xref: /freebsd/sys/kern/vfs_syscalls.c (revision 788ca347b816afd83b2885e0c79aeeb88649b2ab)
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_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
98 static int kern_chflagsat(struct thread *td, int fd, const char *path,
99     enum uio_seg pathseg, u_long flags, int atflag);
100 static int setfflags(struct thread *td, struct vnode *, u_long);
101 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
102 static int getutimens(const struct timespec *, enum uio_seg,
103     struct timespec *, int *);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 
109 /*
110  * The module initialization routine for POSIX asynchronous I/O will
111  * set this to the version of AIO that it implements.  (Zero means
112  * that it is not implemented.)  This value is used here by pathconf()
113  * and in kern_descrip.c by fpathconf().
114  */
115 int async_io_version;
116 
117 /*
118  * Sync each mounted filesystem.
119  */
120 #ifndef _SYS_SYSPROTO_H_
121 struct sync_args {
122 	int     dummy;
123 };
124 #endif
125 /* ARGSUSED */
126 int
127 sys_sync(td, uap)
128 	struct thread *td;
129 	struct sync_args *uap;
130 {
131 	struct mount *mp, *nmp;
132 	int save;
133 
134 	mtx_lock(&mountlist_mtx);
135 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
136 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
137 			nmp = TAILQ_NEXT(mp, mnt_list);
138 			continue;
139 		}
140 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
141 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
142 			save = curthread_pflags_set(TDP_SYNCIO);
143 			vfs_msync(mp, MNT_NOWAIT);
144 			VFS_SYNC(mp, MNT_NOWAIT);
145 			curthread_pflags_restore(save);
146 			vn_finished_write(mp);
147 		}
148 		mtx_lock(&mountlist_mtx);
149 		nmp = TAILQ_NEXT(mp, mnt_list);
150 		vfs_unbusy(mp);
151 	}
152 	mtx_unlock(&mountlist_mtx);
153 	return (0);
154 }
155 
156 /*
157  * Change filesystem quotas.
158  */
159 #ifndef _SYS_SYSPROTO_H_
160 struct quotactl_args {
161 	char *path;
162 	int cmd;
163 	int uid;
164 	caddr_t arg;
165 };
166 #endif
167 int
168 sys_quotactl(td, uap)
169 	struct thread *td;
170 	register struct quotactl_args /* {
171 		char *path;
172 		int cmd;
173 		int uid;
174 		caddr_t arg;
175 	} */ *uap;
176 {
177 	struct mount *mp;
178 	struct nameidata nd;
179 	int error;
180 
181 	AUDIT_ARG_CMD(uap->cmd);
182 	AUDIT_ARG_UID(uap->uid);
183 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
184 		return (EPERM);
185 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
186 	    uap->path, td);
187 	if ((error = namei(&nd)) != 0)
188 		return (error);
189 	NDFREE(&nd, NDF_ONLY_PNBUF);
190 	mp = nd.ni_vp->v_mount;
191 	vfs_ref(mp);
192 	vput(nd.ni_vp);
193 	error = vfs_busy(mp, 0);
194 	vfs_rel(mp);
195 	if (error != 0)
196 		return (error);
197 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
198 
199 	/*
200 	 * Since quota on operation typically needs to open quota
201 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
202 	 * before calling into namei.  Otherwise, unmount might be
203 	 * started between two vfs_busy() invocations (first is our,
204 	 * second is from mount point cross-walk code in lookup()),
205 	 * causing deadlock.
206 	 *
207 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
208 	 * its own, always returning with ubusied mount point.
209 	 */
210 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
211 		vfs_unbusy(mp);
212 	return (error);
213 }
214 
215 /*
216  * Used by statfs conversion routines to scale the block size up if
217  * necessary so that all of the block counts are <= 'max_size'.  Note
218  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
219  * value of 'n'.
220  */
221 void
222 statfs_scale_blocks(struct statfs *sf, long max_size)
223 {
224 	uint64_t count;
225 	int shift;
226 
227 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
228 
229 	/*
230 	 * Attempt to scale the block counts to give a more accurate
231 	 * overview to userland of the ratio of free space to used
232 	 * space.  To do this, find the largest block count and compute
233 	 * a divisor that lets it fit into a signed integer <= max_size.
234 	 */
235 	if (sf->f_bavail < 0)
236 		count = -sf->f_bavail;
237 	else
238 		count = sf->f_bavail;
239 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
240 	if (count <= max_size)
241 		return;
242 
243 	count >>= flsl(max_size);
244 	shift = 0;
245 	while (count > 0) {
246 		shift++;
247 		count >>=1;
248 	}
249 
250 	sf->f_bsize <<= shift;
251 	sf->f_blocks >>= shift;
252 	sf->f_bfree >>= shift;
253 	sf->f_bavail >>= shift;
254 }
255 
256 /*
257  * Get filesystem statistics.
258  */
259 #ifndef _SYS_SYSPROTO_H_
260 struct statfs_args {
261 	char *path;
262 	struct statfs *buf;
263 };
264 #endif
265 int
266 sys_statfs(td, uap)
267 	struct thread *td;
268 	register struct statfs_args /* {
269 		char *path;
270 		struct statfs *buf;
271 	} */ *uap;
272 {
273 	struct statfs sf;
274 	int error;
275 
276 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
277 	if (error == 0)
278 		error = copyout(&sf, uap->buf, sizeof(sf));
279 	return (error);
280 }
281 
282 int
283 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
284     struct statfs *buf)
285 {
286 	struct mount *mp;
287 	struct statfs *sp, sb;
288 	struct nameidata nd;
289 	int error;
290 
291 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
292 	    pathseg, path, td);
293 	error = namei(&nd);
294 	if (error != 0)
295 		return (error);
296 	mp = nd.ni_vp->v_mount;
297 	vfs_ref(mp);
298 	NDFREE(&nd, NDF_ONLY_PNBUF);
299 	vput(nd.ni_vp);
300 	error = vfs_busy(mp, 0);
301 	vfs_rel(mp);
302 	if (error != 0)
303 		return (error);
304 #ifdef MAC
305 	error = mac_mount_check_stat(td->td_ucred, mp);
306 	if (error != 0)
307 		goto out;
308 #endif
309 	/*
310 	 * Set these in case the underlying filesystem fails to do so.
311 	 */
312 	sp = &mp->mnt_stat;
313 	sp->f_version = STATFS_VERSION;
314 	sp->f_namemax = NAME_MAX;
315 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
316 	error = VFS_STATFS(mp, sp);
317 	if (error != 0)
318 		goto out;
319 	if (priv_check(td, PRIV_VFS_GENERATION)) {
320 		bcopy(sp, &sb, sizeof(sb));
321 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
322 		prison_enforce_statfs(td->td_ucred, mp, &sb);
323 		sp = &sb;
324 	}
325 	*buf = *sp;
326 out:
327 	vfs_unbusy(mp);
328 	return (error);
329 }
330 
331 /*
332  * Get filesystem statistics.
333  */
334 #ifndef _SYS_SYSPROTO_H_
335 struct fstatfs_args {
336 	int fd;
337 	struct statfs *buf;
338 };
339 #endif
340 int
341 sys_fstatfs(td, uap)
342 	struct thread *td;
343 	register struct fstatfs_args /* {
344 		int fd;
345 		struct statfs *buf;
346 	} */ *uap;
347 {
348 	struct statfs sf;
349 	int error;
350 
351 	error = kern_fstatfs(td, uap->fd, &sf);
352 	if (error == 0)
353 		error = copyout(&sf, uap->buf, sizeof(sf));
354 	return (error);
355 }
356 
357 int
358 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
359 {
360 	struct file *fp;
361 	struct mount *mp;
362 	struct statfs *sp, sb;
363 	struct vnode *vp;
364 	cap_rights_t rights;
365 	int error;
366 
367 	AUDIT_ARG_FD(fd);
368 	error = getvnode(td->td_proc->p_fd, fd,
369 	    cap_rights_init(&rights, CAP_FSTATFS), &fp);
370 	if (error != 0)
371 		return (error);
372 	vp = fp->f_vnode;
373 	vn_lock(vp, LK_SHARED | LK_RETRY);
374 #ifdef AUDIT
375 	AUDIT_ARG_VNODE1(vp);
376 #endif
377 	mp = vp->v_mount;
378 	if (mp)
379 		vfs_ref(mp);
380 	VOP_UNLOCK(vp, 0);
381 	fdrop(fp, td);
382 	if (mp == NULL) {
383 		error = EBADF;
384 		goto out;
385 	}
386 	error = vfs_busy(mp, 0);
387 	vfs_rel(mp);
388 	if (error != 0)
389 		return (error);
390 #ifdef MAC
391 	error = mac_mount_check_stat(td->td_ucred, mp);
392 	if (error != 0)
393 		goto out;
394 #endif
395 	/*
396 	 * Set these in case the underlying filesystem fails to do so.
397 	 */
398 	sp = &mp->mnt_stat;
399 	sp->f_version = STATFS_VERSION;
400 	sp->f_namemax = NAME_MAX;
401 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
402 	error = VFS_STATFS(mp, sp);
403 	if (error != 0)
404 		goto out;
405 	if (priv_check(td, PRIV_VFS_GENERATION)) {
406 		bcopy(sp, &sb, sizeof(sb));
407 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
408 		prison_enforce_statfs(td->td_ucred, mp, &sb);
409 		sp = &sb;
410 	}
411 	*buf = *sp;
412 out:
413 	if (mp)
414 		vfs_unbusy(mp);
415 	return (error);
416 }
417 
418 /*
419  * Get statistics on all filesystems.
420  */
421 #ifndef _SYS_SYSPROTO_H_
422 struct getfsstat_args {
423 	struct statfs *buf;
424 	long bufsize;
425 	int flags;
426 };
427 #endif
428 int
429 sys_getfsstat(td, uap)
430 	struct thread *td;
431 	register struct getfsstat_args /* {
432 		struct statfs *buf;
433 		long bufsize;
434 		int flags;
435 	} */ *uap;
436 {
437 	size_t count;
438 	int error;
439 
440 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
441 	    UIO_USERSPACE, uap->flags);
442 	if (error == 0)
443 		td->td_retval[0] = count;
444 	return (error);
445 }
446 
447 /*
448  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
449  *	The caller is responsible for freeing memory which will be allocated
450  *	in '*buf'.
451  */
452 int
453 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
454     size_t *countp, enum uio_seg bufseg, int flags)
455 {
456 	struct mount *mp, *nmp;
457 	struct statfs *sfsp, *sp, sb;
458 	size_t count, maxcount;
459 	int error;
460 
461 	maxcount = bufsize / sizeof(struct statfs);
462 	if (bufsize == 0)
463 		sfsp = NULL;
464 	else if (bufseg == UIO_USERSPACE)
465 		sfsp = *buf;
466 	else /* if (bufseg == UIO_SYSSPACE) */ {
467 		count = 0;
468 		mtx_lock(&mountlist_mtx);
469 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
470 			count++;
471 		}
472 		mtx_unlock(&mountlist_mtx);
473 		if (maxcount > count)
474 			maxcount = count;
475 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
476 		    M_WAITOK);
477 	}
478 	count = 0;
479 	mtx_lock(&mountlist_mtx);
480 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
481 		if (prison_canseemount(td->td_ucred, mp) != 0) {
482 			nmp = TAILQ_NEXT(mp, mnt_list);
483 			continue;
484 		}
485 #ifdef MAC
486 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
487 			nmp = TAILQ_NEXT(mp, mnt_list);
488 			continue;
489 		}
490 #endif
491 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
492 			nmp = TAILQ_NEXT(mp, mnt_list);
493 			continue;
494 		}
495 		if (sfsp && count < maxcount) {
496 			sp = &mp->mnt_stat;
497 			/*
498 			 * Set these in case the underlying filesystem
499 			 * fails to do so.
500 			 */
501 			sp->f_version = STATFS_VERSION;
502 			sp->f_namemax = NAME_MAX;
503 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
504 			/*
505 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
506 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
507 			 * overrides MNT_WAIT.
508 			 */
509 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
510 			    (flags & MNT_WAIT)) &&
511 			    (error = VFS_STATFS(mp, sp))) {
512 				mtx_lock(&mountlist_mtx);
513 				nmp = TAILQ_NEXT(mp, mnt_list);
514 				vfs_unbusy(mp);
515 				continue;
516 			}
517 			if (priv_check(td, PRIV_VFS_GENERATION)) {
518 				bcopy(sp, &sb, sizeof(sb));
519 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
520 				prison_enforce_statfs(td->td_ucred, mp, &sb);
521 				sp = &sb;
522 			}
523 			if (bufseg == UIO_SYSSPACE)
524 				bcopy(sp, sfsp, sizeof(*sp));
525 			else /* if (bufseg == UIO_USERSPACE) */ {
526 				error = copyout(sp, sfsp, sizeof(*sp));
527 				if (error != 0) {
528 					vfs_unbusy(mp);
529 					return (error);
530 				}
531 			}
532 			sfsp++;
533 		}
534 		count++;
535 		mtx_lock(&mountlist_mtx);
536 		nmp = TAILQ_NEXT(mp, mnt_list);
537 		vfs_unbusy(mp);
538 	}
539 	mtx_unlock(&mountlist_mtx);
540 	if (sfsp && count > maxcount)
541 		*countp = maxcount;
542 	else
543 		*countp = count;
544 	return (0);
545 }
546 
547 #ifdef COMPAT_FREEBSD4
548 /*
549  * Get old format filesystem statistics.
550  */
551 static void cvtstatfs(struct statfs *, struct ostatfs *);
552 
553 #ifndef _SYS_SYSPROTO_H_
554 struct freebsd4_statfs_args {
555 	char *path;
556 	struct ostatfs *buf;
557 };
558 #endif
559 int
560 freebsd4_statfs(td, uap)
561 	struct thread *td;
562 	struct freebsd4_statfs_args /* {
563 		char *path;
564 		struct ostatfs *buf;
565 	} */ *uap;
566 {
567 	struct ostatfs osb;
568 	struct statfs sf;
569 	int error;
570 
571 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
572 	if (error != 0)
573 		return (error);
574 	cvtstatfs(&sf, &osb);
575 	return (copyout(&osb, uap->buf, sizeof(osb)));
576 }
577 
578 /*
579  * Get filesystem statistics.
580  */
581 #ifndef _SYS_SYSPROTO_H_
582 struct freebsd4_fstatfs_args {
583 	int fd;
584 	struct ostatfs *buf;
585 };
586 #endif
587 int
588 freebsd4_fstatfs(td, uap)
589 	struct thread *td;
590 	struct freebsd4_fstatfs_args /* {
591 		int fd;
592 		struct ostatfs *buf;
593 	} */ *uap;
594 {
595 	struct ostatfs osb;
596 	struct statfs sf;
597 	int error;
598 
599 	error = kern_fstatfs(td, uap->fd, &sf);
600 	if (error != 0)
601 		return (error);
602 	cvtstatfs(&sf, &osb);
603 	return (copyout(&osb, uap->buf, sizeof(osb)));
604 }
605 
606 /*
607  * Get statistics on all filesystems.
608  */
609 #ifndef _SYS_SYSPROTO_H_
610 struct freebsd4_getfsstat_args {
611 	struct ostatfs *buf;
612 	long bufsize;
613 	int flags;
614 };
615 #endif
616 int
617 freebsd4_getfsstat(td, uap)
618 	struct thread *td;
619 	register struct freebsd4_getfsstat_args /* {
620 		struct ostatfs *buf;
621 		long bufsize;
622 		int flags;
623 	} */ *uap;
624 {
625 	struct statfs *buf, *sp;
626 	struct ostatfs osb;
627 	size_t count, size;
628 	int error;
629 
630 	count = uap->bufsize / sizeof(struct ostatfs);
631 	size = count * sizeof(struct statfs);
632 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
633 	    uap->flags);
634 	if (size > 0) {
635 		sp = buf;
636 		while (count > 0 && error == 0) {
637 			cvtstatfs(sp, &osb);
638 			error = copyout(&osb, uap->buf, sizeof(osb));
639 			sp++;
640 			uap->buf++;
641 			count--;
642 		}
643 		free(buf, M_TEMP);
644 	}
645 	if (error == 0)
646 		td->td_retval[0] = count;
647 	return (error);
648 }
649 
650 /*
651  * Implement fstatfs() for (NFS) file handles.
652  */
653 #ifndef _SYS_SYSPROTO_H_
654 struct freebsd4_fhstatfs_args {
655 	struct fhandle *u_fhp;
656 	struct ostatfs *buf;
657 };
658 #endif
659 int
660 freebsd4_fhstatfs(td, uap)
661 	struct thread *td;
662 	struct freebsd4_fhstatfs_args /* {
663 		struct fhandle *u_fhp;
664 		struct ostatfs *buf;
665 	} */ *uap;
666 {
667 	struct ostatfs osb;
668 	struct statfs sf;
669 	fhandle_t fh;
670 	int error;
671 
672 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
673 	if (error != 0)
674 		return (error);
675 	error = kern_fhstatfs(td, fh, &sf);
676 	if (error != 0)
677 		return (error);
678 	cvtstatfs(&sf, &osb);
679 	return (copyout(&osb, uap->buf, sizeof(osb)));
680 }
681 
682 /*
683  * Convert a new format statfs structure to an old format statfs structure.
684  */
685 static void
686 cvtstatfs(nsp, osp)
687 	struct statfs *nsp;
688 	struct ostatfs *osp;
689 {
690 
691 	statfs_scale_blocks(nsp, LONG_MAX);
692 	bzero(osp, sizeof(*osp));
693 	osp->f_bsize = nsp->f_bsize;
694 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
695 	osp->f_blocks = nsp->f_blocks;
696 	osp->f_bfree = nsp->f_bfree;
697 	osp->f_bavail = nsp->f_bavail;
698 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
699 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
700 	osp->f_owner = nsp->f_owner;
701 	osp->f_type = nsp->f_type;
702 	osp->f_flags = nsp->f_flags;
703 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
704 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
705 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
706 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
707 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
708 	    MIN(MFSNAMELEN, OMFSNAMELEN));
709 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
710 	    MIN(MNAMELEN, OMNAMELEN));
711 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
712 	    MIN(MNAMELEN, OMNAMELEN));
713 	osp->f_fsid = nsp->f_fsid;
714 }
715 #endif /* COMPAT_FREEBSD4 */
716 
717 /*
718  * Change current working directory to a given file descriptor.
719  */
720 #ifndef _SYS_SYSPROTO_H_
721 struct fchdir_args {
722 	int	fd;
723 };
724 #endif
725 int
726 sys_fchdir(td, uap)
727 	struct thread *td;
728 	struct fchdir_args /* {
729 		int fd;
730 	} */ *uap;
731 {
732 	register struct filedesc *fdp = td->td_proc->p_fd;
733 	struct vnode *vp, *tdp, *vpold;
734 	struct mount *mp;
735 	struct file *fp;
736 	cap_rights_t rights;
737 	int error;
738 
739 	AUDIT_ARG_FD(uap->fd);
740 	error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
741 	    &fp);
742 	if (error != 0)
743 		return (error);
744 	vp = fp->f_vnode;
745 	VREF(vp);
746 	fdrop(fp, td);
747 	vn_lock(vp, LK_SHARED | LK_RETRY);
748 	AUDIT_ARG_VNODE1(vp);
749 	error = change_dir(vp, td);
750 	while (!error && (mp = vp->v_mountedhere) != NULL) {
751 		if (vfs_busy(mp, 0))
752 			continue;
753 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
754 		vfs_unbusy(mp);
755 		if (error != 0)
756 			break;
757 		vput(vp);
758 		vp = tdp;
759 	}
760 	if (error != 0) {
761 		vput(vp);
762 		return (error);
763 	}
764 	VOP_UNLOCK(vp, 0);
765 	FILEDESC_XLOCK(fdp);
766 	vpold = fdp->fd_cdir;
767 	fdp->fd_cdir = vp;
768 	FILEDESC_XUNLOCK(fdp);
769 	vrele(vpold);
770 	return (0);
771 }
772 
773 /*
774  * Change current working directory (``.'').
775  */
776 #ifndef _SYS_SYSPROTO_H_
777 struct chdir_args {
778 	char	*path;
779 };
780 #endif
781 int
782 sys_chdir(td, uap)
783 	struct thread *td;
784 	struct chdir_args /* {
785 		char *path;
786 	} */ *uap;
787 {
788 
789 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
790 }
791 
792 int
793 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
794 {
795 	register struct filedesc *fdp = td->td_proc->p_fd;
796 	struct nameidata nd;
797 	struct vnode *vp;
798 	int error;
799 
800 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
801 	    pathseg, path, td);
802 	if ((error = namei(&nd)) != 0)
803 		return (error);
804 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
805 		vput(nd.ni_vp);
806 		NDFREE(&nd, NDF_ONLY_PNBUF);
807 		return (error);
808 	}
809 	VOP_UNLOCK(nd.ni_vp, 0);
810 	NDFREE(&nd, NDF_ONLY_PNBUF);
811 	FILEDESC_XLOCK(fdp);
812 	vp = fdp->fd_cdir;
813 	fdp->fd_cdir = nd.ni_vp;
814 	FILEDESC_XUNLOCK(fdp);
815 	vrele(vp);
816 	return (0);
817 }
818 
819 /*
820  * Helper function for raised chroot(2) security function:  Refuse if
821  * any filedescriptors are open directories.
822  */
823 static int
824 chroot_refuse_vdir_fds(fdp)
825 	struct filedesc *fdp;
826 {
827 	struct vnode *vp;
828 	struct file *fp;
829 	int fd;
830 
831 	FILEDESC_LOCK_ASSERT(fdp);
832 
833 	for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
834 		fp = fget_locked(fdp, fd);
835 		if (fp == NULL)
836 			continue;
837 		if (fp->f_type == DTYPE_VNODE) {
838 			vp = fp->f_vnode;
839 			if (vp->v_type == VDIR)
840 				return (EPERM);
841 		}
842 	}
843 	return (0);
844 }
845 
846 /*
847  * This sysctl determines if we will allow a process to chroot(2) if it
848  * has a directory open:
849  *	0: disallowed for all processes.
850  *	1: allowed for processes that were not already chroot(2)'ed.
851  *	2: allowed for all processes.
852  */
853 
854 static int chroot_allow_open_directories = 1;
855 
856 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
857      &chroot_allow_open_directories, 0,
858      "Allow a process to chroot(2) if it has a directory open");
859 
860 /*
861  * Change notion of root (``/'') directory.
862  */
863 #ifndef _SYS_SYSPROTO_H_
864 struct chroot_args {
865 	char	*path;
866 };
867 #endif
868 int
869 sys_chroot(td, uap)
870 	struct thread *td;
871 	struct chroot_args /* {
872 		char *path;
873 	} */ *uap;
874 {
875 	struct nameidata nd;
876 	int error;
877 
878 	error = priv_check(td, PRIV_VFS_CHROOT);
879 	if (error != 0)
880 		return (error);
881 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
882 	    UIO_USERSPACE, uap->path, td);
883 	error = namei(&nd);
884 	if (error != 0)
885 		goto error;
886 	error = change_dir(nd.ni_vp, td);
887 	if (error != 0)
888 		goto e_vunlock;
889 #ifdef MAC
890 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
891 	if (error != 0)
892 		goto e_vunlock;
893 #endif
894 	VOP_UNLOCK(nd.ni_vp, 0);
895 	error = change_root(nd.ni_vp, td);
896 	vrele(nd.ni_vp);
897 	NDFREE(&nd, NDF_ONLY_PNBUF);
898 	return (error);
899 e_vunlock:
900 	vput(nd.ni_vp);
901 error:
902 	NDFREE(&nd, NDF_ONLY_PNBUF);
903 	return (error);
904 }
905 
906 /*
907  * Common routine for chroot and chdir.  Callers must provide a locked vnode
908  * instance.
909  */
910 int
911 change_dir(vp, td)
912 	struct vnode *vp;
913 	struct thread *td;
914 {
915 #ifdef MAC
916 	int error;
917 #endif
918 
919 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
920 	if (vp->v_type != VDIR)
921 		return (ENOTDIR);
922 #ifdef MAC
923 	error = mac_vnode_check_chdir(td->td_ucred, vp);
924 	if (error != 0)
925 		return (error);
926 #endif
927 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
928 }
929 
930 /*
931  * Common routine for kern_chroot() and jail_attach().  The caller is
932  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
933  * authorize this operation.
934  */
935 int
936 change_root(vp, td)
937 	struct vnode *vp;
938 	struct thread *td;
939 {
940 	struct filedesc *fdp;
941 	struct vnode *oldvp;
942 	int error;
943 
944 	fdp = td->td_proc->p_fd;
945 	FILEDESC_XLOCK(fdp);
946 	if (chroot_allow_open_directories == 0 ||
947 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
948 		error = chroot_refuse_vdir_fds(fdp);
949 		if (error != 0) {
950 			FILEDESC_XUNLOCK(fdp);
951 			return (error);
952 		}
953 	}
954 	oldvp = fdp->fd_rdir;
955 	fdp->fd_rdir = vp;
956 	VREF(fdp->fd_rdir);
957 	if (!fdp->fd_jdir) {
958 		fdp->fd_jdir = vp;
959 		VREF(fdp->fd_jdir);
960 	}
961 	FILEDESC_XUNLOCK(fdp);
962 	vrele(oldvp);
963 	return (0);
964 }
965 
966 static __inline void
967 flags_to_rights(int flags, cap_rights_t *rightsp)
968 {
969 
970 	if (flags & O_EXEC) {
971 		cap_rights_set(rightsp, CAP_FEXECVE);
972 	} else {
973 		switch ((flags & O_ACCMODE)) {
974 		case O_RDONLY:
975 			cap_rights_set(rightsp, CAP_READ);
976 			break;
977 		case O_RDWR:
978 			cap_rights_set(rightsp, CAP_READ);
979 			/* FALLTHROUGH */
980 		case O_WRONLY:
981 			cap_rights_set(rightsp, CAP_WRITE);
982 			if (!(flags & (O_APPEND | O_TRUNC)))
983 				cap_rights_set(rightsp, CAP_SEEK);
984 			break;
985 		}
986 	}
987 
988 	if (flags & O_CREAT)
989 		cap_rights_set(rightsp, CAP_CREATE);
990 
991 	if (flags & O_TRUNC)
992 		cap_rights_set(rightsp, CAP_FTRUNCATE);
993 
994 	if (flags & (O_SYNC | O_FSYNC))
995 		cap_rights_set(rightsp, CAP_FSYNC);
996 
997 	if (flags & (O_EXLOCK | O_SHLOCK))
998 		cap_rights_set(rightsp, CAP_FLOCK);
999 }
1000 
1001 /*
1002  * Check permissions, allocate an open file structure, and call the device
1003  * open routine if any.
1004  */
1005 #ifndef _SYS_SYSPROTO_H_
1006 struct open_args {
1007 	char	*path;
1008 	int	flags;
1009 	int	mode;
1010 };
1011 #endif
1012 int
1013 sys_open(td, uap)
1014 	struct thread *td;
1015 	register struct open_args /* {
1016 		char *path;
1017 		int flags;
1018 		int mode;
1019 	} */ *uap;
1020 {
1021 
1022 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1023 	    uap->flags, uap->mode));
1024 }
1025 
1026 #ifndef _SYS_SYSPROTO_H_
1027 struct openat_args {
1028 	int	fd;
1029 	char	*path;
1030 	int	flag;
1031 	int	mode;
1032 };
1033 #endif
1034 int
1035 sys_openat(struct thread *td, struct openat_args *uap)
1036 {
1037 
1038 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1039 	    uap->mode));
1040 }
1041 
1042 int
1043 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1044     int flags, int mode)
1045 {
1046 	struct proc *p = td->td_proc;
1047 	struct filedesc *fdp = p->p_fd;
1048 	struct file *fp;
1049 	struct vnode *vp;
1050 	struct nameidata nd;
1051 	cap_rights_t rights;
1052 	int cmode, error, indx;
1053 
1054 	indx = -1;
1055 
1056 	AUDIT_ARG_FFLAGS(flags);
1057 	AUDIT_ARG_MODE(mode);
1058 	/* XXX: audit dirfd */
1059 	cap_rights_init(&rights, CAP_LOOKUP);
1060 	flags_to_rights(flags, &rights);
1061 	/*
1062 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1063 	 * may be specified.
1064 	 */
1065 	if (flags & O_EXEC) {
1066 		if (flags & O_ACCMODE)
1067 			return (EINVAL);
1068 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1069 		return (EINVAL);
1070 	} else {
1071 		flags = FFLAGS(flags);
1072 	}
1073 
1074 	/*
1075 	 * Allocate the file descriptor, but don't install a descriptor yet.
1076 	 */
1077 	error = falloc_noinstall(td, &fp);
1078 	if (error != 0)
1079 		return (error);
1080 	/*
1081 	 * An extra reference on `fp' has been held for us by
1082 	 * falloc_noinstall().
1083 	 */
1084 	/* Set the flags early so the finit in devfs can pick them up. */
1085 	fp->f_flag = flags & FMASK;
1086 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1087 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1088 	    &rights, td);
1089 	td->td_dupfd = -1;		/* XXX check for fdopen */
1090 	error = vn_open(&nd, &flags, cmode, fp);
1091 	if (error != 0) {
1092 		/*
1093 		 * If the vn_open replaced the method vector, something
1094 		 * wonderous happened deep below and we just pass it up
1095 		 * pretending we know what we do.
1096 		 */
1097 		if (error == ENXIO && fp->f_ops != &badfileops)
1098 			goto success;
1099 
1100 		/*
1101 		 * Handle special fdopen() case. bleh.
1102 		 *
1103 		 * Don't do this for relative (capability) lookups; we don't
1104 		 * understand exactly what would happen, and we don't think
1105 		 * that it ever should.
1106 		 */
1107 		if (nd.ni_strictrelative == 0 &&
1108 		    (error == ENODEV || error == ENXIO) &&
1109 		    td->td_dupfd >= 0) {
1110 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1111 			    &indx);
1112 			if (error == 0)
1113 				goto success;
1114 		}
1115 
1116 		goto bad;
1117 	}
1118 	td->td_dupfd = 0;
1119 	NDFREE(&nd, NDF_ONLY_PNBUF);
1120 	vp = nd.ni_vp;
1121 
1122 	/*
1123 	 * Store the vnode, for any f_type. Typically, the vnode use
1124 	 * count is decremented by direct call to vn_closefile() for
1125 	 * files that switched type in the cdevsw fdopen() method.
1126 	 */
1127 	fp->f_vnode = vp;
1128 	/*
1129 	 * If the file wasn't claimed by devfs bind it to the normal
1130 	 * vnode operations here.
1131 	 */
1132 	if (fp->f_ops == &badfileops) {
1133 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1134 		fp->f_seqcount = 1;
1135 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1136 		    DTYPE_VNODE, vp, &vnops);
1137 	}
1138 
1139 	VOP_UNLOCK(vp, 0);
1140 	if (flags & O_TRUNC) {
1141 		error = fo_truncate(fp, 0, td->td_ucred, td);
1142 		if (error != 0)
1143 			goto bad;
1144 	}
1145 success:
1146 	/*
1147 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1148 	 */
1149 	if (indx == -1) {
1150 		struct filecaps *fcaps;
1151 
1152 #ifdef CAPABILITIES
1153 		if (nd.ni_strictrelative == 1)
1154 			fcaps = &nd.ni_filecaps;
1155 		else
1156 #endif
1157 			fcaps = NULL;
1158 		error = finstall(td, fp, &indx, flags, fcaps);
1159 		/* On success finstall() consumes fcaps. */
1160 		if (error != 0) {
1161 			filecaps_free(&nd.ni_filecaps);
1162 			goto bad;
1163 		}
1164 	} else {
1165 		filecaps_free(&nd.ni_filecaps);
1166 	}
1167 
1168 	/*
1169 	 * Release our private reference, leaving the one associated with
1170 	 * the descriptor table intact.
1171 	 */
1172 	fdrop(fp, td);
1173 	td->td_retval[0] = indx;
1174 	return (0);
1175 bad:
1176 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1177 	fdrop(fp, td);
1178 	return (error);
1179 }
1180 
1181 #ifdef COMPAT_43
1182 /*
1183  * Create a file.
1184  */
1185 #ifndef _SYS_SYSPROTO_H_
1186 struct ocreat_args {
1187 	char	*path;
1188 	int	mode;
1189 };
1190 #endif
1191 int
1192 ocreat(td, uap)
1193 	struct thread *td;
1194 	register struct ocreat_args /* {
1195 		char *path;
1196 		int mode;
1197 	} */ *uap;
1198 {
1199 
1200 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1201 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1202 }
1203 #endif /* COMPAT_43 */
1204 
1205 /*
1206  * Create a special file.
1207  */
1208 #ifndef _SYS_SYSPROTO_H_
1209 struct mknod_args {
1210 	char	*path;
1211 	int	mode;
1212 	int	dev;
1213 };
1214 #endif
1215 int
1216 sys_mknod(td, uap)
1217 	struct thread *td;
1218 	register struct mknod_args /* {
1219 		char *path;
1220 		int mode;
1221 		int dev;
1222 	} */ *uap;
1223 {
1224 
1225 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1226 	    uap->mode, uap->dev));
1227 }
1228 
1229 #ifndef _SYS_SYSPROTO_H_
1230 struct mknodat_args {
1231 	int	fd;
1232 	char	*path;
1233 	mode_t	mode;
1234 	dev_t	dev;
1235 };
1236 #endif
1237 int
1238 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1239 {
1240 
1241 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1242 	    uap->dev));
1243 }
1244 
1245 int
1246 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1247     int mode, int dev)
1248 {
1249 	struct vnode *vp;
1250 	struct mount *mp;
1251 	struct vattr vattr;
1252 	struct nameidata nd;
1253 	cap_rights_t rights;
1254 	int error, whiteout = 0;
1255 
1256 	AUDIT_ARG_MODE(mode);
1257 	AUDIT_ARG_DEV(dev);
1258 	switch (mode & S_IFMT) {
1259 	case S_IFCHR:
1260 	case S_IFBLK:
1261 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1262 		break;
1263 	case S_IFMT:
1264 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1265 		break;
1266 	case S_IFWHT:
1267 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1268 		break;
1269 	case S_IFIFO:
1270 		if (dev == 0)
1271 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1272 		/* FALLTHROUGH */
1273 	default:
1274 		error = EINVAL;
1275 		break;
1276 	}
1277 	if (error != 0)
1278 		return (error);
1279 restart:
1280 	bwillwrite();
1281 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1282 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1283 	    td);
1284 	if ((error = namei(&nd)) != 0)
1285 		return (error);
1286 	vp = nd.ni_vp;
1287 	if (vp != NULL) {
1288 		NDFREE(&nd, NDF_ONLY_PNBUF);
1289 		if (vp == nd.ni_dvp)
1290 			vrele(nd.ni_dvp);
1291 		else
1292 			vput(nd.ni_dvp);
1293 		vrele(vp);
1294 		return (EEXIST);
1295 	} else {
1296 		VATTR_NULL(&vattr);
1297 		vattr.va_mode = (mode & ALLPERMS) &
1298 		    ~td->td_proc->p_fd->fd_cmask;
1299 		vattr.va_rdev = dev;
1300 		whiteout = 0;
1301 
1302 		switch (mode & S_IFMT) {
1303 		case S_IFMT:	/* used by badsect to flag bad sectors */
1304 			vattr.va_type = VBAD;
1305 			break;
1306 		case S_IFCHR:
1307 			vattr.va_type = VCHR;
1308 			break;
1309 		case S_IFBLK:
1310 			vattr.va_type = VBLK;
1311 			break;
1312 		case S_IFWHT:
1313 			whiteout = 1;
1314 			break;
1315 		default:
1316 			panic("kern_mknod: invalid mode");
1317 		}
1318 	}
1319 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1320 		NDFREE(&nd, NDF_ONLY_PNBUF);
1321 		vput(nd.ni_dvp);
1322 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1323 			return (error);
1324 		goto restart;
1325 	}
1326 #ifdef MAC
1327 	if (error == 0 && !whiteout)
1328 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1329 		    &nd.ni_cnd, &vattr);
1330 #endif
1331 	if (error == 0) {
1332 		if (whiteout)
1333 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1334 		else {
1335 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1336 						&nd.ni_cnd, &vattr);
1337 			if (error == 0)
1338 				vput(nd.ni_vp);
1339 		}
1340 	}
1341 	NDFREE(&nd, NDF_ONLY_PNBUF);
1342 	vput(nd.ni_dvp);
1343 	vn_finished_write(mp);
1344 	return (error);
1345 }
1346 
1347 /*
1348  * Create a named pipe.
1349  */
1350 #ifndef _SYS_SYSPROTO_H_
1351 struct mkfifo_args {
1352 	char	*path;
1353 	int	mode;
1354 };
1355 #endif
1356 int
1357 sys_mkfifo(td, uap)
1358 	struct thread *td;
1359 	register struct mkfifo_args /* {
1360 		char *path;
1361 		int mode;
1362 	} */ *uap;
1363 {
1364 
1365 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1366 	    uap->mode));
1367 }
1368 
1369 #ifndef _SYS_SYSPROTO_H_
1370 struct mkfifoat_args {
1371 	int	fd;
1372 	char	*path;
1373 	mode_t	mode;
1374 };
1375 #endif
1376 int
1377 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1378 {
1379 
1380 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1381 	    uap->mode));
1382 }
1383 
1384 int
1385 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1386     int mode)
1387 {
1388 	struct mount *mp;
1389 	struct vattr vattr;
1390 	struct nameidata nd;
1391 	cap_rights_t rights;
1392 	int error;
1393 
1394 	AUDIT_ARG_MODE(mode);
1395 restart:
1396 	bwillwrite();
1397 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1398 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1399 	    td);
1400 	if ((error = namei(&nd)) != 0)
1401 		return (error);
1402 	if (nd.ni_vp != NULL) {
1403 		NDFREE(&nd, NDF_ONLY_PNBUF);
1404 		if (nd.ni_vp == nd.ni_dvp)
1405 			vrele(nd.ni_dvp);
1406 		else
1407 			vput(nd.ni_dvp);
1408 		vrele(nd.ni_vp);
1409 		return (EEXIST);
1410 	}
1411 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1412 		NDFREE(&nd, NDF_ONLY_PNBUF);
1413 		vput(nd.ni_dvp);
1414 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1415 			return (error);
1416 		goto restart;
1417 	}
1418 	VATTR_NULL(&vattr);
1419 	vattr.va_type = VFIFO;
1420 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1421 #ifdef MAC
1422 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1423 	    &vattr);
1424 	if (error != 0)
1425 		goto out;
1426 #endif
1427 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1428 	if (error == 0)
1429 		vput(nd.ni_vp);
1430 #ifdef MAC
1431 out:
1432 #endif
1433 	vput(nd.ni_dvp);
1434 	vn_finished_write(mp);
1435 	NDFREE(&nd, NDF_ONLY_PNBUF);
1436 	return (error);
1437 }
1438 
1439 /*
1440  * Make a hard file link.
1441  */
1442 #ifndef _SYS_SYSPROTO_H_
1443 struct link_args {
1444 	char	*path;
1445 	char	*link;
1446 };
1447 #endif
1448 int
1449 sys_link(td, uap)
1450 	struct thread *td;
1451 	register struct link_args /* {
1452 		char *path;
1453 		char *link;
1454 	} */ *uap;
1455 {
1456 
1457 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1458 	    UIO_USERSPACE, FOLLOW));
1459 }
1460 
1461 #ifndef _SYS_SYSPROTO_H_
1462 struct linkat_args {
1463 	int	fd1;
1464 	char	*path1;
1465 	int	fd2;
1466 	char	*path2;
1467 	int	flag;
1468 };
1469 #endif
1470 int
1471 sys_linkat(struct thread *td, struct linkat_args *uap)
1472 {
1473 	int flag;
1474 
1475 	flag = uap->flag;
1476 	if (flag & ~AT_SYMLINK_FOLLOW)
1477 		return (EINVAL);
1478 
1479 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1480 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1481 }
1482 
1483 int hardlink_check_uid = 0;
1484 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1485     &hardlink_check_uid, 0,
1486     "Unprivileged processes cannot create hard links to files owned by other "
1487     "users");
1488 static int hardlink_check_gid = 0;
1489 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1490     &hardlink_check_gid, 0,
1491     "Unprivileged processes cannot create hard links to files owned by other "
1492     "groups");
1493 
1494 static int
1495 can_hardlink(struct vnode *vp, struct ucred *cred)
1496 {
1497 	struct vattr va;
1498 	int error;
1499 
1500 	if (!hardlink_check_uid && !hardlink_check_gid)
1501 		return (0);
1502 
1503 	error = VOP_GETATTR(vp, &va, cred);
1504 	if (error != 0)
1505 		return (error);
1506 
1507 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1508 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1509 		if (error != 0)
1510 			return (error);
1511 	}
1512 
1513 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1514 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1515 		if (error != 0)
1516 			return (error);
1517 	}
1518 
1519 	return (0);
1520 }
1521 
1522 int
1523 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1524     enum uio_seg segflg, int follow)
1525 {
1526 	struct vnode *vp;
1527 	struct mount *mp;
1528 	struct nameidata nd;
1529 	cap_rights_t rights;
1530 	int error;
1531 
1532 again:
1533 	bwillwrite();
1534 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1535 
1536 	if ((error = namei(&nd)) != 0)
1537 		return (error);
1538 	NDFREE(&nd, NDF_ONLY_PNBUF);
1539 	vp = nd.ni_vp;
1540 	if (vp->v_type == VDIR) {
1541 		vrele(vp);
1542 		return (EPERM);		/* POSIX */
1543 	}
1544 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2 |
1545 	    NOCACHE, segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT),
1546 	    td);
1547 	if ((error = namei(&nd)) == 0) {
1548 		if (nd.ni_vp != NULL) {
1549 			NDFREE(&nd, NDF_ONLY_PNBUF);
1550 			if (nd.ni_dvp == nd.ni_vp)
1551 				vrele(nd.ni_dvp);
1552 			else
1553 				vput(nd.ni_dvp);
1554 			vrele(nd.ni_vp);
1555 			vrele(vp);
1556 			return (EEXIST);
1557 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1558 			/*
1559 			 * Cross-device link.  No need to recheck
1560 			 * vp->v_type, since it cannot change, except
1561 			 * to VBAD.
1562 			 */
1563 			NDFREE(&nd, NDF_ONLY_PNBUF);
1564 			vput(nd.ni_dvp);
1565 			vrele(vp);
1566 			return (EXDEV);
1567 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1568 			error = can_hardlink(vp, td->td_ucred);
1569 #ifdef MAC
1570 			if (error == 0)
1571 				error = mac_vnode_check_link(td->td_ucred,
1572 				    nd.ni_dvp, vp, &nd.ni_cnd);
1573 #endif
1574 			if (error != 0) {
1575 				vput(vp);
1576 				vput(nd.ni_dvp);
1577 				NDFREE(&nd, NDF_ONLY_PNBUF);
1578 				return (error);
1579 			}
1580 			error = vn_start_write(vp, &mp, V_NOWAIT);
1581 			if (error != 0) {
1582 				vput(vp);
1583 				vput(nd.ni_dvp);
1584 				NDFREE(&nd, NDF_ONLY_PNBUF);
1585 				error = vn_start_write(NULL, &mp,
1586 				    V_XSLEEP | PCATCH);
1587 				if (error != 0)
1588 					return (error);
1589 				goto again;
1590 			}
1591 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1592 			VOP_UNLOCK(vp, 0);
1593 			vput(nd.ni_dvp);
1594 			vn_finished_write(mp);
1595 			NDFREE(&nd, NDF_ONLY_PNBUF);
1596 		} else {
1597 			vput(nd.ni_dvp);
1598 			NDFREE(&nd, NDF_ONLY_PNBUF);
1599 			vrele(vp);
1600 			goto again;
1601 		}
1602 	}
1603 	vrele(vp);
1604 	return (error);
1605 }
1606 
1607 /*
1608  * Make a symbolic link.
1609  */
1610 #ifndef _SYS_SYSPROTO_H_
1611 struct symlink_args {
1612 	char	*path;
1613 	char	*link;
1614 };
1615 #endif
1616 int
1617 sys_symlink(td, uap)
1618 	struct thread *td;
1619 	register struct symlink_args /* {
1620 		char *path;
1621 		char *link;
1622 	} */ *uap;
1623 {
1624 
1625 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1626 	    UIO_USERSPACE));
1627 }
1628 
1629 #ifndef _SYS_SYSPROTO_H_
1630 struct symlinkat_args {
1631 	char	*path;
1632 	int	fd;
1633 	char	*path2;
1634 };
1635 #endif
1636 int
1637 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1638 {
1639 
1640 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1641 	    UIO_USERSPACE));
1642 }
1643 
1644 int
1645 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1646     enum uio_seg segflg)
1647 {
1648 	struct mount *mp;
1649 	struct vattr vattr;
1650 	char *syspath;
1651 	struct nameidata nd;
1652 	int error;
1653 	cap_rights_t rights;
1654 
1655 	if (segflg == UIO_SYSSPACE) {
1656 		syspath = path1;
1657 	} else {
1658 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1659 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1660 			goto out;
1661 	}
1662 	AUDIT_ARG_TEXT(syspath);
1663 restart:
1664 	bwillwrite();
1665 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1666 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1667 	    td);
1668 	if ((error = namei(&nd)) != 0)
1669 		goto out;
1670 	if (nd.ni_vp) {
1671 		NDFREE(&nd, NDF_ONLY_PNBUF);
1672 		if (nd.ni_vp == nd.ni_dvp)
1673 			vrele(nd.ni_dvp);
1674 		else
1675 			vput(nd.ni_dvp);
1676 		vrele(nd.ni_vp);
1677 		error = EEXIST;
1678 		goto out;
1679 	}
1680 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1681 		NDFREE(&nd, NDF_ONLY_PNBUF);
1682 		vput(nd.ni_dvp);
1683 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1684 			goto out;
1685 		goto restart;
1686 	}
1687 	VATTR_NULL(&vattr);
1688 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1689 #ifdef MAC
1690 	vattr.va_type = VLNK;
1691 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1692 	    &vattr);
1693 	if (error != 0)
1694 		goto out2;
1695 #endif
1696 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1697 	if (error == 0)
1698 		vput(nd.ni_vp);
1699 #ifdef MAC
1700 out2:
1701 #endif
1702 	NDFREE(&nd, NDF_ONLY_PNBUF);
1703 	vput(nd.ni_dvp);
1704 	vn_finished_write(mp);
1705 out:
1706 	if (segflg != UIO_SYSSPACE)
1707 		uma_zfree(namei_zone, syspath);
1708 	return (error);
1709 }
1710 
1711 /*
1712  * Delete a whiteout from the filesystem.
1713  */
1714 int
1715 sys_undelete(td, uap)
1716 	struct thread *td;
1717 	register struct undelete_args /* {
1718 		char *path;
1719 	} */ *uap;
1720 {
1721 	struct mount *mp;
1722 	struct nameidata nd;
1723 	int error;
1724 
1725 restart:
1726 	bwillwrite();
1727 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1728 	    UIO_USERSPACE, uap->path, td);
1729 	error = namei(&nd);
1730 	if (error != 0)
1731 		return (error);
1732 
1733 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1734 		NDFREE(&nd, NDF_ONLY_PNBUF);
1735 		if (nd.ni_vp == nd.ni_dvp)
1736 			vrele(nd.ni_dvp);
1737 		else
1738 			vput(nd.ni_dvp);
1739 		if (nd.ni_vp)
1740 			vrele(nd.ni_vp);
1741 		return (EEXIST);
1742 	}
1743 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1744 		NDFREE(&nd, NDF_ONLY_PNBUF);
1745 		vput(nd.ni_dvp);
1746 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1747 			return (error);
1748 		goto restart;
1749 	}
1750 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1751 	NDFREE(&nd, NDF_ONLY_PNBUF);
1752 	vput(nd.ni_dvp);
1753 	vn_finished_write(mp);
1754 	return (error);
1755 }
1756 
1757 /*
1758  * Delete a name from the filesystem.
1759  */
1760 #ifndef _SYS_SYSPROTO_H_
1761 struct unlink_args {
1762 	char	*path;
1763 };
1764 #endif
1765 int
1766 sys_unlink(td, uap)
1767 	struct thread *td;
1768 	struct unlink_args /* {
1769 		char *path;
1770 	} */ *uap;
1771 {
1772 
1773 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1774 }
1775 
1776 #ifndef _SYS_SYSPROTO_H_
1777 struct unlinkat_args {
1778 	int	fd;
1779 	char	*path;
1780 	int	flag;
1781 };
1782 #endif
1783 int
1784 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1785 {
1786 	int flag = uap->flag;
1787 	int fd = uap->fd;
1788 	char *path = uap->path;
1789 
1790 	if (flag & ~AT_REMOVEDIR)
1791 		return (EINVAL);
1792 
1793 	if (flag & AT_REMOVEDIR)
1794 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1795 	else
1796 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1797 }
1798 
1799 int
1800 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1801     ino_t oldinum)
1802 {
1803 	struct mount *mp;
1804 	struct vnode *vp;
1805 	struct nameidata nd;
1806 	struct stat sb;
1807 	cap_rights_t rights;
1808 	int error;
1809 
1810 restart:
1811 	bwillwrite();
1812 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1813 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1814 	if ((error = namei(&nd)) != 0)
1815 		return (error == EINVAL ? EPERM : error);
1816 	vp = nd.ni_vp;
1817 	if (vp->v_type == VDIR && oldinum == 0) {
1818 		error = EPERM;		/* POSIX */
1819 	} else if (oldinum != 0 &&
1820 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1821 		  sb.st_ino != oldinum) {
1822 			error = EIDRM;	/* Identifier removed */
1823 	} else {
1824 		/*
1825 		 * The root of a mounted filesystem cannot be deleted.
1826 		 *
1827 		 * XXX: can this only be a VDIR case?
1828 		 */
1829 		if (vp->v_vflag & VV_ROOT)
1830 			error = EBUSY;
1831 	}
1832 	if (error == 0) {
1833 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1834 			NDFREE(&nd, NDF_ONLY_PNBUF);
1835 			vput(nd.ni_dvp);
1836 			if (vp == nd.ni_dvp)
1837 				vrele(vp);
1838 			else
1839 				vput(vp);
1840 			if ((error = vn_start_write(NULL, &mp,
1841 			    V_XSLEEP | PCATCH)) != 0)
1842 				return (error);
1843 			goto restart;
1844 		}
1845 #ifdef MAC
1846 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1847 		    &nd.ni_cnd);
1848 		if (error != 0)
1849 			goto out;
1850 #endif
1851 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1852 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1853 #ifdef MAC
1854 out:
1855 #endif
1856 		vn_finished_write(mp);
1857 	}
1858 	NDFREE(&nd, NDF_ONLY_PNBUF);
1859 	vput(nd.ni_dvp);
1860 	if (vp == nd.ni_dvp)
1861 		vrele(vp);
1862 	else
1863 		vput(vp);
1864 	return (error);
1865 }
1866 
1867 /*
1868  * Reposition read/write file offset.
1869  */
1870 #ifndef _SYS_SYSPROTO_H_
1871 struct lseek_args {
1872 	int	fd;
1873 	int	pad;
1874 	off_t	offset;
1875 	int	whence;
1876 };
1877 #endif
1878 int
1879 sys_lseek(td, uap)
1880 	struct thread *td;
1881 	register struct lseek_args /* {
1882 		int fd;
1883 		int pad;
1884 		off_t offset;
1885 		int whence;
1886 	} */ *uap;
1887 {
1888 	struct file *fp;
1889 	cap_rights_t rights;
1890 	int error;
1891 
1892 	AUDIT_ARG_FD(uap->fd);
1893 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1894 	if (error != 0)
1895 		return (error);
1896 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1897 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1898 	fdrop(fp, td);
1899 	return (error);
1900 }
1901 
1902 #if defined(COMPAT_43)
1903 /*
1904  * Reposition read/write file offset.
1905  */
1906 #ifndef _SYS_SYSPROTO_H_
1907 struct olseek_args {
1908 	int	fd;
1909 	long	offset;
1910 	int	whence;
1911 };
1912 #endif
1913 int
1914 olseek(td, uap)
1915 	struct thread *td;
1916 	register struct olseek_args /* {
1917 		int fd;
1918 		long offset;
1919 		int whence;
1920 	} */ *uap;
1921 {
1922 	struct lseek_args /* {
1923 		int fd;
1924 		int pad;
1925 		off_t offset;
1926 		int whence;
1927 	} */ nuap;
1928 
1929 	nuap.fd = uap->fd;
1930 	nuap.offset = uap->offset;
1931 	nuap.whence = uap->whence;
1932 	return (sys_lseek(td, &nuap));
1933 }
1934 #endif /* COMPAT_43 */
1935 
1936 #if defined(COMPAT_FREEBSD6)
1937 /* Version with the 'pad' argument */
1938 int
1939 freebsd6_lseek(td, uap)
1940 	struct thread *td;
1941 	register struct freebsd6_lseek_args *uap;
1942 {
1943 	struct lseek_args ouap;
1944 
1945 	ouap.fd = uap->fd;
1946 	ouap.offset = uap->offset;
1947 	ouap.whence = uap->whence;
1948 	return (sys_lseek(td, &ouap));
1949 }
1950 #endif
1951 
1952 /*
1953  * Check access permissions using passed credentials.
1954  */
1955 static int
1956 vn_access(vp, user_flags, cred, td)
1957 	struct vnode	*vp;
1958 	int		user_flags;
1959 	struct ucred	*cred;
1960 	struct thread	*td;
1961 {
1962 	accmode_t accmode;
1963 	int error;
1964 
1965 	/* Flags == 0 means only check for existence. */
1966 	if (user_flags == 0)
1967 		return (0);
1968 
1969 	accmode = 0;
1970 	if (user_flags & R_OK)
1971 		accmode |= VREAD;
1972 	if (user_flags & W_OK)
1973 		accmode |= VWRITE;
1974 	if (user_flags & X_OK)
1975 		accmode |= VEXEC;
1976 #ifdef MAC
1977 	error = mac_vnode_check_access(cred, vp, accmode);
1978 	if (error != 0)
1979 		return (error);
1980 #endif
1981 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1982 		error = VOP_ACCESS(vp, accmode, cred, td);
1983 	return (error);
1984 }
1985 
1986 /*
1987  * Check access permissions using "real" credentials.
1988  */
1989 #ifndef _SYS_SYSPROTO_H_
1990 struct access_args {
1991 	char	*path;
1992 	int	amode;
1993 };
1994 #endif
1995 int
1996 sys_access(td, uap)
1997 	struct thread *td;
1998 	register struct access_args /* {
1999 		char *path;
2000 		int amode;
2001 	} */ *uap;
2002 {
2003 
2004 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2005 	    0, uap->amode));
2006 }
2007 
2008 #ifndef _SYS_SYSPROTO_H_
2009 struct faccessat_args {
2010 	int	dirfd;
2011 	char	*path;
2012 	int	amode;
2013 	int	flag;
2014 }
2015 #endif
2016 int
2017 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2018 {
2019 
2020 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2021 	    uap->amode));
2022 }
2023 
2024 int
2025 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2026     int flag, int amode)
2027 {
2028 	struct ucred *cred, *usecred;
2029 	struct vnode *vp;
2030 	struct nameidata nd;
2031 	cap_rights_t rights;
2032 	int error;
2033 
2034 	if (flag & ~AT_EACCESS)
2035 		return (EINVAL);
2036 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2037 		return (EINVAL);
2038 
2039 	/*
2040 	 * Create and modify a temporary credential instead of one that
2041 	 * is potentially shared (if we need one).
2042 	 */
2043 	cred = td->td_ucred;
2044 	if ((flag & AT_EACCESS) == 0 &&
2045 	    ((cred->cr_uid != cred->cr_ruid ||
2046 	    cred->cr_rgid != cred->cr_groups[0]))) {
2047 		usecred = crdup(cred);
2048 		usecred->cr_uid = cred->cr_ruid;
2049 		usecred->cr_groups[0] = cred->cr_rgid;
2050 		td->td_ucred = usecred;
2051 	} else
2052 		usecred = cred;
2053 	AUDIT_ARG_VALUE(amode);
2054 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2055 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2056 	    td);
2057 	if ((error = namei(&nd)) != 0)
2058 		goto out;
2059 	vp = nd.ni_vp;
2060 
2061 	error = vn_access(vp, amode, usecred, td);
2062 	NDFREE(&nd, NDF_ONLY_PNBUF);
2063 	vput(vp);
2064 out:
2065 	if (usecred != cred) {
2066 		td->td_ucred = cred;
2067 		crfree(usecred);
2068 	}
2069 	return (error);
2070 }
2071 
2072 /*
2073  * Check access permissions using "effective" credentials.
2074  */
2075 #ifndef _SYS_SYSPROTO_H_
2076 struct eaccess_args {
2077 	char	*path;
2078 	int	amode;
2079 };
2080 #endif
2081 int
2082 sys_eaccess(td, uap)
2083 	struct thread *td;
2084 	register struct eaccess_args /* {
2085 		char *path;
2086 		int amode;
2087 	} */ *uap;
2088 {
2089 
2090 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2091 	    AT_EACCESS, uap->amode));
2092 }
2093 
2094 #if defined(COMPAT_43)
2095 /*
2096  * Get file status; this version follows links.
2097  */
2098 #ifndef _SYS_SYSPROTO_H_
2099 struct ostat_args {
2100 	char	*path;
2101 	struct ostat *ub;
2102 };
2103 #endif
2104 int
2105 ostat(td, uap)
2106 	struct thread *td;
2107 	register struct ostat_args /* {
2108 		char *path;
2109 		struct ostat *ub;
2110 	} */ *uap;
2111 {
2112 	struct stat sb;
2113 	struct ostat osb;
2114 	int error;
2115 
2116 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2117 	    &sb, NULL);
2118 	if (error != 0)
2119 		return (error);
2120 	cvtstat(&sb, &osb);
2121 	return (copyout(&osb, uap->ub, sizeof (osb)));
2122 }
2123 
2124 /*
2125  * Get file status; this version does not follow links.
2126  */
2127 #ifndef _SYS_SYSPROTO_H_
2128 struct olstat_args {
2129 	char	*path;
2130 	struct ostat *ub;
2131 };
2132 #endif
2133 int
2134 olstat(td, uap)
2135 	struct thread *td;
2136 	register struct olstat_args /* {
2137 		char *path;
2138 		struct ostat *ub;
2139 	} */ *uap;
2140 {
2141 	struct stat sb;
2142 	struct ostat osb;
2143 	int error;
2144 
2145 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2146 	    UIO_USERSPACE, &sb, NULL);
2147 	if (error != 0)
2148 		return (error);
2149 	cvtstat(&sb, &osb);
2150 	return (copyout(&osb, uap->ub, sizeof (osb)));
2151 }
2152 
2153 /*
2154  * Convert from an old to a new stat structure.
2155  */
2156 void
2157 cvtstat(st, ost)
2158 	struct stat *st;
2159 	struct ostat *ost;
2160 {
2161 
2162 	ost->st_dev = st->st_dev;
2163 	ost->st_ino = st->st_ino;
2164 	ost->st_mode = st->st_mode;
2165 	ost->st_nlink = st->st_nlink;
2166 	ost->st_uid = st->st_uid;
2167 	ost->st_gid = st->st_gid;
2168 	ost->st_rdev = st->st_rdev;
2169 	if (st->st_size < (quad_t)1 << 32)
2170 		ost->st_size = st->st_size;
2171 	else
2172 		ost->st_size = -2;
2173 	ost->st_atim = st->st_atim;
2174 	ost->st_mtim = st->st_mtim;
2175 	ost->st_ctim = st->st_ctim;
2176 	ost->st_blksize = st->st_blksize;
2177 	ost->st_blocks = st->st_blocks;
2178 	ost->st_flags = st->st_flags;
2179 	ost->st_gen = st->st_gen;
2180 }
2181 #endif /* COMPAT_43 */
2182 
2183 /*
2184  * Get file status; this version follows links.
2185  */
2186 #ifndef _SYS_SYSPROTO_H_
2187 struct stat_args {
2188 	char	*path;
2189 	struct stat *ub;
2190 };
2191 #endif
2192 int
2193 sys_stat(td, uap)
2194 	struct thread *td;
2195 	register struct stat_args /* {
2196 		char *path;
2197 		struct stat *ub;
2198 	} */ *uap;
2199 {
2200 	struct stat sb;
2201 	int error;
2202 
2203 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2204 	    &sb, NULL);
2205 	if (error == 0)
2206 		error = copyout(&sb, uap->ub, sizeof (sb));
2207 	return (error);
2208 }
2209 
2210 #ifndef _SYS_SYSPROTO_H_
2211 struct fstatat_args {
2212 	int	fd;
2213 	char	*path;
2214 	struct stat	*buf;
2215 	int	flag;
2216 }
2217 #endif
2218 int
2219 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2220 {
2221 	struct stat sb;
2222 	int error;
2223 
2224 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2225 	    UIO_USERSPACE, &sb, NULL);
2226 	if (error == 0)
2227 		error = copyout(&sb, uap->buf, sizeof (sb));
2228 	return (error);
2229 }
2230 
2231 int
2232 kern_statat(struct thread *td, int flag, int fd, char *path,
2233     enum uio_seg pathseg, struct stat *sbp,
2234     void (*hook)(struct vnode *vp, struct stat *sbp))
2235 {
2236 	struct nameidata nd;
2237 	struct stat sb;
2238 	cap_rights_t rights;
2239 	int error;
2240 
2241 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2242 		return (EINVAL);
2243 
2244 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2245 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2246 	    cap_rights_init(&rights, CAP_FSTAT), td);
2247 
2248 	if ((error = namei(&nd)) != 0)
2249 		return (error);
2250 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2251 	if (error == 0) {
2252 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2253 		if (S_ISREG(sb.st_mode))
2254 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2255 		if (__predict_false(hook != NULL))
2256 			hook(nd.ni_vp, &sb);
2257 	}
2258 	NDFREE(&nd, NDF_ONLY_PNBUF);
2259 	vput(nd.ni_vp);
2260 	if (error != 0)
2261 		return (error);
2262 	*sbp = sb;
2263 #ifdef KTRACE
2264 	if (KTRPOINT(td, KTR_STRUCT))
2265 		ktrstat(&sb);
2266 #endif
2267 	return (0);
2268 }
2269 
2270 /*
2271  * Get file status; this version does not follow links.
2272  */
2273 #ifndef _SYS_SYSPROTO_H_
2274 struct lstat_args {
2275 	char	*path;
2276 	struct stat *ub;
2277 };
2278 #endif
2279 int
2280 sys_lstat(td, uap)
2281 	struct thread *td;
2282 	register struct lstat_args /* {
2283 		char *path;
2284 		struct stat *ub;
2285 	} */ *uap;
2286 {
2287 	struct stat sb;
2288 	int error;
2289 
2290 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2291 	    UIO_USERSPACE, &sb, NULL);
2292 	if (error == 0)
2293 		error = copyout(&sb, uap->ub, sizeof (sb));
2294 	return (error);
2295 }
2296 
2297 /*
2298  * Implementation of the NetBSD [l]stat() functions.
2299  */
2300 void
2301 cvtnstat(sb, nsb)
2302 	struct stat *sb;
2303 	struct nstat *nsb;
2304 {
2305 
2306 	bzero(nsb, sizeof *nsb);
2307 	nsb->st_dev = sb->st_dev;
2308 	nsb->st_ino = sb->st_ino;
2309 	nsb->st_mode = sb->st_mode;
2310 	nsb->st_nlink = sb->st_nlink;
2311 	nsb->st_uid = sb->st_uid;
2312 	nsb->st_gid = sb->st_gid;
2313 	nsb->st_rdev = sb->st_rdev;
2314 	nsb->st_atim = sb->st_atim;
2315 	nsb->st_mtim = sb->st_mtim;
2316 	nsb->st_ctim = sb->st_ctim;
2317 	nsb->st_size = sb->st_size;
2318 	nsb->st_blocks = sb->st_blocks;
2319 	nsb->st_blksize = sb->st_blksize;
2320 	nsb->st_flags = sb->st_flags;
2321 	nsb->st_gen = sb->st_gen;
2322 	nsb->st_birthtim = sb->st_birthtim;
2323 }
2324 
2325 #ifndef _SYS_SYSPROTO_H_
2326 struct nstat_args {
2327 	char	*path;
2328 	struct nstat *ub;
2329 };
2330 #endif
2331 int
2332 sys_nstat(td, uap)
2333 	struct thread *td;
2334 	register struct nstat_args /* {
2335 		char *path;
2336 		struct nstat *ub;
2337 	} */ *uap;
2338 {
2339 	struct stat sb;
2340 	struct nstat nsb;
2341 	int error;
2342 
2343 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2344 	    &sb, NULL);
2345 	if (error != 0)
2346 		return (error);
2347 	cvtnstat(&sb, &nsb);
2348 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2349 }
2350 
2351 /*
2352  * NetBSD lstat.  Get file status; this version does not follow links.
2353  */
2354 #ifndef _SYS_SYSPROTO_H_
2355 struct lstat_args {
2356 	char	*path;
2357 	struct stat *ub;
2358 };
2359 #endif
2360 int
2361 sys_nlstat(td, uap)
2362 	struct thread *td;
2363 	register struct nlstat_args /* {
2364 		char *path;
2365 		struct nstat *ub;
2366 	} */ *uap;
2367 {
2368 	struct stat sb;
2369 	struct nstat nsb;
2370 	int error;
2371 
2372 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2373 	    UIO_USERSPACE, &sb, NULL);
2374 	if (error != 0)
2375 		return (error);
2376 	cvtnstat(&sb, &nsb);
2377 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2378 }
2379 
2380 /*
2381  * Get configurable pathname variables.
2382  */
2383 #ifndef _SYS_SYSPROTO_H_
2384 struct pathconf_args {
2385 	char	*path;
2386 	int	name;
2387 };
2388 #endif
2389 int
2390 sys_pathconf(td, uap)
2391 	struct thread *td;
2392 	register struct pathconf_args /* {
2393 		char *path;
2394 		int name;
2395 	} */ *uap;
2396 {
2397 
2398 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2399 }
2400 
2401 #ifndef _SYS_SYSPROTO_H_
2402 struct lpathconf_args {
2403 	char	*path;
2404 	int	name;
2405 };
2406 #endif
2407 int
2408 sys_lpathconf(td, uap)
2409 	struct thread *td;
2410 	register struct lpathconf_args /* {
2411 		char *path;
2412 		int name;
2413 	} */ *uap;
2414 {
2415 
2416 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2417 	    NOFOLLOW));
2418 }
2419 
2420 int
2421 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2422     u_long flags)
2423 {
2424 	struct nameidata nd;
2425 	int error;
2426 
2427 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2428 	    pathseg, path, td);
2429 	if ((error = namei(&nd)) != 0)
2430 		return (error);
2431 	NDFREE(&nd, NDF_ONLY_PNBUF);
2432 
2433 	/* If asynchronous I/O is available, it works for all files. */
2434 	if (name == _PC_ASYNC_IO)
2435 		td->td_retval[0] = async_io_version;
2436 	else
2437 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2438 	vput(nd.ni_vp);
2439 	return (error);
2440 }
2441 
2442 /*
2443  * Return target name of a symbolic link.
2444  */
2445 #ifndef _SYS_SYSPROTO_H_
2446 struct readlink_args {
2447 	char	*path;
2448 	char	*buf;
2449 	size_t	count;
2450 };
2451 #endif
2452 int
2453 sys_readlink(td, uap)
2454 	struct thread *td;
2455 	register struct readlink_args /* {
2456 		char *path;
2457 		char *buf;
2458 		size_t count;
2459 	} */ *uap;
2460 {
2461 
2462 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2463 	    uap->buf, UIO_USERSPACE, uap->count));
2464 }
2465 #ifndef _SYS_SYSPROTO_H_
2466 struct readlinkat_args {
2467 	int	fd;
2468 	char	*path;
2469 	char	*buf;
2470 	size_t	bufsize;
2471 };
2472 #endif
2473 int
2474 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2475 {
2476 
2477 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2478 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2479 }
2480 
2481 int
2482 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2483     char *buf, enum uio_seg bufseg, size_t count)
2484 {
2485 	struct vnode *vp;
2486 	struct iovec aiov;
2487 	struct uio auio;
2488 	struct nameidata nd;
2489 	int error;
2490 
2491 	if (count > IOSIZE_MAX)
2492 		return (EINVAL);
2493 
2494 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2495 	    pathseg, path, fd, td);
2496 
2497 	if ((error = namei(&nd)) != 0)
2498 		return (error);
2499 	NDFREE(&nd, NDF_ONLY_PNBUF);
2500 	vp = nd.ni_vp;
2501 #ifdef MAC
2502 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2503 	if (error != 0) {
2504 		vput(vp);
2505 		return (error);
2506 	}
2507 #endif
2508 	if (vp->v_type != VLNK)
2509 		error = EINVAL;
2510 	else {
2511 		aiov.iov_base = buf;
2512 		aiov.iov_len = count;
2513 		auio.uio_iov = &aiov;
2514 		auio.uio_iovcnt = 1;
2515 		auio.uio_offset = 0;
2516 		auio.uio_rw = UIO_READ;
2517 		auio.uio_segflg = bufseg;
2518 		auio.uio_td = td;
2519 		auio.uio_resid = count;
2520 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2521 		td->td_retval[0] = count - auio.uio_resid;
2522 	}
2523 	vput(vp);
2524 	return (error);
2525 }
2526 
2527 /*
2528  * Common implementation code for chflags() and fchflags().
2529  */
2530 static int
2531 setfflags(td, vp, flags)
2532 	struct thread *td;
2533 	struct vnode *vp;
2534 	u_long flags;
2535 {
2536 	struct mount *mp;
2537 	struct vattr vattr;
2538 	int error;
2539 
2540 	/* We can't support the value matching VNOVAL. */
2541 	if (flags == VNOVAL)
2542 		return (EOPNOTSUPP);
2543 
2544 	/*
2545 	 * Prevent non-root users from setting flags on devices.  When
2546 	 * a device is reused, users can retain ownership of the device
2547 	 * if they are allowed to set flags and programs assume that
2548 	 * chown can't fail when done as root.
2549 	 */
2550 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2551 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2552 		if (error != 0)
2553 			return (error);
2554 	}
2555 
2556 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2557 		return (error);
2558 	VATTR_NULL(&vattr);
2559 	vattr.va_flags = flags;
2560 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2561 #ifdef MAC
2562 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2563 	if (error == 0)
2564 #endif
2565 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2566 	VOP_UNLOCK(vp, 0);
2567 	vn_finished_write(mp);
2568 	return (error);
2569 }
2570 
2571 /*
2572  * Change flags of a file given a path name.
2573  */
2574 #ifndef _SYS_SYSPROTO_H_
2575 struct chflags_args {
2576 	const char *path;
2577 	u_long	flags;
2578 };
2579 #endif
2580 int
2581 sys_chflags(td, uap)
2582 	struct thread *td;
2583 	register struct chflags_args /* {
2584 		const char *path;
2585 		u_long flags;
2586 	} */ *uap;
2587 {
2588 
2589 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2590 	    uap->flags, 0));
2591 }
2592 
2593 #ifndef _SYS_SYSPROTO_H_
2594 struct chflagsat_args {
2595 	int	fd;
2596 	const char *path;
2597 	u_long	flags;
2598 	int	atflag;
2599 }
2600 #endif
2601 int
2602 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2603 {
2604 	int fd = uap->fd;
2605 	const char *path = uap->path;
2606 	u_long flags = uap->flags;
2607 	int atflag = uap->atflag;
2608 
2609 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2610 		return (EINVAL);
2611 
2612 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2613 }
2614 
2615 /*
2616  * Same as chflags() but doesn't follow symlinks.
2617  */
2618 int
2619 sys_lchflags(td, uap)
2620 	struct thread *td;
2621 	register struct lchflags_args /* {
2622 		const char *path;
2623 		u_long flags;
2624 	} */ *uap;
2625 {
2626 
2627 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2628 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2629 }
2630 
2631 static int
2632 kern_chflagsat(struct thread *td, int fd, const char *path,
2633     enum uio_seg pathseg, u_long flags, int atflag)
2634 {
2635 	struct nameidata nd;
2636 	cap_rights_t rights;
2637 	int error, follow;
2638 
2639 	AUDIT_ARG_FFLAGS(flags);
2640 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2641 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2642 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2643 	if ((error = namei(&nd)) != 0)
2644 		return (error);
2645 	NDFREE(&nd, NDF_ONLY_PNBUF);
2646 	error = setfflags(td, nd.ni_vp, flags);
2647 	vrele(nd.ni_vp);
2648 	return (error);
2649 }
2650 
2651 /*
2652  * Change flags of a file given a file descriptor.
2653  */
2654 #ifndef _SYS_SYSPROTO_H_
2655 struct fchflags_args {
2656 	int	fd;
2657 	u_long	flags;
2658 };
2659 #endif
2660 int
2661 sys_fchflags(td, uap)
2662 	struct thread *td;
2663 	register struct fchflags_args /* {
2664 		int fd;
2665 		u_long flags;
2666 	} */ *uap;
2667 {
2668 	struct file *fp;
2669 	cap_rights_t rights;
2670 	int error;
2671 
2672 	AUDIT_ARG_FD(uap->fd);
2673 	AUDIT_ARG_FFLAGS(uap->flags);
2674 	error = getvnode(td->td_proc->p_fd, uap->fd,
2675 	    cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2676 	if (error != 0)
2677 		return (error);
2678 #ifdef AUDIT
2679 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2680 	AUDIT_ARG_VNODE1(fp->f_vnode);
2681 	VOP_UNLOCK(fp->f_vnode, 0);
2682 #endif
2683 	error = setfflags(td, fp->f_vnode, uap->flags);
2684 	fdrop(fp, td);
2685 	return (error);
2686 }
2687 
2688 /*
2689  * Common implementation code for chmod(), lchmod() and fchmod().
2690  */
2691 int
2692 setfmode(td, cred, vp, mode)
2693 	struct thread *td;
2694 	struct ucred *cred;
2695 	struct vnode *vp;
2696 	int mode;
2697 {
2698 	struct mount *mp;
2699 	struct vattr vattr;
2700 	int error;
2701 
2702 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2703 		return (error);
2704 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2705 	VATTR_NULL(&vattr);
2706 	vattr.va_mode = mode & ALLPERMS;
2707 #ifdef MAC
2708 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2709 	if (error == 0)
2710 #endif
2711 		error = VOP_SETATTR(vp, &vattr, cred);
2712 	VOP_UNLOCK(vp, 0);
2713 	vn_finished_write(mp);
2714 	return (error);
2715 }
2716 
2717 /*
2718  * Change mode of a file given path name.
2719  */
2720 #ifndef _SYS_SYSPROTO_H_
2721 struct chmod_args {
2722 	char	*path;
2723 	int	mode;
2724 };
2725 #endif
2726 int
2727 sys_chmod(td, uap)
2728 	struct thread *td;
2729 	register struct chmod_args /* {
2730 		char *path;
2731 		int mode;
2732 	} */ *uap;
2733 {
2734 
2735 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2736 	    uap->mode, 0));
2737 }
2738 
2739 #ifndef _SYS_SYSPROTO_H_
2740 struct fchmodat_args {
2741 	int	dirfd;
2742 	char	*path;
2743 	mode_t	mode;
2744 	int	flag;
2745 }
2746 #endif
2747 int
2748 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2749 {
2750 	int flag = uap->flag;
2751 	int fd = uap->fd;
2752 	char *path = uap->path;
2753 	mode_t mode = uap->mode;
2754 
2755 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2756 		return (EINVAL);
2757 
2758 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2759 }
2760 
2761 /*
2762  * Change mode of a file given path name (don't follow links.)
2763  */
2764 #ifndef _SYS_SYSPROTO_H_
2765 struct lchmod_args {
2766 	char	*path;
2767 	int	mode;
2768 };
2769 #endif
2770 int
2771 sys_lchmod(td, uap)
2772 	struct thread *td;
2773 	register struct lchmod_args /* {
2774 		char *path;
2775 		int mode;
2776 	} */ *uap;
2777 {
2778 
2779 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2780 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2781 }
2782 
2783 int
2784 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2785     mode_t mode, int flag)
2786 {
2787 	struct nameidata nd;
2788 	cap_rights_t rights;
2789 	int error, follow;
2790 
2791 	AUDIT_ARG_MODE(mode);
2792 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2793 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2794 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2795 	if ((error = namei(&nd)) != 0)
2796 		return (error);
2797 	NDFREE(&nd, NDF_ONLY_PNBUF);
2798 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2799 	vrele(nd.ni_vp);
2800 	return (error);
2801 }
2802 
2803 /*
2804  * Change mode of a file given a file descriptor.
2805  */
2806 #ifndef _SYS_SYSPROTO_H_
2807 struct fchmod_args {
2808 	int	fd;
2809 	int	mode;
2810 };
2811 #endif
2812 int
2813 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2814 {
2815 	struct file *fp;
2816 	cap_rights_t rights;
2817 	int error;
2818 
2819 	AUDIT_ARG_FD(uap->fd);
2820 	AUDIT_ARG_MODE(uap->mode);
2821 
2822 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2823 	if (error != 0)
2824 		return (error);
2825 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2826 	fdrop(fp, td);
2827 	return (error);
2828 }
2829 
2830 /*
2831  * Common implementation for chown(), lchown(), and fchown()
2832  */
2833 int
2834 setfown(td, cred, vp, uid, gid)
2835 	struct thread *td;
2836 	struct ucred *cred;
2837 	struct vnode *vp;
2838 	uid_t uid;
2839 	gid_t gid;
2840 {
2841 	struct mount *mp;
2842 	struct vattr vattr;
2843 	int error;
2844 
2845 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2846 		return (error);
2847 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2848 	VATTR_NULL(&vattr);
2849 	vattr.va_uid = uid;
2850 	vattr.va_gid = gid;
2851 #ifdef MAC
2852 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2853 	    vattr.va_gid);
2854 	if (error == 0)
2855 #endif
2856 		error = VOP_SETATTR(vp, &vattr, cred);
2857 	VOP_UNLOCK(vp, 0);
2858 	vn_finished_write(mp);
2859 	return (error);
2860 }
2861 
2862 /*
2863  * Set ownership given a path name.
2864  */
2865 #ifndef _SYS_SYSPROTO_H_
2866 struct chown_args {
2867 	char	*path;
2868 	int	uid;
2869 	int	gid;
2870 };
2871 #endif
2872 int
2873 sys_chown(td, uap)
2874 	struct thread *td;
2875 	register struct chown_args /* {
2876 		char *path;
2877 		int uid;
2878 		int gid;
2879 	} */ *uap;
2880 {
2881 
2882 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2883 	    uap->gid, 0));
2884 }
2885 
2886 #ifndef _SYS_SYSPROTO_H_
2887 struct fchownat_args {
2888 	int fd;
2889 	const char * path;
2890 	uid_t uid;
2891 	gid_t gid;
2892 	int flag;
2893 };
2894 #endif
2895 int
2896 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2897 {
2898 	int flag;
2899 
2900 	flag = uap->flag;
2901 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2902 		return (EINVAL);
2903 
2904 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2905 	    uap->gid, uap->flag));
2906 }
2907 
2908 int
2909 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2910     int uid, int gid, int flag)
2911 {
2912 	struct nameidata nd;
2913 	cap_rights_t rights;
2914 	int error, follow;
2915 
2916 	AUDIT_ARG_OWNER(uid, gid);
2917 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2918 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2919 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2920 
2921 	if ((error = namei(&nd)) != 0)
2922 		return (error);
2923 	NDFREE(&nd, NDF_ONLY_PNBUF);
2924 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2925 	vrele(nd.ni_vp);
2926 	return (error);
2927 }
2928 
2929 /*
2930  * Set ownership given a path name, do not cross symlinks.
2931  */
2932 #ifndef _SYS_SYSPROTO_H_
2933 struct lchown_args {
2934 	char	*path;
2935 	int	uid;
2936 	int	gid;
2937 };
2938 #endif
2939 int
2940 sys_lchown(td, uap)
2941 	struct thread *td;
2942 	register struct lchown_args /* {
2943 		char *path;
2944 		int uid;
2945 		int gid;
2946 	} */ *uap;
2947 {
2948 
2949 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2950 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2951 }
2952 
2953 /*
2954  * Set ownership given a file descriptor.
2955  */
2956 #ifndef _SYS_SYSPROTO_H_
2957 struct fchown_args {
2958 	int	fd;
2959 	int	uid;
2960 	int	gid;
2961 };
2962 #endif
2963 int
2964 sys_fchown(td, uap)
2965 	struct thread *td;
2966 	register struct fchown_args /* {
2967 		int fd;
2968 		int uid;
2969 		int gid;
2970 	} */ *uap;
2971 {
2972 	struct file *fp;
2973 	cap_rights_t rights;
2974 	int error;
2975 
2976 	AUDIT_ARG_FD(uap->fd);
2977 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2978 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2979 	if (error != 0)
2980 		return (error);
2981 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2982 	fdrop(fp, td);
2983 	return (error);
2984 }
2985 
2986 /*
2987  * Common implementation code for utimes(), lutimes(), and futimes().
2988  */
2989 static int
2990 getutimes(usrtvp, tvpseg, tsp)
2991 	const struct timeval *usrtvp;
2992 	enum uio_seg tvpseg;
2993 	struct timespec *tsp;
2994 {
2995 	struct timeval tv[2];
2996 	const struct timeval *tvp;
2997 	int error;
2998 
2999 	if (usrtvp == NULL) {
3000 		vfs_timestamp(&tsp[0]);
3001 		tsp[1] = tsp[0];
3002 	} else {
3003 		if (tvpseg == UIO_SYSSPACE) {
3004 			tvp = usrtvp;
3005 		} else {
3006 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3007 				return (error);
3008 			tvp = tv;
3009 		}
3010 
3011 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3012 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3013 			return (EINVAL);
3014 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3015 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3016 	}
3017 	return (0);
3018 }
3019 
3020 /*
3021  * Common implementation code for futimens(), utimensat().
3022  */
3023 #define	UTIMENS_NULL	0x1
3024 #define	UTIMENS_EXIT	0x2
3025 static int
3026 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3027     struct timespec *tsp, int *retflags)
3028 {
3029 	struct timespec tsnow;
3030 	int error;
3031 
3032 	vfs_timestamp(&tsnow);
3033 	*retflags = 0;
3034 	if (usrtsp == NULL) {
3035 		tsp[0] = tsnow;
3036 		tsp[1] = tsnow;
3037 		*retflags |= UTIMENS_NULL;
3038 		return (0);
3039 	}
3040 	if (tspseg == UIO_SYSSPACE) {
3041 		tsp[0] = usrtsp[0];
3042 		tsp[1] = usrtsp[1];
3043 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3044 		return (error);
3045 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3046 		*retflags |= UTIMENS_EXIT;
3047 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3048 		*retflags |= UTIMENS_NULL;
3049 	if (tsp[0].tv_nsec == UTIME_OMIT)
3050 		tsp[0].tv_sec = VNOVAL;
3051 	else if (tsp[0].tv_nsec == UTIME_NOW)
3052 		tsp[0] = tsnow;
3053 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3054 		return (EINVAL);
3055 	if (tsp[1].tv_nsec == UTIME_OMIT)
3056 		tsp[1].tv_sec = VNOVAL;
3057 	else if (tsp[1].tv_nsec == UTIME_NOW)
3058 		tsp[1] = tsnow;
3059 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3060 		return (EINVAL);
3061 
3062 	return (0);
3063 }
3064 
3065 /*
3066  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3067  * and utimensat().
3068  */
3069 static int
3070 setutimes(td, vp, ts, numtimes, nullflag)
3071 	struct thread *td;
3072 	struct vnode *vp;
3073 	const struct timespec *ts;
3074 	int numtimes;
3075 	int nullflag;
3076 {
3077 	struct mount *mp;
3078 	struct vattr vattr;
3079 	int error, setbirthtime;
3080 
3081 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3082 		return (error);
3083 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3084 	setbirthtime = 0;
3085 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3086 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3087 		setbirthtime = 1;
3088 	VATTR_NULL(&vattr);
3089 	vattr.va_atime = ts[0];
3090 	vattr.va_mtime = ts[1];
3091 	if (setbirthtime)
3092 		vattr.va_birthtime = ts[1];
3093 	if (numtimes > 2)
3094 		vattr.va_birthtime = ts[2];
3095 	if (nullflag)
3096 		vattr.va_vaflags |= VA_UTIMES_NULL;
3097 #ifdef MAC
3098 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3099 	    vattr.va_mtime);
3100 #endif
3101 	if (error == 0)
3102 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3103 	VOP_UNLOCK(vp, 0);
3104 	vn_finished_write(mp);
3105 	return (error);
3106 }
3107 
3108 /*
3109  * Set the access and modification times of a file.
3110  */
3111 #ifndef _SYS_SYSPROTO_H_
3112 struct utimes_args {
3113 	char	*path;
3114 	struct	timeval *tptr;
3115 };
3116 #endif
3117 int
3118 sys_utimes(td, uap)
3119 	struct thread *td;
3120 	register struct utimes_args /* {
3121 		char *path;
3122 		struct timeval *tptr;
3123 	} */ *uap;
3124 {
3125 
3126 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3127 	    uap->tptr, UIO_USERSPACE));
3128 }
3129 
3130 #ifndef _SYS_SYSPROTO_H_
3131 struct futimesat_args {
3132 	int fd;
3133 	const char * path;
3134 	const struct timeval * times;
3135 };
3136 #endif
3137 int
3138 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3139 {
3140 
3141 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3142 	    uap->times, UIO_USERSPACE));
3143 }
3144 
3145 int
3146 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3147     struct timeval *tptr, enum uio_seg tptrseg)
3148 {
3149 	struct nameidata nd;
3150 	struct timespec ts[2];
3151 	cap_rights_t rights;
3152 	int error;
3153 
3154 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3155 		return (error);
3156 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3157 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3158 
3159 	if ((error = namei(&nd)) != 0)
3160 		return (error);
3161 	NDFREE(&nd, NDF_ONLY_PNBUF);
3162 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3163 	vrele(nd.ni_vp);
3164 	return (error);
3165 }
3166 
3167 /*
3168  * Set the access and modification times of a file.
3169  */
3170 #ifndef _SYS_SYSPROTO_H_
3171 struct lutimes_args {
3172 	char	*path;
3173 	struct	timeval *tptr;
3174 };
3175 #endif
3176 int
3177 sys_lutimes(td, uap)
3178 	struct thread *td;
3179 	register struct lutimes_args /* {
3180 		char *path;
3181 		struct timeval *tptr;
3182 	} */ *uap;
3183 {
3184 
3185 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3186 	    UIO_USERSPACE));
3187 }
3188 
3189 int
3190 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3191     struct timeval *tptr, enum uio_seg tptrseg)
3192 {
3193 	struct timespec ts[2];
3194 	struct nameidata nd;
3195 	int error;
3196 
3197 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3198 		return (error);
3199 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3200 	if ((error = namei(&nd)) != 0)
3201 		return (error);
3202 	NDFREE(&nd, NDF_ONLY_PNBUF);
3203 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3204 	vrele(nd.ni_vp);
3205 	return (error);
3206 }
3207 
3208 /*
3209  * Set the access and modification times of a file.
3210  */
3211 #ifndef _SYS_SYSPROTO_H_
3212 struct futimes_args {
3213 	int	fd;
3214 	struct	timeval *tptr;
3215 };
3216 #endif
3217 int
3218 sys_futimes(td, uap)
3219 	struct thread *td;
3220 	register struct futimes_args /* {
3221 		int  fd;
3222 		struct timeval *tptr;
3223 	} */ *uap;
3224 {
3225 
3226 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3227 }
3228 
3229 int
3230 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3231     enum uio_seg tptrseg)
3232 {
3233 	struct timespec ts[2];
3234 	struct file *fp;
3235 	cap_rights_t rights;
3236 	int error;
3237 
3238 	AUDIT_ARG_FD(fd);
3239 	error = getutimes(tptr, tptrseg, ts);
3240 	if (error != 0)
3241 		return (error);
3242 	error = getvnode(td->td_proc->p_fd, fd,
3243 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3244 	if (error != 0)
3245 		return (error);
3246 #ifdef AUDIT
3247 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3248 	AUDIT_ARG_VNODE1(fp->f_vnode);
3249 	VOP_UNLOCK(fp->f_vnode, 0);
3250 #endif
3251 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3252 	fdrop(fp, td);
3253 	return (error);
3254 }
3255 
3256 int
3257 sys_futimens(struct thread *td, struct futimens_args *uap)
3258 {
3259 
3260 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3261 }
3262 
3263 int
3264 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3265     enum uio_seg tptrseg)
3266 {
3267 	struct timespec ts[2];
3268 	struct file *fp;
3269 	cap_rights_t rights;
3270 	int error, flags;
3271 
3272 	AUDIT_ARG_FD(fd);
3273 	error = getutimens(tptr, tptrseg, ts, &flags);
3274 	if (error != 0)
3275 		return (error);
3276 	if (flags & UTIMENS_EXIT)
3277 		return (0);
3278 	error = getvnode(td->td_proc->p_fd, fd,
3279 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3280 	if (error != 0)
3281 		return (error);
3282 #ifdef AUDIT
3283 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3284 	AUDIT_ARG_VNODE1(fp->f_vnode);
3285 	VOP_UNLOCK(fp->f_vnode, 0);
3286 #endif
3287 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3288 	fdrop(fp, td);
3289 	return (error);
3290 }
3291 
3292 int
3293 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3294 {
3295 
3296 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3297 	    uap->times, UIO_USERSPACE, uap->flag));
3298 }
3299 
3300 int
3301 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3302     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3303 {
3304 	struct nameidata nd;
3305 	struct timespec ts[2];
3306 	cap_rights_t rights;
3307 	int error, flags;
3308 
3309 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3310 		return (EINVAL);
3311 
3312 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3313 		return (error);
3314 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3315 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3316 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3317 	if ((error = namei(&nd)) != 0)
3318 		return (error);
3319 	/*
3320 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3321 	 * POSIX states:
3322 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3323 	 * "Search permission is denied by a component of the path prefix."
3324 	 */
3325 	NDFREE(&nd, NDF_ONLY_PNBUF);
3326 	if ((flags & UTIMENS_EXIT) == 0)
3327 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3328 	vrele(nd.ni_vp);
3329 	return (error);
3330 }
3331 
3332 /*
3333  * Truncate a file given its path name.
3334  */
3335 #ifndef _SYS_SYSPROTO_H_
3336 struct truncate_args {
3337 	char	*path;
3338 	int	pad;
3339 	off_t	length;
3340 };
3341 #endif
3342 int
3343 sys_truncate(td, uap)
3344 	struct thread *td;
3345 	register struct truncate_args /* {
3346 		char *path;
3347 		int pad;
3348 		off_t length;
3349 	} */ *uap;
3350 {
3351 
3352 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3353 }
3354 
3355 int
3356 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3357 {
3358 	struct mount *mp;
3359 	struct vnode *vp;
3360 	void *rl_cookie;
3361 	struct vattr vattr;
3362 	struct nameidata nd;
3363 	int error;
3364 
3365 	if (length < 0)
3366 		return(EINVAL);
3367 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3368 	if ((error = namei(&nd)) != 0)
3369 		return (error);
3370 	vp = nd.ni_vp;
3371 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3372 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3373 		vn_rangelock_unlock(vp, rl_cookie);
3374 		vrele(vp);
3375 		return (error);
3376 	}
3377 	NDFREE(&nd, NDF_ONLY_PNBUF);
3378 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3379 	if (vp->v_type == VDIR)
3380 		error = EISDIR;
3381 #ifdef MAC
3382 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3383 	}
3384 #endif
3385 	else if ((error = vn_writechk(vp)) == 0 &&
3386 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3387 		VATTR_NULL(&vattr);
3388 		vattr.va_size = length;
3389 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3390 	}
3391 	VOP_UNLOCK(vp, 0);
3392 	vn_finished_write(mp);
3393 	vn_rangelock_unlock(vp, rl_cookie);
3394 	vrele(vp);
3395 	return (error);
3396 }
3397 
3398 #if defined(COMPAT_43)
3399 /*
3400  * Truncate a file given its path name.
3401  */
3402 #ifndef _SYS_SYSPROTO_H_
3403 struct otruncate_args {
3404 	char	*path;
3405 	long	length;
3406 };
3407 #endif
3408 int
3409 otruncate(td, uap)
3410 	struct thread *td;
3411 	register struct otruncate_args /* {
3412 		char *path;
3413 		long length;
3414 	} */ *uap;
3415 {
3416 	struct truncate_args /* {
3417 		char *path;
3418 		int pad;
3419 		off_t length;
3420 	} */ nuap;
3421 
3422 	nuap.path = uap->path;
3423 	nuap.length = uap->length;
3424 	return (sys_truncate(td, &nuap));
3425 }
3426 #endif /* COMPAT_43 */
3427 
3428 #if defined(COMPAT_FREEBSD6)
3429 /* Versions with the pad argument */
3430 int
3431 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3432 {
3433 	struct truncate_args ouap;
3434 
3435 	ouap.path = uap->path;
3436 	ouap.length = uap->length;
3437 	return (sys_truncate(td, &ouap));
3438 }
3439 
3440 int
3441 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3442 {
3443 	struct ftruncate_args ouap;
3444 
3445 	ouap.fd = uap->fd;
3446 	ouap.length = uap->length;
3447 	return (sys_ftruncate(td, &ouap));
3448 }
3449 #endif
3450 
3451 /*
3452  * Sync an open file.
3453  */
3454 #ifndef _SYS_SYSPROTO_H_
3455 struct fsync_args {
3456 	int	fd;
3457 };
3458 #endif
3459 int
3460 sys_fsync(td, uap)
3461 	struct thread *td;
3462 	struct fsync_args /* {
3463 		int fd;
3464 	} */ *uap;
3465 {
3466 	struct vnode *vp;
3467 	struct mount *mp;
3468 	struct file *fp;
3469 	cap_rights_t rights;
3470 	int error, lock_flags;
3471 
3472 	AUDIT_ARG_FD(uap->fd);
3473 	error = getvnode(td->td_proc->p_fd, uap->fd,
3474 	    cap_rights_init(&rights, CAP_FSYNC), &fp);
3475 	if (error != 0)
3476 		return (error);
3477 	vp = fp->f_vnode;
3478 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3479 	if (error != 0)
3480 		goto drop;
3481 	if (MNT_SHARED_WRITES(mp) ||
3482 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3483 		lock_flags = LK_SHARED;
3484 	} else {
3485 		lock_flags = LK_EXCLUSIVE;
3486 	}
3487 	vn_lock(vp, lock_flags | LK_RETRY);
3488 	AUDIT_ARG_VNODE1(vp);
3489 	if (vp->v_object != NULL) {
3490 		VM_OBJECT_WLOCK(vp->v_object);
3491 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3492 		VM_OBJECT_WUNLOCK(vp->v_object);
3493 	}
3494 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3495 
3496 	VOP_UNLOCK(vp, 0);
3497 	vn_finished_write(mp);
3498 drop:
3499 	fdrop(fp, td);
3500 	return (error);
3501 }
3502 
3503 /*
3504  * Rename files.  Source and destination must either both be directories, or
3505  * both not be directories.  If target is a directory, it must be empty.
3506  */
3507 #ifndef _SYS_SYSPROTO_H_
3508 struct rename_args {
3509 	char	*from;
3510 	char	*to;
3511 };
3512 #endif
3513 int
3514 sys_rename(td, uap)
3515 	struct thread *td;
3516 	register struct rename_args /* {
3517 		char *from;
3518 		char *to;
3519 	} */ *uap;
3520 {
3521 
3522 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3523 	    uap->to, UIO_USERSPACE));
3524 }
3525 
3526 #ifndef _SYS_SYSPROTO_H_
3527 struct renameat_args {
3528 	int	oldfd;
3529 	char	*old;
3530 	int	newfd;
3531 	char	*new;
3532 };
3533 #endif
3534 int
3535 sys_renameat(struct thread *td, struct renameat_args *uap)
3536 {
3537 
3538 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3539 	    UIO_USERSPACE));
3540 }
3541 
3542 int
3543 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3544     enum uio_seg pathseg)
3545 {
3546 	struct mount *mp = NULL;
3547 	struct vnode *tvp, *fvp, *tdvp;
3548 	struct nameidata fromnd, tond;
3549 	cap_rights_t rights;
3550 	int error;
3551 
3552 again:
3553 	bwillwrite();
3554 #ifdef MAC
3555 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3556 	    AUDITVNODE1, pathseg, old, oldfd,
3557 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3558 #else
3559 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3560 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3561 #endif
3562 
3563 	if ((error = namei(&fromnd)) != 0)
3564 		return (error);
3565 #ifdef MAC
3566 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3567 	    fromnd.ni_vp, &fromnd.ni_cnd);
3568 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3569 	if (fromnd.ni_dvp != fromnd.ni_vp)
3570 		VOP_UNLOCK(fromnd.ni_vp, 0);
3571 #endif
3572 	fvp = fromnd.ni_vp;
3573 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3574 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3575 	    cap_rights_init(&rights, CAP_LINKAT), td);
3576 	if (fromnd.ni_vp->v_type == VDIR)
3577 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3578 	if ((error = namei(&tond)) != 0) {
3579 		/* Translate error code for rename("dir1", "dir2/."). */
3580 		if (error == EISDIR && fvp->v_type == VDIR)
3581 			error = EINVAL;
3582 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3583 		vrele(fromnd.ni_dvp);
3584 		vrele(fvp);
3585 		goto out1;
3586 	}
3587 	tdvp = tond.ni_dvp;
3588 	tvp = tond.ni_vp;
3589 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3590 	if (error != 0) {
3591 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3592 		NDFREE(&tond, NDF_ONLY_PNBUF);
3593 		if (tvp != NULL)
3594 			vput(tvp);
3595 		if (tdvp == tvp)
3596 			vrele(tdvp);
3597 		else
3598 			vput(tdvp);
3599 		vrele(fromnd.ni_dvp);
3600 		vrele(fvp);
3601 		vrele(tond.ni_startdir);
3602 		if (fromnd.ni_startdir != NULL)
3603 			vrele(fromnd.ni_startdir);
3604 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3605 		if (error != 0)
3606 			return (error);
3607 		goto again;
3608 	}
3609 	if (tvp != NULL) {
3610 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3611 			error = ENOTDIR;
3612 			goto out;
3613 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3614 			error = EISDIR;
3615 			goto out;
3616 		}
3617 #ifdef CAPABILITIES
3618 		if (newfd != AT_FDCWD) {
3619 			/*
3620 			 * If the target already exists we require CAP_UNLINKAT
3621 			 * from 'newfd'.
3622 			 */
3623 			error = cap_check(&tond.ni_filecaps.fc_rights,
3624 			    cap_rights_init(&rights, CAP_UNLINKAT));
3625 			if (error != 0)
3626 				goto out;
3627 		}
3628 #endif
3629 	}
3630 	if (fvp == tdvp) {
3631 		error = EINVAL;
3632 		goto out;
3633 	}
3634 	/*
3635 	 * If the source is the same as the destination (that is, if they
3636 	 * are links to the same vnode), then there is nothing to do.
3637 	 */
3638 	if (fvp == tvp)
3639 		error = -1;
3640 #ifdef MAC
3641 	else
3642 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3643 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3644 #endif
3645 out:
3646 	if (error == 0) {
3647 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3648 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3649 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3650 		NDFREE(&tond, NDF_ONLY_PNBUF);
3651 	} else {
3652 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3653 		NDFREE(&tond, NDF_ONLY_PNBUF);
3654 		if (tvp != NULL)
3655 			vput(tvp);
3656 		if (tdvp == tvp)
3657 			vrele(tdvp);
3658 		else
3659 			vput(tdvp);
3660 		vrele(fromnd.ni_dvp);
3661 		vrele(fvp);
3662 	}
3663 	vrele(tond.ni_startdir);
3664 	vn_finished_write(mp);
3665 out1:
3666 	if (fromnd.ni_startdir)
3667 		vrele(fromnd.ni_startdir);
3668 	if (error == -1)
3669 		return (0);
3670 	return (error);
3671 }
3672 
3673 /*
3674  * Make a directory file.
3675  */
3676 #ifndef _SYS_SYSPROTO_H_
3677 struct mkdir_args {
3678 	char	*path;
3679 	int	mode;
3680 };
3681 #endif
3682 int
3683 sys_mkdir(td, uap)
3684 	struct thread *td;
3685 	register struct mkdir_args /* {
3686 		char *path;
3687 		int mode;
3688 	} */ *uap;
3689 {
3690 
3691 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3692 	    uap->mode));
3693 }
3694 
3695 #ifndef _SYS_SYSPROTO_H_
3696 struct mkdirat_args {
3697 	int	fd;
3698 	char	*path;
3699 	mode_t	mode;
3700 };
3701 #endif
3702 int
3703 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3704 {
3705 
3706 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3707 }
3708 
3709 int
3710 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3711     int mode)
3712 {
3713 	struct mount *mp;
3714 	struct vnode *vp;
3715 	struct vattr vattr;
3716 	struct nameidata nd;
3717 	cap_rights_t rights;
3718 	int error;
3719 
3720 	AUDIT_ARG_MODE(mode);
3721 restart:
3722 	bwillwrite();
3723 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3724 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3725 	    td);
3726 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3727 	if ((error = namei(&nd)) != 0)
3728 		return (error);
3729 	vp = nd.ni_vp;
3730 	if (vp != NULL) {
3731 		NDFREE(&nd, NDF_ONLY_PNBUF);
3732 		/*
3733 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3734 		 * the strange behaviour of leaving the vnode unlocked
3735 		 * if the target is the same vnode as the parent.
3736 		 */
3737 		if (vp == nd.ni_dvp)
3738 			vrele(nd.ni_dvp);
3739 		else
3740 			vput(nd.ni_dvp);
3741 		vrele(vp);
3742 		return (EEXIST);
3743 	}
3744 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3745 		NDFREE(&nd, NDF_ONLY_PNBUF);
3746 		vput(nd.ni_dvp);
3747 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3748 			return (error);
3749 		goto restart;
3750 	}
3751 	VATTR_NULL(&vattr);
3752 	vattr.va_type = VDIR;
3753 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3754 #ifdef MAC
3755 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3756 	    &vattr);
3757 	if (error != 0)
3758 		goto out;
3759 #endif
3760 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3761 #ifdef MAC
3762 out:
3763 #endif
3764 	NDFREE(&nd, NDF_ONLY_PNBUF);
3765 	vput(nd.ni_dvp);
3766 	if (error == 0)
3767 		vput(nd.ni_vp);
3768 	vn_finished_write(mp);
3769 	return (error);
3770 }
3771 
3772 /*
3773  * Remove a directory file.
3774  */
3775 #ifndef _SYS_SYSPROTO_H_
3776 struct rmdir_args {
3777 	char	*path;
3778 };
3779 #endif
3780 int
3781 sys_rmdir(td, uap)
3782 	struct thread *td;
3783 	struct rmdir_args /* {
3784 		char *path;
3785 	} */ *uap;
3786 {
3787 
3788 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3789 }
3790 
3791 int
3792 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3793 {
3794 	struct mount *mp;
3795 	struct vnode *vp;
3796 	struct nameidata nd;
3797 	cap_rights_t rights;
3798 	int error;
3799 
3800 restart:
3801 	bwillwrite();
3802 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3803 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3804 	if ((error = namei(&nd)) != 0)
3805 		return (error);
3806 	vp = nd.ni_vp;
3807 	if (vp->v_type != VDIR) {
3808 		error = ENOTDIR;
3809 		goto out;
3810 	}
3811 	/*
3812 	 * No rmdir "." please.
3813 	 */
3814 	if (nd.ni_dvp == vp) {
3815 		error = EINVAL;
3816 		goto out;
3817 	}
3818 	/*
3819 	 * The root of a mounted filesystem cannot be deleted.
3820 	 */
3821 	if (vp->v_vflag & VV_ROOT) {
3822 		error = EBUSY;
3823 		goto out;
3824 	}
3825 #ifdef MAC
3826 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3827 	    &nd.ni_cnd);
3828 	if (error != 0)
3829 		goto out;
3830 #endif
3831 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3832 		NDFREE(&nd, NDF_ONLY_PNBUF);
3833 		vput(vp);
3834 		if (nd.ni_dvp == vp)
3835 			vrele(nd.ni_dvp);
3836 		else
3837 			vput(nd.ni_dvp);
3838 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3839 			return (error);
3840 		goto restart;
3841 	}
3842 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3843 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3844 	vn_finished_write(mp);
3845 out:
3846 	NDFREE(&nd, NDF_ONLY_PNBUF);
3847 	vput(vp);
3848 	if (nd.ni_dvp == vp)
3849 		vrele(nd.ni_dvp);
3850 	else
3851 		vput(nd.ni_dvp);
3852 	return (error);
3853 }
3854 
3855 #ifdef COMPAT_43
3856 /*
3857  * Read a block of directory entries in a filesystem independent format.
3858  */
3859 #ifndef _SYS_SYSPROTO_H_
3860 struct ogetdirentries_args {
3861 	int	fd;
3862 	char	*buf;
3863 	u_int	count;
3864 	long	*basep;
3865 };
3866 #endif
3867 int
3868 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3869 {
3870 	long loff;
3871 	int error;
3872 
3873 	error = kern_ogetdirentries(td, uap, &loff);
3874 	if (error == 0)
3875 		error = copyout(&loff, uap->basep, sizeof(long));
3876 	return (error);
3877 }
3878 
3879 int
3880 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3881     long *ploff)
3882 {
3883 	struct vnode *vp;
3884 	struct file *fp;
3885 	struct uio auio, kuio;
3886 	struct iovec aiov, kiov;
3887 	struct dirent *dp, *edp;
3888 	cap_rights_t rights;
3889 	caddr_t dirbuf;
3890 	int error, eofflag, readcnt;
3891 	long loff;
3892 	off_t foffset;
3893 
3894 	/* XXX arbitrary sanity limit on `count'. */
3895 	if (uap->count > 64 * 1024)
3896 		return (EINVAL);
3897 	error = getvnode(td->td_proc->p_fd, uap->fd,
3898 	    cap_rights_init(&rights, CAP_READ), &fp);
3899 	if (error != 0)
3900 		return (error);
3901 	if ((fp->f_flag & FREAD) == 0) {
3902 		fdrop(fp, td);
3903 		return (EBADF);
3904 	}
3905 	vp = fp->f_vnode;
3906 	foffset = foffset_lock(fp, 0);
3907 unionread:
3908 	if (vp->v_type != VDIR) {
3909 		foffset_unlock(fp, foffset, 0);
3910 		fdrop(fp, td);
3911 		return (EINVAL);
3912 	}
3913 	aiov.iov_base = uap->buf;
3914 	aiov.iov_len = uap->count;
3915 	auio.uio_iov = &aiov;
3916 	auio.uio_iovcnt = 1;
3917 	auio.uio_rw = UIO_READ;
3918 	auio.uio_segflg = UIO_USERSPACE;
3919 	auio.uio_td = td;
3920 	auio.uio_resid = uap->count;
3921 	vn_lock(vp, LK_SHARED | LK_RETRY);
3922 	loff = auio.uio_offset = foffset;
3923 #ifdef MAC
3924 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3925 	if (error != 0) {
3926 		VOP_UNLOCK(vp, 0);
3927 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3928 		fdrop(fp, td);
3929 		return (error);
3930 	}
3931 #endif
3932 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3933 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3934 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3935 			    NULL, NULL);
3936 			foffset = auio.uio_offset;
3937 		} else
3938 #	endif
3939 	{
3940 		kuio = auio;
3941 		kuio.uio_iov = &kiov;
3942 		kuio.uio_segflg = UIO_SYSSPACE;
3943 		kiov.iov_len = uap->count;
3944 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3945 		kiov.iov_base = dirbuf;
3946 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3947 			    NULL, NULL);
3948 		foffset = kuio.uio_offset;
3949 		if (error == 0) {
3950 			readcnt = uap->count - kuio.uio_resid;
3951 			edp = (struct dirent *)&dirbuf[readcnt];
3952 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3953 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3954 					/*
3955 					 * The expected low byte of
3956 					 * dp->d_namlen is our dp->d_type.
3957 					 * The high MBZ byte of dp->d_namlen
3958 					 * is our dp->d_namlen.
3959 					 */
3960 					dp->d_type = dp->d_namlen;
3961 					dp->d_namlen = 0;
3962 #				else
3963 					/*
3964 					 * The dp->d_type is the high byte
3965 					 * of the expected dp->d_namlen,
3966 					 * so must be zero'ed.
3967 					 */
3968 					dp->d_type = 0;
3969 #				endif
3970 				if (dp->d_reclen > 0) {
3971 					dp = (struct dirent *)
3972 					    ((char *)dp + dp->d_reclen);
3973 				} else {
3974 					error = EIO;
3975 					break;
3976 				}
3977 			}
3978 			if (dp >= edp)
3979 				error = uiomove(dirbuf, readcnt, &auio);
3980 		}
3981 		free(dirbuf, M_TEMP);
3982 	}
3983 	if (error != 0) {
3984 		VOP_UNLOCK(vp, 0);
3985 		foffset_unlock(fp, foffset, 0);
3986 		fdrop(fp, td);
3987 		return (error);
3988 	}
3989 	if (uap->count == auio.uio_resid &&
3990 	    (vp->v_vflag & VV_ROOT) &&
3991 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3992 		struct vnode *tvp = vp;
3993 		vp = vp->v_mount->mnt_vnodecovered;
3994 		VREF(vp);
3995 		fp->f_vnode = vp;
3996 		fp->f_data = vp;
3997 		foffset = 0;
3998 		vput(tvp);
3999 		goto unionread;
4000 	}
4001 	VOP_UNLOCK(vp, 0);
4002 	foffset_unlock(fp, foffset, 0);
4003 	fdrop(fp, td);
4004 	td->td_retval[0] = uap->count - auio.uio_resid;
4005 	if (error == 0)
4006 		*ploff = loff;
4007 	return (error);
4008 }
4009 #endif /* COMPAT_43 */
4010 
4011 /*
4012  * Read a block of directory entries in a filesystem independent format.
4013  */
4014 #ifndef _SYS_SYSPROTO_H_
4015 struct getdirentries_args {
4016 	int	fd;
4017 	char	*buf;
4018 	u_int	count;
4019 	long	*basep;
4020 };
4021 #endif
4022 int
4023 sys_getdirentries(td, uap)
4024 	struct thread *td;
4025 	register struct getdirentries_args /* {
4026 		int fd;
4027 		char *buf;
4028 		u_int count;
4029 		long *basep;
4030 	} */ *uap;
4031 {
4032 	long base;
4033 	int error;
4034 
4035 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4036 	    NULL, UIO_USERSPACE);
4037 	if (error != 0)
4038 		return (error);
4039 	if (uap->basep != NULL)
4040 		error = copyout(&base, uap->basep, sizeof(long));
4041 	return (error);
4042 }
4043 
4044 int
4045 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4046     long *basep, ssize_t *residp, enum uio_seg bufseg)
4047 {
4048 	struct vnode *vp;
4049 	struct file *fp;
4050 	struct uio auio;
4051 	struct iovec aiov;
4052 	cap_rights_t rights;
4053 	long loff;
4054 	int error, eofflag;
4055 	off_t foffset;
4056 
4057 	AUDIT_ARG_FD(fd);
4058 	if (count > IOSIZE_MAX)
4059 		return (EINVAL);
4060 	auio.uio_resid = count;
4061 	error = getvnode(td->td_proc->p_fd, fd,
4062 	    cap_rights_init(&rights, CAP_READ), &fp);
4063 	if (error != 0)
4064 		return (error);
4065 	if ((fp->f_flag & FREAD) == 0) {
4066 		fdrop(fp, td);
4067 		return (EBADF);
4068 	}
4069 	vp = fp->f_vnode;
4070 	foffset = foffset_lock(fp, 0);
4071 unionread:
4072 	if (vp->v_type != VDIR) {
4073 		error = EINVAL;
4074 		goto fail;
4075 	}
4076 	aiov.iov_base = buf;
4077 	aiov.iov_len = count;
4078 	auio.uio_iov = &aiov;
4079 	auio.uio_iovcnt = 1;
4080 	auio.uio_rw = UIO_READ;
4081 	auio.uio_segflg = bufseg;
4082 	auio.uio_td = td;
4083 	vn_lock(vp, LK_SHARED | LK_RETRY);
4084 	AUDIT_ARG_VNODE1(vp);
4085 	loff = auio.uio_offset = foffset;
4086 #ifdef MAC
4087 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4088 	if (error == 0)
4089 #endif
4090 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4091 		    NULL);
4092 	foffset = auio.uio_offset;
4093 	if (error != 0) {
4094 		VOP_UNLOCK(vp, 0);
4095 		goto fail;
4096 	}
4097 	if (count == auio.uio_resid &&
4098 	    (vp->v_vflag & VV_ROOT) &&
4099 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4100 		struct vnode *tvp = vp;
4101 
4102 		vp = vp->v_mount->mnt_vnodecovered;
4103 		VREF(vp);
4104 		fp->f_vnode = vp;
4105 		fp->f_data = vp;
4106 		foffset = 0;
4107 		vput(tvp);
4108 		goto unionread;
4109 	}
4110 	VOP_UNLOCK(vp, 0);
4111 	*basep = loff;
4112 	if (residp != NULL)
4113 		*residp = auio.uio_resid;
4114 	td->td_retval[0] = count - auio.uio_resid;
4115 fail:
4116 	foffset_unlock(fp, foffset, 0);
4117 	fdrop(fp, td);
4118 	return (error);
4119 }
4120 
4121 #ifndef _SYS_SYSPROTO_H_
4122 struct getdents_args {
4123 	int fd;
4124 	char *buf;
4125 	size_t count;
4126 };
4127 #endif
4128 int
4129 sys_getdents(td, uap)
4130 	struct thread *td;
4131 	register struct getdents_args /* {
4132 		int fd;
4133 		char *buf;
4134 		u_int count;
4135 	} */ *uap;
4136 {
4137 	struct getdirentries_args ap;
4138 
4139 	ap.fd = uap->fd;
4140 	ap.buf = uap->buf;
4141 	ap.count = uap->count;
4142 	ap.basep = NULL;
4143 	return (sys_getdirentries(td, &ap));
4144 }
4145 
4146 /*
4147  * Set the mode mask for creation of filesystem nodes.
4148  */
4149 #ifndef _SYS_SYSPROTO_H_
4150 struct umask_args {
4151 	int	newmask;
4152 };
4153 #endif
4154 int
4155 sys_umask(td, uap)
4156 	struct thread *td;
4157 	struct umask_args /* {
4158 		int newmask;
4159 	} */ *uap;
4160 {
4161 	register struct filedesc *fdp;
4162 
4163 	FILEDESC_XLOCK(td->td_proc->p_fd);
4164 	fdp = td->td_proc->p_fd;
4165 	td->td_retval[0] = fdp->fd_cmask;
4166 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4167 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4168 	return (0);
4169 }
4170 
4171 /*
4172  * Void all references to file by ripping underlying filesystem away from
4173  * vnode.
4174  */
4175 #ifndef _SYS_SYSPROTO_H_
4176 struct revoke_args {
4177 	char	*path;
4178 };
4179 #endif
4180 int
4181 sys_revoke(td, uap)
4182 	struct thread *td;
4183 	register struct revoke_args /* {
4184 		char *path;
4185 	} */ *uap;
4186 {
4187 	struct vnode *vp;
4188 	struct vattr vattr;
4189 	struct nameidata nd;
4190 	int error;
4191 
4192 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4193 	    uap->path, td);
4194 	if ((error = namei(&nd)) != 0)
4195 		return (error);
4196 	vp = nd.ni_vp;
4197 	NDFREE(&nd, NDF_ONLY_PNBUF);
4198 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4199 		error = EINVAL;
4200 		goto out;
4201 	}
4202 #ifdef MAC
4203 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4204 	if (error != 0)
4205 		goto out;
4206 #endif
4207 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4208 	if (error != 0)
4209 		goto out;
4210 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4211 		error = priv_check(td, PRIV_VFS_ADMIN);
4212 		if (error != 0)
4213 			goto out;
4214 	}
4215 	if (vcount(vp) > 1)
4216 		VOP_REVOKE(vp, REVOKEALL);
4217 out:
4218 	vput(vp);
4219 	return (error);
4220 }
4221 
4222 /*
4223  * Convert a user file descriptor to a kernel file entry and check that, if it
4224  * is a capability, the correct rights are present. A reference on the file
4225  * entry is held upon returning.
4226  */
4227 int
4228 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4229 {
4230 	struct file *fp;
4231 	int error;
4232 
4233 	error = fget_unlocked(fdp, fd, rightsp, &fp, NULL);
4234 	if (error != 0)
4235 		return (error);
4236 
4237 	/*
4238 	 * The file could be not of the vnode type, or it may be not
4239 	 * yet fully initialized, in which case the f_vnode pointer
4240 	 * may be set, but f_ops is still badfileops.  E.g.,
4241 	 * devfs_open() transiently create such situation to
4242 	 * facilitate csw d_fdopen().
4243 	 *
4244 	 * Dupfdopen() handling in kern_openat() installs the
4245 	 * half-baked file into the process descriptor table, allowing
4246 	 * other thread to dereference it. Guard against the race by
4247 	 * checking f_ops.
4248 	 */
4249 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4250 		fdrop(fp, curthread);
4251 		return (EINVAL);
4252 	}
4253 	*fpp = fp;
4254 	return (0);
4255 }
4256 
4257 
4258 /*
4259  * Get an (NFS) file handle.
4260  */
4261 #ifndef _SYS_SYSPROTO_H_
4262 struct lgetfh_args {
4263 	char	*fname;
4264 	fhandle_t *fhp;
4265 };
4266 #endif
4267 int
4268 sys_lgetfh(td, uap)
4269 	struct thread *td;
4270 	register struct lgetfh_args *uap;
4271 {
4272 	struct nameidata nd;
4273 	fhandle_t fh;
4274 	register struct vnode *vp;
4275 	int error;
4276 
4277 	error = priv_check(td, PRIV_VFS_GETFH);
4278 	if (error != 0)
4279 		return (error);
4280 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4281 	    uap->fname, td);
4282 	error = namei(&nd);
4283 	if (error != 0)
4284 		return (error);
4285 	NDFREE(&nd, NDF_ONLY_PNBUF);
4286 	vp = nd.ni_vp;
4287 	bzero(&fh, sizeof(fh));
4288 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4289 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4290 	vput(vp);
4291 	if (error == 0)
4292 		error = copyout(&fh, uap->fhp, sizeof (fh));
4293 	return (error);
4294 }
4295 
4296 #ifndef _SYS_SYSPROTO_H_
4297 struct getfh_args {
4298 	char	*fname;
4299 	fhandle_t *fhp;
4300 };
4301 #endif
4302 int
4303 sys_getfh(td, uap)
4304 	struct thread *td;
4305 	register struct getfh_args *uap;
4306 {
4307 	struct nameidata nd;
4308 	fhandle_t fh;
4309 	register struct vnode *vp;
4310 	int error;
4311 
4312 	error = priv_check(td, PRIV_VFS_GETFH);
4313 	if (error != 0)
4314 		return (error);
4315 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4316 	    uap->fname, td);
4317 	error = namei(&nd);
4318 	if (error != 0)
4319 		return (error);
4320 	NDFREE(&nd, NDF_ONLY_PNBUF);
4321 	vp = nd.ni_vp;
4322 	bzero(&fh, sizeof(fh));
4323 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4324 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4325 	vput(vp);
4326 	if (error == 0)
4327 		error = copyout(&fh, uap->fhp, sizeof (fh));
4328 	return (error);
4329 }
4330 
4331 /*
4332  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4333  * open descriptor.
4334  *
4335  * warning: do not remove the priv_check() call or this becomes one giant
4336  * security hole.
4337  */
4338 #ifndef _SYS_SYSPROTO_H_
4339 struct fhopen_args {
4340 	const struct fhandle *u_fhp;
4341 	int flags;
4342 };
4343 #endif
4344 int
4345 sys_fhopen(td, uap)
4346 	struct thread *td;
4347 	struct fhopen_args /* {
4348 		const struct fhandle *u_fhp;
4349 		int flags;
4350 	} */ *uap;
4351 {
4352 	struct mount *mp;
4353 	struct vnode *vp;
4354 	struct fhandle fhp;
4355 	struct file *fp;
4356 	int fmode, error;
4357 	int indx;
4358 
4359 	error = priv_check(td, PRIV_VFS_FHOPEN);
4360 	if (error != 0)
4361 		return (error);
4362 	indx = -1;
4363 	fmode = FFLAGS(uap->flags);
4364 	/* why not allow a non-read/write open for our lockd? */
4365 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4366 		return (EINVAL);
4367 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4368 	if (error != 0)
4369 		return(error);
4370 	/* find the mount point */
4371 	mp = vfs_busyfs(&fhp.fh_fsid);
4372 	if (mp == NULL)
4373 		return (ESTALE);
4374 	/* now give me my vnode, it gets returned to me locked */
4375 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4376 	vfs_unbusy(mp);
4377 	if (error != 0)
4378 		return (error);
4379 
4380 	error = falloc_noinstall(td, &fp);
4381 	if (error != 0) {
4382 		vput(vp);
4383 		return (error);
4384 	}
4385 	/*
4386 	 * An extra reference on `fp' has been held for us by
4387 	 * falloc_noinstall().
4388 	 */
4389 
4390 #ifdef INVARIANTS
4391 	td->td_dupfd = -1;
4392 #endif
4393 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4394 	if (error != 0) {
4395 		KASSERT(fp->f_ops == &badfileops,
4396 		    ("VOP_OPEN in fhopen() set f_ops"));
4397 		KASSERT(td->td_dupfd < 0,
4398 		    ("fhopen() encountered fdopen()"));
4399 
4400 		vput(vp);
4401 		goto bad;
4402 	}
4403 #ifdef INVARIANTS
4404 	td->td_dupfd = 0;
4405 #endif
4406 	fp->f_vnode = vp;
4407 	fp->f_seqcount = 1;
4408 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4409 	    &vnops);
4410 	VOP_UNLOCK(vp, 0);
4411 	if ((fmode & O_TRUNC) != 0) {
4412 		error = fo_truncate(fp, 0, td->td_ucred, td);
4413 		if (error != 0)
4414 			goto bad;
4415 	}
4416 
4417 	error = finstall(td, fp, &indx, fmode, NULL);
4418 bad:
4419 	fdrop(fp, td);
4420 	td->td_retval[0] = indx;
4421 	return (error);
4422 }
4423 
4424 /*
4425  * Stat an (NFS) file handle.
4426  */
4427 #ifndef _SYS_SYSPROTO_H_
4428 struct fhstat_args {
4429 	struct fhandle *u_fhp;
4430 	struct stat *sb;
4431 };
4432 #endif
4433 int
4434 sys_fhstat(td, uap)
4435 	struct thread *td;
4436 	register struct fhstat_args /* {
4437 		struct fhandle *u_fhp;
4438 		struct stat *sb;
4439 	} */ *uap;
4440 {
4441 	struct stat sb;
4442 	struct fhandle fh;
4443 	int error;
4444 
4445 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4446 	if (error != 0)
4447 		return (error);
4448 	error = kern_fhstat(td, fh, &sb);
4449 	if (error == 0)
4450 		error = copyout(&sb, uap->sb, sizeof(sb));
4451 	return (error);
4452 }
4453 
4454 int
4455 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4456 {
4457 	struct mount *mp;
4458 	struct vnode *vp;
4459 	int error;
4460 
4461 	error = priv_check(td, PRIV_VFS_FHSTAT);
4462 	if (error != 0)
4463 		return (error);
4464 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4465 		return (ESTALE);
4466 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4467 	vfs_unbusy(mp);
4468 	if (error != 0)
4469 		return (error);
4470 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4471 	vput(vp);
4472 	return (error);
4473 }
4474 
4475 /*
4476  * Implement fstatfs() for (NFS) file handles.
4477  */
4478 #ifndef _SYS_SYSPROTO_H_
4479 struct fhstatfs_args {
4480 	struct fhandle *u_fhp;
4481 	struct statfs *buf;
4482 };
4483 #endif
4484 int
4485 sys_fhstatfs(td, uap)
4486 	struct thread *td;
4487 	struct fhstatfs_args /* {
4488 		struct fhandle *u_fhp;
4489 		struct statfs *buf;
4490 	} */ *uap;
4491 {
4492 	struct statfs sf;
4493 	fhandle_t fh;
4494 	int error;
4495 
4496 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4497 	if (error != 0)
4498 		return (error);
4499 	error = kern_fhstatfs(td, fh, &sf);
4500 	if (error != 0)
4501 		return (error);
4502 	return (copyout(&sf, uap->buf, sizeof(sf)));
4503 }
4504 
4505 int
4506 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4507 {
4508 	struct statfs *sp;
4509 	struct mount *mp;
4510 	struct vnode *vp;
4511 	int error;
4512 
4513 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4514 	if (error != 0)
4515 		return (error);
4516 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4517 		return (ESTALE);
4518 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4519 	if (error != 0) {
4520 		vfs_unbusy(mp);
4521 		return (error);
4522 	}
4523 	vput(vp);
4524 	error = prison_canseemount(td->td_ucred, mp);
4525 	if (error != 0)
4526 		goto out;
4527 #ifdef MAC
4528 	error = mac_mount_check_stat(td->td_ucred, mp);
4529 	if (error != 0)
4530 		goto out;
4531 #endif
4532 	/*
4533 	 * Set these in case the underlying filesystem fails to do so.
4534 	 */
4535 	sp = &mp->mnt_stat;
4536 	sp->f_version = STATFS_VERSION;
4537 	sp->f_namemax = NAME_MAX;
4538 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4539 	error = VFS_STATFS(mp, sp);
4540 	if (error == 0)
4541 		*buf = *sp;
4542 out:
4543 	vfs_unbusy(mp);
4544 	return (error);
4545 }
4546 
4547 int
4548 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4549 {
4550 	struct file *fp;
4551 	struct mount *mp;
4552 	struct vnode *vp;
4553 	cap_rights_t rights;
4554 	off_t olen, ooffset;
4555 	int error;
4556 
4557 	if (offset < 0 || len <= 0)
4558 		return (EINVAL);
4559 	/* Check for wrap. */
4560 	if (offset > OFF_MAX - len)
4561 		return (EFBIG);
4562 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4563 	if (error != 0)
4564 		return (error);
4565 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4566 		error = ESPIPE;
4567 		goto out;
4568 	}
4569 	if ((fp->f_flag & FWRITE) == 0) {
4570 		error = EBADF;
4571 		goto out;
4572 	}
4573 	if (fp->f_type != DTYPE_VNODE) {
4574 		error = ENODEV;
4575 		goto out;
4576 	}
4577 	vp = fp->f_vnode;
4578 	if (vp->v_type != VREG) {
4579 		error = ENODEV;
4580 		goto out;
4581 	}
4582 
4583 	/* Allocating blocks may take a long time, so iterate. */
4584 	for (;;) {
4585 		olen = len;
4586 		ooffset = offset;
4587 
4588 		bwillwrite();
4589 		mp = NULL;
4590 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4591 		if (error != 0)
4592 			break;
4593 		error = vn_lock(vp, LK_EXCLUSIVE);
4594 		if (error != 0) {
4595 			vn_finished_write(mp);
4596 			break;
4597 		}
4598 #ifdef MAC
4599 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4600 		if (error == 0)
4601 #endif
4602 			error = VOP_ALLOCATE(vp, &offset, &len);
4603 		VOP_UNLOCK(vp, 0);
4604 		vn_finished_write(mp);
4605 
4606 		if (olen + ooffset != offset + len) {
4607 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4608 			    ooffset, olen, offset, len);
4609 		}
4610 		if (error != 0 || len == 0)
4611 			break;
4612 		KASSERT(olen > len, ("Iteration did not make progress?"));
4613 		maybe_yield();
4614 	}
4615  out:
4616 	fdrop(fp, td);
4617 	return (error);
4618 }
4619 
4620 int
4621 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4622 {
4623 
4624 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4625 	    uap->len);
4626 	return (0);
4627 }
4628 
4629 /*
4630  * Unlike madvise(2), we do not make a best effort to remember every
4631  * possible caching hint.  Instead, we remember the last setting with
4632  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4633  * region of any current setting.
4634  */
4635 int
4636 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4637     int advice)
4638 {
4639 	struct fadvise_info *fa, *new;
4640 	struct file *fp;
4641 	struct vnode *vp;
4642 	cap_rights_t rights;
4643 	off_t end;
4644 	int error;
4645 
4646 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4647 		return (EINVAL);
4648 	switch (advice) {
4649 	case POSIX_FADV_SEQUENTIAL:
4650 	case POSIX_FADV_RANDOM:
4651 	case POSIX_FADV_NOREUSE:
4652 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4653 		break;
4654 	case POSIX_FADV_NORMAL:
4655 	case POSIX_FADV_WILLNEED:
4656 	case POSIX_FADV_DONTNEED:
4657 		new = NULL;
4658 		break;
4659 	default:
4660 		return (EINVAL);
4661 	}
4662 	/* XXX: CAP_POSIX_FADVISE? */
4663 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4664 	if (error != 0)
4665 		goto out;
4666 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4667 		error = ESPIPE;
4668 		goto out;
4669 	}
4670 	if (fp->f_type != DTYPE_VNODE) {
4671 		error = ENODEV;
4672 		goto out;
4673 	}
4674 	vp = fp->f_vnode;
4675 	if (vp->v_type != VREG) {
4676 		error = ENODEV;
4677 		goto out;
4678 	}
4679 	if (len == 0)
4680 		end = OFF_MAX;
4681 	else
4682 		end = offset + len - 1;
4683 	switch (advice) {
4684 	case POSIX_FADV_SEQUENTIAL:
4685 	case POSIX_FADV_RANDOM:
4686 	case POSIX_FADV_NOREUSE:
4687 		/*
4688 		 * Try to merge any existing non-standard region with
4689 		 * this new region if possible, otherwise create a new
4690 		 * non-standard region for this request.
4691 		 */
4692 		mtx_pool_lock(mtxpool_sleep, fp);
4693 		fa = fp->f_advice;
4694 		if (fa != NULL && fa->fa_advice == advice &&
4695 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4696 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4697 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4698 			if (offset < fa->fa_start)
4699 				fa->fa_start = offset;
4700 			if (end > fa->fa_end)
4701 				fa->fa_end = end;
4702 		} else {
4703 			new->fa_advice = advice;
4704 			new->fa_start = offset;
4705 			new->fa_end = end;
4706 			new->fa_prevstart = 0;
4707 			new->fa_prevend = 0;
4708 			fp->f_advice = new;
4709 			new = fa;
4710 		}
4711 		mtx_pool_unlock(mtxpool_sleep, fp);
4712 		break;
4713 	case POSIX_FADV_NORMAL:
4714 		/*
4715 		 * If a the "normal" region overlaps with an existing
4716 		 * non-standard region, trim or remove the
4717 		 * non-standard region.
4718 		 */
4719 		mtx_pool_lock(mtxpool_sleep, fp);
4720 		fa = fp->f_advice;
4721 		if (fa != NULL) {
4722 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4723 				new = fa;
4724 				fp->f_advice = NULL;
4725 			} else if (offset <= fa->fa_start &&
4726 			    end >= fa->fa_start)
4727 				fa->fa_start = end + 1;
4728 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4729 				fa->fa_end = offset - 1;
4730 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4731 				/*
4732 				 * If the "normal" region is a middle
4733 				 * portion of the existing
4734 				 * non-standard region, just remove
4735 				 * the whole thing rather than picking
4736 				 * one side or the other to
4737 				 * preserve.
4738 				 */
4739 				new = fa;
4740 				fp->f_advice = NULL;
4741 			}
4742 		}
4743 		mtx_pool_unlock(mtxpool_sleep, fp);
4744 		break;
4745 	case POSIX_FADV_WILLNEED:
4746 	case POSIX_FADV_DONTNEED:
4747 		error = VOP_ADVISE(vp, offset, end, advice);
4748 		break;
4749 	}
4750 out:
4751 	if (fp != NULL)
4752 		fdrop(fp, td);
4753 	free(new, M_FADVISE);
4754 	return (error);
4755 }
4756 
4757 int
4758 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4759 {
4760 
4761 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4762 	    uap->len, uap->advice);
4763 	return (0);
4764 }
4765