1615 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1616 	NDFREE(&nd, NDF_ONLY_PNBUF);
1617 	vput(nd.ni_dvp);
1618 	vn_finished_write(mp);
1619 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "undelete");
1620 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "undelete");
1621 	return (error);
1622 }
1623 
1624 /*
1625  * Delete a name from the filesystem.
1626  */
1627 #ifndef _SYS_SYSPROTO_H_
1628 struct unlink_args {
1629 	char	*path;
1630 };
1631 #endif
1632 /* ARGSUSED */
1633 int
1634 unlink(td, uap)
1635 	struct thread *td;
1636 	struct unlink_args /* {
1637 		syscallarg(char *) path;
1638 	} */ *uap;
1639 {
1640 	struct mount *mp;
1641 	struct vnode *vp;
1642 	int error;
1643 	struct nameidata nd;
1644 
1645 restart:
1646 	bwillwrite();
1647 	NDINIT(&nd, DELETE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), td);
1648 	if ((error = namei(&nd)) != 0)
1649 		return (error);
1650 	vp = nd.ni_vp;
1651 	if (vp->v_type == VDIR)
1652 		error = EPERM;		/* POSIX */
1653 	else {
1654 		/*
1655 		 * The root of a mounted filesystem cannot be deleted.
1656 		 *
1657 		 * XXX: can this only be a VDIR case?
1658 		 */
1659 		if (vp->v_flag & VROOT)
1660 			error = EBUSY;
1661 	}
1662 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1663 		NDFREE(&nd, NDF_ONLY_PNBUF);
1664 		vrele(vp);
1665 		vput(nd.ni_dvp);
1666 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1667 			return (error);
1668 		goto restart;
1669 	}
1670 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
1671 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1672 	if (!error) {
1673 		VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
1674 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1675 	}
1676 	NDFREE(&nd, NDF_ONLY_PNBUF);
1677 	vput(nd.ni_dvp);
1678 	vput(vp);
1679 	vn_finished_write(mp);
1680 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "unlink");
1681 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "unlink");
1682 	return (error);
1683 }
1684 
1685 /*
1686  * Reposition read/write file offset.
1687  */
1688 #ifndef _SYS_SYSPROTO_H_
1689 struct lseek_args {
1690 	int	fd;
1691 	int	pad;
1692 	off_t	offset;
1693 	int	whence;
1694 };
1695 #endif
1696 int
1697 lseek(td, uap)
1698 	struct thread *td;
1699 	register struct lseek_args /* {
1700 		syscallarg(int) fd;
1701 		syscallarg(int) pad;
1702 		syscallarg(off_t) offset;
1703 		syscallarg(int) whence;
1704 	} */ *uap;
1705 {
1706 	struct ucred *cred = td->td_proc->p_ucred;
1707 	struct file *fp;
1708 	struct vnode *vp;
1709 	struct vattr vattr;
1710 	off_t offset;
1711 	int error, noneg;
1712 
1713 	if ((error = fget(td, uap->fd, &fp)) != 0)
1714 		return (error);
1715 	if (fp->f_type != DTYPE_VNODE) {
1716 		fdrop(fp, td);
1717 		return (ESPIPE);
1718 	}
1719 	vp = (struct vnode *)fp->f_data;
1720 	noneg = (vp->v_type != VCHR);
1721 	offset = SCARG(uap, offset);
1722 	switch (SCARG(uap, whence)) {
1723 	case L_INCR:
1724 		if (noneg &&
1725 		    (fp->f_offset < 0 ||
1726 		     (offset > 0 && fp->f_offset > OFF_MAX - offset)))
1727 			return (EOVERFLOW);
1728 		offset += fp->f_offset;
1729 		break;
1730 	case L_XTND:
1731 		error = VOP_GETATTR(vp, &vattr, cred, td);
1732 		if (error)
1733 			return (error);
1734 		if (noneg &&
1735 		    (vattr.va_size > OFF_MAX ||
1736 		     (offset > 0 && vattr.va_size > OFF_MAX - offset)))
1737 			return (EOVERFLOW);
1738 		offset += vattr.va_size;
1739 		break;
1740 	case L_SET:
1741 		break;
1742 	default:
1743 		fdrop(fp, td);
1744 		return (EINVAL);
1745 	}
1746 	if (noneg && offset < 0)
1747 		return (EINVAL);
1748 	fp->f_offset = offset;
1749 	*(off_t *)(td->td_retval) = fp->f_offset;
1750 	fdrop(fp, td);
1751 	return (0);
1752 }
1753 
1754 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
1755 /*
1756  * Reposition read/write file offset.
1757  */
1758 #ifndef _SYS_SYSPROTO_H_
1759 struct olseek_args {
1760 	int	fd;
1761 	long	offset;
1762 	int	whence;
1763 };
1764 #endif
1765 int
1766 olseek(td, uap)
1767 	struct thread *td;
1768 	register struct olseek_args /* {
1769 		syscallarg(int) fd;
1770 		syscallarg(long) offset;
1771 		syscallarg(int) whence;
1772 	} */ *uap;
1773 {
1774 	struct lseek_args /* {
1775 		syscallarg(int) fd;
1776 		syscallarg(int) pad;
1777 		syscallarg(off_t) offset;
1778 		syscallarg(int) whence;
1779 	} */ nuap;
1780 	int error;
1781 
1782 	SCARG(&nuap, fd) = SCARG(uap, fd);
1783 	SCARG(&nuap, offset) = SCARG(uap, offset);
1784 	SCARG(&nuap, whence) = SCARG(uap, whence);
1785 	error = lseek(td, &nuap);
1786 	return (error);
1787 }
1788 #endif /* COMPAT_43 */
1789 
1790 /*
1791  * Check access permissions using passed credentials.
1792  */
1793 static int
1794 vn_access(vp, user_flags, cred, td)
1795 	struct vnode	*vp;
1796 	int		user_flags;
1797 	struct ucred	*cred;
1798 	struct thread	*td;
1799 {
1800 	int error, flags;
1801 
1802 	/* Flags == 0 means only check for existence. */
1803 	error = 0;
1804 	if (user_flags) {
1805 		flags = 0;
1806 		if (user_flags & R_OK)
1807 			flags |= VREAD;
1808 		if (user_flags & W_OK)
1809 			flags |= VWRITE;
1810 		if (user_flags & X_OK)
1811 			flags |= VEXEC;
1812 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1813 			error = VOP_ACCESS(vp, flags, cred, td);
1814 	}
1815 	return (error);
1816 }
1817 
1818 /*
1819  * Check access permissions using "real" credentials.
1820  */
1821 #ifndef _SYS_SYSPROTO_H_
1822 struct access_args {
1823 	char	*path;
1824 	int	flags;
1825 };
1826 #endif
1827 int
1828 access(td, uap)
1829 	struct thread *td;
1830 	register struct access_args /* {
1831 		syscallarg(char *) path;
1832 		syscallarg(int) flags;
1833 	} */ *uap;
1834 {
1835 	struct ucred *cred, *tmpcred;
1836 	register struct vnode *vp;
1837 	int error;
1838 	struct nameidata nd;
1839 
1840 	cred = td->td_proc->p_ucred;
1841 	/*
1842 	 * Create and modify a temporary credential instead of one that
1843 	 * is potentially shared.  This could also mess up socket
1844 	 * buffer accounting which can run in an interrupt context.
1845 	 *
1846 	 * XXX - Depending on how "threads" are finally implemented, it
1847 	 * may be better to explicitly pass the credential to namei()
1848 	 * rather than to modify the potentially shared process structure.
1849 	 */
1850 	tmpcred = crdup(cred);
1851 	tmpcred->cr_uid = cred->cr_ruid;
1852 	tmpcred->cr_groups[0] = cred->cr_rgid;
1853 	td->td_proc->p_ucred = tmpcred;
1854 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1855 	    SCARG(uap, path), td);
1856 	if ((error = namei(&nd)) != 0)
1857 		goto out1;
1858 	vp = nd.ni_vp;
1859 
1860 	error = vn_access(vp, SCARG(uap, flags), tmpcred, td);
1861 	NDFREE(&nd, NDF_ONLY_PNBUF);
1862 	vput(vp);
1863 out1:
1864 	td->td_proc->p_ucred = cred;
1865 	crfree(tmpcred);
1866 	return (error);
1867 }
1868 
1869 /*
1870  * Check access permissions using "effective" credentials.
1871  */
1872 #ifndef _SYS_SYSPROTO_H_
1873 struct eaccess_args {
1874 	char	*path;
1875 	int	flags;
1876 };
1877 #endif
1878 int
1879 eaccess(td, uap)
1880 	struct thread *td;
1881 	register struct eaccess_args /* {
1882 		syscallarg(char *) path;
1883 		syscallarg(int) flags;
1884 	} */ *uap;
1885 {
1886 	struct nameidata nd;
1887 	struct vnode *vp;
1888 	int error;
1889 
1890 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1891 	    SCARG(uap, path), td);
1892 	if ((error = namei(&nd)) != 0)
1893 		return (error);
1894 	vp = nd.ni_vp;
1895 
1896 	error = vn_access(vp, SCARG(uap, flags), td->td_proc->p_ucred, td);
1897 	NDFREE(&nd, NDF_ONLY_PNBUF);
1898 	vput(vp);
1899 	return (error);
1900 }
1901 
1902 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
1903 /*
1904  * Get file status; this version follows links.
1905  */
1906 #ifndef _SYS_SYSPROTO_H_
1907 struct ostat_args {
1908 	char	*path;
1909 	struct ostat *ub;
1910 };
1911 #endif
1912 /* ARGSUSED */
1913 int
1914 ostat(td, uap)
1915 	struct thread *td;
1916 	register struct ostat_args /* {
1917 		syscallarg(char *) path;
1918 		syscallarg(struct ostat *) ub;
1919 	} */ *uap;
1920 {
1921 	struct stat sb;
1922 	struct ostat osb;
1923 	int error;
1924 	struct nameidata nd;
1925 
1926 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1927 	    SCARG(uap, path), td);
1928 	if ((error = namei(&nd)) != 0)
1929 		return (error);
1930 	NDFREE(&nd, NDF_ONLY_PNBUF);
1931 	error = vn_stat(nd.ni_vp, &sb, td);
1932 	vput(nd.ni_vp);
1933 	if (error)
1934 		return (error);
1935 	cvtstat(&sb, &osb);
1936 	error = copyout((caddr_t)&osb, (caddr_t)SCARG(uap, ub), sizeof (osb));
1937 	return (error);
1938 }
1939 
1940 /*
1941  * Get file status; this version does not follow links.
1942  */
1943 #ifndef _SYS_SYSPROTO_H_
1944 struct olstat_args {
1945 	char	*path;
1946 	struct ostat *ub;
1947 };
1948 #endif
1949 /* ARGSUSED */
1950 int
1951 olstat(td, uap)
1952 	struct thread *td;
1953 	register struct olstat_args /* {
1954 		syscallarg(char *) path;
1955 		syscallarg(struct ostat *) ub;
1956 	} */ *uap;
1957 {
1958 	struct vnode *vp;
1959 	struct stat sb;
1960 	struct ostat osb;
1961 	int error;
1962 	struct nameidata nd;
1963 
1964 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
1965 	    SCARG(uap, path), td);
1966 	if ((error = namei(&nd)) != 0)
1967 		return (error);
1968 	vp = nd.ni_vp;
1969 	error = vn_stat(vp, &sb, td);
1970 	NDFREE(&nd, NDF_ONLY_PNBUF);
1971 	vput(vp);
1972 	if (error)
1973 		return (error);
1974 	cvtstat(&sb, &osb);
1975 	error = copyout((caddr_t)&osb, (caddr_t)SCARG(uap, ub), sizeof (osb));
1976 	return (error);
1977 }
1978 
1979 /*
1980  * Convert from an old to a new stat structure.
1981  */
1982 void
1983 cvtstat(st, ost)
1984 	struct stat *st;
1985 	struct ostat *ost;
1986 {
1987 
1988 	ost->st_dev = st->st_dev;
1989 	ost->st_ino = st->st_ino;
1990 	ost->st_mode = st->st_mode;
1991 	ost->st_nlink = st->st_nlink;
1992 	ost->st_uid = st->st_uid;
1993 	ost->st_gid = st->st_gid;
1994 	ost->st_rdev = st->st_rdev;
1995 	if (st->st_size < (quad_t)1 << 32)
1996 		ost->st_size = st->st_size;
1997 	else
1998 		ost->st_size = -2;
1999 	ost->st_atime = st->st_atime;
2000 	ost->st_mtime = st->st_mtime;
2001 	ost->st_ctime = st->st_ctime;
2002 	ost->st_blksize = st->st_blksize;
2003 	ost->st_blocks = st->st_blocks;
2004 	ost->st_flags = st->st_flags;
2005 	ost->st_gen = st->st_gen;
2006 }
2007 #endif /* COMPAT_43 || COMPAT_SUNOS */
2008 
2009 /*
2010  * Get file status; this version follows links.
2011  */
2012 #ifndef _SYS_SYSPROTO_H_
2013 struct stat_args {
2014 	char	*path;
2015 	struct stat *ub;
2016 };
2017 #endif
2018 /* ARGSUSED */
2019 int
2020 stat(td, uap)
2021 	struct thread *td;
2022 	register struct stat_args /* {
2023 		syscallarg(char *) path;
2024 		syscallarg(struct stat *) ub;
2025 	} */ *uap;
2026 {
2027 	struct stat sb;
2028 	int error;
2029 	struct nameidata nd;
2030 
2031 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2032 	    SCARG(uap, path), td);
2033 	if ((error = namei(&nd)) != 0)
2034 		return (error);
2035 	error = vn_stat(nd.ni_vp, &sb, td);
2036 	NDFREE(&nd, NDF_ONLY_PNBUF);
2037 	vput(nd.ni_vp);
2038 	if (error)
2039 		return (error);
2040 	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
2041 	return (error);
2042 }
2043 
2044 /*
2045  * Get file status; this version does not follow links.
2046  */
2047 #ifndef _SYS_SYSPROTO_H_
2048 struct lstat_args {
2049 	char	*path;
2050 	struct stat *ub;
2051 };
2052 #endif
2053 /* ARGSUSED */
2054 int
2055 lstat(td, uap)
2056 	struct thread *td;
2057 	register struct lstat_args /* {
2058 		syscallarg(char *) path;
2059 		syscallarg(struct stat *) ub;
2060 	} */ *uap;
2061 {
2062 	int error;
2063 	struct vnode *vp;
2064 	struct stat sb;
2065 	struct nameidata nd;
2066 
2067 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2068 	    SCARG(uap, path), td);
2069 	if ((error = namei(&nd)) != 0)
2070 		return (error);
2071 	vp = nd.ni_vp;
2072 	error = vn_stat(vp, &sb, td);
2073 	NDFREE(&nd, NDF_ONLY_PNBUF);
2074 	vput(vp);
2075 	if (error)
2076 		return (error);
2077 	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
2078 	return (error);
2079 }
2080 
2081 /*
2082  * Implementation of the NetBSD stat() function.
2083  * XXX This should probably be collapsed with the FreeBSD version,
2084  * as the differences are only due to vn_stat() clearing spares at
2085  * the end of the structures.  vn_stat could be split to avoid this,
2086  * and thus collapse the following to close to zero code.
2087  */
2088 void
2089 cvtnstat(sb, nsb)
2090 	struct stat *sb;
2091 	struct nstat *nsb;
2092 {
2093 	nsb->st_dev = sb->st_dev;
2094 	nsb->st_ino = sb->st_ino;
2095 	nsb->st_mode = sb->st_mode;
2096 	nsb->st_nlink = sb->st_nlink;
2097 	nsb->st_uid = sb->st_uid;
2098 	nsb->st_gid = sb->st_gid;
2099 	nsb->st_rdev = sb->st_rdev;
2100 	nsb->st_atimespec = sb->st_atimespec;
2101 	nsb->st_mtimespec = sb->st_mtimespec;
2102 	nsb->st_ctimespec = sb->st_ctimespec;
2103 	nsb->st_size = sb->st_size;
2104 	nsb->st_blocks = sb->st_blocks;
2105 	nsb->st_blksize = sb->st_blksize;
2106 	nsb->st_flags = sb->st_flags;
2107 	nsb->st_gen = sb->st_gen;
2108 	nsb->st_qspare[0] = sb->st_qspare[0];
2109 	nsb->st_qspare[1] = sb->st_qspare[1];
2110 }
2111 
2112 #ifndef _SYS_SYSPROTO_H_
2113 struct nstat_args {
2114 	char	*path;
2115 	struct nstat *ub;
2116 };
2117 #endif
2118 /* ARGSUSED */
2119 int
2120 nstat(td, uap)
2121 	struct thread *td;
2122 	register struct nstat_args /* {
2123 		syscallarg(char *) path;
2124 		syscallarg(struct nstat *) ub;
2125 	} */ *uap;
2126 {
2127 	struct stat sb;
2128 	struct nstat nsb;
2129 	int error;
2130 	struct nameidata nd;
2131 
2132 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2133 	    SCARG(uap, path), td);
2134 	if ((error = namei(&nd)) != 0)
2135 		return (error);
2136 	NDFREE(&nd, NDF_ONLY_PNBUF);
2137 	error = vn_stat(nd.ni_vp, &sb, td);
2138 	vput(nd.ni_vp);
2139 	if (error)
2140 		return (error);
2141 	cvtnstat(&sb, &nsb);
2142 	error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb));
2143 	return (error);
2144 }
2145 
2146 /*
2147  * NetBSD lstat.  Get file status; this version does not follow links.
2148  */
2149 #ifndef _SYS_SYSPROTO_H_
2150 struct lstat_args {
2151 	char	*path;
2152 	struct stat *ub;
2153 };
2154 #endif
2155 /* ARGSUSED */
2156 int
2157 nlstat(td, uap)
2158 	struct thread *td;
2159 	register struct nlstat_args /* {
2160 		syscallarg(char *) path;
2161 		syscallarg(struct nstat *) ub;
2162 	} */ *uap;
2163 {
2164 	int error;
2165 	struct vnode *vp;
2166 	struct stat sb;
2167 	struct nstat nsb;
2168 	struct nameidata nd;
2169 
2170 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2171 	    SCARG(uap, path), td);
2172 	if ((error = namei(&nd)) != 0)
2173 		return (error);
2174 	vp = nd.ni_vp;
2175 	NDFREE(&nd, NDF_ONLY_PNBUF);
2176 	error = vn_stat(vp, &sb, td);
2177 	vput(vp);
2178 	if (error)
2179 		return (error);
2180 	cvtnstat(&sb, &nsb);
2181 	error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb));
2182 	return (error);
2183 }
2184 
2185 /*
2186  * Get configurable pathname variables.
2187  */
2188 #ifndef _SYS_SYSPROTO_H_
2189 struct pathconf_args {
2190 	char	*path;
2191 	int	name;
2192 };
2193 #endif
2194 /* ARGSUSED */
2195 int
2196 pathconf(td, uap)
2197 	struct thread *td;
2198 	register struct pathconf_args /* {
2199 		syscallarg(char *) path;
2200 		syscallarg(int) name;
2201 	} */ *uap;
2202 {
2203 	int error;
2204 	struct nameidata nd;
2205 
2206 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2207 	    SCARG(uap, path), td);
2208 	if ((error = namei(&nd)) != 0)
2209 		return (error);
2210 	NDFREE(&nd, NDF_ONLY_PNBUF);
2211 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), td->td_retval);
2212 	vput(nd.ni_vp);
2213 	return (error);
2214 }
2215 
2216 /*
2217  * Return target name of a symbolic link.
2218  */
2219 #ifndef _SYS_SYSPROTO_H_
2220 struct readlink_args {
2221 	char	*path;
2222 	char	*buf;
2223 	int	count;
2224 };
2225 #endif
2226 /* ARGSUSED */
2227 int
2228 readlink(td, uap)
2229 	struct thread *td;
2230 	register struct readlink_args /* {
2231 		syscallarg(char *) path;
2232 		syscallarg(char *) buf;
2233 		syscallarg(int) count;
2234 	} */ *uap;
2235 {
2236 	register struct vnode *vp;
2237 	struct iovec aiov;
2238 	struct uio auio;
2239 	int error;
2240 	struct nameidata nd;
2241 
2242 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
2243 	    SCARG(uap, path), td);
2244 	if ((error = namei(&nd)) != 0)
2245 		return (error);
2246 	NDFREE(&nd, NDF_ONLY_PNBUF);
2247 	vp = nd.ni_vp;
2248 	if (vp->v_type != VLNK)
2249 		error = EINVAL;
2250 	else {
2251 		aiov.iov_base = SCARG(uap, buf);
2252 		aiov.iov_len = SCARG(uap, count);
2253 		auio.uio_iov = &aiov;
2254 		auio.uio_iovcnt = 1;
2255 		auio.uio_offset = 0;
2256 		auio.uio_rw = UIO_READ;
2257 		auio.uio_segflg = UIO_USERSPACE;
2258 		auio.uio_td = td;
2259 		auio.uio_resid = SCARG(uap, count);
2260 		error = VOP_READLINK(vp, &auio, td->td_proc->p_ucred);
2261 	}
2262 	vput(vp);
2263 	td->td_retval[0] = SCARG(uap, count) - auio.uio_resid;
2264 	return (error);
2265 }
2266 
2267 /*
2268  * Common implementation code for chflags() and fchflags().
2269  */
2270 static int
2271 setfflags(td, vp, flags)
2272 	struct thread *td;
2273 	struct vnode *vp;
2274 	int flags;
2275 {
2276 	int error;
2277 	struct mount *mp;
2278 	struct vattr vattr;
2279 
2280 	/*
2281 	 * Prevent non-root users from setting flags on devices.  When
2282 	 * a device is reused, users can retain ownership of the device
2283 	 * if they are allowed to set flags and programs assume that
2284 	 * chown can't fail when done as root.
2285 	 */
2286 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2287 		error = suser_xxx(td->td_proc->p_ucred, td->td_proc,
2288 		    PRISON_ROOT);
2289 		if (error)
2290 			return (error);
2291 	}
2292 
2293 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2294 		return (error);
2295 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
2296 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2297 	VATTR_NULL(&vattr);
2298 	vattr.va_flags = flags;
2299 	error = VOP_SETATTR(vp, &vattr, td->td_proc->p_ucred, td);
2300 	VOP_UNLOCK(vp, 0, td);
2301 	vn_finished_write(mp);
2302 	return (error);
2303 }
2304 
2305 /*
2306  * Change flags of a file given a path name.
2307  */
2308 #ifndef _SYS_SYSPROTO_H_
2309 struct chflags_args {
2310 	char	*path;
2311 	int	flags;
2312 };
2313 #endif
2314 /* ARGSUSED */
2315 int
2316 chflags(td, uap)
2317 	struct thread *td;
2318 	register struct chflags_args /* {
2319 		syscallarg(char *) path;
2320 		syscallarg(int) flags;
2321 	} */ *uap;
2322 {
2323 	int error;
2324 	struct nameidata nd;
2325 
2326 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2327 	if ((error = namei(&nd)) != 0)
2328 		return (error);
2329 	NDFREE(&nd, NDF_ONLY_PNBUF);
2330 	error = setfflags(td, nd.ni_vp, SCARG(uap, flags));
2331 	vrele(nd.ni_vp);
2332 	return error;
2333 }
2334 
2335 /*
2336  * Change flags of a file given a file descriptor.
2337  */
2338 #ifndef _SYS_SYSPROTO_H_
2339 struct fchflags_args {
2340 	int	fd;
2341 	int	flags;
2342 };
2343 #endif
2344 /* ARGSUSED */
2345 int
2346 fchflags(td, uap)
2347 	struct thread *td;
2348 	register struct fchflags_args /* {
2349 		syscallarg(int) fd;
2350 		syscallarg(int) flags;
2351 	} */ *uap;
2352 {
2353 	struct file *fp;
2354 	int error;
2355 
2356 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2357 		return (error);
2358 	error = setfflags(td, (struct vnode *) fp->f_data, SCARG(uap, flags));
2359 	fdrop(fp, td);
2360 	return (error);
2361 }
2362 
2363 /*
2364  * Common implementation code for chmod(), lchmod() and fchmod().
2365  */
2366 static int
2367 setfmode(td, vp, mode)
2368 	struct thread *td;
2369 	struct vnode *vp;
2370 	int mode;
2371 {
2372 	int error;
2373 	struct mount *mp;
2374 	struct vattr vattr;
2375 
2376 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2377 		return (error);
2378 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
2379 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2380 	VATTR_NULL(&vattr);
2381 	vattr.va_mode = mode & ALLPERMS;
2382 	error = VOP_SETATTR(vp, &vattr, td->td_proc->p_ucred, td);
2383 	VOP_UNLOCK(vp, 0, td);
2384 	vn_finished_write(mp);
2385 	return error;
2386 }
2387 
2388 /*
2389  * Change mode of a file given path name.
2390  */
2391 #ifndef _SYS_SYSPROTO_H_
2392 struct chmod_args {
2393 	char	*path;
2394 	int	mode;
2395 };
2396 #endif
2397 /* ARGSUSED */
2398 int
2399 chmod(td, uap)
2400 	struct thread *td;
2401 	register struct chmod_args /* {
2402 		syscallarg(char *) path;
2403 		syscallarg(int) mode;
2404 	} */ *uap;
2405 {
2406 	int error;
2407 	struct nameidata nd;
2408 
2409 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2410 	if ((error = namei(&nd)) != 0)
2411 		return (error);
2412 	NDFREE(&nd, NDF_ONLY_PNBUF);
2413 	error = setfmode(td, nd.ni_vp, SCARG(uap, mode));
2414 	vrele(nd.ni_vp);
2415 	return error;
2416 }
2417 
2418 /*
2419  * Change mode of a file given path name (don't follow links.)
2420  */
2421 #ifndef _SYS_SYSPROTO_H_
2422 struct lchmod_args {
2423 	char	*path;
2424 	int	mode;
2425 };
2426 #endif
2427 /* ARGSUSED */
2428 int
2429 lchmod(td, uap)
2430 	struct thread *td;
2431 	register struct lchmod_args /* {
2432 		syscallarg(char *) path;
2433 		syscallarg(int) mode;
2434 	} */ *uap;
2435 {
2436 	int error;
2437 	struct nameidata nd;
2438 
2439 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2440 	if ((error = namei(&nd)) != 0)
2441 		return (error);
2442 	NDFREE(&nd, NDF_ONLY_PNBUF);
2443 	error = setfmode(td, nd.ni_vp, SCARG(uap, mode));
2444 	vrele(nd.ni_vp);
2445 	return error;
2446 }
2447 
2448 /*
2449  * Change mode of a file given a file descriptor.
2450  */
2451 #ifndef _SYS_SYSPROTO_H_
2452 struct fchmod_args {
2453 	int	fd;
2454 	int	mode;
2455 };
2456 #endif
2457 /* ARGSUSED */
2458 int
2459 fchmod(td, uap)
2460 	struct thread *td;
2461 	register struct fchmod_args /* {
2462 		syscallarg(int) fd;
2463 		syscallarg(int) mode;
2464 	} */ *uap;
2465 {
2466 	struct file *fp;
2467 	struct vnode *vp;
2468 	int error;
2469 
2470 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2471 		return (error);
2472 	vp = (struct vnode *)fp->f_data;
2473 	error = setfmode(td, (struct vnode *)fp->f_data, SCARG(uap, mode));
2474 	fdrop(fp, td);
2475 	return (error);
2476 }
2477 
2478 /*
2479  * Common implementation for chown(), lchown(), and fchown()
2480  */
2481 static int
2482 setfown(td, vp, uid, gid)
2483 	struct thread *td;
2484 	struct vnode *vp;
2485 	uid_t uid;
2486 	gid_t gid;
2487 {
2488 	int error;
2489 	struct mount *mp;
2490 	struct vattr vattr;
2491 
2492 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2493 		return (error);
2494 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
2495 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2496 	VATTR_NULL(&vattr);
2497 	vattr.va_uid = uid;
2498 	vattr.va_gid = gid;
2499 	error = VOP_SETATTR(vp, &vattr, td->td_proc->p_ucred, td);
2500 	VOP_UNLOCK(vp, 0, td);
2501 	vn_finished_write(mp);
2502 	return error;
2503 }
2504 
2505 /*
2506  * Set ownership given a path name.
2507  */
2508 #ifndef _SYS_SYSPROTO_H_
2509 struct chown_args {
2510 	char	*path;
2511 	int	uid;
2512 	int	gid;
2513 };
2514 #endif
2515 /* ARGSUSED */
2516 int
2517 chown(td, uap)
2518 	struct thread *td;
2519 	register struct chown_args /* {
2520 		syscallarg(char *) path;
2521 		syscallarg(int) uid;
2522 		syscallarg(int) gid;
2523 	} */ *uap;
2524 {
2525 	int error;
2526 	struct nameidata nd;
2527 
2528 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2529 	if ((error = namei(&nd)) != 0)
2530 		return (error);
2531 	NDFREE(&nd, NDF_ONLY_PNBUF);
2532 	error = setfown(td, nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid));
2533 	vrele(nd.ni_vp);
2534 	return (error);
2535 }
2536 
2537 /*
2538  * Set ownership given a path name, do not cross symlinks.
2539  */
2540 #ifndef _SYS_SYSPROTO_H_
2541 struct lchown_args {
2542 	char	*path;
2543 	int	uid;
2544 	int	gid;
2545 };
2546 #endif
2547 /* ARGSUSED */
2548 int
2549 lchown(td, uap)
2550 	struct thread *td;
2551 	register struct lchown_args /* {
2552 		syscallarg(char *) path;
2553 		syscallarg(int) uid;
2554 		syscallarg(int) gid;
2555 	} */ *uap;
2556 {
2557 	int error;
2558 	struct nameidata nd;
2559 
2560 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2561 	if ((error = namei(&nd)) != 0)
2562 		return (error);
2563 	NDFREE(&nd, NDF_ONLY_PNBUF);
2564 	error = setfown(td, nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid));
2565 	vrele(nd.ni_vp);
2566 	return (error);
2567 }
2568 
2569 /*
2570  * Set ownership given a file descriptor.
2571  */
2572 #ifndef _SYS_SYSPROTO_H_
2573 struct fchown_args {
2574 	int	fd;
2575 	int	uid;
2576 	int	gid;
2577 };
2578 #endif
2579 /* ARGSUSED */
2580 int
2581 fchown(td, uap)
2582 	struct thread *td;
2583 	register struct fchown_args /* {
2584 		syscallarg(int) fd;
2585 		syscallarg(int) uid;
2586 		syscallarg(int) gid;
2587 	} */ *uap;
2588 {
2589 	struct file *fp;
2590 	struct vnode *vp;
2591 	int error;
2592 
2593 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2594 		return (error);
2595 	vp = (struct vnode *)fp->f_data;
2596 	error = setfown(td, (struct vnode *)fp->f_data,
2597 		SCARG(uap, uid), SCARG(uap, gid));
2598 	fdrop(fp, td);
2599 	return (error);
2600 }
2601 
2602 /*
2603  * Common implementation code for utimes(), lutimes(), and futimes().
2604  */
2605 static int
2606 getutimes(usrtvp, tsp)
2607 	const struct timeval *usrtvp;
2608 	struct timespec *tsp;
2609 {
2610 	struct timeval tv[2];
2611 	int error;
2612 
2613 	if (usrtvp == NULL) {
2614 		microtime(&tv[0]);
2615 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2616 		tsp[1] = tsp[0];
2617 	} else {
2618 		if ((error = copyin(usrtvp, tv, sizeof (tv))) != 0)
2619 			return (error);
2620 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2621 		TIMEVAL_TO_TIMESPEC(&tv[1], &tsp[1]);
2622 	}
2623 	return 0;
2624 }
2625 
2626 /*
2627  * Common implementation code for utimes(), lutimes(), and futimes().
2628  */
2629 static int
2630 setutimes(td, vp, ts, nullflag)
2631 	struct thread *td;
2632 	struct vnode *vp;
2633 	const struct timespec *ts;
2634 	int nullflag;
2635 {
2636 	int error;
2637 	struct mount *mp;
2638 	struct vattr vattr;
2639 
2640 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2641 		return (error);
2642 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
2643 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2644 	VATTR_NULL(&vattr);
2645 	vattr.va_atime = ts[0];
2646 	vattr.va_mtime = ts[1];
2647 	if (nullflag)
2648 		vattr.va_vaflags |= VA_UTIMES_NULL;
2649 	error = VOP_SETATTR(vp, &vattr, td->td_proc->p_ucred, td);
2650 	VOP_UNLOCK(vp, 0, td);
2651 	vn_finished_write(mp);
2652 	return error;
2653 }
2654 
2655 /*
2656  * Set the access and modification times of a file.
2657  */
2658 #ifndef _SYS_SYSPROTO_H_
2659 struct utimes_args {
2660 	char	*path;
2661 	struct	timeval *tptr;
2662 };
2663 #endif
2664 /* ARGSUSED */
2665 int
2666 utimes(td, uap)
2667 	struct thread *td;
2668 	register struct utimes_args /* {
2669 		syscallarg(char *) path;
2670 		syscallarg(struct timeval *) tptr;
2671 	} */ *uap;
2672 {
2673 	struct timespec ts[2];
2674 	struct timeval *usrtvp;
2675 	int error;
2676 	struct nameidata nd;
2677 
2678 	usrtvp = SCARG(uap, tptr);
2679 	if ((error = getutimes(usrtvp, ts)) != 0)
2680 		return (error);
2681 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2682 	if ((error = namei(&nd)) != 0)
2683 		return (error);
2684 	NDFREE(&nd, NDF_ONLY_PNBUF);
2685 	error = setutimes(td, nd.ni_vp, ts, usrtvp == NULL);
2686 	vrele(nd.ni_vp);
2687 	return (error);
2688 }
2689 
2690 /*
2691  * Set the access and modification times of a file.
2692  */
2693 #ifndef _SYS_SYSPROTO_H_
2694 struct lutimes_args {
2695 	char	*path;
2696 	struct	timeval *tptr;
2697 };
2698 #endif
2699 /* ARGSUSED */
2700 int
2701 lutimes(td, uap)
2702 	struct thread *td;
2703 	register struct lutimes_args /* {
2704 		syscallarg(char *) path;
2705 		syscallarg(struct timeval *) tptr;
2706 	} */ *uap;
2707 {
2708 	struct timespec ts[2];
2709 	struct timeval *usrtvp;
2710 	int error;
2711 	struct nameidata nd;
2712 
2713 	usrtvp = SCARG(uap, tptr);
2714 	if ((error = getutimes(usrtvp, ts)) != 0)
2715 		return (error);
2716 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2717 	if ((error = namei(&nd)) != 0)
2718 		return (error);
2719 	NDFREE(&nd, NDF_ONLY_PNBUF);
2720 	error = setutimes(td, nd.ni_vp, ts, usrtvp == NULL);
2721 	vrele(nd.ni_vp);
2722 	return (error);
2723 }
2724 
2725 /*
2726  * Set the access and modification times of a file.
2727  */
2728 #ifndef _SYS_SYSPROTO_H_
2729 struct futimes_args {
2730 	int	fd;
2731 	struct	timeval *tptr;
2732 };
2733 #endif
2734 /* ARGSUSED */
2735 int
2736 futimes(td, uap)
2737 	struct thread *td;
2738 	register struct futimes_args /* {
2739 		syscallarg(int ) fd;
2740 		syscallarg(struct timeval *) tptr;
2741 	} */ *uap;
2742 {
2743 	struct timespec ts[2];
2744 	struct file *fp;
2745 	struct timeval *usrtvp;
2746 	int error;
2747 
2748 	usrtvp = SCARG(uap, tptr);
2749 	if ((error = getutimes(usrtvp, ts)) != 0)
2750 		return (error);
2751 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2752 		return (error);
2753 	error = setutimes(td, (struct vnode *)fp->f_data, ts, usrtvp == NULL);
2754 	fdrop(fp, td);
2755 	return (error);
2756 }
2757 
2758 /*
2759  * Truncate a file given its path name.
2760  */
2761 #ifndef _SYS_SYSPROTO_H_
2762 struct truncate_args {
2763 	char	*path;
2764 	int	pad;
2765 	off_t	length;
2766 };
2767 #endif
2768 /* ARGSUSED */
2769 int
2770 truncate(td, uap)
2771 	struct thread *td;
2772 	register struct truncate_args /* {
2773 		syscallarg(char *) path;
2774 		syscallarg(int) pad;
2775 		syscallarg(off_t) length;
2776 	} */ *uap;
2777 {
2778 	struct mount *mp;
2779 	struct vnode *vp;
2780 	struct vattr vattr;
2781 	int error;
2782 	struct nameidata nd;
2783 
2784 	if (uap->length < 0)
2785 		return(EINVAL);
2786 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
2787 	if ((error = namei(&nd)) != 0)
2788 		return (error);
2789 	vp = nd.ni_vp;
2790 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
2791 		vrele(vp);
2792 		return (error);
2793 	}
2794 	NDFREE(&nd, NDF_ONLY_PNBUF);
2795 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
2796 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2797 	if (vp->v_type == VDIR)
2798 		error = EISDIR;
2799 	else if ((error = vn_writechk(vp)) == 0 &&
2800 	    (error = VOP_ACCESS(vp, VWRITE, td->td_proc->p_ucred, td)) == 0) {
2801 		VATTR_NULL(&vattr);
2802 		vattr.va_size = SCARG(uap, length);
2803 		error = VOP_SETATTR(vp, &vattr, td->td_proc->p_ucred, td);
2804 	}
2805 	vput(vp);
2806 	vn_finished_write(mp);
2807 	return (error);
2808 }
2809 
2810 /*
2811  * Truncate a file given a file descriptor.
2812  */
2813 #ifndef _SYS_SYSPROTO_H_
2814 struct ftruncate_args {
2815 	int	fd;
2816 	int	pad;
2817 	off_t	length;
2818 };
2819 #endif
2820 /* ARGSUSED */
2821 int
2822 ftruncate(td, uap)
2823 	struct thread *td;
2824 	register struct ftruncate_args /* {
2825 		syscallarg(int) fd;
2826 		syscallarg(int) pad;
2827 		syscallarg(off_t) length;
2828 	} */ *uap;
2829 {
2830 	struct mount *mp;
2831 	struct vattr vattr;
2832 	struct vnode *vp;
2833 	struct file *fp;
2834 	int error;
2835 
2836 	if (uap->length < 0)
2837 		return(EINVAL);
2838 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2839 		return (error);
2840 	if ((fp->f_flag & FWRITE) == 0) {
2841 		fdrop(fp, td);
2842 		return (EINVAL);
2843 	}
2844 	vp = (struct vnode *)fp->f_data;
2845 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
2846 		fdrop(fp, td);
2847 		return (error);
2848 	}
2849 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
2850 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2851 	if (vp->v_type == VDIR)
2852 		error = EISDIR;
2853 	else if ((error = vn_writechk(vp)) == 0) {
2854 		VATTR_NULL(&vattr);
2855 		vattr.va_size = SCARG(uap, length);
2856 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
2857 	}
2858 	VOP_UNLOCK(vp, 0, td);
2859 	vn_finished_write(mp);
2860 	fdrop(fp, td);
2861 	return (error);
2862 }
2863 
2864 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
2865 /*
2866  * Truncate a file given its path name.
2867  */
2868 #ifndef _SYS_SYSPROTO_H_
2869 struct otruncate_args {
2870 	char	*path;
2871 	long	length;
2872 };
2873 #endif
2874 /* ARGSUSED */
2875 int
2876 otruncate(td, uap)
2877 	struct thread *td;
2878 	register struct otruncate_args /* {
2879 		syscallarg(char *) path;
2880 		syscallarg(long) length;
2881 	} */ *uap;
2882 {
2883 	struct truncate_args /* {
2884 		syscallarg(char *) path;
2885 		syscallarg(int) pad;
2886 		syscallarg(off_t) length;
2887 	} */ nuap;
2888 
2889 	SCARG(&nuap, path) = SCARG(uap, path);
2890 	SCARG(&nuap, length) = SCARG(uap, length);
2891 	return (truncate(td, &nuap));
2892 }
2893 
2894 /*
2895  * Truncate a file given a file descriptor.
2896  */
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct oftruncate_args {
2899 	int	fd;
2900 	long	length;
2901 };
2902 #endif
2903 /* ARGSUSED */
2904 int
2905 oftruncate(td, uap)
2906 	struct thread *td;
2907 	register struct oftruncate_args /* {
2908 		syscallarg(int) fd;
2909 		syscallarg(long) length;
2910 	} */ *uap;
2911 {
2912 	struct ftruncate_args /* {
2913 		syscallarg(int) fd;
2914 		syscallarg(int) pad;
2915 		syscallarg(off_t) length;
2916 	} */ nuap;
2917 
2918 	SCARG(&nuap, fd) = SCARG(uap, fd);
2919 	SCARG(&nuap, length) = SCARG(uap, length);
2920 	return (ftruncate(td, &nuap));
2921 }
2922 #endif /* COMPAT_43 || COMPAT_SUNOS */
2923 
2924 /*
2925  * Sync an open file.
2926  */
2927 #ifndef _SYS_SYSPROTO_H_
2928 struct fsync_args {
2929 	int	fd;
2930 };
2931 #endif
2932 /* ARGSUSED */
2933 int
2934 fsync(td, uap)
2935 	struct thread *td;
2936 	struct fsync_args /* {
2937 		syscallarg(int) fd;
2938 	} */ *uap;
2939 {
2940 	struct vnode *vp;
2941 	struct mount *mp;
2942 	struct file *fp;
2943 	vm_object_t obj;
2944 	int error;
2945 
2946 	GIANT_REQUIRED;
2947 
2948 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2949 		return (error);
2950 	vp = (struct vnode *)fp->f_data;
2951 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
2952 		fdrop(fp, td);
2953 		return (error);
2954 	}
2955 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2956 	if (VOP_GETVOBJECT(vp, &obj) == 0) {
2957 		vm_object_page_clean(obj, 0, 0, 0);
2958 	}
2959 	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, td);
2960 #ifdef SOFTUPDATES
2961 	if (error == 0 && vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2962 	    error = softdep_fsync(vp);
2963 #endif
2964 
2965 	VOP_UNLOCK(vp, 0, td);
2966 	vn_finished_write(mp);
2967 	fdrop(fp, td);
2968 	return (error);
2969 }
2970 
2971 /*
2972  * Rename files.  Source and destination must either both be directories,
2973  * or both not be directories.  If target is a directory, it must be empty.
2974  */
2975 #ifndef _SYS_SYSPROTO_H_
2976 struct rename_args {
2977 	char	*from;
2978 	char	*to;
2979 };
2980 #endif
2981 /* ARGSUSED */
2982 int
2983 rename(td, uap)
2984 	struct thread *td;
2985 	register struct rename_args /* {
2986 		syscallarg(char *) from;
2987 		syscallarg(char *) to;
2988 	} */ *uap;
2989 {
2990 	struct mount *mp;
2991 	struct vnode *tvp, *fvp, *tdvp;
2992 	struct nameidata fromnd, tond;
2993 	int error;
2994 
2995 	bwillwrite();
2996 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2997 	    SCARG(uap, from), td);
2998 	if ((error = namei(&fromnd)) != 0)
2999 		return (error);
3000 	fvp = fromnd.ni_vp;
3001 	if ((error = vn_start_write(fvp, &mp, V_WAIT | PCATCH)) != 0) {
3002 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3003 		vrele(fromnd.ni_dvp);
3004 		vrele(fvp);
3005 		goto out1;
3006 	}
3007 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | NOOBJ,
3008 	    UIO_USERSPACE, SCARG(uap, to), td);
3009 	if (fromnd.ni_vp->v_type == VDIR)
3010 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3011 	if ((error = namei(&tond)) != 0) {
3012 		/* Translate error code for rename("dir1", "dir2/."). */
3013 		if (error == EISDIR && fvp->v_type == VDIR)
3014 			error = EINVAL;
3015 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3016 		vrele(fromnd.ni_dvp);
3017 		vrele(fvp);
3018 		goto out1;
3019 	}
3020 	tdvp = tond.ni_dvp;
3021 	tvp = tond.ni_vp;
3022 	if (tvp != NULL) {
3023 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3024 			error = ENOTDIR;
3025 			goto out;
3026 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3027 			error = EISDIR;
3028 			goto out;
3029 		}
3030 	}
3031 	if (fvp == tdvp)
3032 		error = EINVAL;
3033 	/*
3034 	 * If source is the same as the destination (that is the
3035 	 * same inode number with the same name in the same directory),
3036 	 * then there is nothing to do.
3037 	 */
3038 	if (fvp == tvp && fromnd.ni_dvp == tdvp &&
3039 	    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3040 	    !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
3041 	      fromnd.ni_cnd.cn_namelen))
3042 		error = -1;
3043 out:
3044 	if (!error) {
3045 		VOP_LEASE(tdvp, td, td->td_proc->p_ucred, LEASE_WRITE);
3046 		if (fromnd.ni_dvp != tdvp) {
3047 			VOP_LEASE(fromnd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
3048 		}
3049 		if (tvp) {
3050 			VOP_LEASE(tvp, td, td->td_proc->p_ucred, LEASE_WRITE);
3051 		}
3052 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3053 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3054 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3055 		NDFREE(&tond, NDF_ONLY_PNBUF);
3056 	} else {
3057 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3058 		NDFREE(&tond, NDF_ONLY_PNBUF);
3059 		if (tdvp == tvp)
3060 			vrele(tdvp);
3061 		else
3062 			vput(tdvp);
3063 		if (tvp)
3064 			vput(tvp);
3065 		vrele(fromnd.ni_dvp);
3066 		vrele(fvp);
3067 	}
3068 	vrele(tond.ni_startdir);
3069 	vn_finished_write(mp);
3070 	ASSERT_VOP_UNLOCKED(fromnd.ni_dvp, "rename");
3071 	ASSERT_VOP_UNLOCKED(fromnd.ni_vp, "rename");
3072 	ASSERT_VOP_UNLOCKED(tond.ni_dvp, "rename");
3073 	ASSERT_VOP_UNLOCKED(tond.ni_vp, "rename");
3074 out1:
3075 	if (fromnd.ni_startdir)
3076 		vrele(fromnd.ni_startdir);
3077 	if (error == -1)
3078 		return (0);
3079 	return (error);
3080 }
3081 
3082 /*
3083  * Make a directory file.
3084  */
3085 #ifndef _SYS_SYSPROTO_H_
3086 struct mkdir_args {
3087 	char	*path;
3088 	int	mode;
3089 };
3090 #endif
3091 /* ARGSUSED */
3092 int
3093 mkdir(td, uap)
3094 	struct thread *td;
3095 	register struct mkdir_args /* {
3096 		syscallarg(char *) path;
3097 		syscallarg(int) mode;
3098 	} */ *uap;
3099 {
3100 
3101 	return vn_mkdir(uap->path, uap->mode, UIO_USERSPACE, td);
3102 }
3103 
3104 int
3105 vn_mkdir(path, mode, segflg, td)
3106 	char *path;
3107 	int mode;
3108 	enum uio_seg segflg;
3109 	struct thread *td;
3110 {
3111 	struct mount *mp;
3112 	struct vnode *vp;
3113 	struct vattr vattr;
3114 	int error;
3115 	struct nameidata nd;
3116 
3117 restart:
3118 	bwillwrite();
3119 	NDINIT(&nd, CREATE, LOCKPARENT, segflg, path, td);
3120 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3121 	if ((error = namei(&nd)) != 0)
3122 		return (error);
3123 	vp = nd.ni_vp;
3124 	if (vp != NULL) {
3125 		NDFREE(&nd, NDF_ONLY_PNBUF);
3126 		vrele(vp);
3127 		vput(nd.ni_dvp);
3128 		return (EEXIST);
3129 	}
3130 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3131 		NDFREE(&nd, NDF_ONLY_PNBUF);
3132 		vput(nd.ni_dvp);
3133 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3134 			return (error);
3135 		goto restart;
3136 	}
3137 	VATTR_NULL(&vattr);
3138 	vattr.va_type = VDIR;
3139 	FILEDESC_LOCK(td->td_proc->p_fd);
3140 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3141 	FILEDESC_UNLOCK(td->td_proc->p_fd);
3142 	VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
3143 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3144 	NDFREE(&nd, NDF_ONLY_PNBUF);
3145 	vput(nd.ni_dvp);
3146 	if (!error)
3147 		vput(nd.ni_vp);
3148 	vn_finished_write(mp);
3149 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mkdir");
3150 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "mkdir");
3151 	return (error);
3152 }
3153 
3154 /*
3155  * Remove a directory file.
3156  */
3157 #ifndef _SYS_SYSPROTO_H_
3158 struct rmdir_args {
3159 	char	*path;
3160 };
3161 #endif
3162 /* ARGSUSED */
3163 int
3164 rmdir(td, uap)
3165 	struct thread *td;
3166 	struct rmdir_args /* {
3167 		syscallarg(char *) path;
3168 	} */ *uap;
3169 {
3170 	struct mount *mp;
3171 	struct vnode *vp;
3172 	int error;
3173 	struct nameidata nd;
3174 
3175 restart:
3176 	bwillwrite();
3177 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3178 	    SCARG(uap, path), td);
3179 	if ((error = namei(&nd)) != 0)
3180 		return (error);
3181 	vp = nd.ni_vp;
3182 	if (vp->v_type != VDIR) {
3183 		error = ENOTDIR;
3184 		goto out;
3185 	}
3186 	/*
3187 	 * No rmdir "." please.
3188 	 */
3189 	if (nd.ni_dvp == vp) {
3190 		error = EINVAL;
3191 		goto out;
3192 	}
3193 	/*
3194 	 * The root of a mounted filesystem cannot be deleted.
3195 	 */
3196 	if (vp->v_flag & VROOT) {
3197 		error = EBUSY;
3198 		goto out;
3199 	}
3200 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3201 		NDFREE(&nd, NDF_ONLY_PNBUF);
3202 		if (nd.ni_dvp == vp)
3203 			vrele(nd.ni_dvp);
3204 		else
3205 			vput(nd.ni_dvp);
3206 		vput(vp);
3207 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3208 			return (error);
3209 		goto restart;
3210 	}
3211 	VOP_LEASE(nd.ni_dvp, td, td->td_proc->p_ucred, LEASE_WRITE);
3212 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
3213 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3214 	vn_finished_write(mp);
3215 out:
3216 	NDFREE(&nd, NDF_ONLY_PNBUF);
3217 	if (nd.ni_dvp == vp)
3218 		vrele(nd.ni_dvp);
3219 	else
3220 		vput(nd.ni_dvp);
3221 	vput(vp);
3222 	ASSERT_VOP_UNLOCKED(nd.ni_dvp, "rmdir");
3223 	ASSERT_VOP_UNLOCKED(nd.ni_vp, "rmdir");
3224 	return (error);
3225 }
3226 
3227 #ifdef COMPAT_43
3228 /*
3229  * Read a block of directory entries in a file system independent format.
3230  */
3231 #ifndef _SYS_SYSPROTO_H_
3232 struct ogetdirentries_args {
3233 	int	fd;
3234 	char	*buf;
3235 	u_int	count;
3236 	long	*basep;
3237 };
3238 #endif
3239 int
3240 ogetdirentries(td, uap)
3241 	struct thread *td;
3242 	register struct ogetdirentries_args /* {
3243 		syscallarg(int) fd;
3244 		syscallarg(char *) buf;
3245 		syscallarg(u_int) count;
3246 		syscallarg(long *) basep;
3247 	} */ *uap;
3248 {
3249 	struct vnode *vp;
3250 	struct file *fp;
3251 	struct uio auio, kuio;
3252 	struct iovec aiov, kiov;
3253 	struct dirent *dp, *edp;
3254 	caddr_t dirbuf;
3255 	int error, eofflag, readcnt;
3256 	long loff;
3257 
3258 	/* XXX arbitrary sanity limit on `count'. */
3259 	if (SCARG(uap, count) > 64 * 1024)
3260 		return (EINVAL);
3261 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
3262 		return (error);
3263 	if ((fp->f_flag & FREAD) == 0) {
3264 		fdrop(fp, td);
3265 		return (EBADF);
3266 	}
3267 	vp = (struct vnode *)fp->f_data;
3268 unionread:
3269 	if (vp->v_type != VDIR) {
3270 		fdrop(fp, td);
3271 		return (EINVAL);
3272 	}
3273 	aiov.iov_base = SCARG(uap, buf);
3274 	aiov.iov_len = SCARG(uap, count);
3275 	auio.uio_iov = &aiov;
3276 	auio.uio_iovcnt = 1;
3277 	auio.uio_rw = UIO_READ;
3278 	auio.uio_segflg = UIO_USERSPACE;
3279 	auio.uio_td = td;
3280 	auio.uio_resid = SCARG(uap, count);
3281 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
3282 	loff = auio.uio_offset = fp->f_offset;
3283 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3284 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3285 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3286 			    NULL, NULL);
3287 			fp->f_offset = auio.uio_offset;
3288 		} else
3289 #	endif
3290 	{
3291 		kuio = auio;
3292 		kuio.uio_iov = &kiov;
3293 		kuio.uio_segflg = UIO_SYSSPACE;
3294 		kiov.iov_len = SCARG(uap, count);
3295 		MALLOC(dirbuf, caddr_t, SCARG(uap, count), M_TEMP, M_WAITOK);
3296 		kiov.iov_base = dirbuf;
3297 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3298 			    NULL, NULL);
3299 		fp->f_offset = kuio.uio_offset;
3300 		if (error == 0) {
3301 			readcnt = SCARG(uap, count) - kuio.uio_resid;
3302 			edp = (struct dirent *)&dirbuf[readcnt];
3303 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3304 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3305 					/*
3306 					 * The expected low byte of
3307 					 * dp->d_namlen is our dp->d_type.
3308 					 * The high MBZ byte of dp->d_namlen
3309 					 * is our dp->d_namlen.
3310 					 */
3311 					dp->d_type = dp->d_namlen;
3312 					dp->d_namlen = 0;
3313 #				else
3314 					/*
3315 					 * The dp->d_type is the high byte
3316 					 * of the expected dp->d_namlen,
3317 					 * so must be zero'ed.
3318 					 */
3319 					dp->d_type = 0;
3320 #				endif
3321 				if (dp->d_reclen > 0) {
3322 					dp = (struct dirent *)
3323 					    ((char *)dp + dp->d_reclen);
3324 				} else {
3325 					error = EIO;
3326 					break;
3327 				}
3328 			}
3329 			if (dp >= edp)
3330 				error = uiomove(dirbuf, readcnt, &auio);
3331 		}
3332 		FREE(dirbuf, M_TEMP);
3333 	}
3334 	VOP_UNLOCK(vp, 0, td);
3335 	if (error) {
3336 		fdrop(fp, td);
3337 		return (error);
3338 	}
3339 	if (SCARG(uap, count) == auio.uio_resid) {
3340 		if (union_dircheckp) {
3341 			error = union_dircheckp(td, &vp, fp);
3342 			if (error == -1)
3343 				goto unionread;
3344 			if (error) {
3345 				fdrop(fp, td);
3346 				return (error);
3347 			}
3348 		}
3349 		if ((vp->v_flag & VROOT) &&
3350 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
3351 			struct vnode *tvp = vp;
3352 			vp = vp->v_mount->mnt_vnodecovered;
3353 			VREF(vp);
3354 			fp->f_data = (caddr_t) vp;
3355 			fp->f_offset = 0;
3356 			vrele(tvp);
3357 			goto unionread;
3358 		}
3359 	}
3360 	error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
3361 	    sizeof(long));
3362 	fdrop(fp, td);
3363 	td->td_retval[0] = SCARG(uap, count) - auio.uio_resid;
3364 	return (error);
3365 }
3366 #endif /* COMPAT_43 */
3367 
3368 /*
3369  * Read a block of directory entries in a file system independent format.
3370  */
3371 #ifndef _SYS_SYSPROTO_H_
3372 struct getdirentries_args {
3373 	int	fd;
3374 	char	*buf;
3375 	u_int	count;
3376 	long	*basep;
3377 };
3378 #endif
3379 int
3380 getdirentries(td, uap)
3381 	struct thread *td;
3382 	register struct getdirentries_args /* {
3383 		syscallarg(int) fd;
3384 		syscallarg(char *) buf;
3385 		syscallarg(u_int) count;
3386 		syscallarg(long *) basep;
3387 	} */ *uap;
3388 {
3389 	struct vnode *vp;
3390 	struct file *fp;
3391 	struct uio auio;
3392 	struct iovec aiov;
3393 	long loff;
3394 	int error, eofflag;
3395 
3396 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
3397 		return (error);
3398 	if ((fp->f_flag & FREAD) == 0) {
3399 		fdrop(fp, td);
3400 		return (EBADF);
3401 	}
3402 	vp = (struct vnode *)fp->f_data;
3403 unionread:
3404 	if (vp->v_type != VDIR) {
3405 		fdrop(fp, td);
3406 		return (EINVAL);
3407 	}
3408 	aiov.iov_base = SCARG(uap, buf);
3409 	aiov.iov_len = SCARG(uap, count);
3410 	auio.uio_iov = &aiov;
3411 	auio.uio_iovcnt = 1;
3412 	auio.uio_rw = UIO_READ;
3413 	auio.uio_segflg = UIO_USERSPACE;
3414 	auio.uio_td = td;
3415 	auio.uio_resid = SCARG(uap, count);
3416 	/* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
3417 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
3418 	loff = auio.uio_offset = fp->f_offset;
3419 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
3420 	fp->f_offset = auio.uio_offset;
3421 	VOP_UNLOCK(vp, 0, td);
3422 	if (error) {
3423 		fdrop(fp, td);
3424 		return (error);
3425 	}
3426 	if (SCARG(uap, count) == auio.uio_resid) {
3427 		if (union_dircheckp) {
3428 			error = union_dircheckp(td, &vp, fp);
3429 			if (error == -1)
3430 				goto unionread;
3431 			if (error) {
3432 				fdrop(fp, td);
3433 				return (error);
3434 			}
3435 		}
3436 		if ((vp->v_flag & VROOT) &&
3437 		    (vp->v_mount->mnt_flag & MNT_UNION)) {
3438 			struct vnode *tvp = vp;
3439 			vp = vp->v_mount->mnt_vnodecovered;
3440 			VREF(vp);
3441 			fp->f_data = (caddr_t) vp;
3442 			fp->f_offset = 0;
3443 			vrele(tvp);
3444 			goto unionread;
3445 		}
3446 	}
3447 	if (SCARG(uap, basep) != NULL) {
3448 		error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
3449 		    sizeof(long));
3450 	}
3451 	td->td_retval[0] = SCARG(uap, count) - auio.uio_resid;
3452 	fdrop(fp, td);
3453 	return (error);
3454 }
3455 #ifndef _SYS_SYSPROTO_H_
3456 struct getdents_args {
3457 	int fd;
3458 	char *buf;
3459 	size_t count;
3460 };
3461 #endif
3462 int
3463 getdents(td, uap)
3464 	struct thread *td;
3465 	register struct getdents_args /* {
3466 		syscallarg(int) fd;
3467 		syscallarg(char *) buf;
3468 		syscallarg(u_int) count;
3469 	} */ *uap;
3470 {
3471 	struct getdirentries_args ap;
3472 	ap.fd = uap->fd;
3473 	ap.buf = uap->buf;
3474 	ap.count = uap->count;
3475 	ap.basep = NULL;
3476 	return getdirentries(td, &ap);
3477 }
3478 
3479 /*
3480  * Set the mode mask for creation of filesystem nodes.
3481  *
3482  * MP SAFE
3483  */
3484 #ifndef _SYS_SYSPROTO_H_
3485 struct umask_args {
3486 	int	newmask;
3487 };
3488 #endif
3489 int
3490 umask(td, uap)
3491 	struct thread *td;
3492 	struct umask_args /* {
3493 		syscallarg(int) newmask;
3494 	} */ *uap;
3495 {
3496 	register struct filedesc *fdp;
3497 
3498 	FILEDESC_LOCK(td->td_proc->p_fd);
3499 	fdp = td->td_proc->p_fd;
3500 	td->td_retval[0] = fdp->fd_cmask;
3501 	fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
3502 	FILEDESC_UNLOCK(td->td_proc->p_fd);
3503 	return (0);
3504 }
3505 
3506 /*
3507  * Void all references to file by ripping underlying filesystem
3508  * away from vnode.
3509  */
3510 #ifndef _SYS_SYSPROTO_H_
3511 struct revoke_args {
3512 	char	*path;
3513 };
3514 #endif
3515 /* ARGSUSED */
3516 int
3517 revoke(td, uap)
3518 	struct thread *td;
3519 	register struct revoke_args /* {
3520 		syscallarg(char *) path;
3521 	} */ *uap;
3522 {
3523 	struct mount *mp;
3524 	struct vnode *vp;
3525 	struct vattr vattr;
3526 	int error;
3527 	struct nameidata nd;
3528 
3529 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
3530 	if ((error = namei(&nd)) != 0)
3531 		return (error);
3532 	vp = nd.ni_vp;
3533 	NDFREE(&nd, NDF_ONLY_PNBUF);
3534 	if (vp->v_type != VCHR) {
3535 		error = EINVAL;
3536 		goto out;
3537 	}
3538 	error = VOP_GETATTR(vp, &vattr, td->td_proc->p_ucred, td);
3539 	if (error)
3540 		goto out;
3541 	if (td->td_proc->p_ucred->cr_uid != vattr.va_uid) {
3542 		error = suser_xxx(0, td->td_proc, PRISON_ROOT);
3543 		if (error)
3544 			goto out;
3545 	}
3546 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3547 		goto out;
3548 	if (vcount(vp) > 1)
3549 		VOP_REVOKE(vp, REVOKEALL);
3550 	vn_finished_write(mp);
3551 out:
3552 	vrele(vp);
3553 	return (error);
3554 }
3555 
3556 /*
3557  * Convert a user file descriptor to a kernel file entry.
3558  * The file entry is locked upon returning.
3559  */
3560 int
3561 getvnode(fdp, fd, fpp)
3562 	struct filedesc *fdp;
3563 	int fd;
3564 	struct file **fpp;
3565 {
3566 	int error;
3567 	struct file *fp;
3568 
3569 	fp = NULL;
3570 	if (fdp == NULL)
3571 		error = EBADF;
3572 	else {
3573 		FILEDESC_LOCK(fdp);
3574 		if ((u_int)fd >= fdp->fd_nfiles ||
3575 		    (fp = fdp->fd_ofiles[fd]) == NULL)
3576 			error = EBADF;
3577 		else if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO) {
3578 			fp = NULL;
3579 			error = EINVAL;
3580 		} else {
3581 			fhold(fp);
3582 			error = 0;
3583 		}
3584 		FILEDESC_UNLOCK(fdp);
3585 	}
3586 	*fpp = fp;
3587 	return (error);
3588 }
3589 /*
3590  * Get (NFS) file handle
3591  */
3592 #ifndef _SYS_SYSPROTO_H_
3593 struct getfh_args {
3594 	char	*fname;
3595 	fhandle_t *fhp;
3596 };
3597 #endif
3598 int
3599 getfh(td, uap)
3600 	struct thread *td;
3601 	register struct getfh_args *uap;
3602 {
3603 	struct nameidata nd;
3604 	fhandle_t fh;
3605 	register struct vnode *vp;
3606 	int error;
3607 
3608 	/*
3609 	 * Must be super user
3610 	 */
3611 	error = suser_td(td);
3612 	if (error)
3613 		return (error);
3614 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, td);
3615 	error = namei(&nd);
3616 	if (error)
3617 		return (error);
3618 	NDFREE(&nd, NDF_ONLY_PNBUF);
3619 	vp = nd.ni_vp;
3620 	bzero(&fh, sizeof(fh));
3621 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3622 	error = VFS_VPTOFH(vp, &fh.fh_fid);
3623 	vput(vp);
3624 	if (error)
3625 		return (error);
3626 	error = copyout(&fh, uap->fhp, sizeof (fh));
3627 	return (error);
3628 }
3629 
3630 /*
3631  * syscall for the rpc.lockd to use to translate a NFS file handle into
3632  * an open descriptor.
3633  *
3634  * warning: do not remove the suser() call or this becomes one giant
3635  * security hole.
3636  */
3637 #ifndef _SYS_SYSPROTO_H_
3638 struct fhopen_args {
3639 	const struct fhandle *u_fhp;
3640 	int flags;
3641 };
3642 #endif
3643 int
3644 fhopen(td, uap)
3645 	struct thread *td;
3646 	struct fhopen_args /* {
3647 		syscallarg(const struct fhandle *) u_fhp;
3648 		syscallarg(int) flags;
3649 	} */ *uap;
3650 {
3651 	struct proc *p = td->td_proc;
3652 	struct mount *mp;
3653 	struct vnode *vp;
3654 	struct fhandle fhp;
3655 	struct vattr vat;
3656 	struct vattr *vap = &vat;
3657 	struct flock lf;
3658 	struct file *fp;
3659 	register struct filedesc *fdp = p->p_fd;
3660 	int fmode, mode, error, type;
3661 	struct file *nfp;
3662 	int indx;
3663 
3664 	/*
3665 	 * Must be super user
3666 	 */
3667 	error = suser_td(td);
3668 	if (error)
3669 		return (error);
3670 
3671 	fmode = FFLAGS(SCARG(uap, flags));
3672 	/* why not allow a non-read/write open for our lockd? */
3673 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3674 		return (EINVAL);
3675 	error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp));
3676 	if (error)
3677 		return(error);
3678 	/* find the mount point */
3679 	mp = vfs_getvfs(&fhp.fh_fsid);
3680 	if (mp == NULL)
3681 		return (ESTALE);
3682 	/* now give me my vnode, it gets returned to me locked */
3683 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3684 	if (error)
3685 		return (error);
3686  	/*
3687 	 * from now on we have to make sure not
3688 	 * to forget about the vnode
3689 	 * any error that causes an abort must vput(vp)
3690 	 * just set error = err and 'goto bad;'.
3691 	 */
3692 
3693 	/*
3694 	 * from vn_open
3695 	 */
3696 	if (vp->v_type == VLNK) {
3697 		error = EMLINK;
3698 		goto bad;
3699 	}
3700 	if (vp->v_type == VSOCK) {
3701 		error = EOPNOTSUPP;
3702 		goto bad;
3703 	}
3704 	mode = 0;
3705 	if (fmode & (FWRITE | O_TRUNC)) {
3706 		if (vp->v_type == VDIR) {
3707 			error = EISDIR;
3708 			goto bad;
3709 		}
3710 		error = vn_writechk(vp);
3711 		if (error)
3712 			goto bad;
3713 		mode |= VWRITE;
3714 	}
3715 	if (fmode & FREAD)
3716 		mode |= VREAD;
3717 	if (mode) {
3718 		error = VOP_ACCESS(vp, mode, p->p_ucred, td);
3719 		if (error)
3720 			goto bad;
3721 	}
3722 	if (fmode & O_TRUNC) {
3723 		VOP_UNLOCK(vp, 0, td);				/* XXX */
3724 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
3725 			vrele(vp);
3726 			return (error);
3727 		}
3728 		VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
3729 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);	/* XXX */
3730 		VATTR_NULL(vap);
3731 		vap->va_size = 0;
3732 		error = VOP_SETATTR(vp, vap, p->p_ucred, td);
3733 		vn_finished_write(mp);
3734 		if (error)
3735 			goto bad;
3736 	}
3737 	error = VOP_OPEN(vp, fmode, p->p_ucred, td);
3738 	if (error)
3739 		goto bad;
3740 	/*
3741 	 * Make sure that a VM object is created for VMIO support.
3742 	 */
3743 	if (vn_canvmio(vp) == TRUE) {
3744 		if ((error = vfs_object_create(vp, td, p->p_ucred)) != 0)
3745 			goto bad;
3746 	}
3747 	if (fmode & FWRITE)
3748 		vp->v_writecount++;
3749 
3750 	/*
3751 	 * end of vn_open code
3752 	 */
3753 
3754 	if ((error = falloc(td, &nfp, &indx)) != 0) {
3755 		if (fmode & FWRITE)
3756 			vp->v_writecount--;
3757 		goto bad;
3758 	}
3759 	fp = nfp;
3760 
3761 	/*
3762 	 * Hold an extra reference to avoid having fp ripped out
3763 	 * from under us while we block in the lock op
3764 	 */
3765 	fhold(fp);
3766 	nfp->f_data = (caddr_t)vp;
3767 	nfp->f_flag = fmode & FMASK;
3768 	nfp->f_ops = &vnops;
3769 	nfp->f_type = DTYPE_VNODE;
3770 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
3771 		lf.l_whence = SEEK_SET;
3772 		lf.l_start = 0;
3773 		lf.l_len = 0;
3774 		if (fmode & O_EXLOCK)
3775 			lf.l_type = F_WRLCK;
3776 		else
3777 			lf.l_type = F_RDLCK;
3778 		type = F_FLOCK;
3779 		if ((fmode & FNONBLOCK) == 0)
3780 			type |= F_WAIT;
3781 		VOP_UNLOCK(vp, 0, td);
3782 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3783 			/*
3784 			 * The lock request failed.  Normally close the
3785 			 * descriptor but handle the case where someone might
3786 			 * have dup()d or close()d it when we weren't looking.
3787 			 */
3788 			FILEDESC_LOCK(fdp);
3789 			if (fdp->fd_ofiles[indx] == fp) {
3790 				fdp->fd_ofiles[indx] = NULL;
3791 				FILEDESC_UNLOCK(fdp);
3792 				fdrop(fp, td);
3793 			} else
3794 				FILEDESC_UNLOCK(fdp);
3795 			/*
3796 			 * release our private reference
3797 			 */
3798 			fdrop(fp, td);
3799 			return(error);
3800 		}
3801 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
3802 		fp->f_flag |= FHASLOCK;
3803 	}
3804 	if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
3805 		vfs_object_create(vp, td, p->p_ucred);
3806 
3807 	VOP_UNLOCK(vp, 0, td);
3808 	fdrop(fp, td);
3809 	td->td_retval[0] = indx;
3810 	return (0);
3811 
3812 bad:
3813 	vput(vp);
3814 	return (error);
3815 }
3816 
3817 /*
3818  * Stat an (NFS) file handle.
3819  */
3820 #ifndef _SYS_SYSPROTO_H_
3821 struct fhstat_args {
3822 	struct fhandle *u_fhp;
3823 	struct stat *sb;
3824 };
3825 #endif
3826 int
3827 fhstat(td, uap)
3828 	struct thread *td;
3829 	register struct fhstat_args /* {
3830 		syscallarg(struct fhandle *) u_fhp;
3831 		syscallarg(struct stat *) sb;
3832 	} */ *uap;
3833 {
3834 	struct stat sb;
3835 	fhandle_t fh;
3836 	struct mount *mp;
3837 	struct vnode *vp;
3838 	int error;
3839 
3840 	/*
3841 	 * Must be super user
3842 	 */
3843 	error = suser_td(td);
3844 	if (error)
3845 		return (error);
3846 
3847 	error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t));
3848 	if (error)
3849 		return (error);
3850 
3851 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3852 		return (ESTALE);
3853 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3854 		return (error);
3855 	error = vn_stat(vp, &sb, td);
3856 	vput(vp);
3857 	if (error)
3858 		return (error);
3859 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
3860 	return (error);
3861 }
3862 
3863 /*
3864  * Implement fstatfs() for (NFS) file handles.
3865  */
3866 #ifndef _SYS_SYSPROTO_H_
3867 struct fhstatfs_args {
3868 	struct fhandle *u_fhp;
3869 	struct statfs *buf;
3870 };
3871 #endif
3872 int
3873 fhstatfs(td, uap)
3874 	struct thread *td;
3875 	struct fhstatfs_args /* {
3876 		syscallarg(struct fhandle) *u_fhp;
3877 		syscallarg(struct statfs) *buf;
3878 	} */ *uap;
3879 {
3880 	struct statfs *sp;
3881 	struct mount *mp;
3882 	struct vnode *vp;
3883 	struct statfs sb;
3884 	fhandle_t fh;
3885 	int error;
3886 
3887 	/*
3888 	 * Must be super user
3889 	 */
3890 	error = suser_td(td);
3891 	if (error)
3892 		return (error);
3893 
3894 	if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0)
3895 		return (error);
3896 
3897 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3898 		return (ESTALE);
3899 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3900 		return (error);
3901 	mp = vp->v_mount;
3902 	sp = &mp->mnt_stat;
3903 	vput(vp);
3904 	if ((error = VFS_STATFS(mp, sp, td)) != 0)
3905 		return (error);
3906 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3907 	if (suser_xxx(td->td_proc->p_ucred, 0, 0)) {
3908 		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
3909 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3910 		sp = &sb;
3911 	}
3912 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
3913 }
3914 
3915 /*
3916  * Syscall to push extended attribute configuration information into the
3917  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
3918  * a command (int cmd), and attribute name and misc data.  For now, the
3919  * attribute name is left in userspace for consumption by the VFS_op.
3920  * It will probably be changed to be copied into sysspace by the
3921  * syscall in the future, once issues with various consumers of the
3922  * attribute code have raised their hands.
3923  *
3924  * Currently this is used only by UFS Extended Attributes.
3925  */
3926 int
3927 extattrctl(td, uap)
3928 	struct thread *td;
3929 	struct extattrctl_args *uap;
3930 {
3931 	struct vnode *filename_vp;
3932 	struct nameidata nd;
3933 	struct mount *mp;
3934 	char attrname[EXTATTR_MAXNAMELEN];
3935 	int error;
3936 
3937 	/*
3938 	 * SCARG(uap, attrname) not always defined.  We check again later
3939 	 * when we invoke the VFS call so as to pass in NULL there if needed.
3940 	 */
3941 	if (SCARG(uap, attrname) != NULL) {
3942 		error = copyinstr(SCARG(uap, attrname), attrname,
3943 		    EXTATTR_MAXNAMELEN, NULL);
3944 		if (error)
3945 			return (error);
3946 	}
3947 
3948 	/*
3949 	 * SCARG(uap, filename) not always defined.  If it is, grab
3950 	 * a vnode lock, which VFS_EXTATTRCTL() will later release.
3951 	 */
3952 	filename_vp = NULL;
3953 	if (SCARG(uap, filename) != NULL) {
3954 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3955 		    SCARG(uap, filename), td);
3956 		if ((error = namei(&nd)) != 0)
3957 			return (error);
3958 		filename_vp = nd.ni_vp;
3959 		NDFREE(&nd, NDF_NO_VP_RELE | NDF_NO_VP_UNLOCK);
3960 	}
3961 
3962 	/* SCARG(uap, path) always defined. */
3963 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
3964 	if ((error = namei(&nd)) != 0)
3965 		return (error);
3966 	error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH);
3967 	NDFREE(&nd, 0);
3968 	if (error) {
3969 		if (filename_vp)
3970 			vrele(filename_vp);
3971 		return (error);
3972 	}
3973 
3974 	if (SCARG(uap, attrname) != NULL) {
3975 		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), filename_vp,
3976 		    SCARG(uap, attrnamespace), attrname, td);
3977 	} else {
3978 		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), filename_vp,
3979 		    SCARG(uap, attrnamespace), NULL, td);
3980 	}
3981 
3982 	vn_finished_write(mp);
3983 	/*
3984 	 * VFS_EXTATTRCTL will have unlocked, but not de-ref'd,
3985 	 * filename_vp, so vrele it if it is defined.
3986 	 */
3987 	if (filename_vp != NULL)
3988 		vrele(filename_vp);
3989 
3990 	return (error);
3991 }
3992 
3993 /*
3994  * extattr_set_vp(): Set a named extended attribute on a file or directory
3995  *
3996  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
3997  *            kernelspace string pointer "attrname",
3998  *            userspace iovec array pointer "iovp", unsigned int iovcnt
3999  *            proc "p"
4000  * Returns: 0 on success, an error number otherwise
4001  * Locks: none
4002  * References: vp must be a valid reference for the duration of the call
4003  */
4004 static int
4005 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
4006     struct iovec *iovp, unsigned iovcnt, struct thread *td)
4007 {
4008 	struct mount *mp;
4009 	struct uio auio;
4010 	struct iovec *iov, *needfree = NULL, aiov[UIO_SMALLIOV];
4011 	u_int iovlen, cnt;
4012 	int error, i;
4013 
4014 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
4015 		return (error);
4016 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
4017 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
4018 
4019 	iovlen = iovcnt * sizeof(struct iovec);
4020 	if (iovcnt > UIO_SMALLIOV) {
4021 		if (iovcnt > UIO_MAXIOV) {
4022 			error = EINVAL;
4023 			goto done;
4024 		}
4025 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
4026 		needfree = iov;
4027 	} else
4028 		iov = aiov;
4029 	auio.uio_iov = iov;
4030 	auio.uio_iovcnt = iovcnt;
4031 	auio.uio_rw = UIO_WRITE;
4032 	auio.uio_segflg = UIO_USERSPACE;
4033 	auio.uio_td = td;
4034 	auio.uio_offset = 0;
4035 	if ((error = copyin((caddr_t)iovp, (caddr_t)iov, iovlen)))
4036 		goto done;
4037 	auio.uio_resid = 0;
4038 	for (i = 0; i < iovcnt; i++) {
4039 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
4040 			error = EINVAL;
4041 			goto done;
4042 		}
4043 		auio.uio_resid += iov->iov_len;
4044 		iov++;
4045 	}
4046 	cnt = auio.uio_resid;
4047 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
4048 	    td->td_proc->p_ucred, td);
4049 	cnt -= auio.uio_resid;
4050 	td->td_retval[0] = cnt;
4051 done:
4052 	if (needfree)
4053 		FREE(needfree, M_IOV);
4054 	VOP_UNLOCK(vp, 0, td);
4055 	vn_finished_write(mp);
4056 	return (error);
4057 }
4058 
4059 int
4060 extattr_set_file(td, uap)
4061 	struct thread *td;
4062 	struct extattr_set_file_args *uap;
4063 {
4064 	struct nameidata nd;
4065 	char attrname[EXTATTR_MAXNAMELEN];
4066 	int error;
4067 
4068 	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
4069 	    NULL);
4070 	if (error)
4071 		return (error);
4072 
4073 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
4074 	if ((error = namei(&nd)) != 0)
4075 		return (error);
4076 	NDFREE(&nd, NDF_ONLY_PNBUF);
4077 
4078 	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4079 	    SCARG(uap, iovp), SCARG(uap, iovcnt), td);
4080 
4081 	vrele(nd.ni_vp);
4082 	return (error);
4083 }
4084 
4085 int
4086 extattr_set_fd(td, uap)
4087 	struct thread *td;
4088 	struct extattr_set_fd_args *uap;
4089 {
4090 	struct file *fp;
4091 	char attrname[EXTATTR_MAXNAMELEN];
4092 	int error;
4093 
4094 	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
4095 	    NULL);
4096 	if (error)
4097 		return (error);
4098 
4099 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
4100 		return (error);
4101 
4102 	error = extattr_set_vp((struct vnode *)fp->f_data,
4103 	    SCARG(uap, attrnamespace), attrname, SCARG(uap, iovp),
4104 	    SCARG(uap, iovcnt), td);
4105 	fdrop(fp, td);
4106 
4107 	return (error);
4108 }
4109 
4110 /*
4111  * extattr_get_vp(): Get a named extended attribute on a file or directory
4112  *
4113  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
4114  *            kernelspace string pointer "attrname",
4115  *            userspace iovec array pointer "iovp", unsigned int iovcnt,
4116  *            proc "p"
4117  * Returns: 0 on success, an error number otherwise
4118  * Locks: none
4119  * References: vp must be a valid reference for the duration of the call
4120  */
4121 static int
4122 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
4123     struct iovec *iovp, unsigned iovcnt, struct thread *td)
4124 {
4125 	struct uio auio;
4126 	struct iovec *iov, *needfree = NULL, aiov[UIO_SMALLIOV];
4127 	u_int iovlen, cnt;
4128 	int error, i;
4129 
4130 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_READ);
4131 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
4132 
4133 	iovlen = iovcnt * sizeof (struct iovec);
4134 	if (iovcnt > UIO_SMALLIOV) {
4135 		if (iovcnt > UIO_MAXIOV) {
4136 			error = EINVAL;
4137 			goto done;
4138 		}
4139 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
4140 		needfree = iov;
4141 	} else
4142 		iov = aiov;
4143 	auio.uio_iov = iov;
4144 	auio.uio_iovcnt = iovcnt;
4145 	auio.uio_rw = UIO_READ;
4146 	auio.uio_segflg = UIO_USERSPACE;
4147 	auio.uio_td = td;
4148 	auio.uio_offset = 0;
4149 	if ((error = copyin((caddr_t)iovp, (caddr_t)iov, iovlen)))
4150 		goto done;
4151 	auio.uio_resid = 0;
4152 	for (i = 0; i < iovcnt; i++) {
4153 		if (iov->iov_len > INT_MAX - auio.uio_resid) {
4154 			error = EINVAL;
4155 			goto done;
4156 		}
4157 		auio.uio_resid += iov->iov_len;
4158 		iov++;
4159 	}
4160 	cnt = auio.uio_resid;
4161 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio,
4162 	    td->td_proc->p_ucred, td);
4163 	cnt -= auio.uio_resid;
4164 	td->td_retval[0] = cnt;
4165 done:
4166 	if (needfree)
4167 		FREE(needfree, M_IOV);
4168 	VOP_UNLOCK(vp, 0, td);
4169 	return (error);
4170 }
4171 
4172 int
4173 extattr_get_file(td, uap)
4174 	struct thread *td;
4175 	struct extattr_get_file_args *uap;
4176 {
4177 	struct nameidata nd;
4178 	char attrname[EXTATTR_MAXNAMELEN];
4179 	int error;
4180 
4181 	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
4182 	    NULL);
4183 	if (error)
4184 		return (error);
4185 
4186 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
4187 	if ((error = namei(&nd)) != 0)
4188 		return (error);
4189 	NDFREE(&nd, NDF_ONLY_PNBUF);
4190 
4191 	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4192 	    SCARG(uap, iovp), SCARG(uap, iovcnt), td);
4193 
4194 	vrele(nd.ni_vp);
4195 	return (error);
4196 }
4197 
4198 int
4199 extattr_get_fd(td, uap)
4200 	struct thread *td;
4201 	struct extattr_get_fd_args *uap;
4202 {
4203 	struct file *fp;
4204 	char attrname[EXTATTR_MAXNAMELEN];
4205 	int error;
4206 
4207 	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
4208 	    NULL);
4209 	if (error)
4210 		return (error);
4211 
4212 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
4213 		return (error);
4214 
4215 	error = extattr_get_vp((struct vnode *)fp->f_data,
4216 	    SCARG(uap, attrnamespace), attrname, SCARG(uap, iovp),
4217 	    SCARG(uap, iovcnt), td);
4218 
4219 	fdrop(fp, td);
4220 	return (error);
4221 }
4222 
4223 /*
4224  * extattr_delete_vp(): Delete a named extended attribute on a file or
4225  *                      directory
4226  *
4227  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
4228  *            kernelspace string pointer "attrname", proc "p"
4229  * Returns: 0 on success, an error number otherwise
4230  * Locks: none
4231  * References: vp must be a valid reference for the duration of the call
4232  */
4233 static int
4234 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
4235     struct thread *td)
4236 {
4237 	struct mount *mp;
4238 	int error;
4239 
4240 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
4241 		return (error);
4242 	VOP_LEASE(vp, td, td->td_proc->p_ucred, LEASE_WRITE);
4243 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
4244 
4245 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
4246 	    td->td_proc->p_ucred, td);
4247 
4248 	VOP_UNLOCK(vp, 0, td);
4249 	vn_finished_write(mp);
4250 	return (error);
4251 }
4252 
4253 int
4254 extattr_delete_file(td, uap)
4255 	struct thread *td;
4256 	struct extattr_delete_file_args *uap;
4257 {
4258 	struct nameidata nd;
4259 	char attrname[EXTATTR_MAXNAMELEN];
4260 	int error;
4261 
4262 	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
4263 	     NULL);
4264 	if (error)
4265 		return(error);
4266 
4267 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), td);
4268 	if ((error = namei(&nd)) != 0)
4269 		return(error);
4270 	NDFREE(&nd, NDF_ONLY_PNBUF);
4271 
4272 	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4273 	    attrname, td);
4274 
4275 	vrele(nd.ni_vp);
4276 	return(error);
4277 }
4278 
4279 int
4280 extattr_delete_fd(td, uap)
4281 	struct thread *td;
4282 	struct extattr_delete_fd_args *uap;
4283 {
4284 	struct file *fp;
4285 	struct vnode *vp;
4286 	char attrname[EXTATTR_MAXNAMELEN];
4287 	int error;
4288 
4289 	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
4290 	    NULL);
4291 	if (error)
4292 		return (error);
4293 
4294 	if ((error = getvnode(td->td_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
4295 		return (error);
4296 	vp = (struct vnode *)fp->f_data;
4297 
4298 	error = extattr_delete_vp((struct vnode *)fp->f_data,
4299 	    SCARG(uap, attrnamespace), attrname, td);
4300 
4301 	fdrop(fp, td);
4302 	return (error);
4303 }
4304