xref: /freebsd/sys/kern/vfs_syscalls.c (revision b3aaa0cc21c63d388230c7ef2a80abd631ff20d5)
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_kdtrace.h"
42 #include "opt_ktrace.h"
43 #include "opt_mac.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/sysent.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/sysproto.h>
54 #include <sys/namei.h>
55 #include <sys/filedesc.h>
56 #include <sys/kernel.h>
57 #include <sys/fcntl.h>
58 #include <sys/file.h>
59 #include <sys/filio.h>
60 #include <sys/limits.h>
61 #include <sys/linker.h>
62 #include <sys/sdt.h>
63 #include <sys/stat.h>
64 #include <sys/sx.h>
65 #include <sys/unistd.h>
66 #include <sys/vnode.h>
67 #include <sys/priv.h>
68 #include <sys/proc.h>
69 #include <sys/dirent.h>
70 #include <sys/jail.h>
71 #include <sys/syscallsubr.h>
72 #include <sys/sysctl.h>
73 #ifdef KTRACE
74 #include <sys/ktrace.h>
75 #endif
76 
77 #include <machine/stdarg.h>
78 
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
81 
82 #include <vm/vm.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/uma.h>
86 
87 SDT_PROVIDER_DEFINE(vfs);
88 SDT_PROBE_DEFINE(vfs, , stat, mode);
89 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
90 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
91 SDT_PROBE_DEFINE(vfs, , stat, reg);
92 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
93 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
94 
95 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
96 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
97 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
98 static int setfmode(struct thread *td, struct vnode *, int);
99 static int setfflags(struct thread *td, struct vnode *, int);
100 static int setutimes(struct thread *td, struct vnode *,
101     const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
103     struct thread *td);
104 
105 /*
106  * The module initialization routine for POSIX asynchronous I/O will
107  * set this to the version of AIO that it implements.  (Zero means
108  * that it is not implemented.)  This value is used here by pathconf()
109  * and in kern_descrip.c by fpathconf().
110  */
111 int async_io_version;
112 
113 #ifdef DEBUG
114 static int syncprt = 0;
115 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
116 #endif
117 
118 /*
119  * Sync each mounted filesystem.
120  */
121 #ifndef _SYS_SYSPROTO_H_
122 struct sync_args {
123 	int     dummy;
124 };
125 #endif
126 /* ARGSUSED */
127 int
128 sync(td, uap)
129 	struct thread *td;
130 	struct sync_args *uap;
131 {
132 	struct mount *mp, *nmp;
133 	int vfslocked;
134 
135 	mtx_lock(&mountlist_mtx);
136 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
137 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
138 			nmp = TAILQ_NEXT(mp, mnt_list);
139 			continue;
140 		}
141 		vfslocked = VFS_LOCK_GIANT(mp);
142 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
143 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
144 			MNT_ILOCK(mp);
145 			mp->mnt_noasync++;
146 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
147 			MNT_IUNLOCK(mp);
148 			vfs_msync(mp, MNT_NOWAIT);
149 			VFS_SYNC(mp, MNT_NOWAIT, td);
150 			MNT_ILOCK(mp);
151 			mp->mnt_noasync--;
152 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
153 			    mp->mnt_noasync == 0)
154 				mp->mnt_kern_flag |= MNTK_ASYNC;
155 			MNT_IUNLOCK(mp);
156 			vn_finished_write(mp);
157 		}
158 		VFS_UNLOCK_GIANT(vfslocked);
159 		mtx_lock(&mountlist_mtx);
160 		nmp = TAILQ_NEXT(mp, mnt_list);
161 		vfs_unbusy(mp);
162 	}
163 	mtx_unlock(&mountlist_mtx);
164 	return (0);
165 }
166 
167 /* XXX PRISON: could be per prison flag */
168 static int prison_quotas;
169 #if 0
170 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
171 #endif
172 
173 /*
174  * Change filesystem quotas.
175  */
176 #ifndef _SYS_SYSPROTO_H_
177 struct quotactl_args {
178 	char *path;
179 	int cmd;
180 	int uid;
181 	caddr_t arg;
182 };
183 #endif
184 int
185 quotactl(td, uap)
186 	struct thread *td;
187 	register struct quotactl_args /* {
188 		char *path;
189 		int cmd;
190 		int uid;
191 		caddr_t arg;
192 	} */ *uap;
193 {
194 	struct mount *mp;
195 	int vfslocked;
196 	int error;
197 	struct nameidata nd;
198 
199 	AUDIT_ARG(cmd, uap->cmd);
200 	AUDIT_ARG(uid, uap->uid);
201 	if (jailed(td->td_ucred) && !prison_quotas)
202 		return (EPERM);
203 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
204 	   UIO_USERSPACE, uap->path, td);
205 	if ((error = namei(&nd)) != 0)
206 		return (error);
207 	vfslocked = NDHASGIANT(&nd);
208 	NDFREE(&nd, NDF_ONLY_PNBUF);
209 	mp = nd.ni_vp->v_mount;
210 	vfs_ref(mp);
211 	vput(nd.ni_vp);
212 	error = vfs_busy(mp, 0);
213 	vfs_rel(mp);
214 	if (error) {
215 		VFS_UNLOCK_GIANT(vfslocked);
216 		return (error);
217 	}
218 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, td);
219 	vfs_unbusy(mp);
220 	VFS_UNLOCK_GIANT(vfslocked);
221 	return (error);
222 }
223 
224 /*
225  * Used by statfs conversion routines to scale the block size up if
226  * necessary so that all of the block counts are <= 'max_size'.  Note
227  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
228  * value of 'n'.
229  */
230 void
231 statfs_scale_blocks(struct statfs *sf, long max_size)
232 {
233 	uint64_t count;
234 	int shift;
235 
236 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
237 
238 	/*
239 	 * Attempt to scale the block counts to give a more accurate
240 	 * overview to userland of the ratio of free space to used
241 	 * space.  To do this, find the largest block count and compute
242 	 * a divisor that lets it fit into a signed integer <= max_size.
243 	 */
244 	if (sf->f_bavail < 0)
245 		count = -sf->f_bavail;
246 	else
247 		count = sf->f_bavail;
248 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
249 	if (count <= max_size)
250 		return;
251 
252 	count >>= flsl(max_size);
253 	shift = 0;
254 	while (count > 0) {
255 		shift++;
256 		count >>=1;
257 	}
258 
259 	sf->f_bsize <<= shift;
260 	sf->f_blocks >>= shift;
261 	sf->f_bfree >>= shift;
262 	sf->f_bavail >>= shift;
263 }
264 
265 /*
266  * Get filesystem statistics.
267  */
268 #ifndef _SYS_SYSPROTO_H_
269 struct statfs_args {
270 	char *path;
271 	struct statfs *buf;
272 };
273 #endif
274 int
275 statfs(td, uap)
276 	struct thread *td;
277 	register struct statfs_args /* {
278 		char *path;
279 		struct statfs *buf;
280 	} */ *uap;
281 {
282 	struct statfs sf;
283 	int error;
284 
285 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
286 	if (error == 0)
287 		error = copyout(&sf, uap->buf, sizeof(sf));
288 	return (error);
289 }
290 
291 int
292 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
293     struct statfs *buf)
294 {
295 	struct mount *mp;
296 	struct statfs *sp, sb;
297 	int vfslocked;
298 	int error;
299 	struct nameidata nd;
300 
301 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
302 	    AUDITVNODE1, pathseg, path, td);
303 	error = namei(&nd);
304 	if (error)
305 		return (error);
306 	vfslocked = NDHASGIANT(&nd);
307 	mp = nd.ni_vp->v_mount;
308 	vfs_ref(mp);
309 	NDFREE(&nd, NDF_ONLY_PNBUF);
310 	vput(nd.ni_vp);
311 	error = vfs_busy(mp, 0);
312 	vfs_rel(mp);
313 	if (error) {
314 		VFS_UNLOCK_GIANT(vfslocked);
315 		return (error);
316 	}
317 #ifdef MAC
318 	error = mac_mount_check_stat(td->td_ucred, mp);
319 	if (error)
320 		goto out;
321 #endif
322 	/*
323 	 * Set these in case the underlying filesystem fails to do so.
324 	 */
325 	sp = &mp->mnt_stat;
326 	sp->f_version = STATFS_VERSION;
327 	sp->f_namemax = NAME_MAX;
328 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
329 	error = VFS_STATFS(mp, sp, td);
330 	if (error)
331 		goto out;
332 	if (priv_check(td, PRIV_VFS_GENERATION)) {
333 		bcopy(sp, &sb, sizeof(sb));
334 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
335 		prison_enforce_statfs(td->td_ucred, mp, &sb);
336 		sp = &sb;
337 	}
338 	*buf = *sp;
339 out:
340 	vfs_unbusy(mp);
341 	VFS_UNLOCK_GIANT(vfslocked);
342 	return (error);
343 }
344 
345 /*
346  * Get filesystem statistics.
347  */
348 #ifndef _SYS_SYSPROTO_H_
349 struct fstatfs_args {
350 	int fd;
351 	struct statfs *buf;
352 };
353 #endif
354 int
355 fstatfs(td, uap)
356 	struct thread *td;
357 	register struct fstatfs_args /* {
358 		int fd;
359 		struct statfs *buf;
360 	} */ *uap;
361 {
362 	struct statfs sf;
363 	int error;
364 
365 	error = kern_fstatfs(td, uap->fd, &sf);
366 	if (error == 0)
367 		error = copyout(&sf, uap->buf, sizeof(sf));
368 	return (error);
369 }
370 
371 int
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
373 {
374 	struct file *fp;
375 	struct mount *mp;
376 	struct statfs *sp, sb;
377 	int vfslocked;
378 	struct vnode *vp;
379 	int error;
380 
381 	AUDIT_ARG(fd, fd);
382 	error = getvnode(td->td_proc->p_fd, fd, &fp);
383 	if (error)
384 		return (error);
385 	vp = fp->f_vnode;
386 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
387 	vn_lock(vp, LK_SHARED | LK_RETRY);
388 #ifdef AUDIT
389 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
390 #endif
391 	mp = vp->v_mount;
392 	if (mp)
393 		vfs_ref(mp);
394 	VOP_UNLOCK(vp, 0);
395 	fdrop(fp, td);
396 	if (mp == NULL) {
397 		error = EBADF;
398 		goto out;
399 	}
400 	error = vfs_busy(mp, 0);
401 	vfs_rel(mp);
402 	if (error) {
403 		VFS_UNLOCK_GIANT(vfslocked);
404 		return (error);
405 	}
406 #ifdef MAC
407 	error = mac_mount_check_stat(td->td_ucred, mp);
408 	if (error)
409 		goto out;
410 #endif
411 	/*
412 	 * Set these in case the underlying filesystem fails to do so.
413 	 */
414 	sp = &mp->mnt_stat;
415 	sp->f_version = STATFS_VERSION;
416 	sp->f_namemax = NAME_MAX;
417 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
418 	error = VFS_STATFS(mp, sp, td);
419 	if (error)
420 		goto out;
421 	if (priv_check(td, PRIV_VFS_GENERATION)) {
422 		bcopy(sp, &sb, sizeof(sb));
423 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
424 		prison_enforce_statfs(td->td_ucred, mp, &sb);
425 		sp = &sb;
426 	}
427 	*buf = *sp;
428 out:
429 	if (mp)
430 		vfs_unbusy(mp);
431 	VFS_UNLOCK_GIANT(vfslocked);
432 	return (error);
433 }
434 
435 /*
436  * Get statistics on all filesystems.
437  */
438 #ifndef _SYS_SYSPROTO_H_
439 struct getfsstat_args {
440 	struct statfs *buf;
441 	long bufsize;
442 	int flags;
443 };
444 #endif
445 int
446 getfsstat(td, uap)
447 	struct thread *td;
448 	register struct getfsstat_args /* {
449 		struct statfs *buf;
450 		long bufsize;
451 		int flags;
452 	} */ *uap;
453 {
454 
455 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
456 	    uap->flags));
457 }
458 
459 /*
460  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
461  * 	The caller is responsible for freeing memory which will be allocated
462  *	in '*buf'.
463  */
464 int
465 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
466     enum uio_seg bufseg, int flags)
467 {
468 	struct mount *mp, *nmp;
469 	struct statfs *sfsp, *sp, sb;
470 	size_t count, maxcount;
471 	int vfslocked;
472 	int error;
473 
474 	maxcount = bufsize / sizeof(struct statfs);
475 	if (bufsize == 0)
476 		sfsp = NULL;
477 	else if (bufseg == UIO_USERSPACE)
478 		sfsp = *buf;
479 	else /* if (bufseg == UIO_SYSSPACE) */ {
480 		count = 0;
481 		mtx_lock(&mountlist_mtx);
482 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
483 			count++;
484 		}
485 		mtx_unlock(&mountlist_mtx);
486 		if (maxcount > count)
487 			maxcount = count;
488 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
489 		    M_WAITOK);
490 	}
491 	count = 0;
492 	mtx_lock(&mountlist_mtx);
493 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
494 		if (prison_canseemount(td->td_ucred, mp) != 0) {
495 			nmp = TAILQ_NEXT(mp, mnt_list);
496 			continue;
497 		}
498 #ifdef MAC
499 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
500 			nmp = TAILQ_NEXT(mp, mnt_list);
501 			continue;
502 		}
503 #endif
504 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
505 			nmp = TAILQ_NEXT(mp, mnt_list);
506 			continue;
507 		}
508 		vfslocked = VFS_LOCK_GIANT(mp);
509 		if (sfsp && count < maxcount) {
510 			sp = &mp->mnt_stat;
511 			/*
512 			 * Set these in case the underlying filesystem
513 			 * fails to do so.
514 			 */
515 			sp->f_version = STATFS_VERSION;
516 			sp->f_namemax = NAME_MAX;
517 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
518 			/*
519 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
520 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
521 			 * overrides MNT_WAIT.
522 			 */
523 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
524 			    (flags & MNT_WAIT)) &&
525 			    (error = VFS_STATFS(mp, sp, td))) {
526 				VFS_UNLOCK_GIANT(vfslocked);
527 				mtx_lock(&mountlist_mtx);
528 				nmp = TAILQ_NEXT(mp, mnt_list);
529 				vfs_unbusy(mp);
530 				continue;
531 			}
532 			if (priv_check(td, PRIV_VFS_GENERATION)) {
533 				bcopy(sp, &sb, sizeof(sb));
534 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
535 				prison_enforce_statfs(td->td_ucred, mp, &sb);
536 				sp = &sb;
537 			}
538 			if (bufseg == UIO_SYSSPACE)
539 				bcopy(sp, sfsp, sizeof(*sp));
540 			else /* if (bufseg == UIO_USERSPACE) */ {
541 				error = copyout(sp, sfsp, sizeof(*sp));
542 				if (error) {
543 					vfs_unbusy(mp);
544 					VFS_UNLOCK_GIANT(vfslocked);
545 					return (error);
546 				}
547 			}
548 			sfsp++;
549 		}
550 		VFS_UNLOCK_GIANT(vfslocked);
551 		count++;
552 		mtx_lock(&mountlist_mtx);
553 		nmp = TAILQ_NEXT(mp, mnt_list);
554 		vfs_unbusy(mp);
555 	}
556 	mtx_unlock(&mountlist_mtx);
557 	if (sfsp && count > maxcount)
558 		td->td_retval[0] = maxcount;
559 	else
560 		td->td_retval[0] = count;
561 	return (0);
562 }
563 
564 #ifdef COMPAT_FREEBSD4
565 /*
566  * Get old format filesystem statistics.
567  */
568 static void cvtstatfs(struct statfs *, struct ostatfs *);
569 
570 #ifndef _SYS_SYSPROTO_H_
571 struct freebsd4_statfs_args {
572 	char *path;
573 	struct ostatfs *buf;
574 };
575 #endif
576 int
577 freebsd4_statfs(td, uap)
578 	struct thread *td;
579 	struct freebsd4_statfs_args /* {
580 		char *path;
581 		struct ostatfs *buf;
582 	} */ *uap;
583 {
584 	struct ostatfs osb;
585 	struct statfs sf;
586 	int error;
587 
588 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
589 	if (error)
590 		return (error);
591 	cvtstatfs(&sf, &osb);
592 	return (copyout(&osb, uap->buf, sizeof(osb)));
593 }
594 
595 /*
596  * Get filesystem statistics.
597  */
598 #ifndef _SYS_SYSPROTO_H_
599 struct freebsd4_fstatfs_args {
600 	int fd;
601 	struct ostatfs *buf;
602 };
603 #endif
604 int
605 freebsd4_fstatfs(td, uap)
606 	struct thread *td;
607 	struct freebsd4_fstatfs_args /* {
608 		int fd;
609 		struct ostatfs *buf;
610 	} */ *uap;
611 {
612 	struct ostatfs osb;
613 	struct statfs sf;
614 	int error;
615 
616 	error = kern_fstatfs(td, uap->fd, &sf);
617 	if (error)
618 		return (error);
619 	cvtstatfs(&sf, &osb);
620 	return (copyout(&osb, uap->buf, sizeof(osb)));
621 }
622 
623 /*
624  * Get statistics on all filesystems.
625  */
626 #ifndef _SYS_SYSPROTO_H_
627 struct freebsd4_getfsstat_args {
628 	struct ostatfs *buf;
629 	long bufsize;
630 	int flags;
631 };
632 #endif
633 int
634 freebsd4_getfsstat(td, uap)
635 	struct thread *td;
636 	register struct freebsd4_getfsstat_args /* {
637 		struct ostatfs *buf;
638 		long bufsize;
639 		int flags;
640 	} */ *uap;
641 {
642 	struct statfs *buf, *sp;
643 	struct ostatfs osb;
644 	size_t count, size;
645 	int error;
646 
647 	count = uap->bufsize / sizeof(struct ostatfs);
648 	size = count * sizeof(struct statfs);
649 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
650 	if (size > 0) {
651 		count = td->td_retval[0];
652 		sp = buf;
653 		while (count > 0 && error == 0) {
654 			cvtstatfs(sp, &osb);
655 			error = copyout(&osb, uap->buf, sizeof(osb));
656 			sp++;
657 			uap->buf++;
658 			count--;
659 		}
660 		free(buf, M_TEMP);
661 	}
662 	return (error);
663 }
664 
665 /*
666  * Implement fstatfs() for (NFS) file handles.
667  */
668 #ifndef _SYS_SYSPROTO_H_
669 struct freebsd4_fhstatfs_args {
670 	struct fhandle *u_fhp;
671 	struct ostatfs *buf;
672 };
673 #endif
674 int
675 freebsd4_fhstatfs(td, uap)
676 	struct thread *td;
677 	struct freebsd4_fhstatfs_args /* {
678 		struct fhandle *u_fhp;
679 		struct ostatfs *buf;
680 	} */ *uap;
681 {
682 	struct ostatfs osb;
683 	struct statfs sf;
684 	fhandle_t fh;
685 	int error;
686 
687 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
688 	if (error)
689 		return (error);
690 	error = kern_fhstatfs(td, fh, &sf);
691 	if (error)
692 		return (error);
693 	cvtstatfs(&sf, &osb);
694 	return (copyout(&osb, uap->buf, sizeof(osb)));
695 }
696 
697 /*
698  * Convert a new format statfs structure to an old format statfs structure.
699  */
700 static void
701 cvtstatfs(nsp, osp)
702 	struct statfs *nsp;
703 	struct ostatfs *osp;
704 {
705 
706 	statfs_scale_blocks(nsp, LONG_MAX);
707 	bzero(osp, sizeof(*osp));
708 	osp->f_bsize = nsp->f_bsize;
709 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
710 	osp->f_blocks = nsp->f_blocks;
711 	osp->f_bfree = nsp->f_bfree;
712 	osp->f_bavail = nsp->f_bavail;
713 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
714 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
715 	osp->f_owner = nsp->f_owner;
716 	osp->f_type = nsp->f_type;
717 	osp->f_flags = nsp->f_flags;
718 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
719 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
720 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
721 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
722 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
723 	    MIN(MFSNAMELEN, OMFSNAMELEN));
724 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
725 	    MIN(MNAMELEN, OMNAMELEN));
726 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
727 	    MIN(MNAMELEN, OMNAMELEN));
728 	osp->f_fsid = nsp->f_fsid;
729 }
730 #endif /* COMPAT_FREEBSD4 */
731 
732 /*
733  * Change current working directory to a given file descriptor.
734  */
735 #ifndef _SYS_SYSPROTO_H_
736 struct fchdir_args {
737 	int	fd;
738 };
739 #endif
740 int
741 fchdir(td, uap)
742 	struct thread *td;
743 	struct fchdir_args /* {
744 		int fd;
745 	} */ *uap;
746 {
747 	register struct filedesc *fdp = td->td_proc->p_fd;
748 	struct vnode *vp, *tdp, *vpold;
749 	struct mount *mp;
750 	struct file *fp;
751 	int vfslocked;
752 	int error;
753 
754 	AUDIT_ARG(fd, uap->fd);
755 	if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
756 		return (error);
757 	vp = fp->f_vnode;
758 	VREF(vp);
759 	fdrop(fp, td);
760 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
761 	vn_lock(vp, LK_SHARED | LK_RETRY);
762 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
763 	error = change_dir(vp, td);
764 	while (!error && (mp = vp->v_mountedhere) != NULL) {
765 		int tvfslocked;
766 		if (vfs_busy(mp, 0))
767 			continue;
768 		tvfslocked = VFS_LOCK_GIANT(mp);
769 		error = VFS_ROOT(mp, LK_SHARED, &tdp, td);
770 		vfs_unbusy(mp);
771 		if (error) {
772 			VFS_UNLOCK_GIANT(tvfslocked);
773 			break;
774 		}
775 		vput(vp);
776 		VFS_UNLOCK_GIANT(vfslocked);
777 		vp = tdp;
778 		vfslocked = tvfslocked;
779 	}
780 	if (error) {
781 		vput(vp);
782 		VFS_UNLOCK_GIANT(vfslocked);
783 		return (error);
784 	}
785 	VOP_UNLOCK(vp, 0);
786 	VFS_UNLOCK_GIANT(vfslocked);
787 	FILEDESC_XLOCK(fdp);
788 	vpold = fdp->fd_cdir;
789 	fdp->fd_cdir = vp;
790 	FILEDESC_XUNLOCK(fdp);
791 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
792 	vrele(vpold);
793 	VFS_UNLOCK_GIANT(vfslocked);
794 	return (0);
795 }
796 
797 /*
798  * Change current working directory (``.'').
799  */
800 #ifndef _SYS_SYSPROTO_H_
801 struct chdir_args {
802 	char	*path;
803 };
804 #endif
805 int
806 chdir(td, uap)
807 	struct thread *td;
808 	struct chdir_args /* {
809 		char *path;
810 	} */ *uap;
811 {
812 
813 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
814 }
815 
816 int
817 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
818 {
819 	register struct filedesc *fdp = td->td_proc->p_fd;
820 	int error;
821 	struct nameidata nd;
822 	struct vnode *vp;
823 	int vfslocked;
824 
825 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
826 	    MPSAFE, pathseg, path, td);
827 	if ((error = namei(&nd)) != 0)
828 		return (error);
829 	vfslocked = NDHASGIANT(&nd);
830 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
831 		vput(nd.ni_vp);
832 		VFS_UNLOCK_GIANT(vfslocked);
833 		NDFREE(&nd, NDF_ONLY_PNBUF);
834 		return (error);
835 	}
836 	VOP_UNLOCK(nd.ni_vp, 0);
837 	VFS_UNLOCK_GIANT(vfslocked);
838 	NDFREE(&nd, NDF_ONLY_PNBUF);
839 	FILEDESC_XLOCK(fdp);
840 	vp = fdp->fd_cdir;
841 	fdp->fd_cdir = nd.ni_vp;
842 	FILEDESC_XUNLOCK(fdp);
843 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
844 	vrele(vp);
845 	VFS_UNLOCK_GIANT(vfslocked);
846 	return (0);
847 }
848 
849 /*
850  * Helper function for raised chroot(2) security function:  Refuse if
851  * any filedescriptors are open directories.
852  */
853 static int
854 chroot_refuse_vdir_fds(fdp)
855 	struct filedesc *fdp;
856 {
857 	struct vnode *vp;
858 	struct file *fp;
859 	int fd;
860 
861 	FILEDESC_LOCK_ASSERT(fdp);
862 
863 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
864 		fp = fget_locked(fdp, fd);
865 		if (fp == NULL)
866 			continue;
867 		if (fp->f_type == DTYPE_VNODE) {
868 			vp = fp->f_vnode;
869 			if (vp->v_type == VDIR)
870 				return (EPERM);
871 		}
872 	}
873 	return (0);
874 }
875 
876 /*
877  * This sysctl determines if we will allow a process to chroot(2) if it
878  * has a directory open:
879  *	0: disallowed for all processes.
880  *	1: allowed for processes that were not already chroot(2)'ed.
881  *	2: allowed for all processes.
882  */
883 
884 static int chroot_allow_open_directories = 1;
885 
886 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
887      &chroot_allow_open_directories, 0, "");
888 
889 /*
890  * Change notion of root (``/'') directory.
891  */
892 #ifndef _SYS_SYSPROTO_H_
893 struct chroot_args {
894 	char	*path;
895 };
896 #endif
897 int
898 chroot(td, uap)
899 	struct thread *td;
900 	struct chroot_args /* {
901 		char *path;
902 	} */ *uap;
903 {
904 	int error;
905 	struct nameidata nd;
906 	int vfslocked;
907 
908 	error = priv_check(td, PRIV_VFS_CHROOT);
909 	if (error)
910 		return (error);
911 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
912 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
913 	error = namei(&nd);
914 	if (error)
915 		goto error;
916 	vfslocked = NDHASGIANT(&nd);
917 	if ((error = change_dir(nd.ni_vp, td)) != 0)
918 		goto e_vunlock;
919 #ifdef MAC
920 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
921 		goto e_vunlock;
922 #endif
923 	VOP_UNLOCK(nd.ni_vp, 0);
924 	error = change_root(nd.ni_vp, td);
925 	vrele(nd.ni_vp);
926 	VFS_UNLOCK_GIANT(vfslocked);
927 	NDFREE(&nd, NDF_ONLY_PNBUF);
928 	return (error);
929 e_vunlock:
930 	vput(nd.ni_vp);
931 	VFS_UNLOCK_GIANT(vfslocked);
932 error:
933 	NDFREE(&nd, NDF_ONLY_PNBUF);
934 	return (error);
935 }
936 
937 /*
938  * Common routine for chroot and chdir.  Callers must provide a locked vnode
939  * instance.
940  */
941 int
942 change_dir(vp, td)
943 	struct vnode *vp;
944 	struct thread *td;
945 {
946 	int error;
947 
948 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
949 	if (vp->v_type != VDIR)
950 		return (ENOTDIR);
951 #ifdef MAC
952 	error = mac_vnode_check_chdir(td->td_ucred, vp);
953 	if (error)
954 		return (error);
955 #endif
956 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
957 	return (error);
958 }
959 
960 /*
961  * Common routine for kern_chroot() and jail_attach().  The caller is
962  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
963  * authorize this operation.
964  */
965 int
966 change_root(vp, td)
967 	struct vnode *vp;
968 	struct thread *td;
969 {
970 	struct filedesc *fdp;
971 	struct vnode *oldvp;
972 	int vfslocked;
973 	int error;
974 
975 	VFS_ASSERT_GIANT(vp->v_mount);
976 	fdp = td->td_proc->p_fd;
977 	FILEDESC_XLOCK(fdp);
978 	if (chroot_allow_open_directories == 0 ||
979 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
980 		error = chroot_refuse_vdir_fds(fdp);
981 		if (error) {
982 			FILEDESC_XUNLOCK(fdp);
983 			return (error);
984 		}
985 	}
986 	oldvp = fdp->fd_rdir;
987 	fdp->fd_rdir = vp;
988 	VREF(fdp->fd_rdir);
989 	if (!fdp->fd_jdir) {
990 		fdp->fd_jdir = vp;
991 		VREF(fdp->fd_jdir);
992 	}
993 	FILEDESC_XUNLOCK(fdp);
994 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
995 	vrele(oldvp);
996 	VFS_UNLOCK_GIANT(vfslocked);
997 	return (0);
998 }
999 
1000 /*
1001  * Check permissions, allocate an open file structure, and call the device
1002  * open routine if any.
1003  */
1004 #ifndef _SYS_SYSPROTO_H_
1005 struct open_args {
1006 	char	*path;
1007 	int	flags;
1008 	int	mode;
1009 };
1010 #endif
1011 int
1012 open(td, uap)
1013 	struct thread *td;
1014 	register struct open_args /* {
1015 		char *path;
1016 		int flags;
1017 		int mode;
1018 	} */ *uap;
1019 {
1020 
1021 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1022 }
1023 
1024 #ifndef _SYS_SYSPROTO_H_
1025 struct openat_args {
1026 	int	fd;
1027 	char	*path;
1028 	int	flag;
1029 	int	mode;
1030 };
1031 #endif
1032 int
1033 openat(struct thread *td, struct openat_args *uap)
1034 {
1035 
1036 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1037 	    uap->mode));
1038 }
1039 
1040 int
1041 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1042     int mode)
1043 {
1044 
1045 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1046 }
1047 
1048 int
1049 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1050     int flags, int mode)
1051 {
1052 	struct proc *p = td->td_proc;
1053 	struct filedesc *fdp = p->p_fd;
1054 	struct file *fp;
1055 	struct vnode *vp;
1056 	struct vattr vat;
1057 	struct mount *mp;
1058 	int cmode;
1059 	struct file *nfp;
1060 	int type, indx, error;
1061 	struct flock lf;
1062 	struct nameidata nd;
1063 	int vfslocked;
1064 
1065 	AUDIT_ARG(fflags, flags);
1066 	AUDIT_ARG(mode, mode);
1067 	/* XXX: audit dirfd */
1068 	/*
1069 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1070 	 * be specified.
1071 	 */
1072 	if (flags & O_EXEC) {
1073 		if (flags & O_ACCMODE)
1074 			return (EINVAL);
1075 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1076 		return (EINVAL);
1077 	else
1078 		flags = FFLAGS(flags);
1079 
1080 	error = falloc(td, &nfp, &indx);
1081 	if (error)
1082 		return (error);
1083 	/* An extra reference on `nfp' has been held for us by falloc(). */
1084 	fp = nfp;
1085 	/* Set the flags early so the finit in devfs can pick them up. */
1086 	fp->f_flag = flags & FMASK;
1087 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1088 	NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1089 	    td);
1090 	td->td_dupfd = -1;		/* XXX check for fdopen */
1091 	error = vn_open(&nd, &flags, cmode, fp);
1092 	if (error) {
1093 		/*
1094 		 * If the vn_open replaced the method vector, something
1095 		 * wonderous happened deep below and we just pass it up
1096 		 * pretending we know what we do.
1097 		 */
1098 		if (error == ENXIO && fp->f_ops != &badfileops) {
1099 			fdrop(fp, td);
1100 			td->td_retval[0] = indx;
1101 			return (0);
1102 		}
1103 
1104 		/*
1105 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1106 		 * responsible for dropping the old contents of ofiles[indx]
1107 		 * if it succeeds.
1108 		 */
1109 		if ((error == ENODEV || error == ENXIO) &&
1110 		    td->td_dupfd >= 0 &&		/* XXX from fdopen */
1111 		    (error =
1112 			dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1113 			td->td_retval[0] = indx;
1114 			fdrop(fp, td);
1115 			return (0);
1116 		}
1117 		/*
1118 		 * Clean up the descriptor, but only if another thread hadn't
1119 		 * replaced or closed it.
1120 		 */
1121 		fdclose(fdp, fp, indx, td);
1122 		fdrop(fp, td);
1123 
1124 		if (error == ERESTART)
1125 			error = EINTR;
1126 		return (error);
1127 	}
1128 	td->td_dupfd = 0;
1129 	vfslocked = NDHASGIANT(&nd);
1130 	NDFREE(&nd, NDF_ONLY_PNBUF);
1131 	vp = nd.ni_vp;
1132 
1133 	fp->f_vnode = vp;	/* XXX Does devfs need this? */
1134 	/*
1135 	 * If the file wasn't claimed by devfs bind it to the normal
1136 	 * vnode operations here.
1137 	 */
1138 	if (fp->f_ops == &badfileops) {
1139 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1140 		fp->f_seqcount = 1;
1141 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1142 	}
1143 
1144 	VOP_UNLOCK(vp, 0);
1145 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1146 		lf.l_whence = SEEK_SET;
1147 		lf.l_start = 0;
1148 		lf.l_len = 0;
1149 		if (flags & O_EXLOCK)
1150 			lf.l_type = F_WRLCK;
1151 		else
1152 			lf.l_type = F_RDLCK;
1153 		type = F_FLOCK;
1154 		if ((flags & FNONBLOCK) == 0)
1155 			type |= F_WAIT;
1156 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1157 			    type)) != 0)
1158 			goto bad;
1159 		atomic_set_int(&fp->f_flag, FHASLOCK);
1160 	}
1161 	if (flags & O_TRUNC) {
1162 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1163 			goto bad;
1164 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1165 		VATTR_NULL(&vat);
1166 		vat.va_size = 0;
1167 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1168 #ifdef MAC
1169 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1170 		if (error == 0)
1171 #endif
1172 			error = VOP_SETATTR(vp, &vat, td->td_ucred);
1173 		VOP_UNLOCK(vp, 0);
1174 		vn_finished_write(mp);
1175 		if (error)
1176 			goto bad;
1177 	}
1178 	VFS_UNLOCK_GIANT(vfslocked);
1179 	/*
1180 	 * Release our private reference, leaving the one associated with
1181 	 * the descriptor table intact.
1182 	 */
1183 	fdrop(fp, td);
1184 	td->td_retval[0] = indx;
1185 	return (0);
1186 bad:
1187 	VFS_UNLOCK_GIANT(vfslocked);
1188 	fdclose(fdp, fp, indx, td);
1189 	fdrop(fp, td);
1190 	return (error);
1191 }
1192 
1193 #ifdef COMPAT_43
1194 /*
1195  * Create a file.
1196  */
1197 #ifndef _SYS_SYSPROTO_H_
1198 struct ocreat_args {
1199 	char	*path;
1200 	int	mode;
1201 };
1202 #endif
1203 int
1204 ocreat(td, uap)
1205 	struct thread *td;
1206 	register struct ocreat_args /* {
1207 		char *path;
1208 		int mode;
1209 	} */ *uap;
1210 {
1211 
1212 	return (kern_open(td, uap->path, UIO_USERSPACE,
1213 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1214 }
1215 #endif /* COMPAT_43 */
1216 
1217 /*
1218  * Create a special file.
1219  */
1220 #ifndef _SYS_SYSPROTO_H_
1221 struct mknod_args {
1222 	char	*path;
1223 	int	mode;
1224 	int	dev;
1225 };
1226 #endif
1227 int
1228 mknod(td, uap)
1229 	struct thread *td;
1230 	register struct mknod_args /* {
1231 		char *path;
1232 		int mode;
1233 		int dev;
1234 	} */ *uap;
1235 {
1236 
1237 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1238 }
1239 
1240 #ifndef _SYS_SYSPROTO_H_
1241 struct mknodat_args {
1242 	int	fd;
1243 	char	*path;
1244 	mode_t	mode;
1245 	dev_t	dev;
1246 };
1247 #endif
1248 int
1249 mknodat(struct thread *td, struct mknodat_args *uap)
1250 {
1251 
1252 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1253 	    uap->dev));
1254 }
1255 
1256 int
1257 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1258     int dev)
1259 {
1260 
1261 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1262 }
1263 
1264 int
1265 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1266     int mode, int dev)
1267 {
1268 	struct vnode *vp;
1269 	struct mount *mp;
1270 	struct vattr vattr;
1271 	int error;
1272 	int whiteout = 0;
1273 	struct nameidata nd;
1274 	int vfslocked;
1275 
1276 	AUDIT_ARG(mode, mode);
1277 	AUDIT_ARG(dev, dev);
1278 	switch (mode & S_IFMT) {
1279 	case S_IFCHR:
1280 	case S_IFBLK:
1281 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1282 		break;
1283 	case S_IFMT:
1284 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1285 		break;
1286 	case S_IFWHT:
1287 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1288 		break;
1289 	case S_IFIFO:
1290 		if (dev == 0)
1291 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1292 		/* FALLTHROUGH */
1293 	default:
1294 		error = EINVAL;
1295 		break;
1296 	}
1297 	if (error)
1298 		return (error);
1299 restart:
1300 	bwillwrite();
1301 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1302 	    pathseg, path, fd, td);
1303 	if ((error = namei(&nd)) != 0)
1304 		return (error);
1305 	vfslocked = NDHASGIANT(&nd);
1306 	vp = nd.ni_vp;
1307 	if (vp != NULL) {
1308 		NDFREE(&nd, NDF_ONLY_PNBUF);
1309 		if (vp == nd.ni_dvp)
1310 			vrele(nd.ni_dvp);
1311 		else
1312 			vput(nd.ni_dvp);
1313 		vrele(vp);
1314 		VFS_UNLOCK_GIANT(vfslocked);
1315 		return (EEXIST);
1316 	} else {
1317 		VATTR_NULL(&vattr);
1318 		FILEDESC_SLOCK(td->td_proc->p_fd);
1319 		vattr.va_mode = (mode & ALLPERMS) &
1320 		    ~td->td_proc->p_fd->fd_cmask;
1321 		FILEDESC_SUNLOCK(td->td_proc->p_fd);
1322 		vattr.va_rdev = dev;
1323 		whiteout = 0;
1324 
1325 		switch (mode & S_IFMT) {
1326 		case S_IFMT:	/* used by badsect to flag bad sectors */
1327 			vattr.va_type = VBAD;
1328 			break;
1329 		case S_IFCHR:
1330 			vattr.va_type = VCHR;
1331 			break;
1332 		case S_IFBLK:
1333 			vattr.va_type = VBLK;
1334 			break;
1335 		case S_IFWHT:
1336 			whiteout = 1;
1337 			break;
1338 		default:
1339 			panic("kern_mknod: invalid mode");
1340 		}
1341 	}
1342 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1343 		NDFREE(&nd, NDF_ONLY_PNBUF);
1344 		vput(nd.ni_dvp);
1345 		VFS_UNLOCK_GIANT(vfslocked);
1346 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1347 			return (error);
1348 		goto restart;
1349 	}
1350 #ifdef MAC
1351 	if (error == 0 && !whiteout)
1352 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1353 		    &nd.ni_cnd, &vattr);
1354 #endif
1355 	if (!error) {
1356 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1357 		if (whiteout)
1358 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1359 		else {
1360 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1361 						&nd.ni_cnd, &vattr);
1362 			if (error == 0)
1363 				vput(nd.ni_vp);
1364 		}
1365 	}
1366 	NDFREE(&nd, NDF_ONLY_PNBUF);
1367 	vput(nd.ni_dvp);
1368 	vn_finished_write(mp);
1369 	VFS_UNLOCK_GIANT(vfslocked);
1370 	return (error);
1371 }
1372 
1373 /*
1374  * Create a named pipe.
1375  */
1376 #ifndef _SYS_SYSPROTO_H_
1377 struct mkfifo_args {
1378 	char	*path;
1379 	int	mode;
1380 };
1381 #endif
1382 int
1383 mkfifo(td, uap)
1384 	struct thread *td;
1385 	register struct mkfifo_args /* {
1386 		char *path;
1387 		int mode;
1388 	} */ *uap;
1389 {
1390 
1391 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1392 }
1393 
1394 #ifndef _SYS_SYSPROTO_H_
1395 struct mkfifoat_args {
1396 	int	fd;
1397 	char	*path;
1398 	mode_t	mode;
1399 };
1400 #endif
1401 int
1402 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1403 {
1404 
1405 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1406 	    uap->mode));
1407 }
1408 
1409 int
1410 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1411 {
1412 
1413 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1414 }
1415 
1416 int
1417 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1418     int mode)
1419 {
1420 	struct mount *mp;
1421 	struct vattr vattr;
1422 	int error;
1423 	struct nameidata nd;
1424 	int vfslocked;
1425 
1426 	AUDIT_ARG(mode, mode);
1427 restart:
1428 	bwillwrite();
1429 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1430 	    pathseg, path, fd, td);
1431 	if ((error = namei(&nd)) != 0)
1432 		return (error);
1433 	vfslocked = NDHASGIANT(&nd);
1434 	if (nd.ni_vp != NULL) {
1435 		NDFREE(&nd, NDF_ONLY_PNBUF);
1436 		if (nd.ni_vp == nd.ni_dvp)
1437 			vrele(nd.ni_dvp);
1438 		else
1439 			vput(nd.ni_dvp);
1440 		vrele(nd.ni_vp);
1441 		VFS_UNLOCK_GIANT(vfslocked);
1442 		return (EEXIST);
1443 	}
1444 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1445 		NDFREE(&nd, NDF_ONLY_PNBUF);
1446 		vput(nd.ni_dvp);
1447 		VFS_UNLOCK_GIANT(vfslocked);
1448 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1449 			return (error);
1450 		goto restart;
1451 	}
1452 	VATTR_NULL(&vattr);
1453 	vattr.va_type = VFIFO;
1454 	FILEDESC_SLOCK(td->td_proc->p_fd);
1455 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1456 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1457 #ifdef MAC
1458 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1459 	    &vattr);
1460 	if (error)
1461 		goto out;
1462 #endif
1463 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1464 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1465 	if (error == 0)
1466 		vput(nd.ni_vp);
1467 #ifdef MAC
1468 out:
1469 #endif
1470 	vput(nd.ni_dvp);
1471 	vn_finished_write(mp);
1472 	VFS_UNLOCK_GIANT(vfslocked);
1473 	NDFREE(&nd, NDF_ONLY_PNBUF);
1474 	return (error);
1475 }
1476 
1477 /*
1478  * Make a hard file link.
1479  */
1480 #ifndef _SYS_SYSPROTO_H_
1481 struct link_args {
1482 	char	*path;
1483 	char	*link;
1484 };
1485 #endif
1486 int
1487 link(td, uap)
1488 	struct thread *td;
1489 	register struct link_args /* {
1490 		char *path;
1491 		char *link;
1492 	} */ *uap;
1493 {
1494 
1495 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1496 }
1497 
1498 #ifndef _SYS_SYSPROTO_H_
1499 struct linkat_args {
1500 	int	fd1;
1501 	char	*path1;
1502 	int	fd2;
1503 	char	*path2;
1504 	int	flag;
1505 };
1506 #endif
1507 int
1508 linkat(struct thread *td, struct linkat_args *uap)
1509 {
1510 	int flag;
1511 
1512 	flag = uap->flag;
1513 	if (flag & ~AT_SYMLINK_FOLLOW)
1514 		return (EINVAL);
1515 
1516 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1517 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1518 }
1519 
1520 static int hardlink_check_uid = 0;
1521 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1522     &hardlink_check_uid, 0,
1523     "Unprivileged processes cannot create hard links to files owned by other "
1524     "users");
1525 static int hardlink_check_gid = 0;
1526 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1527     &hardlink_check_gid, 0,
1528     "Unprivileged processes cannot create hard links to files owned by other "
1529     "groups");
1530 
1531 static int
1532 can_hardlink(struct vnode *vp, struct ucred *cred)
1533 {
1534 	struct vattr va;
1535 	int error;
1536 
1537 	if (!hardlink_check_uid && !hardlink_check_gid)
1538 		return (0);
1539 
1540 	error = VOP_GETATTR(vp, &va, cred);
1541 	if (error != 0)
1542 		return (error);
1543 
1544 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1545 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1546 		if (error)
1547 			return (error);
1548 	}
1549 
1550 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1551 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1552 		if (error)
1553 			return (error);
1554 	}
1555 
1556 	return (0);
1557 }
1558 
1559 int
1560 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1561 {
1562 
1563 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1564 }
1565 
1566 int
1567 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1568     enum uio_seg segflg, int follow)
1569 {
1570 	struct vnode *vp;
1571 	struct mount *mp;
1572 	struct nameidata nd;
1573 	int vfslocked;
1574 	int lvfslocked;
1575 	int error;
1576 
1577 	bwillwrite();
1578 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1579 	    fd1, td);
1580 
1581 	if ((error = namei(&nd)) != 0)
1582 		return (error);
1583 	vfslocked = NDHASGIANT(&nd);
1584 	NDFREE(&nd, NDF_ONLY_PNBUF);
1585 	vp = nd.ni_vp;
1586 	if (vp->v_type == VDIR) {
1587 		vrele(vp);
1588 		VFS_UNLOCK_GIANT(vfslocked);
1589 		return (EPERM);		/* POSIX */
1590 	}
1591 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1592 		vrele(vp);
1593 		VFS_UNLOCK_GIANT(vfslocked);
1594 		return (error);
1595 	}
1596 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1597 	    segflg, path2, fd2, td);
1598 	if ((error = namei(&nd)) == 0) {
1599 		lvfslocked = NDHASGIANT(&nd);
1600 		if (nd.ni_vp != NULL) {
1601 			if (nd.ni_dvp == nd.ni_vp)
1602 				vrele(nd.ni_dvp);
1603 			else
1604 				vput(nd.ni_dvp);
1605 			vrele(nd.ni_vp);
1606 			error = EEXIST;
1607 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1608 		    == 0) {
1609 			VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1610 			VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1611 			error = can_hardlink(vp, td->td_ucred);
1612 			if (error == 0)
1613 #ifdef MAC
1614 				error = mac_vnode_check_link(td->td_ucred,
1615 				    nd.ni_dvp, vp, &nd.ni_cnd);
1616 			if (error == 0)
1617 #endif
1618 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1619 			VOP_UNLOCK(vp, 0);
1620 			vput(nd.ni_dvp);
1621 		}
1622 		NDFREE(&nd, NDF_ONLY_PNBUF);
1623 		VFS_UNLOCK_GIANT(lvfslocked);
1624 	}
1625 	vrele(vp);
1626 	vn_finished_write(mp);
1627 	VFS_UNLOCK_GIANT(vfslocked);
1628 	return (error);
1629 }
1630 
1631 /*
1632  * Make a symbolic link.
1633  */
1634 #ifndef _SYS_SYSPROTO_H_
1635 struct symlink_args {
1636 	char	*path;
1637 	char	*link;
1638 };
1639 #endif
1640 int
1641 symlink(td, uap)
1642 	struct thread *td;
1643 	register struct symlink_args /* {
1644 		char *path;
1645 		char *link;
1646 	} */ *uap;
1647 {
1648 
1649 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1650 }
1651 
1652 #ifndef _SYS_SYSPROTO_H_
1653 struct symlinkat_args {
1654 	char	*path;
1655 	int	fd;
1656 	char	*path2;
1657 };
1658 #endif
1659 int
1660 symlinkat(struct thread *td, struct symlinkat_args *uap)
1661 {
1662 
1663 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1664 	    UIO_USERSPACE));
1665 }
1666 
1667 int
1668 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1669 {
1670 
1671 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1672 }
1673 
1674 int
1675 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1676     enum uio_seg segflg)
1677 {
1678 	struct mount *mp;
1679 	struct vattr vattr;
1680 	char *syspath;
1681 	int error;
1682 	struct nameidata nd;
1683 	int vfslocked;
1684 
1685 	if (segflg == UIO_SYSSPACE) {
1686 		syspath = path1;
1687 	} else {
1688 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1689 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1690 			goto out;
1691 	}
1692 	AUDIT_ARG(text, syspath);
1693 restart:
1694 	bwillwrite();
1695 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1696 	    segflg, path2, fd, td);
1697 	if ((error = namei(&nd)) != 0)
1698 		goto out;
1699 	vfslocked = NDHASGIANT(&nd);
1700 	if (nd.ni_vp) {
1701 		NDFREE(&nd, NDF_ONLY_PNBUF);
1702 		if (nd.ni_vp == nd.ni_dvp)
1703 			vrele(nd.ni_dvp);
1704 		else
1705 			vput(nd.ni_dvp);
1706 		vrele(nd.ni_vp);
1707 		VFS_UNLOCK_GIANT(vfslocked);
1708 		error = EEXIST;
1709 		goto out;
1710 	}
1711 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1712 		NDFREE(&nd, NDF_ONLY_PNBUF);
1713 		vput(nd.ni_dvp);
1714 		VFS_UNLOCK_GIANT(vfslocked);
1715 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1716 			goto out;
1717 		goto restart;
1718 	}
1719 	VATTR_NULL(&vattr);
1720 	FILEDESC_SLOCK(td->td_proc->p_fd);
1721 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1722 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1723 #ifdef MAC
1724 	vattr.va_type = VLNK;
1725 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1726 	    &vattr);
1727 	if (error)
1728 		goto out2;
1729 #endif
1730 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1731 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1732 	if (error == 0)
1733 		vput(nd.ni_vp);
1734 #ifdef MAC
1735 out2:
1736 #endif
1737 	NDFREE(&nd, NDF_ONLY_PNBUF);
1738 	vput(nd.ni_dvp);
1739 	vn_finished_write(mp);
1740 	VFS_UNLOCK_GIANT(vfslocked);
1741 out:
1742 	if (segflg != UIO_SYSSPACE)
1743 		uma_zfree(namei_zone, syspath);
1744 	return (error);
1745 }
1746 
1747 /*
1748  * Delete a whiteout from the filesystem.
1749  */
1750 int
1751 undelete(td, uap)
1752 	struct thread *td;
1753 	register struct undelete_args /* {
1754 		char *path;
1755 	} */ *uap;
1756 {
1757 	int error;
1758 	struct mount *mp;
1759 	struct nameidata nd;
1760 	int vfslocked;
1761 
1762 restart:
1763 	bwillwrite();
1764 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1765 	    UIO_USERSPACE, uap->path, td);
1766 	error = namei(&nd);
1767 	if (error)
1768 		return (error);
1769 	vfslocked = NDHASGIANT(&nd);
1770 
1771 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1772 		NDFREE(&nd, NDF_ONLY_PNBUF);
1773 		if (nd.ni_vp == nd.ni_dvp)
1774 			vrele(nd.ni_dvp);
1775 		else
1776 			vput(nd.ni_dvp);
1777 		if (nd.ni_vp)
1778 			vrele(nd.ni_vp);
1779 		VFS_UNLOCK_GIANT(vfslocked);
1780 		return (EEXIST);
1781 	}
1782 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1783 		NDFREE(&nd, NDF_ONLY_PNBUF);
1784 		vput(nd.ni_dvp);
1785 		VFS_UNLOCK_GIANT(vfslocked);
1786 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1787 			return (error);
1788 		goto restart;
1789 	}
1790 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1791 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1792 	NDFREE(&nd, NDF_ONLY_PNBUF);
1793 	vput(nd.ni_dvp);
1794 	vn_finished_write(mp);
1795 	VFS_UNLOCK_GIANT(vfslocked);
1796 	return (error);
1797 }
1798 
1799 /*
1800  * Delete a name from the filesystem.
1801  */
1802 #ifndef _SYS_SYSPROTO_H_
1803 struct unlink_args {
1804 	char	*path;
1805 };
1806 #endif
1807 int
1808 unlink(td, uap)
1809 	struct thread *td;
1810 	struct unlink_args /* {
1811 		char *path;
1812 	} */ *uap;
1813 {
1814 
1815 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1816 }
1817 
1818 #ifndef _SYS_SYSPROTO_H_
1819 struct unlinkat_args {
1820 	int	fd;
1821 	char	*path;
1822 	int	flag;
1823 };
1824 #endif
1825 int
1826 unlinkat(struct thread *td, struct unlinkat_args *uap)
1827 {
1828 	int flag = uap->flag;
1829 	int fd = uap->fd;
1830 	char *path = uap->path;
1831 
1832 	if (flag & ~AT_REMOVEDIR)
1833 		return (EINVAL);
1834 
1835 	if (flag & AT_REMOVEDIR)
1836 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1837 	else
1838 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1839 }
1840 
1841 int
1842 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1843 {
1844 
1845 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1846 }
1847 
1848 int
1849 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1850 {
1851 	struct mount *mp;
1852 	struct vnode *vp;
1853 	int error;
1854 	struct nameidata nd;
1855 	int vfslocked;
1856 
1857 restart:
1858 	bwillwrite();
1859 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1860 	    pathseg, path, fd, td);
1861 	if ((error = namei(&nd)) != 0)
1862 		return (error == EINVAL ? EPERM : error);
1863 	vfslocked = NDHASGIANT(&nd);
1864 	vp = nd.ni_vp;
1865 	if (vp->v_type == VDIR)
1866 		error = EPERM;		/* POSIX */
1867 	else {
1868 		/*
1869 		 * The root of a mounted filesystem cannot be deleted.
1870 		 *
1871 		 * XXX: can this only be a VDIR case?
1872 		 */
1873 		if (vp->v_vflag & VV_ROOT)
1874 			error = EBUSY;
1875 	}
1876 	if (error == 0) {
1877 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1878 			NDFREE(&nd, NDF_ONLY_PNBUF);
1879 			vput(nd.ni_dvp);
1880 			if (vp == nd.ni_dvp)
1881 				vrele(vp);
1882 			else
1883 				vput(vp);
1884 			VFS_UNLOCK_GIANT(vfslocked);
1885 			if ((error = vn_start_write(NULL, &mp,
1886 			    V_XSLEEP | PCATCH)) != 0)
1887 				return (error);
1888 			goto restart;
1889 		}
1890 #ifdef MAC
1891 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1892 		    &nd.ni_cnd);
1893 		if (error)
1894 			goto out;
1895 #endif
1896 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1897 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1898 #ifdef MAC
1899 out:
1900 #endif
1901 		vn_finished_write(mp);
1902 	}
1903 	NDFREE(&nd, NDF_ONLY_PNBUF);
1904 	vput(nd.ni_dvp);
1905 	if (vp == nd.ni_dvp)
1906 		vrele(vp);
1907 	else
1908 		vput(vp);
1909 	VFS_UNLOCK_GIANT(vfslocked);
1910 	return (error);
1911 }
1912 
1913 /*
1914  * Reposition read/write file offset.
1915  */
1916 #ifndef _SYS_SYSPROTO_H_
1917 struct lseek_args {
1918 	int	fd;
1919 	int	pad;
1920 	off_t	offset;
1921 	int	whence;
1922 };
1923 #endif
1924 int
1925 lseek(td, uap)
1926 	struct thread *td;
1927 	register struct lseek_args /* {
1928 		int fd;
1929 		int pad;
1930 		off_t offset;
1931 		int whence;
1932 	} */ *uap;
1933 {
1934 	struct ucred *cred = td->td_ucred;
1935 	struct file *fp;
1936 	struct vnode *vp;
1937 	struct vattr vattr;
1938 	off_t offset;
1939 	int error, noneg;
1940 	int vfslocked;
1941 
1942 	if ((error = fget(td, uap->fd, &fp)) != 0)
1943 		return (error);
1944 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1945 		fdrop(fp, td);
1946 		return (ESPIPE);
1947 	}
1948 	vp = fp->f_vnode;
1949 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1950 	noneg = (vp->v_type != VCHR);
1951 	offset = uap->offset;
1952 	switch (uap->whence) {
1953 	case L_INCR:
1954 		if (noneg &&
1955 		    (fp->f_offset < 0 ||
1956 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1957 			error = EOVERFLOW;
1958 			break;
1959 		}
1960 		offset += fp->f_offset;
1961 		break;
1962 	case L_XTND:
1963 		vn_lock(vp, LK_SHARED | LK_RETRY);
1964 		error = VOP_GETATTR(vp, &vattr, cred);
1965 		VOP_UNLOCK(vp, 0);
1966 		if (error)
1967 			break;
1968 		if (noneg &&
1969 		    (vattr.va_size > OFF_MAX ||
1970 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1971 			error = EOVERFLOW;
1972 			break;
1973 		}
1974 		offset += vattr.va_size;
1975 		break;
1976 	case L_SET:
1977 		break;
1978 	case SEEK_DATA:
1979 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1980 		break;
1981 	case SEEK_HOLE:
1982 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1983 		break;
1984 	default:
1985 		error = EINVAL;
1986 	}
1987 	if (error == 0 && noneg && offset < 0)
1988 		error = EINVAL;
1989 	if (error != 0)
1990 		goto drop;
1991 	fp->f_offset = offset;
1992 	*(off_t *)(td->td_retval) = fp->f_offset;
1993 drop:
1994 	fdrop(fp, td);
1995 	VFS_UNLOCK_GIANT(vfslocked);
1996 	return (error);
1997 }
1998 
1999 #if defined(COMPAT_43)
2000 /*
2001  * Reposition read/write file offset.
2002  */
2003 #ifndef _SYS_SYSPROTO_H_
2004 struct olseek_args {
2005 	int	fd;
2006 	long	offset;
2007 	int	whence;
2008 };
2009 #endif
2010 int
2011 olseek(td, uap)
2012 	struct thread *td;
2013 	register struct olseek_args /* {
2014 		int fd;
2015 		long offset;
2016 		int whence;
2017 	} */ *uap;
2018 {
2019 	struct lseek_args /* {
2020 		int fd;
2021 		int pad;
2022 		off_t offset;
2023 		int whence;
2024 	} */ nuap;
2025 
2026 	nuap.fd = uap->fd;
2027 	nuap.offset = uap->offset;
2028 	nuap.whence = uap->whence;
2029 	return (lseek(td, &nuap));
2030 }
2031 #endif /* COMPAT_43 */
2032 
2033 /* Version with the 'pad' argument */
2034 int
2035 freebsd6_lseek(td, uap)
2036 	struct thread *td;
2037 	register struct freebsd6_lseek_args *uap;
2038 {
2039 	struct lseek_args ouap;
2040 
2041 	ouap.fd = uap->fd;
2042 	ouap.offset = uap->offset;
2043 	ouap.whence = uap->whence;
2044 	return (lseek(td, &ouap));
2045 }
2046 
2047 /*
2048  * Check access permissions using passed credentials.
2049  */
2050 static int
2051 vn_access(vp, user_flags, cred, td)
2052 	struct vnode	*vp;
2053 	int		user_flags;
2054 	struct ucred	*cred;
2055 	struct thread	*td;
2056 {
2057 	int error;
2058 	accmode_t accmode;
2059 
2060 	/* Flags == 0 means only check for existence. */
2061 	error = 0;
2062 	if (user_flags) {
2063 		accmode = 0;
2064 		if (user_flags & R_OK)
2065 			accmode |= VREAD;
2066 		if (user_flags & W_OK)
2067 			accmode |= VWRITE;
2068 		if (user_flags & X_OK)
2069 			accmode |= VEXEC;
2070 #ifdef MAC
2071 		error = mac_vnode_check_access(cred, vp, accmode);
2072 		if (error)
2073 			return (error);
2074 #endif
2075 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2076 			error = VOP_ACCESS(vp, accmode, cred, td);
2077 	}
2078 	return (error);
2079 }
2080 
2081 /*
2082  * Check access permissions using "real" credentials.
2083  */
2084 #ifndef _SYS_SYSPROTO_H_
2085 struct access_args {
2086 	char	*path;
2087 	int	flags;
2088 };
2089 #endif
2090 int
2091 access(td, uap)
2092 	struct thread *td;
2093 	register struct access_args /* {
2094 		char *path;
2095 		int flags;
2096 	} */ *uap;
2097 {
2098 
2099 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2100 }
2101 
2102 #ifndef _SYS_SYSPROTO_H_
2103 struct faccessat_args {
2104 	int	dirfd;
2105 	char	*path;
2106 	int	mode;
2107 	int	flag;
2108 }
2109 #endif
2110 int
2111 faccessat(struct thread *td, struct faccessat_args *uap)
2112 {
2113 
2114 	if (uap->flag & ~AT_EACCESS)
2115 		return (EINVAL);
2116 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2117 	    uap->mode));
2118 }
2119 
2120 int
2121 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2122 {
2123 
2124 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2125 }
2126 
2127 int
2128 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2129     int flags, int mode)
2130 {
2131 	struct ucred *cred, *tmpcred;
2132 	struct vnode *vp;
2133 	struct nameidata nd;
2134 	int vfslocked;
2135 	int error;
2136 
2137 	/*
2138 	 * Create and modify a temporary credential instead of one that
2139 	 * is potentially shared.  This could also mess up socket
2140 	 * buffer accounting which can run in an interrupt context.
2141 	 */
2142 	if (!(flags & AT_EACCESS)) {
2143 		cred = td->td_ucred;
2144 		tmpcred = crdup(cred);
2145 		tmpcred->cr_uid = cred->cr_ruid;
2146 		tmpcred->cr_groups[0] = cred->cr_rgid;
2147 		td->td_ucred = tmpcred;
2148 	} else
2149 		cred = tmpcred = td->td_ucred;
2150 	NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2151 	    AUDITVNODE1, pathseg, path, fd, td);
2152 	if ((error = namei(&nd)) != 0)
2153 		goto out1;
2154 	vfslocked = NDHASGIANT(&nd);
2155 	vp = nd.ni_vp;
2156 
2157 	error = vn_access(vp, mode, tmpcred, td);
2158 	NDFREE(&nd, NDF_ONLY_PNBUF);
2159 	vput(vp);
2160 	VFS_UNLOCK_GIANT(vfslocked);
2161 out1:
2162 	if (!(flags & AT_EACCESS)) {
2163 		td->td_ucred = cred;
2164 		crfree(tmpcred);
2165 	}
2166 	return (error);
2167 }
2168 
2169 /*
2170  * Check access permissions using "effective" credentials.
2171  */
2172 #ifndef _SYS_SYSPROTO_H_
2173 struct eaccess_args {
2174 	char	*path;
2175 	int	flags;
2176 };
2177 #endif
2178 int
2179 eaccess(td, uap)
2180 	struct thread *td;
2181 	register struct eaccess_args /* {
2182 		char *path;
2183 		int flags;
2184 	} */ *uap;
2185 {
2186 
2187 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2188 }
2189 
2190 int
2191 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2192 {
2193 
2194 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2195 }
2196 
2197 #if defined(COMPAT_43)
2198 /*
2199  * Get file status; this version follows links.
2200  */
2201 #ifndef _SYS_SYSPROTO_H_
2202 struct ostat_args {
2203 	char	*path;
2204 	struct ostat *ub;
2205 };
2206 #endif
2207 int
2208 ostat(td, uap)
2209 	struct thread *td;
2210 	register struct ostat_args /* {
2211 		char *path;
2212 		struct ostat *ub;
2213 	} */ *uap;
2214 {
2215 	struct stat sb;
2216 	struct ostat osb;
2217 	int error;
2218 
2219 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2220 	if (error)
2221 		return (error);
2222 	cvtstat(&sb, &osb);
2223 	error = copyout(&osb, uap->ub, sizeof (osb));
2224 	return (error);
2225 }
2226 
2227 /*
2228  * Get file status; this version does not follow links.
2229  */
2230 #ifndef _SYS_SYSPROTO_H_
2231 struct olstat_args {
2232 	char	*path;
2233 	struct ostat *ub;
2234 };
2235 #endif
2236 int
2237 olstat(td, uap)
2238 	struct thread *td;
2239 	register struct olstat_args /* {
2240 		char *path;
2241 		struct ostat *ub;
2242 	} */ *uap;
2243 {
2244 	struct stat sb;
2245 	struct ostat osb;
2246 	int error;
2247 
2248 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2249 	if (error)
2250 		return (error);
2251 	cvtstat(&sb, &osb);
2252 	error = copyout(&osb, uap->ub, sizeof (osb));
2253 	return (error);
2254 }
2255 
2256 /*
2257  * Convert from an old to a new stat structure.
2258  */
2259 void
2260 cvtstat(st, ost)
2261 	struct stat *st;
2262 	struct ostat *ost;
2263 {
2264 
2265 	ost->st_dev = st->st_dev;
2266 	ost->st_ino = st->st_ino;
2267 	ost->st_mode = st->st_mode;
2268 	ost->st_nlink = st->st_nlink;
2269 	ost->st_uid = st->st_uid;
2270 	ost->st_gid = st->st_gid;
2271 	ost->st_rdev = st->st_rdev;
2272 	if (st->st_size < (quad_t)1 << 32)
2273 		ost->st_size = st->st_size;
2274 	else
2275 		ost->st_size = -2;
2276 	ost->st_atime = st->st_atime;
2277 	ost->st_mtime = st->st_mtime;
2278 	ost->st_ctime = st->st_ctime;
2279 	ost->st_blksize = st->st_blksize;
2280 	ost->st_blocks = st->st_blocks;
2281 	ost->st_flags = st->st_flags;
2282 	ost->st_gen = st->st_gen;
2283 }
2284 #endif /* COMPAT_43 */
2285 
2286 /*
2287  * Get file status; this version follows links.
2288  */
2289 #ifndef _SYS_SYSPROTO_H_
2290 struct stat_args {
2291 	char	*path;
2292 	struct stat *ub;
2293 };
2294 #endif
2295 int
2296 stat(td, uap)
2297 	struct thread *td;
2298 	register struct stat_args /* {
2299 		char *path;
2300 		struct stat *ub;
2301 	} */ *uap;
2302 {
2303 	struct stat sb;
2304 	int error;
2305 
2306 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2307 	if (error == 0)
2308 		error = copyout(&sb, uap->ub, sizeof (sb));
2309 	return (error);
2310 }
2311 
2312 #ifndef _SYS_SYSPROTO_H_
2313 struct fstatat_args {
2314 	int	fd;
2315 	char	*path;
2316 	struct stat	*buf;
2317 	int	flag;
2318 }
2319 #endif
2320 int
2321 fstatat(struct thread *td, struct fstatat_args *uap)
2322 {
2323 	struct stat sb;
2324 	int error;
2325 
2326 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2327 	    UIO_USERSPACE, &sb);
2328 	if (error == 0)
2329 		error = copyout(&sb, uap->buf, sizeof (sb));
2330 	return (error);
2331 }
2332 
2333 int
2334 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2335 {
2336 
2337 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2338 }
2339 
2340 int
2341 kern_statat(struct thread *td, int flag, int fd, char *path,
2342     enum uio_seg pathseg, struct stat *sbp)
2343 {
2344 
2345 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2346 }
2347 
2348 int
2349 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2350     enum uio_seg pathseg, struct stat *sbp,
2351     void (*hook)(struct vnode *vp, struct stat *sbp))
2352 {
2353 	struct nameidata nd;
2354 	struct stat sb;
2355 	int error, vfslocked;
2356 
2357 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2358 		return (EINVAL);
2359 
2360 	NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2361 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2362 	    path, fd, td);
2363 
2364 	if ((error = namei(&nd)) != 0)
2365 		return (error);
2366 	vfslocked = NDHASGIANT(&nd);
2367 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2368 	if (!error) {
2369 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2370 		if (S_ISREG(sb.st_mode))
2371 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2372 		if (__predict_false(hook != NULL))
2373 			hook(nd.ni_vp, &sb);
2374 	}
2375 	NDFREE(&nd, NDF_ONLY_PNBUF);
2376 	vput(nd.ni_vp);
2377 	VFS_UNLOCK_GIANT(vfslocked);
2378 	if (error)
2379 		return (error);
2380 	*sbp = sb;
2381 #ifdef KTRACE
2382 	if (KTRPOINT(td, KTR_STRUCT))
2383 		ktrstat(&sb);
2384 #endif
2385 	return (0);
2386 }
2387 
2388 /*
2389  * Get file status; this version does not follow links.
2390  */
2391 #ifndef _SYS_SYSPROTO_H_
2392 struct lstat_args {
2393 	char	*path;
2394 	struct stat *ub;
2395 };
2396 #endif
2397 int
2398 lstat(td, uap)
2399 	struct thread *td;
2400 	register struct lstat_args /* {
2401 		char *path;
2402 		struct stat *ub;
2403 	} */ *uap;
2404 {
2405 	struct stat sb;
2406 	int error;
2407 
2408 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2409 	if (error == 0)
2410 		error = copyout(&sb, uap->ub, sizeof (sb));
2411 	return (error);
2412 }
2413 
2414 int
2415 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2416 {
2417 
2418 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2419 	    sbp));
2420 }
2421 
2422 /*
2423  * Implementation of the NetBSD [l]stat() functions.
2424  */
2425 void
2426 cvtnstat(sb, nsb)
2427 	struct stat *sb;
2428 	struct nstat *nsb;
2429 {
2430 	bzero(nsb, sizeof *nsb);
2431 	nsb->st_dev = sb->st_dev;
2432 	nsb->st_ino = sb->st_ino;
2433 	nsb->st_mode = sb->st_mode;
2434 	nsb->st_nlink = sb->st_nlink;
2435 	nsb->st_uid = sb->st_uid;
2436 	nsb->st_gid = sb->st_gid;
2437 	nsb->st_rdev = sb->st_rdev;
2438 	nsb->st_atimespec = sb->st_atimespec;
2439 	nsb->st_mtimespec = sb->st_mtimespec;
2440 	nsb->st_ctimespec = sb->st_ctimespec;
2441 	nsb->st_size = sb->st_size;
2442 	nsb->st_blocks = sb->st_blocks;
2443 	nsb->st_blksize = sb->st_blksize;
2444 	nsb->st_flags = sb->st_flags;
2445 	nsb->st_gen = sb->st_gen;
2446 	nsb->st_birthtimespec = sb->st_birthtimespec;
2447 }
2448 
2449 #ifndef _SYS_SYSPROTO_H_
2450 struct nstat_args {
2451 	char	*path;
2452 	struct nstat *ub;
2453 };
2454 #endif
2455 int
2456 nstat(td, uap)
2457 	struct thread *td;
2458 	register struct nstat_args /* {
2459 		char *path;
2460 		struct nstat *ub;
2461 	} */ *uap;
2462 {
2463 	struct stat sb;
2464 	struct nstat nsb;
2465 	int error;
2466 
2467 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2468 	if (error)
2469 		return (error);
2470 	cvtnstat(&sb, &nsb);
2471 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2472 	return (error);
2473 }
2474 
2475 /*
2476  * NetBSD lstat.  Get file status; this version does not follow links.
2477  */
2478 #ifndef _SYS_SYSPROTO_H_
2479 struct lstat_args {
2480 	char	*path;
2481 	struct stat *ub;
2482 };
2483 #endif
2484 int
2485 nlstat(td, uap)
2486 	struct thread *td;
2487 	register struct nlstat_args /* {
2488 		char *path;
2489 		struct nstat *ub;
2490 	} */ *uap;
2491 {
2492 	struct stat sb;
2493 	struct nstat nsb;
2494 	int error;
2495 
2496 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2497 	if (error)
2498 		return (error);
2499 	cvtnstat(&sb, &nsb);
2500 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2501 	return (error);
2502 }
2503 
2504 /*
2505  * Get configurable pathname variables.
2506  */
2507 #ifndef _SYS_SYSPROTO_H_
2508 struct pathconf_args {
2509 	char	*path;
2510 	int	name;
2511 };
2512 #endif
2513 int
2514 pathconf(td, uap)
2515 	struct thread *td;
2516 	register struct pathconf_args /* {
2517 		char *path;
2518 		int name;
2519 	} */ *uap;
2520 {
2521 
2522 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name));
2523 }
2524 
2525 int
2526 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name)
2527 {
2528 	struct nameidata nd;
2529 	int error, vfslocked;
2530 
2531 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2532 	    AUDITVNODE1, pathseg, path, td);
2533 	if ((error = namei(&nd)) != 0)
2534 		return (error);
2535 	vfslocked = NDHASGIANT(&nd);
2536 	NDFREE(&nd, NDF_ONLY_PNBUF);
2537 
2538 	/* If asynchronous I/O is available, it works for all files. */
2539 	if (name == _PC_ASYNC_IO)
2540 		td->td_retval[0] = async_io_version;
2541 	else
2542 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2543 	vput(nd.ni_vp);
2544 	VFS_UNLOCK_GIANT(vfslocked);
2545 	return (error);
2546 }
2547 
2548 /*
2549  * Return target name of a symbolic link.
2550  */
2551 #ifndef _SYS_SYSPROTO_H_
2552 struct readlink_args {
2553 	char	*path;
2554 	char	*buf;
2555 	size_t	count;
2556 };
2557 #endif
2558 int
2559 readlink(td, uap)
2560 	struct thread *td;
2561 	register struct readlink_args /* {
2562 		char *path;
2563 		char *buf;
2564 		size_t count;
2565 	} */ *uap;
2566 {
2567 
2568 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2569 	    UIO_USERSPACE, uap->count));
2570 }
2571 #ifndef _SYS_SYSPROTO_H_
2572 struct readlinkat_args {
2573 	int	fd;
2574 	char	*path;
2575 	char	*buf;
2576 	size_t	bufsize;
2577 };
2578 #endif
2579 int
2580 readlinkat(struct thread *td, struct readlinkat_args *uap)
2581 {
2582 
2583 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2584 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2585 }
2586 
2587 int
2588 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2589     enum uio_seg bufseg, size_t count)
2590 {
2591 
2592 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2593 	    count));
2594 }
2595 
2596 int
2597 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2598     char *buf, enum uio_seg bufseg, size_t count)
2599 {
2600 	struct vnode *vp;
2601 	struct iovec aiov;
2602 	struct uio auio;
2603 	int error;
2604 	struct nameidata nd;
2605 	int vfslocked;
2606 
2607 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2608 	    AUDITVNODE1, pathseg, path, fd, td);
2609 
2610 	if ((error = namei(&nd)) != 0)
2611 		return (error);
2612 	NDFREE(&nd, NDF_ONLY_PNBUF);
2613 	vfslocked = NDHASGIANT(&nd);
2614 	vp = nd.ni_vp;
2615 #ifdef MAC
2616 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2617 	if (error) {
2618 		vput(vp);
2619 		VFS_UNLOCK_GIANT(vfslocked);
2620 		return (error);
2621 	}
2622 #endif
2623 	if (vp->v_type != VLNK)
2624 		error = EINVAL;
2625 	else {
2626 		aiov.iov_base = buf;
2627 		aiov.iov_len = count;
2628 		auio.uio_iov = &aiov;
2629 		auio.uio_iovcnt = 1;
2630 		auio.uio_offset = 0;
2631 		auio.uio_rw = UIO_READ;
2632 		auio.uio_segflg = bufseg;
2633 		auio.uio_td = td;
2634 		auio.uio_resid = count;
2635 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2636 	}
2637 	vput(vp);
2638 	VFS_UNLOCK_GIANT(vfslocked);
2639 	td->td_retval[0] = count - auio.uio_resid;
2640 	return (error);
2641 }
2642 
2643 /*
2644  * Common implementation code for chflags() and fchflags().
2645  */
2646 static int
2647 setfflags(td, vp, flags)
2648 	struct thread *td;
2649 	struct vnode *vp;
2650 	int flags;
2651 {
2652 	int error;
2653 	struct mount *mp;
2654 	struct vattr vattr;
2655 
2656 	/*
2657 	 * Prevent non-root users from setting flags on devices.  When
2658 	 * a device is reused, users can retain ownership of the device
2659 	 * if they are allowed to set flags and programs assume that
2660 	 * chown can't fail when done as root.
2661 	 */
2662 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2663 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2664 		if (error)
2665 			return (error);
2666 	}
2667 
2668 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2669 		return (error);
2670 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2671 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2672 	VATTR_NULL(&vattr);
2673 	vattr.va_flags = flags;
2674 #ifdef MAC
2675 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2676 	if (error == 0)
2677 #endif
2678 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2679 	VOP_UNLOCK(vp, 0);
2680 	vn_finished_write(mp);
2681 	return (error);
2682 }
2683 
2684 /*
2685  * Change flags of a file given a path name.
2686  */
2687 #ifndef _SYS_SYSPROTO_H_
2688 struct chflags_args {
2689 	char	*path;
2690 	int	flags;
2691 };
2692 #endif
2693 int
2694 chflags(td, uap)
2695 	struct thread *td;
2696 	register struct chflags_args /* {
2697 		char *path;
2698 		int flags;
2699 	} */ *uap;
2700 {
2701 	int error;
2702 	struct nameidata nd;
2703 	int vfslocked;
2704 
2705 	AUDIT_ARG(fflags, uap->flags);
2706 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2707 	    uap->path, td);
2708 	if ((error = namei(&nd)) != 0)
2709 		return (error);
2710 	NDFREE(&nd, NDF_ONLY_PNBUF);
2711 	vfslocked = NDHASGIANT(&nd);
2712 	error = setfflags(td, nd.ni_vp, uap->flags);
2713 	vrele(nd.ni_vp);
2714 	VFS_UNLOCK_GIANT(vfslocked);
2715 	return (error);
2716 }
2717 
2718 /*
2719  * Same as chflags() but doesn't follow symlinks.
2720  */
2721 int
2722 lchflags(td, uap)
2723 	struct thread *td;
2724 	register struct lchflags_args /* {
2725 		char *path;
2726 		int flags;
2727 	} */ *uap;
2728 {
2729 	int error;
2730 	struct nameidata nd;
2731 	int vfslocked;
2732 
2733 	AUDIT_ARG(fflags, uap->flags);
2734 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2735 	    uap->path, td);
2736 	if ((error = namei(&nd)) != 0)
2737 		return (error);
2738 	vfslocked = NDHASGIANT(&nd);
2739 	NDFREE(&nd, NDF_ONLY_PNBUF);
2740 	error = setfflags(td, nd.ni_vp, uap->flags);
2741 	vrele(nd.ni_vp);
2742 	VFS_UNLOCK_GIANT(vfslocked);
2743 	return (error);
2744 }
2745 
2746 /*
2747  * Change flags of a file given a file descriptor.
2748  */
2749 #ifndef _SYS_SYSPROTO_H_
2750 struct fchflags_args {
2751 	int	fd;
2752 	int	flags;
2753 };
2754 #endif
2755 int
2756 fchflags(td, uap)
2757 	struct thread *td;
2758 	register struct fchflags_args /* {
2759 		int fd;
2760 		int flags;
2761 	} */ *uap;
2762 {
2763 	struct file *fp;
2764 	int vfslocked;
2765 	int error;
2766 
2767 	AUDIT_ARG(fd, uap->fd);
2768 	AUDIT_ARG(fflags, uap->flags);
2769 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2770 		return (error);
2771 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2772 #ifdef AUDIT
2773 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2774 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2775 	VOP_UNLOCK(fp->f_vnode, 0);
2776 #endif
2777 	error = setfflags(td, fp->f_vnode, uap->flags);
2778 	VFS_UNLOCK_GIANT(vfslocked);
2779 	fdrop(fp, td);
2780 	return (error);
2781 }
2782 
2783 /*
2784  * Common implementation code for chmod(), lchmod() and fchmod().
2785  */
2786 static int
2787 setfmode(td, vp, mode)
2788 	struct thread *td;
2789 	struct vnode *vp;
2790 	int mode;
2791 {
2792 	int error;
2793 	struct mount *mp;
2794 	struct vattr vattr;
2795 
2796 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2797 		return (error);
2798 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2799 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2800 	VATTR_NULL(&vattr);
2801 	vattr.va_mode = mode & ALLPERMS;
2802 #ifdef MAC
2803 	error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2804 	if (error == 0)
2805 #endif
2806 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2807 	VOP_UNLOCK(vp, 0);
2808 	vn_finished_write(mp);
2809 	return (error);
2810 }
2811 
2812 /*
2813  * Change mode of a file given path name.
2814  */
2815 #ifndef _SYS_SYSPROTO_H_
2816 struct chmod_args {
2817 	char	*path;
2818 	int	mode;
2819 };
2820 #endif
2821 int
2822 chmod(td, uap)
2823 	struct thread *td;
2824 	register struct chmod_args /* {
2825 		char *path;
2826 		int mode;
2827 	} */ *uap;
2828 {
2829 
2830 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2831 }
2832 
2833 #ifndef _SYS_SYSPROTO_H_
2834 struct fchmodat_args {
2835 	int	dirfd;
2836 	char	*path;
2837 	mode_t	mode;
2838 	int	flag;
2839 }
2840 #endif
2841 int
2842 fchmodat(struct thread *td, struct fchmodat_args *uap)
2843 {
2844 	int flag = uap->flag;
2845 	int fd = uap->fd;
2846 	char *path = uap->path;
2847 	mode_t mode = uap->mode;
2848 
2849 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2850 		return (EINVAL);
2851 
2852 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2853 }
2854 
2855 int
2856 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2857 {
2858 
2859 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2860 }
2861 
2862 /*
2863  * Change mode of a file given path name (don't follow links.)
2864  */
2865 #ifndef _SYS_SYSPROTO_H_
2866 struct lchmod_args {
2867 	char	*path;
2868 	int	mode;
2869 };
2870 #endif
2871 int
2872 lchmod(td, uap)
2873 	struct thread *td;
2874 	register struct lchmod_args /* {
2875 		char *path;
2876 		int mode;
2877 	} */ *uap;
2878 {
2879 
2880 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2881 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2882 }
2883 
2884 
2885 int
2886 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2887     mode_t mode, int flag)
2888 {
2889 	int error;
2890 	struct nameidata nd;
2891 	int vfslocked;
2892 	int follow;
2893 
2894 	AUDIT_ARG(mode, mode);
2895 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2896 	NDINIT_AT(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg, path,
2897 	    fd, td);
2898 	if ((error = namei(&nd)) != 0)
2899 		return (error);
2900 	vfslocked = NDHASGIANT(&nd);
2901 	NDFREE(&nd, NDF_ONLY_PNBUF);
2902 	error = setfmode(td, nd.ni_vp, mode);
2903 	vrele(nd.ni_vp);
2904 	VFS_UNLOCK_GIANT(vfslocked);
2905 	return (error);
2906 }
2907 
2908 /*
2909  * Change mode of a file given a file descriptor.
2910  */
2911 #ifndef _SYS_SYSPROTO_H_
2912 struct fchmod_args {
2913 	int	fd;
2914 	int	mode;
2915 };
2916 #endif
2917 int
2918 fchmod(td, uap)
2919 	struct thread *td;
2920 	register struct fchmod_args /* {
2921 		int fd;
2922 		int mode;
2923 	} */ *uap;
2924 {
2925 	struct file *fp;
2926 	int vfslocked;
2927 	int error;
2928 
2929 	AUDIT_ARG(fd, uap->fd);
2930 	AUDIT_ARG(mode, uap->mode);
2931 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2932 		return (error);
2933 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2934 #ifdef AUDIT
2935 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2936 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2937 	VOP_UNLOCK(fp->f_vnode, 0);
2938 #endif
2939 	error = setfmode(td, fp->f_vnode, uap->mode);
2940 	VFS_UNLOCK_GIANT(vfslocked);
2941 	fdrop(fp, td);
2942 	return (error);
2943 }
2944 
2945 /*
2946  * Common implementation for chown(), lchown(), and fchown()
2947  */
2948 static int
2949 setfown(td, vp, uid, gid)
2950 	struct thread *td;
2951 	struct vnode *vp;
2952 	uid_t uid;
2953 	gid_t gid;
2954 {
2955 	int error;
2956 	struct mount *mp;
2957 	struct vattr vattr;
2958 
2959 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2960 		return (error);
2961 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2962 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2963 	VATTR_NULL(&vattr);
2964 	vattr.va_uid = uid;
2965 	vattr.va_gid = gid;
2966 #ifdef MAC
2967 	error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2968 	    vattr.va_gid);
2969 	if (error == 0)
2970 #endif
2971 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2972 	VOP_UNLOCK(vp, 0);
2973 	vn_finished_write(mp);
2974 	return (error);
2975 }
2976 
2977 /*
2978  * Set ownership given a path name.
2979  */
2980 #ifndef _SYS_SYSPROTO_H_
2981 struct chown_args {
2982 	char	*path;
2983 	int	uid;
2984 	int	gid;
2985 };
2986 #endif
2987 int
2988 chown(td, uap)
2989 	struct thread *td;
2990 	register struct chown_args /* {
2991 		char *path;
2992 		int uid;
2993 		int gid;
2994 	} */ *uap;
2995 {
2996 
2997 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2998 }
2999 
3000 #ifndef _SYS_SYSPROTO_H_
3001 struct fchownat_args {
3002 	int fd;
3003 	const char * path;
3004 	uid_t uid;
3005 	gid_t gid;
3006 	int flag;
3007 };
3008 #endif
3009 int
3010 fchownat(struct thread *td, struct fchownat_args *uap)
3011 {
3012 	int flag;
3013 
3014 	flag = uap->flag;
3015 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3016 		return (EINVAL);
3017 
3018 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3019 	    uap->gid, uap->flag));
3020 }
3021 
3022 int
3023 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3024     int gid)
3025 {
3026 
3027 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3028 }
3029 
3030 int
3031 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3032     int uid, int gid, int flag)
3033 {
3034 	struct nameidata nd;
3035 	int error, vfslocked, follow;
3036 
3037 	AUDIT_ARG(owner, uid, gid);
3038 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3039 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
3040 	    fd, td);
3041 
3042 	if ((error = namei(&nd)) != 0)
3043 		return (error);
3044 	vfslocked = NDHASGIANT(&nd);
3045 	NDFREE(&nd, NDF_ONLY_PNBUF);
3046 	error = setfown(td, nd.ni_vp, uid, gid);
3047 	vrele(nd.ni_vp);
3048 	VFS_UNLOCK_GIANT(vfslocked);
3049 	return (error);
3050 }
3051 
3052 /*
3053  * Set ownership given a path name, do not cross symlinks.
3054  */
3055 #ifndef _SYS_SYSPROTO_H_
3056 struct lchown_args {
3057 	char	*path;
3058 	int	uid;
3059 	int	gid;
3060 };
3061 #endif
3062 int
3063 lchown(td, uap)
3064 	struct thread *td;
3065 	register struct lchown_args /* {
3066 		char *path;
3067 		int uid;
3068 		int gid;
3069 	} */ *uap;
3070 {
3071 
3072 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3073 }
3074 
3075 int
3076 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3077     int gid)
3078 {
3079 
3080 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3081 	    AT_SYMLINK_NOFOLLOW));
3082 }
3083 
3084 /*
3085  * Set ownership given a file descriptor.
3086  */
3087 #ifndef _SYS_SYSPROTO_H_
3088 struct fchown_args {
3089 	int	fd;
3090 	int	uid;
3091 	int	gid;
3092 };
3093 #endif
3094 int
3095 fchown(td, uap)
3096 	struct thread *td;
3097 	register struct fchown_args /* {
3098 		int fd;
3099 		int uid;
3100 		int gid;
3101 	} */ *uap;
3102 {
3103 	struct file *fp;
3104 	int vfslocked;
3105 	int error;
3106 
3107 	AUDIT_ARG(fd, uap->fd);
3108 	AUDIT_ARG(owner, uap->uid, uap->gid);
3109 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3110 		return (error);
3111 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3112 #ifdef AUDIT
3113 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3114 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3115 	VOP_UNLOCK(fp->f_vnode, 0);
3116 #endif
3117 	error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3118 	VFS_UNLOCK_GIANT(vfslocked);
3119 	fdrop(fp, td);
3120 	return (error);
3121 }
3122 
3123 /*
3124  * Common implementation code for utimes(), lutimes(), and futimes().
3125  */
3126 static int
3127 getutimes(usrtvp, tvpseg, tsp)
3128 	const struct timeval *usrtvp;
3129 	enum uio_seg tvpseg;
3130 	struct timespec *tsp;
3131 {
3132 	struct timeval tv[2];
3133 	const struct timeval *tvp;
3134 	int error;
3135 
3136 	if (usrtvp == NULL) {
3137 		microtime(&tv[0]);
3138 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
3139 		tsp[1] = tsp[0];
3140 	} else {
3141 		if (tvpseg == UIO_SYSSPACE) {
3142 			tvp = usrtvp;
3143 		} else {
3144 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3145 				return (error);
3146 			tvp = tv;
3147 		}
3148 
3149 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3150 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3151 			return (EINVAL);
3152 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3153 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3154 	}
3155 	return (0);
3156 }
3157 
3158 /*
3159  * Common implementation code for utimes(), lutimes(), and futimes().
3160  */
3161 static int
3162 setutimes(td, vp, ts, numtimes, nullflag)
3163 	struct thread *td;
3164 	struct vnode *vp;
3165 	const struct timespec *ts;
3166 	int numtimes;
3167 	int nullflag;
3168 {
3169 	int error, setbirthtime;
3170 	struct mount *mp;
3171 	struct vattr vattr;
3172 
3173 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3174 		return (error);
3175 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3176 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3177 	setbirthtime = 0;
3178 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3179 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3180 		setbirthtime = 1;
3181 	VATTR_NULL(&vattr);
3182 	vattr.va_atime = ts[0];
3183 	vattr.va_mtime = ts[1];
3184 	if (setbirthtime)
3185 		vattr.va_birthtime = ts[1];
3186 	if (numtimes > 2)
3187 		vattr.va_birthtime = ts[2];
3188 	if (nullflag)
3189 		vattr.va_vaflags |= VA_UTIMES_NULL;
3190 #ifdef MAC
3191 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3192 	    vattr.va_mtime);
3193 #endif
3194 	if (error == 0)
3195 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3196 	VOP_UNLOCK(vp, 0);
3197 	vn_finished_write(mp);
3198 	return (error);
3199 }
3200 
3201 /*
3202  * Set the access and modification times of a file.
3203  */
3204 #ifndef _SYS_SYSPROTO_H_
3205 struct utimes_args {
3206 	char	*path;
3207 	struct	timeval *tptr;
3208 };
3209 #endif
3210 int
3211 utimes(td, uap)
3212 	struct thread *td;
3213 	register struct utimes_args /* {
3214 		char *path;
3215 		struct timeval *tptr;
3216 	} */ *uap;
3217 {
3218 
3219 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3220 	    UIO_USERSPACE));
3221 }
3222 
3223 #ifndef _SYS_SYSPROTO_H_
3224 struct futimesat_args {
3225 	int fd;
3226 	const char * path;
3227 	const struct timeval * times;
3228 };
3229 #endif
3230 int
3231 futimesat(struct thread *td, struct futimesat_args *uap)
3232 {
3233 
3234 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3235 	    uap->times, UIO_USERSPACE));
3236 }
3237 
3238 int
3239 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3240     struct timeval *tptr, enum uio_seg tptrseg)
3241 {
3242 
3243 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3244 }
3245 
3246 int
3247 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3248     struct timeval *tptr, enum uio_seg tptrseg)
3249 {
3250 	struct nameidata nd;
3251 	struct timespec ts[2];
3252 	int error, vfslocked;
3253 
3254 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3255 		return (error);
3256 	NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3257 	    fd, td);
3258 
3259 	if ((error = namei(&nd)) != 0)
3260 		return (error);
3261 	vfslocked = NDHASGIANT(&nd);
3262 	NDFREE(&nd, NDF_ONLY_PNBUF);
3263 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3264 	vrele(nd.ni_vp);
3265 	VFS_UNLOCK_GIANT(vfslocked);
3266 	return (error);
3267 }
3268 
3269 /*
3270  * Set the access and modification times of a file.
3271  */
3272 #ifndef _SYS_SYSPROTO_H_
3273 struct lutimes_args {
3274 	char	*path;
3275 	struct	timeval *tptr;
3276 };
3277 #endif
3278 int
3279 lutimes(td, uap)
3280 	struct thread *td;
3281 	register struct lutimes_args /* {
3282 		char *path;
3283 		struct timeval *tptr;
3284 	} */ *uap;
3285 {
3286 
3287 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3288 	    UIO_USERSPACE));
3289 }
3290 
3291 int
3292 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3293     struct timeval *tptr, enum uio_seg tptrseg)
3294 {
3295 	struct timespec ts[2];
3296 	int error;
3297 	struct nameidata nd;
3298 	int vfslocked;
3299 
3300 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3301 		return (error);
3302 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3303 	if ((error = namei(&nd)) != 0)
3304 		return (error);
3305 	vfslocked = NDHASGIANT(&nd);
3306 	NDFREE(&nd, NDF_ONLY_PNBUF);
3307 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3308 	vrele(nd.ni_vp);
3309 	VFS_UNLOCK_GIANT(vfslocked);
3310 	return (error);
3311 }
3312 
3313 /*
3314  * Set the access and modification times of a file.
3315  */
3316 #ifndef _SYS_SYSPROTO_H_
3317 struct futimes_args {
3318 	int	fd;
3319 	struct	timeval *tptr;
3320 };
3321 #endif
3322 int
3323 futimes(td, uap)
3324 	struct thread *td;
3325 	register struct futimes_args /* {
3326 		int  fd;
3327 		struct timeval *tptr;
3328 	} */ *uap;
3329 {
3330 
3331 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3332 }
3333 
3334 int
3335 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3336     enum uio_seg tptrseg)
3337 {
3338 	struct timespec ts[2];
3339 	struct file *fp;
3340 	int vfslocked;
3341 	int error;
3342 
3343 	AUDIT_ARG(fd, fd);
3344 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3345 		return (error);
3346 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3347 		return (error);
3348 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3349 #ifdef AUDIT
3350 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3351 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3352 	VOP_UNLOCK(fp->f_vnode, 0);
3353 #endif
3354 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3355 	VFS_UNLOCK_GIANT(vfslocked);
3356 	fdrop(fp, td);
3357 	return (error);
3358 }
3359 
3360 /*
3361  * Truncate a file given its path name.
3362  */
3363 #ifndef _SYS_SYSPROTO_H_
3364 struct truncate_args {
3365 	char	*path;
3366 	int	pad;
3367 	off_t	length;
3368 };
3369 #endif
3370 int
3371 truncate(td, uap)
3372 	struct thread *td;
3373 	register struct truncate_args /* {
3374 		char *path;
3375 		int pad;
3376 		off_t length;
3377 	} */ *uap;
3378 {
3379 
3380 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3381 }
3382 
3383 int
3384 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3385 {
3386 	struct mount *mp;
3387 	struct vnode *vp;
3388 	struct vattr vattr;
3389 	int error;
3390 	struct nameidata nd;
3391 	int vfslocked;
3392 
3393 	if (length < 0)
3394 		return(EINVAL);
3395 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3396 	if ((error = namei(&nd)) != 0)
3397 		return (error);
3398 	vfslocked = NDHASGIANT(&nd);
3399 	vp = nd.ni_vp;
3400 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3401 		vrele(vp);
3402 		VFS_UNLOCK_GIANT(vfslocked);
3403 		return (error);
3404 	}
3405 	NDFREE(&nd, NDF_ONLY_PNBUF);
3406 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3407 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3408 	if (vp->v_type == VDIR)
3409 		error = EISDIR;
3410 #ifdef MAC
3411 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3412 	}
3413 #endif
3414 	else if ((error = vn_writechk(vp)) == 0 &&
3415 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3416 		VATTR_NULL(&vattr);
3417 		vattr.va_size = length;
3418 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3419 	}
3420 	vput(vp);
3421 	vn_finished_write(mp);
3422 	VFS_UNLOCK_GIANT(vfslocked);
3423 	return (error);
3424 }
3425 
3426 #if defined(COMPAT_43)
3427 /*
3428  * Truncate a file given its path name.
3429  */
3430 #ifndef _SYS_SYSPROTO_H_
3431 struct otruncate_args {
3432 	char	*path;
3433 	long	length;
3434 };
3435 #endif
3436 int
3437 otruncate(td, uap)
3438 	struct thread *td;
3439 	register struct otruncate_args /* {
3440 		char *path;
3441 		long length;
3442 	} */ *uap;
3443 {
3444 	struct truncate_args /* {
3445 		char *path;
3446 		int pad;
3447 		off_t length;
3448 	} */ nuap;
3449 
3450 	nuap.path = uap->path;
3451 	nuap.length = uap->length;
3452 	return (truncate(td, &nuap));
3453 }
3454 #endif /* COMPAT_43 */
3455 
3456 /* Versions with the pad argument */
3457 int
3458 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3459 {
3460 	struct truncate_args ouap;
3461 
3462 	ouap.path = uap->path;
3463 	ouap.length = uap->length;
3464 	return (truncate(td, &ouap));
3465 }
3466 
3467 int
3468 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3469 {
3470 	struct ftruncate_args ouap;
3471 
3472 	ouap.fd = uap->fd;
3473 	ouap.length = uap->length;
3474 	return (ftruncate(td, &ouap));
3475 }
3476 
3477 /*
3478  * Sync an open file.
3479  */
3480 #ifndef _SYS_SYSPROTO_H_
3481 struct fsync_args {
3482 	int	fd;
3483 };
3484 #endif
3485 int
3486 fsync(td, uap)
3487 	struct thread *td;
3488 	struct fsync_args /* {
3489 		int fd;
3490 	} */ *uap;
3491 {
3492 	struct vnode *vp;
3493 	struct mount *mp;
3494 	struct file *fp;
3495 	int vfslocked;
3496 	int error;
3497 
3498 	AUDIT_ARG(fd, uap->fd);
3499 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3500 		return (error);
3501 	vp = fp->f_vnode;
3502 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3503 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3504 		goto drop;
3505 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3506 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
3507 	if (vp->v_object != NULL) {
3508 		VM_OBJECT_LOCK(vp->v_object);
3509 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3510 		VM_OBJECT_UNLOCK(vp->v_object);
3511 	}
3512 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3513 
3514 	VOP_UNLOCK(vp, 0);
3515 	vn_finished_write(mp);
3516 drop:
3517 	VFS_UNLOCK_GIANT(vfslocked);
3518 	fdrop(fp, td);
3519 	return (error);
3520 }
3521 
3522 /*
3523  * Rename files.  Source and destination must either both be directories, or
3524  * both not be directories.  If target is a directory, it must be empty.
3525  */
3526 #ifndef _SYS_SYSPROTO_H_
3527 struct rename_args {
3528 	char	*from;
3529 	char	*to;
3530 };
3531 #endif
3532 int
3533 rename(td, uap)
3534 	struct thread *td;
3535 	register struct rename_args /* {
3536 		char *from;
3537 		char *to;
3538 	} */ *uap;
3539 {
3540 
3541 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3542 }
3543 
3544 #ifndef _SYS_SYSPROTO_H_
3545 struct renameat_args {
3546 	int	oldfd;
3547 	char	*old;
3548 	int	newfd;
3549 	char	*new;
3550 };
3551 #endif
3552 int
3553 renameat(struct thread *td, struct renameat_args *uap)
3554 {
3555 
3556 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3557 	    UIO_USERSPACE));
3558 }
3559 
3560 int
3561 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3562 {
3563 
3564 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3565 }
3566 
3567 int
3568 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3569     enum uio_seg pathseg)
3570 {
3571 	struct mount *mp = NULL;
3572 	struct vnode *tvp, *fvp, *tdvp;
3573 	struct nameidata fromnd, tond;
3574 	int tvfslocked;
3575 	int fvfslocked;
3576 	int error;
3577 
3578 	bwillwrite();
3579 #ifdef MAC
3580 	NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3581 	    AUDITVNODE1, pathseg, old, oldfd, td);
3582 #else
3583 	NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3584 	    AUDITVNODE1, pathseg, old, oldfd, td);
3585 #endif
3586 
3587 	if ((error = namei(&fromnd)) != 0)
3588 		return (error);
3589 	fvfslocked = NDHASGIANT(&fromnd);
3590 	tvfslocked = 0;
3591 #ifdef MAC
3592 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3593 	    fromnd.ni_vp, &fromnd.ni_cnd);
3594 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3595 	if (fromnd.ni_dvp != fromnd.ni_vp)
3596 		VOP_UNLOCK(fromnd.ni_vp, 0);
3597 #endif
3598 	fvp = fromnd.ni_vp;
3599 	if (error == 0)
3600 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3601 	if (error != 0) {
3602 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3603 		vrele(fromnd.ni_dvp);
3604 		vrele(fvp);
3605 		goto out1;
3606 	}
3607 	NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3608 	    MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3609 	if (fromnd.ni_vp->v_type == VDIR)
3610 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3611 	if ((error = namei(&tond)) != 0) {
3612 		/* Translate error code for rename("dir1", "dir2/."). */
3613 		if (error == EISDIR && fvp->v_type == VDIR)
3614 			error = EINVAL;
3615 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3616 		vrele(fromnd.ni_dvp);
3617 		vrele(fvp);
3618 		vn_finished_write(mp);
3619 		goto out1;
3620 	}
3621 	tvfslocked = NDHASGIANT(&tond);
3622 	tdvp = tond.ni_dvp;
3623 	tvp = tond.ni_vp;
3624 	if (tvp != NULL) {
3625 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3626 			error = ENOTDIR;
3627 			goto out;
3628 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3629 			error = EISDIR;
3630 			goto out;
3631 		}
3632 	}
3633 	if (fvp == tdvp) {
3634 		error = EINVAL;
3635 		goto out;
3636 	}
3637 	/*
3638 	 * If the source is the same as the destination (that is, if they
3639 	 * are links to the same vnode), then there is nothing to do.
3640 	 */
3641 	if (fvp == tvp)
3642 		error = -1;
3643 #ifdef MAC
3644 	else
3645 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3646 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3647 #endif
3648 out:
3649 	if (!error) {
3650 		VOP_LEASE(tdvp, td, td->td_ucred, LEASE_WRITE);
3651 		if (fromnd.ni_dvp != tdvp) {
3652 			VOP_LEASE(fromnd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3653 		}
3654 		if (tvp) {
3655 			VOP_LEASE(tvp, td, td->td_ucred, LEASE_WRITE);
3656 		}
3657 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3658 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3659 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3660 		NDFREE(&tond, NDF_ONLY_PNBUF);
3661 	} else {
3662 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3663 		NDFREE(&tond, NDF_ONLY_PNBUF);
3664 		if (tvp)
3665 			vput(tvp);
3666 		if (tdvp == tvp)
3667 			vrele(tdvp);
3668 		else
3669 			vput(tdvp);
3670 		vrele(fromnd.ni_dvp);
3671 		vrele(fvp);
3672 	}
3673 	vrele(tond.ni_startdir);
3674 	vn_finished_write(mp);
3675 out1:
3676 	if (fromnd.ni_startdir)
3677 		vrele(fromnd.ni_startdir);
3678 	VFS_UNLOCK_GIANT(fvfslocked);
3679 	VFS_UNLOCK_GIANT(tvfslocked);
3680 	if (error == -1)
3681 		return (0);
3682 	return (error);
3683 }
3684 
3685 /*
3686  * Make a directory file.
3687  */
3688 #ifndef _SYS_SYSPROTO_H_
3689 struct mkdir_args {
3690 	char	*path;
3691 	int	mode;
3692 };
3693 #endif
3694 int
3695 mkdir(td, uap)
3696 	struct thread *td;
3697 	register struct mkdir_args /* {
3698 		char *path;
3699 		int mode;
3700 	} */ *uap;
3701 {
3702 
3703 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3704 }
3705 
3706 #ifndef _SYS_SYSPROTO_H_
3707 struct mkdirat_args {
3708 	int	fd;
3709 	char	*path;
3710 	mode_t	mode;
3711 };
3712 #endif
3713 int
3714 mkdirat(struct thread *td, struct mkdirat_args *uap)
3715 {
3716 
3717 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3718 }
3719 
3720 int
3721 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3722 {
3723 
3724 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3725 }
3726 
3727 int
3728 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3729     int mode)
3730 {
3731 	struct mount *mp;
3732 	struct vnode *vp;
3733 	struct vattr vattr;
3734 	int error;
3735 	struct nameidata nd;
3736 	int vfslocked;
3737 
3738 	AUDIT_ARG(mode, mode);
3739 restart:
3740 	bwillwrite();
3741 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3742 	    segflg, path, fd, td);
3743 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3744 	if ((error = namei(&nd)) != 0)
3745 		return (error);
3746 	vfslocked = NDHASGIANT(&nd);
3747 	vp = nd.ni_vp;
3748 	if (vp != NULL) {
3749 		NDFREE(&nd, NDF_ONLY_PNBUF);
3750 		/*
3751 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3752 		 * the strange behaviour of leaving the vnode unlocked
3753 		 * if the target is the same vnode as the parent.
3754 		 */
3755 		if (vp == nd.ni_dvp)
3756 			vrele(nd.ni_dvp);
3757 		else
3758 			vput(nd.ni_dvp);
3759 		vrele(vp);
3760 		VFS_UNLOCK_GIANT(vfslocked);
3761 		return (EEXIST);
3762 	}
3763 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3764 		NDFREE(&nd, NDF_ONLY_PNBUF);
3765 		vput(nd.ni_dvp);
3766 		VFS_UNLOCK_GIANT(vfslocked);
3767 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3768 			return (error);
3769 		goto restart;
3770 	}
3771 	VATTR_NULL(&vattr);
3772 	vattr.va_type = VDIR;
3773 	FILEDESC_SLOCK(td->td_proc->p_fd);
3774 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3775 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
3776 #ifdef MAC
3777 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3778 	    &vattr);
3779 	if (error)
3780 		goto out;
3781 #endif
3782 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3783 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3784 #ifdef MAC
3785 out:
3786 #endif
3787 	NDFREE(&nd, NDF_ONLY_PNBUF);
3788 	vput(nd.ni_dvp);
3789 	if (!error)
3790 		vput(nd.ni_vp);
3791 	vn_finished_write(mp);
3792 	VFS_UNLOCK_GIANT(vfslocked);
3793 	return (error);
3794 }
3795 
3796 /*
3797  * Remove a directory file.
3798  */
3799 #ifndef _SYS_SYSPROTO_H_
3800 struct rmdir_args {
3801 	char	*path;
3802 };
3803 #endif
3804 int
3805 rmdir(td, uap)
3806 	struct thread *td;
3807 	struct rmdir_args /* {
3808 		char *path;
3809 	} */ *uap;
3810 {
3811 
3812 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3813 }
3814 
3815 int
3816 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3817 {
3818 
3819 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3820 }
3821 
3822 int
3823 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3824 {
3825 	struct mount *mp;
3826 	struct vnode *vp;
3827 	int error;
3828 	struct nameidata nd;
3829 	int vfslocked;
3830 
3831 restart:
3832 	bwillwrite();
3833 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3834 	    pathseg, path, fd, td);
3835 	if ((error = namei(&nd)) != 0)
3836 		return (error);
3837 	vfslocked = NDHASGIANT(&nd);
3838 	vp = nd.ni_vp;
3839 	if (vp->v_type != VDIR) {
3840 		error = ENOTDIR;
3841 		goto out;
3842 	}
3843 	/*
3844 	 * No rmdir "." please.
3845 	 */
3846 	if (nd.ni_dvp == vp) {
3847 		error = EINVAL;
3848 		goto out;
3849 	}
3850 	/*
3851 	 * The root of a mounted filesystem cannot be deleted.
3852 	 */
3853 	if (vp->v_vflag & VV_ROOT) {
3854 		error = EBUSY;
3855 		goto out;
3856 	}
3857 #ifdef MAC
3858 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3859 	    &nd.ni_cnd);
3860 	if (error)
3861 		goto out;
3862 #endif
3863 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3864 		NDFREE(&nd, NDF_ONLY_PNBUF);
3865 		vput(vp);
3866 		if (nd.ni_dvp == vp)
3867 			vrele(nd.ni_dvp);
3868 		else
3869 			vput(nd.ni_dvp);
3870 		VFS_UNLOCK_GIANT(vfslocked);
3871 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3872 			return (error);
3873 		goto restart;
3874 	}
3875 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3876 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3877 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3878 	vn_finished_write(mp);
3879 out:
3880 	NDFREE(&nd, NDF_ONLY_PNBUF);
3881 	vput(vp);
3882 	if (nd.ni_dvp == vp)
3883 		vrele(nd.ni_dvp);
3884 	else
3885 		vput(nd.ni_dvp);
3886 	VFS_UNLOCK_GIANT(vfslocked);
3887 	return (error);
3888 }
3889 
3890 #ifdef COMPAT_43
3891 /*
3892  * Read a block of directory entries in a filesystem independent format.
3893  */
3894 #ifndef _SYS_SYSPROTO_H_
3895 struct ogetdirentries_args {
3896 	int	fd;
3897 	char	*buf;
3898 	u_int	count;
3899 	long	*basep;
3900 };
3901 #endif
3902 int
3903 ogetdirentries(td, uap)
3904 	struct thread *td;
3905 	register struct ogetdirentries_args /* {
3906 		int fd;
3907 		char *buf;
3908 		u_int count;
3909 		long *basep;
3910 	} */ *uap;
3911 {
3912 	struct vnode *vp;
3913 	struct file *fp;
3914 	struct uio auio, kuio;
3915 	struct iovec aiov, kiov;
3916 	struct dirent *dp, *edp;
3917 	caddr_t dirbuf;
3918 	int error, eofflag, readcnt, vfslocked;
3919 	long loff;
3920 
3921 	/* XXX arbitrary sanity limit on `count'. */
3922 	if (uap->count > 64 * 1024)
3923 		return (EINVAL);
3924 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3925 		return (error);
3926 	if ((fp->f_flag & FREAD) == 0) {
3927 		fdrop(fp, td);
3928 		return (EBADF);
3929 	}
3930 	vp = fp->f_vnode;
3931 unionread:
3932 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3933 	if (vp->v_type != VDIR) {
3934 		VFS_UNLOCK_GIANT(vfslocked);
3935 		fdrop(fp, td);
3936 		return (EINVAL);
3937 	}
3938 	aiov.iov_base = uap->buf;
3939 	aiov.iov_len = uap->count;
3940 	auio.uio_iov = &aiov;
3941 	auio.uio_iovcnt = 1;
3942 	auio.uio_rw = UIO_READ;
3943 	auio.uio_segflg = UIO_USERSPACE;
3944 	auio.uio_td = td;
3945 	auio.uio_resid = uap->count;
3946 	vn_lock(vp, LK_SHARED | LK_RETRY);
3947 	loff = auio.uio_offset = fp->f_offset;
3948 #ifdef MAC
3949 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3950 	if (error) {
3951 		VOP_UNLOCK(vp, 0);
3952 		VFS_UNLOCK_GIANT(vfslocked);
3953 		fdrop(fp, td);
3954 		return (error);
3955 	}
3956 #endif
3957 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3958 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3959 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3960 			    NULL, NULL);
3961 			fp->f_offset = auio.uio_offset;
3962 		} else
3963 #	endif
3964 	{
3965 		kuio = auio;
3966 		kuio.uio_iov = &kiov;
3967 		kuio.uio_segflg = UIO_SYSSPACE;
3968 		kiov.iov_len = uap->count;
3969 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3970 		kiov.iov_base = dirbuf;
3971 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3972 			    NULL, NULL);
3973 		fp->f_offset = kuio.uio_offset;
3974 		if (error == 0) {
3975 			readcnt = uap->count - kuio.uio_resid;
3976 			edp = (struct dirent *)&dirbuf[readcnt];
3977 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3978 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3979 					/*
3980 					 * The expected low byte of
3981 					 * dp->d_namlen is our dp->d_type.
3982 					 * The high MBZ byte of dp->d_namlen
3983 					 * is our dp->d_namlen.
3984 					 */
3985 					dp->d_type = dp->d_namlen;
3986 					dp->d_namlen = 0;
3987 #				else
3988 					/*
3989 					 * The dp->d_type is the high byte
3990 					 * of the expected dp->d_namlen,
3991 					 * so must be zero'ed.
3992 					 */
3993 					dp->d_type = 0;
3994 #				endif
3995 				if (dp->d_reclen > 0) {
3996 					dp = (struct dirent *)
3997 					    ((char *)dp + dp->d_reclen);
3998 				} else {
3999 					error = EIO;
4000 					break;
4001 				}
4002 			}
4003 			if (dp >= edp)
4004 				error = uiomove(dirbuf, readcnt, &auio);
4005 		}
4006 		free(dirbuf, M_TEMP);
4007 	}
4008 	if (error) {
4009 		VOP_UNLOCK(vp, 0);
4010 		VFS_UNLOCK_GIANT(vfslocked);
4011 		fdrop(fp, td);
4012 		return (error);
4013 	}
4014 	if (uap->count == auio.uio_resid &&
4015 	    (vp->v_vflag & VV_ROOT) &&
4016 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4017 		struct vnode *tvp = vp;
4018 		vp = vp->v_mount->mnt_vnodecovered;
4019 		VREF(vp);
4020 		fp->f_vnode = vp;
4021 		fp->f_data = vp;
4022 		fp->f_offset = 0;
4023 		vput(tvp);
4024 		VFS_UNLOCK_GIANT(vfslocked);
4025 		goto unionread;
4026 	}
4027 	VOP_UNLOCK(vp, 0);
4028 	VFS_UNLOCK_GIANT(vfslocked);
4029 	error = copyout(&loff, uap->basep, sizeof(long));
4030 	fdrop(fp, td);
4031 	td->td_retval[0] = uap->count - auio.uio_resid;
4032 	return (error);
4033 }
4034 #endif /* COMPAT_43 */
4035 
4036 /*
4037  * Read a block of directory entries in a filesystem independent format.
4038  */
4039 #ifndef _SYS_SYSPROTO_H_
4040 struct getdirentries_args {
4041 	int	fd;
4042 	char	*buf;
4043 	u_int	count;
4044 	long	*basep;
4045 };
4046 #endif
4047 int
4048 getdirentries(td, uap)
4049 	struct thread *td;
4050 	register struct getdirentries_args /* {
4051 		int fd;
4052 		char *buf;
4053 		u_int count;
4054 		long *basep;
4055 	} */ *uap;
4056 {
4057 	long base;
4058 	int error;
4059 
4060 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4061 	if (error)
4062 		return (error);
4063 	if (uap->basep != NULL)
4064 		error = copyout(&base, uap->basep, sizeof(long));
4065 	return (error);
4066 }
4067 
4068 int
4069 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4070     long *basep)
4071 {
4072 	struct vnode *vp;
4073 	struct file *fp;
4074 	struct uio auio;
4075 	struct iovec aiov;
4076 	int vfslocked;
4077 	long loff;
4078 	int error, eofflag;
4079 
4080 	AUDIT_ARG(fd, fd);
4081 	if (count > INT_MAX)
4082 		return (EINVAL);
4083 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
4084 		return (error);
4085 	if ((fp->f_flag & FREAD) == 0) {
4086 		fdrop(fp, td);
4087 		return (EBADF);
4088 	}
4089 	vp = fp->f_vnode;
4090 unionread:
4091 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4092 	if (vp->v_type != VDIR) {
4093 		VFS_UNLOCK_GIANT(vfslocked);
4094 		error = EINVAL;
4095 		goto fail;
4096 	}
4097 	aiov.iov_base = buf;
4098 	aiov.iov_len = count;
4099 	auio.uio_iov = &aiov;
4100 	auio.uio_iovcnt = 1;
4101 	auio.uio_rw = UIO_READ;
4102 	auio.uio_segflg = UIO_USERSPACE;
4103 	auio.uio_td = td;
4104 	auio.uio_resid = count;
4105 	vn_lock(vp, LK_SHARED | LK_RETRY);
4106 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
4107 	loff = auio.uio_offset = fp->f_offset;
4108 #ifdef MAC
4109 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4110 	if (error == 0)
4111 #endif
4112 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4113 		    NULL);
4114 	fp->f_offset = auio.uio_offset;
4115 	if (error) {
4116 		VOP_UNLOCK(vp, 0);
4117 		VFS_UNLOCK_GIANT(vfslocked);
4118 		goto fail;
4119 	}
4120 	if (count == auio.uio_resid &&
4121 	    (vp->v_vflag & VV_ROOT) &&
4122 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4123 		struct vnode *tvp = vp;
4124 		vp = vp->v_mount->mnt_vnodecovered;
4125 		VREF(vp);
4126 		fp->f_vnode = vp;
4127 		fp->f_data = vp;
4128 		fp->f_offset = 0;
4129 		vput(tvp);
4130 		VFS_UNLOCK_GIANT(vfslocked);
4131 		goto unionread;
4132 	}
4133 	VOP_UNLOCK(vp, 0);
4134 	VFS_UNLOCK_GIANT(vfslocked);
4135 	*basep = loff;
4136 	td->td_retval[0] = count - auio.uio_resid;
4137 fail:
4138 	fdrop(fp, td);
4139 	return (error);
4140 }
4141 
4142 #ifndef _SYS_SYSPROTO_H_
4143 struct getdents_args {
4144 	int fd;
4145 	char *buf;
4146 	size_t count;
4147 };
4148 #endif
4149 int
4150 getdents(td, uap)
4151 	struct thread *td;
4152 	register struct getdents_args /* {
4153 		int fd;
4154 		char *buf;
4155 		u_int count;
4156 	} */ *uap;
4157 {
4158 	struct getdirentries_args ap;
4159 	ap.fd = uap->fd;
4160 	ap.buf = uap->buf;
4161 	ap.count = uap->count;
4162 	ap.basep = NULL;
4163 	return (getdirentries(td, &ap));
4164 }
4165 
4166 /*
4167  * Set the mode mask for creation of filesystem nodes.
4168  */
4169 #ifndef _SYS_SYSPROTO_H_
4170 struct umask_args {
4171 	int	newmask;
4172 };
4173 #endif
4174 int
4175 umask(td, uap)
4176 	struct thread *td;
4177 	struct umask_args /* {
4178 		int newmask;
4179 	} */ *uap;
4180 {
4181 	register struct filedesc *fdp;
4182 
4183 	FILEDESC_XLOCK(td->td_proc->p_fd);
4184 	fdp = td->td_proc->p_fd;
4185 	td->td_retval[0] = fdp->fd_cmask;
4186 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4187 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4188 	return (0);
4189 }
4190 
4191 /*
4192  * Void all references to file by ripping underlying filesystem away from
4193  * vnode.
4194  */
4195 #ifndef _SYS_SYSPROTO_H_
4196 struct revoke_args {
4197 	char	*path;
4198 };
4199 #endif
4200 int
4201 revoke(td, uap)
4202 	struct thread *td;
4203 	register struct revoke_args /* {
4204 		char *path;
4205 	} */ *uap;
4206 {
4207 	struct vnode *vp;
4208 	struct vattr vattr;
4209 	int error;
4210 	struct nameidata nd;
4211 	int vfslocked;
4212 
4213 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4214 	    UIO_USERSPACE, uap->path, td);
4215 	if ((error = namei(&nd)) != 0)
4216 		return (error);
4217 	vfslocked = NDHASGIANT(&nd);
4218 	vp = nd.ni_vp;
4219 	NDFREE(&nd, NDF_ONLY_PNBUF);
4220 	if (vp->v_type != VCHR) {
4221 		error = EINVAL;
4222 		goto out;
4223 	}
4224 #ifdef MAC
4225 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4226 	if (error)
4227 		goto out;
4228 #endif
4229 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4230 	if (error)
4231 		goto out;
4232 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4233 		error = priv_check(td, PRIV_VFS_ADMIN);
4234 		if (error)
4235 			goto out;
4236 	}
4237 	if (vcount(vp) > 1)
4238 		VOP_REVOKE(vp, REVOKEALL);
4239 out:
4240 	vput(vp);
4241 	VFS_UNLOCK_GIANT(vfslocked);
4242 	return (error);
4243 }
4244 
4245 /*
4246  * Convert a user file descriptor to a kernel file entry.
4247  * A reference on the file entry is held upon returning.
4248  */
4249 int
4250 getvnode(fdp, fd, fpp)
4251 	struct filedesc *fdp;
4252 	int fd;
4253 	struct file **fpp;
4254 {
4255 	int error;
4256 	struct file *fp;
4257 
4258 	fp = NULL;
4259 	if (fdp == NULL)
4260 		error = EBADF;
4261 	else {
4262 		FILEDESC_SLOCK(fdp);
4263 		if ((u_int)fd >= fdp->fd_nfiles ||
4264 		    (fp = fdp->fd_ofiles[fd]) == NULL)
4265 			error = EBADF;
4266 		else if (fp->f_vnode == NULL) {
4267 			fp = NULL;
4268 			error = EINVAL;
4269 		} else {
4270 			fhold(fp);
4271 			error = 0;
4272 		}
4273 		FILEDESC_SUNLOCK(fdp);
4274 	}
4275 	*fpp = fp;
4276 	return (error);
4277 }
4278 
4279 /*
4280  * Get an (NFS) file handle.
4281  */
4282 #ifndef _SYS_SYSPROTO_H_
4283 struct lgetfh_args {
4284 	char	*fname;
4285 	fhandle_t *fhp;
4286 };
4287 #endif
4288 int
4289 lgetfh(td, uap)
4290 	struct thread *td;
4291 	register struct lgetfh_args *uap;
4292 {
4293 	struct nameidata nd;
4294 	fhandle_t fh;
4295 	register struct vnode *vp;
4296 	int vfslocked;
4297 	int error;
4298 
4299 	error = priv_check(td, PRIV_VFS_GETFH);
4300 	if (error)
4301 		return (error);
4302 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4303 	    UIO_USERSPACE, uap->fname, td);
4304 	error = namei(&nd);
4305 	if (error)
4306 		return (error);
4307 	vfslocked = NDHASGIANT(&nd);
4308 	NDFREE(&nd, NDF_ONLY_PNBUF);
4309 	vp = nd.ni_vp;
4310 	bzero(&fh, sizeof(fh));
4311 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4312 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4313 	vput(vp);
4314 	VFS_UNLOCK_GIANT(vfslocked);
4315 	if (error)
4316 		return (error);
4317 	error = copyout(&fh, uap->fhp, sizeof (fh));
4318 	return (error);
4319 }
4320 
4321 #ifndef _SYS_SYSPROTO_H_
4322 struct getfh_args {
4323 	char	*fname;
4324 	fhandle_t *fhp;
4325 };
4326 #endif
4327 int
4328 getfh(td, uap)
4329 	struct thread *td;
4330 	register struct getfh_args *uap;
4331 {
4332 	struct nameidata nd;
4333 	fhandle_t fh;
4334 	register struct vnode *vp;
4335 	int vfslocked;
4336 	int error;
4337 
4338 	error = priv_check(td, PRIV_VFS_GETFH);
4339 	if (error)
4340 		return (error);
4341 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4342 	    UIO_USERSPACE, uap->fname, td);
4343 	error = namei(&nd);
4344 	if (error)
4345 		return (error);
4346 	vfslocked = NDHASGIANT(&nd);
4347 	NDFREE(&nd, NDF_ONLY_PNBUF);
4348 	vp = nd.ni_vp;
4349 	bzero(&fh, sizeof(fh));
4350 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4351 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4352 	vput(vp);
4353 	VFS_UNLOCK_GIANT(vfslocked);
4354 	if (error)
4355 		return (error);
4356 	error = copyout(&fh, uap->fhp, sizeof (fh));
4357 	return (error);
4358 }
4359 
4360 /*
4361  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4362  * open descriptor.
4363  *
4364  * warning: do not remove the priv_check() call or this becomes one giant
4365  * security hole.
4366  */
4367 #ifndef _SYS_SYSPROTO_H_
4368 struct fhopen_args {
4369 	const struct fhandle *u_fhp;
4370 	int flags;
4371 };
4372 #endif
4373 int
4374 fhopen(td, uap)
4375 	struct thread *td;
4376 	struct fhopen_args /* {
4377 		const struct fhandle *u_fhp;
4378 		int flags;
4379 	} */ *uap;
4380 {
4381 	struct proc *p = td->td_proc;
4382 	struct mount *mp;
4383 	struct vnode *vp;
4384 	struct fhandle fhp;
4385 	struct vattr vat;
4386 	struct vattr *vap = &vat;
4387 	struct flock lf;
4388 	struct file *fp;
4389 	register struct filedesc *fdp = p->p_fd;
4390 	int fmode, error, type;
4391 	accmode_t accmode;
4392 	struct file *nfp;
4393 	int vfslocked;
4394 	int indx;
4395 
4396 	error = priv_check(td, PRIV_VFS_FHOPEN);
4397 	if (error)
4398 		return (error);
4399 	fmode = FFLAGS(uap->flags);
4400 	/* why not allow a non-read/write open for our lockd? */
4401 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4402 		return (EINVAL);
4403 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4404 	if (error)
4405 		return(error);
4406 	/* find the mount point */
4407 	mp = vfs_busyfs(&fhp.fh_fsid);
4408 	if (mp == NULL)
4409 		return (ESTALE);
4410 	vfslocked = VFS_LOCK_GIANT(mp);
4411 	/* now give me my vnode, it gets returned to me locked */
4412 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4413 	vfs_unbusy(mp);
4414 	if (error)
4415 		goto out;
4416 	/*
4417 	 * from now on we have to make sure not
4418 	 * to forget about the vnode
4419 	 * any error that causes an abort must vput(vp)
4420 	 * just set error = err and 'goto bad;'.
4421 	 */
4422 
4423 	/*
4424 	 * from vn_open
4425 	 */
4426 	if (vp->v_type == VLNK) {
4427 		error = EMLINK;
4428 		goto bad;
4429 	}
4430 	if (vp->v_type == VSOCK) {
4431 		error = EOPNOTSUPP;
4432 		goto bad;
4433 	}
4434 	accmode = 0;
4435 	if (fmode & (FWRITE | O_TRUNC)) {
4436 		if (vp->v_type == VDIR) {
4437 			error = EISDIR;
4438 			goto bad;
4439 		}
4440 		error = vn_writechk(vp);
4441 		if (error)
4442 			goto bad;
4443 		accmode |= VWRITE;
4444 	}
4445 	if (fmode & FREAD)
4446 		accmode |= VREAD;
4447 	if (fmode & O_APPEND)
4448 		accmode |= VAPPEND;
4449 #ifdef MAC
4450 	error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4451 	if (error)
4452 		goto bad;
4453 #endif
4454 	if (accmode) {
4455 		error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4456 		if (error)
4457 			goto bad;
4458 	}
4459 	if (fmode & O_TRUNC) {
4460 		VOP_UNLOCK(vp, 0);				/* XXX */
4461 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4462 			vrele(vp);
4463 			goto out;
4464 		}
4465 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
4466 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4467 #ifdef MAC
4468 		/*
4469 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4470 		 * should be right.
4471 		 */
4472 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4473 		if (error == 0) {
4474 #endif
4475 			VATTR_NULL(vap);
4476 			vap->va_size = 0;
4477 			error = VOP_SETATTR(vp, vap, td->td_ucred);
4478 #ifdef MAC
4479 		}
4480 #endif
4481 		vn_finished_write(mp);
4482 		if (error)
4483 			goto bad;
4484 	}
4485 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4486 	if (error)
4487 		goto bad;
4488 
4489 	if (fmode & FWRITE)
4490 		vp->v_writecount++;
4491 
4492 	/*
4493 	 * end of vn_open code
4494 	 */
4495 
4496 	if ((error = falloc(td, &nfp, &indx)) != 0) {
4497 		if (fmode & FWRITE)
4498 			vp->v_writecount--;
4499 		goto bad;
4500 	}
4501 	/* An extra reference on `nfp' has been held for us by falloc(). */
4502 	fp = nfp;
4503 	nfp->f_vnode = vp;
4504 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4505 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4506 		lf.l_whence = SEEK_SET;
4507 		lf.l_start = 0;
4508 		lf.l_len = 0;
4509 		if (fmode & O_EXLOCK)
4510 			lf.l_type = F_WRLCK;
4511 		else
4512 			lf.l_type = F_RDLCK;
4513 		type = F_FLOCK;
4514 		if ((fmode & FNONBLOCK) == 0)
4515 			type |= F_WAIT;
4516 		VOP_UNLOCK(vp, 0);
4517 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4518 			    type)) != 0) {
4519 			/*
4520 			 * The lock request failed.  Normally close the
4521 			 * descriptor but handle the case where someone might
4522 			 * have dup()d or close()d it when we weren't looking.
4523 			 */
4524 			fdclose(fdp, fp, indx, td);
4525 
4526 			/*
4527 			 * release our private reference
4528 			 */
4529 			fdrop(fp, td);
4530 			goto out;
4531 		}
4532 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4533 		atomic_set_int(&fp->f_flag, FHASLOCK);
4534 	}
4535 
4536 	VOP_UNLOCK(vp, 0);
4537 	fdrop(fp, td);
4538 	vfs_rel(mp);
4539 	VFS_UNLOCK_GIANT(vfslocked);
4540 	td->td_retval[0] = indx;
4541 	return (0);
4542 
4543 bad:
4544 	vput(vp);
4545 out:
4546 	VFS_UNLOCK_GIANT(vfslocked);
4547 	return (error);
4548 }
4549 
4550 /*
4551  * Stat an (NFS) file handle.
4552  */
4553 #ifndef _SYS_SYSPROTO_H_
4554 struct fhstat_args {
4555 	struct fhandle *u_fhp;
4556 	struct stat *sb;
4557 };
4558 #endif
4559 int
4560 fhstat(td, uap)
4561 	struct thread *td;
4562 	register struct fhstat_args /* {
4563 		struct fhandle *u_fhp;
4564 		struct stat *sb;
4565 	} */ *uap;
4566 {
4567 	struct stat sb;
4568 	fhandle_t fh;
4569 	struct mount *mp;
4570 	struct vnode *vp;
4571 	int vfslocked;
4572 	int error;
4573 
4574 	error = priv_check(td, PRIV_VFS_FHSTAT);
4575 	if (error)
4576 		return (error);
4577 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4578 	if (error)
4579 		return (error);
4580 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4581 		return (ESTALE);
4582 	vfslocked = VFS_LOCK_GIANT(mp);
4583 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4584 	vfs_unbusy(mp);
4585 	if (error) {
4586 		VFS_UNLOCK_GIANT(vfslocked);
4587 		return (error);
4588 	}
4589 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4590 	vput(vp);
4591 	VFS_UNLOCK_GIANT(vfslocked);
4592 	if (error)
4593 		return (error);
4594 	error = copyout(&sb, uap->sb, sizeof(sb));
4595 	return (error);
4596 }
4597 
4598 /*
4599  * Implement fstatfs() for (NFS) file handles.
4600  */
4601 #ifndef _SYS_SYSPROTO_H_
4602 struct fhstatfs_args {
4603 	struct fhandle *u_fhp;
4604 	struct statfs *buf;
4605 };
4606 #endif
4607 int
4608 fhstatfs(td, uap)
4609 	struct thread *td;
4610 	struct fhstatfs_args /* {
4611 		struct fhandle *u_fhp;
4612 		struct statfs *buf;
4613 	} */ *uap;
4614 {
4615 	struct statfs sf;
4616 	fhandle_t fh;
4617 	int error;
4618 
4619 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4620 	if (error)
4621 		return (error);
4622 	error = kern_fhstatfs(td, fh, &sf);
4623 	if (error)
4624 		return (error);
4625 	return (copyout(&sf, uap->buf, sizeof(sf)));
4626 }
4627 
4628 int
4629 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4630 {
4631 	struct statfs *sp;
4632 	struct mount *mp;
4633 	struct vnode *vp;
4634 	int vfslocked;
4635 	int error;
4636 
4637 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4638 	if (error)
4639 		return (error);
4640 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4641 		return (ESTALE);
4642 	vfslocked = VFS_LOCK_GIANT(mp);
4643 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4644 	if (error) {
4645 		vfs_unbusy(mp);
4646 		VFS_UNLOCK_GIANT(vfslocked);
4647 		return (error);
4648 	}
4649 	vput(vp);
4650 	error = prison_canseemount(td->td_ucred, mp);
4651 	if (error)
4652 		goto out;
4653 #ifdef MAC
4654 	error = mac_mount_check_stat(td->td_ucred, mp);
4655 	if (error)
4656 		goto out;
4657 #endif
4658 	/*
4659 	 * Set these in case the underlying filesystem fails to do so.
4660 	 */
4661 	sp = &mp->mnt_stat;
4662 	sp->f_version = STATFS_VERSION;
4663 	sp->f_namemax = NAME_MAX;
4664 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4665 	error = VFS_STATFS(mp, sp, td);
4666 	if (error == 0)
4667 		*buf = *sp;
4668 out:
4669 	vfs_unbusy(mp);
4670 	VFS_UNLOCK_GIANT(vfslocked);
4671 	return (error);
4672 }
4673