xref: /freebsd/sys/kern/vfs_syscalls.c (revision a9148abd9da5db2f1c682fb17bed791845fc41c9)
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  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_compat.h"
41 #include "opt_ktrace.h"
42 #include "opt_mac.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/sysent.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/sysproto.h>
53 #include <sys/namei.h>
54 #include <sys/filedesc.h>
55 #include <sys/kernel.h>
56 #include <sys/fcntl.h>
57 #include <sys/file.h>
58 #include <sys/filio.h>
59 #include <sys/limits.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/priv.h>
66 #include <sys/proc.h>
67 #include <sys/dirent.h>
68 #include <sys/jail.h>
69 #include <sys/syscallsubr.h>
70 #include <sys/sysctl.h>
71 #ifdef KTRACE
72 #include <sys/ktrace.h>
73 #endif
74 
75 #include <machine/stdarg.h>
76 
77 #include <security/audit/audit.h>
78 #include <security/mac/mac_framework.h>
79 
80 #include <vm/vm.h>
81 #include <vm/vm_object.h>
82 #include <vm/vm_page.h>
83 #include <vm/uma.h>
84 
85 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
86 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
87 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
88 static int setfmode(struct thread *td, struct vnode *, int);
89 static int setfflags(struct thread *td, struct vnode *, int);
90 static int setutimes(struct thread *td, struct vnode *,
91     const struct timespec *, int, int);
92 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
93     struct thread *td);
94 
95 /*
96  * The module initialization routine for POSIX asynchronous I/O will
97  * set this to the version of AIO that it implements.  (Zero means
98  * that it is not implemented.)  This value is used here by pathconf()
99  * and in kern_descrip.c by fpathconf().
100  */
101 int async_io_version;
102 
103 #ifdef DEBUG
104 static int syncprt = 0;
105 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
106 #endif
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int vfslocked;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		vfslocked = VFS_LOCK_GIANT(mp);
132 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
133 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
134 			MNT_ILOCK(mp);
135 			mp->mnt_noasync++;
136 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
137 			MNT_IUNLOCK(mp);
138 			vfs_msync(mp, MNT_NOWAIT);
139 			VFS_SYNC(mp, MNT_NOWAIT, td);
140 			MNT_ILOCK(mp);
141 			mp->mnt_noasync--;
142 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
143 			    mp->mnt_noasync == 0)
144 				mp->mnt_kern_flag |= MNTK_ASYNC;
145 			MNT_IUNLOCK(mp);
146 			vn_finished_write(mp);
147 		}
148 		VFS_UNLOCK_GIANT(vfslocked);
149 		mtx_lock(&mountlist_mtx);
150 		nmp = TAILQ_NEXT(mp, mnt_list);
151 		vfs_unbusy(mp);
152 	}
153 	mtx_unlock(&mountlist_mtx);
154 	return (0);
155 }
156 
157 /* XXX PRISON: could be per prison flag */
158 static int prison_quotas;
159 #if 0
160 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
161 #endif
162 
163 /*
164  * Change filesystem quotas.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct quotactl_args {
168 	char *path;
169 	int cmd;
170 	int uid;
171 	caddr_t arg;
172 };
173 #endif
174 int
175 quotactl(td, uap)
176 	struct thread *td;
177 	register struct quotactl_args /* {
178 		char *path;
179 		int cmd;
180 		int uid;
181 		caddr_t arg;
182 	} */ *uap;
183 {
184 	struct mount *mp;
185 	int vfslocked;
186 	int error;
187 	struct nameidata nd;
188 
189 	AUDIT_ARG(cmd, uap->cmd);
190 	AUDIT_ARG(uid, uap->uid);
191 	if (jailed(td->td_ucred) && !prison_quotas)
192 		return (EPERM);
193 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1,
194 	   UIO_USERSPACE, uap->path, td);
195 	if ((error = namei(&nd)) != 0)
196 		return (error);
197 	vfslocked = NDHASGIANT(&nd);
198 	NDFREE(&nd, NDF_ONLY_PNBUF);
199 	mp = nd.ni_vp->v_mount;
200 	if ((error = vfs_busy(mp, 0))) {
201 		vrele(nd.ni_vp);
202 		VFS_UNLOCK_GIANT(vfslocked);
203 		return (error);
204 	}
205 	vrele(nd.ni_vp);
206 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, td);
207 	vfs_unbusy(mp);
208 	VFS_UNLOCK_GIANT(vfslocked);
209 	return (error);
210 }
211 
212 /*
213  * Used by statfs conversion routines to scale the block size up if
214  * necessary so that all of the block counts are <= 'max_size'.  Note
215  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
216  * value of 'n'.
217  */
218 void
219 statfs_scale_blocks(struct statfs *sf, long max_size)
220 {
221 	uint64_t count;
222 	int shift;
223 
224 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
225 
226 	/*
227 	 * Attempt to scale the block counts to give a more accurate
228 	 * overview to userland of the ratio of free space to used
229 	 * space.  To do this, find the largest block count and compute
230 	 * a divisor that lets it fit into a signed integer <= max_size.
231 	 */
232 	if (sf->f_bavail < 0)
233 		count = -sf->f_bavail;
234 	else
235 		count = sf->f_bavail;
236 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
237 	if (count <= max_size)
238 		return;
239 
240 	count >>= flsl(max_size);
241 	shift = 0;
242 	while (count > 0) {
243 		shift++;
244 		count >>=1;
245 	}
246 
247 	sf->f_bsize <<= shift;
248 	sf->f_blocks >>= shift;
249 	sf->f_bfree >>= shift;
250 	sf->f_bavail >>= shift;
251 }
252 
253 /*
254  * Get filesystem statistics.
255  */
256 #ifndef _SYS_SYSPROTO_H_
257 struct statfs_args {
258 	char *path;
259 	struct statfs *buf;
260 };
261 #endif
262 int
263 statfs(td, uap)
264 	struct thread *td;
265 	register struct statfs_args /* {
266 		char *path;
267 		struct statfs *buf;
268 	} */ *uap;
269 {
270 	struct statfs sf;
271 	int error;
272 
273 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
274 	if (error == 0)
275 		error = copyout(&sf, uap->buf, sizeof(sf));
276 	return (error);
277 }
278 
279 int
280 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
281     struct statfs *buf)
282 {
283 	struct mount *mp;
284 	struct statfs *sp, sb;
285 	int vfslocked;
286 	int error;
287 	struct nameidata nd;
288 
289 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
290 	    pathseg, path, td);
291 	error = namei(&nd);
292 	if (error)
293 		return (error);
294 	vfslocked = NDHASGIANT(&nd);
295 	mp = nd.ni_vp->v_mount;
296 	vfs_ref(mp);
297 	NDFREE(&nd, NDF_ONLY_PNBUF);
298 	vput(nd.ni_vp);
299 #ifdef MAC
300 	error = mac_mount_check_stat(td->td_ucred, mp);
301 	if (error)
302 		goto out;
303 #endif
304 	/*
305 	 * Set these in case the underlying filesystem fails to do so.
306 	 */
307 	sp = &mp->mnt_stat;
308 	sp->f_version = STATFS_VERSION;
309 	sp->f_namemax = NAME_MAX;
310 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
311 	error = VFS_STATFS(mp, sp, td);
312 	if (error)
313 		goto out;
314 	if (priv_check(td, PRIV_VFS_GENERATION)) {
315 		bcopy(sp, &sb, sizeof(sb));
316 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
317 		prison_enforce_statfs(td->td_ucred, mp, &sb);
318 		sp = &sb;
319 	}
320 	*buf = *sp;
321 out:
322 	vfs_rel(mp);
323 	VFS_UNLOCK_GIANT(vfslocked);
324 	if (mtx_owned(&Giant))
325 		printf("statfs(%d): %s: %d\n", vfslocked, path, error);
326 	return (error);
327 }
328 
329 /*
330  * Get filesystem statistics.
331  */
332 #ifndef _SYS_SYSPROTO_H_
333 struct fstatfs_args {
334 	int fd;
335 	struct statfs *buf;
336 };
337 #endif
338 int
339 fstatfs(td, uap)
340 	struct thread *td;
341 	register struct fstatfs_args /* {
342 		int fd;
343 		struct statfs *buf;
344 	} */ *uap;
345 {
346 	struct statfs sf;
347 	int error;
348 
349 	error = kern_fstatfs(td, uap->fd, &sf);
350 	if (error == 0)
351 		error = copyout(&sf, uap->buf, sizeof(sf));
352 	return (error);
353 }
354 
355 int
356 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
357 {
358 	struct file *fp;
359 	struct mount *mp;
360 	struct statfs *sp, sb;
361 	int vfslocked;
362 	struct vnode *vp;
363 	int error;
364 
365 	AUDIT_ARG(fd, fd);
366 	error = getvnode(td->td_proc->p_fd, fd, &fp);
367 	if (error)
368 		return (error);
369 	vp = fp->f_vnode;
370 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
371 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
372 #ifdef AUDIT
373 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
374 #endif
375 	mp = vp->v_mount;
376 	if (mp)
377 		vfs_ref(mp);
378 	VOP_UNLOCK(vp, 0);
379 	fdrop(fp, td);
380 	if (vp->v_iflag & VI_DOOMED) {
381 		error = EBADF;
382 		goto out;
383 	}
384 #ifdef MAC
385 	error = mac_mount_check_stat(td->td_ucred, mp);
386 	if (error)
387 		goto out;
388 #endif
389 	/*
390 	 * Set these in case the underlying filesystem fails to do so.
391 	 */
392 	sp = &mp->mnt_stat;
393 	sp->f_version = STATFS_VERSION;
394 	sp->f_namemax = NAME_MAX;
395 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
396 	error = VFS_STATFS(mp, sp, td);
397 	if (error)
398 		goto out;
399 	if (priv_check(td, PRIV_VFS_GENERATION)) {
400 		bcopy(sp, &sb, sizeof(sb));
401 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
402 		prison_enforce_statfs(td->td_ucred, mp, &sb);
403 		sp = &sb;
404 	}
405 	*buf = *sp;
406 out:
407 	if (mp)
408 		vfs_rel(mp);
409 	VFS_UNLOCK_GIANT(vfslocked);
410 	return (error);
411 }
412 
413 /*
414  * Get statistics on all filesystems.
415  */
416 #ifndef _SYS_SYSPROTO_H_
417 struct getfsstat_args {
418 	struct statfs *buf;
419 	long bufsize;
420 	int flags;
421 };
422 #endif
423 int
424 getfsstat(td, uap)
425 	struct thread *td;
426 	register struct getfsstat_args /* {
427 		struct statfs *buf;
428 		long bufsize;
429 		int flags;
430 	} */ *uap;
431 {
432 
433 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
434 	    uap->flags));
435 }
436 
437 /*
438  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
439  * 	The caller is responsible for freeing memory which will be allocated
440  *	in '*buf'.
441  */
442 int
443 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
444     enum uio_seg bufseg, int flags)
445 {
446 	struct mount *mp, *nmp;
447 	struct statfs *sfsp, *sp, sb;
448 	size_t count, maxcount;
449 	int vfslocked;
450 	int error;
451 
452 	maxcount = bufsize / sizeof(struct statfs);
453 	if (bufsize == 0)
454 		sfsp = NULL;
455 	else if (bufseg == UIO_USERSPACE)
456 		sfsp = *buf;
457 	else /* if (bufseg == UIO_SYSSPACE) */ {
458 		count = 0;
459 		mtx_lock(&mountlist_mtx);
460 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
461 			count++;
462 		}
463 		mtx_unlock(&mountlist_mtx);
464 		if (maxcount > count)
465 			maxcount = count;
466 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
467 		    M_WAITOK);
468 	}
469 	count = 0;
470 	mtx_lock(&mountlist_mtx);
471 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
472 		if (prison_canseemount(td->td_ucred, mp) != 0) {
473 			nmp = TAILQ_NEXT(mp, mnt_list);
474 			continue;
475 		}
476 #ifdef MAC
477 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
478 			nmp = TAILQ_NEXT(mp, mnt_list);
479 			continue;
480 		}
481 #endif
482 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
483 			nmp = TAILQ_NEXT(mp, mnt_list);
484 			continue;
485 		}
486 		vfslocked = VFS_LOCK_GIANT(mp);
487 		if (sfsp && count < maxcount) {
488 			sp = &mp->mnt_stat;
489 			/*
490 			 * Set these in case the underlying filesystem
491 			 * fails to do so.
492 			 */
493 			sp->f_version = STATFS_VERSION;
494 			sp->f_namemax = NAME_MAX;
495 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
496 			/*
497 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
498 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
499 			 * overrides MNT_WAIT.
500 			 */
501 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
502 			    (flags & MNT_WAIT)) &&
503 			    (error = VFS_STATFS(mp, sp, td))) {
504 				VFS_UNLOCK_GIANT(vfslocked);
505 				mtx_lock(&mountlist_mtx);
506 				nmp = TAILQ_NEXT(mp, mnt_list);
507 				vfs_unbusy(mp);
508 				continue;
509 			}
510 			if (priv_check(td, PRIV_VFS_GENERATION)) {
511 				bcopy(sp, &sb, sizeof(sb));
512 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
513 				prison_enforce_statfs(td->td_ucred, mp, &sb);
514 				sp = &sb;
515 			}
516 			if (bufseg == UIO_SYSSPACE)
517 				bcopy(sp, sfsp, sizeof(*sp));
518 			else /* if (bufseg == UIO_USERSPACE) */ {
519 				error = copyout(sp, sfsp, sizeof(*sp));
520 				if (error) {
521 					vfs_unbusy(mp);
522 					VFS_UNLOCK_GIANT(vfslocked);
523 					return (error);
524 				}
525 			}
526 			sfsp++;
527 		}
528 		VFS_UNLOCK_GIANT(vfslocked);
529 		count++;
530 		mtx_lock(&mountlist_mtx);
531 		nmp = TAILQ_NEXT(mp, mnt_list);
532 		vfs_unbusy(mp);
533 	}
534 	mtx_unlock(&mountlist_mtx);
535 	if (sfsp && count > maxcount)
536 		td->td_retval[0] = maxcount;
537 	else
538 		td->td_retval[0] = count;
539 	return (0);
540 }
541 
542 #ifdef COMPAT_FREEBSD4
543 /*
544  * Get old format filesystem statistics.
545  */
546 static void cvtstatfs(struct statfs *, struct ostatfs *);
547 
548 #ifndef _SYS_SYSPROTO_H_
549 struct freebsd4_statfs_args {
550 	char *path;
551 	struct ostatfs *buf;
552 };
553 #endif
554 int
555 freebsd4_statfs(td, uap)
556 	struct thread *td;
557 	struct freebsd4_statfs_args /* {
558 		char *path;
559 		struct ostatfs *buf;
560 	} */ *uap;
561 {
562 	struct ostatfs osb;
563 	struct statfs sf;
564 	int error;
565 
566 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
567 	if (error)
568 		return (error);
569 	cvtstatfs(&sf, &osb);
570 	return (copyout(&osb, uap->buf, sizeof(osb)));
571 }
572 
573 /*
574  * Get filesystem statistics.
575  */
576 #ifndef _SYS_SYSPROTO_H_
577 struct freebsd4_fstatfs_args {
578 	int fd;
579 	struct ostatfs *buf;
580 };
581 #endif
582 int
583 freebsd4_fstatfs(td, uap)
584 	struct thread *td;
585 	struct freebsd4_fstatfs_args /* {
586 		int fd;
587 		struct ostatfs *buf;
588 	} */ *uap;
589 {
590 	struct ostatfs osb;
591 	struct statfs sf;
592 	int error;
593 
594 	error = kern_fstatfs(td, uap->fd, &sf);
595 	if (error)
596 		return (error);
597 	cvtstatfs(&sf, &osb);
598 	return (copyout(&osb, uap->buf, sizeof(osb)));
599 }
600 
601 /*
602  * Get statistics on all filesystems.
603  */
604 #ifndef _SYS_SYSPROTO_H_
605 struct freebsd4_getfsstat_args {
606 	struct ostatfs *buf;
607 	long bufsize;
608 	int flags;
609 };
610 #endif
611 int
612 freebsd4_getfsstat(td, uap)
613 	struct thread *td;
614 	register struct freebsd4_getfsstat_args /* {
615 		struct ostatfs *buf;
616 		long bufsize;
617 		int flags;
618 	} */ *uap;
619 {
620 	struct statfs *buf, *sp;
621 	struct ostatfs osb;
622 	size_t count, size;
623 	int error;
624 
625 	count = uap->bufsize / sizeof(struct ostatfs);
626 	size = count * sizeof(struct statfs);
627 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
628 	if (size > 0) {
629 		count = td->td_retval[0];
630 		sp = buf;
631 		while (count > 0 && error == 0) {
632 			cvtstatfs(sp, &osb);
633 			error = copyout(&osb, uap->buf, sizeof(osb));
634 			sp++;
635 			uap->buf++;
636 			count--;
637 		}
638 		free(buf, M_TEMP);
639 	}
640 	return (error);
641 }
642 
643 /*
644  * Implement fstatfs() for (NFS) file handles.
645  */
646 #ifndef _SYS_SYSPROTO_H_
647 struct freebsd4_fhstatfs_args {
648 	struct fhandle *u_fhp;
649 	struct ostatfs *buf;
650 };
651 #endif
652 int
653 freebsd4_fhstatfs(td, uap)
654 	struct thread *td;
655 	struct freebsd4_fhstatfs_args /* {
656 		struct fhandle *u_fhp;
657 		struct ostatfs *buf;
658 	} */ *uap;
659 {
660 	struct ostatfs osb;
661 	struct statfs sf;
662 	fhandle_t fh;
663 	int error;
664 
665 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
666 	if (error)
667 		return (error);
668 	error = kern_fhstatfs(td, fh, &sf);
669 	if (error)
670 		return (error);
671 	cvtstatfs(&sf, &osb);
672 	return (copyout(&osb, uap->buf, sizeof(osb)));
673 }
674 
675 /*
676  * Convert a new format statfs structure to an old format statfs structure.
677  */
678 static void
679 cvtstatfs(nsp, osp)
680 	struct statfs *nsp;
681 	struct ostatfs *osp;
682 {
683 
684 	statfs_scale_blocks(nsp, LONG_MAX);
685 	bzero(osp, sizeof(*osp));
686 	osp->f_bsize = nsp->f_bsize;
687 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
688 	osp->f_blocks = nsp->f_blocks;
689 	osp->f_bfree = nsp->f_bfree;
690 	osp->f_bavail = nsp->f_bavail;
691 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
692 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
693 	osp->f_owner = nsp->f_owner;
694 	osp->f_type = nsp->f_type;
695 	osp->f_flags = nsp->f_flags;
696 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
697 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
698 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
699 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
700 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
701 	    MIN(MFSNAMELEN, OMFSNAMELEN));
702 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
703 	    MIN(MNAMELEN, OMNAMELEN));
704 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
705 	    MIN(MNAMELEN, OMNAMELEN));
706 	osp->f_fsid = nsp->f_fsid;
707 }
708 #endif /* COMPAT_FREEBSD4 */
709 
710 /*
711  * Change current working directory to a given file descriptor.
712  */
713 #ifndef _SYS_SYSPROTO_H_
714 struct fchdir_args {
715 	int	fd;
716 };
717 #endif
718 int
719 fchdir(td, uap)
720 	struct thread *td;
721 	struct fchdir_args /* {
722 		int fd;
723 	} */ *uap;
724 {
725 	register struct filedesc *fdp = td->td_proc->p_fd;
726 	struct vnode *vp, *tdp, *vpold;
727 	struct mount *mp;
728 	struct file *fp;
729 	int vfslocked;
730 	int error;
731 
732 	AUDIT_ARG(fd, uap->fd);
733 	if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
734 		return (error);
735 	vp = fp->f_vnode;
736 	VREF(vp);
737 	fdrop(fp, td);
738 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
739 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
740 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
741 	error = change_dir(vp, td);
742 	while (!error && (mp = vp->v_mountedhere) != NULL) {
743 		int tvfslocked;
744 		if (vfs_busy(mp, 0))
745 			continue;
746 		tvfslocked = VFS_LOCK_GIANT(mp);
747 		error = VFS_ROOT(mp, LK_EXCLUSIVE, &tdp, td);
748 		vfs_unbusy(mp);
749 		if (error) {
750 			VFS_UNLOCK_GIANT(tvfslocked);
751 			break;
752 		}
753 		vput(vp);
754 		VFS_UNLOCK_GIANT(vfslocked);
755 		vp = tdp;
756 		vfslocked = tvfslocked;
757 	}
758 	if (error) {
759 		vput(vp);
760 		VFS_UNLOCK_GIANT(vfslocked);
761 		return (error);
762 	}
763 	VOP_UNLOCK(vp, 0);
764 	VFS_UNLOCK_GIANT(vfslocked);
765 	FILEDESC_XLOCK(fdp);
766 	vpold = fdp->fd_cdir;
767 	fdp->fd_cdir = vp;
768 	FILEDESC_XUNLOCK(fdp);
769 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
770 	vrele(vpold);
771 	VFS_UNLOCK_GIANT(vfslocked);
772 	return (0);
773 }
774 
775 /*
776  * Change current working directory (``.'').
777  */
778 #ifndef _SYS_SYSPROTO_H_
779 struct chdir_args {
780 	char	*path;
781 };
782 #endif
783 int
784 chdir(td, uap)
785 	struct thread *td;
786 	struct chdir_args /* {
787 		char *path;
788 	} */ *uap;
789 {
790 
791 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
792 }
793 
794 int
795 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
796 {
797 	register struct filedesc *fdp = td->td_proc->p_fd;
798 	int error;
799 	struct nameidata nd;
800 	struct vnode *vp;
801 	int vfslocked;
802 
803 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1 | MPSAFE,
804 	    pathseg, path, td);
805 	if ((error = namei(&nd)) != 0)
806 		return (error);
807 	vfslocked = NDHASGIANT(&nd);
808 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
809 		vput(nd.ni_vp);
810 		VFS_UNLOCK_GIANT(vfslocked);
811 		NDFREE(&nd, NDF_ONLY_PNBUF);
812 		return (error);
813 	}
814 	VOP_UNLOCK(nd.ni_vp, 0);
815 	VFS_UNLOCK_GIANT(vfslocked);
816 	NDFREE(&nd, NDF_ONLY_PNBUF);
817 	FILEDESC_XLOCK(fdp);
818 	vp = fdp->fd_cdir;
819 	fdp->fd_cdir = nd.ni_vp;
820 	FILEDESC_XUNLOCK(fdp);
821 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
822 	vrele(vp);
823 	VFS_UNLOCK_GIANT(vfslocked);
824 	return (0);
825 }
826 
827 /*
828  * Helper function for raised chroot(2) security function:  Refuse if
829  * any filedescriptors are open directories.
830  */
831 static int
832 chroot_refuse_vdir_fds(fdp)
833 	struct filedesc *fdp;
834 {
835 	struct vnode *vp;
836 	struct file *fp;
837 	int fd;
838 
839 	FILEDESC_LOCK_ASSERT(fdp);
840 
841 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
842 		fp = fget_locked(fdp, fd);
843 		if (fp == NULL)
844 			continue;
845 		if (fp->f_type == DTYPE_VNODE) {
846 			vp = fp->f_vnode;
847 			if (vp->v_type == VDIR)
848 				return (EPERM);
849 		}
850 	}
851 	return (0);
852 }
853 
854 /*
855  * This sysctl determines if we will allow a process to chroot(2) if it
856  * has a directory open:
857  *	0: disallowed for all processes.
858  *	1: allowed for processes that were not already chroot(2)'ed.
859  *	2: allowed for all processes.
860  */
861 
862 static int chroot_allow_open_directories = 1;
863 
864 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
865      &chroot_allow_open_directories, 0, "");
866 
867 /*
868  * Change notion of root (``/'') directory.
869  */
870 #ifndef _SYS_SYSPROTO_H_
871 struct chroot_args {
872 	char	*path;
873 };
874 #endif
875 int
876 chroot(td, uap)
877 	struct thread *td;
878 	struct chroot_args /* {
879 		char *path;
880 	} */ *uap;
881 {
882 	int error;
883 	struct nameidata nd;
884 	int vfslocked;
885 
886 	error = priv_check(td, PRIV_VFS_CHROOT);
887 	if (error)
888 		return (error);
889 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
890 	    UIO_USERSPACE, uap->path, td);
891 	error = namei(&nd);
892 	if (error)
893 		goto error;
894 	vfslocked = NDHASGIANT(&nd);
895 	if ((error = change_dir(nd.ni_vp, td)) != 0)
896 		goto e_vunlock;
897 #ifdef MAC
898 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
899 		goto e_vunlock;
900 #endif
901 	VOP_UNLOCK(nd.ni_vp, 0);
902 	error = change_root(nd.ni_vp, td);
903 	vrele(nd.ni_vp);
904 	VFS_UNLOCK_GIANT(vfslocked);
905 	NDFREE(&nd, NDF_ONLY_PNBUF);
906 	return (error);
907 e_vunlock:
908 	vput(nd.ni_vp);
909 	VFS_UNLOCK_GIANT(vfslocked);
910 error:
911 	NDFREE(&nd, NDF_ONLY_PNBUF);
912 	return (error);
913 }
914 
915 /*
916  * Common routine for chroot and chdir.  Callers must provide a locked vnode
917  * instance.
918  */
919 int
920 change_dir(vp, td)
921 	struct vnode *vp;
922 	struct thread *td;
923 {
924 	int error;
925 
926 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
927 	if (vp->v_type != VDIR)
928 		return (ENOTDIR);
929 #ifdef MAC
930 	error = mac_vnode_check_chdir(td->td_ucred, vp);
931 	if (error)
932 		return (error);
933 #endif
934 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
935 	return (error);
936 }
937 
938 /*
939  * Common routine for kern_chroot() and jail_attach().  The caller is
940  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
941  * authorize this operation.
942  */
943 int
944 change_root(vp, td)
945 	struct vnode *vp;
946 	struct thread *td;
947 {
948 	struct filedesc *fdp;
949 	struct vnode *oldvp;
950 	int vfslocked;
951 	int error;
952 
953 	VFS_ASSERT_GIANT(vp->v_mount);
954 	fdp = td->td_proc->p_fd;
955 	FILEDESC_XLOCK(fdp);
956 	if (chroot_allow_open_directories == 0 ||
957 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
958 		error = chroot_refuse_vdir_fds(fdp);
959 		if (error) {
960 			FILEDESC_XUNLOCK(fdp);
961 			return (error);
962 		}
963 	}
964 	oldvp = fdp->fd_rdir;
965 	fdp->fd_rdir = vp;
966 	VREF(fdp->fd_rdir);
967 	if (!fdp->fd_jdir) {
968 		fdp->fd_jdir = vp;
969 		VREF(fdp->fd_jdir);
970 	}
971 	FILEDESC_XUNLOCK(fdp);
972 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
973 	vrele(oldvp);
974 	VFS_UNLOCK_GIANT(vfslocked);
975 	return (0);
976 }
977 
978 /*
979  * Check permissions, allocate an open file structure, and call the device
980  * open routine if any.
981  */
982 #ifndef _SYS_SYSPROTO_H_
983 struct open_args {
984 	char	*path;
985 	int	flags;
986 	int	mode;
987 };
988 #endif
989 int
990 open(td, uap)
991 	struct thread *td;
992 	register struct open_args /* {
993 		char *path;
994 		int flags;
995 		int mode;
996 	} */ *uap;
997 {
998 
999 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1000 }
1001 
1002 #ifndef _SYS_SYSPROTO_H_
1003 struct openat_args {
1004 	int	fd;
1005 	char	*path;
1006 	int	flag;
1007 	int	mode;
1008 };
1009 #endif
1010 int
1011 openat(struct thread *td, struct openat_args *uap)
1012 {
1013 
1014 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1015 	    uap->mode));
1016 }
1017 
1018 int
1019 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1020     int mode)
1021 {
1022 
1023 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1024 }
1025 
1026 int
1027 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1028     int flags, int mode)
1029 {
1030 	struct proc *p = td->td_proc;
1031 	struct filedesc *fdp = p->p_fd;
1032 	struct file *fp;
1033 	struct vnode *vp;
1034 	struct vattr vat;
1035 	struct mount *mp;
1036 	int cmode;
1037 	struct file *nfp;
1038 	int type, indx, error;
1039 	struct flock lf;
1040 	struct nameidata nd;
1041 	int vfslocked;
1042 
1043 	AUDIT_ARG(fflags, flags);
1044 	AUDIT_ARG(mode, mode);
1045 	/* XXX: audit dirfd */
1046 	/*
1047 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1048 	 * be specified.
1049 	 */
1050 	if (flags & O_EXEC) {
1051 		if (flags & O_ACCMODE)
1052 			return (EINVAL);
1053 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1054 		return (EINVAL);
1055 	else
1056 		flags = FFLAGS(flags);
1057 
1058 	error = falloc(td, &nfp, &indx);
1059 	if (error)
1060 		return (error);
1061 	/* An extra reference on `nfp' has been held for us by falloc(). */
1062 	fp = nfp;
1063 	/* Set the flags early so the finit in devfs can pick them up. */
1064 	fp->f_flag = flags & FMASK;
1065 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1066 	NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1067 	    td);
1068 	td->td_dupfd = -1;		/* XXX check for fdopen */
1069 	error = vn_open(&nd, &flags, cmode, fp);
1070 	if (error) {
1071 		/*
1072 		 * If the vn_open replaced the method vector, something
1073 		 * wonderous happened deep below and we just pass it up
1074 		 * pretending we know what we do.
1075 		 */
1076 		if (error == ENXIO && fp->f_ops != &badfileops) {
1077 			fdrop(fp, td);
1078 			td->td_retval[0] = indx;
1079 			return (0);
1080 		}
1081 
1082 		/*
1083 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1084 		 * responsible for dropping the old contents of ofiles[indx]
1085 		 * if it succeeds.
1086 		 */
1087 		if ((error == ENODEV || error == ENXIO) &&
1088 		    td->td_dupfd >= 0 &&		/* XXX from fdopen */
1089 		    (error =
1090 			dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1091 			td->td_retval[0] = indx;
1092 			fdrop(fp, td);
1093 			return (0);
1094 		}
1095 		/*
1096 		 * Clean up the descriptor, but only if another thread hadn't
1097 		 * replaced or closed it.
1098 		 */
1099 		fdclose(fdp, fp, indx, td);
1100 		fdrop(fp, td);
1101 
1102 		if (error == ERESTART)
1103 			error = EINTR;
1104 		return (error);
1105 	}
1106 	td->td_dupfd = 0;
1107 	vfslocked = NDHASGIANT(&nd);
1108 	NDFREE(&nd, NDF_ONLY_PNBUF);
1109 	vp = nd.ni_vp;
1110 
1111 	fp->f_vnode = vp;	/* XXX Does devfs need this? */
1112 	/*
1113 	 * If the file wasn't claimed by devfs bind it to the normal
1114 	 * vnode operations here.
1115 	 */
1116 	if (fp->f_ops == &badfileops) {
1117 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1118 		fp->f_seqcount = 1;
1119 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1120 	}
1121 
1122 	VOP_UNLOCK(vp, 0);
1123 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1124 		lf.l_whence = SEEK_SET;
1125 		lf.l_start = 0;
1126 		lf.l_len = 0;
1127 		if (flags & O_EXLOCK)
1128 			lf.l_type = F_WRLCK;
1129 		else
1130 			lf.l_type = F_RDLCK;
1131 		type = F_FLOCK;
1132 		if ((flags & FNONBLOCK) == 0)
1133 			type |= F_WAIT;
1134 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1135 			    type)) != 0)
1136 			goto bad;
1137 		atomic_set_int(&fp->f_flag, FHASLOCK);
1138 	}
1139 	if (flags & O_TRUNC) {
1140 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1141 			goto bad;
1142 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1143 		VATTR_NULL(&vat);
1144 		vat.va_size = 0;
1145 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1146 #ifdef MAC
1147 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1148 		if (error == 0)
1149 #endif
1150 			error = VOP_SETATTR(vp, &vat, td->td_ucred);
1151 		VOP_UNLOCK(vp, 0);
1152 		vn_finished_write(mp);
1153 		if (error)
1154 			goto bad;
1155 	}
1156 	VFS_UNLOCK_GIANT(vfslocked);
1157 	/*
1158 	 * Release our private reference, leaving the one associated with
1159 	 * the descriptor table intact.
1160 	 */
1161 	fdrop(fp, td);
1162 	td->td_retval[0] = indx;
1163 	return (0);
1164 bad:
1165 	VFS_UNLOCK_GIANT(vfslocked);
1166 	fdclose(fdp, fp, indx, td);
1167 	fdrop(fp, td);
1168 	return (error);
1169 }
1170 
1171 #ifdef COMPAT_43
1172 /*
1173  * Create a file.
1174  */
1175 #ifndef _SYS_SYSPROTO_H_
1176 struct ocreat_args {
1177 	char	*path;
1178 	int	mode;
1179 };
1180 #endif
1181 int
1182 ocreat(td, uap)
1183 	struct thread *td;
1184 	register struct ocreat_args /* {
1185 		char *path;
1186 		int mode;
1187 	} */ *uap;
1188 {
1189 
1190 	return (kern_open(td, uap->path, UIO_USERSPACE,
1191 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1192 }
1193 #endif /* COMPAT_43 */
1194 
1195 /*
1196  * Create a special file.
1197  */
1198 #ifndef _SYS_SYSPROTO_H_
1199 struct mknod_args {
1200 	char	*path;
1201 	int	mode;
1202 	int	dev;
1203 };
1204 #endif
1205 int
1206 mknod(td, uap)
1207 	struct thread *td;
1208 	register struct mknod_args /* {
1209 		char *path;
1210 		int mode;
1211 		int dev;
1212 	} */ *uap;
1213 {
1214 
1215 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1216 }
1217 
1218 #ifndef _SYS_SYSPROTO_H_
1219 struct mknodat_args {
1220 	int	fd;
1221 	char	*path;
1222 	mode_t	mode;
1223 	dev_t	dev;
1224 };
1225 #endif
1226 int
1227 mknodat(struct thread *td, struct mknodat_args *uap)
1228 {
1229 
1230 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1231 	    uap->dev));
1232 }
1233 
1234 int
1235 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1236     int dev)
1237 {
1238 
1239 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1240 }
1241 
1242 int
1243 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1244     int mode, int dev)
1245 {
1246 	struct vnode *vp;
1247 	struct mount *mp;
1248 	struct vattr vattr;
1249 	int error;
1250 	int whiteout = 0;
1251 	struct nameidata nd;
1252 	int vfslocked;
1253 
1254 	AUDIT_ARG(mode, mode);
1255 	AUDIT_ARG(dev, dev);
1256 	switch (mode & S_IFMT) {
1257 	case S_IFCHR:
1258 	case S_IFBLK:
1259 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1260 		break;
1261 	case S_IFMT:
1262 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1263 		break;
1264 	case S_IFWHT:
1265 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1266 		break;
1267 	case S_IFIFO:
1268 		if (dev == 0)
1269 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1270 		/* FALLTHROUGH */
1271 	default:
1272 		error = EINVAL;
1273 		break;
1274 	}
1275 	if (error)
1276 		return (error);
1277 restart:
1278 	bwillwrite();
1279 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1280 	    pathseg, path, fd, td);
1281 	if ((error = namei(&nd)) != 0)
1282 		return (error);
1283 	vfslocked = NDHASGIANT(&nd);
1284 	vp = nd.ni_vp;
1285 	if (vp != NULL) {
1286 		NDFREE(&nd, NDF_ONLY_PNBUF);
1287 		if (vp == nd.ni_dvp)
1288 			vrele(nd.ni_dvp);
1289 		else
1290 			vput(nd.ni_dvp);
1291 		vrele(vp);
1292 		VFS_UNLOCK_GIANT(vfslocked);
1293 		return (EEXIST);
1294 	} else {
1295 		VATTR_NULL(&vattr);
1296 		FILEDESC_SLOCK(td->td_proc->p_fd);
1297 		vattr.va_mode = (mode & ALLPERMS) &
1298 		    ~td->td_proc->p_fd->fd_cmask;
1299 		FILEDESC_SUNLOCK(td->td_proc->p_fd);
1300 		vattr.va_rdev = dev;
1301 		whiteout = 0;
1302 
1303 		switch (mode & S_IFMT) {
1304 		case S_IFMT:	/* used by badsect to flag bad sectors */
1305 			vattr.va_type = VBAD;
1306 			break;
1307 		case S_IFCHR:
1308 			vattr.va_type = VCHR;
1309 			break;
1310 		case S_IFBLK:
1311 			vattr.va_type = VBLK;
1312 			break;
1313 		case S_IFWHT:
1314 			whiteout = 1;
1315 			break;
1316 		default:
1317 			panic("kern_mknod: invalid mode");
1318 		}
1319 	}
1320 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1321 		NDFREE(&nd, NDF_ONLY_PNBUF);
1322 		vput(nd.ni_dvp);
1323 		VFS_UNLOCK_GIANT(vfslocked);
1324 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1325 			return (error);
1326 		goto restart;
1327 	}
1328 #ifdef MAC
1329 	if (error == 0 && !whiteout)
1330 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1331 		    &nd.ni_cnd, &vattr);
1332 #endif
1333 	if (!error) {
1334 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1335 		if (whiteout)
1336 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1337 		else {
1338 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1339 						&nd.ni_cnd, &vattr);
1340 			if (error == 0)
1341 				vput(nd.ni_vp);
1342 		}
1343 	}
1344 	NDFREE(&nd, NDF_ONLY_PNBUF);
1345 	vput(nd.ni_dvp);
1346 	vn_finished_write(mp);
1347 	VFS_UNLOCK_GIANT(vfslocked);
1348 	return (error);
1349 }
1350 
1351 /*
1352  * Create a named pipe.
1353  */
1354 #ifndef _SYS_SYSPROTO_H_
1355 struct mkfifo_args {
1356 	char	*path;
1357 	int	mode;
1358 };
1359 #endif
1360 int
1361 mkfifo(td, uap)
1362 	struct thread *td;
1363 	register struct mkfifo_args /* {
1364 		char *path;
1365 		int mode;
1366 	} */ *uap;
1367 {
1368 
1369 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1370 }
1371 
1372 #ifndef _SYS_SYSPROTO_H_
1373 struct mkfifoat_args {
1374 	int	fd;
1375 	char	*path;
1376 	mode_t	mode;
1377 };
1378 #endif
1379 int
1380 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1381 {
1382 
1383 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1384 	    uap->mode));
1385 }
1386 
1387 int
1388 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1389 {
1390 
1391 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1392 }
1393 
1394 int
1395 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1396     int mode)
1397 {
1398 	struct mount *mp;
1399 	struct vattr vattr;
1400 	int error;
1401 	struct nameidata nd;
1402 	int vfslocked;
1403 
1404 	AUDIT_ARG(mode, mode);
1405 restart:
1406 	bwillwrite();
1407 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1408 	    pathseg, path, fd, td);
1409 	if ((error = namei(&nd)) != 0)
1410 		return (error);
1411 	vfslocked = NDHASGIANT(&nd);
1412 	if (nd.ni_vp != NULL) {
1413 		NDFREE(&nd, NDF_ONLY_PNBUF);
1414 		if (nd.ni_vp == nd.ni_dvp)
1415 			vrele(nd.ni_dvp);
1416 		else
1417 			vput(nd.ni_dvp);
1418 		vrele(nd.ni_vp);
1419 		VFS_UNLOCK_GIANT(vfslocked);
1420 		return (EEXIST);
1421 	}
1422 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1423 		NDFREE(&nd, NDF_ONLY_PNBUF);
1424 		vput(nd.ni_dvp);
1425 		VFS_UNLOCK_GIANT(vfslocked);
1426 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1427 			return (error);
1428 		goto restart;
1429 	}
1430 	VATTR_NULL(&vattr);
1431 	vattr.va_type = VFIFO;
1432 	FILEDESC_SLOCK(td->td_proc->p_fd);
1433 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1434 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1435 #ifdef MAC
1436 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1437 	    &vattr);
1438 	if (error)
1439 		goto out;
1440 #endif
1441 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1442 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1443 	if (error == 0)
1444 		vput(nd.ni_vp);
1445 #ifdef MAC
1446 out:
1447 #endif
1448 	vput(nd.ni_dvp);
1449 	vn_finished_write(mp);
1450 	VFS_UNLOCK_GIANT(vfslocked);
1451 	NDFREE(&nd, NDF_ONLY_PNBUF);
1452 	return (error);
1453 }
1454 
1455 /*
1456  * Make a hard file link.
1457  */
1458 #ifndef _SYS_SYSPROTO_H_
1459 struct link_args {
1460 	char	*path;
1461 	char	*link;
1462 };
1463 #endif
1464 int
1465 link(td, uap)
1466 	struct thread *td;
1467 	register struct link_args /* {
1468 		char *path;
1469 		char *link;
1470 	} */ *uap;
1471 {
1472 
1473 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1474 }
1475 
1476 #ifndef _SYS_SYSPROTO_H_
1477 struct linkat_args {
1478 	int	fd1;
1479 	char	*path1;
1480 	int	fd2;
1481 	char	*path2;
1482 	int	flag;
1483 };
1484 #endif
1485 int
1486 linkat(struct thread *td, struct linkat_args *uap)
1487 {
1488 	int flag;
1489 
1490 	flag = uap->flag;
1491 	if (flag & ~AT_SYMLINK_FOLLOW)
1492 		return (EINVAL);
1493 
1494 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1495 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1496 }
1497 
1498 static int hardlink_check_uid = 0;
1499 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1500     &hardlink_check_uid, 0,
1501     "Unprivileged processes cannot create hard links to files owned by other "
1502     "users");
1503 static int hardlink_check_gid = 0;
1504 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1505     &hardlink_check_gid, 0,
1506     "Unprivileged processes cannot create hard links to files owned by other "
1507     "groups");
1508 
1509 static int
1510 can_hardlink(struct vnode *vp, struct ucred *cred)
1511 {
1512 	struct vattr va;
1513 	int error;
1514 
1515 	if (!hardlink_check_uid && !hardlink_check_gid)
1516 		return (0);
1517 
1518 	error = VOP_GETATTR(vp, &va, cred);
1519 	if (error != 0)
1520 		return (error);
1521 
1522 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1523 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1524 		if (error)
1525 			return (error);
1526 	}
1527 
1528 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1529 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1530 		if (error)
1531 			return (error);
1532 	}
1533 
1534 	return (0);
1535 }
1536 
1537 int
1538 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1539 {
1540 
1541 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1542 }
1543 
1544 int
1545 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1546     enum uio_seg segflg, int follow)
1547 {
1548 	struct vnode *vp;
1549 	struct mount *mp;
1550 	struct nameidata nd;
1551 	int vfslocked;
1552 	int lvfslocked;
1553 	int error;
1554 
1555 	bwillwrite();
1556 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1557 	    fd1, td);
1558 
1559 	if ((error = namei(&nd)) != 0)
1560 		return (error);
1561 	vfslocked = NDHASGIANT(&nd);
1562 	NDFREE(&nd, NDF_ONLY_PNBUF);
1563 	vp = nd.ni_vp;
1564 	if (vp->v_type == VDIR) {
1565 		vrele(vp);
1566 		VFS_UNLOCK_GIANT(vfslocked);
1567 		return (EPERM);		/* POSIX */
1568 	}
1569 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1570 		vrele(vp);
1571 		VFS_UNLOCK_GIANT(vfslocked);
1572 		return (error);
1573 	}
1574 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1575 	    segflg, path2, fd2, td);
1576 	if ((error = namei(&nd)) == 0) {
1577 		lvfslocked = NDHASGIANT(&nd);
1578 		if (nd.ni_vp != NULL) {
1579 			if (nd.ni_dvp == nd.ni_vp)
1580 				vrele(nd.ni_dvp);
1581 			else
1582 				vput(nd.ni_dvp);
1583 			vrele(nd.ni_vp);
1584 			error = EEXIST;
1585 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1586 		    == 0) {
1587 			VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1588 			VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1589 			error = can_hardlink(vp, td->td_ucred);
1590 			if (error == 0)
1591 #ifdef MAC
1592 				error = mac_vnode_check_link(td->td_ucred,
1593 				    nd.ni_dvp, vp, &nd.ni_cnd);
1594 			if (error == 0)
1595 #endif
1596 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1597 			VOP_UNLOCK(vp, 0);
1598 			vput(nd.ni_dvp);
1599 		}
1600 		NDFREE(&nd, NDF_ONLY_PNBUF);
1601 		VFS_UNLOCK_GIANT(lvfslocked);
1602 	}
1603 	vrele(vp);
1604 	vn_finished_write(mp);
1605 	VFS_UNLOCK_GIANT(vfslocked);
1606 	return (error);
1607 }
1608 
1609 /*
1610  * Make a symbolic link.
1611  */
1612 #ifndef _SYS_SYSPROTO_H_
1613 struct symlink_args {
1614 	char	*path;
1615 	char	*link;
1616 };
1617 #endif
1618 int
1619 symlink(td, uap)
1620 	struct thread *td;
1621 	register struct symlink_args /* {
1622 		char *path;
1623 		char *link;
1624 	} */ *uap;
1625 {
1626 
1627 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1628 }
1629 
1630 #ifndef _SYS_SYSPROTO_H_
1631 struct symlinkat_args {
1632 	char	*path;
1633 	int	fd;
1634 	char	*path2;
1635 };
1636 #endif
1637 int
1638 symlinkat(struct thread *td, struct symlinkat_args *uap)
1639 {
1640 
1641 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1642 	    UIO_USERSPACE));
1643 }
1644 
1645 int
1646 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1647 {
1648 
1649 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1650 }
1651 
1652 int
1653 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1654     enum uio_seg segflg)
1655 {
1656 	struct mount *mp;
1657 	struct vattr vattr;
1658 	char *syspath;
1659 	int error;
1660 	struct nameidata nd;
1661 	int vfslocked;
1662 
1663 	if (segflg == UIO_SYSSPACE) {
1664 		syspath = path1;
1665 	} else {
1666 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1667 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1668 			goto out;
1669 	}
1670 	AUDIT_ARG(text, syspath);
1671 restart:
1672 	bwillwrite();
1673 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1674 	    segflg, path2, fd, td);
1675 	if ((error = namei(&nd)) != 0)
1676 		goto out;
1677 	vfslocked = NDHASGIANT(&nd);
1678 	if (nd.ni_vp) {
1679 		NDFREE(&nd, NDF_ONLY_PNBUF);
1680 		if (nd.ni_vp == nd.ni_dvp)
1681 			vrele(nd.ni_dvp);
1682 		else
1683 			vput(nd.ni_dvp);
1684 		vrele(nd.ni_vp);
1685 		VFS_UNLOCK_GIANT(vfslocked);
1686 		error = EEXIST;
1687 		goto out;
1688 	}
1689 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1690 		NDFREE(&nd, NDF_ONLY_PNBUF);
1691 		vput(nd.ni_dvp);
1692 		VFS_UNLOCK_GIANT(vfslocked);
1693 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1694 			goto out;
1695 		goto restart;
1696 	}
1697 	VATTR_NULL(&vattr);
1698 	FILEDESC_SLOCK(td->td_proc->p_fd);
1699 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1700 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1701 #ifdef MAC
1702 	vattr.va_type = VLNK;
1703 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1704 	    &vattr);
1705 	if (error)
1706 		goto out2;
1707 #endif
1708 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1709 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1710 	if (error == 0)
1711 		vput(nd.ni_vp);
1712 #ifdef MAC
1713 out2:
1714 #endif
1715 	NDFREE(&nd, NDF_ONLY_PNBUF);
1716 	vput(nd.ni_dvp);
1717 	vn_finished_write(mp);
1718 	VFS_UNLOCK_GIANT(vfslocked);
1719 out:
1720 	if (segflg != UIO_SYSSPACE)
1721 		uma_zfree(namei_zone, syspath);
1722 	return (error);
1723 }
1724 
1725 /*
1726  * Delete a whiteout from the filesystem.
1727  */
1728 int
1729 undelete(td, uap)
1730 	struct thread *td;
1731 	register struct undelete_args /* {
1732 		char *path;
1733 	} */ *uap;
1734 {
1735 	int error;
1736 	struct mount *mp;
1737 	struct nameidata nd;
1738 	int vfslocked;
1739 
1740 restart:
1741 	bwillwrite();
1742 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1743 	    UIO_USERSPACE, uap->path, td);
1744 	error = namei(&nd);
1745 	if (error)
1746 		return (error);
1747 	vfslocked = NDHASGIANT(&nd);
1748 
1749 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1750 		NDFREE(&nd, NDF_ONLY_PNBUF);
1751 		if (nd.ni_vp == nd.ni_dvp)
1752 			vrele(nd.ni_dvp);
1753 		else
1754 			vput(nd.ni_dvp);
1755 		if (nd.ni_vp)
1756 			vrele(nd.ni_vp);
1757 		VFS_UNLOCK_GIANT(vfslocked);
1758 		return (EEXIST);
1759 	}
1760 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1761 		NDFREE(&nd, NDF_ONLY_PNBUF);
1762 		vput(nd.ni_dvp);
1763 		VFS_UNLOCK_GIANT(vfslocked);
1764 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1765 			return (error);
1766 		goto restart;
1767 	}
1768 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1769 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1770 	NDFREE(&nd, NDF_ONLY_PNBUF);
1771 	vput(nd.ni_dvp);
1772 	vn_finished_write(mp);
1773 	VFS_UNLOCK_GIANT(vfslocked);
1774 	return (error);
1775 }
1776 
1777 /*
1778  * Delete a name from the filesystem.
1779  */
1780 #ifndef _SYS_SYSPROTO_H_
1781 struct unlink_args {
1782 	char	*path;
1783 };
1784 #endif
1785 int
1786 unlink(td, uap)
1787 	struct thread *td;
1788 	struct unlink_args /* {
1789 		char *path;
1790 	} */ *uap;
1791 {
1792 
1793 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1794 }
1795 
1796 #ifndef _SYS_SYSPROTO_H_
1797 struct unlinkat_args {
1798 	int	fd;
1799 	char	*path;
1800 	int	flag;
1801 };
1802 #endif
1803 int
1804 unlinkat(struct thread *td, struct unlinkat_args *uap)
1805 {
1806 	int flag = uap->flag;
1807 	int fd = uap->fd;
1808 	char *path = uap->path;
1809 
1810 	if (flag & ~AT_REMOVEDIR)
1811 		return (EINVAL);
1812 
1813 	if (flag & AT_REMOVEDIR)
1814 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1815 	else
1816 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1817 }
1818 
1819 int
1820 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1821 {
1822 
1823 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1824 }
1825 
1826 int
1827 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1828 {
1829 	struct mount *mp;
1830 	struct vnode *vp;
1831 	int error;
1832 	struct nameidata nd;
1833 	int vfslocked;
1834 
1835 restart:
1836 	bwillwrite();
1837 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1838 	    pathseg, path, fd, td);
1839 	if ((error = namei(&nd)) != 0)
1840 		return (error == EINVAL ? EPERM : error);
1841 	vfslocked = NDHASGIANT(&nd);
1842 	vp = nd.ni_vp;
1843 	if (vp->v_type == VDIR)
1844 		error = EPERM;		/* POSIX */
1845 	else {
1846 		/*
1847 		 * The root of a mounted filesystem cannot be deleted.
1848 		 *
1849 		 * XXX: can this only be a VDIR case?
1850 		 */
1851 		if (vp->v_vflag & VV_ROOT)
1852 			error = EBUSY;
1853 	}
1854 	if (error == 0) {
1855 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1856 			NDFREE(&nd, NDF_ONLY_PNBUF);
1857 			vput(nd.ni_dvp);
1858 			if (vp == nd.ni_dvp)
1859 				vrele(vp);
1860 			else
1861 				vput(vp);
1862 			VFS_UNLOCK_GIANT(vfslocked);
1863 			if ((error = vn_start_write(NULL, &mp,
1864 			    V_XSLEEP | PCATCH)) != 0)
1865 				return (error);
1866 			goto restart;
1867 		}
1868 #ifdef MAC
1869 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1870 		    &nd.ni_cnd);
1871 		if (error)
1872 			goto out;
1873 #endif
1874 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1875 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1876 #ifdef MAC
1877 out:
1878 #endif
1879 		vn_finished_write(mp);
1880 	}
1881 	NDFREE(&nd, NDF_ONLY_PNBUF);
1882 	vput(nd.ni_dvp);
1883 	if (vp == nd.ni_dvp)
1884 		vrele(vp);
1885 	else
1886 		vput(vp);
1887 	VFS_UNLOCK_GIANT(vfslocked);
1888 	return (error);
1889 }
1890 
1891 /*
1892  * Reposition read/write file offset.
1893  */
1894 #ifndef _SYS_SYSPROTO_H_
1895 struct lseek_args {
1896 	int	fd;
1897 	int	pad;
1898 	off_t	offset;
1899 	int	whence;
1900 };
1901 #endif
1902 int
1903 lseek(td, uap)
1904 	struct thread *td;
1905 	register struct lseek_args /* {
1906 		int fd;
1907 		int pad;
1908 		off_t offset;
1909 		int whence;
1910 	} */ *uap;
1911 {
1912 	struct ucred *cred = td->td_ucred;
1913 	struct file *fp;
1914 	struct vnode *vp;
1915 	struct vattr vattr;
1916 	off_t offset;
1917 	int error, noneg;
1918 	int vfslocked;
1919 
1920 	if ((error = fget(td, uap->fd, &fp)) != 0)
1921 		return (error);
1922 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1923 		fdrop(fp, td);
1924 		return (ESPIPE);
1925 	}
1926 	vp = fp->f_vnode;
1927 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1928 	noneg = (vp->v_type != VCHR);
1929 	offset = uap->offset;
1930 	switch (uap->whence) {
1931 	case L_INCR:
1932 		if (noneg &&
1933 		    (fp->f_offset < 0 ||
1934 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1935 			error = EOVERFLOW;
1936 			break;
1937 		}
1938 		offset += fp->f_offset;
1939 		break;
1940 	case L_XTND:
1941 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1942 		error = VOP_GETATTR(vp, &vattr, cred);
1943 		VOP_UNLOCK(vp, 0);
1944 		if (error)
1945 			break;
1946 		if (noneg &&
1947 		    (vattr.va_size > OFF_MAX ||
1948 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1949 			error = EOVERFLOW;
1950 			break;
1951 		}
1952 		offset += vattr.va_size;
1953 		break;
1954 	case L_SET:
1955 		break;
1956 	case SEEK_DATA:
1957 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1958 		break;
1959 	case SEEK_HOLE:
1960 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1961 		break;
1962 	default:
1963 		error = EINVAL;
1964 	}
1965 	if (error == 0 && noneg && offset < 0)
1966 		error = EINVAL;
1967 	if (error != 0)
1968 		goto drop;
1969 	fp->f_offset = offset;
1970 	*(off_t *)(td->td_retval) = fp->f_offset;
1971 drop:
1972 	fdrop(fp, td);
1973 	VFS_UNLOCK_GIANT(vfslocked);
1974 	return (error);
1975 }
1976 
1977 #if defined(COMPAT_43)
1978 /*
1979  * Reposition read/write file offset.
1980  */
1981 #ifndef _SYS_SYSPROTO_H_
1982 struct olseek_args {
1983 	int	fd;
1984 	long	offset;
1985 	int	whence;
1986 };
1987 #endif
1988 int
1989 olseek(td, uap)
1990 	struct thread *td;
1991 	register struct olseek_args /* {
1992 		int fd;
1993 		long offset;
1994 		int whence;
1995 	} */ *uap;
1996 {
1997 	struct lseek_args /* {
1998 		int fd;
1999 		int pad;
2000 		off_t offset;
2001 		int whence;
2002 	} */ nuap;
2003 
2004 	nuap.fd = uap->fd;
2005 	nuap.offset = uap->offset;
2006 	nuap.whence = uap->whence;
2007 	return (lseek(td, &nuap));
2008 }
2009 #endif /* COMPAT_43 */
2010 
2011 /* Version with the 'pad' argument */
2012 int
2013 freebsd6_lseek(td, uap)
2014 	struct thread *td;
2015 	register struct freebsd6_lseek_args *uap;
2016 {
2017 	struct lseek_args ouap;
2018 
2019 	ouap.fd = uap->fd;
2020 	ouap.offset = uap->offset;
2021 	ouap.whence = uap->whence;
2022 	return (lseek(td, &ouap));
2023 }
2024 
2025 /*
2026  * Check access permissions using passed credentials.
2027  */
2028 static int
2029 vn_access(vp, user_flags, cred, td)
2030 	struct vnode	*vp;
2031 	int		user_flags;
2032 	struct ucred	*cred;
2033 	struct thread	*td;
2034 {
2035 	int error;
2036 	accmode_t accmode;
2037 
2038 	/* Flags == 0 means only check for existence. */
2039 	error = 0;
2040 	if (user_flags) {
2041 		accmode = 0;
2042 		if (user_flags & R_OK)
2043 			accmode |= VREAD;
2044 		if (user_flags & W_OK)
2045 			accmode |= VWRITE;
2046 		if (user_flags & X_OK)
2047 			accmode |= VEXEC;
2048 #ifdef MAC
2049 		error = mac_vnode_check_access(cred, vp, accmode);
2050 		if (error)
2051 			return (error);
2052 #endif
2053 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2054 			error = VOP_ACCESS(vp, accmode, cred, td);
2055 	}
2056 	return (error);
2057 }
2058 
2059 /*
2060  * Check access permissions using "real" credentials.
2061  */
2062 #ifndef _SYS_SYSPROTO_H_
2063 struct access_args {
2064 	char	*path;
2065 	int	flags;
2066 };
2067 #endif
2068 int
2069 access(td, uap)
2070 	struct thread *td;
2071 	register struct access_args /* {
2072 		char *path;
2073 		int flags;
2074 	} */ *uap;
2075 {
2076 
2077 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2078 }
2079 
2080 #ifndef _SYS_SYSPROTO_H_
2081 struct faccessat_args {
2082 	int	dirfd;
2083 	char	*path;
2084 	int	mode;
2085 	int	flag;
2086 }
2087 #endif
2088 int
2089 faccessat(struct thread *td, struct faccessat_args *uap)
2090 {
2091 
2092 	if (uap->flag & ~AT_EACCESS)
2093 		return (EINVAL);
2094 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2095 	    uap->mode));
2096 }
2097 
2098 int
2099 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2100 {
2101 
2102 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2103 }
2104 
2105 int
2106 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2107     int flags, int mode)
2108 {
2109 	struct ucred *cred, *tmpcred;
2110 	struct vnode *vp;
2111 	struct nameidata nd;
2112 	int vfslocked;
2113 	int error;
2114 
2115 	/*
2116 	 * Create and modify a temporary credential instead of one that
2117 	 * is potentially shared.  This could also mess up socket
2118 	 * buffer accounting which can run in an interrupt context.
2119 	 */
2120 	if (!(flags & AT_EACCESS)) {
2121 		cred = td->td_ucred;
2122 		tmpcred = crdup(cred);
2123 		tmpcred->cr_uid = cred->cr_ruid;
2124 		tmpcred->cr_groups[0] = cred->cr_rgid;
2125 		td->td_ucred = tmpcred;
2126 	} else
2127 		cred = tmpcred = td->td_ucred;
2128 	NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2129 	    pathseg, path, fd, td);
2130 	if ((error = namei(&nd)) != 0)
2131 		goto out1;
2132 	vfslocked = NDHASGIANT(&nd);
2133 	vp = nd.ni_vp;
2134 
2135 	error = vn_access(vp, mode, tmpcred, td);
2136 	NDFREE(&nd, NDF_ONLY_PNBUF);
2137 	vput(vp);
2138 	VFS_UNLOCK_GIANT(vfslocked);
2139 out1:
2140 	if (!(flags & AT_EACCESS)) {
2141 		td->td_ucred = cred;
2142 		crfree(tmpcred);
2143 	}
2144 	return (error);
2145 }
2146 
2147 /*
2148  * Check access permissions using "effective" credentials.
2149  */
2150 #ifndef _SYS_SYSPROTO_H_
2151 struct eaccess_args {
2152 	char	*path;
2153 	int	flags;
2154 };
2155 #endif
2156 int
2157 eaccess(td, uap)
2158 	struct thread *td;
2159 	register struct eaccess_args /* {
2160 		char *path;
2161 		int flags;
2162 	} */ *uap;
2163 {
2164 
2165 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2166 }
2167 
2168 int
2169 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2170 {
2171 
2172 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2173 }
2174 
2175 #if defined(COMPAT_43)
2176 /*
2177  * Get file status; this version follows links.
2178  */
2179 #ifndef _SYS_SYSPROTO_H_
2180 struct ostat_args {
2181 	char	*path;
2182 	struct ostat *ub;
2183 };
2184 #endif
2185 int
2186 ostat(td, uap)
2187 	struct thread *td;
2188 	register struct ostat_args /* {
2189 		char *path;
2190 		struct ostat *ub;
2191 	} */ *uap;
2192 {
2193 	struct stat sb;
2194 	struct ostat osb;
2195 	int error;
2196 
2197 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2198 	if (error)
2199 		return (error);
2200 	cvtstat(&sb, &osb);
2201 	error = copyout(&osb, uap->ub, sizeof (osb));
2202 	return (error);
2203 }
2204 
2205 /*
2206  * Get file status; this version does not follow links.
2207  */
2208 #ifndef _SYS_SYSPROTO_H_
2209 struct olstat_args {
2210 	char	*path;
2211 	struct ostat *ub;
2212 };
2213 #endif
2214 int
2215 olstat(td, uap)
2216 	struct thread *td;
2217 	register struct olstat_args /* {
2218 		char *path;
2219 		struct ostat *ub;
2220 	} */ *uap;
2221 {
2222 	struct stat sb;
2223 	struct ostat osb;
2224 	int error;
2225 
2226 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2227 	if (error)
2228 		return (error);
2229 	cvtstat(&sb, &osb);
2230 	error = copyout(&osb, uap->ub, sizeof (osb));
2231 	return (error);
2232 }
2233 
2234 /*
2235  * Convert from an old to a new stat structure.
2236  */
2237 void
2238 cvtstat(st, ost)
2239 	struct stat *st;
2240 	struct ostat *ost;
2241 {
2242 
2243 	ost->st_dev = st->st_dev;
2244 	ost->st_ino = st->st_ino;
2245 	ost->st_mode = st->st_mode;
2246 	ost->st_nlink = st->st_nlink;
2247 	ost->st_uid = st->st_uid;
2248 	ost->st_gid = st->st_gid;
2249 	ost->st_rdev = st->st_rdev;
2250 	if (st->st_size < (quad_t)1 << 32)
2251 		ost->st_size = st->st_size;
2252 	else
2253 		ost->st_size = -2;
2254 	ost->st_atime = st->st_atime;
2255 	ost->st_mtime = st->st_mtime;
2256 	ost->st_ctime = st->st_ctime;
2257 	ost->st_blksize = st->st_blksize;
2258 	ost->st_blocks = st->st_blocks;
2259 	ost->st_flags = st->st_flags;
2260 	ost->st_gen = st->st_gen;
2261 }
2262 #endif /* COMPAT_43 */
2263 
2264 /*
2265  * Get file status; this version follows links.
2266  */
2267 #ifndef _SYS_SYSPROTO_H_
2268 struct stat_args {
2269 	char	*path;
2270 	struct stat *ub;
2271 };
2272 #endif
2273 int
2274 stat(td, uap)
2275 	struct thread *td;
2276 	register struct stat_args /* {
2277 		char *path;
2278 		struct stat *ub;
2279 	} */ *uap;
2280 {
2281 	struct stat sb;
2282 	int error;
2283 
2284 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2285 	if (error == 0)
2286 		error = copyout(&sb, uap->ub, sizeof (sb));
2287 	return (error);
2288 }
2289 
2290 #ifndef _SYS_SYSPROTO_H_
2291 struct fstatat_args {
2292 	int	fd;
2293 	char	*path;
2294 	struct stat	*buf;
2295 	int	flag;
2296 }
2297 #endif
2298 int
2299 fstatat(struct thread *td, struct fstatat_args *uap)
2300 {
2301 	struct stat sb;
2302 	int error;
2303 
2304 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2305 	    UIO_USERSPACE, &sb);
2306 	if (error == 0)
2307 		error = copyout(&sb, uap->buf, sizeof (sb));
2308 	return (error);
2309 }
2310 
2311 int
2312 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2313 {
2314 
2315 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2316 }
2317 
2318 int
2319 kern_statat(struct thread *td, int flag, int fd, char *path,
2320     enum uio_seg pathseg, struct stat *sbp)
2321 {
2322 	struct nameidata nd;
2323 	struct stat sb;
2324 	int error, vfslocked;
2325 
2326 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2327 		return (EINVAL);
2328 
2329 	NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2330 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2331 	    path, fd, td);
2332 
2333 	if ((error = namei(&nd)) != 0)
2334 		return (error);
2335 	vfslocked = NDHASGIANT(&nd);
2336 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2337 	NDFREE(&nd, NDF_ONLY_PNBUF);
2338 	vput(nd.ni_vp);
2339 	VFS_UNLOCK_GIANT(vfslocked);
2340 	if (mtx_owned(&Giant))
2341 		printf("stat(%d): %s\n", vfslocked, path);
2342 	if (error)
2343 		return (error);
2344 	*sbp = sb;
2345 #ifdef KTRACE
2346 	if (KTRPOINT(td, KTR_STRUCT))
2347 		ktrstat(&sb);
2348 #endif
2349 	return (0);
2350 }
2351 
2352 /*
2353  * Get file status; this version does not follow links.
2354  */
2355 #ifndef _SYS_SYSPROTO_H_
2356 struct lstat_args {
2357 	char	*path;
2358 	struct stat *ub;
2359 };
2360 #endif
2361 int
2362 lstat(td, uap)
2363 	struct thread *td;
2364 	register struct lstat_args /* {
2365 		char *path;
2366 		struct stat *ub;
2367 	} */ *uap;
2368 {
2369 	struct stat sb;
2370 	int error;
2371 
2372 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2373 	if (error == 0)
2374 		error = copyout(&sb, uap->ub, sizeof (sb));
2375 	return (error);
2376 }
2377 
2378 int
2379 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2380 {
2381 
2382 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2383 	    sbp));
2384 }
2385 
2386 /*
2387  * Implementation of the NetBSD [l]stat() functions.
2388  */
2389 void
2390 cvtnstat(sb, nsb)
2391 	struct stat *sb;
2392 	struct nstat *nsb;
2393 {
2394 	bzero(nsb, sizeof *nsb);
2395 	nsb->st_dev = sb->st_dev;
2396 	nsb->st_ino = sb->st_ino;
2397 	nsb->st_mode = sb->st_mode;
2398 	nsb->st_nlink = sb->st_nlink;
2399 	nsb->st_uid = sb->st_uid;
2400 	nsb->st_gid = sb->st_gid;
2401 	nsb->st_rdev = sb->st_rdev;
2402 	nsb->st_atimespec = sb->st_atimespec;
2403 	nsb->st_mtimespec = sb->st_mtimespec;
2404 	nsb->st_ctimespec = sb->st_ctimespec;
2405 	nsb->st_size = sb->st_size;
2406 	nsb->st_blocks = sb->st_blocks;
2407 	nsb->st_blksize = sb->st_blksize;
2408 	nsb->st_flags = sb->st_flags;
2409 	nsb->st_gen = sb->st_gen;
2410 	nsb->st_birthtimespec = sb->st_birthtimespec;
2411 }
2412 
2413 #ifndef _SYS_SYSPROTO_H_
2414 struct nstat_args {
2415 	char	*path;
2416 	struct nstat *ub;
2417 };
2418 #endif
2419 int
2420 nstat(td, uap)
2421 	struct thread *td;
2422 	register struct nstat_args /* {
2423 		char *path;
2424 		struct nstat *ub;
2425 	} */ *uap;
2426 {
2427 	struct stat sb;
2428 	struct nstat nsb;
2429 	int error;
2430 
2431 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2432 	if (error)
2433 		return (error);
2434 	cvtnstat(&sb, &nsb);
2435 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2436 	return (error);
2437 }
2438 
2439 /*
2440  * NetBSD lstat.  Get file status; this version does not follow links.
2441  */
2442 #ifndef _SYS_SYSPROTO_H_
2443 struct lstat_args {
2444 	char	*path;
2445 	struct stat *ub;
2446 };
2447 #endif
2448 int
2449 nlstat(td, uap)
2450 	struct thread *td;
2451 	register struct nlstat_args /* {
2452 		char *path;
2453 		struct nstat *ub;
2454 	} */ *uap;
2455 {
2456 	struct stat sb;
2457 	struct nstat nsb;
2458 	int error;
2459 
2460 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2461 	if (error)
2462 		return (error);
2463 	cvtnstat(&sb, &nsb);
2464 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2465 	return (error);
2466 }
2467 
2468 /*
2469  * Get configurable pathname variables.
2470  */
2471 #ifndef _SYS_SYSPROTO_H_
2472 struct pathconf_args {
2473 	char	*path;
2474 	int	name;
2475 };
2476 #endif
2477 int
2478 pathconf(td, uap)
2479 	struct thread *td;
2480 	register struct pathconf_args /* {
2481 		char *path;
2482 		int name;
2483 	} */ *uap;
2484 {
2485 
2486 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name));
2487 }
2488 
2489 int
2490 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name)
2491 {
2492 	struct nameidata nd;
2493 	int error, vfslocked;
2494 
2495 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2496 	    pathseg, path, td);
2497 	if ((error = namei(&nd)) != 0)
2498 		return (error);
2499 	vfslocked = NDHASGIANT(&nd);
2500 	NDFREE(&nd, NDF_ONLY_PNBUF);
2501 
2502 	/* If asynchronous I/O is available, it works for all files. */
2503 	if (name == _PC_ASYNC_IO)
2504 		td->td_retval[0] = async_io_version;
2505 	else
2506 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2507 	vput(nd.ni_vp);
2508 	VFS_UNLOCK_GIANT(vfslocked);
2509 	return (error);
2510 }
2511 
2512 /*
2513  * Return target name of a symbolic link.
2514  */
2515 #ifndef _SYS_SYSPROTO_H_
2516 struct readlink_args {
2517 	char	*path;
2518 	char	*buf;
2519 	size_t	count;
2520 };
2521 #endif
2522 int
2523 readlink(td, uap)
2524 	struct thread *td;
2525 	register struct readlink_args /* {
2526 		char *path;
2527 		char *buf;
2528 		size_t count;
2529 	} */ *uap;
2530 {
2531 
2532 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2533 	    UIO_USERSPACE, uap->count));
2534 }
2535 #ifndef _SYS_SYSPROTO_H_
2536 struct readlinkat_args {
2537 	int	fd;
2538 	char	*path;
2539 	char	*buf;
2540 	size_t	bufsize;
2541 };
2542 #endif
2543 int
2544 readlinkat(struct thread *td, struct readlinkat_args *uap)
2545 {
2546 
2547 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2548 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2549 }
2550 
2551 int
2552 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2553     enum uio_seg bufseg, size_t count)
2554 {
2555 
2556 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2557 	    count));
2558 }
2559 
2560 int
2561 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2562     char *buf, enum uio_seg bufseg, size_t count)
2563 {
2564 	struct vnode *vp;
2565 	struct iovec aiov;
2566 	struct uio auio;
2567 	int error;
2568 	struct nameidata nd;
2569 	int vfslocked;
2570 
2571 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2572 	    pathseg, path, fd, td);
2573 
2574 	if ((error = namei(&nd)) != 0)
2575 		return (error);
2576 	NDFREE(&nd, NDF_ONLY_PNBUF);
2577 	vfslocked = NDHASGIANT(&nd);
2578 	vp = nd.ni_vp;
2579 #ifdef MAC
2580 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2581 	if (error) {
2582 		vput(vp);
2583 		VFS_UNLOCK_GIANT(vfslocked);
2584 		return (error);
2585 	}
2586 #endif
2587 	if (vp->v_type != VLNK)
2588 		error = EINVAL;
2589 	else {
2590 		aiov.iov_base = buf;
2591 		aiov.iov_len = count;
2592 		auio.uio_iov = &aiov;
2593 		auio.uio_iovcnt = 1;
2594 		auio.uio_offset = 0;
2595 		auio.uio_rw = UIO_READ;
2596 		auio.uio_segflg = bufseg;
2597 		auio.uio_td = td;
2598 		auio.uio_resid = count;
2599 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2600 	}
2601 	vput(vp);
2602 	VFS_UNLOCK_GIANT(vfslocked);
2603 	td->td_retval[0] = count - auio.uio_resid;
2604 	return (error);
2605 }
2606 
2607 /*
2608  * Common implementation code for chflags() and fchflags().
2609  */
2610 static int
2611 setfflags(td, vp, flags)
2612 	struct thread *td;
2613 	struct vnode *vp;
2614 	int flags;
2615 {
2616 	int error;
2617 	struct mount *mp;
2618 	struct vattr vattr;
2619 
2620 	/*
2621 	 * Prevent non-root users from setting flags on devices.  When
2622 	 * a device is reused, users can retain ownership of the device
2623 	 * if they are allowed to set flags and programs assume that
2624 	 * chown can't fail when done as root.
2625 	 */
2626 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2627 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2628 		if (error)
2629 			return (error);
2630 	}
2631 
2632 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2633 		return (error);
2634 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2635 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2636 	VATTR_NULL(&vattr);
2637 	vattr.va_flags = flags;
2638 #ifdef MAC
2639 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2640 	if (error == 0)
2641 #endif
2642 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2643 	VOP_UNLOCK(vp, 0);
2644 	vn_finished_write(mp);
2645 	return (error);
2646 }
2647 
2648 /*
2649  * Change flags of a file given a path name.
2650  */
2651 #ifndef _SYS_SYSPROTO_H_
2652 struct chflags_args {
2653 	char	*path;
2654 	int	flags;
2655 };
2656 #endif
2657 int
2658 chflags(td, uap)
2659 	struct thread *td;
2660 	register struct chflags_args /* {
2661 		char *path;
2662 		int flags;
2663 	} */ *uap;
2664 {
2665 	int error;
2666 	struct nameidata nd;
2667 	int vfslocked;
2668 
2669 	AUDIT_ARG(fflags, uap->flags);
2670 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2671 	    uap->path, td);
2672 	if ((error = namei(&nd)) != 0)
2673 		return (error);
2674 	NDFREE(&nd, NDF_ONLY_PNBUF);
2675 	vfslocked = NDHASGIANT(&nd);
2676 	error = setfflags(td, nd.ni_vp, uap->flags);
2677 	vrele(nd.ni_vp);
2678 	VFS_UNLOCK_GIANT(vfslocked);
2679 	return (error);
2680 }
2681 
2682 /*
2683  * Same as chflags() but doesn't follow symlinks.
2684  */
2685 int
2686 lchflags(td, uap)
2687 	struct thread *td;
2688 	register struct lchflags_args /* {
2689 		char *path;
2690 		int flags;
2691 	} */ *uap;
2692 {
2693 	int error;
2694 	struct nameidata nd;
2695 	int vfslocked;
2696 
2697 	AUDIT_ARG(fflags, uap->flags);
2698 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2699 	    uap->path, td);
2700 	if ((error = namei(&nd)) != 0)
2701 		return (error);
2702 	vfslocked = NDHASGIANT(&nd);
2703 	NDFREE(&nd, NDF_ONLY_PNBUF);
2704 	error = setfflags(td, nd.ni_vp, uap->flags);
2705 	vrele(nd.ni_vp);
2706 	VFS_UNLOCK_GIANT(vfslocked);
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Change flags of a file given a file descriptor.
2712  */
2713 #ifndef _SYS_SYSPROTO_H_
2714 struct fchflags_args {
2715 	int	fd;
2716 	int	flags;
2717 };
2718 #endif
2719 int
2720 fchflags(td, uap)
2721 	struct thread *td;
2722 	register struct fchflags_args /* {
2723 		int fd;
2724 		int flags;
2725 	} */ *uap;
2726 {
2727 	struct file *fp;
2728 	int vfslocked;
2729 	int error;
2730 
2731 	AUDIT_ARG(fd, uap->fd);
2732 	AUDIT_ARG(fflags, uap->flags);
2733 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2734 		return (error);
2735 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2736 #ifdef AUDIT
2737 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2738 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2739 	VOP_UNLOCK(fp->f_vnode, 0);
2740 #endif
2741 	error = setfflags(td, fp->f_vnode, uap->flags);
2742 	VFS_UNLOCK_GIANT(vfslocked);
2743 	fdrop(fp, td);
2744 	return (error);
2745 }
2746 
2747 /*
2748  * Common implementation code for chmod(), lchmod() and fchmod().
2749  */
2750 static int
2751 setfmode(td, vp, mode)
2752 	struct thread *td;
2753 	struct vnode *vp;
2754 	int mode;
2755 {
2756 	int error;
2757 	struct mount *mp;
2758 	struct vattr vattr;
2759 
2760 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2761 		return (error);
2762 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2763 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2764 	VATTR_NULL(&vattr);
2765 	vattr.va_mode = mode & ALLPERMS;
2766 #ifdef MAC
2767 	error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2768 	if (error == 0)
2769 #endif
2770 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2771 	VOP_UNLOCK(vp, 0);
2772 	vn_finished_write(mp);
2773 	return (error);
2774 }
2775 
2776 /*
2777  * Change mode of a file given path name.
2778  */
2779 #ifndef _SYS_SYSPROTO_H_
2780 struct chmod_args {
2781 	char	*path;
2782 	int	mode;
2783 };
2784 #endif
2785 int
2786 chmod(td, uap)
2787 	struct thread *td;
2788 	register struct chmod_args /* {
2789 		char *path;
2790 		int mode;
2791 	} */ *uap;
2792 {
2793 
2794 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2795 }
2796 
2797 #ifndef _SYS_SYSPROTO_H_
2798 struct fchmodat_args {
2799 	int	dirfd;
2800 	char	*path;
2801 	mode_t	mode;
2802 	int	flag;
2803 }
2804 #endif
2805 int
2806 fchmodat(struct thread *td, struct fchmodat_args *uap)
2807 {
2808 	int flag = uap->flag;
2809 	int fd = uap->fd;
2810 	char *path = uap->path;
2811 	mode_t mode = uap->mode;
2812 
2813 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2814 		return (EINVAL);
2815 
2816 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2817 }
2818 
2819 int
2820 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2821 {
2822 
2823 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2824 }
2825 
2826 /*
2827  * Change mode of a file given path name (don't follow links.)
2828  */
2829 #ifndef _SYS_SYSPROTO_H_
2830 struct lchmod_args {
2831 	char	*path;
2832 	int	mode;
2833 };
2834 #endif
2835 int
2836 lchmod(td, uap)
2837 	struct thread *td;
2838 	register struct lchmod_args /* {
2839 		char *path;
2840 		int mode;
2841 	} */ *uap;
2842 {
2843 
2844 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2845 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2846 }
2847 
2848 
2849 int
2850 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2851     mode_t mode, int flag)
2852 {
2853 	int error;
2854 	struct nameidata nd;
2855 	int vfslocked;
2856 	int follow;
2857 
2858 	AUDIT_ARG(mode, mode);
2859 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2860 	NDINIT_AT(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg, path,
2861 	    fd, td);
2862 	if ((error = namei(&nd)) != 0)
2863 		return (error);
2864 	vfslocked = NDHASGIANT(&nd);
2865 	NDFREE(&nd, NDF_ONLY_PNBUF);
2866 	error = setfmode(td, nd.ni_vp, mode);
2867 	vrele(nd.ni_vp);
2868 	VFS_UNLOCK_GIANT(vfslocked);
2869 	return (error);
2870 }
2871 
2872 /*
2873  * Change mode of a file given a file descriptor.
2874  */
2875 #ifndef _SYS_SYSPROTO_H_
2876 struct fchmod_args {
2877 	int	fd;
2878 	int	mode;
2879 };
2880 #endif
2881 int
2882 fchmod(td, uap)
2883 	struct thread *td;
2884 	register struct fchmod_args /* {
2885 		int fd;
2886 		int mode;
2887 	} */ *uap;
2888 {
2889 	struct file *fp;
2890 	int vfslocked;
2891 	int error;
2892 
2893 	AUDIT_ARG(fd, uap->fd);
2894 	AUDIT_ARG(mode, uap->mode);
2895 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2896 		return (error);
2897 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2898 #ifdef AUDIT
2899 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2900 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2901 	VOP_UNLOCK(fp->f_vnode, 0);
2902 #endif
2903 	error = setfmode(td, fp->f_vnode, uap->mode);
2904 	VFS_UNLOCK_GIANT(vfslocked);
2905 	fdrop(fp, td);
2906 	return (error);
2907 }
2908 
2909 /*
2910  * Common implementation for chown(), lchown(), and fchown()
2911  */
2912 static int
2913 setfown(td, vp, uid, gid)
2914 	struct thread *td;
2915 	struct vnode *vp;
2916 	uid_t uid;
2917 	gid_t gid;
2918 {
2919 	int error;
2920 	struct mount *mp;
2921 	struct vattr vattr;
2922 
2923 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2924 		return (error);
2925 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2926 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2927 	VATTR_NULL(&vattr);
2928 	vattr.va_uid = uid;
2929 	vattr.va_gid = gid;
2930 #ifdef MAC
2931 	error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2932 	    vattr.va_gid);
2933 	if (error == 0)
2934 #endif
2935 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2936 	VOP_UNLOCK(vp, 0);
2937 	vn_finished_write(mp);
2938 	return (error);
2939 }
2940 
2941 /*
2942  * Set ownership given a path name.
2943  */
2944 #ifndef _SYS_SYSPROTO_H_
2945 struct chown_args {
2946 	char	*path;
2947 	int	uid;
2948 	int	gid;
2949 };
2950 #endif
2951 int
2952 chown(td, uap)
2953 	struct thread *td;
2954 	register struct chown_args /* {
2955 		char *path;
2956 		int uid;
2957 		int gid;
2958 	} */ *uap;
2959 {
2960 
2961 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2962 }
2963 
2964 #ifndef _SYS_SYSPROTO_H_
2965 struct fchownat_args {
2966 	int fd;
2967 	const char * path;
2968 	uid_t uid;
2969 	gid_t gid;
2970 	int flag;
2971 };
2972 #endif
2973 int
2974 fchownat(struct thread *td, struct fchownat_args *uap)
2975 {
2976 	int flag;
2977 
2978 	flag = uap->flag;
2979 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2980 		return (EINVAL);
2981 
2982 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2983 	    uap->gid, uap->flag));
2984 }
2985 
2986 int
2987 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2988     int gid)
2989 {
2990 
2991 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2992 }
2993 
2994 int
2995 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2996     int uid, int gid, int flag)
2997 {
2998 	struct nameidata nd;
2999 	int error, vfslocked, follow;
3000 
3001 	AUDIT_ARG(owner, uid, gid);
3002 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3003 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
3004 	    fd, td);
3005 
3006 	if ((error = namei(&nd)) != 0)
3007 		return (error);
3008 	vfslocked = NDHASGIANT(&nd);
3009 	NDFREE(&nd, NDF_ONLY_PNBUF);
3010 	error = setfown(td, nd.ni_vp, uid, gid);
3011 	vrele(nd.ni_vp);
3012 	VFS_UNLOCK_GIANT(vfslocked);
3013 	return (error);
3014 }
3015 
3016 /*
3017  * Set ownership given a path name, do not cross symlinks.
3018  */
3019 #ifndef _SYS_SYSPROTO_H_
3020 struct lchown_args {
3021 	char	*path;
3022 	int	uid;
3023 	int	gid;
3024 };
3025 #endif
3026 int
3027 lchown(td, uap)
3028 	struct thread *td;
3029 	register struct lchown_args /* {
3030 		char *path;
3031 		int uid;
3032 		int gid;
3033 	} */ *uap;
3034 {
3035 
3036 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3037 }
3038 
3039 int
3040 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3041     int gid)
3042 {
3043 
3044 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3045 	    AT_SYMLINK_NOFOLLOW));
3046 }
3047 
3048 /*
3049  * Set ownership given a file descriptor.
3050  */
3051 #ifndef _SYS_SYSPROTO_H_
3052 struct fchown_args {
3053 	int	fd;
3054 	int	uid;
3055 	int	gid;
3056 };
3057 #endif
3058 int
3059 fchown(td, uap)
3060 	struct thread *td;
3061 	register struct fchown_args /* {
3062 		int fd;
3063 		int uid;
3064 		int gid;
3065 	} */ *uap;
3066 {
3067 	struct file *fp;
3068 	int vfslocked;
3069 	int error;
3070 
3071 	AUDIT_ARG(fd, uap->fd);
3072 	AUDIT_ARG(owner, uap->uid, uap->gid);
3073 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3074 		return (error);
3075 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3076 #ifdef AUDIT
3077 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
3078 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3079 	VOP_UNLOCK(fp->f_vnode, 0);
3080 #endif
3081 	error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3082 	VFS_UNLOCK_GIANT(vfslocked);
3083 	fdrop(fp, td);
3084 	return (error);
3085 }
3086 
3087 /*
3088  * Common implementation code for utimes(), lutimes(), and futimes().
3089  */
3090 static int
3091 getutimes(usrtvp, tvpseg, tsp)
3092 	const struct timeval *usrtvp;
3093 	enum uio_seg tvpseg;
3094 	struct timespec *tsp;
3095 {
3096 	struct timeval tv[2];
3097 	const struct timeval *tvp;
3098 	int error;
3099 
3100 	if (usrtvp == NULL) {
3101 		microtime(&tv[0]);
3102 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
3103 		tsp[1] = tsp[0];
3104 	} else {
3105 		if (tvpseg == UIO_SYSSPACE) {
3106 			tvp = usrtvp;
3107 		} else {
3108 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3109 				return (error);
3110 			tvp = tv;
3111 		}
3112 
3113 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3114 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3115 			return (EINVAL);
3116 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3117 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3118 	}
3119 	return (0);
3120 }
3121 
3122 /*
3123  * Common implementation code for utimes(), lutimes(), and futimes().
3124  */
3125 static int
3126 setutimes(td, vp, ts, numtimes, nullflag)
3127 	struct thread *td;
3128 	struct vnode *vp;
3129 	const struct timespec *ts;
3130 	int numtimes;
3131 	int nullflag;
3132 {
3133 	int error, setbirthtime;
3134 	struct mount *mp;
3135 	struct vattr vattr;
3136 
3137 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3138 		return (error);
3139 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3140 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3141 	setbirthtime = 0;
3142 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3143 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3144 		setbirthtime = 1;
3145 	VATTR_NULL(&vattr);
3146 	vattr.va_atime = ts[0];
3147 	vattr.va_mtime = ts[1];
3148 	if (setbirthtime)
3149 		vattr.va_birthtime = ts[1];
3150 	if (numtimes > 2)
3151 		vattr.va_birthtime = ts[2];
3152 	if (nullflag)
3153 		vattr.va_vaflags |= VA_UTIMES_NULL;
3154 #ifdef MAC
3155 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3156 	    vattr.va_mtime);
3157 #endif
3158 	if (error == 0)
3159 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3160 	VOP_UNLOCK(vp, 0);
3161 	vn_finished_write(mp);
3162 	return (error);
3163 }
3164 
3165 /*
3166  * Set the access and modification times of a file.
3167  */
3168 #ifndef _SYS_SYSPROTO_H_
3169 struct utimes_args {
3170 	char	*path;
3171 	struct	timeval *tptr;
3172 };
3173 #endif
3174 int
3175 utimes(td, uap)
3176 	struct thread *td;
3177 	register struct utimes_args /* {
3178 		char *path;
3179 		struct timeval *tptr;
3180 	} */ *uap;
3181 {
3182 
3183 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3184 	    UIO_USERSPACE));
3185 }
3186 
3187 #ifndef _SYS_SYSPROTO_H_
3188 struct futimesat_args {
3189 	int fd;
3190 	const char * path;
3191 	const struct timeval * times;
3192 };
3193 #endif
3194 int
3195 futimesat(struct thread *td, struct futimesat_args *uap)
3196 {
3197 
3198 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3199 	    uap->times, UIO_USERSPACE));
3200 }
3201 
3202 int
3203 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3204     struct timeval *tptr, enum uio_seg tptrseg)
3205 {
3206 
3207 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3208 }
3209 
3210 int
3211 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3212     struct timeval *tptr, enum uio_seg tptrseg)
3213 {
3214 	struct nameidata nd;
3215 	struct timespec ts[2];
3216 	int error, vfslocked;
3217 
3218 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3219 		return (error);
3220 	NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3221 	    fd, td);
3222 
3223 	if ((error = namei(&nd)) != 0)
3224 		return (error);
3225 	vfslocked = NDHASGIANT(&nd);
3226 	NDFREE(&nd, NDF_ONLY_PNBUF);
3227 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3228 	vrele(nd.ni_vp);
3229 	VFS_UNLOCK_GIANT(vfslocked);
3230 	return (error);
3231 }
3232 
3233 /*
3234  * Set the access and modification times of a file.
3235  */
3236 #ifndef _SYS_SYSPROTO_H_
3237 struct lutimes_args {
3238 	char	*path;
3239 	struct	timeval *tptr;
3240 };
3241 #endif
3242 int
3243 lutimes(td, uap)
3244 	struct thread *td;
3245 	register struct lutimes_args /* {
3246 		char *path;
3247 		struct timeval *tptr;
3248 	} */ *uap;
3249 {
3250 
3251 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3252 	    UIO_USERSPACE));
3253 }
3254 
3255 int
3256 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3257     struct timeval *tptr, enum uio_seg tptrseg)
3258 {
3259 	struct timespec ts[2];
3260 	int error;
3261 	struct nameidata nd;
3262 	int vfslocked;
3263 
3264 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3265 		return (error);
3266 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3267 	if ((error = namei(&nd)) != 0)
3268 		return (error);
3269 	vfslocked = NDHASGIANT(&nd);
3270 	NDFREE(&nd, NDF_ONLY_PNBUF);
3271 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3272 	vrele(nd.ni_vp);
3273 	VFS_UNLOCK_GIANT(vfslocked);
3274 	return (error);
3275 }
3276 
3277 /*
3278  * Set the access and modification times of a file.
3279  */
3280 #ifndef _SYS_SYSPROTO_H_
3281 struct futimes_args {
3282 	int	fd;
3283 	struct	timeval *tptr;
3284 };
3285 #endif
3286 int
3287 futimes(td, uap)
3288 	struct thread *td;
3289 	register struct futimes_args /* {
3290 		int  fd;
3291 		struct timeval *tptr;
3292 	} */ *uap;
3293 {
3294 
3295 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3296 }
3297 
3298 int
3299 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3300     enum uio_seg tptrseg)
3301 {
3302 	struct timespec ts[2];
3303 	struct file *fp;
3304 	int vfslocked;
3305 	int error;
3306 
3307 	AUDIT_ARG(fd, fd);
3308 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3309 		return (error);
3310 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3311 		return (error);
3312 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3313 #ifdef AUDIT
3314 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
3315 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3316 	VOP_UNLOCK(fp->f_vnode, 0);
3317 #endif
3318 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3319 	VFS_UNLOCK_GIANT(vfslocked);
3320 	fdrop(fp, td);
3321 	return (error);
3322 }
3323 
3324 /*
3325  * Truncate a file given its path name.
3326  */
3327 #ifndef _SYS_SYSPROTO_H_
3328 struct truncate_args {
3329 	char	*path;
3330 	int	pad;
3331 	off_t	length;
3332 };
3333 #endif
3334 int
3335 truncate(td, uap)
3336 	struct thread *td;
3337 	register struct truncate_args /* {
3338 		char *path;
3339 		int pad;
3340 		off_t length;
3341 	} */ *uap;
3342 {
3343 
3344 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3345 }
3346 
3347 int
3348 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3349 {
3350 	struct mount *mp;
3351 	struct vnode *vp;
3352 	struct vattr vattr;
3353 	int error;
3354 	struct nameidata nd;
3355 	int vfslocked;
3356 
3357 	if (length < 0)
3358 		return(EINVAL);
3359 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3360 	if ((error = namei(&nd)) != 0)
3361 		return (error);
3362 	vfslocked = NDHASGIANT(&nd);
3363 	vp = nd.ni_vp;
3364 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3365 		vrele(vp);
3366 		VFS_UNLOCK_GIANT(vfslocked);
3367 		return (error);
3368 	}
3369 	NDFREE(&nd, NDF_ONLY_PNBUF);
3370 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3371 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3372 	if (vp->v_type == VDIR)
3373 		error = EISDIR;
3374 #ifdef MAC
3375 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3376 	}
3377 #endif
3378 	else if ((error = vn_writechk(vp)) == 0 &&
3379 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3380 		VATTR_NULL(&vattr);
3381 		vattr.va_size = length;
3382 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3383 	}
3384 	vput(vp);
3385 	vn_finished_write(mp);
3386 	VFS_UNLOCK_GIANT(vfslocked);
3387 	return (error);
3388 }
3389 
3390 #if defined(COMPAT_43)
3391 /*
3392  * Truncate a file given its path name.
3393  */
3394 #ifndef _SYS_SYSPROTO_H_
3395 struct otruncate_args {
3396 	char	*path;
3397 	long	length;
3398 };
3399 #endif
3400 int
3401 otruncate(td, uap)
3402 	struct thread *td;
3403 	register struct otruncate_args /* {
3404 		char *path;
3405 		long length;
3406 	} */ *uap;
3407 {
3408 	struct truncate_args /* {
3409 		char *path;
3410 		int pad;
3411 		off_t length;
3412 	} */ nuap;
3413 
3414 	nuap.path = uap->path;
3415 	nuap.length = uap->length;
3416 	return (truncate(td, &nuap));
3417 }
3418 #endif /* COMPAT_43 */
3419 
3420 /* Versions with the pad argument */
3421 int
3422 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3423 {
3424 	struct truncate_args ouap;
3425 
3426 	ouap.path = uap->path;
3427 	ouap.length = uap->length;
3428 	return (truncate(td, &ouap));
3429 }
3430 
3431 int
3432 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3433 {
3434 	struct ftruncate_args ouap;
3435 
3436 	ouap.fd = uap->fd;
3437 	ouap.length = uap->length;
3438 	return (ftruncate(td, &ouap));
3439 }
3440 
3441 /*
3442  * Sync an open file.
3443  */
3444 #ifndef _SYS_SYSPROTO_H_
3445 struct fsync_args {
3446 	int	fd;
3447 };
3448 #endif
3449 int
3450 fsync(td, uap)
3451 	struct thread *td;
3452 	struct fsync_args /* {
3453 		int fd;
3454 	} */ *uap;
3455 {
3456 	struct vnode *vp;
3457 	struct mount *mp;
3458 	struct file *fp;
3459 	int vfslocked;
3460 	int error;
3461 
3462 	AUDIT_ARG(fd, uap->fd);
3463 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3464 		return (error);
3465 	vp = fp->f_vnode;
3466 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3467 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3468 		goto drop;
3469 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3470 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
3471 	if (vp->v_object != NULL) {
3472 		VM_OBJECT_LOCK(vp->v_object);
3473 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3474 		VM_OBJECT_UNLOCK(vp->v_object);
3475 	}
3476 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3477 
3478 	VOP_UNLOCK(vp, 0);
3479 	vn_finished_write(mp);
3480 drop:
3481 	VFS_UNLOCK_GIANT(vfslocked);
3482 	fdrop(fp, td);
3483 	return (error);
3484 }
3485 
3486 /*
3487  * Rename files.  Source and destination must either both be directories, or
3488  * both not be directories.  If target is a directory, it must be empty.
3489  */
3490 #ifndef _SYS_SYSPROTO_H_
3491 struct rename_args {
3492 	char	*from;
3493 	char	*to;
3494 };
3495 #endif
3496 int
3497 rename(td, uap)
3498 	struct thread *td;
3499 	register struct rename_args /* {
3500 		char *from;
3501 		char *to;
3502 	} */ *uap;
3503 {
3504 
3505 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3506 }
3507 
3508 #ifndef _SYS_SYSPROTO_H_
3509 struct renameat_args {
3510 	int	oldfd;
3511 	char	*old;
3512 	int	newfd;
3513 	char	*new;
3514 };
3515 #endif
3516 int
3517 renameat(struct thread *td, struct renameat_args *uap)
3518 {
3519 
3520 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3521 	    UIO_USERSPACE));
3522 }
3523 
3524 int
3525 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3526 {
3527 
3528 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3529 }
3530 
3531 int
3532 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3533     enum uio_seg pathseg)
3534 {
3535 	struct mount *mp = NULL;
3536 	struct vnode *tvp, *fvp, *tdvp;
3537 	struct nameidata fromnd, tond;
3538 	int tvfslocked;
3539 	int fvfslocked;
3540 	int error;
3541 
3542 	bwillwrite();
3543 #ifdef MAC
3544 	NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3545 	    AUDITVNODE1, pathseg, old, oldfd, td);
3546 #else
3547 	NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3548 	    AUDITVNODE1, pathseg, old, oldfd, td);
3549 #endif
3550 
3551 	if ((error = namei(&fromnd)) != 0)
3552 		return (error);
3553 	fvfslocked = NDHASGIANT(&fromnd);
3554 	tvfslocked = 0;
3555 #ifdef MAC
3556 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3557 	    fromnd.ni_vp, &fromnd.ni_cnd);
3558 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3559 	if (fromnd.ni_dvp != fromnd.ni_vp)
3560 		VOP_UNLOCK(fromnd.ni_vp, 0);
3561 #endif
3562 	fvp = fromnd.ni_vp;
3563 	if (error == 0)
3564 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3565 	if (error != 0) {
3566 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3567 		vrele(fromnd.ni_dvp);
3568 		vrele(fvp);
3569 		goto out1;
3570 	}
3571 	NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3572 	    MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3573 	if (fromnd.ni_vp->v_type == VDIR)
3574 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3575 	if ((error = namei(&tond)) != 0) {
3576 		/* Translate error code for rename("dir1", "dir2/."). */
3577 		if (error == EISDIR && fvp->v_type == VDIR)
3578 			error = EINVAL;
3579 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3580 		vrele(fromnd.ni_dvp);
3581 		vrele(fvp);
3582 		vn_finished_write(mp);
3583 		goto out1;
3584 	}
3585 	tvfslocked = NDHASGIANT(&tond);
3586 	tdvp = tond.ni_dvp;
3587 	tvp = tond.ni_vp;
3588 	if (tvp != NULL) {
3589 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3590 			error = ENOTDIR;
3591 			goto out;
3592 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3593 			error = EISDIR;
3594 			goto out;
3595 		}
3596 	}
3597 	if (fvp == tdvp) {
3598 		error = EINVAL;
3599 		goto out;
3600 	}
3601 	/*
3602 	 * If the source is the same as the destination (that is, if they
3603 	 * are links to the same vnode), then there is nothing to do.
3604 	 */
3605 	if (fvp == tvp)
3606 		error = -1;
3607 #ifdef MAC
3608 	else
3609 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3610 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3611 #endif
3612 out:
3613 	if (!error) {
3614 		VOP_LEASE(tdvp, td, td->td_ucred, LEASE_WRITE);
3615 		if (fromnd.ni_dvp != tdvp) {
3616 			VOP_LEASE(fromnd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3617 		}
3618 		if (tvp) {
3619 			VOP_LEASE(tvp, td, td->td_ucred, LEASE_WRITE);
3620 		}
3621 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3622 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3623 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3624 		NDFREE(&tond, NDF_ONLY_PNBUF);
3625 	} else {
3626 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3627 		NDFREE(&tond, NDF_ONLY_PNBUF);
3628 		if (tvp)
3629 			vput(tvp);
3630 		if (tdvp == tvp)
3631 			vrele(tdvp);
3632 		else
3633 			vput(tdvp);
3634 		vrele(fromnd.ni_dvp);
3635 		vrele(fvp);
3636 	}
3637 	vrele(tond.ni_startdir);
3638 	vn_finished_write(mp);
3639 out1:
3640 	if (fromnd.ni_startdir)
3641 		vrele(fromnd.ni_startdir);
3642 	VFS_UNLOCK_GIANT(fvfslocked);
3643 	VFS_UNLOCK_GIANT(tvfslocked);
3644 	if (error == -1)
3645 		return (0);
3646 	return (error);
3647 }
3648 
3649 /*
3650  * Make a directory file.
3651  */
3652 #ifndef _SYS_SYSPROTO_H_
3653 struct mkdir_args {
3654 	char	*path;
3655 	int	mode;
3656 };
3657 #endif
3658 int
3659 mkdir(td, uap)
3660 	struct thread *td;
3661 	register struct mkdir_args /* {
3662 		char *path;
3663 		int mode;
3664 	} */ *uap;
3665 {
3666 
3667 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3668 }
3669 
3670 #ifndef _SYS_SYSPROTO_H_
3671 struct mkdirat_args {
3672 	int	fd;
3673 	char	*path;
3674 	mode_t	mode;
3675 };
3676 #endif
3677 int
3678 mkdirat(struct thread *td, struct mkdirat_args *uap)
3679 {
3680 
3681 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3682 }
3683 
3684 int
3685 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3686 {
3687 
3688 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3689 }
3690 
3691 int
3692 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3693     int mode)
3694 {
3695 	struct mount *mp;
3696 	struct vnode *vp;
3697 	struct vattr vattr;
3698 	int error;
3699 	struct nameidata nd;
3700 	int vfslocked;
3701 
3702 	AUDIT_ARG(mode, mode);
3703 restart:
3704 	bwillwrite();
3705 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3706 	    segflg, path, fd, td);
3707 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3708 	if ((error = namei(&nd)) != 0)
3709 		return (error);
3710 	vfslocked = NDHASGIANT(&nd);
3711 	vp = nd.ni_vp;
3712 	if (vp != NULL) {
3713 		NDFREE(&nd, NDF_ONLY_PNBUF);
3714 		/*
3715 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3716 		 * the strange behaviour of leaving the vnode unlocked
3717 		 * if the target is the same vnode as the parent.
3718 		 */
3719 		if (vp == nd.ni_dvp)
3720 			vrele(nd.ni_dvp);
3721 		else
3722 			vput(nd.ni_dvp);
3723 		vrele(vp);
3724 		VFS_UNLOCK_GIANT(vfslocked);
3725 		return (EEXIST);
3726 	}
3727 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3728 		NDFREE(&nd, NDF_ONLY_PNBUF);
3729 		vput(nd.ni_dvp);
3730 		VFS_UNLOCK_GIANT(vfslocked);
3731 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3732 			return (error);
3733 		goto restart;
3734 	}
3735 	VATTR_NULL(&vattr);
3736 	vattr.va_type = VDIR;
3737 	FILEDESC_SLOCK(td->td_proc->p_fd);
3738 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3739 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
3740 #ifdef MAC
3741 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3742 	    &vattr);
3743 	if (error)
3744 		goto out;
3745 #endif
3746 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3747 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3748 #ifdef MAC
3749 out:
3750 #endif
3751 	NDFREE(&nd, NDF_ONLY_PNBUF);
3752 	vput(nd.ni_dvp);
3753 	if (!error)
3754 		vput(nd.ni_vp);
3755 	vn_finished_write(mp);
3756 	VFS_UNLOCK_GIANT(vfslocked);
3757 	return (error);
3758 }
3759 
3760 /*
3761  * Remove a directory file.
3762  */
3763 #ifndef _SYS_SYSPROTO_H_
3764 struct rmdir_args {
3765 	char	*path;
3766 };
3767 #endif
3768 int
3769 rmdir(td, uap)
3770 	struct thread *td;
3771 	struct rmdir_args /* {
3772 		char *path;
3773 	} */ *uap;
3774 {
3775 
3776 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3777 }
3778 
3779 int
3780 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3781 {
3782 
3783 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3784 }
3785 
3786 int
3787 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3788 {
3789 	struct mount *mp;
3790 	struct vnode *vp;
3791 	int error;
3792 	struct nameidata nd;
3793 	int vfslocked;
3794 
3795 restart:
3796 	bwillwrite();
3797 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3798 	    pathseg, path, fd, td);
3799 	if ((error = namei(&nd)) != 0)
3800 		return (error);
3801 	vfslocked = NDHASGIANT(&nd);
3802 	vp = nd.ni_vp;
3803 	if (vp->v_type != VDIR) {
3804 		error = ENOTDIR;
3805 		goto out;
3806 	}
3807 	/*
3808 	 * No rmdir "." please.
3809 	 */
3810 	if (nd.ni_dvp == vp) {
3811 		error = EINVAL;
3812 		goto out;
3813 	}
3814 	/*
3815 	 * The root of a mounted filesystem cannot be deleted.
3816 	 */
3817 	if (vp->v_vflag & VV_ROOT) {
3818 		error = EBUSY;
3819 		goto out;
3820 	}
3821 #ifdef MAC
3822 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3823 	    &nd.ni_cnd);
3824 	if (error)
3825 		goto out;
3826 #endif
3827 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3828 		NDFREE(&nd, NDF_ONLY_PNBUF);
3829 		vput(vp);
3830 		if (nd.ni_dvp == vp)
3831 			vrele(nd.ni_dvp);
3832 		else
3833 			vput(nd.ni_dvp);
3834 		VFS_UNLOCK_GIANT(vfslocked);
3835 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3836 			return (error);
3837 		goto restart;
3838 	}
3839 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3840 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3841 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3842 	vn_finished_write(mp);
3843 out:
3844 	NDFREE(&nd, NDF_ONLY_PNBUF);
3845 	vput(vp);
3846 	if (nd.ni_dvp == vp)
3847 		vrele(nd.ni_dvp);
3848 	else
3849 		vput(nd.ni_dvp);
3850 	VFS_UNLOCK_GIANT(vfslocked);
3851 	return (error);
3852 }
3853 
3854 #ifdef COMPAT_43
3855 /*
3856  * Read a block of directory entries in a filesystem independent format.
3857  */
3858 #ifndef _SYS_SYSPROTO_H_
3859 struct ogetdirentries_args {
3860 	int	fd;
3861 	char	*buf;
3862 	u_int	count;
3863 	long	*basep;
3864 };
3865 #endif
3866 int
3867 ogetdirentries(td, uap)
3868 	struct thread *td;
3869 	register struct ogetdirentries_args /* {
3870 		int fd;
3871 		char *buf;
3872 		u_int count;
3873 		long *basep;
3874 	} */ *uap;
3875 {
3876 	struct vnode *vp;
3877 	struct file *fp;
3878 	struct uio auio, kuio;
3879 	struct iovec aiov, kiov;
3880 	struct dirent *dp, *edp;
3881 	caddr_t dirbuf;
3882 	int error, eofflag, readcnt, vfslocked;
3883 	long loff;
3884 
3885 	/* XXX arbitrary sanity limit on `count'. */
3886 	if (uap->count > 64 * 1024)
3887 		return (EINVAL);
3888 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3889 		return (error);
3890 	if ((fp->f_flag & FREAD) == 0) {
3891 		fdrop(fp, td);
3892 		return (EBADF);
3893 	}
3894 	vp = fp->f_vnode;
3895 unionread:
3896 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3897 	if (vp->v_type != VDIR) {
3898 		VFS_UNLOCK_GIANT(vfslocked);
3899 		fdrop(fp, td);
3900 		return (EINVAL);
3901 	}
3902 	aiov.iov_base = uap->buf;
3903 	aiov.iov_len = uap->count;
3904 	auio.uio_iov = &aiov;
3905 	auio.uio_iovcnt = 1;
3906 	auio.uio_rw = UIO_READ;
3907 	auio.uio_segflg = UIO_USERSPACE;
3908 	auio.uio_td = td;
3909 	auio.uio_resid = uap->count;
3910 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3911 	loff = auio.uio_offset = fp->f_offset;
3912 #ifdef MAC
3913 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3914 	if (error) {
3915 		VOP_UNLOCK(vp, 0);
3916 		VFS_UNLOCK_GIANT(vfslocked);
3917 		fdrop(fp, td);
3918 		return (error);
3919 	}
3920 #endif
3921 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3922 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3923 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3924 			    NULL, NULL);
3925 			fp->f_offset = auio.uio_offset;
3926 		} else
3927 #	endif
3928 	{
3929 		kuio = auio;
3930 		kuio.uio_iov = &kiov;
3931 		kuio.uio_segflg = UIO_SYSSPACE;
3932 		kiov.iov_len = uap->count;
3933 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3934 		kiov.iov_base = dirbuf;
3935 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3936 			    NULL, NULL);
3937 		fp->f_offset = kuio.uio_offset;
3938 		if (error == 0) {
3939 			readcnt = uap->count - kuio.uio_resid;
3940 			edp = (struct dirent *)&dirbuf[readcnt];
3941 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3942 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3943 					/*
3944 					 * The expected low byte of
3945 					 * dp->d_namlen is our dp->d_type.
3946 					 * The high MBZ byte of dp->d_namlen
3947 					 * is our dp->d_namlen.
3948 					 */
3949 					dp->d_type = dp->d_namlen;
3950 					dp->d_namlen = 0;
3951 #				else
3952 					/*
3953 					 * The dp->d_type is the high byte
3954 					 * of the expected dp->d_namlen,
3955 					 * so must be zero'ed.
3956 					 */
3957 					dp->d_type = 0;
3958 #				endif
3959 				if (dp->d_reclen > 0) {
3960 					dp = (struct dirent *)
3961 					    ((char *)dp + dp->d_reclen);
3962 				} else {
3963 					error = EIO;
3964 					break;
3965 				}
3966 			}
3967 			if (dp >= edp)
3968 				error = uiomove(dirbuf, readcnt, &auio);
3969 		}
3970 		free(dirbuf, M_TEMP);
3971 	}
3972 	if (error) {
3973 		VOP_UNLOCK(vp, 0);
3974 		VFS_UNLOCK_GIANT(vfslocked);
3975 		fdrop(fp, td);
3976 		return (error);
3977 	}
3978 	if (uap->count == auio.uio_resid &&
3979 	    (vp->v_vflag & VV_ROOT) &&
3980 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3981 		struct vnode *tvp = vp;
3982 		vp = vp->v_mount->mnt_vnodecovered;
3983 		VREF(vp);
3984 		fp->f_vnode = vp;
3985 		fp->f_data = vp;
3986 		fp->f_offset = 0;
3987 		vput(tvp);
3988 		VFS_UNLOCK_GIANT(vfslocked);
3989 		goto unionread;
3990 	}
3991 	VOP_UNLOCK(vp, 0);
3992 	VFS_UNLOCK_GIANT(vfslocked);
3993 	error = copyout(&loff, uap->basep, sizeof(long));
3994 	fdrop(fp, td);
3995 	td->td_retval[0] = uap->count - auio.uio_resid;
3996 	return (error);
3997 }
3998 #endif /* COMPAT_43 */
3999 
4000 /*
4001  * Read a block of directory entries in a filesystem independent format.
4002  */
4003 #ifndef _SYS_SYSPROTO_H_
4004 struct getdirentries_args {
4005 	int	fd;
4006 	char	*buf;
4007 	u_int	count;
4008 	long	*basep;
4009 };
4010 #endif
4011 int
4012 getdirentries(td, uap)
4013 	struct thread *td;
4014 	register struct getdirentries_args /* {
4015 		int fd;
4016 		char *buf;
4017 		u_int count;
4018 		long *basep;
4019 	} */ *uap;
4020 {
4021 	long base;
4022 	int error;
4023 
4024 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4025 	if (error)
4026 		return (error);
4027 	if (uap->basep != NULL)
4028 		error = copyout(&base, uap->basep, sizeof(long));
4029 	return (error);
4030 }
4031 
4032 int
4033 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4034     long *basep)
4035 {
4036 	struct vnode *vp;
4037 	struct file *fp;
4038 	struct uio auio;
4039 	struct iovec aiov;
4040 	int vfslocked;
4041 	long loff;
4042 	int error, eofflag;
4043 
4044 	AUDIT_ARG(fd, fd);
4045 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
4046 		return (error);
4047 	if ((fp->f_flag & FREAD) == 0) {
4048 		fdrop(fp, td);
4049 		return (EBADF);
4050 	}
4051 	vp = fp->f_vnode;
4052 unionread:
4053 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4054 	if (vp->v_type != VDIR) {
4055 		VFS_UNLOCK_GIANT(vfslocked);
4056 		error = EINVAL;
4057 		goto fail;
4058 	}
4059 	aiov.iov_base = buf;
4060 	aiov.iov_len = count;
4061 	auio.uio_iov = &aiov;
4062 	auio.uio_iovcnt = 1;
4063 	auio.uio_rw = UIO_READ;
4064 	auio.uio_segflg = UIO_USERSPACE;
4065 	auio.uio_td = td;
4066 	auio.uio_resid = count;
4067 	/* vn_lock(vp, LK_SHARED | LK_RETRY); */
4068 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4069 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
4070 	loff = auio.uio_offset = fp->f_offset;
4071 #ifdef MAC
4072 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4073 	if (error == 0)
4074 #endif
4075 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4076 		    NULL);
4077 	fp->f_offset = auio.uio_offset;
4078 	if (error) {
4079 		VOP_UNLOCK(vp, 0);
4080 		VFS_UNLOCK_GIANT(vfslocked);
4081 		goto fail;
4082 	}
4083 	if (count == auio.uio_resid &&
4084 	    (vp->v_vflag & VV_ROOT) &&
4085 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4086 		struct vnode *tvp = vp;
4087 		vp = vp->v_mount->mnt_vnodecovered;
4088 		VREF(vp);
4089 		fp->f_vnode = vp;
4090 		fp->f_data = vp;
4091 		fp->f_offset = 0;
4092 		vput(tvp);
4093 		VFS_UNLOCK_GIANT(vfslocked);
4094 		goto unionread;
4095 	}
4096 	VOP_UNLOCK(vp, 0);
4097 	VFS_UNLOCK_GIANT(vfslocked);
4098 	*basep = loff;
4099 	td->td_retval[0] = count - auio.uio_resid;
4100 fail:
4101 	fdrop(fp, td);
4102 	return (error);
4103 }
4104 
4105 #ifndef _SYS_SYSPROTO_H_
4106 struct getdents_args {
4107 	int fd;
4108 	char *buf;
4109 	size_t count;
4110 };
4111 #endif
4112 int
4113 getdents(td, uap)
4114 	struct thread *td;
4115 	register struct getdents_args /* {
4116 		int fd;
4117 		char *buf;
4118 		u_int count;
4119 	} */ *uap;
4120 {
4121 	struct getdirentries_args ap;
4122 	ap.fd = uap->fd;
4123 	ap.buf = uap->buf;
4124 	ap.count = uap->count;
4125 	ap.basep = NULL;
4126 	return (getdirentries(td, &ap));
4127 }
4128 
4129 /*
4130  * Set the mode mask for creation of filesystem nodes.
4131  */
4132 #ifndef _SYS_SYSPROTO_H_
4133 struct umask_args {
4134 	int	newmask;
4135 };
4136 #endif
4137 int
4138 umask(td, uap)
4139 	struct thread *td;
4140 	struct umask_args /* {
4141 		int newmask;
4142 	} */ *uap;
4143 {
4144 	register struct filedesc *fdp;
4145 
4146 	FILEDESC_XLOCK(td->td_proc->p_fd);
4147 	fdp = td->td_proc->p_fd;
4148 	td->td_retval[0] = fdp->fd_cmask;
4149 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4150 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4151 	return (0);
4152 }
4153 
4154 /*
4155  * Void all references to file by ripping underlying filesystem away from
4156  * vnode.
4157  */
4158 #ifndef _SYS_SYSPROTO_H_
4159 struct revoke_args {
4160 	char	*path;
4161 };
4162 #endif
4163 int
4164 revoke(td, uap)
4165 	struct thread *td;
4166 	register struct revoke_args /* {
4167 		char *path;
4168 	} */ *uap;
4169 {
4170 	struct vnode *vp;
4171 	struct vattr vattr;
4172 	int error;
4173 	struct nameidata nd;
4174 	int vfslocked;
4175 
4176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4177 	    UIO_USERSPACE, uap->path, td);
4178 	if ((error = namei(&nd)) != 0)
4179 		return (error);
4180 	vfslocked = NDHASGIANT(&nd);
4181 	vp = nd.ni_vp;
4182 	NDFREE(&nd, NDF_ONLY_PNBUF);
4183 	if (vp->v_type != VCHR) {
4184 		error = EINVAL;
4185 		goto out;
4186 	}
4187 #ifdef MAC
4188 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4189 	if (error)
4190 		goto out;
4191 #endif
4192 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4193 	if (error)
4194 		goto out;
4195 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4196 		error = priv_check(td, PRIV_VFS_ADMIN);
4197 		if (error)
4198 			goto out;
4199 	}
4200 	if (vcount(vp) > 1)
4201 		VOP_REVOKE(vp, REVOKEALL);
4202 out:
4203 	vput(vp);
4204 	VFS_UNLOCK_GIANT(vfslocked);
4205 	return (error);
4206 }
4207 
4208 /*
4209  * Convert a user file descriptor to a kernel file entry.
4210  * A reference on the file entry is held upon returning.
4211  */
4212 int
4213 getvnode(fdp, fd, fpp)
4214 	struct filedesc *fdp;
4215 	int fd;
4216 	struct file **fpp;
4217 {
4218 	int error;
4219 	struct file *fp;
4220 
4221 	fp = NULL;
4222 	if (fdp == NULL)
4223 		error = EBADF;
4224 	else {
4225 		FILEDESC_SLOCK(fdp);
4226 		if ((u_int)fd >= fdp->fd_nfiles ||
4227 		    (fp = fdp->fd_ofiles[fd]) == NULL)
4228 			error = EBADF;
4229 		else if (fp->f_vnode == NULL) {
4230 			fp = NULL;
4231 			error = EINVAL;
4232 		} else {
4233 			fhold(fp);
4234 			error = 0;
4235 		}
4236 		FILEDESC_SUNLOCK(fdp);
4237 	}
4238 	*fpp = fp;
4239 	return (error);
4240 }
4241 
4242 /*
4243  * Get an (NFS) file handle.
4244  */
4245 #ifndef _SYS_SYSPROTO_H_
4246 struct lgetfh_args {
4247 	char	*fname;
4248 	fhandle_t *fhp;
4249 };
4250 #endif
4251 int
4252 lgetfh(td, uap)
4253 	struct thread *td;
4254 	register struct lgetfh_args *uap;
4255 {
4256 	struct nameidata nd;
4257 	fhandle_t fh;
4258 	register struct vnode *vp;
4259 	int vfslocked;
4260 	int error;
4261 
4262 	error = priv_check(td, PRIV_VFS_GETFH);
4263 	if (error)
4264 		return (error);
4265 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4266 	    UIO_USERSPACE, uap->fname, td);
4267 	error = namei(&nd);
4268 	if (error)
4269 		return (error);
4270 	vfslocked = NDHASGIANT(&nd);
4271 	NDFREE(&nd, NDF_ONLY_PNBUF);
4272 	vp = nd.ni_vp;
4273 	bzero(&fh, sizeof(fh));
4274 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4275 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4276 	vput(vp);
4277 	VFS_UNLOCK_GIANT(vfslocked);
4278 	if (error)
4279 		return (error);
4280 	error = copyout(&fh, uap->fhp, sizeof (fh));
4281 	return (error);
4282 }
4283 
4284 #ifndef _SYS_SYSPROTO_H_
4285 struct getfh_args {
4286 	char	*fname;
4287 	fhandle_t *fhp;
4288 };
4289 #endif
4290 int
4291 getfh(td, uap)
4292 	struct thread *td;
4293 	register struct getfh_args *uap;
4294 {
4295 	struct nameidata nd;
4296 	fhandle_t fh;
4297 	register struct vnode *vp;
4298 	int vfslocked;
4299 	int error;
4300 
4301 	error = priv_check(td, PRIV_VFS_GETFH);
4302 	if (error)
4303 		return (error);
4304 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4305 	    UIO_USERSPACE, uap->fname, td);
4306 	error = namei(&nd);
4307 	if (error)
4308 		return (error);
4309 	vfslocked = NDHASGIANT(&nd);
4310 	NDFREE(&nd, NDF_ONLY_PNBUF);
4311 	vp = nd.ni_vp;
4312 	bzero(&fh, sizeof(fh));
4313 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4314 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4315 	vput(vp);
4316 	VFS_UNLOCK_GIANT(vfslocked);
4317 	if (error)
4318 		return (error);
4319 	error = copyout(&fh, uap->fhp, sizeof (fh));
4320 	return (error);
4321 }
4322 
4323 /*
4324  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4325  * open descriptor.
4326  *
4327  * warning: do not remove the priv_check() call or this becomes one giant
4328  * security hole.
4329  */
4330 #ifndef _SYS_SYSPROTO_H_
4331 struct fhopen_args {
4332 	const struct fhandle *u_fhp;
4333 	int flags;
4334 };
4335 #endif
4336 int
4337 fhopen(td, uap)
4338 	struct thread *td;
4339 	struct fhopen_args /* {
4340 		const struct fhandle *u_fhp;
4341 		int flags;
4342 	} */ *uap;
4343 {
4344 	struct proc *p = td->td_proc;
4345 	struct mount *mp;
4346 	struct vnode *vp;
4347 	struct fhandle fhp;
4348 	struct vattr vat;
4349 	struct vattr *vap = &vat;
4350 	struct flock lf;
4351 	struct file *fp;
4352 	register struct filedesc *fdp = p->p_fd;
4353 	int fmode, error, type;
4354 	accmode_t accmode;
4355 	struct file *nfp;
4356 	int vfslocked;
4357 	int indx;
4358 
4359 	error = priv_check(td, PRIV_VFS_FHOPEN);
4360 	if (error)
4361 		return (error);
4362 	fmode = FFLAGS(uap->flags);
4363 	/* why not allow a non-read/write open for our lockd? */
4364 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4365 		return (EINVAL);
4366 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4367 	if (error)
4368 		return(error);
4369 	/* find the mount point */
4370 	mp = vfs_getvfs(&fhp.fh_fsid);
4371 	if (mp == NULL)
4372 		return (ESTALE);
4373 	vfslocked = VFS_LOCK_GIANT(mp);
4374 	/* now give me my vnode, it gets returned to me locked */
4375 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4376 	if (error)
4377 		goto out;
4378 	/*
4379 	 * from now on we have to make sure not
4380 	 * to forget about the vnode
4381 	 * any error that causes an abort must vput(vp)
4382 	 * just set error = err and 'goto bad;'.
4383 	 */
4384 
4385 	/*
4386 	 * from vn_open
4387 	 */
4388 	if (vp->v_type == VLNK) {
4389 		error = EMLINK;
4390 		goto bad;
4391 	}
4392 	if (vp->v_type == VSOCK) {
4393 		error = EOPNOTSUPP;
4394 		goto bad;
4395 	}
4396 	accmode = 0;
4397 	if (fmode & (FWRITE | O_TRUNC)) {
4398 		if (vp->v_type == VDIR) {
4399 			error = EISDIR;
4400 			goto bad;
4401 		}
4402 		error = vn_writechk(vp);
4403 		if (error)
4404 			goto bad;
4405 		accmode |= VWRITE;
4406 	}
4407 	if (fmode & FREAD)
4408 		accmode |= VREAD;
4409 	if (fmode & O_APPEND)
4410 		accmode |= VAPPEND;
4411 #ifdef MAC
4412 	error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4413 	if (error)
4414 		goto bad;
4415 #endif
4416 	if (accmode) {
4417 		error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4418 		if (error)
4419 			goto bad;
4420 	}
4421 	if (fmode & O_TRUNC) {
4422 		VOP_UNLOCK(vp, 0);				/* XXX */
4423 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4424 			vrele(vp);
4425 			goto out;
4426 		}
4427 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
4428 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4429 #ifdef MAC
4430 		/*
4431 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4432 		 * should be right.
4433 		 */
4434 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4435 		if (error == 0) {
4436 #endif
4437 			VATTR_NULL(vap);
4438 			vap->va_size = 0;
4439 			error = VOP_SETATTR(vp, vap, td->td_ucred);
4440 #ifdef MAC
4441 		}
4442 #endif
4443 		vn_finished_write(mp);
4444 		if (error)
4445 			goto bad;
4446 	}
4447 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4448 	if (error)
4449 		goto bad;
4450 
4451 	if (fmode & FWRITE)
4452 		vp->v_writecount++;
4453 
4454 	/*
4455 	 * end of vn_open code
4456 	 */
4457 
4458 	if ((error = falloc(td, &nfp, &indx)) != 0) {
4459 		if (fmode & FWRITE)
4460 			vp->v_writecount--;
4461 		goto bad;
4462 	}
4463 	/* An extra reference on `nfp' has been held for us by falloc(). */
4464 	fp = nfp;
4465 	nfp->f_vnode = vp;
4466 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4467 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4468 		lf.l_whence = SEEK_SET;
4469 		lf.l_start = 0;
4470 		lf.l_len = 0;
4471 		if (fmode & O_EXLOCK)
4472 			lf.l_type = F_WRLCK;
4473 		else
4474 			lf.l_type = F_RDLCK;
4475 		type = F_FLOCK;
4476 		if ((fmode & FNONBLOCK) == 0)
4477 			type |= F_WAIT;
4478 		VOP_UNLOCK(vp, 0);
4479 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4480 			    type)) != 0) {
4481 			/*
4482 			 * The lock request failed.  Normally close the
4483 			 * descriptor but handle the case where someone might
4484 			 * have dup()d or close()d it when we weren't looking.
4485 			 */
4486 			fdclose(fdp, fp, indx, td);
4487 
4488 			/*
4489 			 * release our private reference
4490 			 */
4491 			fdrop(fp, td);
4492 			goto out;
4493 		}
4494 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4495 		atomic_set_int(&fp->f_flag, FHASLOCK);
4496 	}
4497 
4498 	VOP_UNLOCK(vp, 0);
4499 	fdrop(fp, td);
4500 	vfs_rel(mp);
4501 	VFS_UNLOCK_GIANT(vfslocked);
4502 	td->td_retval[0] = indx;
4503 	return (0);
4504 
4505 bad:
4506 	vput(vp);
4507 out:
4508 	vfs_rel(mp);
4509 	VFS_UNLOCK_GIANT(vfslocked);
4510 	return (error);
4511 }
4512 
4513 /*
4514  * Stat an (NFS) file handle.
4515  */
4516 #ifndef _SYS_SYSPROTO_H_
4517 struct fhstat_args {
4518 	struct fhandle *u_fhp;
4519 	struct stat *sb;
4520 };
4521 #endif
4522 int
4523 fhstat(td, uap)
4524 	struct thread *td;
4525 	register struct fhstat_args /* {
4526 		struct fhandle *u_fhp;
4527 		struct stat *sb;
4528 	} */ *uap;
4529 {
4530 	struct stat sb;
4531 	fhandle_t fh;
4532 	struct mount *mp;
4533 	struct vnode *vp;
4534 	int vfslocked;
4535 	int error;
4536 
4537 	error = priv_check(td, PRIV_VFS_FHSTAT);
4538 	if (error)
4539 		return (error);
4540 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4541 	if (error)
4542 		return (error);
4543 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4544 		return (ESTALE);
4545 	vfslocked = VFS_LOCK_GIANT(mp);
4546 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) {
4547 		vfs_rel(mp);
4548 		VFS_UNLOCK_GIANT(vfslocked);
4549 		return (error);
4550 	}
4551 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4552 	vput(vp);
4553 	vfs_rel(mp);
4554 	VFS_UNLOCK_GIANT(vfslocked);
4555 	if (error)
4556 		return (error);
4557 	error = copyout(&sb, uap->sb, sizeof(sb));
4558 	return (error);
4559 }
4560 
4561 /*
4562  * Implement fstatfs() for (NFS) file handles.
4563  */
4564 #ifndef _SYS_SYSPROTO_H_
4565 struct fhstatfs_args {
4566 	struct fhandle *u_fhp;
4567 	struct statfs *buf;
4568 };
4569 #endif
4570 int
4571 fhstatfs(td, uap)
4572 	struct thread *td;
4573 	struct fhstatfs_args /* {
4574 		struct fhandle *u_fhp;
4575 		struct statfs *buf;
4576 	} */ *uap;
4577 {
4578 	struct statfs sf;
4579 	fhandle_t fh;
4580 	int error;
4581 
4582 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4583 	if (error)
4584 		return (error);
4585 	error = kern_fhstatfs(td, fh, &sf);
4586 	if (error)
4587 		return (error);
4588 	return (copyout(&sf, uap->buf, sizeof(sf)));
4589 }
4590 
4591 int
4592 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4593 {
4594 	struct statfs *sp;
4595 	struct mount *mp;
4596 	struct vnode *vp;
4597 	int vfslocked;
4598 	int error;
4599 
4600 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4601 	if (error)
4602 		return (error);
4603 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4604 		return (ESTALE);
4605 	vfslocked = VFS_LOCK_GIANT(mp);
4606 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4607 	if (error) {
4608 		VFS_UNLOCK_GIANT(vfslocked);
4609 		vfs_rel(mp);
4610 		return (error);
4611 	}
4612 	vput(vp);
4613 	error = prison_canseemount(td->td_ucred, mp);
4614 	if (error)
4615 		goto out;
4616 #ifdef MAC
4617 	error = mac_mount_check_stat(td->td_ucred, mp);
4618 	if (error)
4619 		goto out;
4620 #endif
4621 	/*
4622 	 * Set these in case the underlying filesystem fails to do so.
4623 	 */
4624 	sp = &mp->mnt_stat;
4625 	sp->f_version = STATFS_VERSION;
4626 	sp->f_namemax = NAME_MAX;
4627 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4628 	error = VFS_STATFS(mp, sp, td);
4629 	if (error == 0)
4630 		*buf = *sp;
4631 out:
4632 	vfs_rel(mp);
4633 	VFS_UNLOCK_GIANT(vfslocked);
4634 	return (error);
4635 }
4636